Nursing Research Principles and Methods 7th Edition Polit - PDFCOFFEE.COM (2025)

PREFACE

T

his seventh edition of Nursing Research: Principles and Methods presents many important changes to this textbook. This edition retains the features that have made this an award-winning textbook, while introducing revisions that will make it more relevant in an environment that is increasingly focused on evidence-based nursing practice. New to This Edition • Emphasis on producing and evaluating research evidence. This edition focuses more squarely on the fact that research is an evidencebuilding enterprise. We stress throughout that the decisions researchers make in designing and implementing a study have implications for the quality of evidence the study yields—and that the quality of evidence affects the utility of study findings for nursing practice. We have also expanded the final chapter on using research in an evidence-based nursing practice. • Expanded coverage of qualitative research methods. As in the sixth edition, this textbook discusses the methods associated with naturalistic inquiries (qualitative studies) in a manner roughly parallel to the description of methods more typically used in traditional scientific research (quantitative studies). This edition, which is coauthored by a prominent qualitative nurse researcher, goes even further than previous ones in offering assistance to those embarking on a qualitative research project. For example,

differences in the approaches used by grounded theory researchers, phenomenologists, and ethnographers are noted throughout. • Better “how to” assistance. There are even more “how to” tips in this edition than in previous ones, with many new ones aimed at qualitative researchers. Moreover, these tips are interspersed throughout the text rather than clustered at the end, which not only places them in more relevant locations, but also makes for a livelier presentation. • Improved readability. This textbook has been widely hailed for its clear, concise, and “userfriendly” presentation. In this edition, however, we have gone to great lengths to write in an even simpler, more straightforward fashion, in recognition of the fact that research methods are an inherently complex topic. Additionally, the readability of the text is enhanced by several visual features, such as the use of a new, fullcolor design. • Greater acknowledgment of international efforts. This edition gives better recognition to the contributions of nurse researchers from around the globe. Organization of the Text The content of this edition is organized into seven main parts. • Part I—Foundations of Nursing Research introduces fundamental concepts in nursing research. Chapter 1 summarizes the history and vii

viii

Preface

future of nursing research, introduces the topic of using research evidence for nursing practice, discusses the philosophical underpinnings of qualitative research versus quantitative research, and describes the major purposes of nursing research. Chapter 2 introduces readers to key terms, with new emphasis on terms related to the quality of research evidence. Chapter 3 presents an overview of steps in the research process for both qualitative and quantitative studies. • Part II—Conceptualizing a Research Study further sets the stage for learning about the research process by discussing issues relating to a study’s conceptualization: the formulation of research questions and hypotheses (Chapter 4); the review of relevant research (Chapter 5); and the development of theoretical and conceptual contexts (Chapter 6). • Part III—Designs for Nursing Research presents material on the design of qualitative and quantitative nursing research studies. Chapter 7 discusses issues important to the design of research that is ethically sound. Chapter 8 describes fundamental principles and applications of quantitative research design, while Chapter 9 describes mechanisms of research control for quantitative studies. Chapter 10 examines quantitative research with different purposes. Chapter 11 is devoted to research designs for qualitative studies, with new material added on critical theory, feminist, and participatory action research. Chapter 12 discusses mixed method research designs in which methods for qualitative and quantitative inquiry are blended. Chapter 13 presents designs and strategies for selecting samples of study participants. • Part IV—Measurement and Data Collection deals with the gathering of information in a study. Chapter 14 discusses the overall data collection plan, and the subsequent three chapters present materials on specific data collection methods such as self-reports (Chapter 15), observation (Chapter 16), and biophysiologic and other methods (Chapter 17). Chapter 18 discusses the concept of measurement, and then focuses on methods of assessing data quality.

• Part V—The Analysis of Research Data discusses methods of analyzing qualitative and quantitative data. Chapters 19, 20, and 21 present an overview of univariate, inferential, and multivariate statistical analyses, respectively. Chapter 22 describes the development of an overall analytic strategy for quantitative studies. Chapter 23 discusses methods of doing qualitative analyses, with specific information about grounded theory, phenomenologic, and ethnographic analyses. • Part VI—Communicating Research focuses on two types of research communication. Chapter 24 discusses how to write about research and, greatly expanded in this edition, how to publish a research report and prepare a thesis or dissertation. Chapter 25 offers suggestions and guidelines on preparing research proposals. • Part VII—Using Research Results is intended to sharpen the critical awareness of nurses with regard to the use of research findings by practicing nurses. Chapter 26 discusses the interpretation and appraisal of research reports. The concluding chapter (Chapter 27) offers suggestions on utilizing research to build an evidence-based practice, and includes guidance on performing integrative reviews. Key Features This textbook was designed to be helpful to those who are learning how to do research, as well as to the growing number of nurses who are learning to appraise research reports critically and to use research findings in practice. Many of the features successfully used in previous editions have been retained in this seventh edition. Among the basic principles that helped to shape this and earlier editions of this book are (1) an unswerving conviction that the development of research skills is critical to the nursing profession; (2) a fundamental belief that research is an intellectually and professionally rewarding enterprise; and (3) a judgment that learning about research methods need be neither intimidating nor dull. Consistent with these principles, we have tried to present the fundamentals of research methods in a way

Preface

that both facilitates understanding and arouses curiosity and interest. Key features of our approach include the following: • Research Examples. Each chapter concludes with one or two actual research examples (usually one quantitative and one qualitative study) designed to highlight critical points made in the chapter and to sharpen the reader’s critical thinking skills. In addition, many research examples are used to illustrate key points in the text and to stimulate students’ thinking about a research project. • Clear, “user friendly” style. Our writing style is designed to be easily digestible and nonintimidating. Concepts are introduced carefully and systematically, difficult ideas are presented clearly and from several vantage points, and readers are assumed to have no prior exposure to technical terms. • Specific practical tips on doing research. The textbook is filled with practical guidance on how to translate the abstract notions of research methods into realistic strategies for conducting research. Every chapter includes several tips for applying the chapter’s lessons to real-life situations. The inclusion of these suggestions acknowledges the fact that there is often a large gap between what gets taught in research methods textbooks and what a researcher needs to know in conducting a study. • Aids to student learning. Several features are used to enhance and reinforce learning and to help focus the student’s attention on specific areas of text content, including the following: succinct, bulleted summaries at the end of each chapter; tables and figures that provide examples and graphic materials in support of the text discussion; study suggestions at the end of each chapter; and suggested methodologic and substantive readings for each chapter. Teaching-Learning Package Nursing Research: Principles and Methods, seventh edition, has an ancillary package designed with both students and instructors in mind.

ix

• The Study Guide augments the textbook and provides students with exercises that correspond to each text chapter. Answers to selected exercises are included at the end of the Study Guide. The Study Guide also includes two actual research reports that students can read, analyze, and critique. • Free CD-ROM: The Study Guide also includes a CD-ROM providing hundreds of review questions to assist students in self-testing. This review program provides a rationale for both correct and incorrect answers, helping students to identify areas of strength and areas needing further study. • The Instructor’s Resource CD-ROM includes a chapter corresponding to every chapter in the textbook. Each chapter of the Instructor’s Manual contains a statement of intent, student objectives, new terms in the chapter, comments on selected research examples in the textbook, answers to certain Study Guide exercises, and test questions (true/false and multiple choice) and answers. New to this edition are PowerPoint slides summarizing key points in each chapter, and test questions have been placed into a program that allows instructors to automatically generate a test complete with instructions and an answer key. A gradebook is also included in this program. It is our hope that the content, style, and organization of this book continue to meet the needs of a broad spectrum of nursing students and nurse researchers. We also hope that the book will help to foster enthusiasm for the kinds of discoveries that research can produce, and for the knowledge that will help support an evidence-based nursing practice. DENISE F. POLIT,

PHD

CHERYL TATANO BECK,

DNSC, CNM, FAAN

CONTENTS

Part

1

Foundations of Nursing Research Chapter 1 Introduction to Nursing Research

Major Steps in a Quantitative Study 47 Activities in a Qualitative Study 55 Research Examples 58 Summary Points 60

1

Part

3

Nursing Research in Perspective 3 Nursing Research: Past, Present, and Future Sources of Evidence for Nursing Practice Paradigms for Nursing Research 13 The Purposes of Nursing Research 18 Research Examples 21 Summary Points 22

Conceptualizing a Research Study 5 11

Chapter 2 Key Concepts and Terms in Qualitative and Quantitative Research 26 The Faces and Places of Research 26 The Building Blocks of a Study 29 Key Challenges of Conducting Research Research Examples 41 Summary Points 42

35

Chapter 3 Overview of the Research Process in Qualitative and Quantitative Studies 46 Major Classes of Quantitative and Qualitative Research 46

2

63

Chapter 4 Research Problems, Research Questions, and Hypotheses 65 Overview of Research Problems 65 Sources of Research Problems 67 Development and Refinement of Research Problems 69 Communicating the Research Problem Research Hypotheses 77 Research Examples 84 Summary Points 85 Chapter 5 Reviewing the Literature

73

88

Purposes of a Literature Review 88 Scope of a Literature Search 89 Locating Relevant Literature for a Research Review 91 Reading Research Reports 96 Preparing a Written Literature Review

103 xiii

xiv

Contents

Research Examples of Research Literature Reviews 109 Summary points 111 Chapter 6 Developing a Conceptual Context

114

Theories, Models, and Frameworks 114 The Nature of Theories and Conceptual Models 119 Conceptual Models Used in Nursing Research 120 Testing, Using, and Developing a Theory or Framework 125 Research Examples 133 Summary Points 134

Part

3

Designs for Nursing Research Chapter 7 Designing Ethical Research

139 141

The Need for Ethical Guidelines 141 The Principle of Beneficence 143 The Principle of Respect for Human Dignity 147 The Principle of Justice 149 Informed Consent 150 Vulnerable Subjects 154 External Reviews and the Protection of Human Rights 156 Building Ethics into the Design of the Study 157 Research Examples 158 Summary Points 159 Chapter 8 Designing Quantitative Studies Aspects of Quantitative Research Design 162 Overview of Research Design Types Experiments 168

Quasi-Experiments 181 Nonexperimental Research 188 Designs and Research Evidence 195 Summary Points 196 Chapter 9 Enhancing Rigor in Quantitative Research 201 Controlling the Research Situation 201 Controlling Intrinsic Subject Factors 203 Characteristics of Good Design 209 Research Example 219 Summary Points 220 Chapter 10 Quantitative Research for Various Purposes 223 Studies That Are Typically Experimental or Quasi-Experimental 223 Studies That Can Be Either Experimental or Nonexperimental 230 Studies That Are Typically Nonexperimental Research Examples of Various Types of Quantitative Studies 239 Summary Points 240

234

Chapter 11 Qualitative Research Design and Approaches 245 The Design of Qualitative Studies 245 Qualitative Research Traditions 248 Other Types of Qualitative Research 259 Research With Ideological Perspectives 263 Research Examples 266 Summary Points 268

162

164

Chapter 12 Integration of Qualitative and Quantitative Designs 273 Rationale for Multimethod Research 273 Applications of Multimethod Research 275 Multimethod Research Designs 279 Strategies for Multimethod Research 280

Contents

Obstacles to Multimethod Research Research Examples 284 Summary Points 286 Chapter 13 Sampling Designs

289

Basic Sampling Concepts in Quantitative Studies 289 Nonprobability Sampling 292 Probability Sampling 295 Sample Size in Quantitative Studies Implementing a Sampling Plan in Quantitative Studies 303 Sampling in Qualitative Research Research Examples 310 Summary Points 311

Part

283

305

4

Measurement and Data Collection

315

Chapter 14 Designing and Implementing a Data Collection Plan 317 Existing Data Versus Original Data 317 Dimensions of Data Collection Approaches 318 Major Types of Data Collection Methods 319 Converting Quantitative and Qualitative Data 322 Developing a Data Collection Plan in a Quantitative Study 323 Implementing a Data Collection Plan in a Quantitative Study 330 Data Collection Issues in Qualitative Studies 332 Research Examples 336 Summary Points 337 Chapter 15 Collecting Self-Report Data

340

Qualitative Self-Report Techniques Quantitative Self-Report Instruments

340 349

xv

Using and Preparing Structured Self-Report Instruments 352 Administering Structured Self-Report Instruments 365 Research Examples 368 Summary Points 369 Chapter 16 Collecting Observational Data

300

Observational Issues 375 Qualitative Observational Methods: Participant Observation 378 Observational Methods: Structured Observations 385 Mechanical Aids in Observations Observer Biases 391 Research Examples 392 Summary Points 393

375

390

Chapter 17 Collecting Biophysiologic and Other Data 398 Biophysiologic Measures 398 Records, Documents, and Available Data 402 Q Methodology 403 Projective Techniques 405 Vignettes 407 Cognitive and Neuropsychological Tests 408 Examples of Studies Using Alternative Data Collection Methods 409 Summary Points 410 Chapter 18 Assessing Data Quality

413

Measurement 413 Reliability of Measuring Instruments 416 Validity 422 Other Criteria for Assessing Quantitative Measures 428 Assessment of Qualitative Data and Their Interpretation 430 Research Examples 437 Summary Points 443

xvi

Part

Contents

5

The Analysis of Research Data

449

Chapter 19 Analyzing Quantitative Data: Descriptive Statistics 451 Levels of Measurement 451 Frequency Distributions 455 Central Tendency 459 Variability 460 Bivariate Descriptive Statistics: Contingency Tables and Correlation 465 The Computer and Descriptive Statistics 469 Research Example 473 Summary Points 475 Chapter 20 Analyzing Quantitative Data: Inferential Statistics 477 Sampling Distributions 477 Estimation of Parameters 479 Hypothesis Testing 480 Testing Differences Between Two Group Means 486 Testing Differences Between Three or More Group Means 489 Testing Differences in Proportions 493 Testing Relationships Between Two Variables 494 Power Analysis 495 The Computer and Inferential Statistics 502 Guide to Bivariate Statistical Tests 505 Research Example 507 Summary Points 507 Chapter 21 Analyzing Quantitative Data: Multivariate Statistics 511 Simple Linear Regression 511 Multiple Linear Regression 514 Analysis of Covariance 521 Factor Analysis 526

Other Least-Squares Multivariate Techniques 530 Causal Modeling 532 Other Multivariate Statistical Procedures 536 The Computer and Multivariate Statistics 538 Guide to Multivariate Statistical Tests 540 Research Example 543 Summary Points 543 Chapter 22 Designing and Implementing a Quantitative Analysis Strategy 547 Phases in the Analysis of Quantitative Data 547 Preanalysis Phase 547 Preliminary Assessments and Actions Principal Analyses 560 Interpretation of Results 562 Research Example 566 Summary Points 568 Chapter 23 Analyzing Qualitative Data

570

Introduction to Qualitative Analysis Qualitative Data Management and Organization 572 Analytic Procedures 578 Interpretation of Qualitative Findings Research Examples 592 Summary Points 594

Part

553

570

6

Communicating Research 599 Chapter 24 Summarizing and Sharing Research Findings 601 Getting Started on Dissemination 601 Content of Research Reports 604 The Style of Research Reports 617 Types of Research Reports 618 Summary Points 626

591

Contents

Chapter 25 Writing a Research Proposal

629

Overview of Research Proposals 629 Proposals for Theses and Dissertations 635 Funding for Research Proposals 637 Grant Applications to the National Institutes of Health 639 Research Examples 647 Summary Points 648

Part

7

Using Research Results 651 Chapter 26 Evaluating Research Reports The Research Critique 653 Elements of a Research Critique

653 654

xvii

Conclusion 666 Research Examples 667 Summary Points 669 Chapter 27 Utilizing Research: Putting Research Evidence Into Nursing Practice 671 Research Utilization Versus Evidence-Based Practice 671 Barriers to Using Research in Nursing Practice 677 The Process of Using Research in Nursing Practice 681 Research Integration and Synthesis 690 Research Example 696 Summary Points 697 Appendix Glossary Index

REVIEWERS

Eileen Chasens, DSN, RN

Virginia Nehring, PhD, RN

Assistant Professor Wayne State University Detroit, Michigan

Associate Professor Wright State University Dayton, Ohio

Celia R. Colon-Rivera, RN, PhD

Marlene Reimer, RN, PhD, CNN(C)

Professor University of Puerto Rico-Mayaguez Campus Mayaguez, Puerto Rico

Carol Cornwell, PhD, MS, CS

Director of Nursing Research and Assistant Professor of Nursing Georgia Southern University School of Nursing Statesboro, Georgia

Linda Goodfellow, RN, MNEd Assistant Professor of Nursing Duquesne University Pittsburgh, Pennsylvania

Janice S. Hayes, PhD, RN

Associate Professor Florida Atlantic University College of Nursing Davie, Florida

Joan R. S. McDowell, MN Lecturer University of Glaston Nursing & Midwifery School Glasgow, Scotland

Associate Professor Faculty of Nursing University of Calgary Calgary, Alberta, Canada

Beth L. Rodgers, PhD, RN

Professor and Chair, Foundations of Nursing Department University of Wisconsin-Milwaukee School of Nursing Milwaukee, Wisconsin

Janet S. Secrest, PhD, RN

Assistant Professor University of Tennessee at Chattanooga School of Nursing Chattanooga, Tennessee

Gloria Weber, PhD, RN, CNAA Associate Professor The University of Texas at Tyler College of Nursing Tyler, Texas

xi

APPENDIX

A

Statistical Tables

TABLE A-1 Distribution of t Probability LEVEL OF SIGNIFICANCE FOR ONE-TAILED TEST .10

.05

.025 .01 .005 LEVEL OF SIGNIFICANCE FOR TWO-TAILED TEST

.0005

df

.20

.10

.05

.02

.01

.001

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 40 60 120 ∞

3.078 1.886 1.638 1.533 1.476 1.440 1.415 1.397 1.383 1.372 1.363 1.356 1.350 1.345 1.341 1.337 1.333 1.330 1.328 1.325 1.323 1.321 1.319 1.318 1.316 1.315 1.314 1.313 1.311 1.310 1.303 1.296 1.289 1.282

6.314 2.920 2.353 2.132 2.015 1.953 1.895 1.860 1.833 1.812 1.796 1.782 1.771 1.761 1.753 1.746 1.740 1.734 1.729 1.725 1.721 1.717 1.714 1.711 1.708 1.706 1.703 1.701 1.699 1.697 1.684 1.671 1.658 1.645

12.706 4.303 3.182 2.776 2.571 2.447 2.365 2.306 2.262 2.228 2.201 2.179 2.160 2.145 2.131 2.120 2.110 2.101 2.093 2.086 2.080 2.074 2.069 2.064 2.060 2.056 2.052 2.048 2.045 2.042 2.021 2.000 1.980 1.960

31.821 6.965 4.541 3.747 3.376 3.143 2.998 2.896 2.821 2.765 2.718 2.681 2.650 2.624 2.602 2.583 2.567 2.552 2.539 2.528 2.518 2.508 2.500 2.492 2.485 2.479 2.473 2.467 2.462 2.457 2.423 2.390 2.358 2.326

63.657 9.925 5.841 4.604 4.032 3.707 3.449 3.355 3.250 3.169 3.106 3.055 3.012 2.977 2.947 2.921 2.898 2.878 2.861 2.845 2.831 2.819 2.807 2.797 2.787 2.779 2.771 2.763 2.756 2.750 2.704 2.660 2.617 2.576

636.619 31.598 12.941 8.610 6.859 5.959 5.405 5.041 4.781 4.587 4.437 4.318 4.221 4.140 4.073 4.015 3.965 3.922 3.883 3.850 3.819 3.792 3.767 3.745 3.725 3.707 3.690 3.674 3.659 3.646 3.551 3.460 3.373 3.291

706

APPENDIX A

TABLE A-2 Significant Values of F ! = .05 (Two-Tailed)

! = .025 (one-tailed)

dfB dfW

1

2

3

4

5

6

8

12

24

1 2 3 4 5

161.4 18.51 10.13 7.71 6.61

199.5 19.00 9.55 6.94 5.79

215.7 19.16 9.28 6.59 5.41

224.6 19.25 9.12 6.39 5.19

230.2 19.30 9.01 6.26 5.05

234.0 19.33 8.94 6.16 4.95

238.9 19.37 8.84 6.04 4.82

243.9 19.41 8.74 5.91 4.68

249.0 19.45 8.64 5.77 4.53

254.3 19.50 8.53 5.63 4.36

6 7 8 9 10

5.99 5.59 5.32 5.12 4.96

5.14 4.74 4.46 4.26 4.10

4.76 4.35 4.07 3.86 3.71

4.53 4.12 3.84 3.63 3.48

4.39 3.97 3.69 3.48 3.33

4.28 3.87 3.58 3.37 3.22

4.15 3.73 3.44 3.23 3.07

4.00 3.57 3.28 3.07 2.91

3.84 3.41 3.12 2.90 2.74

3.67 3.23 2.93 2.71 2.54

11 12 13 14 15

4.84 4.75 4.67 4.60 4.54

3.98 3.88 3.80 3.74 3.68

3.59 3.49 3.41 3.34 3.29

3.36 3.26 3.18 3.11 3.06

3.20 3.11 3.02 2.96 2.90

3.09 3.00 2.92 2.85 2.79

2.95 2.85 2.77 2.70 2.64

2.79 2.69 2.60 2.53 2.48

2.61 2.50 2.42 2.35 2.29

2.40 2.30 2.21 2.13 2.07

16 17 18 19 20

4.49 4.45 4.41 4.38 4.35

3.63 3.59 3.55 3.52 3.49

3.24 3.20 3.16 3.13 3.10

3.01 2.96 2.93 2.90 2.87

2.85 2.81 2.77 2.74 2.71

2.74 2.70 2.66 2.63 2.60

2.59 2.55 2.51 2.48 2.45

2.42 2.38 2.34 2.31 2.28

2.24 2.19 2.15 2.11 2.08

2.01 1.96 1.92 1.88 1.84

21 22 23 24 25

4.32 4.30 4.28 4.26 4.24

3.47 3.44 3.42 3.40 3.38

3.07 3.05 3.03 3.01 2.99

2.84 2.82 2.80 2.78 2.76

2.68 2.66 2.64 2.62 2.60

2.57 2.55 2.53 2.51 2.49

2.42 2.40 2.38 2.36 2.34

2.25 2.23 2.20 2.18 2.16

2.05 2.03 2.00 1.98 1.96

1.81 1.78 1.76 1.73 1.71

26 27 28 29 30

4.22 4.21 4.20 4.18 4.17

3.37 3.35 3.34 3.33 3.32

2.98 2.96 2.95 2.93 2.92

2.74 2.73 2.71 2.70 2.69

2.59 2.57 2.56 2.54 2.53

2.47 2.46 2.44 2.43 2.42

2.32 2.30 2.29 2.28 2.27

2.15 2.13 2.12 2.10 2.09

1.95 1.93 1.91 1.90 1.89

1.69 1.67 1.65 1.64 1.62

40 60 120 ∞

4.08 4.00 3.92 3.84

3.23 3.15 3.07 2.99

2.84 2.76 2.68 2.60

2.61 2.52 2.45 2.37

2.45 2.37 2.29 2.21

2.34 2.25 2.17 2.09

2.18 2.10 2.02 1.94

2.00 1.92 1.83 1.75

1.79 1.70 1.61 1.52

1.51 1.39 1.25 1.00

(continued)

APPENDIX A

TABLE A-2 Significant Values of F (continued) !=.01 (Two-Tailed)

707

! = .005(one-tailed)

dfB dfW

1

2

3

4

5

6

8

12

24

1 2 3 4 5

4052 98.49 34.12 21.20 16.26

4999 99.00 30.81 18.00 13.27

5403 99.17 29.46 16.69 12.06

5625 99.25 28.71 15.98 11.39

5764 99.30 28.24 15.52 10.97

5859 99.33 27.91 15.21 10.67

5981 99.36 27.49 14.80 10.29

6106 99.42 27.05 14.37 9.89

6234 99.46 26.60 13.93 9.47

6366 99.50 26.12 13.46 9.02

6 7 8 9 10

13.74 12.25 11.26 10.56 10.04

10.92 9.55 8.65 8.02 7.56

9.78 8.45 7.59 6.99 6.55

9.15 7.85 7.01 6.42 5.99

8.75 7.46 6.63 6.06 5.64

8.47 7.19 6.37 5.80 5.39

8.10 6.84 6.03 5.47 5.06

7.72 6.47 5.67 5.11 4.71

7.31 6.07 5.28 4.73 4.33

6.88 5.65 4.86 4.31 3.91

11 12 13 14 15

9.65 9.33 9.07 8.86 8.68

7.20 6.93 6.70 6.51 6.36

6.22 5.95 5.74 5.56 5.42

5.67 5.41 5.20 5.03 4.89

5.32 5.06 4.86 4.69 4.56

5.07 4.82 4.62 4.46 4.32

4.74 4.50 4.30 4.14 4.00

4.40 4.16 3.96 3.80 3.67

4.02 3.78 3.59 3.43 3.29

3.60 3.36 3.16 3.00 2.87

16 17 18 19 20

8.53 8.40 8.28 8.18 8.10

6.23 6.11 6.01 5.93 5.85

5.29 5.18 5.09 5.01 4.94

4.77 4.67 4.58 4.50 4.43

4.44 4.34 4.29 4.17 4.10

4.20 4.10 4.01 3.94 3.87

3.89 3.78 3.71 3.63 3.56

3.55 3.45 3.37 3.30 3.23

3.18 3.08 3.00 2.92 2.86

2.75 2.65 2.57 2.49 2.42

21 22 23 24 25

8.02 7.94 7.88 7.82 7.77

5.78 5.72 5.66 5.61 5.57

4.87 4.82 4.76 4.72 4.68

4.37 4.31 4.26 4.22 4.18

4.04 3.99 3.94 3.90 3.86

3.81 3.76 3.71 3.67 3.63

3.51 3.45 3.41 3.36 3.32

3.17 3.12 3.07 3.03 2.99

2.80 2.75 2.70 2.66 2.62

2.36 2.31 2.26 2.21 2.17

26 27 28 29 30

7.72 7.68 7.64 7.60 7.56

5.53 5.49 5.45 5.42 5.39

4.64 4.60 4.57 4.54 4.51

4.14 4.11 4.07 4.04 4.02

3.82 3.78 3.75 3.73 3.70

3.59 3.56 3.53 3.50 3.47

3.29 3.26 3.23 3.20 3.17

2.96 2.93 2.90 2.87 2.84

2.58 2.55 2.52 2.49 2.47

2.13 2.10 2.06 2.03 2.01

40 60 120 ∞

7.31 7.08 6.85 6.64

5.18 4.98 4.79 4.60

4.31 4.13 3.95 3.78

3.83 3.65 3.48 3.32

3.51 3.34 3.17 3.02

3.29 3.12 2.96 2.80

2.99 2.82 2.66 2.51

2.66 2.50 2.34 2.18

2.29 2.12 1.95 1.79

1.80 1.60 1.38 1.00

(continued)

708

APPENDIX A

TABLE A-2 Significant Values of F (continued) ! = .001 (Two-Tailed)

! = .0005 (one-tailed)

dfB dfW

1

2

3

4

5

6

8

12

24

1 405284 500000 540379 562500 576405 585937 598144 610667 623497 636619 2 998.5 999.0 999.2 999.2 999.3 999.3 999.4 999.4 999.5 999.5 3 167.5 148.5 141.1 137.1 134.6 132.8 130.6 128.3 125.9 123.5 4 74.14 61.25 56.18 53.44 51.71 50.53 49.00 47.41 45.77 44.05 5 47.04 36.61 33.20 31.09 29.75 28.84 27.64 26.42 25.14 23.78 6 7 8 9 10

35.51 29.22 25.42 22.86 21.04

27.00 21.69 18.49 16.39 14.91

23.70 18.77 15.83 13.90 12.55

21.90 17.19 14.39 12.56 11.28

20.81 16.21 13.49 11.71 10.48

20.03 15.52 12.86 11.13 9.92

19.03 14.63 17.04 10.37 9.20

17.99 13.71 11.19 9.57 8.45

16.89 12.73 10.30 8.72 7.64

15.75 11.69 9.34 7.81 6.76

11 12 13 14 15

19.69 18.64 17.81 17.14 16.59

13.81 12.97 12.31 11.78 11.34

11.56 10.80 10.21 9.73 9.34

10.35 9.63 9.07 8.62 8.25

9.58 8.89 8.35 7.92 7.57

9.05 8.38 7.86 7.43 7.09

8.35 7.71 7.21 6.80 6.47

7.63 7.00 6.52 6.13 5.81

6.85 6.25 5.78 5.41 5.10

6.00 5.42 4.97 4.60 4.31

16 17 18 19 20

16.12 15.72 15.38 15.08 14.82

10.97 10.66 10.39 10.16 9.95

9.00 8.73 8.49 8.28 8.10

7.94 7.68 7.46 7.26 7.10

7.27 7.02 6.81 6.61 6.46

6.81 6.56 6.35 6.18 6.02

6.19 5.96 5.76 5.59 5.44

5.55 5.32 5.13 4.97 4.82

4.85 4.63 4.45 4.29 4.15

4.06 3.85 3.67 3.52 3.38

21 22 23 24 25

14.59 14.38 14.19 14.03 13.88

9.77 9.61 9.47 9.34 9.22

7.94 7.80 7.67 7.55 7.45

6.95 6.81 6.69 6.59 6.49

6.32 6.19 6.08 5.98 5.88

5.88 5.76 5.65 5.55 5.46

5.31 5.19 5.09 4.99 4.91

4.70 4.58 4.48 4.39 4.31

4.03 3.92 3.82 3.74 3.66

3.26 3.15 3.05 2.97 2.89

26 27 28 29 30

13.74 13.61 13.50 13.39 13.29

9.12 9.02 8.93 8.85 8.77

7.36 7.27 7.19 7.12 7.05

6.41 6.33 6.25 6.19 6.12

5.80 5.73 5.66 5.59 5.53

5.38 5.31 5.24 5.18 5.12

4.83 4.76 4.69 4.64 4.58

4.24 4.17 4.11 4.05 4.00

3.59 3.52 3.46 3.41 3.36

2.82 2.75 2.70 2.64 2.59

40 60 120 ∞

12.61 11.97 11.38 10.83

8.25 7.76 7.31 6.91

6.60 6.17 5.79 5.42

5.70 5.31 4.95 4.62

5.13 4.76 4.42 4.10

4.73 4.37 4.04 3.74

4.21 3.87 3.55 3.27

3.64 3.31 3.02 2.74

3.01 2.69 2.40 2.13

2.23 1.90 1.56 1.00

APPENDIX A

TABLE A-3 Distribution of "2 Probability LEVEL OF SIGNIFICANCE df

.10

.05

.02

.01

.001

1 2 3 4 5

2.71 4.61 6.25 7.78 9.24

3.84 5.99 7.82 9.49 11.07

5.41 7.82 9.84 11.67 13.39

6.63 9.21 11.34 13.28 15.09

10.83 13.82 16.27 18.46 20.52

6 7 8 9 10

10.64 12.02 13.36 14.68 15.99

12.59 14.07 15.51 16.92 18.31

15.03 16.62 18.17 19.68 21.16

16.81 18.48 20.09 21.67 23.21

22.46 24.32 26.12 27.88 29.59

11 12 13 14 15

17.28 18.55 19.81 21.06 22.31

19.68 21.03 22.36 23.68 25.00

22.62 24.05 25.47 26.87 28.26

24.72 26.22 27.69 29.14 30.58

31.26 32.91 34.53 36.12 37.70

16 17 18 19 20

23.54 24.77 25.99 27.20 28.41

26.30 27.59 28.87 30.14 31.41

29.63 31.00 32.35 33.69 35.02

32.00 33.41 34.81 36.19 37.57

39.25 40.79 42.31 43.82 45.32

21 22 23 24 25

29.62 30.81 32.01 33.20 34.38

32.67 33.92 35.17 36.42 37.65

36.34 37.66 38.97 40.27 41.57

38.93 40.29 41.64 42.98 44.31

46.80 48.27 49.73 51.18 52.62

26 27 28 29 30

35.56 36.74 37.92 39.09 40.26

38.89 40.11 41.34 42.56 43.77

42.86 44.14 45.42 46.69 47.96

45.64 46.96 48.28 49.59 50.89

54.05 55.48 56.89 58.30 59.70

709

710

APPENDIX A

TABLE A-4 Significant Values of the Correlation Coefficient LEVEL OF SIGNIFICANCE FOR ONE-TAILED TEST .05 df

.025 .01 .005 .0005 LEVEL OF SIGNIFICANCE FOR TWO-TAILED TEST

.10

.05

.02

.01

.001

1 2 3 4 5

.98769 .90000 .8054 .7293 .6694

.99692 .95000 .8783 .8114 .7545

.999507 .98000 .93433 .8822 .8329

.999877 .990000 .95873 .91720 .8745

.9999988 .99900 .99116 .97406 .95074

6 7 8 9 10

.6215 .5822 .5494 .5214 .4973

.7067 .6664 .6319 .6021 .5760

.7887 .7498 .7155 .6851 .6581

.8343 .7977 .7646 .7348 .7079

.92493 .8982 .8721 .8471 .8233

11 12 13 14 15

.4762 .4575 .4409 .4259 .4124

.5529 .5324 .5139 .4973 .4821

.6339 .6120 .5923 .5742 .5577

.6835 .6614 .5411 .6226 .6055

.8010 .7800 .7603 .7420 .7246

16 17 18 19 20

.4000 .3887 .3783 .3687 .3598

.4683 .4555 .4438 .4329 .4227

.5425 .5285 .5155 .5034 .4921

.5897 .5751 .5614 .5487 .5368

.7084 .6932 .5687 .6652 .6524

25 30 35 40 45

.3233 .2960 .2746 .2573 .2428

.3809 .3494 .3246 .3044 .2875

.4451 .4093 .3810 .3578 .3384

.5869 .4487 .4182 .3932 .3721

.5974 .5541 .5189 .4896 .4648

50 60 70 80 90 100

.2306 .2108 .1954 .1829 .1726 .1638

.2732 .2500 .2319 .2172 .2050 .1946

.3218 .2948 .2737 .2565 .2422 .2301

.3541 .3248 .3017 .2830 .2673 .2540

.4433 .4078 .3799 .3568 .3375 .3211

PA R T

1

Foundations of Nursing Research

1

Introduction to Nursing Research

NURSING RESEARCH IN PERSPECTIVE It is an exciting—and challenging—time to be a nurse. Nurses are managing their clinical responsibilities at a time when the nursing profession and the larger health care system require an extraordinary range of skills and talents of them. Nurses are expected to deliver the highest possible quality of care in a compassionate manner, while also being mindful of costs. To accomplish these diverse (and sometimes conflicting) goals, nurses must access and evaluate extensive clinical information, and incorporate it into their clinical decision-making. In today’s world, nurses must become lifelong learners, capable of reflecting on, evaluating, and modifying their clinical practice based on new knowledge. And, nurses are increasingly expected to become producers of new knowledge through nursing research. What Is Nursing Research? Research is systematic inquiry that uses disciplined methods to answer questions or solve problems. The ultimate goal of research is to develop, refine, and expand a body of knowledge. Nurses are increasingly engaged in disciplined studies that benefit the profession and its patients,

and that contribute to improvements in the entire health care system. Nursing research is systematic inquiry designed to develop knowledge about issues of importance to the nursing profession, including nursing practice, education, administration, and informatics. In this book, we emphasize clinical nursing research, that is, research designed to generate knowledge to guide nursing practice and to improve the health and quality of life of nurses’ clients. Nursing research has experienced remarkable growth in the past three decades, providing nurses with an increasingly sound base of knowledge from which to practice. Yet as we proceed into the 21st century, many questions endure and much remains to be done to incorporate research-based knowledge into nursing practice. Examples of nursing research questions: • What are the factors that determine the length of stay of patients in the intensive care unit undergoing coronary artery bypass graft surgery (Doering, Esmailian, Imperial-Perez, & Monsein, 2001)? • How do adults with acquired brain injury perceive their social interactions and relationships (Paterson & Stewart, 2002)?

4

PART 1 Foundations of Nursing Research

The Importance of Research in Nursing Nurses increasingly are expected to adopt an evidence-based practice (EBP), which is broadly defined as the use of the best clinical evidence in making patient care decisions. Although there is not a consensus about what types of “evidence” are appropriate for EBP (Goode, 2000), there is general agreement that research findings from rigorous studies constitute the best type of evidence for informing nurses’ decisions, actions, and interactions with clients. Nurses are accepting the need to base specific nursing actions and decisions on evidence indicating that the actions are clinically appropriate, cost-effective, and result in positive outcomes for clients. Nurses who incorporate high-quality research evidence into their clinical decisions and advice are being professionally accountable to their clients. They are also reinforcing the identity of nursing as a profession. Another reason for nurses to engage in and use research involves the spiraling costs of health care and the cost-containment practices being instituted in health care facilities. Now, more than ever, nurses need to document the social relevance and effectiveness of their practice, not only to the profession but to nursing care consumers, health care administrators, third-party payers (e.g., insurance companies), and government agencies. Some research findings will help eliminate nursing actions that do not achieve desired outcomes. Other findings will help nurses identify practices that improve health care outcomes and contain costs as well. Nursing research is essential if nurses are to understand the varied dimensions of their profession. Research enables nurses to describe the characteristics of a particular nursing situation about which little is known; to explain phenomena that must be considered in planning nursing care; to predict the probable outcomes of certain nursing decisions; to control the occurrence of undesired outcomes; and to initiate activities to promote desired client behavior.

Example of an EBP project: • The Association of Women’s Health, Obstetric, and Neonatal Nurses (AWHONN) is one nursing organization that has demonstrated a strong commitment to evidence-based nursing practice. For example, AWHONN undertook a project that developed and tested an evidencebased protocol for urinary incontinence in women, and then designed procedures to facilitate the protocol’s implementation into clinical practice (Samselle et al., 2000a, 2000b). More recently, AWHONN and the National Association of Neonatal Nurses designed and tested an evidence-based protocol for neonatal skin care, and also instituted procedures for implementing it (Lund, Kuller, Lane, Lott, Raines, & Thomas, 2001; Lund, Osborne, Kuller, Lane, Lott, & Raines, 2001). The Consumer–Producer Continuum in Nursing Research With the current emphasis on EBP, it has become every nurse’s responsibility to engage in one or more roles along a continuum of research participation. At one end of the continuum are those nurses whose involvement in research is indirect. Consumers of nursing research read research reports to develop new skills and to keep up to date on relevant findings that may affect their practice. Nurses increasingly are expected to maintain this level of involvement with research, at a minimum. Research utilization—the use of research findings in a practice setting— depends on intelligent nursing research consumers. At the other end of the continuum are the producers of nursing research: nurses who actively participate in designing and implementing research studies. At one time, most nurse researchers were academics who taught in schools of nursing, but research is increasingly being conducted by practicing nurses who want to find what works best for their patients. Between these two end points on the continuum lie a rich variety of research activities in which nurses engage as a way of improving their effec-

CHAPTER 1 Introduction to Nursing Research

tiveness and enhancing their professional lives. These activities include the following: • Participating in a journal club in a practice setting, which involves regular meetings among nurses to discuss and critique research articles • Attending research presentations at professional conferences • Discussing the implications and relevance of research findings with clients • Giving clients information and advice about participation in studies • Assisting in the collection of research information (e.g., distributing questionnaires to patients) • Reviewing a proposed research plan with respect to its feasibility in a clinical setting and offering clinical expertise to improve the plan • Collaborating in the development of an idea for a clinical research project • Participating on an institutional committee that reviews the ethical aspects of proposed research before it is undertaken • Evaluating completed research for its possible use in practice, and using it when appropriate In all these activities, nurses with some research skills are in a better position than those without them to make a contribution to nursing knowledge. An understanding of nursing research can improve the depth and breadth of every nurse’s professional practice. NURSING RESEARCH: PA S T, P R E S E N T, A N D FUTURE Although nursing research has not always had the prominence and importance it enjoys today, its long and interesting history portends a distinguished future. Table 1-1 summarizes some of the key events in the historical evolution of nursing research. The Early Years: From Nightingale to the 1950s Most people would agree that research in nursing began with Florence Nightingale. Her landmark

5

publication, Notes on Nursing (1859), describes her early interest in environmental factors that promote physical and emotional well-being—an interest that continues among nurses nearly 150 years later. Nightingale’s most widely known research contribution involved her data collection and analysis relating to factors affecting soldier mortality and morbidity during the Crimean War. Based on her skillful analyses and presentations, she was successful in effecting some changes in nursing care— and, more generally, in public health. For many years after Nightingale’s work, the nursing literature contained little research. Some attribute this absence to the apprenticeship nature of nursing. The pattern of nursing research that eventually emerged at the turn of the century was closely aligned to the problems confronting nurses. Most studies conducted between 1900 and 1940 concerned nurses’ education. For example, in 1923, a group called the Committee for the Study of Nursing Education studied the educational preparation of nurse teachers, administrators, and public health nurses and the clinical experiences of nursing students. The committee issued what has become known as the Goldmark Report, which identified many inadequacies in the educational backgrounds of the groups studied and concluded that advanced educational preparation was essential. As more nurses received university-based education, studies concerning nursing students—their differential characteristics, problems, and satisfactions— became more numerous. During the 1940s, studies concerning nursing education continued, spurred on by the unprecedented demand for nursing personnel that resulted from World War II. For example, Brown (1948) reassessed nursing education in a study initiated at the request of the National Nursing Council for War Service. The findings from the study, like those of the Goldmark Report, revealed numerous inadequacies in nursing education. Brown recommended that the education of nurses occur in collegiate settings. Many subsequent research investigations concerning the functions performed by nurses, nurses’ roles and attitudes, hospital environments, and nurse—patient interactions stemmed from the Brown report.

TABLE 1.1 Historical Landmarks Affecting Nursing Research YEAR

EVENT

1859

Nightingale’s Notes on Nursing published

1900

American Nursing Journal begins publication

1923

Columbia University establishes first doctoral program for nurses Goldmark Report with recommendations for nursing education published

1930s

American Journal of Nursing publishes clinical cases studies

1948

Brown publishes report on inadequacies of nursing education

1952

The journal Nursing Research begins publication

1955

Inception of the American Nurses’ Foundation to sponsor nursing research

1957

Establishment of nursing research center at Walter Reed Army Institute of Research

1963

International Journal of Nursing Studies begins publication

1965

American Nurses’ Association (ANA) begins sponsoring nursing research conferences

1966

Nursing history archive established at Mugar Library, Boston University

1968

Canadian Journal of Nursing Research begins publication

1971

ANA establishes a Commission on Research

1972

ANA establishes its Council of Nurse Researchers

1976

Stetler and Marram publish guidelines on assessing research for use in practice

1978

The journals Research in Nursing & Health and Advances in Nursing Science begin publication

1979

Western Journal of Nursing Research begins publication

1982

The Conduct and Utilization of Research in Nursing (CURN) project publishes report

1983

Annual Review of Nursing Research begins publication

1985

ANA Cabinet on Nursing Research establishes research priorities

1986

National Center for Nursing Research (NCNR) established within U.S. National Institutes of Health

1987

The journal Scholarly Inquiry for Nursing Practice begins publication

1988

The journals Applied Nursing Research and Nursing Science Quarterly begin publication Conference on Research Priorities (CORP #1) in convened by NCNR

1989

U.S. Agency for Health Care Policy and Research (AHCPR) is established

1992

The journal Clinical Nursing Research begins publication

1993

NCNR becomes a full institute, the National Institute of Nursing Research (NINR) CORP #2 is convened to establish priorities for 1995–1999 The Cochrane Collaboration is established The journal Journal of Nursing Measurement begins publication

1994

The journal Qualitative Health Research begins publication

1997

Canadian Health Services Research Foundation is established with federal funding

1999

AHCPR is renamed Agency for Healthcare Research and Quality (AHRQ)

2000

NINR issues funding priorities for 2000–2004; annual funding exceeds $100 million The Canadian Institute of Health Research is launched The journal Biological Research for Nursing begins publication

6

CHAPTER 1 Introduction to Nursing Research

A number of forces combined during the 1950s to put nursing research on a rapidly accelerating upswing. An increase in the number of nurses with advanced educational degrees, the establishment of a nursing research center at the Walter Reed Army Institute of Research, an increase in the availability of funds from the government and private foundations, and the inception of the American Nurses’ Foundation—which is devoted exclusively to the promotion of nursing research—were forces providing impetus to nursing research during this period. Until the 1950s, nurse researchers had few outlets for reporting their studies to the nursing community. The American Journal of Nursing, first published in 1900, began on a limited basis to publish some studies in the 1930s. The increasing number of studies being conducted during the 1950s, however, created the need for a journal in which findings could be published; thus, Nursing Research came into being in 1952. Nursing research took a twist in the 1950s not experienced by research in other professions, at least not to the same extent as in nursing. Nurses studied themselves: Who is the nurse? What does the nurse do? Why do individuals choose to enter nursing? What are the characteristics of the ideal nurse? How do other groups perceive the nurse? Nursing Research in the 1960s Knowledge development through research in nursing began in earnest only about 40 years ago, in the 1960s. Nursing leaders began to express concern about the lack of research in nursing practice. Several professional nursing organizations, such as the Western Interstate Council for Higher Education in Nursing, established priorities for research investigations during this period. Practice-oriented research on various clinical topics began to emerge in the literature. The 1960s was the period during which terms such as conceptual framework, conceptual model, nursing process, and theoretical base of nursing practice began to appear in the literature and to influence views about the role of theory in nursing research. Funding continued to be available both

7

for the educational preparation of nurses and, increasingly, for nursing research. Nursing research began to advance worldwide in the 1960s. The International Journal of Nursing Studies began publication in 1963, and the Canadian Journal of Nursing Research was first published in 1968. Example of nursing research breakthroughs in the 1960s: • Jeanne Quint Benoliel began a program of research that had a major impact on medicine, medical sociology, and nursing. Quint explored the subjective experiences of patients after diagnosis with a life-threatening illness (1967). Of particular note, physicians in the early 1960s usually did not advise women that they had breast cancer, even after a mastectomy. Quint’s (1962, 1963) seminal study of the personal experiences of women after radical mastectomy contributed to changes in communication and information control by physicians and nurses. Nursing Research in the 1970s By the 1970s, the growing number of nurses conducting research studies and the discussions of theoretical and contextual issues surrounding nursing research created the need for additional communication outlets. Several additional journals that focus on nursing research were established in the 1970s, including Advances in Nursing Science, Research in Nursing & Health, the Western Journal of Nursing Research, and the Journal of Advanced Nursing. In the 1970s, there was a decided change in emphasis in nursing research from areas such as teaching, curriculum, and nurses themselves to the improvement of client care—signifying a growing awareness by nurses of the need for a scientific base from which to practice. Nursing leaders strongly endorsed this direction for nursing studies. Lindeman (1975), for example, conducted a study to ascertain the views of nursing leaders concerning the focus of nursing studies; clinical problems were identified as the highest priorities. Nurses also began to pay attention to the utilization of research findings

8

PART 1 Foundations of Nursing Research

in nursing practice. A seminal article by Stetler and Marram (1976) offered guidance on assessing research for application in practice settings. In the United States, research skills among nurses continued to improve in the 1970s. The cadre of nurses with earned doctorates steadily increased, especially during the later 1970s. The availability of both predoctoral and postdoctoral research fellowships facilitated the development of advanced research skills. Example of nursing research breakthroughs in the 1970s: • Kathryn Barnard’s research led to breakthroughs in the area of neonatal and young child development. Her research program focused on the identification and assessment of children at risk of developmental and health problems, such as abused and neglected children and failure-to-thrive children (Barnard, 1973, 1976; Barnard & Collar, 1973; Barnard, Wenner, Weber, Gray, & Peterson, 1977). Her research contributed to work on early interventions for children with disabilities, and also to the field of developmental psychology. Nursing Research in the 1980s The 1980s brought nursing research to a new level of development. An increase in the number of qualified nurse researchers, the widespread availability of computers for the collection and analysis of information, and an ever-growing recognition that research is an integral part of professional nursing led nursing leaders to raise new issues and concerns. More attention was paid to the types of questions being asked, the methods of collecting and analyzing information being used, the linking of research to theory, and the utilization of research findings in practice. Several events provided impetus for nursing research in this decade. For example, the first volume of the Annual Review of Nursing Research was published in 1983. These annual reviews include summaries of current research knowledge on selected areas of research practice and encourage utilization of research findings.

Of particular importance in the United States was the establishment in 1986 of the National Center for Nursing Research (NCNR) at the National Institutes of Health (NIH) by congressional mandate, despite a presidential veto that was overridden largely as a result of nurse-scientists’ successful lobbying efforts. The purpose of NCNR was to promote—and financially support— research training and research projects relating to patient care. In addition, the Center for Research for Nursing was created in 1983 by the American Nurses’ Association. The Center’s mission is to develop and coordinate a research program to serve as the source of national information for the profession. Meanwhile, funding for nursing research became available in Canada in the 1980s through the National Health Research Development Program (NHRDP). Several nursing groups developed priorities for nursing research during the 1980s. For example, in 1985, the American Nurses’ Association Cabinet on Nursing Research established priorities that helped focus research more precisely on aspects of nursing practice. Also in the 1980s, nurses began to conduct formal projects designed to increase research utilization. Finally, specialty journals such as Heart & Lung and Cancer Nursing began to expand their coverage of research reports, and several new research-related journals were established: Applied Nursing Research, Scholarly Inquiry for Nursing Practice, and Nursing Science Quarterly. The journal Applied Nursing Research is notable for its intended audience: it includes research reports on studies of special relevance to practicing nurses. Several forces outside of the nursing profession in the late 1980s helped to shape today’s nursing research landscape. A group from the McMaster Medical School in Canada designed a clinical learning strategy that was called evidencebased medicine (EBM). EBM, which promulgated the view that scientific research findings were far superior to the opinions of authorities as a basis for clinical decisions, constituted a profound shift for medical education and practice, and has had a major effect on all health care professions.

CHAPTER 1 Introduction to Nursing Research

In 1989, the U.S. government established the Agency for Health Care Policy and Research (AHCPR). AHCPR (which was renamed the Agency for Healthcare Research and Quality, or AHRQ, in 1999) is the federal agency that has been charged with supporting research specifically designed to improve the quality of health care, reduce health costs, and enhance patient safety, and thus plays a pivotal role in the promulgation of EBP. Example of nursing research breakthroughs in the 1980s: • A team of researchers headed by Dorothy Brooten engaged in studies that led to the development and testing of a model of site transitional care. Brooten and her colleagues (1986, 1988, 1989), for example, conducted studies of nurse specialist–managed home follow-up services for very-low-birth-weight infants who were discharged early from the hospital, and later expanded to other high-risk patients (1994). The site transitional care model, which was developed in anticipation of government cost-cutting measures of the 1980s, has been used as a framework for patients who are at health risk as a result of early discharge from hospitals, and has been recognized by numerous health care disciplines. Nursing Research in the 1990s Nursing science came into its maturity during the 1990s. As but one example, nursing research was strengthened and given more national visibility in the United States when NCNR was promoted to full institute status within the NIH: in 1993, the National Institute of Nursing Research (NINR) was born. The birth of NINR helps put nursing research more into the mainstream of research activities enjoyed by other health disciplines. Funding for nursing research has also grown. In 1986, the NCNR had a budget of $16.2 million, whereas 16 years later in fiscal year 2002, the budget for NINR was over $120 million. Funding opportunities for nursing research expanded in other countries as well during the 1990s. For example,

9

the Canadian Health Services Research Foundation was established in 1997 with an endowment from federal funds, and plans for the Canadian Institute for Health Research were underway. Several research journals were established during the 1990s, including Qualitative Health Research, Clinical Nursing Research, Clinical Effectiveness, and Outcomes Management for Nursing Practice. These journals emerged in response to the growth in clinically oriented and indepth research among nurses, and interest in EBP. Another major contribution to EBP was inaugurated in 1993: the Cochrane Collaboration, an international network of institutions and individuals, maintains and updates systematic reviews of hundreds of clinical interventions to facilitate EBP. Some current nursing research is guided by priorities established by prominent nurse researchers in the 1990s, who were brought together by NCNR for two Conferences on Research Priorities (CORP). The priorities established by the first CORP for research through 1994 included low birth weight, human immunodeficiency virus (HIV) infection, long-term care, symptom management, nursing informatics, health promotion, and technology dependence. In 1993, the second CORP established the following research emphases for a portion of NINR’s funding from 1995 through 1999: developing and testing community-based nursing models; assessing the effectiveness of nursing interventions in HIV/AIDS; developing and testing approaches to remediating cognitive impairment; testing interventions for coping with chronic illness; and identifying biobehavioral factors and testing interventions to promote immunocompetence. Example of nursing research breakthroughs in the 1990s: • Many studies that Donaldson (2000) identified as breakthroughs in nursing research were conducted in the 1990s. This reflects, in part, the growth of research programs in which teams of researchers engage in a series of related research on important topics, rather than discrete and unconnected studies. As but one example,

10

PART 1 Foundations of Nursing Research

several nurse researchers had breakthroughs during the 1990s in the area of psychoneuroimmunology, which has been adopted as the model of mind—body interactions. Barbara Swanson and Janice Zeller, for example, conducted several studies relating to HIV infection and neuropsychological function (Swanson, Cronin-Stubbs, Zeller, Kessler, & Bielauskas, 1993; Swanson, Zeller, & Spear, 1998) that have led to discoveries in environmental management as a means of improving immune system status. Future Directions for Nursing Research Nursing research continues to develop at a rapid pace and will undoubtedly flourish in the 21st century. Broadly speaking, the priority for nursing research in the future will be the promotion of excellence in nursing science. Toward this end, nurse researchers and practicing nurses will be sharpening their research skills, and using those skills to address emerging issues of importance to the profession and its clientele. Certain trends for the beginning of the 21st century are evident from developments taking shape in the 1990s: • Increased focus on outcomes research. Outcomes research is designed to assess and document the effectiveness of health care services. The increasing number of studies that can be characterized as outcomes research has been stimulated by the need for cost-effective care that achieves positive outcomes without compromising quality. Nurses are increasingly engaging in outcomes research focused both on patients and on the overall delivery system. • Increased focus on biophysiologic research. Nurse researchers have begun increasingly to study biologic and physiologic phenomena as part of the effort to develop better clinical evidence. Consistent with this trend, a new journal called Biological Research for Nursing was launched in 2000. • Promotion of evidence-based practice. Concerted efforts to translate research findings

into practice will continue and nurses at all levels will be encouraged to engage in evidencebased patient care. In turn, improvements will be needed both in the quality of nursing studies and in nurses’ skills in understanding, critiquing, and using study results. • Development of a stronger knowledge base through multiple, confirmatory strategies. Practicing nurses cannot be expected to change a procedure or adopt an innovation on the basis of a single, isolated study. Confirmation is usually needed through the deliberate replication (i.e., the repeating) of studies with different clients, in different clinical settings, and at different times to ensure that the findings are robust. Replication in different settings is especially important now because the primary setting for health care delivery is shifting from inpatient hospitals to ambulatory settings, the community, and homes. Another confirmatory strategy is the conduct of multiple-site investigations by researchers in several locations. • Strengthening of multidisciplinary collaboration. Interdisciplinary collaboration of nurses with researchers in related fields (as well as intradisciplinary collaboration among nurse researchers) is likely to continue to expand in the 21st century as researchers address fundamental problems at the biobehavioral and psychobiologic interface. As one example, there are likely to be vast opportunities for nurses and other health care researchers to integrate breakthroughs in human genetics into lifestyle and health care interventions. In turn, such collaborative efforts could lead to nurse researchers playing a more prominent role in national and international health care policies. • Expanded dissemination of research findings. The Internet and other electronic communication have a big impact on the dissemination of research information, which in turn may help to promote EBP. Through such technological advances as online publishing (e.g., the Online Journal of Knowledge Synthesis for Nursing, the Online Journal of Clinical Innovation); on-line resources such as Lippincott’s NursingCenter.com; elec-

CHAPTER 1 Introduction to Nursing Research

tronic document retrieval and delivery; e-mail; and electronic mailing lists, information about innovations can be communicated more widely and more quickly than ever before. • Increasing the visibility of nursing research. The 21st century is likely to witness efforts to increase the visibility of nursing research, the onus for which will fall on the shoulders of nurse researchers themselves. Most people are unaware that nurses are scholars and researchers. Nurse researchers must market themselves and their research to professional organizations, consumer organizations, and the corporate world to increase support for their research. They also need to educate upper-level managers and corporate executives about the importance of clinical outcomes research. As Baldwin and Nail (2000) have noted, nurse researchers are one of the best-qualified groups to meet the need in today’s world for clinical outcomes research, but they are not recognized for their expertise. Priorities and goals for the future are also under discussion. NINR has established scientific goals and objectives for the 5-year period of 2000 to 2004. The four broad goals are: (1) to identify and support research opportunities that will achieve scientific distinction and produce significant contributions to health; (2) to identify and support future areas of opportunity to advance research on high quality, cost-effective care and to contribute to the scientific base for nursing practice; (3) to communicate and disseminate research findings resulting from NINR-funded research; and (4) to enhance the development of nurse researchers through training and career development opportunities. For the years 2000, 2001, and 2002, topics identified by NINR as special areas of research opportunity included: • Chronic illnesses or conditions (e.g., management of chronic pain; care of children with asthma; adherence to diabetes self-management) • Behavioral changes and interventions (e.g., research in informal caregiving; disparities of infant mortality; effective sleep in health and illness)

11

• Responding to compelling public health concerns (e.g., reducing health disparities in cancer screening; end-of-life/palliative care) SOURCES OF EVIDENCE FOR NURSING PRACTICE Nursing students are taught how best to practice nursing, and best-practice learning continues throughout nurses’ careers. Some of what students and nurses learn is based on systematic research, but much of it is not. In fact, Millenson (1997) estimated that 85% of health care practice has not been scientifically validated. Clinical nursing practice relies on a collage of information sources that vary in dependability and validity. Increasingly there are discussions of evidence hierarchies that acknowledge that certain types of evidence and knowledge are superior to others. A brief discussion of some alternative sources of evidence shows how research-based information is different. Tradition Many questions are answered and problems solved based on inherited customs or tradition. Within each culture, certain “truths” are accepted as given. For example, as citizens of democratic societies, most of us accept, without proof, that democracy is the highest form of government. This type of knowledge often is so much a part of our heritage that few of us seek verification. Tradition offers some advantages. It is efficient as an information source: each individual is not required to begin anew in an attempt to understand the world or certain aspects of it. Tradition or custom also facilitates communication by providing a common foundation of accepted truth. Nevertheless, tradition poses some problems because many traditions have never been evaluated for their validity. Indeed, by their nature, traditions may interfere with the ability to perceive alternatives. Walker’s (1967) research on ritualistic practices in nursing suggests that some traditional nursing practices, such as the routine taking of a patient’s temperature, pulse, and

12

PART 1 Foundations of Nursing Research

respirations, may be dysfunctional. The Walker study illustrates the potential value of critical appraisal of custom and tradition before accepting them as truth. There is growing concern that many nursing interventions are based on tradition, customs, and “unit culture” rather than on sound evidence (e.g., French, 1999). Authority In our complex society, there are authorities—people with specialized expertise—in every field. We are constantly faced with making decisions about matters with which we have had no direct experience; therefore, it seems natural to place our trust in the judgment of people who are authoritative on an issue by virtue of specialized training or experience. As a source of understanding, however, authority has shortcomings. Authorities are not infallible, particularly if their expertise is based primarily on personal experience; yet, like tradition, their knowledge often goes unchallenged. Although nursing practice would flounder if every piece of advice from nursing educators were challenged by students, nursing education would be incomplete if students never had occasion to pose such questions as: How does the authority (the instructor) know? What evidence is there that what I am learning is valid? Clinical Experience, Trial and Error, and Intuition Our own clinical experiences represent a familiar and functional source of knowledge. The ability to generalize, to recognize regularities, and to make predictions based on observations is an important characteristic of the human mind. Despite the obvious value of clinical expertise, it has limitations as a type of evidence. First, each individual’s experience is fairly restricted. A nurse may notice, for example, that two or three cardiac patients follow similar postoperative sleep patterns. This observation may lead to some interesting discoveries with implications for nursing interventions, but does one nurse’s observations justify broad changes in nursing care? A second limitation of experience is

that the same objective event is usually experienced or perceived differently by two individuals. Related to clinical experience is the method of trial and error. In this approach, alternatives are tried successively until a solution to a problem is found. We likely have all used trial and error in our lives, including in our professional work. For example, many patients dislike the taste of potassium chloride solution. Nurses try to disguise the taste of the medication in various ways until one method meets with the approval of the patient. Trial and error may offer a practical means of securing knowledge, but it is fallible. This method is haphazard, and the knowledge obtained is often unrecorded and, hence, inaccessible in subsequent clinical situations. Finally, intuition is a type of knowledge that cannot be explained on the basis of reasoning or prior instruction. Although intuition and hunches undoubtedly play a role in nursing practice—as they do in the conduct of research—it is difficult to develop policies and practices for nurses on the basis of intuition. Logical Reasoning Solutions to many perplexing problems are developed by logical thought processes. Logical reasoning as a method of knowing combines experience, intellectual faculties, and formal systems of thought. Inductive reasoning is the process of developing generalizations from specific observations. For example, a nurse may observe the anxious behavior of (specific) hospitalized children and conclude that (in general) children’s separation from their parents is stressful. Deductive reasoning is the process of developing specific predictions from general principles. For example, if we assume that separation anxiety occurs in hospitalized children (in general), then we might predict that (specific) children in Memorial Hospital whose parents do not room-in will manifest symptoms of stress. Both systems of reasoning are useful as a means of understanding and organizing phenomena, and both play a role in nursing research. However, reasoning in and of itself is limited because the validity of reasoning depends on the accuracy of the information (or premises) with which one starts, and

CHAPTER 1 Introduction to Nursing Research

reasoning may be an insufficient basis for evaluating accuracy. Assembled Information In making clinical decisions, health care professionals also rely on information that has been assembled for a variety of purposes. For example, local, national, and international bench-marking data provide information on such issues as the rates of using various procedures (e.g., rates of cesarean deliveries) or rates of infection (e.g., nosocomial pneumonia rates), and can serve as a guide in evaluating clinical practices. Cost data— that is, information on the costs associated with certain procedures, policies, or practices—are sometimes used as a factor in clinical decisionmaking. Quality improvement and risk data, such as medication error reports and evidence on the incidence and prevalence of skin breakdown, can be used to assess practices and determine the need for practice changes. Such sources, although offering some information that can be used in practice, provide no mechanism for determining whether improvements in patient outcomes result from their use. Disciplined Research Research conducted within a disciplined format is the most sophisticated method of acquiring evidence that humans have developed. Nursing research combines aspects of logical reasoning with other features to create evidence that, although fallible, tends to be more reliable than other methods of knowledge acquisition. The current emphasis on evidence-based health care requires nurses to base their clinical practice to the greatest extent possible on research-based findings rather than on tradition, authority, intuition, or personal experience. Findings from rigorous research investigations are considered to be at the pinnacle of the evidence hierarchy for establishing an EBP. As we discuss next, disciplined research in nursing is richly diverse with regard to questions asked and methods used.

13

PA R A D I G M S F O R NURSING RESEARCH A paradigm is a world view, a general perspective on the complexities of the real world. Paradigms for human inquiry are often characterized in terms of the ways in which they respond to basic philosophical questions: • Ontologic: What is the nature of reality? • Epistemologic: What is the relationship between the inquirer and that being studied? • Axiologic: What is the role of values in the inquiry? • Methodologic: How should the inquirer obtain knowledge? Disciplined inquiry in the field of nursing is being conducted mainly within two broad paradigms, both of which have legitimacy for nursing research. This section describes the two alternative paradigms and outlines their associated methodologies. The Positivist Paradigm One paradigm for nursing research is known as positivism. Positivism is rooted in 19th century thought, guided by such philosophers as Comte, Mill, Newton, and Locke. Positivism is a reflection of a broader cultural phenomenon that, in the humanities, is referred to as modernism, which emphasizes the rational and the scientific. Although strict positivist thinking—sometimes referred to as logical positivism—has been challenged and undermined, a modified positivist position remains a dominant force in scientific research. The fundamental ontologic assumption of positivists is that there is a reality out there that can be studied and known (an assumption refers to a basic principle that is believed to be true without proof or verification). Adherents of the positivist approach assume that nature is basically ordered and regular and that an objective reality exists independent of human observation. In other words, the world is assumed not to be merely a creation of the human mind. The related assumption of determinism refers to the belief that phenomena are not haphazard or random events but rather have antecedent causes. If a person has a cerebrovascular accident,

14

PART 1 Foundations of Nursing Research

the scientist in a positivist tradition assumes that there must be one or more reasons that can be potentially identified and understood. Much of the activity in which a researcher in a positivist paradigm is engaged is directed at understanding the underlying causes of natural phenomena. Because of their fundamental belief in an objective reality, positivists seek to be as objective as possible in their pursuit of knowledge. Positivists attempt to hold their personal beliefs and biases in check insofar as possible during their research to avoid contaminating the phenomena under investigation. The positivists’ scientific approach involves the use of orderly, disciplined procedures that are designed to test researchers’ hunches about the nature of phenomena being studied and relationships among them.

The Naturalistic Paradigm The naturalistic paradigm began as a countermovement to positivism with writers such as Weber and Kant. Just as positivism reflects the cultural phenomenon of modernism that burgeoned in the wake of the industrial revolution, naturalism is an outgrowth of the pervasive cultural transformation that is usually referred to as postmodernism. Postmodern thinking emphasizes the value of deconstruction—that is, of taking apart old ideas and structures—and reconstruction—that is, putting ideas and structures together in new ways. The naturalistic paradigm represents a major alternative system for conducting disciplined research in nursing. Table 1-2 compares the major assumptions of the positivist and naturalistic paradigms.

TABLE 1.2 Major Assumptions of the Positivist and Naturalistic Paradigms ASSUMPTION

POSITIVIST PARADIGM

NATURALISTIC PARADIGM

Ontologic (What is the nature of reality?)

Reality exists; there is a real world driven by real natural causes.

Reality is multiple and subjective, mentally constructed by individuals.

Epistemologic (How is the inquirer related to those being researched?)

The inquirer is independent from those being researched; findings are not influenced by the researcher.

The inquirer interacts with those being researched; findings are the creation of the interactive process.

Axiologic (What is the role of values in the inquiry?)

Values and biases are to be held in check; objectivity is sought.

Subjectivity and values are inevitable and desirable.

Methodologic (How is knowledge obtained?)

Deductive processes Emphasis on discrete, specific concepts Verification of researchers’ hunches Fixed design Tight controls over context Emphasis on measured, quantitative information; statistical analysis Seeks generalizations

Inductive processes Emphasis on entirety of some phenomenon, holistic Emerging interpretations grounded in participants’ experiences Flexible design Context-bound Emphasis on narrative information; qualitative analysis Seeks patterns

CHAPTER 1 Introduction to Nursing Research

For the naturalistic inquirer, reality is not a fixed entity but rather a construction of the individuals participating in the research; reality exists within a context, and many constructions are possible. Naturalists thus take the position of relativism: if there are always multiple interpretations of reality that exist in people’s minds, then there is no process by which the ultimate truth or falsity of the constructions can be determined. Epistemologically, the naturalistic paradigm assumes that knowledge is maximized when the distance between the inquirer and the participants in the study is minimized. The voices and interpretations of those under study are crucial to understanding the phenomenon of interest, and subjective interactions are the primary way to access them. The findings from a naturalistic inquiry are the product of the interaction between the inquirer and the participants. Paradigms and Methods: Quantitative and Qualitative Research Broadly speaking, research methods are the techniques used by researchers to structure a study and to gather and analyze information relevant to the research question. The two alternative paradigms have strong implications for the research methods to be used. The methodologic distinction typically focuses on differences between quantitative research, which is most closely allied with the positivist tradition, and qualitative research, which is most often associated with naturalistic inquiry— although positivists sometimes engage in qualitative studies, and naturalistic researchers sometimes collect quantitative information. This section provides an overview of the methods associated with the two alternative paradigms. Note that this discussion accentuates differences in methods as a heuristic device; in reality, there is often greater overlap of methods than this introductory discussion implies. The “Scientific Method” and Quantitative Research The traditional, positivist “scientific method” refers to a general set of orderly, disciplined procedures used to acquire information. Quantitative

15

researchers use deductive reasoning to generate hunches that are tested in the real world. They typically move in an orderly and systematic fashion from the definition of a problem and the selection of concepts on which to focus, through the design of the study and collection of information, to the solution of the problem. By systematic, we mean that the investigator progresses logically through a series of steps, according to a prespecified plan of action. Quantitative researchers use mechanisms designed to control the study. Control involves imposing conditions on the research situation so that biases are minimized and precision and validity are maximized. The problems that are of interest to nurse researchers—for example, obesity, compliance with a regimen, or pain—are highly complicated phenomena, often representing the effects of various forces. In trying to isolate relationships between phenomena, quantitative researchers attempt to control factors that are not under direct investigation. For example, if a scientist is interested in exploring the relationship between diet and heart disease, steps are usually taken to control other potential contributors to coronary disorders, such as stress and cigarette smoking, as well as additional factors that might be relevant, such as a person’s age and gender. Control mechanisms are discussed at length in this book. Quantitative researchers gather empirical evidence—evidence that is rooted in objective reality and gathered directly or indirectly through the senses. Empirical evidence, then, consists of observations gathered through sight, hearing, taste, touch, or smell. Observations of the presence or absence of skin inflammation, the heart rate of a patient, or the weight of a newborn infant are all examples of empirical observations. The requirement to use empirical evidence as the basis for knowledge means that findings are grounded in reality rather than in researchers’ personal beliefs. Evidence for a study in the positivist paradigm is gathered according to a specified plan, using formal instruments to collect needed information. Usually (but not always) the information gathered in such a study is quantitative—that is, numeric

16

PART 1 Foundations of Nursing Research

information that results from some type of formal measurement and that is analyzed with statistical procedures. An important goal of a traditional scientific study is to understand phenomena, not in isolated circumstances, but in a broad, general sense. For example, quantitative researchers are typically not as interested in understanding why Ann Jones has cervical cancer as in understanding what general factors lead to this carcinoma in Ann and others. The desire to go beyond the specifics of the situation is an important feature of the traditional scientific approach. In fact, the degree to which research findings can be generalized to individuals other than those who participated in the study (referred to as the generalizability of the research) is a widely used criterion for assessing the quality of quantitative studies. The traditional scientific approach used by quantitative researchers has enjoyed considerable stature as a method of inquiry, and it has been used productively by nurse researchers studying a wide range of nursing problems. This is not to say, however, that this approach can solve all nursing problems. One important limitation—common to both quantitative and qualitative research—is that research methods cannot be used to answer moral or ethical questions. Many of our most persistent and intriguing questions about the human experience fall into this area—questions such as whether euthanasia should be practiced or abortion should be legal. Given the many moral issues that are linked to health care, it is inevitable that the nursing process will never rely exclusively on scientific information. The traditional research approach also must contend with problems of measurement. To study a phenomenon, quantitative researchers attempt to measure it. For example, if the phenomenon of interest is patient morale, researchers might want to assess if morale is high or low, or higher under certain conditions than under others. Although there are reasonably accurate measures of physiologic phenomena, such as blood pressure and body temperature, comparably accurate measures of such psychological phenomena as patient morale, pain, or self-image have not been developed.

Another issue is that nursing research tends to focus on human beings, who are inherently complex and diverse. Traditional quantitative methods typically focus on a relatively small portion of the human experience (e.g., weight gain, depression, chemical dependency) in a single study. Complexities tend to be controlled and, if possible, eliminated, rather than studied directly, and this narrowness of focus can sometimes obscure insights. Finally and relatedly, quantitative research conducted in the positivist paradigm has sometimes been accused of a narrowness and inflexibility of vision, a problem that has been called a sedimented view of the world that does not fully capture the reality of human experience. Naturalistic Methods and Qualitative Research Naturalistic methods of inquiry attempt to deal with the issue of human complexity by exploring it directly. Researchers in the naturalistic tradition emphasize the inherent complexity of humans, their ability to shape and create their own experiences, and the idea that truth is a composite of realities. Consequently, naturalistic investigations place a heavy emphasis on understanding the human experience as it is lived, usually through the careful collection and analysis of qualitative materials that are narrative and subjective. Researchers who reject the traditional “scientific method” believe that a major limitation of the classical model is that it is reductionist—that is, it reduces human experience to only the few concepts under investigation, and those concepts are defined in advance by the researcher rather than emerging from the experiences of those under study. Naturalistic researchers tend to emphasize the dynamic, holistic, and individual aspects of human experience and attempt to capture those aspects in their entirety, within the context of those who are experiencing them. Flexible, evolving procedures are used to capitalize on findings that emerge in the course of the study. Naturalistic inquiry always takes place in the field (i.e., in naturalistic settings), often over an extended period of time, while quantitative research

CHAPTER 1 Introduction to Nursing Research

takes place both in natural as well as in contrived laboratory settings. In naturalistic research, the collection of information and its analysis typically progress concurrently; as researchers sift through information, insights are gained, new questions emerge, and further evidence is sought to amplify or confirm the insights. Through an inductive process, researchers integrate information to develop a theory or description that helps explicate processes under observation. Naturalistic studies result in rich, in-depth information that has the potential to elucidate varied dimensions of a complicated phenomenon. Because of this feature—and the relative ease with which qualitative findings can be communicated to lay audiences—it has been argued that qualitative methods will play a more prominent role in health care policy and development in the future (Carey, 1997). The findings from in-depth qualitative research are rarely superficial, but there are several limitations of the approach. Human beings are used directly as the instrument through which information is gathered, and humans are extremely intelligent and sensitive—but fallible—tools. The subjectivity that enriches the analytic insights of skillful researchers can yield trivial “findings” among less competent inquirers. The subjective nature of naturalistic inquiry sometimes causes concerns about the idiosyncratic nature of the conclusions. Would two naturalistic researchers studying the same phenomenon in the same setting arrive at the same results? The situation is further complicated by the fact that most naturalistic studies involve a relatively small group of people under study. Questions about the generalizability of findings from naturalistic inquiries sometimes arise. Multiple Paradigms and Nursing Research Paradigms should be viewed as lenses that help to sharpen our focus on a phenomenon of interest, not as blinders that limit intellectual curiosity. The emergence of alternative paradigms for the study of nursing problems is, in our view, a healthy and

17

desirable trend in the pursuit of new evidence for practice. Although researchers’ world view may be paradigmatic, knowledge itself is not. Nursing knowledge would be thin, indeed, if there were not a rich array of methods available within the two paradigms—methods that are often complementary in their strengths and limitations. We believe that intellectual pluralism should be encouraged and fostered. Thus far, we have emphasized differences between the two paradigms and their associated methods so that their distinctions would be easy to understand. Subsequent chapters of this book will further elaborate on differences in terminology, methods, and research products. It is equally important, however, to note that these two paradigms have many features in common, only some of which are mentioned here: • Ultimate goals. The ultimate aim of disciplined inquiry, regardless of the underlying paradigm, is to gain understanding about phenomena. Both quantitative and qualitative researchers seek to capture the truth with regard to an aspect of the world in which they are interested, and both groups can make significant contributions to nursing knowledge. Moreover, qualitative studies often serve as a crucial starting point for more controlled quantitative studies. • External evidence. Although the word empiricism has come to be allied with the traditional scientific approach, it is nevertheless the case that researchers in both traditions gather and analyze external evidence that is collected through their senses. Neither qualitative nor quantitative researchers are armchair analysts, relying on their own beliefs and views of the world for their conclusions. Information is gathered from others in a deliberate fashion. • Reliance on human cooperation. Because evidence for nursing research comes primarily from human participants, the need for human cooperation is inevitable. To understand people’s characteristics and experiences, researchers must persuade them to participate in the investigation and to act and speak candidly. For certain

18

PART 1 Foundations of Nursing Research

topics, the need for candor and cooperation is a challenging requirement—for researchers in either tradition. • Ethical constraints. Research with human beings is guided by ethical principles that sometimes interfere with research goals. For example, if researchers want to test a potentially beneficial intervention, is it ethical to withhold the treatment from some people to see what happens? As discussed later in the book (see Chapter 7), ethical dilemmas often confront researchers, regardless of their paradigmatic orientation. • Fallibility of disciplined research. Virtually all studies—in either paradigm—have some limitations. Every research question can be addressed in many different ways, and inevitably there are tradeoffs. Financial constraints are universal, but limitations exist even when resources are abundant. This does not mean that small, simple studies have no value. It means that no single study can ever definitively answer a research question. Each completed study adds to a body of accumulated knowledge. If several researchers pose the same question and if each obtains the same or similar results, increased confidence can be placed in the answer to the question. The fallibility of any single study makes it important to understand the tradeoffs and decisions that investigators make when evaluating the adequacy of those decisions. Thus, despite philosophic and methodologic differences, researchers using the traditional quantitative approach or naturalistic methods often share overall goals and face many similar constraints and challenges. The selection of an appropriate method depends on researchers’ personal taste and philosophy, and also on the research question. If a researcher asks, “What are the effects of surgery on circadian rhythms (biologic cycles)?” the researcher really needs to express the effects through the careful quantitative measurement of various bodily properties subject to rhythmic variation. On the other hand, if a researcher asks, “What is the process by which parents learn to cope with the death of a child?” the researcher

would be hard pressed to quantify such a process. Personal world views of researchers help to shape their questions. In reading about the alternative paradigms for nursing research, you likely were more attracted to one of the two paradigms—the one that corresponds most closely to your view of the world and of reality. It is important, however, to learn about and respect both approaches to disciplined inquiry, and to recognize their respective strengths and limitations. In this textbook, we describe methods associated with both qualitative and quantitative research. THE PURPOSES OF NURSING RESEARCH The general purpose of nursing research is to answer questions or solve problems of relevance to the nursing profession. Sometimes a distinction is made between basic and applied research. As traditionally defined, basic research is undertaken to extend the base of knowledge in a discipline, or to formulate or refine a theory. For example, a researcher may perform an in-depth study to better understand normal grieving processes, without having explicit nursing applications in mind. Applied research focuses on finding solutions to existing problems. For example, a study to determine the effectiveness of a nursing intervention to ease grieving would be applied research. Basic research is appropriate for discovering general principles of human behavior and biophysiologic processes; applied research is designed to indicate how these principles can be used to solve problems in nursing practice. In nursing, the findings from applied research may pose questions for basic research, and the results of basic research often suggest clinical applications. The specific purposes of nursing research include identification, description, exploration, explanation, prediction, and control. Within each purpose, various types of question are addressed by nurse researchers; certain questions are more amenable to qualitative than to quantitative inquiry, and vice versa.

CHAPTER 1 Introduction to Nursing Research

Identification and Description Qualitative researchers sometimes study phenomena about which little is known. In some cases, so little is known that the phenomenon has yet to be clearly identified or named or has been inadequately defined or conceptualized. The in-depth, probing nature of qualitative research is well suited to the task of answering such questions as, “What is this phenomenon?” and “What is its name?” (Table 1-3). In

19

quantitative research, by contrast, the researcher begins with a phenomenon that has been previously studied or defined—sometimes in a qualitative study. Thus, in quantitative research, identification typically precedes the inquiry. Qualitative example of identification: Weiss and Hutchinson (2000) investigated people with diabetes and hypertension to discover the basic social problem that affects their adherence

TABLE 1.3 Research Purposes and Research Questions

PURPOSE

TYPES OF QUESTIONS: QUANTITATIVE RESEARCH

Identification

TYPES OF QUESTIONS: QUALITATIVE RESEARCH What is this phenomenon? What is its name?

Description

How prevalent is the phenomenon? How often does the phenomenon occur? What are the characteristics of the phenomenon?

What are the dimensions of the phenomenon? What variations exist? What is important about the phenomenon?

Exploration

What factors are related to the phenomenon? What are the antecedents of the phenomenon?

What is the full nature of the phenomenon? What is really going on here? What is the process by which the phenomenon evolves or is experienced?

Explanation

What are the measurable associations between phenomena? What factors cause the phenomenon? Does the theory explain the phenomenon?

How does the phenomenon work? Why does the phenomenon exist? What is the meaning of the phenomenon? How did the phenomenon occur?

Prediction

What will happen if we alter a phenomenon or introduce an intervention? If phenomenon X occurs, will phenomenon Y follow?

Control

How can we make the phenomenon , happen or alter its nature or prevalence? Can the occurrence of the phenomenon be prevented or controlled?

20

PART 1 Foundations of Nursing Research

to health care directives. Through in-depth interviews with 21 clients, the researchers identified that warnings of vulnerability was the basic problem undermining adherence. Description of phenomena is another important purpose of research. In a descriptive study, researchers observe, count, delineate, and classify. Nurse researchers have described a wide variety of phenomena. Examples include patients’ stress and coping, pain management, adaptation processes, health beliefs, rehabilitation success, and time patterns of temperature readings. Description can be a major purpose for both qualitative and quantitative researchers. Quantitative description focuses on the prevalence, incidence, size, and measurable attributes of phenomena. Qualitative researchers, on the other hand, use indepth methods to describe the dimensions, variations, and importance of phenomena. Table 1-3 compares descriptive questions posed by quantitative and qualitative researchers. Quantitative example of description: Bohachick, Taylor, Sereika, Reeder, and Anton (2002) conducted a study to describe quantitative changes in psychological well-being and psychological resources 6 months after a heart transplantation. Qualitative example of description: Bournes and Mitchell (2002) undertook an in-depth study to describe the experience of waiting in a critical care waiting room. Exploration Like descriptive research, exploratory research begins with a phenomenon of interest; but rather than simply observing and describing it, exploratory research investigates the full nature of the phenomenon, the manner in which it is manifested, and the other factors to which it is related. For example, a descriptive quantitative study of patients’ preoperative stress might seek to document the degree of stress patients experience before surgery and the percentage of patients who actually experience it. An exploratory study might ask the following: What factors diminish or

increase a patient’s stress? Is a patient’s stress related to behaviors of the nursing staff? Is stress related to the patient’s cultural backgrounds? Qualitative methods are especially useful for exploring the full nature of a little-understood phenomenon. Exploratory qualitative research is designed to shed light on the various ways in which a phenomenon is manifested and on underlying processes. Quantitative example of exploration: Reynolds and Neidig (2002) studied the incidence and severity of nausea accompanying combinative antiretroviral therapies among HIVinfected patients, and explored patterns of nausea in relation to patient characteristics. Qualitative example of exploration: Through in-depth interviews, Sadala and Mendes (2000) explored the experiences of 18 nurses who cared for patients who had been pronounced brain dead but kept alive to serve as organ donors. Explanation The goals of explanatory research are to understand the underpinnings of specific natural phenomena, and to explain systematic relationships among phenomena. Explanatory research is often linked to theories, which represent a method of deriving, organizing, and integrating ideas about the manner in which phenomena are interrelated. Whereas descriptive research provides new information, and exploratory research provides promising insights, explanatory research attempts to offer understanding of the underlying causes or full nature of a phenomenon. In quantitative research, theories or prior findings are used deductively as the basis for generating explanations that are then tested empirically. That is, based on a previously developed theory or body of evidence, researchers make specific predictions that, if upheld by the findings, add credibility to the explanation. In qualitative studies, researchers may search for explanations about how or why a phenomenon exists or what a phenomenon means as a basis for developing a theory that is grounded in rich, in-depth, experiential evidence.

CHAPTER 1 Introduction to Nursing Research

Quantitative example of explanation: Resnick, Orwig, Maganizer, and Wynne (2002) tested a model to explain exercise behavior among older adults on the basis of social support, age, and self-efficacy expectations. Qualitative example of explanation: Hupcey (2000) undertook a study that involved the development of a model explaining the psychosocial needs of patients in the intensive care unit. Feeling safe was the overwhelming need of patients in the intensive care unit. Prediction and Control Many phenomena defy explanation. Yet it is frequently possible to make predictions and to control phenomena based on research findings, even in the absence of complete understanding. For example, research has shown that the incidence of Down syndrome in infants increases with the age of the mother. We can predict that a woman aged 40 years is at higher risk of bearing a child with Down syndrome than is a woman aged 25 years. We can partially control the outcome by educating women about the risks and offering amniocentesis to women older than 35 years of age. Note, however, that the ability to predict and control in this example does not depend on an explanation of why older women are at a higher risk of having an abnormal child. In many examples of nursing and health-related studies—typically, quantitative ones—prediction and control are key objectives. Studies designed to test the efficacy of a nursing intervention are ultimately concerned with controlling patient outcomes or the costs of care. Quantitative example of prediction: Lindeke, Stanley, Else, and Mills (2002) used neonatal data to predict academic performance and the need for special services among school-aged children who had been in a level 3 neonatal intensive care unit. RESEARCH EXAMPLES Each chapter of this book presents brief descriptions of actual studies conducted by nurse researchers. The descriptions focus on aspects of

21

the study emphasized in the chapter. A review of the full journal article likely would prove useful. Research Example of a Quantitative Study McDonald, Freeland, Thomas, and Moore (2001) conducted a study to determine the effectiveness of a preoperative pain management intervention for relieving pain among elders undergoing surgery. Their report appeared in the journal Research in Nursing & Health. McDonald (who had conducted earlier research on the topic of pain and pain management) and her colleagues developed a preoperative intervention that taught pain management and pain communication skills. The content was specifically geared to older adults undergoing surgery. Forty elders, all older than age 65 years, were recruited to participate in the study. Half of these elders were assigned, at random, to participate in the special intervention; the remaining half got usual preoperative care. Postoperative pain was measured for both groups on the evening of the surgery, on postoperative day 1, and on postoperative day 2. The results supported the researchers’ predictions that (a) pain in both groups would decline over time; and (b) those receiving the special intervention would experience greater decreases in pain over time. The researchers noted that further research is needed to determine whether the intervention’s effect resulted from instruction in pain management or in pain communication skills (and, indeed, McDonald reported being in the process of conducting such a study). They also noted that the study was based on elders undergoing certain types of surgery at a single site, acknowledging that the findings need confirmation in other settings and contexts. Nevertheless, this study offers evidence that pain responses of elderly surgical patients can be lowered through a nursing intervention. The strength of this evidence lies in several factors— several of which you will appreciate more as you become familiar with research methods. Most important, this study was quite rigorous. The intervention itself was based on a formal theory of communication accommodation, which addresses how people adjust communication to their own needs. The researchers took care to ensure that the two groups being compared were equivalent in terms of background and

22

PART 1 Foundations of Nursing Research

medical characteristics, so that group differences in pain responses reflected the intervention and not some spurious factor. The research team members who measured pain responses were not aware of whether the elders were in the intervention group, so as not to bias the measurements. Finally, the findings are more persuasive because the team of researchers who conducted the study have developed a solid program of research on pain, and their research has contributed incrementally to understanding pain responses and appropriate nursing interventions.

Research Example of a Qualitative Study Cheek and Ballantyne (2001) undertook a study to describe the search and selection process for an aged care facility after discharge of a family member from acute hospital settings in Australia, and to explore the effects the process had on the individuals and their families. Twelve residents and 20 of their sponsors (the primary contact person responsible for the resident) participated in the study. Face-to-face in-depth interviews were conducted with residents in the aged care facilities and with family members in their homes. They were all asked to talk about their personal experiences of the search and selection process and its effect on their well-being. These interviews were audiotaped and then transcribed. Analysis of the interview transcripts revealed five themes. One theme, for example, was labeled “dealing with the system—cutting through the maze.” Dealing with the system was perceived as being in the middle of a war zone. This sense of battle was related to confusion, lack of control, and the feeling of being at the system’s mercy. Contributing to this perception of being at war with the system was the stress of having to deal with multiple aged care facilities on an individual basis. A second major theme was labeled “Urgency—moving them on and in.” Sponsors felt a sense of urgency in finding a suitable facility to have their family member transferred to from the acute setting. Sponsors felt pressured to make on-the-spot decisions to accept or reject a place in a facility once it had become available. This thorough and careful study provides a firsthand perspective on the experiences of people going through the process of selecting an appropriate longterm care facility for aging family members. One of

the central implications for practice of this study concerns the need to revise the search and selection process to make it more efficient in terms of time and effort of the sponsors and residents. In addition, the study suggests that increased communication—from the acute setting to the aged care facilities being considered—could play an important role in decreasing the stress of this guilt-ridden experience. The clinical implications of the study are strengthened by the fact that the researchers took steps to ensure its rigor. For example, the transcripts of these interviews were read by at least two members of the research team who individually identified themes from each interview. The researchers then compared and discussed the themes from all the interviews until consensus was reached. Moreover, the researchers took steps to weigh their evidence for their thematic conclusions against potentially competing explanations of the data.

S U M M A RY P O I N T S • Nursing research is systematic inquiry to develop knowledge about issues of importance to nurses. • Nurses in various settings are adopting an evidence-based practice that incorporates research findings into their decisions and their interactions with clients. • Knowledge of nursing research enhances the professional practice of all nurses—including both consumers of research (who read and evaluate studies) and producers of research (who design and undertake research studies). • Nursing research began with Florence Nightingale but developed slowly until its rapid acceleration in the 1950s. Since the 1970s, nursing research has focused on problems relating to clinical practice. • The National Institute of Nursing Research (NINR), established at the U.S. National Institutes of Health in 1993, affirms the stature of nursing research in the United States. • Future emphases of nursing research are likely to include outcomes research, research utilization projects, replications of research, multisite studies, and expanded dissemination efforts.

CHAPTER 1 Introduction to Nursing Research

• Disciplined research stands in contrast to other sources of evidence for nursing practice, such as tradition, voices of authority, personal experience, trial and error, intuition, and logical reasoning; rigorous research is at the pinnacle of the evidence hierarchy as a basis for making clinical decisions. • Disciplined inquiry in nursing is conducted within two broad paradigms—world views with underlying assumptions about the complexities of reality: the positivist paradigm and the naturalistic paradigm. • In the positivist paradigm, it is assumed that there is an objective reality and that natural phenomena are regular and orderly. The related assumption of determinism refers to the belief that phenomena are the result of prior causes and are not haphazard. • In the naturalistic paradigm, it is assumed that reality is not a fixed entity but is rather a construction of human minds—and thus “truth” is a composite of multiple constructions of reality. • The positivist paradigm is associated with quantitative research—the collection and analysis of numeric information. Quantitative research is typically conducted within the traditional “scientific method,” which is a systematic and controlled process. Quantitative researchers base their findings on empirical evidence (evidence collected by way of the human senses) and strive for generalizability of their findings beyond a single setting or situation. • Researchers within the naturalistic paradigm emphasize understanding the human experience as it is lived through the collection and analysis of subjective, narrative materials using flexible procedures that evolve in the field; this paradigm is associated with qualitative research. • Basic research is designed to extend the base of information for the sake of knowledge. Applied research focuses on discovering solutions to immediate problems. • Research purposes for nursing research include identification, description, exploration, explanation, prediction, and control.

23

STUDY ACTIVITIES Chapter 1 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing the concepts presented in this chapter. In addition, the following study questions can be addressed: 1. What are some of the current changes occurring in the health care delivery system, and how could these changes influence nursing research? 2. Is your world view closer to the positivist or the naturalistic paradigm? Explore the aspects of the two paradigms that are especially consistent with your world view. 3. How does the assumption of scientific determinism conflict with or coincide with superstitious thinking? Take, as an example, the superstition associated with four-leaf clovers or a rabbit’s foot. 4. How does the ability to predict phenomena offer the possibility of their control? SUGGESTED READINGS Methodologic and Theoretical References American Nurses’ Association Cabinet on Nursing Research. (1985). Directions for nursing research: Toward the twenty-first century. Kansas City, MO: American Nurses’ Association. Baldwin, K. M., & Nail, L. M. (2000). Opportunities and challenges in clinical nursing research. Journal of Nursing Scholarship, 32, 163–166. Brown, E. L. (1948). Nursing for the future. New York: Russell Sage. Carey, M. A. (1997). Qualitative research in policy development. In J. M. Morse (Ed). Completing a qualitative project: Details and dialogue (pp. 345–354). Thousand Oaks, CA: Sage. D’Antonio, P. (1997). Toward a history of research in nursing. Nursing Research, 46, 105–110. Donaldson, S. K. (2000). Breakthroughs in scientific research: The discipline of nursing, 1960–1999. Annual Review of Nursing Research, 18, 247–311. French, P. (1999). The development of evidence-based nursing. Journal of Advanced Nursing, 29, 72–78.

24

PART 1 Foundations of Nursing Research

Goode, C. J. (2000). What constitutes “evidence” in evidence-based practice? Applied Nursing Research, 13, 222–225. Guba, E. G. (Ed.). (1990). The paradigm dialog. Newbury Park, CA: Sage. Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. Beverly Hills, CA: Sage. Lindeman, C. A. (1975). Delphi survey of priorities in clinical nursing research. Nursing Research, 24, 434–441. Millenson, M. L. (1997). Demanding medical evidence. Chicago: University of Chicago Press. Nightingale, F. (1859). Notes on nursing: What it is, and what it is not. Philadelphia: J. B. Lippincott. Stetler, C. B., & Marram, G. (1976). Evaluating research findings for applicability in practice. Nursing Outlook, 24, 559–563. Walker, V. H. (1967). Nursing and ritualistic practice. New York: Macmillan.

Studies Cited in Chapter 1 Barnard, K. E. (1973). The effects of stimulation on the sleep behavior of the premature infant. In M. Batey (Ed.), Communicating nursing research (Vol. 6, pp. 12–33). Boulder, CO: WICHE. Barnard, K. E. (1976). The state of the art: Nursing and early intervention with handicapped infants. In T. Tjossem (Ed.), Proceedings of the 1974 President’s Committee on Mental Retardation. Baltimore, MD: University Park Press. Barnard, K. E., & Collar, B. S. (1973). Early diagnosis, interpretation, and intervention. Annals of the New York Academy of Sciences, 205, 373–382. Barnard, K. E., Wenner, W., Weber, B., Gray, C., & Peterson, A. (1977). Premature infant refocus. In P. Miller (Ed.), Research to practice in mental retardation: Vol. 3, Biomedical aspects. Baltimore, MD: University Park Press. Bohachick, P., Taylor, M., Sereika, S., Reeder, S., & Anton, B. (2002). Social support, personal control, and psychosocial recovery following heart transplantation. Clinical Nursing Research, 11, 34–51. Bournes, D. A., & Mitchell, G. J. (2002). Waiting: The experience of persons in a critical care waiting room. Research in Nursing & Health, 25, 58–67. Brooten, D., Brown, L. P., Munro, B. H., York, R., Cohen, S., Roncoli, M., & Hollingsworth, A. (1988). Early discharge and specialist transitional care. Image: Journal of Nursing Scholarship, 20, 64–68.

Brooten, D., Gennaro, S., Knapp, H., Brown, L. P., & York, R. (1989). Clinical specialist pre- and postdischarge teaching of parents of very low birthweight infants. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 18, 316–322. Brooten, D., Kumar, S., Brown, L. P., Butts, P., Finkler, S., Bakewell-Sachs, S., Gibbons, S., & DelivoriaPapadopoulos, M. (1986). A randomized clinical trail of early hospital discharge and home follow-up of very low birthweight infants. New England Journal of Medicine, 315, 934–939. Brooten, D., Roncoli, M., Finkler, S., Arnold, L., Cohen, A., & Mennuti, M. (1994). A randomized clinical trial of early hospital discharge and home follow-up of women having cesarean birth. Obstetrics and Gynecology, 84, 832–838. Cheek, J. & Ballantyne, A. (2001). Moving them on and in: The process of searching for and selecting an aged care facility. Qualitative Health Research, 11, 221–237. Doering, L. V., Esmailian, F., Imperial-Perez, F., & Monsein, S. (2001). Determinants of intensive care length of stay after coronary artery bypass graft surgery. Heart & Lung, 30, 9–17. Hupcey, J. E. (2000). Feeling safe: The psychosocial needs of ICU patients. Journal of Nursing Scholarship, 32, 361–367. Lindeke, L. L., Stanley, J. R., Else, B. S., & Mills, M. M. (2002). Neonatal predictors of school-based services used by NICU graduates at school age. Journal of Maternal—Child Nursing, 27, 41–46. Lund, C. H., Kuller, J., Lane, A. T., Lott, J., Raines, D., & Thomas, K. (2001). Neonatal skin care: Evaluation of the AWHONN/NANN research-based practice project on knowledge and skin care practices. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 30, 30–40. Lund, C. H., Osborne, J., Kuller, J., Lane, A. T., Lott, J., & Raines, D. (2001). Neonatal skin care: Clinical outcomes of the AWHONN/NANN evidence-based clinical practice guideline. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 30, 41–51. McDonald, D. D., Freeland, M., Thomas, G., & Moore, J. (2001). Testing a preoperative pain management intervention for elders. Research in Nursing & Health, 24, 402–409. Paterson, J., & Stewart, J. (2002). Adults with acquired brain injury: Perceptions of their social world. Rehabilitation Nursing, 27, 13–18. Quint, J. C. (1962). Delineation of qualitative aspects of nursing care. Nursing Research, 11, 204–206.

CHAPTER 1 Introduction to Nursing Research Quint, J. C. (1963). The impact of mastectomy. American Journal of Nursing, 63, 88–91. Quint, J. C. (1967). The nurse and the dying patient. New York: Macmillan. Resnick, B., Orwig, D., Maganizer, J., & Wynne, C. (2002). The effect of social support on exercise behavior in older adults. Clinical Nursing Research, 11, 52–70. Reynolds, N. R., & Neidig, J. L. (2002). Characteristics of nausea reported by HIV-infected patients initiating combination antiretroviral regimens. Clinical Nursing Research, 11, 71–88. Sadala, M. L. A., & Mendes, H. W. B. (2000). Caring for organ donors: The intensive care unit nurses’ view. Qualitative Health Research, 10, 788–805. Samselle, C. M., Wyman, J. F., Thomas, K. K., Newman, D. K., Gray, M., Dougherty, M., & Burns, P. A. (2000a). Continence for women: Evaluation of AWHONN’s third research utilization project. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 29, 9–17.

25

Samselle, C. M., Wyman, J. F., Thomas, K. K., Newman, D. K., Gray, M., Dougherty, M., & Burns, P. A. (2000b). Continence for women: A test of AWHONN’s evidence-based practice protocol in clinical practice. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 29, 18–26. Swanson, B., Cronin-Stubbs, D., Zeller, J. M., Kessler, H. A., & Bielauskas, L. A. (1993). Characterizing the neuropsychological functioning of persons with human immunodeficiency virus infection. Archives of Psychiatric Nursing, 7, 82–90. Swanson, B., Zeller, J. M., & Spear, G. (1998). Cortisol upregulates HIV p24 antigen in cultured human monocyte-derived macrophages. Journal of the Association of Nurses in AIDS care, 9, 78–83. Weiss, J., & Hutchinson, S. A. (2000). Warnings about vulnerability in clients with diabetes and hypertension. Qualitative Health Research, 10, 521–537.

2

Key Concepts and Terms in Qualitative and Quantitative Research

R

esearch, like nursing or any other discipline, has its own language and terminology—its own jargon. Some terms are used by both qualitative and quantitative researchers (although in some cases, the connotations differ), whereas others are used predominantly by one or the other group. New terms are introduced throughout this textbook, but we devote this chapter to some fundamental terms and concepts so that more complex ideas can be more readily grasped. T H E FA C E S A N D P L A C E S OF RESEARCH When researchers address a problem or answer a question through disciplined research—regardless of the underlying paradigm—they are doing a study (or an investigation or research project). Studies involve various people working together in different roles. Roles on a Research Project Studies with humans involve two sets of people: those who do the research and those who provide the information. In a quantitative study, the people who are being studied are referred to as subjects or study participants, as shown in Table 2-1. (Subjects who provide information to researchers

by answering questions directly—e.g., by filling out a questionnaire—may be called respondents.) The term subjects implies that people are acted upon by researchers (i.e., are subject to research protocols), and usually is avoided by qualitative researchers. In a qualitative study, the individuals cooperating in the study play an active rather than a passive role in the research, and are usually referred to as study participants, informants, or key informants. Collectively, both in qualitative and quantitative studies, study participants comprise the sample. The person who undertakes the research is the researcher or investigator (or sometimes, especially in quantitative studies, the scientist). Studies are often undertaken by several people rather than by a single researcher. Collaborative research involving a team of nurses with both clinical and methodologic expertise (or involving different members of a health care team) is increasingly common in addressing problems of clinical relevance. When a study is undertaken by a research team, the person directing the investigation is referred to as the project director or principal investigator (PI). Two or three researchers collaborating equally are co-investigators. When specialized expertise is needed on a short-term basis (e.g., for statistical analysis), projects may involve one or more consultants. In a large-scale project,

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

27

TABLE 2.1 Key Terms Used in Quantitative and Qualitative Research CONCEPT

QUANTITATIVE TERM

QUALITATIVE TERM

Person Contributing Information

Subject Study participant Respondent

— Study participant Informant, key informant

Person Undertaking the Study

Researcher Investigator Scientist

Researcher Investigator —

That Which Is Being Investigated

— Concepts Constructs Variables

Phenomena Concepts — —

System of Organizing Concepts

Theory, theoretical framework Conceptual framework, conceptual model

Theory Conceptual framework, sensitizing framework

Information Gathered

Data (numerical values)

Data (narrative descriptions)

Connections Between Concepts

Relationships (cause-and-effect, functional)

Patterns of association

Quality of the Evidence

Reliability Validity Generalizability Objectivity

Dependability Credibility Transferability Confirmability

dozens of individuals may be involved in planning the study, producing research-related materials, collecting and analyzing the information, and managing the flow of work. The examples of staffing configurations that follow span the continuum from an extremely large project to a more modest one. Examples of staffing: Example of Staffing on a Quantitative Study The first author of this book has been involved in a complex, multicomponent 6-year study of poor women living in four major cities (Cleveland, Los Angeles, Miami, and Philadelphia). As part of the study, she and two colleagues prepared a book-length

report documenting the health problems and health care concerns of about 4000 welfare mothers who were interviewed in 1998 and again in 2001 (Polit, London, & Martinez, 2001). The total project staff for this research involves well over 100 people, including two co-investigators; lead investigators of the 6 project components (Polit was one of these); a dozen other senior-level researchers; over 50 interviewers; 5 interview supervisors; and dozens of research assistants, computer programmers, secretaries, editors, and other support staff. Several health consultants, including a prominent nurse researcher, were reviewers of the report. The project was funded by a consortium of government agencies and private foundations.

28

PART 1 Foundations of Nursing Research

Example of Staffing on a Qualitative Study Beck (2002) conducted a qualitative study focusing on the experiences of mothers of twins. The team included Beck as the PI (who gathered and analyzed all the information herself); a childbirth educator (who helped to recruit mothers into the study); an administrative assistant (who handled a variety of administrative tasks, like paying stipends to the mothers); a transcriber (who listened to tape-recorded conversations with the mothers and typed them up verbatim); and a secretary (who handled correspondence). This study had some financial support through Beck’s university. In addition to participants and researchers, other parties sometimes are involved in studies. When financial assistance is obtained to pay for research costs, the organization providing the money is the funder or sponsor. Reviewers are sometimes called on to critique various aspects of a study and offer feedback. If these people are at a similar level of experience as the researchers, they may be called peer reviewers. Student projects are more likely to be reviewed by faculty advisors. Sometimes students or young researchers get advice and support from mentors, who not only give direct feedback but model standards of excellence in research. Research Settings Research can be conducted in a wide variety of locales—in health care facilities, in people’s homes, in classrooms, and so on. Researchers make decisions about where to conduct a study based on the nature of the research question and the type of information needed to address it. Generally speaking, the site is the overall location for the research—it could be an entire community (e.g., a Haitian neighborhood in Miami) or an institution within a community (e.g., a hospital in Boston). Researchers sometimes engage in multisite studies because the use of multiple sites usually offers a larger or more diverse sample of study participants. For example, in a study of a new nursing intervention, researchers may wish to implement the

intervention in both public and private hospitals or in urban and rural locations. Settings are the more specific places where data collection occurs. In some cases, the setting and the site are the same, as when the selected site is a large hospital, and information is collected exclusively within that setting. When the site is a larger community, however, the researcher must decide where data should be collected—in nursing homes, homeless shelters, and so on. Because the nature of the setting can influence the way people behave or feel and how they respond to questions, the selection of an appropriate setting is important. Some studies take place in naturalistic settings (in the field), such as in people’s homes or offices. In-depth qualitative studies are especially likely to be done in natural settings because qualitative researchers are interested in studying the context of participants’ experiences. When researchers go into the field to collect their information, they are engaged in fieldwork. In qualitative studies, fieldwork may take months or even years to complete. Qualitative fieldwork often involves studying participants in multiple settings within the selected site (e.g., in their homes, at meetings, and so on). At the other extreme, studies sometimes are conducted in highly controlled laboratory settings that may or may not have elaborate scientific equipment installed. Both human and nonhuman research can occur in laboratory settings. For nurse researchers, studies are often conducted in quasi-natural settings, such as hospitals or other similar facilities. These are settings that are not necessarily natural to the participants (unless the participants are nurses or other health care personnel), but neither are they highly contrived and controlled research laboratories. Example of a study in a naturalistic setting: Carlisle (2000) studied the search for meaning in the caregiving experience among informal carers of people living with HIV and AIDS. The researcher gathered in-depth information from carers in their homes and in HIV/AIDS volunteer organizations.

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

Example of a study in a laboratory setting: Pierce and Clancy (2001) studied the effects of hypoxia on diaphragm activity in anesthetized rats. THE BUILDING BLOCKS OF A STUDY Phenomena, Concepts, and Constructs Research focuses on abstract rather than tangible phenomena. For example, the terms pain, coping, grief, and resilience are all abstractions of particular aspects of human behavior and characteristics. These abstractions are referred to as concepts or, in qualitative studies, phenomena. Researchers (especially quantitative researchers) also use the term construct. Like a concept, a construct refers to an abstraction or mental representation inferred from situations or behaviors. Kerlinger and Lee (2000) distinguish concepts from constructs by noting that constructs are abstractions that are deliberately and systematically invented (or constructed) by researchers for a specific purpose. For example, self-care in Orem’s model of health maintenance is a construct. The terms construct and concept are sometimes used interchangeably, although by convention, a construct often refers to a more complex abstraction than a concept. Theories and Conceptual Models A theory is a systematic, abstract explanation of some aspect of reality. In a theory, concepts are knitted together into a coherent system to describe or explain some aspect of the world. Theories play a role in both qualitative and quantitative research. In a quantitative study, researchers often start with a theory, framework, or conceptual model (the distinctions are discussed in Chapter 6). On the basis of theory, researchers make predictions about how phenomena will behave in the real world if the theory is true. In other words, researchers use deductive reasoning to develop from the general theory specific predictions that can be tested empirically. The results of the

29

research are used to reject, modify, or lend credence to the theory. In qualitative research, theories may be used in various ways (Sandelowski, 1993). Sometimes conceptual or sensitizing frameworks—derived from various disciplines or qualitative research traditions that will be described in Chapter 3— provide an impetus for a study or offer an orienting world view with clear conceptual underpinnings. In such studies, the framework may help in interpreting information gathered by researchers. In other qualitative studies, theory is the product of the research: The investigators use information from the participants inductively as the basis for developing a theory firmly rooted in the participants’ experiences. The participants’ input is the starting point from which the researcher begins to conceptualize, seeking to explain patterns, commonalities, and relationships emerging from the researcher— participant interactions. The goal in such studies is to arrive at a theory that explains phenomena as they occur, not as they are preconceived. Inductively generated theories from qualitative studies are sometimes subjected to more controlled confirmation through quantitative research. Variables In quantitative studies, concepts are usually referred to as variables. A variable, as the name implies, is something that varies. Weight, anxiety levels, income, and body temperature are all variables (i.e., each of these properties varies from one person to another). To quantitative researchers, nearly all aspects of human beings and their environment are variables. For example, if everyone weighed 150 pounds, weight would not be a variable. If it rained continuously and the temperature was always 70!F, weather would not be a variable, it would be a constant. But it is precisely because people and conditions do vary that research is conducted. Most quantitative researchers seek to understand how or why things vary, and to learn how differences in one variable are related to differences in another. For example, lung cancer research is concerned with the variable of lung

30

PART 1 Foundations of Nursing Research

cancer. It is a variable because not everybody has this disease. Researchers have studied what variables might be linked to lung cancer and have discovered that cigarette smoking is related. Smoking is also a variable because not everyone smokes. A variable, then, is any quality of a person, group, or situation that varies or takes on different values. Variables are the central building blocks of quantitative studies. There are different types of variables, as discussed next. Continuous, Discrete, and Categorical Variables Sometimes variables take on a wide range of values. A person’s age, for instance, can take on values from zero to more than 100, and the values are not restricted to whole numbers. Such continuous variables have values that can be represented on a continuum. In theory, a continuous variable can assume an infinite number of values between two points. For example, consider the continuous variable weight: between 1 and 2 pounds, the number of values is limitless: 1.005, 1.7, 1.33333, and so on. By contrast, a discrete variable is one that has a finite number of values between any two points, representing discrete quantities. For example, if people were asked how many children they had, they might answer 0, 1, 2, 3, or more. The value for number of children is discrete, because a number such as 1.5 is not a meaningful value. Between the values 1 and 3, the only possible value is 2. Other variables take on a small range of values that do not inherently represent a quantity. The variable gender, for example, has only two values (male and female). Variables that take on only a handful of discrete nonquantitative values are categorical variables. Another example is blood type (A, B, AB, and O). When categorical variables take on only two values, they are sometimes referred to as dichotomous variables. Some examples of dichotomous variables are pregnant/not pregnant, HIV positive/HIV negative, and alive/dead. Active Versus Attribute Variables Variables are often characteristics of research subjects, such as their age, health beliefs, or weight. Variables such as these are attribute variables. In

many research situations, however, the investigator creates a variable. For example, if a researcher is interested in testing the effectiveness of patientcontrolled analgesia as opposed to intramuscular analgesia in relieving pain after surgery, some patients would be given patient-controlled analgesia and others would receive intramuscular analgesia. In the context of this study, method of pain management is a variable because different patients are given different analgesic methods. Kerlinger and Lee (2000) refer to variables that the researcher creates as active variables. Note that an active variable in one study could be an attribute variable in another. For example, a researcher might create an “active” salt-intake variable by exposing two groups of people to different amounts of salt in their diets. Another researcher could examine the salt-intake “attributes” of a sample by asking about their consumption of salt. Dependent Versus Independent Variables Many studies are aimed at unraveling and understanding causes of phenomena. Does a nursing intervention cause more rapid recovery? Does smoking cause lung cancer? The presumed cause is the independent variable, and the presumed effect is the dependent variable. (Note that some researchers use the term criterion variable rather than dependent variable. In studies that analyze the consequences of an intervention, it is usually necessary to establish criteria against which the intervention’s success can be assessed—hence, the origin of the term criterion variable. Others use the term outcome variable—the variable capturing the outcome of interest—in lieu of dependent variable. The term dependent variable, however, is more general and is the term used throughout this book.) Variability in the dependent variable is presumed to depend on variability in the independent variable. For example, researchers investigate the extent to which lung cancer (the dependent variable) depends on smoking (the independent variable). Or, investigators may be concerned with the extent to which patients’ perception of pain (the dependent variable) depends on different nursing actions (the independent variable).

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

Frequently, the terms independent variable and dependent variable are used to indicate direction of influence rather than causal link. For example, suppose a researcher studied the behaviors of people caring for cognitively impaired elders and found that the patient’s age and the caregivers’ use of social touch were related: the older the patient, the less social touch the caregiver used. The researcher would likely not conclude that patient age caused reductions in social touch. Yet the direction of influence clearly runs from age to touch: it makes no sense to suggest that caregivers’ social touch influenced elders’ age! Although in this example the researcher does not infer a causeand-effect connection, it is appropriate to conceptualize social touch as the dependent variable and age as the independent variable, because it is the caregivers’ use of social touch that the researcher is interested in understanding, explaining, or predicting. Many dependent variables studied by nurse researchers have multiple causes or antecedents. If we were interested in studying factors that influence people’s weight, for example, we might consider their height, physical activity, and diet as independent variables. Multiple dependent variables also may be of interest to researchers. For example, an investigator may be concerned with comparing the effectiveness of two methods of nursing care for children with cystic fibrosis. Several dependent variables could be used as criteria of treatment effectiveness, such as length of hospital stay, number of recurrent respiratory infections, presence of cough, and so forth. In short, it is common to design studies with multiple independent and dependent variables. Variables are not inherently dependent or independent. A dependent variable in one study could be an independent variable in another study. For example, a study might examine the effect of nurses’ contraceptive counseling (the independent variable) on unwanted births (the dependent variable). Another study might investigate the effect of unwanted births (the independent variable) on the incidence of child abuse (the dependent variable). In short, whether a variable is independent or dependent is a

31

function of the role that it plays in a particular study. Example of independent and dependent variables: Varda and Behnke (2000) asked, What is the effect of the timing of an initial bath on temperature in newborns? Their independent variable was timing of the infant’s initial bath (1 hour versus 2 hours after birth). Their dependent variable was axillary temperature. Heterogeneity A term frequently used in connection with variables is heterogeneity. When an attribute is extremely varied in the group under investigation, the group is said to be heterogeneous with respect to that variable. If, on the other hand, the amount of variability is limited, the group is described as relatively homogeneous. For example, for the variable height, a group of 2-year-old children is likely to be more homogeneous than a group of 18-yearold adolescents. The degree of variability or heterogeneity of a group of subjects has implications for study design. Definitions of Concepts and Variables Concepts in a study need to be defined and explicated, and dictionary definitions are almost never adequate. Two types of definitions are of particular relevance in a study—conceptual and operational. The concepts in which researchers are interested are, as noted, abstractions of observable phenomena. Researchers’ world view and their outlook on nursing shape how those concepts are defined. A conceptual definition presents the abstract or theoretical meaning of the concepts being studied. Conceptual meanings are based on theoretical formulations, on a firm understanding of relevant literature, or on researchers’ clinical experience (or on a combination of these). Even seemingly straightforward terms need to be conceptually defined by researchers. The classic example of this is the concept of caring. Morse and her colleagues (1990) scrutinized the works of numerous nurse

32

PART 1 Foundations of Nursing Research

researchers and theorists to determine how caring was defined, and identified five different categories of conceptual definitions: as a human trait; a moral imperative; an affect; an interpersonal relationship; and a therapeutic intervention. Researchers undertaking studies concerned with caring need to make clear which conceptual definition of caring they have adopted—both to themselves and to their audience of readers. In qualitative studies, conceptual definitions of key phenomena may be the major end product of the endeavor, reflecting an intent to have the meaning of concepts defined by those being studied. In quantitative studies, however, researchers need to clarify and define the research concepts at the outset. This is necessary because quantitative researchers must indicate how the variables will be observed and measured in the actual research situation. An operational definition of a concept specifies the operations that researchers must perform to collect the required information. Operational definitions should correspond to conceptual definitions. Variables differ in the ease with which they can be operationalized. The variable weight, for example, is easy to define and measure. We might operationally define weight as follows: the amount that an object weighs in pounds, to the nearest full pound. Note that this definition designates that weight will be determined with one measuring system (pounds) rather than another (grams). The operational definition might also specify that subjects’ weight will be measured to the nearest pound using a spring scale with subjects fully undressed after 10 hours of fasting. This operational definition clearly indicates what is meant by the variable weight. Unfortunately, few variables of interest in nursing research are operationalized as easily as weight. There are multiple methods of measuring most variables, and researchers must choose the method that best captures the variables as they conceptualize them. Take, for example, anxiety, which can be defined in terms of both physiologic and psychological functioning. For researchers choosing to emphasize physiologic aspects of anxiety, the operational definition might involve a physiologic measure such as the Palmar Sweat Index. If,

on the other hand, researchers conceptualize anxiety as primarily a psychological state, the operational definition might involve a paper-and-pencil measure such as the State Anxiety Scale. Readers of research reports may not agree with how investigators conceptualized and operationalized variables, but precision in defining terms has the advantage of communicating exactly what terms mean within the context of the study. Example of conceptual and operational definitions: Beck and Gable (2001) conceptually defined various aspects of postpartum depression and then described how the definitions were linked operationally to a measure Beck developed, the Postpartum Depression Screening Scale (PDSS). For example, one aspect of postpartum depression is cognitive impairment, conceptually defined as “a mother’s loss of control over her thought processes leaves her frightened that she may be losing her mind.” Operationally, the PDSS captured this dimension by having women indicate their level of agreement with such statements as, “I could not stop the thoughts that kept racing in my mind.” Data Research data (singular, datum) are the pieces of information obtained in the course of the investigation. In quantitative studies, researchers identify the variables of interest, develop operational definitions of those variables, and then collect relevant data from subjects. The actual values of the study variables constitute the data for the project. Quantitative researchers collect primarily quantitative data— that is, information in numeric form. As an example, suppose we were conducting a quantitative study in which a key variable was depression; we would need to measure how depressed study participants were. We might ask, “Thinking about the past week, how depressed would you say you have been on a scale from 0 to 10, where 0 means ‘not at all’ and 10 means ‘the most possible’?” Box 2-1 presents some quantitative data for three fictitious

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research BOX 2.1 Example of Quantitative Data Question:

Data:

Thinking about the past week, how depressed would you say you have been on a scale from 0 to 10, where 0 means “not at all” and 10 means “the most possible?” 9 (Subject 1) 0 (Subject 2) 4 (Subject 3)

respondents. The subjects have provided a number corresponding to their degree of depression—9 for subject 1 (a high level of depression), 0 for subject 2 (no depression), and 4 for subject 3 (little depression). The numeric values for all subjects in the study, collectively, would comprise the data on depression. In qualitative studies, the researcher collects primarily qualitative data, that is, narrative descriptions. Narrative information can be obtained by having conversations with the participants, by making detailed notes about how participants behave in naturalistic settings, or by obtaining narrative records from participants, such as diaries.

33

Suppose we were studying depression qualitatively. Box 2-2 presents qualitative data for three participants responding conversationally to the question, “Tell me about how you’ve been feeling lately— have you felt sad or depressed at all, or have you generally been in good spirits?” Here, the data consist of rich narrative descriptions of each participant’s emotional state. Typically, an operation known as coding is required to make research data amenable to analysis. In quantitative studies, coding is the process of translating verbal data into numeric form. For example, answers to a question about a subject’s gender might be coded “1” for female and “2” for male (or vice versa). In qualitative coding, researchers develop coding categories that represent important themes in the data. Relationships Researchers are rarely interested in a single isolated concept or phenomenon, except in descriptive studies. As an example of a descriptive study, a researcher might do research to determine the percentage of patients receiving intravenous (IV) therapy who experience IV infiltration. In this example, the variable is IV infiltration versus no infiltration. Usually, however, researchers study phenomena in relation to other phenomena—that is, they explore

BOX 2.2 Example of Qualitative Data Question: Data:

Tell me about how you’ve been feeling lately—have you felt sad or depressed at all, or have you generally been in good spirits? Well, actually, I’ve been pretty depressed lately, to tell you the truth. I wake up each morning and I can’t seem to think of anything to look forward to. I mope around the house all day, kind of in despair. I just can’t seem to shake the blues, and I’ve begun to think I need to go see a shrink. (Participant 1) I can’t remember ever feeling better in my life. I just got promoted to a new job that makes me feel like I can really get ahead in my company. And I’ve just gotten engaged to a really great guy who is very special. (Participant 2) I’ve had a few ups and downs the past week, but basically things are on a pretty even keel. I don’t have too many complaints. (Participant 3)

34

PART 1 Foundations of Nursing Research

or test relationships. A relationship is a bond or a connection between phenomena. For example, researchers repeatedly have found a relationship between cigarette smoking and lung cancer. Both qualitative and quantitative studies examine relationships, but in different ways. In quantitative studies, researchers are primarily interested in the relationship between the independent variables and dependent variables. The research question focuses on whether variation in the dependent variable is systematically related to variation in the independent variable. Relationships are usually expressed in quantitative terms, such as more than, less than, and so on. For example, let us consider as our dependent variable a person’s body weight. What variables are related to (associated with) a person’s weight? Some possibilities are height, caloric intake, and exercise. For each of these independent variables, we can make a prediction about the nature of the relationship to the dependent variable: Height: Taller people will weigh more than shorter people. Caloric intake: People with higher caloric intake will be heavier than those with lower caloric intake. Exercise: The lower the amount of exercise, the greater will be the person’s weight. Each statement expresses a predicted relationship between weight (the dependent variable) and a measurable independent variable. Terms such as more than and heavier than imply that as we observe a change in one variable, we are likely to observe a corresponding change in weight. If Nate is taller than Tom, we would predict (in the absence of any other information) that Nate is also heavier than Tom. Most quantitative studies are undertaken to determine whether relationships exist among variables. Quantitative studies typically address one or more of the following questions about relationships: • Does a relationship between variables exist? (e.g., is cigarette smoking related to lung cancer?) • What is the direction of the relationship between variables? (e.g., are people who smoke

more likely or less likely to get lung cancer than those who do not?) • How strong is the relationship between the variables? (e.g., how powerful is the relationship between smoking and lung cancer? How probable is it that smokers will be lung cancer victims?) • What is the nature of the relationship between variables? (e.g., does smoking cause lung cancer? Does some other factor cause both smoking and lung cancer?) As this last question suggests, quantitative variables can be related to one another in different ways. One type of relationship is referred to as a cause-and-effect (or causal) relationship. Within the positivist paradigm, natural phenomena are assumed not to be random or haphazard; if phenomena have antecedent factors or causes, they are presumably discoverable. For instance, in our example about a person’s weight, we might speculate that there is a causal relationship between caloric intake and weight: consuming more calories causes weight gain. Example of a study of causal relationships: Keller and Trevi˜no (2001) studied whether a regimen of walking (and different frequencies of walking) caused reductions in cardiovascular risk factors, such as obesity and high blood lipids, in Mexican-American women. Not all relationships between variables can be interpreted as cause-and-effect relationships. There is a relationship, for example, between a person’s pulmonary artery and tympanic temperatures: people with high readings on one tend to have high readings on the other. We cannot say, however, that pulmonary artery temperature caused tympanic temperature, nor that tympanic temperature caused pulmonary artery temperature, despite the relationship that exists between the two variables. This type of relationship is sometimes referred to as a functional relationship (or an associative relationship) rather than as a causal relationship. Example of a study of functional relationships: Pressler and Hepworth (2002) examined the relationship between preterm neonate’s behavioral

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

competence on the one hand, and the infant’s gender and race on the other. Qualitative researchers are not concerned with quantifying relationships, nor in testing and confirming causal relationships. Rather, qualitative researchers seek patterns of association as a way of illuminating the underlying meaning and dimensionality of phenomena of interest. Patterns of interconnected themes and processes are identified as a means of understanding the whole. In some qualitative studies, theories are generated by identifying relationships between emerging categories. These new connections help to “weave the fractured story back together after the data have been analyzed” (Glaser, 1978, p. 72). Example of a qualitative study of patterns: Lam and Mackenzie (2002) explored Chinese parents’ experiences in parenting a child with Down syndrome. One major theme that emerged in the indepth interviews was parental acceptance of the child. Although the researchers had not specifically sought to examine differences between mothers and fathers, they noted that mothers and fathers did not accept their child at the same pace. KEY CHALLENGES OF CONDUCTING RESEARCH Researchers face numerous challenges in conducting research, including the following: • Conceptual challenges (How should key concepts be defined? What are the theoretical underpinnings of the study?) • Financial challenges (How will the study be paid for? Will available resources be adequate?) • Administrative challenges (Is there sufficient time to complete the study? Can the flow of tasks be adequately managed?) • Practical challenges (Will there be enough study participants? Will institutions cooperate in the study?) • Ethical challenges (Can the study achieve its goals without infringing on human or animal rights?)

35

• Clinical challenges (Will the research goals conflict with clinical goals? What difficulties will be encountered in doing research with vulnerable or frail patients?) • Methodologic challenges (Will the methods used to address the research question yield accurate and valid results?) Most of this book provides guidance relating to the last question, and this section highlights key methodologic challenges. However, other challenges are also discussed in this book.* Reliability, Validity, and Trustworthiness Researchers want their findings to reflect the truth. Research cannot contribute evidence to guide clinical practice if the findings are inaccurate, biased, fail adequately to represent the experiences of the target group, or are based on a misinterpretation of the data. Consumers of research need to assess the quality of evidence offered in a study by evaluating the conceptual and methodologic decisions the researchers made, and producers of research need to strive to make good decisions to produce evidence of the highest possible quality. Quantitative researchers use several criteria to assess the quality of a study, and two of the most important criteria are reliability and validity. Reliability refers to the accuracy and consistency of information obtained in a study. The term is most often associated with the methods used to measure research variables. For example, if a thermometer measured Bob’s temperature as 98.1!F one minute and as 102.5!F the next minute, the reliability of the thermometer would be highly suspect. The concept of reliability is also important in interpreting the results of statistical analyses. Statistical reliability refers to the probability that the same results would be obtained with a completely new sample of subjects—that is, that the results are an *The

following chapters present relevant materials: conceptual issues—Chapter 6; financial issues—Chapter 25; administrative, practical, and clinical issues—Chapter 4; and ethical issues—Chapter 7.

36

PART 1 Foundations of Nursing Research

accurate reflection of a wider group than just the particular people who participated in the study. Validity is a more complex concept that broadly concerns the soundness of the study’s evidence—that is, whether the findings are cogent, convincing, and well grounded. Like reliability, validity is an important criterion for assessing the methods of measuring variables. In this context, the validity question is whether there is evidence to support the assertion that the methods are really measuring the abstract concepts that they purport to measure. Is a paper-andpencil measure of depression really measuring depression? Or is it measuring something else, such as loneliness, low self-esteem, or stress? The importance of having solid conceptual definitions of research variables—as well as high-quality methods to operationalize them—should be apparent. Another aspect of validity concerns the quality of the researcher’s evidence regarding the effect of the independent variable on the dependent variable. Did a nursing intervention really bring about improvements in patients’ outcomes—or were other factors responsible for patients’ progress? Researchers make numerous methodologic decisions that can influence this type of study validity. Qualitative researchers use somewhat different criteria (and different terminology) in evaluating a study’s quality. In general, qualitative researchers discuss methods of enhancing the trustworthiness of the study’s data (Lincoln & Guba, 1985). Trustworthiness encompasses several different dimensions—credibility, transferability (discussed later in the chapter), confirmability, and dependability. Dependability refers to evidence that is consistent and stable. Confirmability is similar to objectivity; it is the degree to which study results are derived from characteristics of participants and the study context, not from researcher biases. Credibility, an especially important aspect of trustworthiness, is achieved to the extent that the research methods engender confidence in the truth of the data and in the researchers’ interpretations of the data. Credibility in a qualitative study can be enhanced through various approaches (see Chapter 18), but one in particular merits early discussion because it has implications for the design of all

studies, including quantitative ones. Triangulation is the use of multiple sources or referents to draw conclusions about what constitutes the truth. In a quantitative study, this might mean having alternative operational definitions of a dependent variable to determine if predicted effects are consistent across the two. In a qualitative study, triangulation might involve trying to understand the full complexity of a poorly understood phenomenon by using multiple means of data collection to converge on the truth (e.g., having in-depth discussions with study participants, as well as watching their behavior in natural settings). Nurse researchers are also beginning to triangulate across paradigms—that is, to integrate both qualitative and quantitative data in a single study to offset the shortcomings of each approach. Example of triangulation: Tarzian (2000) used triangulation of data methods in her qualitative study on caring for dying patients with air hunger. Tarzian interviewed 10 nurses who had cared for air-hungry patients and, to complement the nurses’ accounts, two family members who witnessed spouses suffering from air hunger. Trustworthiness of the study findings was enhanced because family members confirmed important themes. For example, nurses disclosed that air hunger evoked a physical effect, such as feeling out of breath just watching patients struggling to breathe. Family members supported this theme. One husband recalled, “My chest hurt just watching her, breathing like that all day long” (p. 139). Nurse researchers need to design their studies in such a way that threats to the reliability, validity, and trustworthiness of their studies are minimized. This book offers advice on how to do this. Bias Bias is a major concern in designing a study because it can threaten the study’s validity and trustworthiness. In general, a bias is an influence that produces a distortion in the study results. Biases can affect the quality of evidence in both qualitative and quantitative studies.

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

Bias can result from a number of factors, including the following: • Study participants’ candor. Sometimes people distort their behavior or their self-disclosures (consciously or subconsciously) in an effort to present themselves in the best possible light. • Subjectivity of the researcher. Investigators may distort information in the direction of their preconceptions, or in line with their own experiences. • Sample characteristics. The sample itself may be biased; for example, if a researcher studies abortion attitudes but includes only members of right-to-life (or pro-choice) groups in the sample, the results would be distorted. • Faulty methods of data collection. An inadequate method of capturing key concepts can lead to biases; for example, a flawed paper-andpencil measure of patient satisfaction with nursing care may exaggerate or underestimate patients’ complaints. • Faulty study design. A researcher may not have structured the study in such a way that an unbiased answer to the research question can be achieved. To some extent, bias can never be avoided totally because the potential for its occurrence is so pervasive. Some bias is haphazard and affects only small segments of the data. As an example of such random bias, a handful of study participants might fail to provide totally accurate information as a result of extreme fatigue at the time the data were collected. Systematic bias, on the other hand, results when the bias is consistent or uniform. For example, if a spring scale consistently measured people’s weights as being 2 pounds heavier than their true weight, there would be systematic bias in the data on weight. Rigorous research methods aim to eliminate or minimize systematic bias—or, at least, to detect its presence so it can be taken into account in interpreting the data. Researchers adopt a variety of strategies to address bias. Triangulation is one such approach, the idea being that multiple sources of information or points of view can help counterbalance biases and offer avenues to identify them. Quantitative

37

researchers use various methods to combat the effects of bias, and many of these entail research control. Research Control One of the central features of quantitative studies is that they typically involve efforts to control tightly various aspects of the research. Research control involves holding constant other influences on the dependent variable so that the true relationship between the independent and dependent variables can be understood. In other words, research control attempts to eliminate contaminating factors that might cloud the relationship between the variables that are of central interest. The issue of contaminating factors—or extraneous variables, as they are called—can best be illustrated with an example. Suppose we were interested in studying whether teenage women are at higher risk of having low-birth-weight infants than are older mothers because of their age. In other words, we want to test whether there is something about women’s maturational development that causes differences in birth weight. Existing studies have shown that, in fact, teenagers have a higher rate of low-birth-weight babies than women in their 20s. The question here is whether maternal age itself (the independent variable) causes differences in birth weight (the dependent variable), or whether there are other mechanisms that account for the relationship between age and birth weight. We need to design a study so as to control other influences on the dependent variable—influences that are also related to the independent variable. Two variables of interest are the mother’s nutritional habits and her prenatal care. Teenagers tend to be less careful than older women about their eating patterns during pregnancy, and are also less likely to obtain adequate prenatal care. Both nutrition and the amount of care could, in turn, affect the baby’s birth weight. Thus, if these two factors are not controlled, then any observed relationship between mother’s age and her baby’s weight at birth could be caused by the mother’s age itself, her diet, or her prenatal care. It would be impossible to know what the underlying cause really is.

38

PART 1 Foundations of Nursing Research

These three possible explanations might be portrayed schematically as follows: 1. Mother’s ageSinfant birth weight 2. Mother’s age Sprenatal care Sinfant birth weight 3. Mother’s age Snutrition Sinfant birth weight The arrows here symbolize a causal mechanism or an influence. In examples 2 and 3, the effect of maternal age on infant birth weight is mediated by prenatal care and nutrition, respectively; these variables would be considered mediating variables in these last two models. Some research is specifically designed to test paths of mediation, but in the present example these variables are extraneous to the research question. Our task is to design a study so that the first explanation can be tested. Both nutrition and prenatal care must be controlled if our goal is to learn if explanation 1 is valid. How can we impose such control? There are a number of ways, as discussed in Chapter 9, but the general principle underlying each alternative is the same: the extraneous variables of the study must be held constant. The extraneous variables must somehow be handled so that, in the context of the study, they are not related to the independent or dependent variable. As an example, let us say we want to compare the birth weights of infants born to two groups of women: those aged 15 to 19 years and those aged 25 to 29 years. We must then design a

study in such a way that the nutritional and prenatal health care practices of the two groups are comparable, even though, in general, the two groups are not comparable in these respects. Table 2-2 illustrates how we might deliberately select subjects for the study in such a way that both older and younger mothers had similar eating habits and amounts of prenatal attention; the two groups have been matched in terms of the two extraneous variables; one third of both groups have the same nutrition ratings and amount of prenatal care. By building in this comparability, nutrition and prenatal care have been held constant in the two groups. If groups differ in birth weight (as they, in fact, do in Table 2-2), then we might infer that age (and not diet or prenatal care) influenced the infants’ birth weights. If the two groups did not differ, however, we might tentatively conclude that it is not mother’s age per se that causes young women to have a higher percentage of lowbirth-weight babies, but rather some other variable, such as nutrition or prenatal care. It is important to note that although we have designated prenatal care and nutrition as extraneous variables in this particular study, they are not at all extraneous to a full understanding of the factors that influence birth weight; in other studies, nutritional practices and frequency of prenatal care might be key independent variables. By exercising control in this example, we have taken a step toward explaining the relationship between variables. The world is complex, and many

TABLE 2.2 Fictitious Example: Control of Two Extraneous Variables AGE OF MOTHER (YEARS)

NUTRITIONAL PRACTICES

NO. OF PRENATAL VISITS

INFANT BIRTH WEIGHT

15–19

33% rated “good” 33% rated “fair” 33% rated “poor”

33% 1–3 visits 33% 4–6 visits 33% # 6 visits

20% " 2500 g; 80% # 2500 g

25–29

33% rated “good” 33% rated “fair” 33% rated “poor”

33% 1–3 visits 33% 4–6 visits 33% # 6 visits

9% " 2500 g; 91% # 2500 g

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

variables are interrelated in complicated ways. When studying a particular problem within the positivist paradigm, it is difficult to examine this complexity directly; researchers must usually analyze a couple of relationships at a time and put pieces together like a jigsaw puzzle. That is why even modest studies can make contributions to knowledge. The extent of the contribution in a quantitative study, however, is often directly related to how well researchers control contaminating influences. In the present example, we identified three variables that could affect birth weight, but dozens of others might be relevant, such as maternal stress, mothers’ use of drugs or alcohol during pregnancy, and so on. Researchers need to isolate the independent and dependent variables in which they are interested and then pinpoint from dozens of possible candidates those extraneous variables that need to be controlled. Example of control through matching: Mackey, Williams, and Tiller (2000) compared the stress and birth outcomes of women who experienced preterm labor during pregnancy with those who did not. To keep the groups similar, the groups were matched in terms of age, race, parity, gestational age, and method of hospital payment.

Maternal age

Maternal nutritional practices

39

It is often impossible to control all variables that affect the dependent variable, and not even necessary to do so. Extraneous variables need to be controlled only if they simultaneously are related to both the dependent and independent variables. This notion is illustrated in Figure 2-1, which has the following elements: • Each circle represents all the variability associated with a particular variable. • The large circle in the center stands for the dependent variable, infant birth weight. • Smaller circles stand for factors contributing to infant birth weight. • Overlapping circles indicate the degree to which the variables are related to each other. In this hypothetical example, four variables are related to infant birth weight: mother’s age, amount of prenatal care, nutritional practices, and smoking during pregnancy. The first three of these variables are also interrelated; this is shown by the fact that these three circles overlap not only with infant birth weight but also with each other. That is, younger mothers tend to have different patterns of prenatal care and nutrition than older mothers. The mother’s prenatal use of cigarettes, however, is unrelated to these three

Prenatal care

Infant birth weight

Maternal use of cigarettes

FIGURE 2.1 Hypothetical representation of factors affecting infant birth weight.

40

PART 1 Foundations of Nursing Research

variables. In other words, women who smoke during their pregnancies (according to this fictitious representation) are as likely to be young as old, to eat properly as not, and to get adequate prenatal care as not. If this representation were accurate, then maternal smoking would not be need to be controlled to study the effect of maternal age on infant birth weight. If this scheme is incorrect—if teenage mothers smoke more or less than older mothers—then maternal smoking practices should be controlled. Figure 2-1 does not represent infant birth weight as being totally determined by the four other variables. The darkened area of the birth weight circle designates “unexplained” variability in infant birth weight. That is, other determinants of birth weight are needed for us to understand fully what causes babies to be born weighing different amounts. Genetic characteristics, events occurring during the pregnancy, and medical treatments administered to pregnant women are examples of other factors that contribute to an infant’s weight at birth. Dozens, and perhaps hundreds, of circles would need to be sketched onto Figure 2-1 for us to understand factors affecting infant birth weight. In designing a study, quantitative researchers should attempt to control those variables that overlap with both independent and dependent variables to understand fully the relationship between the main variables of interest. Research control in quantitative studies is viewed as a critical tool for managing bias and for enhancing the validity of researchers’ conclusions. There are situations, however, in which too much control can introduce bias. For example, if researchers tightly control the ways in which key study variables can manifest themselves, it is possible that the true nature of those variables will be obscured. When the key concepts are phenomena that are poorly understood or the dimensions of which have not been clarified, then an approach that allows some flexibility is better suited to the study aims—such as in a qualitative study. Research rooted in the naturalistic paradigm does not impose controls. With their emphasis on holism and the individuality of human experience, qualitative researchers typically adhere to the view that to

impose controls on a research setting is to remove irrevocably some of the meaning of reality. Randomness For quantitative researchers, a powerful tool for eliminating bias concerns the concept of randomness— having certain features of the study established by chance rather than by design or personal preference. When people are selected at random to participate in the study, for example, each person has an equal probability of being selected. This in turn means that there are no systematic biases in the make-up of the sample. Men are as likely to be selected as women, for example. Randomness is a compelling method of controlling extraneous variables. Qualitative researchers almost never consider randomness a desirable tool for fully understanding a phenomenon. Qualitative researchers tend to use information obtained early in the study in a purposive (nonrandom) fashion to guide their inquiry and to pursue information-rich sources that can help them expand or refine their conceptualizations. Researchers’ judgments are viewed as indispensable vehicles for uncovering the complexities of the phenomena of interest. Generalizability and Transferability Nurses increasingly rely on evidence from disciplined research as a guide in their clinical practice. If study findings are totally unique to the people, places, or circumstances of the original research, can they be used as a basis for changes in practice? The answer, clearly, is no. As noted in Chapter 1, generalizability is the criterion used in a quantitative study to assess the extent to which the findings can be applied to other groups and settings. How do researchers enhance the generalizability of a study? First and foremost, they must design studies strong in reliability and validity. There is little point in wondering whether results are generalizable if they are not accurate or valid. In selecting subjects, researchers must also give thought to the types of people to

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

whom the results might be generalized—and then select them in such a way that a nonbiased sample is obtained. If a study is intended to have implications for male and female patients, then men and women should be included as participants. If an intervention is intended to benefit patients in urban and rural hospitals, then perhaps a multisite study is warranted. Chapter 10 describes other issues to consider in evaluating generalizability. Qualitative researchers do not specifically seek to make their findings generalizable. Nevertheless, qualitative researchers often seek understandings that might prove useful in other situations. Lincoln and Guba (1985), in their highly influential book on naturalistic inquiry, discuss the concept of transferability, the extent to which qualitative findings can be transferred to other settings, as another aspect of a study’s trustworthiness. An important mechanism for promoting transferability is the amount of information qualitative researchers provide about the contexts of their studies. Thick description, a widely used term among qualitative researchers, refers to a rich and thorough description of the research setting and of observed transactions and processes. Quantitative researchers, like qualitative researchers, need to describe their study participants and their research settings thoroughly so that the utility of the evidence for others can be assessed. Replication Virtually every study has flaws or limitations. Even the most rigorous study is likely to contain some bias, or to engender unresolved questions about the validity or trustworthiness of the findings. And few studies are broad enough that findings can be generalized to all groups or settings of interest. Nursing practice is almost never changed on the basis of a single study, no matter how sound. Evidence-based practice generally builds on accumulated evidence. Replications are attempts to validate the findings from one study in an independent inquiry. Replication is, in effect, a form of triangulation—the use of multiple sources and referents (multiple findings) to draw conclusions

41

about the validity or truth of findings. Replication research is critical for the development of nursing science. Yet, remarkably, there is a dearth of replication studies—or, at least, published replication studies. This may reflect a strong preference on the part of researchers, editors, and funders for originality and “breaking new ground.” “Paving the way,” however, is just as critical as breaking new ground, and wellplanned and well-executed replication studies are an important paving tool on the road to evidencebased practice. Some strategies for replication are described in Chapter 10. RESEARCH EXAMPLES This section presents brief overviews of a quantitative and a qualitative study. These overviews deal primarily with key concepts that were presented in this chapter. You may wish to consult the full research report in thinking about differences in style and content of qualitative and quantitative reports. Research Example of a Quantitative Study Health care strategies for urinary incontinence (UI) have emerged and been tested in several studies of community-dwelling women. Dougherty and her coresearchers (2002) noted, however, that health care strategies designed for adults in urban settings do not always transfer well to rural environments. They designed a study to implement and test the efficacy of a behavioral management for continence (BMC) intervention for older women with UI in seven rural counties in north Florida. The intervention involved self-monitoring, bladder training, and pelvic muscle exercise with biofeedback in the women’s homes. Over a 2-year period, 218 women aged 55 years and older who had regular involuntary urine loss were recruited for the study. Half the subjects were selected, at random, to receive the intervention. This procedure permitted a rigorous comparison of the outcomes of the two groups who, because selection into them was random, were presumably alike in all respects—except for receipt of the intervention.

42

PART 1 Foundations of Nursing Research

Group membership (i.e., whether a woman was in the BMC group) was the independent variable. Both groups received follow-up visits, during which time outcome data were gathered, every 6 months for up to 2 years. The primary dependent variable was urine loss. This was operationalized as the amount of urine lost in grams per 24 hours, as measured by the change in weight of incontinence pads. Secondary dependent variables relating to urinary outcomes included measures obtained from 3-day bladder diaries that subjects maintained (e.g., micturition frequency and episodes of urine loss). In addition, the researchers assessed the effect of the intervention on subjects’ quality of life. This concept was operationalized using a paper-and-pencil instrument known as the Incontinence Impact Questionnaire (IIQ). The IIQ, which involved 26 questions about the extent to which incontinence affected functioning in various areas (e.g., daily living, social interactions, self-perception), previously had been shown to be a reliable and valid indicator of quality of life. The findings were encouraging. Over the 2 years in which the women were followed, the BMC group sustained UI improvement, whereas those in the other group experienced worsening severity in urine loss. The two groups also differed at follow-up with regard to episodes of urine loss and quality of life. The study was methodologically strong. Half the women, selected at random, received the special intervention and the other half did not. This is a particularly powerful way to control extraneous variables. Although the number of subjects was fairly small— and therefore replications are clearly needed—it is noteworthy that the sample was drawn from seven different rural counties.

Research Example of a Qualitative Study Wise (2002) examined the experience of children who received liver transplants from the time before transplantation, through the surgery, and after. The sample consisted of nine children between the ages of 7 and 15 years. Wise conducted all the interviews with the children herself either in their homes or in an outpatient setting. These conversations ranged in length from 20 to 40 minutes. The interviews were audiotaped and transcribed.

Before the interviews, Wise asked the children if they would draw two pictures of themselves, one before the transplantation and one that reflected their present status. The purpose of this artwork was to help the children relax and also to provide an opening for the interviews. An art therapist who interpreted the children’s artwork served as a consultant for this qualitative study; the artwork thereby provided an opportunity for triangulation. The qualitative data obtained from the interviews were analyzed and interpreted to discover the underlying themes of the children’s experiences. Wise used thick description in reporting her results. Four themes emerged that described the essence of the phenomenon of pediatric liver transplantation: (1) search for connections with peers before and after transplantation, and also for connections with the donor, (2) ordinary and extraordinary experiences of hospitalization, (3) painful responses and feelings of being out of control, and (4) parents’ responses to the illness. The following quote illustrates this fourth theme and is an example of Wise’s thick description: I will never tell my Mom how I feel about anything. I don’t think I would ever tell the truth because I would never want to upset her. I can just see the statement on her face. I know how she feels . . . she has been through so much stuff with me. I basically worry if she is all right instead of me (p. 86).

Wise engaged in a number of activities to establish the rigor of her study. To enhance trustworthiness, for instance, she maintained a journal in which she documented her observations, analysis decisions, and so on. Credibility was established by having an older adolescent validate the themes and also by having an advisor and three colleagues review her findings.

S U M M A RY P O I N T S • A research study (or investigation or research project) is undertaken by one or more researchers (or investigators or scientists). The people who provide information to the researchers are referred to as subjects, study participants, or respondents (in quantitative research) or study participants or informants in qualitative research; collectively they comprise the sample.

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

• Collaborative research involving a team of nurses with both clinical and methodologic expertise is increasingly common in addressing problems of clinical relevance. • The site is the overall location for the research; researchers sometimes engage in multisite studies. Settings are the more specific places where data collection will occur. Settings for nursing research can range from totally naturalistic environments to formal laboratories. • Researchers investigate concepts and phenomena (or constructs), which are abstractions or mental representations inferred from behavior or events. • Concepts are the building blocks of theories, which are systematic explanations of some aspect of the real world. • In quantitative studies, concepts are referred to as variables. A variable is a characteristic or quality that takes on different values (i.e., a variable varies from one person or object to another). • Variables that are inherent characteristics of a person that the researcher measures or observes are attribute variables. When a researcher actively creates a variable, as when a special intervention is introduced, the variable is an active variable. • Variables that can take on an infinite range of values along a continuum are continuous variables (e.g., height and weight). A discrete variable, by contrast, is one that has a finite number of values between two points (e.g., number of children). Variables with distinct categories that do not represent a quantity are categorical variables (e.g., gender and blood type). • The dependent variable is the behavior, characteristic, or outcome the researcher is interested in understanding, explaining, predicting, or affecting. The independent variable is the presumed cause of, antecedent to, or influence on the dependent variable. • Groups that are highly varied with respect to some attribute are described as heterogeneous; groups with limited variability are described as homogeneous. • A conceptual definition elucidates the abstract or theoretical meaning of the concepts being

43

studied. An operational definition is the specification of the procedures and tools required to measure a variable. • Data—the information collected during the course of a study—may take the form of narrative information (qualitative data) or numeric values (quantitative data). • Researchers often focus on relationships between two or more concepts. A relationship is a bond or connection (or pattern of association) between two phenomena. Quantitative researchers focus on the relationship between the independent variables and dependent variables. • When the independent variable causes or affects the dependent variable, the relationship is a cause-and-effect (or causal) relationship. In a functional or associative relationship, variables are related in a noncausal way. • Researchers face numerous conceptual, practical, ethical, and methodologic challenges. The major methodologic challenge is designing studies that are reliable and valid (quantitative studies) or trustworthy (qualitative studies). • Reliability refers to the accuracy and consistency of information obtained in a study. Validity is a more complex concept that broadly concerns the soundness of the study’s evidence—that is, whether the findings are cogent, convincing, and well grounded. • Trustworthiness in qualitative research encompasses several different dimensions. Dependability refers to evidence that is believable, consistent, and stable over time. Confirmability refers to evidence of the researcher’s objectivity. Credibility is achieved to the extent that the research methods engender confidence in the truth of the data and in the researchers’ interpretations of the data. • Triangulation, the use of multiple sources or referents to draw conclusions about what constitutes the truth, is one approach to establishing credibility. • A bias is an influence that produces a distortion in the study results. Systematic bias results when a bias is consistent or uniform across study participants or situations.

44

PART 1 Foundations of Nursing Research

• In quantitative studies, research control is used to hold constant outside influences on the dependent variable so that the relationship between the independent and dependent variables can be better understood. • The external influences the researcher seeks to control are extraneous variables—extraneous to the purpose of a specific study. There are a number of ways to control such influences, but the general principle is that the extraneous variables must be held constant. • For a quantitative researcher, a powerful tool to eliminate bias concerns randomness—having certain features of the study established by chance rather than by design or personal preference. • Generalizability is the criterion used in a quantitative study to assess the extent to which the findings can be applied to other groups and settings. A similar concept in qualitative studies is transferability, the extent to which qualitative findings can be transferred to other settings. An important mechanism for promoting transferability is thick description, the rich and thorough description of the research setting or context so that others can make inferences about contextual similarities • Replications, which are attempts to validate the findings from one study in an independent inquiry, are a crucial form of triangulation. Replication research is essential for the development of nursing science and evidence-based practice. STUDY ACTIVITIES Chapter 2 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Suggest ways of conceptually and operationally defining the following concepts: nursing competency, aggressive behavior, pain, home health hazards, postsurgical recovery, and body image.

2. Name five continuous, five discrete, and five categorical variables; identify which, if any, are dichotomous. 3. Identify which of the following variables could be active variables and which are attribute variables (some may be both): height, degree of fatigue, cooperativeness, noise level on hospital units, length of stay in hospital, educational attainment, self-esteem, nurses’ job satisfaction. 4. In the following research problems, identify the independent and dependent variables: a. How do nurses and physicians differ in the ways they view the extended role concept for nurses? b. Does problem-oriented recording lead to more effective patient care than other recording methods? c. Do elderly patients have lower pain thresholds than younger patients? d. How are the sleeping patterns of infants affected by different forms of stimulation? e. Can home visits by nurses to released psychiatric patients reduce readmission rates? SUGGESTED READINGS Methodologic References Glaser, B. (1978). Theoretical sensitivity. Mill Valley, CA: The Sociology Press. Kerlinger, F. N., & Lee, H. B. (2000). Foundations of behavioral research (4th ed.). Orlando, FL: Harcourt College Publishers. Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. Newbury Park, CA: Sage. Morse, J. M., Solberg, S. M., Neander, W. L., Bottorff, J. L., & Johnson, J. L. (1990). Concepts of caring and caring as a concept. Advances in Nursing Science, 13, 1–14. Morse, J. M., & Field, P. A. (1995). Qualitative research methods for health professionals (2nd ed.). Thousand Oaks, CA: Sage. Sandelowski, M. (1993). Theory unmasked: The uses and guises of theory in qualitative research. Research in Nursing & Health, 16, 213–218.

CHAPTER 2 Key Concepts and Terms in Qualitative and Quantitative Research

Studies Cited in Chapter 2 Beck, C. T. (2002). Releasing the pause button: Mothering twins during the first year of life, Qualitative Health Research, 12, 593–608. Beck, C. T., & Gable, R. K. (2001). Ensuring content validity: An illustration of the process. Journal of Nursing Measurement, 9, 201–215. Carlisle, C. (2000). The search for meaning in HIV and AIDS: The carer’s experience. Qualitative Health Research, 10, 750–765. Dougherty, M., Dwyer, J., Pendergast, J., Boyington, A., Tomlinson, B., Coward, R., Duncan, R. P., Vogel, B., & Rooks, L. (2002). A randomized trial of behavioral management for continence with older rural women. Research in Nursing & Health, 25, 3–13. Keller, C., & Trevi˜no, R. P. (2001). Effects of two frequencies of walking on cardiovascular risk factor reduction in Mexican American women. Research in Nursing & Health, 24, 390–401. Lam, L., & Mackenzie, A. E. (2002). Coping with a child with Down syndrome. Qualitative Health Research, 12, 223–237. Mackey, M. C., Williams, C. A., & Tiller, C. M. (2000). Stress, pre-term labour and birth outcomes. Journal of Advanced Nursing, 32, 666–674.

45

Pierce, J. D. & Clancy, R. L. (2001). Effects of hypoxia on diaphragm activity in anesthetized rats. Journal of Perianesthesia Nursing, 16, 181–186. Polit, D. F., London, A., & Martinez, J. (2001). The health of poor urban women. New York: Manpower Demonstration Research Corporation. (Report available online at: www.mdrc.org.) Pressler, J. L., & Hepworth, J. T. (2002). A quantitative use of the NIDCAP® tool. Clinical Nursing Research, 11, 89–102. Tarzian, A. J. (2000). Caring for dying patients who have air hunger. Journal of Nursing Scholarship, 32, 137–143. Varda, K. E., & Behnke, R. S. (2000). The effect of timing of initial bath on newborn’s temperature. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 29, 27–32. Wise, B. (2002). In their own words: The lived experience of pediatric liver transplantation. Qualitative Health Research, 12, 74–90.

3

Overview of the Research Process in Qualitative and Quantitative Studies

R

esearchers usually work within a paradigm that is consistent with their world view, and that gives rise to the types of question that excite their curiosity. The maturity of the concept of interest also may lead to one or the other paradigm: when little is known about a topic, a qualitative approach is often more fruitful than a quantitative one. The progression of activities differs for qualitative and quantitative researchers; we discuss the flow of both in this chapter. First, however, we briefly describe broad categories of quantitative and qualitative research. MAJOR CLASSES OF Q U A N T I TAT I V E A N D Q U A L I TAT I V E R E S E A R C H Experimental and Nonexperimental Studies in Quantitative Research A basic distinction in quantitative studies is the difference between experimental and nonexperimental research. In experimental research, researchers actively introduce an intervention or treatment. In nonexperimental research, on the other hand, researchers collect data without making changes or introducing treatments. For example, if a researcher gave bran flakes to one group of subjects and prune juice to another to evaluate which method facilitated elimination more effectively, the

study would be experimental because the researcher intervened in the normal course of things. In this example, the researcher created an “active variable” involving a dietary intervention. If, on the other hand, a researcher compared elimination patterns of two groups of people whose regular eating patterns differed—for example, some normally took foods that stimulated bowel elimination and others did not—there is no intervention. Such a study focuses on existing attributes and is nonexperimental. Experimental studies are explicitly designed to test causal relationships. Sometimes nonexperimental studies also seek to elucidate or detect causal relationships, but doing so is tricky and usually is less conclusive. Experimental studies offer the possibility of greater control over extraneous variables than nonexperimental studies. Example of experimental research: Johnson (2001) tested the effects of a submaximal exercise protocol, in comparison with a near-maximal voluntary contraction protocol, on continence control and muscle contraction strength among women with genuine stress urinary incontinence. In this example, the researcher intervened by designating that some women would receive the submaximal exercise protocol and others would not. In other words, the researcher controlled the

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

47

independent variable, which in this case was the type of protocol.

process of managing late stages of breastfeeding and weaning the child from the breast.

Example of nonexperimental research: Wong and her co-researchers (2002) searched for factors that contributed to hospital readmission in a Hong Kong hospital. A readmitted group was compared with a nonreadmitted group of patients in terms of demographic characteristics and health conditions upon admission.

Phenomenology, which has its disciplinary roots in both philosophy and psychology and is rooted in a philosophical tradition developed by Husserl and Heidegger, is concerned with the lived experiences of humans. Phenomenology is an approach to thinking about what life experiences of people are like and what they mean. The phenomenological researcher asks the questions: What is the essence of this phenomenon as experienced by these people? Or, What is the meaning of the phenomena to those who experience it?

In this nonexperimental study, the researchers did not intervene in any way; they observed and measured subjects’ attributes. They explored whether there were identifiable characteristics and conditions that distinguished the two groups of patients, with the aim of discovering opportunities to reduce readmissions. Research Traditions in Qualitative Research Qualitative studies are often rooted in research traditions that originate in the disciplines of anthropology, sociology, and psychology. Three such traditions have had especially strong influences on qualitative nursing research and are briefly describe here. Chapter 11 provides a fuller discussion of alternative research traditions and the methods associated with them. The grounded theory tradition, which has its roots in sociology, seeks to describe and understand the key social psychological and structural processes that occur in a social setting. Grounded theory was developed in the 1960s by two sociologists, Glaser and Strauss (1967). The focus of most grounded theory studies is on a developing social experience—the social and psychological stages and phases that characterize a particular event or episode. A major component of grounded theory is the discovery of a core variable that is central in explaining what is going on in that social scene. Grounded theory researchers strive to generate comprehensive explanations of phenomena that are grounded in reality. Example of a grounded theory study: Hauck and Irurita (2002) conducted a grounded theory study to explain the maternal

Example of a phenomenological study: Sundin, Norberg, and Jansson (2001) conducted a phenomenological study to illuminate the lived experiences of care providers who were highly skilled communicators in their relationships with patients with stroke and aphasia. Ethnography is the primary research tradition within anthropology, and provides a framework for studying the meanings, patterns, and experiences of a defined cultural group in a holistic fashion. Ethnographers typically engage in extensive fieldwork, often participating to the extent possible in the life of the culture under study. Ethnographic research is in some cases concerned with broadly defined cultures (e.g., Haitian refugee communities), but sometimes focuses on more narrowly defined cultures (e.g., the culture of emergency departments). The aim of ethnographers is to learn from (rather than to study) members of a cultural group, to understand their world view as they perceive and live it. Example of an ethnographic study: Powers (2001) undertook an ethnographic analysis of a nursing home residence, focusing on the ethical issues of daily living affecting nursing home residents with dementia. MAJOR STEPS IN A Q U A N T I TAT I V E S T U D Y In quantitative studies, researchers move from the beginning point of a study (the posing of a question)

48

PART 1 Foundations of Nursing Research

to the end point (the obtaining of an answer) in a fairly linear sequence of steps that is broadly similar across studies. In some studies, the steps overlap, whereas in others, certain steps are unnecessary. Still, there is a general flow of activities that is typical of a quantitative study. This section describes that flow, and the next section describes how qualitative studies differ. Phase 1: The Conceptual Phase The early steps in a quantitative research project typically involve activities with a strong conceptual or intellectual element. These activities include reading, conceptualizing, theorizing, reconceptualizing, and reviewing ideas with colleagues or advisers. During this phase, researchers call on such skills as creativity, deductive reasoning, insight, and a firm grounding in previous research on the topic of interest. Step 1: Formulating and Delimiting the Problem One of the first things a researcher must do is develop a research problem and research questions. Good research depends to a great degree on good questions. Without a significant, interesting problem, the most carefully and skillfully designed research project is of little value. Quantitative researchers usually proceed from the selection of a broad problem area to the development of specific questions that are amenable to empirical inquiry. In developing a research question to be studied, nurse researchers must pay close attention to substantive issues (Is this research question significant, given the existing base of knowledge?); clinical issues (Could findings from this research be useful in clinical practice?); and methodologic issues (How can this question best be studied to yield high-quality evidence?). The identification of research questions must also take into consideration practical and ethical concerns. TIP: A critical ingredient in developing good research questions is personal interest. We offer this advice to those of you who plan to undertake a research project: Begin with topics that

fascinate you or about which you have a passionate interest or curiosity. Step 2: Reviewing the Related Literature Quantitative research is typically conducted within the context of previous knowledge. To build on existing theory or research, quantitative researchers strive to understand what is already known about a research problem. A thorough literature review provides a foundation on which to base new knowledge and usually is conducted well before any data are collected in quantitative studies. For clinical problems, it would likely also be necessary to learn as much as possible about the “status quo” of current procedures relating to the topic, and to review existing practice guidelines or protocols. A familiarization with previous studies can also be useful in suggesting research topics or in identifying aspects of a problem about which more research is needed. Thus, a literature review sometimes precedes the delineation of the research problem. Step 3: Undertaking Clinical Fieldwork In addition to refreshing or updating clinical knowledge based on written work, researchers embarking on a clinical nursing study benefit from spending time in clinical settings, discussing the topic with clinicians and health care administrators, and observing current practices. Sterling (2001) notes that such clinical fieldwork can provide perspectives on recent clinical trends, current diagnostic procedures, and relevant health care delivery models; it can also help researchers better understand affected clients and the settings in which care is provided. In addition to expanding the researchers’ clinical and conceptual knowledge, such fieldwork can be valuable in developing methodologic tools for strengthening the study. For example, in the course of clinical fieldwork researchers might learn what extraneous variables need to be controlled, or might discover the need for Spanish-speaking research assistants. As with literature reviews, clinical fieldwork may serve as a stimulus for developing research questions and may be the first step in the process for some researchers.

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

Step 4: Defining the Framework and Developing Conceptual Definitions Theory is the ultimate aim of science in that it transcends the specifics of a particular time, place, and group of people and aims to identify regularities in the relationships among variables. When quantitative research is performed within the context of a theoretical framework—that is, when previous theory is used as a basis for generating predictions that can be tested through empirical research—the findings may have broader significance and utility. Even when the research question is not embedded in a theory, researchers must have a clear sense of the concepts under study. Thus, an important task in the initial phase of a project is the development of conceptual definitions. Step 5: Formulating Hypotheses A hypothesis is a statement of the researcher’s expectations about relationships between the variables under investigation. Hypotheses, in other words, are predictions of expected outcomes; they state the relationships researchers expect to find as a result of the study. The research question identifies the concepts under investigation and asks how the concepts might be related; a hypothesis is the predicted answer. For example, the initial research question might be phrased as follows: Is preeclamptic toxemia in pregnant women associated with stress factors present during pregnancy? This might be translated into the following hypothesis: Pregnant women with a higher incidence of stressful events during pregnancy will be more likely than women with a lower incidence of stress to experience preeclamptic toxemia. Most quantitative studies are designed to test hypotheses through statistical analysis. Phase 2: The Design and Planning Phase In the second major phase of a quantitative research project, researchers make decisions about the methods and procedures to be used to address the research question, and plan for the actual

49

collection of data. Sometimes the nature of the question dictates the methods to be used, but more often than not, researchers have considerable flexibility to be creative and make many decisions. These methodologic decisions usually have crucial implications for the validity and reliability of the study findings. If the methods used to collect and analyze research data are seriously flawed, then the evidence from the study may be of little value. Step 6: Selecting a Research Design The research design is the overall plan for obtaining answers to the questions being studied and for handling some of the difficulties encountered during the research process. A wide variety of research designs is available for quantitative studies, including numerous experimental and nonexperimental designs. In designing the study, researchers specify which specific design will be adopted and what controls will be used to minimize bias and enhance the interpretability of results. In quantitative studies, research designs tend to be highly structured, with tight controls over extraneous variables. Research designs also indicate other aspects of the research—for example, how often subjects will be measured or observed, what types of comparisons will be made, and where the study will take place. The research design is essentially the architectural backbone of the study. Step 7: Developing Protocols for the Intervention In experimental research, researchers actively intervene and create the independent variable, which means that people in the sample will be exposed to different treatments or conditions. For example, if we were interested in testing the effect of biofeedback in treating hypertension, the independent variable would be biofeedback compared with either an alternative treatment (e.g., relaxation therapy), or with no treatment. The intervention protocol for the study would need to be developed, specifying exactly what the biofeedback treatment would entail (e.g., who would administer it, how frequently and over how long a period the treatment would last, what specific equipment would be used,

50

PART 1 Foundations of Nursing Research

and so on) and what the alternative condition would be. The goal of well-articulated protocols is to have all subjects in each group treated in the same way. (In nonexperimental research, of course, this step would not be necessary.) Step 8: Identifying the Population to be Studied Before selecting subjects, quantitative researchers need to know what characteristics participants should possess. Researchers and others using the findings also need to know to whom study results can be generalized. Thus, during the planning phase of quantitative studies, researchers must identify the population to be studied. The term population refers to the aggregate or totality of those conforming to a set of specifications. For example, we might specify nurses (RNs) and residence in the United States as attributes of interest; the study population would then consist of all licensed RNs who reside in the United States. We could in a similar fashion define a population consisting of all children younger than 10 years of age with muscular dystrophy in Canada, or all the change-of-shift reports for the year 2002 in Massachusetts General Hospital. Step 9: Designing the Sampling Plan Research studies almost always rely on a sample of subjects, who are a subset of the population. It is clearly more practical and less costly to collect data from a sample than from an entire population. The risk, however, is that the sample might not adequately reflect the population’s behaviors, traits, symptoms, or beliefs. Various methods of obtaining samples are available. These methods vary in cost, effort, and skills required, but their adequacy is assessed by the same criterion: the representativeness of the selected sample. That is, the quality of the sample for quantitative studies depends on how typical, or representative, the sample is of the population with respect to the variables of concern in the study. Sophisticated sampling procedures can produce samples that have a high likelihood of being representative. The most sophisticated methods are

probability sampling methods, which use random procedures for selecting subjects. In a probability sample, every member of the population has an equal probability of being included in the sample. With nonprobability sampling, by contrast, there is no way of ensuring that each member of the population could be selected; consequently, the risk of a biased (unrepresentative) sample is greater. The design of a sampling plan includes the selection of a sampling method, the specification of the sample size (i.e., number of subjects), and the development of procedures for recruiting subjects. Step 10: Specifying Methods to Measure the Research Variables Quantitative researchers must develop methods to observe or measure the research variables as accurately as possible. Based on the conceptual definitions, the researcher selects or designs appropriate methods of operationalizing the variables and collecting data. A variety of quantitative data collection approaches exist. Biophysiologic measurements often play an important role in clinical nursing research. Through self-reports, another popular method of data collection, subjects are asked directly about their feelings, behaviors, attitudes, and personal traits (for example, in an interview with research personnel). Another technique is observation, wherein researchers collect data by observing and recording aspects of people’s behavior. Data collection methods vary in the degree of structure imposed on subjects. Quantitative approaches tend to be fairly structured, involving the use of a formal instrument that elicits the same information from every subject. Sometimes researchers need to develop their own instruments, but more often they use or adapt measuring instruments that have been developed by others. The task of measuring research variables and developing a data collection plan is a complex and challenging process that permits a great deal of creativity and choice. Before finalizing the data collection plan, researchers must carefully evaluate whether the chosen methods capture key concepts accurately.

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

Step 11: Developing Methods for Safeguarding Human/Animal Rights Most nursing research involves human subjects, although some studies involve animals. In either case, procedures need to be developed to ensure that the study adheres to ethical principles. For example, forms often need to be developed to document that subjects’ participation in the study was voluntary. Each aspect of the study plan needs to be reviewed to determine whether the rights of subjects have been adequately protected. Often that review involves a formal presentation to an external committee. Step 12: Finalizing and Reviewing the Research Plan Before actually collecting research data, researchers often perform a number of “tests” to ensure that plans will work smoothly. For example, they may evaluate the readability of any written materials to determine if people with below-average reading skills can comprehend them, or they may need to test whether technical equipment is functioning properly. If questionnaires are used, it is important to know whether respondents understand questions or find certain ones objectionable; this is usually referred to as pretesting the questionnaire. During final study preparations, researchers also have to determine the type of training to provide to those responsible for collecting data. If researchers have concerns about their study plans, they may undertake a pilot study, which is a small-scale version or trial run of the major study. Normally, researchers have their research plan critiqued by peers, consultants, or other reviewers to obtain substantive, clinical, or methodologic feedback before implementing the plan. When researchers seek financial support for the study, a proposal typically is submitted to a funding source, and reviewers of the proposed plan usually suggest improvements. Students conducting a study as part of a course or degree requirement have their plans reviewed by faculty advisers. Even under other circumstances, however, researchers are well advised to ask individuals external to the project to assess preliminary plans.

51

Experienced researchers with fresh perspectives can often be invaluable in identifying pitfalls and shortcomings that otherwise might not have been recognized. Phase 3: The Empirical Phase The empirical portion of quantitative studies involves collecting research data and preparing those data for analysis. In many studies, the empirical phase is one of the most time-consuming parts of the investigation, although the amount of time spent collecting data varies considerably from one study to the next. If data are collected by distributing a written questionnaire to intact groups, this task may be accomplished in a matter of days. More often, however, data collection requires several weeks, or even months, of work. Step 13: Collecting the Data The actual collection of data in a quantitative study often proceeds according to a preestablished plan. The researcher’s plan typically specifies procedures for the actual collection of data (e.g., where and when the data will be gathered); for describing the study to participants; and for recording information. Technological advances in the past few decades have expanded possibilities for automating data collection. A considerable amount of both clerical and administrative work is required during data collection. Researchers typically must be sure, for example, that enough materials are available to complete the study; that participants are informed of the time and place that their presence may be required; that research personnel (such as interviewers) are conscientious in keeping their appointments; that schedules do not conflict; and that a suitable system of maintaining confidentiality of information has been implemented. Step 14: Preparing the Data for Analysis After data are collected, a few preliminary activities must be performed before data analysis begins. For instance, it is normally necessary to look through questionnaires to determine if they are usable.

52

PART 1 Foundations of Nursing Research

Sometimes forms are left almost entirely blank or contain other indications of misinterpretation or noncompliance. Another step is to assign identification numbers to the responses or observations of different subjects, if this was not done previously. Coding of the data is typically needed at this point. As noted in Chapter 2, coding involves the translation of verbal data into numeric form, according to a specified plan. This might mean assigning numeric codes to categorical variables such as gender (e.g., 1 for females and 2 for males). Coding might also be needed to categorize narrative responses to certain questions. For example, patients’ verbatim responses to a question about the quality of nursing care they received during hospitalization might be coded into positive reactions (1), negative reactions (2), neutral reactions (3), or mixed reactions (4). Another preliminary step involves transferring the data from written documents onto computer files for subsequent analysis. Phase 4: The Analytic Phase Quantitative data gathered in the empirical phase are not reported in raw form. They are subjected to analysis and interpretation, which occurs in the fourth major phase of a project. Step 15: Analyzing the Data The data themselves do not provide answers to research questions. Ordinarily, the amount of data collected in a study is rather extensive; research questions cannot be answered by a simple perusal of numeric information. Data need to be processed and analyzed in an orderly, coherent fashion. Quantitative information is usually analyzed through statistical procedures. Statistical analyses cover a broad range of techniques, from simple procedures that we all use regularly (e.g., computing an average) to complex and sophisticated methods. Although some methods are computationally formidable, the underlying logic of statistical tests is relatively easy to grasp, and computers have eliminated the need to get bogged down with detailed mathematic operations.

Step 16: Interpreting the Results Before the results of a study can be communicated effectively, they must be systematically interpreted. Interpretation is the process of making sense of the results and of examining their implications. The process of interpretation begins with an attempt to explain the findings within the context of the theoretical framework, prior empirical knowledge, and clinical experience. If research hypotheses have been supported, an explanation of the results may be straightforward because the findings fit into a previously conceived argument. If hypotheses are not supported, researchers must explain why this might be so. Is the underlying conceptualization wrong, or was it inappropriate for the research problem? Or do the findings reflect problems with the research methods rather than the framework (e.g., was the measuring tool inappropriate)? To provide sound explanations, researchers not only must be familiar with clinical issues, prior research, and conceptual underpinnings, but must be able to understand methodologic limitations of the study. In other words, the interpretation of the findings must take into account all available evidence about the study’s reliability and validity. Researchers need to evaluate critically the decisions they made in designing the study and to recommend alternatives to others interested in the same research problem. Phase 5: The Dissemination Phase The analytic phase brings researchers full circle: it provides answers to the questions posed in the first phase of the project. However, researchers’ responsibilities are not complete until the study results are disseminated. Step 17: Communicating the Findings A study cannot contribute evidence to nursing practice if the results are not communicated. The most compelling hypothesis, the most rigorous study, the most dramatic results are of no value to the nursing community if they are unknown. Another—and often final—task of a research project, therefore, is

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

the preparation of a research report that can be shared with others. Research reports can take various forms: term papers, dissertations, journal articles, presentations at professional conferences, and so on. Journal articles—reports appearing in such professional journals as Nursing Research—usually are the most useful because they are available to a broad, international audience. There is also a growing number of outlets for research dissemination on the Internet. Step 18: Utilizing the Findings in Practice Many interesting studies have been conducted by nurses without having any effect on nursing practice or nursing education. Ideally, the concluding step of a high-quality study is to plan for its utilization in practice settings. Although nurse researchers may not themselves be in a position to implement a plan for utilizing research findings, they can contribute to the process by including in their research reports recommendations regarding how the evidence from the study could be incorporated into the practice of nursing and by vigorously pursuing opportunities to disseminate the findings to practicing nurses. Organization of a Quantitative Research Project The steps described in the preceding section represent an idealized conception of what researchers do. The research process rarely follows a neatly prescribed pattern of sequential procedures. Developments in one step, for example, may require alterations in a previously completed activity. Nevertheless, for the quantitative researcher, careful organization is very important. Almost all research projects are conducted under some time pressure. Students in research courses have end-of-term deadlines; governmentsponsored research involves funds granted for a specified time. Those who may not have such formal time constraints (e.g., graduate students working on theses or dissertations) normally have their own goals for project completion. Setting up a timetable in advance may be an important means of

53

meeting such goals. Having deadlines for tasks— even tentative ones—helps to impose order and delimits tasks that might otherwise continue indefinitely, such as problem selection and literature reviews. It is not possible to give even approximate figures for the relative percentage of time that should be spent on each task in quantitative studies. Some projects require many months to develop and pretest the measuring instruments, whereas other studies use previously existing ones, for example. The write-up of the study may take many months or only a few days. Clearly, not all steps are equally time-consuming. It would make little sense simply to divide the available time by the number of tasks. Let us suppose a researcher was studying the following problem: Is a woman’s decision to have an annual mammogram related to her perceived susceptibility to breast cancer? Using the organization of steps outlined earlier, here are some of the tasks that might be undertaken:* 1. The researcher, who lost her mother to breast cancer, is concerned that many older women do not get a mammogram regularly. Her specific research question is whether mammogram practices are different for women who have different views about their susceptibility to breast cancer. 2. The researcher reviews the research literature on mammograms, factors affecting mammography decisions, and interventions designed to promote it. 3. The researcher does clinical fieldwork by discussing the problem with nurses and other health care professionals in various clinical settings (health clinics, private obstetrics and gynecology practices) and by informally discussing the problem with women in a support group for breast cancer victims. 4. The researcher examines frameworks for conceptualizing the problem. She finds that the

*This

is, of course, only a partial list of tasks and is designed to illustrate the flow of activities; the flow in this example is more orderly than would ordinarily be true.

54

PART 1 Foundations of Nursing Research

Health Belief Model (see Chapter 6) is relevant, and this helps her to develop a conceptual definition of susceptibility to breast cancer. 5. Based on what the researcher has learned, the following hypothesis is developed: Women who perceive themselves as not susceptible to breast cancer are less likely than other women to get an annual mammogram. 6. The researcher adopts a nonexperimental research design that involves collecting data from subjects at a single point in time. She designs the study to control the extraneous variables of age, marital status, and general health status. 7. There is no intervention in this study (the design is nonexperimental) and so this step does not need to be undertaken. 8. The researcher designates that the population of interest is women between the ages of 50 and 65 years living in Canada who have not been previously diagnosed as having any form of cancer. 9. The researcher decides to recruit for the research sample 200 women living in Toronto; they are identified at random using a telephone procedure known as random-digit dialing. 10. The research variables will be measured through self-report; that is, the independent variable (perceived susceptibility), dependent variable (mammogram history), and extraneous variables will be measured by asking the subjects a series of questions. The researcher decides to use existing measures of key variables, rather than developing new ones. 11. A human subjects committee at the researcher’s institution is asked to review the research plans to determine whether the study adheres to ethical standards. 12. Plans for the study are finalized: the methods are reviewed and refined by colleagues with clinical and methodologic expertise; the data collection instruments are pretested; and interviewers who will collect the data are trained.

13. Data are collected by conducting telephone interviews with the research sample. 14. Data are prepared for analysis by coding them and entering them onto a computer file. 15. Data are analyzed using a statistical software package. 16. The results indicate that the hypothesis is supported; however, the researcher’s interpretation must take into consideration that many women who were asked to participate in the study declined to do so. Moreover, the analysis revealed that mammogram use in the sample was substantially higher than had been reported in earlier studies. 17. The researcher presents an early report on her findings and interpretations at a conference of Sigma Theta Tau International. She subsequently publishes the report in the Western Journal of Nursing Research. 18. The researcher seeks out clinicians to discuss how the study findings can be utilized in practice. The researcher in this study wants to conduct this study over a 2-year period. Figure 3-1 presents a hypothetical schedule for the research tasks to be completed. (The selection of the problem is not included because the research topic has already been identified.) Note that many steps overlap or are undertaken concurrently. Some steps are projected to involve little time, whereas others require months of work. In developing a time schedule of this sort, a number of considerations should be kept in mind, including researchers’ level of knowledge and methodologic competence. Resources available to researchers, in terms of research funds and personnel, greatly influence time estimates. In the present example, the researcher almost certainly would have required funding from a sponsor to help pay for the cost of hiring interviewers, unless she were able to depend on colleagues or students. It is also important to consider the practical aspects of performing the study, which were not all enumerated in the preceding section. Obtaining supplies, securing permissions, getting approval for using

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies 0

6

4

2

8

10

12

14

16

18

20

55

24

22

Conceptual Phase Step 2 Step 3 Step 4 Step 5 Planning Phase Step 6 Step 8* Step 9 Step 10 Step 11 Step 12 Empirical Phase Step 13 Step 14 Analytic Phase

Step 15 Step 16

Dissemination Phase

Step 17 Step 18

2

4

6

8

10

12

14

16

18

20

22

24

* Note that Step 7 was not necessary because this study did not involve an intervention.

FIGURE 3.1

Project timetable in calendar months.

forms or instruments, hiring staff, and holding meetings are all time-consuming, but necessary, activities. Individuals differ in the kinds of tasks that appeal to them. Some people enjoy the preliminary phase, which has a strong intellectual component, whereas others are more eager to collect the data, a task that is more interpersonal. Researchers should, however, allocate a reasonable amount of time to do justice to each activity. ACTIVITIES IN A Q U A L I TAT I V E S T U D Y As we have just seen, quantitative research involves a fairly linear progression of tasks— researchers plan in advance the steps to be taken to maximize study integrity and then follow those steps as faithfully as possible. In qualitative studies, by contrast, the progression is closer to a circle than to a straight line—qualitative researchers are

continually examining and interpreting data and making decisions about how to proceed based on what has already been discovered. Because qualitative researchers have a flexible approach to the collection and analysis of data, it is impossible to define the flow of activities precisely—the flow varies from one study to another, and researchers themselves do not know ahead of time exactly how the study will proceed. The following sections provide a sense of how qualitative studies are conducted by describing some major activities and indicating how and when they might be performed. Conceptualizing and Planning a Qualitative Study Identifying the Research Problem Like quantitative researchers, qualitative researchers usually begin with a broad topic area to be studied.

56

PART 1 Foundations of Nursing Research

However, qualitative researchers usually focus on an aspect of a topic that is poorly understood and about which little is known. Therefore, they do not develop hypotheses or pose highly refined research questions before going into the field. The general topic area may be narrowed and clarified on the basis of selfreflection and discussion with colleagues (or clients), but researchers may proceed with a fairly broad research question that allows the focus to be sharpened and delineated more clearly once the study is underway. (Qualitative researchers may also decide to focus on a topic that has been extensively researched quantitatively, but has had little qualitative attention.) Doing Literature Reviews There are conflicting opinions among qualitative researchers about doing a literature review at the outset of a study. At one extreme are those who believe that researchers should not consult the literature before collecting data. Their concern is that prior studies or clinical writings might influence researchers’ conceptualization of the phenomena under study. According to this view, the phenomena should be elucidated based on participants’ viewpoints rather than on any prior information. Those sharing this viewpoint often do a literature review at the end of the study rather than at the beginning. Others feel that researchers should conduct at least a preliminary up-front literature review to obtain some possible guidance (including guidance in identifying the kinds of biases that have emerged in studying the topic). Still others believe that a full up-front literature review is appropriate. In any case, qualitative researchers typically find a relatively small body of relevant previous work because of the types of question they ask.

perhaps some clinical fieldwork) to identify the most suitable and information-rich environment for the conduct of the study. For a qualitative researcher, an ideal site is one in which (1) entry is possible; (2) a rich mix of people, interactions, and situations relating to the research question is present; and (3) the researcher can adopt—and maintain—an appropriate role vis-à-vis study participants. It is critical to appraise the suitability of the site (and the settings within the site where data will be collecting) before entering the field. In some cases, researchers may have access to the site selected for the study. In others, however, researchers need to gain entrée into the site or settings within it. A site may be well suited to the needs of the research, but if researchers cannot “get in,” the study cannot proceed. Gaining entrée typically involves negotiations with gatekeepers who have the authority to permit entry into their world. Gaining entrée requires strong interpersonal skills, as well as familiarity with the customs and language of the site. In addition, certain strategies are more likely to succeed than others. For example, gatekeepers might be persuaded to be cooperative if it can be demonstrated that there will be direct benefits to them or their constituents—or if a great humanitarian purpose will be served. Researchers also need to gain the gatekeepers’ trust, and that can only occur if researchers are congenial, persuasive, forthright about research requirements (e.g., how much time the fieldwork will require), and—perhaps most important—express genuine interest in and concern for the situations of the people in the site. In qualitative research, gaining entrée is likely to be an ongoing process of establishing relationships and rapport with gatekeepers and others at the site, including prospective informants.

Selecting and Gaining Entrée Into Research Sites During the planning phase, qualitative researchers must also select a site that is consistent with the topic under study. For example, if the topic is the health beliefs of the urban poor, an inner-city neighborhood with a high percentage of low-income residents must be identified. In making such a decision, researchers may need to engage in anticipatory fieldwork (and

Research Design in Qualitative Studies As we have seen, quantitative researchers do not collect data until the research design has been finalized. In a qualitative study, by contrast, the research design is often referred to as an emergent design—a design that emerges during the course of data collection. Certain design features are guided by the qualitative research tradition within which the researcher is working, but nevertheless few

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

qualitative studies have rigidly structured designs that prohibit changes while in the field. As previously noted, qualitative designs are not concerned with the control of extraneous variables. The full context of the phenomenon is considered an important factor in understanding how it plays out in the lives of people experiencing it. Although qualitative researchers do not always know in advance exactly how the study will progress in the field, they nevertheless must have some sense of how much time is available for field work and must also arrange for and test needed equipment, such as tape recorders or videotaping equipment. Other planning activities include such tasks as hiring and training interviewers to assist in the collection of data; securing interpreters if the informants speak a different language; and hiring appropriate consultants, transcribers, and support staff.

AQ4

Addressing Ethical Issues Qualitative researchers, like quantitative researchers, must also develop plans for addressing ethical issues—and, indeed, there are special concerns in qualitative studies because of the more intimate nature of the relationship that typically develops between researchers and study participants. Chapter 7 describes some of these concerns. Conducting the Qualitative Study In qualitative studies, the tasks of sampling, data collection, data analysis, and interpretation typically take place iteratively. Qualitative researchers begin by talking with or observing a few people who have first-hand experience with the phenomenon under study. The discussions and observations are loosely structured, allowing for the expression of a full range of beliefs, feelings, and behaviors. Analysis and interpretation are ongoing, concurrent activities that guide choices about the kinds of people to sample next and the types of questions to ask or observations to make. The actual process of data analysis involves clustering together related types of narrative information into a coherent scheme. The analysis of qualitative data is an intensive, time-consuming activity.

57

As analysis and interpretation progress, researchers begin to identify themes and categories, which are used to build a rich description or theory of the phenomenon. The kinds of data obtained and the people selected as participants tend to become increasingly focused and purposeful as the conceptualization is developed and refined. Concept development and verification shape the sampling process—as a conceptualization or theory develops, the researcher seeks participants who can confirm and enrich the theoretical understandings, as well as participants who can potentially challenge them and lead to further theoretical development. Quantitative researchers decide in advance how many subjects to include in the study, but qualitative researchers’ sampling decisions are guided by the data themselves. Many qualitative researchers use the principle of data saturation, which occurs when themes and categories in the data become repetitive and redundant, such that no new information can be gleaned by further data collection. In quantitative studies, researchers seek to collect high-quality data by using measuring instruments that have been demonstrated to be accurate and valid. Qualitative researchers, by contrast, must take steps to demonstrate the trustworthiness of the data while in the field. The central feature of these efforts is to confirm that the findings accurately reflect the experiences and viewpoints of participants, rather than perceptions of the researchers. One confirmatory activity, for example, involves going back to participants and sharing preliminary interpretations with them so that they can evaluate whether the researcher’s thematic analysis is consistent with their experiences. Another strategy is to use triangulation to converge on a thorough depiction of the target phenomena. An issue that qualitative researchers sometimes need to address is the development of appropriate strategies for leaving the field. Because qualitative researchers may develop strong relationships with study participants and entire communities, they need to be sensitive to the fact that their departure from the field might seem like a form of rejection or abandonment. Graceful departures and methods of achieving closure are important.

58

PART 1 Foundations of Nursing Research

Disseminating Qualitative Findings

RESEARCH EXAMPLES

Qualitative nursing researchers also strive to share their findings with others at conferences and in journal articles. Qualitative findings, because of their depth and richness, also lend themselves more readily to book-length manuscripts than do quantitative findings. Regardless of researchers’ position about when a literature review should be conducted, they usually include a summary of prior research in their reports as a means of providing context for the study. Quantitative reports almost never present raw data—that is, data in the form they were collected, which are numeric values. Qualitative reports, by contrast, are usually filled with rich verbatim passages directly from participants. The excerpts are used in an evidentiary fashion to support or illustrate researchers’ interpretations and theoretical formulations.

In this section, we illustrate the progression of activities and discuss the time schedule of two studies (one quantitative and the other qualitative) conducted by the second author of this book.

Example of raw data in a qualitative report: Scannell-Desch (2000) studied the hardships and personal strategies of 24 female Vietnam war nurses. One of the emotional hardships they experienced had to do with the youth of the patients and the severity of their injuries. The researcher supported this with the following quote from an army nurse:

Project Schedule for a Quantitative Study Beck and Gable (2001) undertook a study to evaluate the accuracy of the newly developed Postpartum Depression Screening Scale (PDSS) in screening new mothers for this mood disorder. Phase 1. Conceptual Phase: 1 Month This phase was the shortest, in large part because much of the conceptual work had been done in Beck and Gable’s (2000) first study, in which they actually developed the screening scale. The literature had already been reviewed, so all that was needed was to update the review. The same framework and conceptual definitions that had been used in the first study were used in the new study.

I had to amputate the leg of one patient. That was the first time I ever had to do that. His leg was hanging by a tissue band. I was new here, and the doctor yelled at me to “get the damn thing off.” Doctors take legs off, nurses don’t do that. He yelled at me again and said, “You do it.” (pp. 533–534).

Phase 2. Design and Planning Phase: 6 Months The second phase was time-consuming. It included not only fine-tuning the research design, but gaining entrée into the hospital where subjects were recruited and obtaining approval of the hospital’s human subjects review committee. During this period, Beck met with statistical consultants and an instrument development consultant numerous times to finalize the study design.

Like quantitative researchers, qualitative nurse researchers want their findings used in nursing practice and subsequent research. Qualitative findings often are the basis for formulating hypotheses that are tested by quantitative researchers, and for developing measuring instruments for both research and clinical purposes. Qualitative findings can also provide a foundation for designing effective nursing interventions. Qualitative studies help to shape nurses’ perceptions of a problem or situation and their conceptualizations of potential solutions.

Phase 3. Empirical Phase: 11 Months Data collection took almost a year to complete. The design called for administering the PDSS to 150 mothers who were 6 weeks postpartum, and then scheduling a psychiatric diagnostic interview for them to determine if they were suffering from postpartum depression. Women were recruited into the study during prepared childbirth classes. Recruitment began 4 months before data collection because the researchers had to wait until 6 weeks after delivery to gather data. The nurse

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

psychotherapist, who had her own clinical practice, was able to come to the hospital (a 2-hour drive for her) only 1 day a week to conduct the diagnostic interviews; this contributed to the time required to achieve the desired sample size. Phase 4. Analytic Phase: 3 Months Statistical tests were performed to determine a cutoff score on the PDSS above which mothers would be identified as having screened positive for postpartum depression. Data analysis also was undertaken to determine the accuracy of the PDSS in predicting diagnosed postpartum depression. During this phase, Beck met with the statisticians and instrument development consultant to interpret results. Phase 5. Dissemination Phase: 18 Months The researchers prepared a research report and submitted the manuscript to the journal Nursing Research for possible publication. Within 4 months it was accepted for publication, but it was “in press” (awaiting publication) for 14 months before being published. During this period, the authors presented their findings at regional and international conferences. The researchers also had to prepare a summary report for submission to the agency that funded the research. Project Schedule for a Qualitative Study Beck (2002) conducted a grounded theory study on mothering twins during the first year after delivery. Total time from start to finish was approximately 2 years. Phase 1. Conceptual Phase: 3 Months Beck became interested in mothers of multiples as a result of her quantitative studies on postpartum depression. The findings of these studies had revealed a much higher prevalence of postpartum depression among mothers of multiples than among those of singletons. Beck had never studied multiple births before, so she needed to review that literature carefully. Gaining entrée into the research site

59

(a hospital) did not take long, however, because she had previously conducted a study there and was known to the hospital’s gatekeepers. The key gatekeeper was a nurse who was in charge of the hospital’s support group for parents of multiples—a nurse with whom Beck had developed an excellent rapport in the previous study (the nurse was one of the childbirth educators who had helped recruit mothers for the postpartum depression study). Phase 2. Design and Planning Phase: 4 Months After reviewing the literature in the conceptual phase, a grounded theory design was selected. The researcher met with the nurse who headed the support group to plan the best approach for recruiting mothers of twins into the study. Plans were also made for the researcher to attend the monthly meetings of the support group. Once the design was finalized, the research proposal was submitted to and approved by both the hospital’s and university’s human subjects review committees. Phase 3. Empirical/Analytic Phases 10 months Data collection and data analysis phases occurred simultaneously in this grounded theory study. Beck attended the “parents of multiples” support group for 10 months. During that period, she conducted in-depth interviews with 16 mothers of twins in their homes, and analyzed her rich and extensive data. Beck’s analysis indicated that “life on hold” was the basic problem mothers of twins experienced during the first year of their twins’ lives. As mothers attempted to resume their own lives, they progressed through a four-stage process that Beck called “releasing the pause button.” Phase 4 Dissemination Phase: 6+ Months A manuscript was written describing this study and submitted for publication in a journal. The manuscript was published in 2002 in the journal Qualitative Health Research. In addition to disseminating the results as a journal article, Beck presented the findings at a regional nursing research conference.

60

PART 1 Foundations of Nursing Research

S U M M A RY P O I N T S • A basic distinction in quantitative studies is between experimental and nonexperimental research. In experimental research, researchers actively intervene or introduce a treatment, whereas in nonexperimental research, researchers make observations of existing situations and characteristics without intervening. • Qualitative research often is strongly rooted in research traditions that originate in the disciplines of anthropology, sociology, and psychology. Three such traditions have had strong influence on qualitative nursing research: grounded theory, phenomenology, and ethnography. • Grounded theory seeks to describe and understand key social psychological and structural processes that occur in a social setting. • Phenomenology is concerned with the lived experiences of humans and is an approach to thinking about what the life experiences of people are like. • Ethnography provides a framework for studying the meanings, patterns, and experiences of a defined cultural group in a holistic fashion. • The steps involved in conducting a quantitative study are fairly standard; researchers usually progress in a linear fashion from asking research questions to answering them. • The main phases and steps in a quantitative study are the conceptual, planning, empirical, analytic, and dissemination phases. • The conceptual phase involves (1) defining the problem to be studied; (2) doing a literature review; (3) engaging in clinical fieldwork for clinical studies; (4) developing a framework and conceptual definitions; and (5) formulating hypotheses to be tested. • The planning phase entails (6) selecting a research design; (7) developing intervention protocols if the study is experimental; (8) specifying the population; (9) developing a sampling plan; (10) specifying methods to measure the research variables, through such approaches as self-report, observation, or the use of biophysiologic methods; (11) undertaking

steps to safeguard the rights of subjects; and (12) finalizing the research plan, by conferring with colleagues, pretesting instruments, and, in some cases, conducting a pilot study. • The empirical phase involves (13) collecting data; and (14) preparing data for analysis. • The analytic phase involves (15) analyzing data through statistical analysis; and (16) interpreting the results. • The dissemination phase entails (17) communicating the findings through the preparation of research reports that can be presented orally or published in written form, most often as journal articles; and (18) efforts to promote the use of the study evidence in nursing practice. • The conduct of quantitative studies requires careful planning and organization. The preparation of a timetable with expected deadlines for task completion is recommended. • The flow of activities in a qualitative study is more flexible and less linear. • Qualitative researchers begin with a broad question regarding the phenomenon of interest, often focusing on a little-studied aspect. • In the early phase of a qualitative study, researchers select a site and seek to gain entrée into it and into the specific settings in which data collection will occur. Gaining entrée typically involves enlisting the cooperation of gatekeepers within the site. • The research design of qualitative studies is typically an emergent design. Once in the field, researchers select informants, collect data, and then analyze and interpret them in an iterative fashion; field experiences help in an ongoing fashion to shape the design of the study. • Early analysis leads to refinements in sampling and data collection, until saturation (redundancy of information) is achieved. • Qualitative researchers conclude by disseminating findings that can subsequently be used to (1) shape the direction of further studies (including more highly controlled quantitative studies); (2) guide the development of structured measuring tools for clinical and research purposes; and (3) shape nurses’ perceptions of a problem or situa-

CHAPTER 3 Overview of the Research Process in Qualitative and Quantitative Studies

tion and their conceptualizations of potential solutions. STUDY ACTIVITIES Chapter 3 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing concepts presented in this chapter. In addition, the following study questions can be addressed: 1. In quantitative studies, the same measurements are made of all subjects. What do you think researchers are trying to achieve by this degree of structure? Why might such structure not be appropriate in qualitative studies? 2. Which type of research do you think is easier to conduct—qualitative or quantitative research? Defend your response. 3. Suppose you were interested in studying fatigue in patients on chemotherapy. (This could involve either a quantitative or a qualitative approach.) Suggest some possible clinical fieldwork activities that would help you conceptualize the problem and develop a research strategy. SUGGESTED READINGS Methodologic References Creswell, J. W. (1998). Qualitative inquiry and research design: Choosing among five traditions. Thousand Oaks, CA: Sage. Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory: Strategies for qualitative research. Chicago: Aldine.

61

Kerlinger, F. N., & Lee, H. B. (2000). Foundations of behavioral research. (4th ed.). Orlando, FL: Harcourt College Publishers. Sterling, Y. M. (2001). The clinical imperative in clinical nursing research. Applied Nursing Research, 14, 44–47.

Studies Cited in Chapter 3 Beck, C. T. (2002). Releasing the pause button: Mothering twins during the first year of life. Qualitative Health Research, 12, 593–608. Beck, C. T., & Gable, R. K. (2000). Postpartum Depression Screening Scale: Development and psychometric testing, Nursing Research, 49, 272–282. Beck, C. T., & Gable, R. K. (2001). Further validation of the Postpartum Depression Screening Scale. Nursing Research, 50, 155–164. Hauck, Y. L., & Irurita, V. F. (2002). Constructing compatibility: Managing breast-feeding and weaning from the mother’s perspective. Qualitative Health Research, 12, 897–914. Johnson, V. Y. (2001). Effects of a submaximal exercise protocol to recondition the pelvic floor musculature. Nursing Research, 50, 33–41. Powers, B. A. (2001). Ethnographic analysis of everyday ethics in the care of nursing home residents with dementia. Nursing Research, 50, 332–339. Scannell-Desch, E. A. (2000). Hardships and personal strategies of Vietnam war nurses. Western Journal of Nursing Research, 22, 526–550. Sundin, K., Norberg, A., & Jansson, L. (2001). The meaning of skilled care providers’ relationship with stroke and aphasia patients. Qualitative Health Research, 11, 308–321. Wong, F., Ho, M., Chiu, I., Lui, W., Chan, C., & Lee, K. (2002). Factors contributing to hospital readmission in a Hong Kong regional hospital. Nursing Research, 51, 40–49.

PA R T

2

Conceptualizing a Research Study

4

Research Problems, Research Questions, and Hypotheses

OVERVIEW OF RESEARCH PROBLEMS Studies begin as problems that researchers want to solve or as questions they want to answer. This chapter discusses the formulation and development of research problems. We begin by clarifying some relevant terms. Basic Terminology At the most general level, a researcher selects a topic or a phenomenon on which to focus. Examples of research topics are adolescent smoking, patient compliance, coping with disability, and pain management. Within each of these broad topics are many potential research problems. In this section, we illustrate various terms using the topic side effects of chemotherapy. A research problem is an enigmatic, perplexing, or troubling condition. Both qualitative and quantitative researchers identify a research problem within a broad topic area of interest. The purpose of research is to “solve” the problem—or to contribute to its solution—by accumulating relevant information. A problem statement articulates the problem to be addressed and indicates the need for a study. Table 4-1 presents a problem statement related to the topic of side effects of chemotherapy.

Research questions are the specific queries researchers want to answer in addressing the research problem. Research questions guide the types of data to be collected in a study. Researchers who make specific predictions regarding answers to the research question pose hypotheses that are tested empirically. Many reports include a statement of purpose (or purpose statement), which is the researcher’s summary of the overall goal of a study. A researcher might also identify several research aims or objectives—the specific accomplishments the researcher hopes to achieve by conducting the study. The objectives include obtaining answers to research questions or testing research hypotheses but may also encompass some broader aims (e.g., developing recommendations for changes to nursing practice based on the study results). These terms are not always consistently defined in research methods textbooks, and differences between the terms are often subtle. Table 4-1 illustrates the interrelationships among terms as we define them. Research Problems and Paradigms Some research problems are better suited for studies using qualitative versus quantitative methods. Quantitative studies usually involve concepts that

66

PART 2 Conceptualizing a Research Study

TABLE 4.1 Example of Terms Relating to Research Problems TERM

EXAMPLE

Topic/focus

Side effects of chemotherapy

Research problem

Nausea and vomiting are common side effects among patients on chemotherapy, and interventions to date have been only moderately successful in reducing these effects. New interventions that can reduce or prevent these side effects need to be identified.

Statement of purpose

The purpose of the study is to test an intervention to reduce chemotherapy-induced side effects—specifically, to compare the effectiveness of patient-controlled and nurse-administered antiemetic therapy for controlling nausea and vomiting in patients on chemotherapy.

Research question

What is the relative effectiveness of patient-controlled antiemetic therapy versus nurse-controlled antiemetic therapy with regard to (a) medication consumption and (b) control of nausea and vomiting in patients on chemotherapy?

Hypotheses

(1) Subjects receiving antiemetic therapy by a patient-controlled pump will report less nausea than subjects receiving the therapy by nurse administration; (2) subjects receiving antiemetic therapy by a patient-controlled pump will vomit less than subjects receiving the therapy by nurse administration; (3) subjects receiving antiemetic therapy by a patient-controlled pump will consume less medication than subjects receiving the therapy by nurse administration.

Aims/objectives

This study has as its aim the following objectives: (1) to develop and implement two alternative procedures for administering antiemetic therapy for patients receiving moderate emetogenic chemotherapy (patient controlled versus nurse controlled); (2) to test three hypotheses concerning the relative effectiveness of the alternative procedures on medication consumption and control of side effects; and (3) to use the findings to develop recommendations for possible changes to therapeutic procedures.

are fairly well developed, about which there is an existing body of literature, and for which reliable methods of measurement have been developed. For example, a quantitative study might be undertaken to determine if postpartum depression is higher among women who are employed 6 months after delivery than among those who stay home with their babies. There are relatively accurate measures of postpartum depression that would yield quantitative information about the level of depression in a sample of employed and nonemployed postpartum women.

Qualitative studies are often undertaken because some aspect of a phenomenon is poorly understood, and the researcher wants to develop a rich, comprehensive, and context-bound understanding of it. Qualitative studies are usually initiated to heighten awareness and create a dialogue about a phenomenon. In the example of postpartum depression, qualitative methods would not be well suited to comparing levels of depression among the two groups of women, but they would be ideal for exploring, for example, the meaning of postpartum

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

depression among new mothers. Thus, the nature of the research question is closely allied to paradigms and research traditions within paradigms. SOURCES OF RESEARCH PROBLEMS Students are sometimes puzzled about the origins of research problems. Where do ideas for research problems come from? How do researchers select topic areas and develop research questions? At the most basic level, research topics originate with researchers’ interests. Because research is a timeconsuming enterprise, curiosity about and interest in a topic are essential to a project’s success. Explicit sources that might fuel researchers’ curiosity include experience, the nursing literature, social issues, theories, and ideas from others. Experience and Clinical Fieldwork The nurse’s everyday clinical experience is a rich source of ideas for research problems. As you are performing your nursing functions, you are bound to find a wealth of research ideas if you are curious about why things are the way they are or about how things could be improved if something were to change. You may be well along the way to developing a research idea if you have ever asked the following kinds of questions: Why are things done this way? What information would help to solve this problem? What is the process by which this situation arose? What would happen if ... ? For beginning researchers in particular, clinical experience (or clinical coursework) is often the most compelling source for topics. Immediate problems that need a solution or that excite the curiosity are relevant and interesting and, thus, may generate more enthusiasm than abstract and distant problems inferred from a theory. Clinical fieldwork before a study may also help to identify clinical problems. TIP: Personal experiences in clinical settings are a provocative source of research ideas. Here are some hints on how to proceed: • Watch for recurring problems and see if you can discern a pattern in situations that lead to the problem.

67

Example: Why do many patients complain of being tired after being transferred from a coronary care unit to a progressive care unit? • Think about aspects of your work that are irksome, frustrating, or do not result in the intended outcome—then try to identify factors contributing to the problem that could be changed. Example: Why is suppertime so frustrating in a nursing home? • Critically examine some decisions you make in performing your functions. Are these decisions based on tradition, or are they based on systematic evidence that supports their efficacy? Many practices in nursing that have become custom might be challenged. Example: What would happen if visiting hours in the intensive care unit were changed from 10 minutes every hour to the regularly scheduled hours existing in the rest of the hospital? Nursing Literature Ideas for research projects often come from reading the nursing literature. Beginning nurse researchers can profit from regularly reading nursing journals, either clinical specialty journals or research journals such as Nursing Research or the Western Journal of Nursing Research. Nonresearch articles can be helpful in alerting researchers to clinical trends and issues of importance in clinical settings. Published research reports may suggest problem areas indirectly by stimulating the imagination and directly by specifying further areas in need of investigation. Example of a direct suggestion for further research: Stranahan (2001) studied the relationship between nurse practitioners’ attitudes about spiritual care and their spiritual care practices. She made several recommendations for further research in her report, such as the following: “Studies should be conducted to determine reasons why nurse practitioners do not practice spiritual care in the primary care setting” (p. 87). Inconsistencies in the findings reported in nursing literature sometimes generate ideas for

68

PART 2 Conceptualizing a Research Study

studies. For example, there are inconsistencies regarding which type of tactile stimulation or touch (e.g., gentle touch, stroking, rubbing) has the most beneficial physiologic and behavioral effects on preterm infants. Such discrepancies can lead to the design of a study to resolve the matter. Researchers may also wonder whether a study similar to one reported in a journal article would yield comparable results if applied in a different setting or with a different population. Replications are needed to establish the validity and generalizability of previous findings. In summary, a familiarity with existing research, or with problematic and controversial nursing issues that have yet to be understood and investigated systematically, is an important route to developing a research topic. Students who are actively seeking a problem to study will find it useful to read widely in areas of interest. In Chapter 5, we deal more extensively with the conduct of research literature reviews. TIP: In a pinch, do not hesitate to replicate a study that is reported in the research literature. Replications are a valuable learning experience and can make important contributions if they corroborate (or even if they challenge) earlier findings. Social Issues Sometimes, topics are suggested by more global contemporary social or political issues of relevance to the health care community. For example, the feminist movement has raised questions about such topics as sexual harassment, domestic violence, and gender equity in health care and in research. The civil rights movement has led to research on minority health problems, access to health care, and culturally sensitive interventions. Thus, an idea for a study may stem from a familiarity with social concerns or controversial social problems. Theory The fourth major source of research problems lies in the theories and conceptual schemes that have

been developed in nursing and related disciplines. To be useful in nursing practice, theories must be tested through research for their applicability to hospital units, clinics, classrooms, and other nursing environments. When researchers decide to base a study on an existing theory, deductions from the theory must be developed. Essentially, researchers must ask the following questions: If this theory is correct, what kind of behavior would I expect to find in certain situations or under certain conditions? What kind of evidence would support this theory? This process, which is described more fully in Chapter 6, would eventually result in a specific problem that could be subjected to systematic investigation. Ideas From External Sources External sources can sometimes provide the impetus for a research idea. In some cases, a research topic may be given as a direct suggestion. For example, a faculty member may give students a list of topics from which to choose or may actually assign a specific topic to be studied. Organizations that sponsor funded research, such as government agencies, often identify topics on which research proposals are encouraged. Ideas for research are also being noted on various websites on the internet (see, for example, Duffy, 2001). Research ideas sometimes represent a response to priorities that are established within the nursing profession, examples of which were discussed in Chapter 1. Priorities for nursing research have been established by many nursing specialty practices. Priority lists can often serve as a useful starting point for exploring research topics. Often, ideas for studies emerge as a result of a brainstorming session. By discussing possible research topics with peers, advisers or mentors, or researchers with advanced skills, ideas often become clarified and sharpened or enriched and more fully developed. Professional conferences often provide an excellent opportunity for such discussions.

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

DEVELOPMENT AND REFINEMENT OF RESEARCH PROBLEMS Unless a research problem is developed on the basis of theory or an explicit suggestion from an external source, the actual procedures for developing a research topic are difficult to describe. The process is rarely a smooth and orderly one; there are likely to be false starts, inspirations, and setbacks in the process of developing a research problem statement. The few suggestions offered here are not intended to imply that there are techniques for making this first step easy but rather to encourage beginning researchers to persevere in the absence of instant success. Selecting a Topic The development of a research problem is a creative process that depends on imagination and ingenuity. In the early stages, when research ideas are being generated, it is wise not to be critical of them immediately. It is better to begin by relaxing and jotting down general areas of interest as they come to mind. At this point, it matters little if the terms used to remind you of your ideas are abstract or concrete, broad or specific, technical, or colloquial—the important point is to put some ideas on paper. Examples of some broad topics that may come to mind include nurse—patient communication, pain in patients with cancer, and postoperative loss of orientation. After this first step, the ideas can be sorted in terms of interest, knowledge about the topics, and the perceived feasibility of turning the topics into a research project. When the most fruitful idea has been selected, the rest of the list should not be discarded; it may be necessary to return to it. Narrowing the Topic Once researchers have identified a topic of interest, they need to ask questions that lead to a researchable problem. Examples of question stems that may help to focus an inquiry include the following:

69

• What is going on with ...? • What is the process by which ...? • What is the meaning of ...? • Why do ...? • When do ...? • How do ...? • What can be done to solve ...? • What is the extent of ...? • How intense are ...? • What influences ...? • What causes ...? • What characteristics are associated with ...? • What differences exist between ...? • What are the consequences of ...? • What is the relationship between ...? • What factors contribute to ...? • What conditions prevail before ...? • How effective is ...? Here again, early criticism of ideas is often counterproductive. Try not to jump to the conclusion that an idea sounds trivial or uninspired without giving it more careful consideration or without exploring it with advisers or colleagues. Beginning researchers often develop problems that are too broad in scope or too complex and unwieldy for their level of methodologic expertise. The transformation of the general topic into a workable problem is typically accomplished in a number of uneven steps, involving a series of successive approximations. Each step should result in progress toward the goals of narrowing the scope of the problem and sharpening and defining the concepts. As researchers move from general topics to more specific researchable problems, more than one potential problem area can emerge. Let us consider the following example. Suppose you were working on a medical unit and were puzzled by that fact that some patients always complained about having to wait for pain medication when certain nurses were assigned to them and, yet, these same patients offered no complaints with other nurses. The general problem area is discrepancy in complaints from patients regarding pain medications administered by different nurses. You might ask the following: What

70

PART 2 Conceptualizing a Research Study

accounts for this discrepancy? How can I improve the situation? Such questions are not actual research questions; they are too broad and vague. They may, however, lead you to ask other questions, such as the following: How do the two groups of nurses differ? What characteristics are unique to each group of nurses? What characteristics do the group of complaining patients share? At this point, you may observe that the ethnic background of the patients and nurses appears to be a relevant factor. This may direct you to a review of the literature for studies concerning ethnicity in relation to nursing care, or it may provoke you to discuss the observations with others. The result of these efforts may be several researchable questions, such as the following: • What is the essence of patient complaints among patients of different ethnic backgrounds? • What is the patient’s experience of waiting for pain medication? • How do complaints by patients of different ethnic backgrounds get expressed by patients and perceived by nurses? • Is the ethnic background of nurses related to the frequency with which they dispense pain medication? • Is the ethnic background of patients related to the frequency and intensity of complaints when waiting for pain medication? • Does the number of patient complaints increase when patients are of dissimilar ethnic backgrounds as opposed to when they are of the same ethnic background as nurses? • Do nurses’ dispensing behaviors change as a function of the similarity between their own ethnic background and that of patients? All these questions stem from the same general problem, yet each would be studied differently—for example, some suggest a qualitative approach and others suggest a quantitative one. A quantitative researcher might become curious about nurses’ dispensing behaviors, based on some interesting evidence in the literature regarding ethnic differences. Both ethnicity and nurses’ dispensing behaviors are variables that can be measured in a straightforward and reliable manner. A qualitative researcher who

noticed differences in patient complaints would likely be more interested in understanding the essence of the complaints, the patients’ experience of frustration, the process by which the problem got resolved, or the full nature of the nurse—patient interactions regarding the dispensing of medications. These are aspects of the research problem that would be difficult to quantify. Researchers choose the final problem to be studied based on several factors, including its inherent interest to them and its compatibility with a paradigm of preference. In addition, tentative problems usually vary in their feasibility and worth. It is at this point that a critical evaluation of ideas is appropriate. Evaluating Research Problems There are no rules for making a final selection of a research problem. Some criteria, however, should be kept in mind in the decision-making process. The four most important considerations are the significance, researchability, and feasibility of the problem, and its interest to the researcher. Significance of the Problem A crucial factor in selecting a problem to be studied is its significance to nursing—especially to nursing practice. Evidence from the study should have the potential of contributing meaningfully to nursing knowledge. Researchers should pose the following kinds of questions: Is the problem an important one? Will patients, nurses, or the broader health care community or society benefit from the evidence that will be produced? Will the results lead to practical applications? Will the results have theoretical relevance? Will the findings challenge (or lend support to) untested assumptions? Will the study help to formulate or alter nursing practices or policies? If the answer to all these questions is “no,” then the problem should be abandoned. Researchability of the Problem Not all problems are amenable to study through scientific investigation. Problems or questions of a moral or ethical nature, although provocative, are

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

incapable of being researched. Take, for example, the following: Should assisted suicide be legalized? The answer to such a question is based on a person’s values. There are no right or wrong answers, only points of view. The problem is suitable to debate, not to research. To be sure, it is possible to ask related questions that could be researched. For instance, each of the following questions could be investigated in a research project: • What are nurses’ attitudes toward assisted suicide? • Do oncology nurses hold more favorable opinions of assisted suicide than other nurses? • What moral dilemmas are perceived by nurses who might be involved in assisted suicide? • What are the attitudes of terminally ill patients toward assisted suicide? • Do terminally ill patients living with a high level of pain hold more favorable attitudes toward assisted suicide than those with less pain? • How do family members experience the loss of a loved one through assisted suicide? The findings from these hypothetical projects would have no bearing, of course, on whether assisted suicide should be legalized, but the information could be useful in developing a better understanding of the issues. In quantitative studies, researchable problems are ones involving variables that can be precisely defined and measured. For example, suppose a researcher is trying to determine what effect early discharge has on patient well-being. Well-being is too vague a concept for a study. The researcher would have to sharpen and define the concept so that it could be observed and measured. That is, the researcher would have to establish criteria against which patients’ progress toward well-being could be assessed. When a new area of inquiry is being pursued, however, it may be impossible to define the concepts of interest in precise terms. In such cases, it may be appropriate to address the problem using in-depth qualitative research. The problem may then be stated in fairly broad terms to permit full exploration of the concept of interest.

71

Feasibility of Addressing the Problem A problem that is both significant and researchable may still be inappropriate if a study designed to address it is not feasible. The issue of feasibility encompasses various considerations. Not all of the following factors are relevant for every problem, but they should be kept in mind in making a final decision. Time and Timing. Most studies have deadlines or at least goals for completion. Therefore, the problem must be one that can be adequately studied within the time allotted. This means that the scope of the problem should be sufficiently restricted that enough time will be available for the various steps and activities reviewed in Chapter 3. It is wise to be conservative in estimating time for various tasks because research activities often require more time to accomplish than anticipated. Qualitative studies may be especially time-consuming. A related consideration is the timing of the project. Some of the research steps—especially data collection—may be more readily performed at certain times of the day, week, or year than at other times. For example, if the problem focused on patients with peptic ulcers, the research might be more easily conducted in the fall and spring because of the increase in the number of patients with peptic ulcers during these seasons. When the timing requirements of the tasks do not match the time available for their performance, the feasibility of the project may be jeopardized. Availability of Study Participants. In any study involving humans, researchers need to consider whether individuals with the desired characteristics will be available and willing to cooperate. Securing people’s cooperation may in some cases be easy (e.g., getting nursing students to complete a questionnaire in a classroom), but other situations may pose more difficulties. Some people may not have the time, others may have no interest in a study that has little personal benefit, and others may not feel well enough to participate. Fortunately, people usually are willing to cooperate if research demands are minimal. Researchers may need to exert extra effort in recruiting participants—or may

72

PART 2 Conceptualizing a Research Study

have to offer a monetary incentive—if the research is time-consuming or demanding. An additional problem may be that of identifying and locating people with needed characteristics. For example, if we were interested in studying the coping strategies of people who had lost a family member through suicide, we would have to develop a plan for identifying prospective participants from this distinct and inconspicuous population. Cooperation of Others. Often, it is insufficient to obtain the cooperation of prospective study participants alone. If the sample includes children, mentally incompetent people, or senile individuals, it would be necessary to secure the permission of parents or guardians, an issue discussed in the chapter on ethics (see Chapter 7). In institutional or organizational settings (e.g., hospitals), access to clients, members, personnel, or records usually requires administrative authorization. Many health care facilities require that any project be presented to a panel of reviewers for approval. As noted in Chapter 3, a critical requirement in many qualitative studies is gaining entrée into an appropriate community, setting, or group, and developing the trust of gatekeepers. Facilities and Equipment. All studies have resource requirements, although in some cases, needs may be modest. It is prudent to consider what facilities and equipment will be needed and whether they will be available before embarking on a project to avoid disappointment and frustration. The following is a partial list of considerations: • Will assistants be needed, and are such assistants available? • If technical equipment and apparatus are needed, can they be secured, and are they functioning properly? Will audiotaping or videotaping equipment be required, and is it of sufficient sensitivity for the research conditions? Will laboratory facilities be required, and are they available? • Will space be required, and can it be obtained? • Will telephones, office equipment, or other supplies be required?

• Are duplicating or printing services available, and are they reliable? • Will transportation needs pose any difficulties? Money. Monetary requirements for research projects vary widely, ranging from $10 to $20 for small student projects to hundreds of thousands (or even millions) of dollars for large-scale, government-sponsored research. The investigator on a limited budget should think carefully about projected expenses before making the final selection of a problem. Some major categories of research-related expenditures are the following: • Literature costs—computerized literature search and retrieval service charges, Internet access charges, reproduction costs, index cards, books and journals • Personnel costs—payments to individuals hired to help with the data collection (e.g., for conducting interviews, coding, data entry, transcribing, word processing) • Study participant costs—payment to participants as an incentive for their cooperation or to offset their own expenses (e.g., transportation or baby-sitting costs) • Supplies—paper, envelopes, computer disks, postage, audiotapes, and so forth • Printing and duplication costs—expenditures for printing forms, questionnaires, participant recruitment notices, and so on • Equipment—laboratory apparatus, audio- or video-recorders, calculators, and the like • Computer-related expenses (e.g., purchasing software) • Laboratory fees for the analysis of biophysiologic data • Transportation costs Experience of the Researcher. The problem should be chosen from a field about which investigators have some prior knowledge or experience. Researchers have difficulty adequately developing a study on a topic that is totally new and unfamiliar— although clinical fieldwork before launching the study may make up for certain deficiencies. In addition to substantive knowledge, the issue of technical

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

expertise should not be overlooked. Beginning researchers with limited methodologic skills should avoid research problems that might require the development of sophisticated measuring instruments or that involve complex data analyses. Ethical Considerations. A research problem may not be feasible because the investigation of the problem would pose unfair or unethical demands on participants. The ethical responsibilities of researchers should not be taken lightly. People engaged in research activities should be thoroughly knowledgeable about the rights of human or animal subjects. An overview of major ethical considerations concerning human study participants is presented in Chapter 7 and should be reviewed when considering the feasibility of a prospective project. Interest to the Researcher Even if the tentative problem is researchable, significant, and feasible, there is one more criterion: the researcher’s own interest in the problem. Genuine interest in and curiosity about the chosen research problem are critical prerequisites to a successful study. A great deal of time and energy are expended in a study; there is little sense devoting these personal resources to a project that does not generate enthusiasm. TIP: Beginning researchers often seek out suggestions on topic areas, and such assistance may be helpful in getting started. Nevertheless, it is rarely wise to be talked into a topic toward which you are not personally inclined. If you do not find a problem attractive or stimulating during the beginning phases of a study—when opportunities for creativity and intellectual enjoyment are at their highest—then you are bound to regret your choice later. C O M M U N I C AT I N G T H E RESEARCH PROBLEM It is clear that a study cannot progress without the choice of a problem; it is less clear, but nonetheless true, that the problem and research questions should be carefully stated in writing before pro-

73

ceeding with the design of the study or with field work. Putting one’s ideas in writing is often sufficient to illuminate ambiguities and uncertainties. This section discusses the wording of problem statements, statements of purpose, and research questions, and the following major section discusses hypotheses. Problem Statements A problem statement is an expression of the dilemma or disturbing situation that needs investigation for the purposes of providing understanding and direction. A problem statement identifies the nature of the problem that is being addressed in the study and, typically, its context and significance. In general, the problem statement should be broad enough to include central concerns, but narrow enough in scope to serve as a guide to study design. Example of a problem statement from a quantitative study: Women account for an increasing percentage of adults with human immunodeficiency virus (HIV).... Most of these HIV-infected women are in their childbearing years. As a result, approximately 7,000 infants are exposed prenatally each year.... All infants exposed to HIV prenatally are at risk for developmental problems.... Little is known about the quality of parental caregiving for infants of mothers with HIV, because few studies have examined parenting in this vulnerable group.... The purpose of this report is to describe the development of infants of mothers with HIV and to determine the extent to which child characteristics, parental caregiver characteristics, family characteristics, and parenting quality influence development (Holditch-Davis, Miles, Burchinal, O’Donnell, McKinney, & Lim, 2001, pp. 5–6).

In this example, the general topic could be described as infant development among at-risk children. The investigators’ more specific focus is on four factors that influence infant development among children exposed to HIV prenatally. The problem statement asserts the nature of the problem (these children are at risk of developmental problems) and indicates its breadth (7000 infants annually). It also provides a justification for conducting

74

PART 2 Conceptualizing a Research Study

a new study: the dearth of existing studies on parenting in this population. The problem statement for a qualitative study similarly expresses the nature of the problem, its context, and its significance, as in the following example: Example of a problem statement from a qualitative study: Members of cultural minority groups may find themselves surrounded by people whose values, beliefs, and interpretations differ from their own during hospitalization. This is often the case for Canada’s aboriginal population, as many live in culturally distinct communities.... To promote healing among clients from minority cultural communities, it is important for nurses to understand the phenomenon of receiving care in an unfamiliar culture. This exploratory study examined how members of the Big Cove Mi’kmaq First Nation Community ... subjectively experienced being cared for in a nonaboriginal institution (Baker & Daigle, 2000, p. 8).

As in the previous example, these qualitative researchers clearly articulated the nature of the problem and the justification for a new study. Qualitative studies that are embedded in a particular research tradition usually incorporate terms and concepts in their problem statements that foreshadow their tradition of inquiry (Creswell, 1998). For example, the problem statement in a grounded theory study might refer to the need to generate a theory relating to social processes. A problem statement for a phenomenological study might note the need to know more about people’s experiences (as in the preceding example) or the meanings they attribute to those experiences. And an ethnographer might indicate the desire to describe how cultural forces affect people’s behavior. Problem statements usually appear early in a research report and are often interwoven with a review of the literature, which provides context by documenting knowledge gaps. Statements of Purpose Many researchers first articulate their research goals formally as a statement of purpose, worded in the declarative form. The statement captures—usually in

one or two clear sentences—the essence of the study. The purpose statement establishes the general direction of the inquiry. The words purpose or goal usually appear in a purpose statement (e.g., The purpose of this study was..., or, The goal of this study was...), but sometimes the words intent, aim, or objective are used instead. Unfortunately, some research reports leave the statement of purpose implicit, placing an unnecessary burden on readers to make inferences about the goals. In a quantitative study, a statement of purpose identifies the key study variables and their possible interrelationships, as well as the nature of the population of interest. Example of a statement of purpose from a quantitative study: “The purpose of this study was to determine whether viewing a video of an actual pediatric inhalation induction would reduce the level of parental anxiety” (Zuwala & Barber, 2001, p. 21). This statement identifies the population of interest (parents whose child required inhalation induction), the independent variable (viewing a video of such an induction, versus not viewing the video), and the dependent variable (parental anxiety). In qualitative studies, the statement of purpose indicates the nature of the inquiry, the key concept or phenomenon, and the group, community, or setting under study. Example of a statement of purpose from a qualitative study: Gallagher and Pierce (2002) designed their qualitative study for the following two purposes: “to gain the family caregivers’ perspective of dealing with UI [urinary incontinence] for the care recipient who lives in a home setting, and to gain care recipients’ perspective on the UI care given by family caregivers in the home setting” (p. 25). This statement indicates that the central phenomenon of interest is perspectives on caregiving and that the groups under study are UI patients in home settings and the family caregivers caring for them. Often, the statement of purpose specifically mentions the underlying research tradition, if this is relevant.

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

Example of a statement of purpose from a grounded theory study: The purpose is “to generate a grounded substantive theory of the process of forgiveness in patients with cancer” (Mickley and Cowles, 2001, p. 31). The statement of purpose communicates more than just the nature of the problem. Through researchers’ selection of verbs, a statement of purpose suggests the manner in which they sought to solve the problem, or the state of knowledge on the topic. That is, a study whose purpose is to explore or describe some phenomenon is likely to be an investigation of a little-researched topic, often involving a qualitative approach such as a phenomenology or ethnography. A statement of purpose for a qualitative study—especially a grounded theory study—may also use verbs such as understand, discover, develop, or generate. Creswell (1998) notes that the statements of purpose in qualitative studies often “encode” the tradition of inquiry not only through the researcher’s choice of verbs but also through the use of certain terms or “buzz words” associated with those traditions, as follows: • Grounded theory: Processes; social structures; social interactions • Phenomenological studies: Experience; lived experience; meaning; essence • Ethnographic studies: Culture; roles; myths; cultural behavior Quantitative researchers also suggest the nature of the inquiry through their selection of verbs. A purpose statement indicating that the study purpose is to test or determine or evaluate the effectiveness of an intervention suggests an experimental design, for example. A study whose purpose is to examine or assess the relationship between two variables is more likely to refer to a nonexperimental quantitative design. In some cases, the verb is ambiguous: a purpose statement indicating that the researcher’s intent is to compare could be referring to a comparison of alternative treatments (using an experimental approach) or a comparison of two preexisting groups (using a nonexperimental approach). In any event, verbs such as test, evaluate, and compare

75

suggest an existing knowledge base, quantifiable variables, and designs with tight scientific controls. Note that the choice of verbs in a statement of purpose should connote objectivity. A statement of purpose indicating that the intent of the study was to prove, demonstrate, or show something suggests a bias. TIP: In wording your research questions or statement of purpose, look at published research reports for models. You may find, however, that some reports fail to state unambiguously the study purpose or specific research questions. Thus, in some studies, you may have to infer the research problem from several sources, such as the title of the report. In other reports, the purpose or questions are clearly stated but may be difficult to find. Researchers most often state their purpose or questions at the end of the introductory section of the report. Research Questions Research questions are, in some cases, direct rewordings of statements of purpose, phrased interrogatively rather than declaratively, as in the following example: • The purpose of this study is to assess the relationship between the dependency level of renal transplant recipients and their rate of recovery. • What is the relationship between the dependency level of renal transplant recipients and their rate of recovery? The question form has the advantage of simplicity and directness. Questions invite an answer and help to focus attention on the kinds of data that would have to be collected to provide that answer. Some research reports thus omit a statement of purpose and state only research questions. Other researchers use a set of research questions to clarify or lend greater specificity to the purpose statement. Example of research questions clarifying a statement of purpose: Statement of Purpose: The purpose of this study was to explore the relationship between method of pain management during labor and specific labor and birth outcomes.

76

PART 2 Conceptualizing a Research Study

Research Questions: (1) Are nonepidural and epidural methods of pain relief associated with augmentation during the first stage of labor? (2) Is the length of second stage labor associated with epidural and nonepidural methods of pain relief? (3) Are newborn Apgar scores at 1 minute and 5 minutes associated with method of pain relief? (4) Does epidural anesthesia affect maternal temperature? (Walker & O’Brien, 1999) In this example, the statement of purpose provides a global message about the researchers’ goal to explore relationships among several variables. The research questions identified the two methods of pain management (the independent variable) and the specific labor and birth outcomes of interest (the dependent variables). Research Questions in Quantitative Studies In quantitative studies, research questions identify the key variables (especially the independent and dependent variables), the relationships among them, and the population under study. The variables are all measurable concepts, and the questions suggest quantification. For example, a descriptive research question might ask about the frequency or prevalence of variables, or their average values (What percentage of women breastfeed their infants? or What is the average interstitial fluid volume at 60 minutes after intravenous infiltration following treatment with cold applications?). Most quantitative studies, however, ask questions about relationships between variables. In Chapter 2, we noted that researchers ask various questions about relationships. These can be illustrated with an example of women’s emotional responses to miscarriage: 1. Existence of relationship: Is there a relationship between miscarriage and depression— that is, are there differences in depression levels of pregnant women who miscarry compared with those who do not? 2. Direction of relationship: Do women who miscarry exhibit higher (or lower) levels of depression than pregnant women who do not?

3. Strength of relationship: How strong is the risk of depression among women who miscarry? 4. Nature of relationship: Does having a miscarriage contribute to depression? Does depression contribute to a miscarriage? Or does some other factor influence both? 5. Moderated relationship: Are levels of depression among women who miscarry moderated by whether the woman has previously given birth? (i.e., Is the relationship between depression and miscarriage different for primiparas and multiparas?) 6. Mediated relationship: Does a miscarriage directly affect depression or does depression occur because the miscarriage had a negative effect on marital relations? The last two research questions involve mediator and moderator variables, which are variables of interest to the researcher (i.e., that are not extraneous) and that affect the relationship between the independent and dependent variables. A moderator variable is a variable that affects the strength or direction of an association between the independent and dependent variable. The independent variable is said to interact with the moderator: the independent variable’s relationship with the dependent variable is stronger or weaker for different values of the moderator variable (Bennett, 2000). In the preceding example, it might be that the risk of depression after a miscarriage is low among women who had previously given birth (i.e., when the moderating variable parity is greater than 0), but high among women who do not have children (i.e., when parity equals 0). When all women are considered together without taking parity into account, the relationship between experiencing a miscarriage (the independent variable) and levels of depression (the dependent variable) might appear moderate. Therefore, identifying parity as a key moderator is important in understanding when to expect a relationship between miscarriage and depression, and this understanding has clinical relevance. Research questions that involve mediator variables concern the identification of causal pathways. A mediator variable is a variable that

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

intervenes between the independent and dependent variable and helps to explain why the relationship exists. In our hypothetical example, we are asking whether depression levels among women who have experienced a miscarriage are influenced by the negative effect of the miscarriage on marital relations. In research questions involving mediators, researchers are typically more interested in the mediators than in the independent variable, because the mediators are key explanatory mechanisms. In summary, except for questions of a purely descriptive nature, research questions in quantitative research focus on unraveling relationships among variables. Example of a research question from a quantitative study: Watt-Watson, Garfinkel, Gallop, Stevens, and Streiner (2000) conducted a study about acute care nurses’ empathy and its effects on patients. Their primary research question was about the existence and direction of a relationship: Do nurses with greater empathy have patients experiencing less pain and receiving adequate analgesia than those with less empathy? Research Questions in Qualitative Studies Researchers in the various qualitative traditions vary in their conceptualization of what types of questions are important. Grounded theory researchers are likely to ask process questions, phenomenologists tend to ask meaning questions, and ethnographers generally ask descriptive questions about cultures. The terms associated with the various traditions, discussed previously in connection with purpose statements, are likely to be incorporated into the research questions. Example of a research question from a phenomenological study: What is the lived experience of caring for a family member with Alzheimer’s disease at home? (Butcher, Holkup, & Buckwalter, 2001) It is important to note, however, that not all qualitative studies are rooted in a specific research

77

tradition. Many researchers use naturalistic methods to describe or explore phenomena without focusing on cultures, meaning, or social processes. Example of a research question from a qualitative study: Wilson and Williams (2000) undertook a qualitative study that explored the potential effects of visualism (a prejudice in favor of the seen) on the perceived legitimacy of telephone work in community nursing. Among the specific research questions that guided their in-depth interviews with community nurses were the following: Is telephone consultation considered real work? Is it considered real communication? Can telephone consultation bring the community and its nursing services into close relationship? In qualitative studies, research questions sometimes evolve over the course of the study. The researcher begins with a focus that defines the general boundaries of the inquiry. However, the boundaries are not cast in stone; the boundaries “can be altered and, in the typical naturalistic inquiry, will be” (Lincoln & Guba, 1985, p. 228). The naturalist begins with a research question that provides a general starting point but does not prohibit discovery; qualitative researchers are often sufficiently flexible that the question can be modified as new information makes it relevant to do so. RESEARCH HYPOTHESES A hypothesis is a prediction about the relationship between two or more variables. A hypothesis thus translates a quantitative research question into a precise prediction of expected outcomes. In qualitative studies, researchers do not begin with a hypothesis, in part because there is usually too little known about the topic to justify a hypothesis, and in part because qualitative researchers want the inquiry to be guided by participants’ viewpoints rather than by their own. Thus, this discussion focuses on hypotheses used to guide quantitative inquiries (some of which are generated within qualitative studies).

78

PART 2 Conceptualizing a Research Study

Function of Hypotheses in Quantitative Research Research questions, as we have seen, are usually queries about relationships between variables. Hypotheses are proposed solutions or answers to these queries. For instance, the research question might ask: Does history of sexual abuse in childhood affect the development of irritable bowel syndrome in women? The researcher might predict the following: Women who were sexually abused in childhood have a higher incidence of irritable bowel syndrome than women who were not. Hypotheses sometimes follow directly from a theoretical framework. Scientists reason from theories to hypotheses and test those hypotheses in the real world. The validity of a theory is never examined directly. Rather, it is through hypothesis testing that the worth of a theory can be evaluated. Let us take as an example the theory of reinforcement. This theory maintains that behavior that is positively reinforced (rewarded) tends to be learned or repeated. The theory itself is too abstract to be put to an empirical test, but if the theory is valid, it should be possible to make predictions about certain kinds of behavior. For example, the following hypotheses have been deduced from reinforcement theory: • Elderly patients who are praised (reinforced) by nursing personnel for self-feeding require less assistance in feeding than patients who are not praised. • Pediatric patients who are given a reward (e.g., a balloon or permission to watch television) when they cooperate during nursing procedures tend to be more obliging during those procedures than nonrewarded peers. Both of these propositions can be put to a test in the real world. The theory gains support if the hypotheses are confirmed. Not all hypotheses are derived from theory. Even in the absence of a theory, well-conceived hypotheses offer direction and suggest explanations. Perhaps an example will clarify this point. Suppose we hypothesized that nurses who have received a baccalaureate education are more likely

to experience stress in their first nursing job than are nurses with a diploma-school education. We could justify our speculation based on theory (e.g., role conflict theory, cognitive dissonance theory), earlier studies, personal observations, or on the basis of some combination of these. The development of predictions in and of itself forces researchers to think logically, to exercise critical judgment, and to tie together earlier research findings.

Now let us suppose the preceding hypothesis is not confirmed by the evidence collected; that is, we find that baccalaureate and diploma nurses demonstrate comparable stress in their first job. The failure of data to support a prediction forces researchers to analyze theory or previous research critically, to carefully review the limitations of the study’s methods, and to explore alternative explanations for the findings.

The use of hypotheses in quantitative studies tends to induce critical thinking and to facilitate understanding and interpretation of the data. To illustrate further the utility of hypotheses, suppose we conducted the study guided only by the research question, Is there a relationship between nurses’ basic preparation and the degree of stress experienced on the first job? The investigator without a hypothesis is, apparently, prepared to accept any results. The problem is that it is almost always possible to explain something superficially after the fact, no matter what the findings are. Hypotheses guard against superficiality and minimize the possibility that spurious results will be misconstrued. Characteristics of Testable Hypotheses Testable research hypotheses state expected relationships between the independent variable (the presumed cause or antecedent) and the dependent variable (the presumed effect or outcome) within a population. Example of a research hypothesis: Cardiac patients receiving an intervention involving “vicarious experience” through support

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

from former patients have (1) less anxiety; (2) higher self-efficacy expectation; and (3) higher self-reported activity than other patients (Parent & Fortin, 2000). In this example, the independent variable is receipt versus nonreceipt of the intervention, and the dependent variables are anxiety, self-efficacy expectation, and activity. The hypothesis predicts better outcomes among patients who receive the intervention. Unfortunately, researchers occasionally present hypotheses that fail to make a relational statement. For example, the following prediction is not an acceptable research hypothesis: Pregnant women who receive prenatal instruction regarding postpartum experiences are not likely to experience postpartum depression.

This statement expresses no anticipated relationship; in fact, there is only one variable (postpartum depression), and a relationship by definition requires at least two variables. When a prediction does not express an anticipated relationship, it cannot be tested. In our example, how would we know whether the hypothesis was supported—what absolute standard could be used to decide whether to accept or reject the hypothesis? To illustrate the problem more concretely, suppose we asked a group of mothers who had been given instruction on postpartum experiences the following question 1 month after delivery: On the whole, how depressed have you been since you gave birth? Would you say (1) extremely depressed, (2) moderately depressed, (3) somewhat depressed, or (4) not at all depressed? Based on responses to this question, how could we compare the actual outcome with the predicted outcome? Would all the women have to say they were “not at all depressed?” Would the prediction be supported if 51% of the women said they were “not at all depressed” or “somewhat depressed?” There is no adequate way of testing the accuracy of the prediction. A test is simple, however, if we modify the prediction to the following: Pregnant women who receive prenatal instruction are less likely to experi-

79

ence postpartum depression than those with no prenatal instruction. Here, the dependent variable is the women’s depression, and the independent variable is their receipt versus nonreceipt of prenatal instruction. The relational aspect of the prediction is embodied in the phrase less than. If a hypothesis lacks a phrase such as more than, less than, greater than, different from, related to, associated with, or something similar, it is not amenable to testing in a quantitative study. To test this revised hypothesis, we could ask two groups of women with different prenatal instruction experiences to respond to the question on depression and then compare the groups’ responses. The absolute degree of depression of either group would not be at issue. Hypotheses, ideally, should be based on sound, justifiable rationales. The most defensible hypotheses follow from previous research findings or are deduced from a theory. When a relatively new area is being investigated, the researcher may have to turn to logical reasoning or personal experience to justify the predictions. There are, however, few problems for which research evidence is totally lacking. The Derivation of Hypotheses Many students ask the question, How do I go about developing hypotheses? Two basic processes— induction and deduction—constitute the intellectual machinery involved in deriving hypotheses. An inductive hypothesis is a generalization based on observed relationships. Researchers observe certain patterns, trends, or associations among phenomena and then use the observations as a basis for predictions. Related literature should be examined to learn what is already known on a topic, but an important source for inductive hypotheses is personal experiences, combined with intuition and critical analysis. For example, a nurse might notice that presurgical patients who ask a lot of questions relating to pain or who express many pain-related apprehensions have a more difficult time in learning appropriate postoperative procedures. The nurse could then formulate a hypothesis, such as the following, that could be tested through more rigorous

80

PART 2 Conceptualizing a Research Study

procedures: Patients who are stressed by fears of pain will have more difficulty in deep breathing and coughing after their surgery than patients who are not stressed. Qualitative studies are an important source of inspiration for inductive hypotheses. Example of deriving an inductive hypothesis: In Beck’s (1998) qualitative study of postpartum-onset panic disorder, one of her findings was a theme relating to self-esteem: “As a result of recurring panic attacks, negative changes in women’s lifestyles ensued—lowering their selfesteem and leaving them to bear the burden of disappointing not only themselves but also their families” (p. 134). A hypothesis that can be derived from this qualitative finding might be as follows: Women who experience postpartum onset panic disorder have lower self-esteem than women who do not experience this disorder. The other mechanism for deriving hypotheses is through deduction. Theories of how phenomena behave and interrelate cannot be tested directly. Through deductive reasoning, a researcher can develop hypotheses based on general theoretical principles. Inductive hypotheses begin with specific observations and move toward generalizations; deductive hypotheses have as a starting point theories that are applied to particular situations. The following syllogism illustrates the reasoning process involved: • All human beings have red and white blood cells. • John Doe is a human being. • Therefore, John Doe has red and white blood cells. In this simple example, the hypothesis is that John Doe does, in fact, have red and white blood cells, a deduction that could be verified. Theories thus can serve as a valuable point of departure for hypothesis development. Researchers must ask: If this theory is valid, what are the implications for a phenomenon of interest? In other words, researchers deduce that if the general theory is true, then certain outcomes or consequences can be expected. Specific predictions derived from general principles must then be subjected to testing

through the collection of empirical data. If these data are congruent with hypothesized outcomes, then the theory is strengthened. The advancement of nursing knowledge depends on both inductive and deductive hypotheses. Ideally, a cyclical process is set in motion wherein observations are made (e.g., in a qualitative study); inductive hypotheses are formulated; systematic and controlled observations are made to test the hypotheses; theoretical systems are developed on the basis of the results; deductive hypotheses are formulated from the theory; new data are gathered; theories are modified, and so forth. Researchers need to be organizers of concepts (think inductively), logicians (think deductively), and, above all, critics and skeptics of resulting formulations, constantly demanding evidence. Wording of Hypotheses A good hypothesis is worded in simple, clear, and concise language. Although it is cumbersome to include conceptual or operational definitions of terms directly in the hypothesis statement, it should be specific enough so that readers understand what the variables are and whom researchers will be studying. Simple Versus Complex Hypotheses For the purpose of this book, we define a simple hypothesis as a hypothesis that expresses an expected relationship between one independent and one dependent variable. A complex hypothesis is a prediction of a relationship between two (or more) independent variables and/or two (or more) dependent variables. Complex hypotheses sometimes are referred to as multivariate hypotheses because they involve multiple variables. We give some concrete examples of both types of hypotheses, but let us first explain the differences in abstract terms. Simple hypotheses state a relationship between a single independent variable, which we will call X, and a single dependent variable, which we will label Y. Y is the predicted effect, outcome, or consequence of X, which is the presumed cause, antecedent, or precondition. The nature of this relationship is presented in Figure 4-1A. The

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

A.

81

B.

x

x1

y

y x2

C.

D.

y1

y1 x1

x y2

FIGURE 4.1

x2 y2

Schematic representation of various hypothetical relationships.

hatched area of the circles, which represent variables X and Y, signifies the strength of the relationship between them. If there were a one-to-one correspondence between variables X and Y, the two circles would completely overlap, and the entire area would be hatched. If the variables were totally unrelated, the circles would not overlap at all. Example of a simple hypothesis: Patients receiving a warmed solution for body cavity irrigation during surgical procedures [X] will maintain a higher core body temperature [Y] than patients receiving a room temperature solution (Kelly, Doughty, Hasselbeck, & Vacchiano, 2000). Most phenomena are the result not of one variable but of a complex array of variables. A person’s

weight, for example, is affected simultaneously by such factors as the person’s height, diet, bone structure, activity level, and metabolism. If Y in Figure 4-1A was weight, and X was a person’s caloric intake, we would not be able to explain or understand individual variation in weight completely. For example, knowing that Dave Harper’s daily caloric intake averaged 2500 calories would not allow us a perfect prediction of his weight. Knowledge of other factors, such as his height, would improve the accuracy with which his weight could be predicted. Figure 4-1B presents a schematic representation of the effect of two independent variables on one dependent variable. The complex hypothesis would state the nature of the relationship between Y on the one hand and X1 and X2 on the other. To

82

PART 2 Conceptualizing a Research Study

pursue the preceding example, the hypothesis might be: Taller people (X1) and people with higher caloric intake (X2) weigh more (Y) than shorter people and those with lower caloric intake. As the figure shows, a larger proportion of the area of Y is hatched when there are two independent variables than when there is only one. This means that caloric intake and height do a better job in helping us explaining variations in weight (Y) than caloric intake alone. Complex hypotheses have the advantage of allowing researchers to capture some of the complexity of the real world. It is not always possible to design a study with complex hypotheses. Practical considerations (e.g., researchers’ technical skills and resources) may make it difficult to test complex hypotheses. An important goal of research, however, is to explain the dependent variable as thoroughly as possible, and two or more independent variables are typically more successful than one alone. Example of a complex hypothesis—multiple independent variables: Among breast cancer survivors, emotional wellbeing [Y] is influenced by the women’s self-esteem [X1], their resourcefulness [X2] and their degree of social support [X3] (Dirksen, 2000). Just as a phenomenon can result from more than one independent variable, so a single independent variable can have an effect on, or be antecedent to, more than one phenomenon. Figure 4-1C illustrates this type of relationship. A number of studies have found, for example, that cigarette smoking (the independent variable, X), can lead to both lung cancer (Y1) and coronary disorders (Y2). This type of complex hypothesis is common in studies that try to assess the impact of a nursing intervention on a variety of criterion measures of patient well-being. Example of a complex hypothesis—multiple dependent variables: The implementation of an evidence-based protocol for urinary incontinence [X] will result in decreased frequency of urinary incontinence episodes (Y1), decreased urine loss per episode [Y2], and

decreased avoidance of activities [Y3] among women in ambulatory care settings (Sampselle et al., 2000). Finally, a more complex type of hypothesis, which links two or more independent variables to two or more dependent variables, is shown in Figure 4-1D. An example might be a hypothesis that smoking and the consumption of alcohol during pregnancy might lead to lower birth weights and lower Apgar scores in infants. Hypotheses are also complex if mediator or moderator variables are included in the prediction. For example, it might be hypothesized that the effect of caloric intake (X) on weight (Y) is moderated by gender (Z)—that is, the relationship between height and weight is different for men and women. Example of a complex hypothesis with mediator: The quality of life of a family [Y] during the survivor phase after cancer diagnosis is affected by family resources [X1] and illness survival stressors such as fear of recurrence [X2], through the mediating variable, the family meaning of the illness [Z] (Mellon & Northouse, 2001). In general, hypotheses should be worded in the present tense. Researchers make predictions about relationships that exist in the population, and not just about a relationship that will be revealed in a particular sample. Hypotheses can be stated in various ways as long as the researcher specifies or implies the relationship to be tested. Here are examples: 1. Older patients are more at risk of experiencing a fall than younger patients. 2. There is a relationship between the age of a patient and the risk of falling. 3. The older the patient, the greater the risk that she or he will fall. 4. Older patients differ from younger ones with respect to their risk of falling. 5. Younger patients tend to be less at risk of a fall than older patients. 6. The risk of falling increases with the age of the patient.

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

Other variations are also possible. The important point to remember is that the hypothesis must specify the independent variable (here, patients’ age) and the dependent variables (here, risk of falling) and the anticipated relationship between them. Directional Versus Nondirectional Hypotheses Sometimes hypotheses are described as being either directional or nondirectional. A directional hypothesis is one that specifies not only the existence but the expected direction of the relationship between variables. In the six versions of the hypothesis in the preceding list, versions 1, 3, 5, and 6 are directional because there is an explicit prediction that older patients are at greater risk of falling than younger ones. A nondirectional hypothesis, by contrast, does not stipulate the direction of the relationship. Versions 2 and 4 in the example illustrate the wording of nondirectional hypotheses. These hypotheses state the prediction that a patient’s age and the risk of falling are related; they do not stipulate, however, whether the researcher thinks that older patients or younger ones are at greater risk. Hypotheses derived from theory are almost always directional because theories explain phenomena, thus providing a rationale for expecting variables to be related in certain ways. Existing studies also offer a basis for directional hypotheses. When there is no theory or related research, when the findings of related studies are contradictory, or when researchers’ own experience leads to ambivalence, nondirectional hypotheses may be appropriate. Some people argue, in fact, that nondirectional hypotheses are preferable because they connote a degree of impartiality. Directional hypotheses, it is said, imply that researchers are intellectually committed to certain outcomes, and such a commitment might lead to bias. This argument fails to recognize that researchers typically do have hunches about outcomes, whether they state those expectations explicitly or not. We prefer directional hypotheses—when there is a reasonable basis for them—because they clarify the study’s framework and demonstrate that researchers have thought critically about the

83

phenomena under study. Directional hypotheses may also permit a more sensitive statistical test through the use of a one-tailed test—a rather fine point that is discussed in Chapter 20. Research Versus Null Hypotheses Hypotheses are sometimes classified as being either research hypotheses or null hypotheses. Research hypotheses (also referred to as substantive, declarative, or scientific hypotheses) are statements of expected relationships between variables. All the hypotheses presented thus far are research hypotheses that indicate researchers’ actual expectations. The logic of statistical inference operates on principles that are somewhat confusing to many beginning students. This logic requires that hypotheses be expressed such that no relationship is expected. Null hypotheses (or statistical hypotheses) state that there is no relationship between the independent and dependent variables. The null form of the hypothesis used in our preceding examples would be a statement such as: “Patients’ age is unrelated to their risk of falling” or “Older patients are just as likely as younger patients to fall.” The null hypothesis might be compared with the assumption of innocence of an accused criminal in our system of justice: the variables are assumed to be “innocent” of any relationship until they can be shown “guilty” through appropriate statistical procedures. The null hypothesis represents the formal statement of this assumption of innocence. TIP: If you formulate hypotheses, avoid stating them in null form. When statistical tests are performed, the underlying null hypothesis is assumed without being explicitly stated. Stating hypotheses in the null form gives an amateurish impression. Hypothesis Testing Hypotheses are formally tested through statistical procedures; researchers seek to determine through statistics whether their hypotheses have a high

84

PART 2 Conceptualizing a Research Study

probability of being correct. However, hypotheses are never proved through hypothesis testing; rather, they are accepted or supported. Findings are always tentative. Certainly, if the same results are replicated in numerous investigations, then greater confidence can be placed in the conclusions. Hypotheses come to be increasingly supported with mounting evidence. Let us look more closely at why this is so. Suppose we hypothesized that height and weight are related. We predict that, on average, tall people weigh more than short people. We then obtain height and weight measurements from a sample and analyze the data. Now suppose we happened by chance to choose a sample that consisted of short, heavy people, and tall, thin people. Our results might indicate that there is no relationship between a person’s height and weight. Would we then be justified in stating that this study proved or demonstrated that height and weight in humans are unrelated? As another example, suppose we hypothesized that tall nurses are more effective than short ones. This hypothesis is used here only to illustrate a point because, in reality, we would expect no relationship between height and a nurse’s job performance. Now suppose that, by chance again, we drew a sample of nurses in which tall nurses received better job evaluations than short ones. Could we conclude definitively that height is related to a nurse’s performance? These two examples illustrate the difficulty of using observations from a sample to generalize to a population. Other issues, such as the accuracy of the measures, the effects of uncontrolled extraneous variables, and the validity of underlying assumptions prevent researchers from concluding with finality that hypotheses are proved. TIP: If a researcher uses any statistical tests (as is true in most quantitative studies), it means that there are underlying hypotheses— regardless of whether the researcher explicitly states them—because statistical tests are designed to test hypotheses. In planning a quantitative study of your own, do not be afraid to make a prediction, that is, to state a hypothesis.

RESEARCH EXAMPLES This section describes how the research problem and research questions were communicated in two nursing studies, one quantitative and one qualitative. Research Example of a Quantitative Study Van Servellen, Aguirre, Sarna, and Brecht (2002) studied emotional distress in HIV-infected men and women. The researchers noted that, despite the fact that AIDS rates have been dropping for men but increasing for women, few studies have described the health experiences of HIV-infected women or compared them with those of men. This situation was viewed as especially troubling because of certain evidence indicating that, once HIV infected, women may be at greater risk than men for illness-related morbidity and adverse outcomes. As stated by the researchers, the purpose of their study was “to describe and compare patterns of emotional distress in men and women with symptomatic HIV seeking care in community-based treatment centers” (p. 50). The researchers went on to note that understanding gender differences and similarities in relation to sociodemographic characteristics, health status, and stress-resistant resources could “provide important information in designing gender-specific programs to improve quality of life and reduce emotional distress in clients affected by HIV” (p. 50). The conceptual framework for the study was attribution theory, which offers explanations of links between life stressors and emotional distress. This framework guided the development of the four study hypotheses, which were as follows: Hypothesis 1: Sociodemographic vulnerability (less than high school education, etc.) will be associated with emotional distress in both men and women. Hypothesis 2: Poor physical and functional health status will be associated with emotional distress in both men and women. Hypothesis 3: Optimism and social support will be associated with positive mental health outcomes ... in both men and women. Hypothesis 4: Women will have higher levels of emotional distress than men (pp. 53–54).

CHAPTER 4 Research Problems, Research Questions, and Hypotheses

Data for the study were collected from 82 men and 44 women with HIV disease in Los Angeles. The results indicated that women had greater disruptions in physical and psychosocial well-being than men, consistent with the fourth hypothesis. Physical health and optimism were the primary predictors of emotional distress in both men and women, supporting hypotheses 2 and 3. However, the first hypothesis was not supported in this low-income sample: there were no significant relationships between any sociodemographic vulnerability indicators and the subjects’ level of anxiety or depression.

Research Example of a Qualitative Study Beery, Sommers, and Hall (2002) studied the experiences of women with permanent cardiac pacemakers. The researchers stated that biotechnical devices such as pacemakers are increasingly being implanted into people to manage an array of disorders, yet relatively little research has examined the emotional impact of such an experience. They further noted that women may have distinctive responses to implanted devices because of cultural messages about the masculinity of technology, but little was known about women’s unique responses to permanent cardiac pacemakers. The purpose of Beery and colleagues’ study was to explore women’s responses to pacemaker implementation, using in-depth interviews to solicit the women’s life stories. The researchers identified two specific research questions for their study: “What is the experience of women living with permanent cardiac pacemakers?” and “How do women incorporate permanent cardiac pacemakers into their lives and bodies?” (p. 8). A sample of 11 women who were patients at the cardiology service of a large hospital participated in the study. During interviews, the women were asked a series of questions regarding life events that led up to, and occurred during and after, their pacemaker’s implantation. Each woman participated in two interviews. An example of the questions asked in the initial interview is: “What has living with a pacemaker been like for you?” (p. 12). In the follow-up interviews, more specific questions were asked, such as, “How often do you think about the pacemaker?” and “When might you be reminded of it?” (p. 12). The researchers’ analysis revealed eight themes that emerged from the interview data: relinquishing care, owning the pacemaker, experiencing fears and

85

resistance, imaging their body, normalizing, positioning as caregivers, finding resilience, and sensing omnipotence.

S U M M A RY P O I N T S • A research problem is a perplexing or enigmatic situation that a researcher wants to address through disciplined inquiry. • Researchers usually identify a broad topic, narrow the scope of the problem, and then identify questions consistent with a paradigm of choice. • The most common sources of ideas for nursing research problems are experience, relevant literature, social issues, theory, and external sources. • Various criteria should be considered in assessing the value of a research problem. The problem should be clinically significant; researchable (questions of a moral or ethical nature are inappropriate); feasible; and of personal interest. • Feasibility involves the issues of time, cooperation of study participants and other people, availability of facilities and equipment, researcher experience, and ethical considerations. • Researchers communicate their aims in research reports as problem statements, statements of purpose, research questions, or hypotheses. The problem statement articulates the nature, context, and significance of a problem to be studied. • A statement of purpose summarizes the overall study goal; in both qualitative and quantitative studies, the purpose statement identifies the key concepts (variables) and the study group or population. • Purpose statements often communicate, through the use of verbs and other key terms, the underlying research tradition of qualitative studies, or whether study is experimental or nonexperimental in quantitative ones. • A research question is the specific query researchers want to answer in addressing the research problem. In quantitative studies, research questions usually are about the existence, nature, strength, and direction of relationships. • Some research questions are about moderating variables that affect the strength or direction of

86

PART 2 Conceptualizing a Research Study

a relationship between the independent and dependent variables; others are about mediating variables that intervene between the independent and dependent variable and help to explain why the relationship exists. • In quantitative studies, a hypothesis is a statement of predicted relationships between two or more variables. A testable hypothesis states the anticipated association between one or more independent and one or more dependent variables. • Simple hypotheses express a predicted relationship between one independent variable and one dependent variable, whereas complex hypotheses state an anticipated relationship between two or more independent variables and two or more dependent variables (or state predictions about mediating or moderating variables). • Directional hypotheses predict the direction of a relationship; nondirectional hypotheses predict the existence of relationships, not their direction. • Research hypotheses predict the existence of relationships; statistical or null hypotheses express the absence of a relationship. • Hypotheses are never proved or disproved in an ultimate sense—they are accepted or rejected, supported or not supported by the data. STUDY ACTIVITIES Chapter 4 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing the concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Think of a frustrating experience you have had as a nursing student or as a practicing nurse. Identify the problem area. Ask yourself a series of questions until you have one that you think is researchable. Evaluate the problem in terms of the evaluation criteria discussed in this chapter. 2. Examine the following five problem statements. Are they researchable problems as stated? Why or why not? If a problem statement is not researchable, modify it in such a way that the problem could be studied scientifically.

a. What are the factors affecting the attrition rate of nursing students? b. What is the relationship between atmospheric humidity and heart rate in humans? c. Should nurses be responsible for inserting nasogastric tubes? d. How effective are walk-in clinics? e. What is the best approach for conducting patient interviews? 3. Examine a recent issue of a nursing research journal. Find an article that does not present a formal, well-articulated statement of purpose. Write a statement of purpose (or research questions) for that study. 4. Below are four hypotheses. For each hypothesis: (1) identify the independent and dependent variables; (2) indicate whether the hypothesis is simple or complex, and directional or nondirectional; and (3) state the hypotheses in null form. a. Patients who are not told their diagnoses report more subjective feelings of stress than do patients who are told their diagnosis. b. Patients receiving intravenous therapy report greater nighttime sleep pattern disturbances than patients not receiving intravenous therapy. c. Patients with roommates call for a nurse less often than patients without roommates. d. Women who have participated in Lamaze classes request pain medication during labor less often than women who have not taken these classes. SUGGESTED READINGS Methodologic References Bennett, J. A. (2000). Mediator and moderator variables in nursing research: Conceptual and statistical differences. Research in Nursing & Health, 23, 415–420. Creswell, J. W. (1998). Qualitative inquiry and research design: Choosing among five traditions. Thousand Oaks, CA: Sage. Duffy, M. (2001). Getting qualitative research ideas and help on-line. In P. L. Munhall (Ed.), Nursing Research: A qualitative perspective (3rd ed., pp. 639–645). Sudbury, MA: Jones & Bartlett.

CHAPTER 4 Research Problems, Research Questions, and Hypotheses Kerlinger, F. N., & Lee, H. B. (2000). Foundations of behavioral research (4th ed.). Orlando, FL: Harcourt College Publishers. Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. Newbury Park, CA: Sage. Martin, P. A. (1994). The utility of the research problem statement. Applied Nursing Research, 7, 47–49.

Studies Cited in Chapter 4 Baker, C., & Daigle, M. C. (2000). Cross-cultural hospital care as experienced by Mi’kmaq clients. Western Journal of Nursing Research, 22, 8–28. Beck, C. T. (1998). Postpartum onset panic disorder. Image: Journal of Nursing Scholarship, 30, 131–135. Beery, T. A., Sommers, M. S., & Hall, J. (2002). Focused life stories of women with cardiac pacemakers. Western Journal of Nursing Research, 24, 7–27. Butcher, H. K., Holkup, P. A., & Buckwalter, K. C. (2001). The experience of caring for a family member with Alzheimer’s disease. Western Journal of Nursing Research, 23, 33–55. Dirksen, S. R. (2000). Predicting well-being among breast cancer survivors. Journal of Advanced Nursing, 32, 937–943. Gallagher, M., & Pierce, L. L. (2002). Caregivers’ and care recipients’ perceptions of dealing with urinary incontinence. Rehabilitation Nursing, 27, 25–31. Holditch-Davis, D., Miles, M. S., Burchinal, M., O’Donnell, K., McKinney, R., & Lim, W. (2001). Parental caregiving and developmental outcomes of infants of mothers with HIV. Nursing Research, 50, 5–14. Kelly, J. A., Doughty, J. K., Hasselbeck, A. N., & Vacchiano, C. (2000). The effect of arthroscopic irrigation fluid warming on body temperature. Journal of Perianesthesia Nursing, 15, 245–252. Mellon, S., & Northouse, L. L. (2001). Family survivorship and quality of life following a cancer diagnosis. Research in Nursing & Health, 24, 446–459.

87

Mickley, J. R., & Cowles, K. (2001). Ameliorating the tension: The use of forgiveness for healing. Oncology Nursing Forum, 28, 31–37. Parent, N., & Fortin, F. (2000). A randomized, controlled trial of vicarious experience through peer support for male first-time cardiac surgery patients. Heart & Lung, 29, 389–400. Sampselle, C. M., Wyman, J., Thomas, K., Newman, D. K., Gray, M., Dougherty, M., & Burns, P. A. (2000). Continence for women: A test of AWHONN’s evidence-based protocol in clinical practice. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 29, 18–26. Stranahan, S. (2001). Spiritual perception, attitudes about spiritual care, and spiritual care practices among nurse practitioners. Western Journal of Nursing Research, 23, 90–104. van Servellen, G., Aguirre, M., Sarna, L., & Brecht, M. (2002). Differential predictors of emotional distress in HIV-infected men and women. Western Journal of Nursing Research, 24, 49–72. Walker, N. C., & O’Brien, B. (1999). The relationship between method of pain management during labor and birth outcomes. Clinical Nursing Research, 8, 119–134. Watt-Watson, J., Garfinkel, P., Gallop, R., Stevens, B., & Streiner, D. (2000). The impact of nurses’ empathic responses on patients’ pain management in acute care. Nursing Research, 49, 191–200. Wilson, K., & Williams, W. A. (2000). Visualism in community nursing: Implications for telephone work with service users. Qualitative Health Research, 10, 507–520. Zuwala R., & Barber K.R. (2001). Reducing anxiety in parents before and during pediatric anesthesia induction. AANA Journal, 69, 21–25.

5

Reviewing the Literature

R

esearchers almost never conduct a study in an intellectual vacuum; their studies are usually undertaken within the context of an existing knowledge base. Researchers undertake a literature review to familiarize themselves with that knowledge base—although, as noted in Chapter 3, some qualitative researchers deliberately bypass an in-depth literature search before entering the field to avoid having their inquiries constrained or biased by prior work on the topic. This chapter discusses the functions that a literature review can play in a research project and the kinds of material covered in a literature review. Suggestions are provided on finding references, reading research reports, recording information, and organizing and drafting a written review. Because research reports are not always easy to digest, a section of this chapter offers suggestions on reading them. PURPOSES OF A L I T E R AT U R E R E V I E W Literature reviews can serve a number of important functions in the research process—as well as important functions for nurses seeking to develop an evidence-based practice. For researchers, acquaintance with relevant research literature and the state of current knowledge can help with the following:

• Identification of a research problem and development or refinement of research questions or hypotheses • Orientation to what is known and not known about an area of inquiry, to ascertain what research can best make a contribution to the existing base of evidence • Determination of any gaps or inconsistencies in a body of research • Determination of a need to replicate a prior study in a different setting or with a different study population • Identification or development of new or refined clinical interventions to test through empirical research • Identification of relevant theoretical or conceptual frameworks for a research problem • Identification of suitable designs and data collection methods for a study • For those developing research proposals for funding, identification of experts in the field who could be used as consultants • Assistance in interpreting study findings and in developing implications and recommendations A literature review helps to lay the foundation for a study, and can also inspire new research ideas. A literature review also plays a role at the end of the study, when researchers are trying to make

CHAPTER 5 Reviewing the Literature

sense of their findings. Most research reports include summaries of relevant literature in the introduction. A literature review early in the report provides readers with a background for understanding current knowledge on a topic and illuminates the significance of the new study. Written research reviews are also included in research proposals that describe what a researcher is planning to study and how the study will be conducted. Of course, research reviews are not undertaken exclusively by researchers. Both consumers and producers of nursing research need to acquire skills for reviewing research critically. Nursing students, nursing faculty, clinical nurses, nurse administrators, and nurses involved in policy-making organizations also need to review and synthesize evidence-based information. S C O P E O F A L I T E R AT U R E SEARCH You undoubtedly have some skills in locating and organizing information. However, a review of research literature differs in many respects from other kinds of term papers that students prepare. In this section, the type of information that should be sought in conducting a research review is examined, and other issues relating to the breadth and depth of the review are considered—including differences among the main qualitative research traditions. Types of Information to Seek Written materials vary considerably in their quality, their intended audience, and the kind of information they contain. Researchers performing a review of the literature ordinarily come in contact with a wide range of material and have to decide what to read or what to include in a written review. We offer some suggestions that may help in making such decisions. The appropriateness of a reference concerns both its content (i.e., its relevance to the topic of the review) and the nature of the information it contains. The most important type of information for a research review are findings from empirical investi-

89

gations. Cumulatively, research reports sum up what is known on a topic, but the information from such reports is of greatest value when the findings are integrated in a critical synthesis. For a literature review, you should rely mostly on primary source research reports, which are descriptions of studies written by the researchers who conducted them. Secondary source research documents are descriptions of studies prepared by someone other than the original researcher. Literature review summaries, then, are secondary sources. Such reviews, if they exist and are recent, are an especially good place to begin a literature search because they provide a quick summary of the literature, and the bibliography is helpful. For many clinical topics, the reviews prepared by the Cochrane Collaboration are a particularly good resource. However, secondary descriptions of studies should not be considered substitutes for primary sources for a new literature review. Secondary sources typically fail to provide much detail about studies, and they are seldom completely objective. Our own values and biases are a filter through which information passes (although we should make efforts to control such biases), but we should not accept as a second filter the biases of the person who prepared a secondary source summary of research studies. Examples of primary and secondary sources: • Secondary source, a review of the literature on patient experiences in the ICU: Stein-Parbury, J. & McKinley, S. (2000). Patients’ experiences of being in an intensive care unit: A select literature review. American Journal of Critical Care, 9, 20–27. • Primary source, an original qualitative study on patient experiences in the ICU: Hupcey, J. E. (2000). Feeling safe: The psychosocial needs of ICU patients. Journal of Nursing Scholarship, 32, 361–367. In addition to locating empirical references, you may find in your search various nonresearch references, including opinion articles, case reports, anecdotes, and clinical descriptions. Some qualitative

90

PART 2 Conceptualizing a Research Study

researchers also review relevant literary or artistic work, to gain insights about human experiences. Such materials may serve to broaden understanding of a research problem, illustrate a point, demonstrate a need for research, or describe aspects of clinical practice. Such writings may thus may play a very important role in formulating research ideas—or may even suggest ways to broaden or focus the literature search—but they usually have limited utility in written research reviews because they are subjective and do not address the central question of written reviews: What is the current state of knowledge on this research problem? Depth and Breadth of Literature Coverage Some students worry about how broad their literature search should be. Of course, there is no convenient formula for the number of references that should be tracked down, or how many pages the written review should be. The extensiveness of the review depends on a number of factors. For written reviews, a major determinant is the nature of the document being prepared. The major types of research reviews include the following: • A review included in a research report. As we discuss later in this chapter, research reports published in journals usually include brief literature reviews in their introductions. These reviews are succinct and have two major goals: to provide readers with a quick overview of the state of knowledge on the research problem being addressed; and to document the need for the new study and demonstrate how it will contribute to existing evidence. These reviews are usually only two to four double-spaced pages, and therefore only a limited number of references can be cited. This does not mean, of course, that researchers have not conducted a more thorough review, but rather that they are summarizing only what readers need to know to understand the study context. (If the report is published in a book or other format, the literature review section may be longer.)

• A review included in a research proposal. Research proposals designed to persuade funders (or advisors) about the merits of a proposed study usually include a literature review section. As with a review in a research report, a review in a proposal provides a knowledge context and confirms the need for and significance of new research. In a proposal, however, a review also demonstrates the writer’s command of the literature. The length of such reviews may be established in proposal guidelines, but they are often 5 to 10 pages long. • A review in a thesis or dissertation. Doctoral dissertations often include a thorough review covering materials directly and indirectly related to the problem area. Often, an entire chapter is devoted to a summary of the literature, and such chapters are frequently 15 to 25 pages in length. • Free-standing literature reviews. Increasingly, nurses are preparing literature reviews that critically appraise and summarize a body of research on a topic, and such reviews play a powerful role in the development of an evidence-based practice. Students are sometimes asked to prepare a written research review for a course, and nurses sometimes do literature reviews as part of utilization projects. Researchers who are experts in a field also may do integrative reviews that are published as journal articles or that contribute to major evidence-based practice projects. Such integrative reviews are discussed in more detail in Chapter 27. Free-standing literature reviews designed to appraise a body of research critically are usually at least 15 to 25 pages long. The breadth of a literature review also depends on the topic. For some topics, it may be necessary to review research findings in the non-nursing literature, such as sociology, psychology, biology, or medicine. Breadth of the review may also be affected by how extensive the research on the topic has been. If there have been 15 published studies on a specific problem, it would be difficult to draw conclusions about the current state of knowledge on that topic without reading all 15 reports. However,

CHAPTER 5 Reviewing the Literature

it is not necessarily true that the literature task is easier for little-researched topics. Literature reviews on new areas of inquiry may need to include studies of peripherally related topics to develop a meaningful context. Relevance and quality are the key criteria for including references in a written review of the literature. With respect to depth in describing studies in a written review, the most important criteria are relevance and type of review. Research that is highly related to the problem usually merits more detailed coverage. Studies that are only indirectly related can often be summarized in a sentence or two, or omitted entirely if there are page restrictions. Literature Reviews in Qualitative Research Traditions As indicated in Chapter 3, qualitative researchers have different views about reviewing the literature as part of a new study. Some of the differences reflect viewpoints associated with various qualitative research traditions. In grounded theory studies, researchers typically collect data in the field before reviewing the literature. As the data are analyzed, the grounded theory begins to take shape. Once the theory appears to be sufficiently developed, researchers then turn to the literature, seeking to relate prior findings to the theory. Glaser (1978) warns that “It’s hard enough to generate one’s own ideas without the ‘rich’ detailment provided by literature in the same field” (p. 31). Thus, grounded theory researchers defer the literature review, but then determine how previous research fits with or extends the emerging theory. Phenomenologists often undertake a search for relevant materials at the outset of a study. In reviewing the literature for a phenomenological study, researchers look for experiential descriptions of the phenomenon being studied (Munhall, 2001). The purpose is to expand the researcher’s understanding of the phenomenon from multiple perspectives. Van Manen (1990) suggests that, in addition to past research studies, artistic sources of experiential descriptions should be located such as poetry, novels, plays, films, and art. These artistic

91

sources can offer powerful examples and images of the experience under study. Even though “ethnography starts with a conscious attitude of almost complete ignorance” (Spradley, 1979, p. 4), a review of the literature that led to the choice of the cultural problem to be studied is often done before data collection. Munhall (2001) points out that this literature review may be more conceptual than data-based. A second, more thorough literature review is often done during data analysis and interpretation so that findings can be compared with previous literature. L O C A T I N G R E L E VA N T L I T E R AT U R E F O R A RESEARCH REVIEW The ability to identify and locate documents on a research topic is an important skill that requires adaptability— rapid technological changes, such as the expanding use of the Internet, are making manual methods of finding information from print resources obsolete, and more sophisticated methods of searching the literature are being introduced continuously. We urge you to consult with librarians at your institution or to search the Internet for updated information. One caveat should be mentioned. You may be tempted to do a search through an Internet search engine, such as Yahoo, Google, or Alta Vista. Such a search might provide you with interesting information about interest groups, support groups, advocacy organizations, and the like. However, such Internet searches are not likely to give you comprehensive bibliographic information on the research literature on your topic—and you might become frustrated with searching through the vast number of websites now available. TIP: Locating all relevant information on a research question is a bit like being a detective. The various electronic and print literature retrieval tools are a tremendous aid, but there inevitably needs to be some digging for, and a lot of sifting and sorting of, the clues to knowledge on a topic. Be prepared for sleuthing! And don’t hesitate to ask your reference librarians for help in your detective work.

92

PART 2 Conceptualizing a Research Study

Electronic Literature Searches In most college and university libraries, students can perform their own searches of electronic databases—huge bibliographic files that can be accessed by computer. Most of the electronic databases of interest to nurses can be accessed either through an online search (i.e., by directly communicating with a host computer over telephone lines or the Internet) or by CD-ROM (compact disks that store the bibliographic information). Several competing commercial vendors offer information retrieval services for bibliographic databases. Currently, the most widely used service providers for accessing bibliographic files are the following: • Aries Knowledge Finder (www.ariessys.com) • Ebsco Information Services (www.ebsco.com) • Ovid Technologies (www.ovid.com) • PaperChase (www.paperchase.com) • SilverPlatter Information (www.silverplatter.com) All of these services provide user-friendly retrieval of bibliographic information—they offer menu-driven systems with on-screen support so that retrieval can usually proceed with minimal instruction. However, the services vary with regard to a number of factors, such as number of databases covered, cost, online help, ease of use, special features, methods of access, and mapping capabilities. Mapping is a feature that allows you to search for topics in your own words, rather than needing to enter a term that is exactly the same as a subject heading in the database. The vendor’s software translates (“maps”) the topic you enter into the most plausible subject heading. TIP: Even when there are mapping capabilities, it may prove useful in your search to learn the subject headings of the database or the key words that researchers themselves identify to classify their studies. Subject headings for databases can be located in the database’s thesaurus. A good place to find key words is in a journal article once you have found a relevant reference. Several other electronic resources should be mentioned. First, books and other holdings of libraries can almost always be scanned electronically

using online catalog systems. Moreover, through the Internet, the catalog holdings of libraries across the country can be searched. Finally, it may be useful to search through Sigma Theta Tau International’s Registry of Nursing Research on the Internet. This registry is an electronic research database with over 12,000 studies that can be searched by key words, variables, and researchers’ names. The registry provides access to studies that have not yet been published, which cuts down the publication lag time; however, caution is needed because these studies have not been subjected to peer review (i.e., critical review by other experts in the field). Electronic publishing in general is expanding at a rapid pace; librarians and faculty should be consulted for the most useful websites. TIP: It is rarely possible to identify all relevant studies exclusively through automated literature retrieval mechanisms. An excellent method of identifying additional references is to examine citations in recently published studies or published literature reviews. Key Electronic Databases for Nurse Researchers The two electronic databases that are most likely to be useful to nurse researchers are CINAHL (Cumulative Index to Nursing and Allied Health Literature) and MEDLINE® (Medical Literature On-Line). Other potentially useful bibliographic databases for nurse researchers include: • AIDSLINE (AIDS Information On-Line) • CancerLit (Cancer Literature) • CHID (Combined Health Information Database) • EMBASE (the Excerpta Medica database) • ETOH (Alcohol and Alcohol Problems Science Database) • HealthSTAR (Health Services, Technology, Administration, and Research) • PsycINFO (Psychology Information) • Rndex (Nursing and managed care database) The CINAHL Database The CINAHL database is the most important electronic database for nurses. This database covers

CHAPTER 5 Reviewing the Literature

references to virtually all English-language nursing and allied health journals, as well as to books, book chapters, nursing dissertations, and selected conference proceedings in nursing and allied health fields. References from more than 1200 journals are included in CINAHL. The CINAHL database covers materials dating from 1982 to the present and contains more than 420,000 records. In addition to providing bibliographic information for locating references (i.e., the author, title, journal, year of publication, volume, and page numbers), this database provides abstracts (brief summaries) of articles for more than 300 journals. Supplementary information, such as names of data collection instruments, is available for many records in the database. Documents of interest can typically be ordered electronically. CINAHL can be accessed online or by CDROM, either directly through CINAHL or through one of the commercial vendors cited earlier. Information about CINAHL’s own online service can be obtained through the CINAHL website (www.cinahl.com). We will use the CINAHL database to illustrate some of the features of an electronic search. Our example relied on the Ovid Search Software for CD-ROM, but similar features are available through other vendors’ software. Most searches are likely to begin with a subject search (i.e., a search for references on a specific topic). For such a search, you would type in a word or phrase that captures the essence of the topic (or the subject heading, if you know it), and the computer would then proceed with the search. An important alternative to a subject search is a textword search that looks for the words you enter in text fields of each record, including the title and the abstract. If you know the name of a researcher who has worked on the topic, an author search might be productive. TIP: If you want to identify all major research reports on a topic, you need to be flexible and to think broadly about the key words that could be related to your topic. For example, if you are interested in anorexia nervosa, you might look under anorexia, eating disorders, and weight

93

loss, and perhaps under appetite, eating behavior, food habits, bulimia, and body weight changes. In a subject search, after you enter the topic the computer might give you a message through scope notes about the definition of the CINAHL subject heading so you could determine whether the mapping procedure produced the right match. For example, if you typed in the subject “baby blues,” the software would lead you to the subject heading “postpartum depression,” and the scope note would give the following definition: “Any depressive disorder associated with the postpartum period. Severity may range from a mild case of ‘baby blues’ to a psychotic state.” If this is the topic you had in mind, you would then learn the number of “hits” there are in the database for postpartum depression—that is, how many matches there are for that topic. TIP: If your topic includes independent and dependent variables, you may need to do separate searches for each. For example, if you were interested in learning about the effect of health beliefs on compliance behaviors among patients with AIDS, you might want to read about health beliefs (in general) and about compliance behaviors (in general). Moreover, you might also want to access research on patients with AIDS and their circumstances. In most cases, the number of hits in a subject search initially is rather large, and you will want to refine the search to ensure that you retrieve the most appropriate references. You can delimit retrieved documents in a number of ways. For example, you can restrict the search to those references for which your topic is the main focus of the document. For most subject headings, you also can select from a number of subheadings specific to the topic you are searching. You might also want to limit the references to a certain type of document (e.g., only research reports); specific journal subsets (e.g., only ones published in nursing journals); certain features of the document (e.g., only ones with abstracts); restricted publication dates (e.g., only those after 1999); certain languages (e.g., only those written in English); or certain study participant characteristics (e.g., only adolescents).

94

PART 2 Conceptualizing a Research Study

TIP: When searching a database through a commercial vendor, it is usually possible to combine searches to find only those references that include two or more topics. For example, if we were interested in the effect of stress on substance abuse, we could do independent searches for the two topics and then combine the searches to identify studies involving both variables. To illustrate how searches can be delimited with a concrete example, suppose we were interested in recent research on brain injury, which is the term we enter in a subject search. Here is an example of how many hits there were on successive restrictions to the search, using the CINAHL database through June 2001: Search Topic/Restriction

Brain injuries Restrict to main focus Limit to research reports Limit to nursing journals Limit to 1999 through 2001 publications

Hits

1459 1263 481 28 12

This narrowing of the search—from 1459 initial references on brain injuries to 12 references for recent nursing research reports on brain injuries—took under 1 minute to perform. Next, we would display the information for these 12 references on the monitor, and we could then print full bibliographic information for the ones that appeared especially promising. An example of one of the CINAHL record entries retrieved through this search on brain injuries is presented in Box 5-1. Each entry shows an accession number that is the unique identifier for each record in the database (the number that can be used to order the full text. Then, the authors and title of the reference are displayed, followed by source information. The source indicates the following: • Name of the journal (Critical Care Nursing Quarterly) • Volume (23) • Issue (4) • Page numbers (42–51) • Year and month of publication (2001 Feb.) • Number of cited references (23)

The printout also shows all the CINAHL subject headings that were coded for this particular entry; any of these headings could have been used in the subject search process to retrieve this particular reference. Note that the subject headings include both substantive/topical headings (e.g., brain injuries, quality of life) and methodologic headings (e.g., interviews). Next, when formal, named instruments are used in the study, these are printed under Instrumentation. Finally, the abstract for the study is presented. Based on the abstract, we would then decide whether this reference was pertinent to our inquiry. Once relevant references are identified, the full research reports can be obtained and reviewed. All the documents referenced in the database can be ordered by mail or facsimile (fax), so it is not necessary for your library to subscribe to the referenced journal. Many of the retrieval service providers (such as Ovid) offer full text online services, so that, for certain journals, documents can be browsed directly, linked to other documents, and downloaded. The MEDLINE®Database The MEDLINE® database was developed by the U.S. National Library of Medicine (NLM), and is widely recognized as the premier source for bibliographic coverage of the biomedical literature. MEDLINE® incorporates information from Index Medicus, International Nursing Index, and other sources. MEDLINE® covers more than 4300 journals and contains more than 11 million records. In 1999, abstracts of reviews from the Cochrane Collaboration became available through MEDLINE® (they are also available directly on the internet at www.hcn.net.au/cochrane). Because the MEDLINE® database is so large, it is often useful to access a subset of the database rather than the unabridged version that has references dating from 1966 to the present. For example, some subsets of the database cover only references within the previous 5 years. Other subsets include core medical journals, specialty journals, and nursing journals. The MEDLINE® database can be accessed online or by CD-ROM through a commercial vendor (e.g., Ovid, Aries Knowledge Finder) for a fee. This

BOX 5.1 Example of a Printout From a CINAHL Search ACCESSION NUMBER 2001035983. SPECIAL FIELDS CONTAINED Fields available in this record: abstract. AUTHORS DePalma JA. INSTITUTION Senior Research Associate, Oncology Nursing Society, Pittsburgh, Pennsylvania. TITLE Measuring quality of life of patients of traumatic brain injury. SOURCE Critical Care Nursing Quarterly, 23(4):42–51, 2001 Feb. (23 ref) ABBREVIATED SOURCE CRIT CARE NURS Q, 23(4):42–51, 2001 Feb. (23 ref) CINAHL SUBJECT HEADINGS Brain Injuries/pr [Prognosis] Brain Injuries/rh [Rehabilitation] *Brain Injuries/pf [Psychosocial Factors] Clinical Nursing Research Data Collection Family/pf [Psychosocial Factors] Inpatients Interviews

*Quality of Life/ev [Evaluation] Questionnaires Recovery Research Instruments Research Subjects Survivors/pf [Psychosocial Factors] Trauma Severity Indices Treatment Outcomes

INSTRUMENTATION Sickness Impact Profile (SIP). ABSTRACT Quality of life (QOL) is recognized as an important indicator of health care and the patient’s ability to cope with illness, treatment, and recuperation. Issues that need to be addressed in any proposed QOL research include a clear definition of QOL, a sound rationale for the choice of a measurement instrument, and the value of qualitative data. Measuring QOL in a patient population that has experienced traumatic brain injury (TBI) raises special concerns associated with the physical, behavioral, and cognitive limitations inherent with the specific TBI population. These pertinent issues are discussed with a focus that should be helpful for persons planning QOL projects and those reading and critiquing related literature. A study conducted by the author with patients with severe trauma injury will be used as an example of the impact of these issues on an actual project. Copyright © 2001 by Aspen Publishers, Inc. (23 ref)

96

PART 2 Conceptualizing a Research Study

database can also be accessed for free through the Internet at various websites, including the following: • PubMed (http://www.ncbi.nlm.nih.gov/PubMed) • Infotrieve (http://www4.infotrieve.com/newmedline/search.asp) The advantage of accessing the database through commercial vendors is that they offer superior search capabilities and special features. Print Resources Print-based resources that must be searched manually are rapidly being overshadowed by electronic databases, but their availability should not be ignored. It is sometimes necessary to refer to printed resources to perform a search to include early literature on a topic. For example, the CINAHL database does not include references to research reports published before 1982. Print indexes are books that are used to locate articles in journals and periodicals, books, dissertations, publications of professional organizations, and government documents. Indexes that are particularly useful to nurses are the International Nursing Index, Cumulative Index to Nursing and Allied Health Literature (the “red books”), Nursing Studies Index, Index Medicus, and Hospital Literature Index. Indexes are published periodically throughout the year (e.g., quarterly), with an annual cumulative index. When using a print index, you usually first need to identify the appropriate subject heading. Subject headings can be located in the index’s thesaurus. Once the proper subject heading is determined, you can proceed to the subject section of the index, which lists the actual references. Abstract journals summarize articles that have appeared in other journals. Abstracting services are in general more useful than indexes because they provide a summary of a study rather than just a title. Two important abstract sources for the nursing literature are Nursing Abstracts and Psychological Abstracts. TIP: If you are doing a completely manual search, it is a wise practice to begin the search with the most recent issue of the index or abstract journal and then to proceed backward.

(Most electronic databases are organized chronologically, with the most recent references appearing at the beginning of a listing.) READING RESEARCH REPORTS Once you have identified potential references, you can proceed to locate the documents. For research literature reviews, relevant information will be found mainly in research reports in professional journals, such as Nursing Research. Before discussing how to prepare a written review, we briefly present some suggestions on how to read research reports in journals. What Are Research Journal Articles? Research journal articles are reports that summarize a study or one aspect of a complex study. Because journal space is limited, the typical research article is relatively brief—usually only 15 to 25 manuscript pages, double-spaced. This means that the researcher must condense a lot of information into a short space. Research reports are accepted by journals on a competitive basis and are critically reviewed before acceptance for publication. Readers of research journal articles thus have some assurance that the studies have already been scrutinized for their scientific merit. Nevertheless, the publication of an article does not mean that the findings can be uncritically accepted as true, because most studies have some limitations that have implications for the validity of the findings. This is why consumers as well as producers of research can profit from understanding research methods. Research reports in journals tend to be organized in certain format and written in a particular style. The next two sections discuss the content and style of research journal articles.* Content of Research Reports Research reports typically consist of four major sections (introduction, method section, results section, discussion section), plus an abstract and references.

*A more detailed discussion of the structure of journal articles is presented in Chapter 24.

CHAPTER 5 Reviewing the Literature

The Abstract The abstract is a brief description of the study placed at the beginning of the journal article. The abstract answers, in about 100 to 200 words, the following questions: What were the research questions? What methods did the researcher use to address those questions? What did the researcher find? and What are the implications for nursing practice? Readers can review an abstract to assess whether the entire report is of interest. Some journals have moved from having traditional abstracts—which are single paragraphs summarizing the main features of the study—to slightly longer, more structured, and more informative

97

abstracts with specific headings. For example, abstracts in Nursing Research after 1997 present information about the study organized under the following headings: Background, Objectives, Method, Results, Conclusions, and Key Words. Box 5-2 presents abstracts from two actual studies. The first is a “new style” abstract for a quantitative study entitled “Family reports of barriers to optimal care of the dying” (Tolle, Tilden, Rosenfeld, & Hickman, 2000). The second is a more traditional abstract for a qualitative study entitled “Families of origin of homeless and never-homeless women” (Anderson & Imle, 2001). These two studies are used as illustrations throughout this section.

BOX 5.2 Examples of Abstracts From Published Research QUANTITATIVE STUDY Background: In response to intense national pressure to improve care of the dying, efforts have been made to determine problems or barriers to optimal care. However, prior research is limited by such factors as setting, focus, and sampling. Objectives: The purpose of this study was to identify barriers to optimal care of a population-based representative sample of decedents across a full range of settings in which death occurred. Method: Families were contacted 2 to 5 months after decedents’ deaths by using data on their death certificates. Over a 14-month period, telephone interviews were conducted with 475 family informants who had been involved in caring for the patient in the last month of life. Interviews were standardized by use of a 58-item structured questionnaire. Results: Data show a high frequency of advance planning (68%) and a high level of respect by clinicians for patient—family preferences about end-of-life location and treatment decisions. Family satisfaction with care was generally high, even though pain was a problem in one-third of the sample of decedents. Conclusion: Barriers to optimal care of the dying remain, despite a generally positive overall profile; barriers include level of pain and management of pain, as well as some dissatisfaction with physician availability (Tolle, Tilden, Rosenfeld, & Hickman, 2000). QUALITATIVE STUDY Naturalistic inquiry was used to compare the characteristics of families of origin of homeless women with never-homeless women. The women’s experiences in their families of origin were explored during in-depth interviews using Lofland and Lofland’s conceptions of meanings, practices, episodes, roles, and relationships to guide the analysis. The two groups were similar with respect to family abuse history, transience, and loss. The never-homeless women had support from an extended family member who provided unconditional love, protection, a sense of connection, and age-appropriate expectations, as contrasted with homeless women who described themselves as being without, disconnected, and having to be little adults in their families of origin. The experiences of family love and connection seemed to protect never-homeless women from the effects of traumatic life events in childhood. These findings provide support for the influence of a woman’s family of origin as a precursor to homelessness (Anderson & Imle, 2001).

98

PART 2 Conceptualizing a Research Study

The Introduction The introduction acquaints readers with the research problem and its context. The introduction, which may or may not be specifically labeled “Introduction,” follows immediately after the abstract. This section usually describes the following: • The central phenomena, concepts, or variables under study. The problem area under investigation is identified. • The statement of purpose, and research questions or hypotheses to be tested. The reader is told what the researcher set out to accomplish in the study. • A review of the related literature. Current knowledge relating to the study problem is briefly described so readers can understand how the study fits in with previous findings and can assess the contribution of the new study. • The theoretical framework. In theoretically driven studies, the framework is usually presented in the introduction. • The significance of and need for the study. The introduction to most research reports includes an explanation of why the study is important to nursing. Thus, the introduction sets the stage for a description of what the researcher did and what was learned. Examples from an introductory paragraph: The homeless in the United States continue to increase in numbers and in diversity.... An estimated 760,000 people experience homelessness at some time during a one-year period.... Women and families make up the fastest-growing segment of the homeless population, and women head an estimated 90% of homeless families. The purpose of this study was to compare the characteristics of families of origin of homeless women with the families of origin of neverhomeless women whose childhood experiences placed them at risk for homelessness (Anderson & Imle, 2001, p. 394).

In this paragraph, the researchers described the background of the problem, the population of primary interest (homeless women), and the study purpose. The Method Section The method section describes the method the researcher used to answer the research questions. The

method section tells readers about major methodologic decisions, and may offer rationales for those decisions. For example, a report for a qualitative study often explains why a qualitative approach was considered to be especially appropriate and fruitful. In a quantitative study, the method section usually describes the following, which may be presented as labeled subsections: • The research design. A description of the research design focuses on the overall plan for the collection of data, often including the steps the researcher took to minimize biases and enhance the interpretability of the results by instituting various controls. • The subjects. Quantitative research reports usually describe the population under study, specifying the criteria by which the researcher decided whether a person would be eligible for the study. The method section also describes the actual sample, indicating how people were selected and the number of subjects in the sample. • Measures and data collection. In the method section, researchers describe the methods and procedures used to collect the data, including how the critical research variables were operationalized; they may also present information concerning the quality of the measuring tools. • Study procedures. The method section contains a description of the procedures used to conduct the study, including a description of any intervention. The researcher’s efforts to protect the rights of human subjects may also be documented in the method section. Table 5-1 presents excerpts from the method section of the quantitative study by Tolle and her colleagues (2000). Qualitative researchers discuss many of the same issues, but with different emphases. For example, a qualitative study often provides more information about the research setting and the context of the study, and less information on sampling. Also, because formal instruments are not used to collect qualitative data, there is little discussion about data collection methods, but there may be more information on data collection procedures. Increasingly, reports of qualitative studies are including descriptions of the

CHAPTER 5 Reviewing the Literature

99

TABLE 5.1 Excerpts From Method Section, Quantitative Report METHODOLOGIC ELEMENT

EXCERPT FROM TOLLE ET AL.’S STUDY, 2000*

Research design

Telephone surveys were conducted with 475 family respondents 2 to 5 months after decedents’ deaths (p. 311).

Sample

...Death certificates for all Oregon deaths occurring in the 14 months between November 1996 and December 1997 were systematically randomly sampled, excluding decedents under the age of 18 years and deaths attributable to suicide, homicide, accident, or those undergoing medical examiner review. Out of a sampling frame of N = 24,074, the systematic random sample yielded 1,458 death certificates (p. 311).

Data collection

Family respondents were interviewed by use of a 58-item questionnaire developed by the investigators.... Telephone interviews were conducted by graduate student research assistants, intensively trained to standardize survey administration (pp. 311–312).

Procedure

In the initial telephone call, the study was explained. If the potential respondent gave their informed consent to participate, an appointment for the interview was made for the following week.... Those who agreed to participate were mailed a letter of introduction, an information sheet explaining the study, and a postcard to return if the person subsequently decided to decline participation (p. 311).

* From, Tolle, S. W., Tilden, V. P., Rosenfeld, A. G., & Hickman, S. E. (2000). Family reports of barriers to optimal care of the dying. Nursing Research, 49, 310–317.

researchers’ efforts to ensure the trustworthiness of the data. Some qualitative reports also have a subsection on data analysis. There are fairly standard ways of analyzing quantitative data, but such standardization does not exist for qualitative data, so qualitative researchers may describe their analytic approach. Table 5-2 presents excerpts from the method section of the study by Anderson and Imle (2001). The Results Section The results section presents the research findings (i.e., the results obtained in the analyses of the data). The text summarizes the findings, often accompanied by tables or figures that highlight the most noteworthy results.

Virtually all results sections contain basic descriptive information, including a description of the study participants (e.g., their average age). In quantitative studies, the researcher provides basic descriptive information for the key variables, using simple statistics. For example, in a study of the effect of prenatal drug exposure on the birth outcomes of infants, the results section might begin by describing the average birth weights and Apgar scores of the infants, or the percentage who were of low birth weight (under 2500 g). In quantitative studies, the results section also reports the following information relating to the statistical analyses performed: • The names of statistical tests used. A statistical test is a procedure for testing hypotheses and

100

PART 2 Conceptualizing a Research Study

TABLE 5.2 Excerpts From Method Section, Qualitative Report METHODOLOGIC ELEMENT

EXCERPT FROM ANDERSON & IMLE’S STUDY, 2001*

Design

Naturalistic inquiry was used to explore the families of origin of homeless and never-homeless women from their perspectives (p. 397).

Sample

The criteria for inclusion in the study were that the women had never been homeless, had experienced traumatic childhoods, were at least 18 years of age and spoke English.... The inclusion criteria for the study of homeless women were that the women had experienced homelessness, had taken steps toward moving away from life on the streets, were age 18 or over, and spoke English. The homeless (n = 12) and never-homeless (n = 16) women were similar in age, number of persons in the family of origin...education, ethnicity, and abuse histories (p. 398).

Data collection

One to three in-depth interviews, lasting 45 minutes to 2 hours, were conducted with both homeless and never-homeless women. Intensive interviewing is especially well suited to a retrospective study that relies on the participant to recall their memories because the setting described no longer exists.... All interviews were conducted by the author (p. 397).

Data analysis

Social units of analysis...were used to organize and assist in the coding and analysis of the interview data. The 5 social units that emerged during the interviews with the homeless sample were analyzed in the sample of neverhomeless women and themes were identified (p. 400).

* From Anderson, D. G. & Imle. (2001). Families of origin of homeless and never-homeless women. Western Journal of Nursing Research, 23, 394–413.

evaluating the believability of the findings. For example, if the percentage of low-birth-weight infants in the sample of drug-exposed infants is computed, how probable is it that the percentage is accurate? If the researcher finds that the average birth weight of drug-exposed infants in the sample is lower than that of nonexposed infants, how probable is it that the same would be true for other infants not in the sample? That is, is the relationship between prenatal drug exposure and infant birth weight real and likely to be replicated with a new sample of infants—or does the result reflect a peculiarity of the sample? Statistical tests answer such questions. Statistical tests are based on common principles; you do not have to

know the names of all statistical tests (there are dozens) to comprehend the findings. • The value of the calculated statistic. Computers are used to compute a numeric value for the particular statistical test used. The value allows the researchers to draw conclusions about the meaning of the results. The actual numeric value of the statistic, however, is not inherently meaningful and need not concern you. • The significance. The most important information is whether the results of the statistical tests were significant (not to be confused with important or clinically relevant). If a researcher reports that the results are statistically significant, it means the findings are probably valid

CHAPTER 5 Reviewing the Literature

and replicable with a new sample of subjects. Research reports also indicate the level of significance, which is an index of how probable it is that the findings are reliable. For example, if a report indicates that a finding was significant at the .05 level, this means that only 5 times out of 100 (5 ! 100 " .05) would the obtained result be spurious or haphazard. In other words, 95 times out of 100, similar results would be obtained with a new sample. Readers can therefore have a high degree of confidence—but not total assurance—that the findings are reliable. Example from the results section of a quantitative study: Overall, 71% of family respondents reported knowing the decedent’s preference for place of death, and 68% of those families believed that the decedent had died in their preferred location. Decedents who got their wish differed significantly by location of death, #2 (2, n " 337) " 131.2, p < .001. Of decedents whose wishes were known, 100% of those who died at home wanted to die at home, and 40% of decedents who died in nursing homes and 45% of those who died in hospitals got their wish for location of death” (Tolle et al., 2000, p. 313).AQ5

In this excerpt, the authors indicated that decedents who died at home were much more likely to have died in their preferred location than decedents who died in nursing homes or hospitals, and that the probability (p) that this finding is spurious is less than 1 in 1000 (1 ! 1000 " .001). Thus, the finding is highly reliable. Note that to comprehend this finding, you do not need to understand what the #2 statistic is, nor to concern yourself with the actual value of the statistic, 131.2. In qualitative reports, the researcher often organizes findings according to the major themes, processes, or categories that were identified in the data. The results section of qualitative reports sometimes has several subsections, the headings of which correspond to the researcher’s labels for the themes. Excerpts from the raw data are presented to support and provide a rich description of the thematic analysis. The results section of qualitative studies may also present the researcher’s emerging theory about the phenomenon under study, although this may appear in the concluding section of the report.

101

Example from the results section of a qualitative study: The homeless people interviewed did not have a sense of connectedness.... In contrast, the never-homeless women had connections to family, friends, and to larger social systems that lasted into adulthood and the foreseeable future.... Many of the never-homeless also described tangible links to their past. Robin, for example, described her dining room set that had belonged to her adopted grandparents: “I was always told that this table came with them on a covered wagon.... They paid $35 for this set; that includes the chairs and buffet.... There are places on here that have [her brother’s] teeth marks. I used to play house under here” (Anderson & Imle, 2001, p. 409).

In this excerpt, the researchers illustrate their finding that never-homeless women maintained rich connections with their past with a direct quote from a study participant. The Discussion Section In the discussion section, the researcher draws conclusions about the meaning and implications of the findings. This section tries to unravel what the results mean, why things turned out the way they did, and how the results can be used in practice. The discussion in both qualitative and quantitative reports may incorporate the following elements: • An interpretation of the results. The interpretation involves the translation of findings into practical, conceptual, or theoretical meaning. • Implications. Researchers often offer suggestions for how their findings could be used to improve nursing, and they may also make recommendations on how best to advance knowledge in the area through additional research. • Study limitations. The researcher is in the best position possible to discuss study limitations, such as sample deficiencies, design problems, weaknesses in data collection, and so forth. A discussion section that presents these limitations demonstrates to readers that the author was aware of these limitations and probably took them into account in interpreting the findings.

102

PART 2 Conceptualizing a Research Study

Example from a discussion section of a quantitative report: Overall, one third of the sample of family respondents indicated moderate to severe decedent pain in the final week of life. Although this rate is somewhat better than rates reported elsewhere, it still raises concern that control of pain for dying patients is simply not good enough.... Interestingly, families had more complaints about the management of pain for decedents who died at home, even though they did not report higher levels of pain. Perhaps this is because in the home setting, family members are more aware of pain management problems and bear more responsibility for direct care of such needs (Tolle et al., 2000, p. 315).

References Research journal articles conclude with a list of the books, reports, and journal articles that were referenced in the text of the report. For those interested in pursuing additional reading on a substantive topic, the reference list of a current research study is an excellent place to begin. The Style of Research Journal Articles Research reports tell a story. However, the style in which many research journal articles are written— especially reports of quantitative studies—makes it difficult for beginning research consumers to become interested in the story. To unaccustomed audiences, research reports may seem stuffy, pedantic, and bewildering. Four factors contribute to this impression: 1. Compactness. Journal space is limited, so authors try to compress many ideas and concepts into a short space. Interesting, personalized aspects of the investigation often cannot be reported. And, in qualitative studies, only a handful of supporting quotes can be included. 2. Jargon. The authors of both qualitative and quantitative reports use research terms that are assumed to be part of readers’ vocabulary, but that may seem esoteric. 3. Objectivity. Quantitative researchers normally avoid any impression of subjectivity and thus

research stories are told in a way that makes them sound impersonal. For example, most quantitative research reports are written in the passive voice (i.e., personal pronouns are avoided). Use of the passive voice tends to make a report less inviting and lively than the use of the active voice, and it tends to give the impression that the researcher did not play an active role in conducting the study. (Qualitative reports, by contrast, are more subjective and personal, and written in a more conversational style.) 4. Statistical information. In quantitative reports, numbers and statistical symbols may intimidate readers who do not have strong mathematic interest or training. Most nursing studies are quantitative, and thus most research reports summarize the results of statistical analyses. Indeed, nurse researchers have become increasingly sophisticated during the past decade and have begun to use more powerful and complex statistical tools. A major goal of this textbook is to assist nurses in dealing with these issues. Tips on Reading Research Reports As you progress through this textbook, you will acquire skills for evaluating various aspects of research reports critically. Some preliminary hints on digesting research reports and dealing with the issues previously described follow. • Grow accustomed to the style of research reports by reading them frequently, even though you may not yet understand all the technical points. Try to keep the underlying rationale for the style of research reports in mind as you are reading. • Read from a report that has been photocopied. Then you will be able to use a highlighter, underline portions of the article, write questions or notes in the margins, and so on. • Read journal articles slowly. It may be useful to skim the article first to get the major points and then read the article more carefully a second time. • On the second or later reading of a journal article, train yourself to become an active reader.

CHAPTER 5 Reviewing the Literature

Reading actively means that you are constantly monitoring yourself to determine whether you understand what you are reading. If you have comprehension problems, go back and reread difficult passages or make notes about your confusion so that you can ask someone for clarification. In most cases, that “someone” will be your research instructor or another faculty member, but also consider contacting the researchers themselves. The postal and e-mail addresses of the researchers are usually included in the journal article, and researchers are generally more than willing to discuss their research with others. • Keep this textbook with you as a reference while you are reading articles initially. This will enable you to look up unfamiliar terms in the glossary at the end of the book, or in the index. • Try not to get bogged down in (or scared away by) statistical information. Try to grasp the gist of the story without letting formulas and numbers frustrate you. • Until you become accustomed to the style and jargon of research journal articles, you may want to “translate” them mentally or in writing. You can do this by expanding compact paragraphs into looser constructions, by translating jargon into more familiar terms, by recasting the report into an active voice, and by summarizing the findings with words rather than numbers. As an example, Box 5-3 presents a summary of a fictitious study, written in the style typically found in research reports. Terms that can be looked up in the glossary of this book are underlined, and the notes in the margins indicate the type of information the author is communicating. Box 5-4 presents a “translation” of this summary, recasting the information into language that is more digestible. Note that it is not just the jargon specific to research methods that makes the original version complicated (e.g., “sequelae” is more obscure than “consequences”). Thus, a dictionary might also be needed when reading research reports. • Although it is certainly important to read research reports with understanding, it is also

103

important to read them critically, especially when you are preparing a written literature review. A critical reading involves an evaluation of the researcher’s major conceptual and methodologic decisions. Unfortunately, it is difficult for students to criticize these decisions before they have gained some conceptual and methodologic skills themselves. These skills will be strengthened as you progress through this book, but sometimes common sense and thoughtful analysis may suggest flaws in a study, even to beginning students. Some of the key questions to ask include the following: Does the way the researcher conceptualized the problem make sense—for example, do the hypotheses seem sensible? Did the researcher conduct a quantitative study when a qualitative one would have been more appropriate? In a quantitative study, were the research variables measured in a reasonable way, or would an alternative method have been better? Additional guidelines for critiquing various aspects of a research report are presented in Chapter 26. P R E PA R I N G A W R I T T E N L I T E R AT U R E R E V I E W A number of steps are involved in preparing a written review, as summarized in Figure 5-1. As the figure shows, after identifying potential sources, you need to locate the references and screen them for their relevancy. Screening References References that have been identified through the literature search need to be screened. One screen is totally practical—is the reference readily accessible? For example, although abstracts of dissertations may be easy to retrieve, full dissertations are not; some references may be written in a language you do not read. A second screen is the relevance of the reference, which you can usually (but not always) surmise by reading the abstract. When abstracts are not available, you will need to take a

104

PART 2 Conceptualizing a Research Study

BOX 5.3 Summary of a Fictitious Study for Translation

Purpose of the study

Research design

Research instruments

Data analysis procedure

Implications

The potentially negative sequelae of having an abortion on the psychological adjustment of adolescents have not been adequately studied. The present study sought to determine whether alternative pregnancy resolution decisions have different long-term effects on the psychological functioning of young women.

Need further study

Three groups of low-income pregnant teenagers attending an inner-city clinic were the subjects in this study: those who delivered and kept the baby; those who delivered and relinquished the baby for adoption; and those who had an abortion. There were 25 subjects in each group. The study instruments included a selfadministered questionnaire and a battery of psychological tests measuring depression, anxiety, and psychosomatic symptoms. The instruments were administered upon entry into the study (when the subjects first came to the clinic) and then 1 year after termination of the pregnancy.

Study population

The data were analyzed using analysis of variance (ANOVA). The ANOVA tests indicated that the three groups did not differ significantly in terms of depression, anxiety, or psychosomatic symptoms at the initial testing. At the post-test, however, the abortion group had significantly higher scores on the depression scale, and these girls were significantly more likely than the two delivery groups to report severe tension headaches. There were no significant differences on any of the dependent variables for the two delivery groups. The results of this study suggest that young women who elect to have an abortion may experience a number of long-term negative consequences. It would appear that appropriate efforts should be made to follow up abortion patients to determine their need for suitable treatment.

Research sample

Results

Interpretation

CHAPTER 5 Reviewing the Literature

105

BOX 5.4 Translated Version of Fictitious Research Study As researchers, we wondered whether young women who had an abortion had any emotional problems in the long run. It seemed to us that not enough research had been done to know whether any psychological harm resulted from an abortion. We decided to study this question ourselves by comparing the experiences of three types of teenagers who became pregnant—first, girls who delivered and kept their babies; second, those who delivered the babies but gave them up for adoption; and third, those who elected to have an abortion. All teenagers in the sample were poor, and all were patients at an inner-city clinic. Altogether, we studied 75 girls—25 in each of the three groups. We evaluated the teenagers’ emotional states by asking them to fill out a questionnaire and to take several psychological tests. These tests allowed us to assess things such as the girls’ degree of depression and anxiety and whether they had any complaints of a psychosomatic nature. We asked them to fill out the forms twice: once when they came into the clinic, and then again a year after the abortion or the delivery. We learned that the three groups of teenagers looked pretty much alike in terms of their emotional states when they first filled out the forms. But when we compared how the three groups looked a year later, we found that the teenagers who had abortions were more depressed and were more likely to say they had severe tension headaches than teenagers in the other two groups. The teenagers who kept their babies and those who gave their babies up for adoption looked pretty similar one year after their babies were born, at least in terms of depression, anxiety, and psychosomatic complaints. Thus, it seems that we might be right in having some concerns about the emotional effects of having an abortion. Nurses should be aware of these long-term emotional effects, and it even may be advisable to institute some type of follow-up procedure to find out if these young women need additional help.

Discard irrelevant and inappropriate references

Identify key words and concepts to be searched

Identify potential references through electronic or manual search

Retrieve promising references

Screen references for relevance and appropriateness

Read relevant reference and take notes

Identify new references through citations

FIGURE 5.1

Flow of tasks in a literature review.

Organize references

Analyze and integrate materials

Write review

106

PART 2 Conceptualizing a Research Study

guess about relevance based on the title. For critical integrated reviews (see Chapter 27), a third criterion is the study’s methodologic quality—that is, the quality of evidence the study yields. Abstracting and Recording Notes Once a document has been determined to be relevant, you should read the entire report carefully and critically, identifying material that is sufficiently important to warrant note taking and observing flaws in the study or gaps in the report. As noted earlier, it is useful to work with photocopied articles so that you can highlight or underline critical information. Even with a copied article, we recommend taking notes or writing a summary of the report’s strengths and limitations. A formal protocol is sometimes helpful for recording information in a systematic fashion. An example of such a protocol is presented in Figure 5-2. Although many of the terms on this protocol are probably not familiar to you at this point, you will learn their meaning as you progress through this book. Organizing the Review Organization of information is a critical task in preparing a written review. When the literature on a topic is extensive, we recommend preparing a summary table. The table could include columns with headings such as Author, Type of Study (Qualitative versus Quantitative), Sample, Design, Data Collection Approach, and Key Findings. Such a table provides a quick overview that allows you to make sense of a mass of information. Example of a summary table: Abercrombie (2001) reviewed research related to strategies that have been found to improve follow-up after an abnormal Papanicolaou (Pap) smear test. Her review included a table that summarized nine studies. The headings in her columns were: Author and date; sample size; objectives; design/intervention; and results. Most writers find it helpful to work from an outline when preparing a written review. If the review is lengthy and complex, it is useful to write

out the outline; a mental outline may be sufficient for shorter reviews. The important point is to work out a structure before starting to write so that the presentation has a meaningful and understandable flow. Lack of organization is a common weakness in students’ first attempts at writing a research literature review. Although the specifics of the organization differ from topic to topic, the overall goal is to structure the review in such a way that the presentation is logical, demonstrates meaningful integration, and leads to a conclusion about what is known and not known about the topic. TIP: An important principle in organizing a review is to figure out a way to cluster and compare studies. For example, you could contrast studies that have similar findings with studies that have conflicting or inconclusive findings, making sure to analyze why the discrepancies may have occurred. Or you might want to cluster studies that have operationalized key variables in similar ways. Other reviews might have as an organizing theme the nature of the setting or the sample if research findings vary according to key characteristics (e.g., comparing research with female subjects and research with male subjects, if the results differ.) Doing a research review is a little bit like doing a qualitative study—you must search for important themes. Once the main topics and their order of presentation have been determined, a review of the notes is in order. This not only will help you recall materials read earlier but also will lay the groundwork for decisions about where (if at all) a particular reference fits in terms of your outline. If certain references do not seem to fit anywhere, the outline may need to be revised or the reference discarded. Writing a Literature Review At this point, you will have completed the most difficult tasks of the literature review process, but that process is not complete until you have drafted and edited a written product. Although it is beyond the scope of this textbook to offer detailed guidance on writing research reviews, we offer a few comments on their content and style. Additional assistance is

CHAPTER 5 Reviewing the Literature Citation:

Type of Study:

Authors: Title: Journal: Year:

Issue

Volume

Quantitative

107

Pages: Both

Qualitative

Location/setting: Key Concepts/ Variables:

Concepts: Intervention/Independent Variable: Dependent Variable: Controlled Variables:

Design Type:

Experimental Specific Design: Descrip. of Intervention:

Quasi-experimental

Longitudinal/prospective Comparison group(s):

Cross-sectional

Qual. Tradition:

Grounded theory

Phenomenology

Sample:

Size: Sample characteristics:

Sampling method:

Data Sources:

Self-report Type: Description of measures:

Observational

Pre-experimental

Nonexperimental

No. of data collection points:

Ethnography

Other:

Biophysiologic

Other:

Pearson's r Factor analysis

Other: Other:

Data Quality: Statistical Tests:

ANOVA T test Bivariate: Multiple regression Multivariate:

Findings:

Recommendations: Strengths:

Weaknesses:

FIGURE 5.2

Example of a literature review protocol.

Chi-square MANOVA

108

PART 2 Conceptualizing a Research Study

provided in books such as those by Fink (1998) and Galvan (1999). Content of the Written Literature Review A written research review should provide readers with an objective, well-organized summary of the current state of knowledge on a topic. A literature review should be neither a series of quotes nor a series of abstracts. The central tasks are to summarize and critically evaluate the evidence so as to reveal the current state of knowledge on a topic—not simply to describe what researchers have done. The review should point out both consistencies and contradictions in the literature, and offer possible explanations for inconsistencies (e.g., different conceptualizations or data collection methods). Although important studies should be described in some detail, it is not necessary to provide extensive coverage for every reference (especially if there are page constraints). Reports of lesser significance that result in comparable findings can be summarized together. Example of grouped studies: McCullagh, Lusk, and Ronis (2002, p. 33) summarized several studies as follows: “Although noise-induced hearing loss is preventable through appropriate use of hearing protection devices, studies among farmers consistently show a low level of use (Broste et al., 1989; Engstrand, 1995; Hallet, 1987; Karlovich et al.,1988; Langsford et al., 1995).” The literature should be summarized in your own words. The review should demonstrate that consideration has been given to the cumulative significance of the body of research. Stringing together quotes from various documents fails to show that previous research has been assimilated and understood. The review should be objective, to the extent possible. Studies that conflict with personal values or hunches should not be omitted. The review also should not deliberately ignore a study because its findings contradict other studies. Inconsistent results should be analyzed and the supporting evidence evaluated objectively.

The literature review should conclude with a summary of the state of the art of knowledge on the topic. The summary should recap study findings and indicate how credible they are; it should also make note of gaps or areas of research inactivity. The summary thus requires critical judgment about the extensiveness and dependability of the evidence on a topic. If the literature review is conducted as part of a new study, this critical summary should demonstrate the need for the research and should clarify the context within which any hypotheses were developed. TIP: The literature review section of a research report (or research proposal) usually includes information not only about what is known about the problem and relevant interventions (if any), but about how prevalent the problem is. In research reports and proposals, the authors are trying to “build a case” for their new study. As you progress through this book, you will become increasingly proficient in critically evaluating the research literature. We hope you will understand the mechanics of writing a research review once you have completed this chapter, but we do not expect that you will be in a position to write a state-of-theart review until you have acquired more skills in research methods. Style of a Research Review Students preparing their first written research review often have trouble adjusting to the standard style of such reviews. For example, some students accept research results without criticism or reservation, reflecting a common misunderstanding about the conclusiveness of empirical research. You should keep in mind that no hypothesis or theory can be proved or disproved by empirical testing, and no research question can be definitely answered in a single study. Every study has some limitations, the severity of which is affected by the researcher’s methodologic decisions. The fact that theories and hypotheses cannot be ultimately proved or disproved does not, of course, mean that we must disregard evidence or challenge every idea

CHAPTER 5 Reviewing the Literature

109

TABLE 5.3 Examples of Stylistic Difficulties for Research Reviews* INAPPROPRIATE STYLE OR WORDING

RECOMMENDED CHANGE

1. It is known that unmet expectations engender anxiety.

Several experts (Greenberg, 2001; Cameron, 2000) have asserted that unmet expectations engender anxiety.

2. The woman who does not participate Previous studies have indicated that women who in childbirth preparation classes tends to participate in preparation for childbirth classes manifest manifest a high degree of stress during labor. less stress during labor than those who do not (Klotz, 2002; Mirling, 2000; McTygue, 2001). 3. Studies have proved that doctors and nurses do not fully understand the psychobiologic dynamics of recovery from a myocardial infarction.

The studies by Sacks (2000) and Carter (2001) suggest that doctors and nurses do not fully understand the psychobiologic dynamics of recovery from a myocardial infarction.

4. Attitudes cannot be changed quickly.

Attitudes have been found to be relatively enduring attributes that cannot be changed quickly (Dodge-Hanson, 2000; Woodward, 2001).

5. Responsibility is an intrinsic stressor.

According to Doctor A. Cassard, an authority on stress, responsibility is an intrinsic stressor (Cassard, 2000, 2001).

* All references are fictitious.

we encounter—especially if results have been replicated. The problem is partly a semantic one: hypotheses are not proved, they are supported by research findings; theories are not verified, but they may be tentatively accepted if there a substantial body of evidence demonstrates their legitimacy. TIP: When describing study findings, you should generally use phrases indicating tentativeness of the results, such as the following: • Several studies have found • Findings thus far suggest • Results from a landmark study indicated • The data supported the hypothesis . . . • There appears to be strong evidence that A related stylistic problem is an inclination of novice reviewers to interject opinions (their own or someone else’s) into the review. The review should include opinions sparingly and should be explicit

about their source. Reviewers’ own opinions do not belong in a review, with the exception of assessments of study quality. The left-hand column of Table 5-3 presents several examples of stylistic flaws. The right-hand column offers recommendations for rewording the sentences to conform to a more acceptable form for a research literature review. Many alternative wordings are possible. RESEARCH EXAMPLES OF RESEARCH L I T E R AT U R E R E V I E W S The best way to learn about the style, content, and organization of a research literature review is to read several reviews that appear in the nursing literature. We present two excerpts from reviews here and urge you to read other reviews on a topic of interest to you.

110

PART 2 Conceptualizing a Research Study

Research Example From a Quantitative Research Report Teel, Duncan, and Lai (2001) conducted a study about the experiences of 83 caregivers of patients who had had a stroke. A segment of their literature review that was included in their introduction follows (Teel et al., 2001, pp. 53–54). Over half a million Americans suffer strokes each year. Approximately 75% survive, yet most have residual neurologic impairment that requires supportive care (Gresham et al., 1995)*. Long-term assistance for many stroke patients is provided in home settings, by family caregivers who must acquire a number of new skills to successfully manage the outcomes of stroke... (Biegel et al., 1991; Evans et al., 1992; Jacob, 1991; Matson, 1994). Family home care management after stroke is essential, yet it is often stressful and demanding. The physical care requirements, vigilance, and altered roles that are often part of the stroke sequelae contribute to caregiving stress (Davis & Grant, 1994). Caregiving demands can have negative emotional and physical consequences for the family caregivers, which can, in turn, have negative implications for continuation of the caregiving role. Because of the potential effects, many caregiver outcomes have been studied. Mental health outcomes, including depression, perceived burden and strain, anxiety, and alternations in mood, have been examined relative to caregiving (Matson, 1994; Periard & Ames, 1993). Stroke caregivers were found to have higher depression scores than noncaregivers, and the elevated levels persisted at 1 year post-stroke (Schultz et al., 1988). For caregiving wives, increased social support was correlated with less depression (Robinson & Kaye, 1994). Physical health outcomes for caregivers have included assessments of general health and chronic illness (George & Gwyther, 1986), number of physician visits or days hospitalized (Cattanach & Tebes, 1991), assessment of immune function (Kiecolt-Glaser et al., 1991), and fatigue (Jensen & Given, 1991; Nygaard, 1988; Rabins et al., 1982; Teel & Press, 1999). Stroke caregivers have reported impaired physical health (Deimling & Bass, 1986), yet Rees and colleagues (1994) found no immunologic alterations in stroke caregivers who had been caring for at least 6 months compared with caregivers in a cross-sectional analysis of immune function.

* Consult the full research report for references cited in this excerpted literature review.

Evans and colleagues (1992) have suggested that the influence of the family also may affect stroke outcome. For example, the family can have a buffering effect on patient coping, with family emotional, informational, and practical support enhancing post-stroke coping ability. Patients categorized as having suboptimal home environments at 1 year after stroke had caregivers who were more likely to be depressed, less likely to be a spouse caregiver, had below average knowledge about stroke care, and reported more family dysfunction than caregivers in the optimal group (Evans et al., 1991). Because each of these variables has been associated with stroke outcome, post-stroke evaluation and treatment should include attention to minimizing caregiver depression and family dysfunction, while promoting knowledge about stroke care (Evans et al., 1991, 1992). Overall, the research literature is extremely limited in reports of outcomes for caregivers of stroke patients. In particular, studies about caregiving in the first several months after a stroke, a time in which there can be significant change in patient condition and caregiving routine, is virtually absent from the literature.

Research Example From a Qualitative Research Report Boydell, Goering, and Morrell-Bellai (2000) conducted a study of the experiences of 29 homeless individuals. A portion of the literature review for their research report follows (Boydell et al., 2000, pp. 26–27). Studies show that homelessness involves much more than not having a place to live. Individuals often lose their sense of identity, self-worth, and self-efficacy (Buckner, Bassuk, & Zima, 1993). Hallebone (1997) studied 38 homeless men ethnographically and found that psychosocial identities tend to be fragmented.... Taylor’s (1993) study involving qualitative interviews with 10 homeless women indicates that participants shared experiences of depersonalization and stigmatization and the subsequent effects on their personhood. It was found that being or appearing unclean and having an identity without certification (paper proof) greatly affected the women’s sense of self-esteem and personhood. Snow and Anderson (1993) report that those recently dislocated expressed a strong aversion to other homeless individuals. In contrast, those who had been homeless for extended periods of time were more likely than those re-

CHAPTER 5 Reviewing the Literature cently dislocated to embrace self-concepts such as tramp and bum. These unconventional self-concepts may be acquired and reinforced, at least in part, through social comparisons and identification with other homeless people (Grigsby et al., 1990).... Montgomery (1994) found that homeless women felt that their hard times contributed to the creation of a new and more positive self.... The literature also suggests that there is a spiritual dimension to the experience of homelessness that is often ignored. Matousek (1991) describes how the profound loss of self, which is associated with homelessness, presents a spiritual challenge to define one’s very existence.

S U M M A RY P O I N T S • A research literature review is a written summary of the state of existing knowledge on a research problem. The task of reviewing research literature involves the identification, selection, critical analysis, and written description of existing information on a topic. • Researchers review the research literature to develop research ideas, to determine knowledge on a topic of interest, to provide a context for a study, and to justify the need for a study; consumers review and synthesize evidence-based information to gain knowledge and improve nursing practice. • The most important type of information for a research review are findings from empirical studies. Various nonresearch references—including opinion articles, case reports, anecdotes, and clinical descriptions—may serve to broaden understanding of a research problem or demonstrate a need for research, but in general they have limited utility in written research reviews. • A primary source with respect to the research literature is the original description of a study prepared by the researcher who conducted it; a secondary source is a description of the study by a person unconnected with it. Primary sources should be consulted whenever possible in performing a literature review. • An important bibliographic development for locating references for a research review is the widespread availability of various electronic databases, many of which can be accessed

111

through an online search or by way of CDROM. For nurses, the CINAHL and MEDLINE ® databases are especially useful. • In searching a bibliographic database, users usually perform a subject search for a topic of interest, but other types of searches (e.g., textword search, author search) are available. • Although electronic information retrieval is widespread, print resources such as print indexes and abstract journals are also available. • References that have been identified must be screened for relevance and then read critically. For research reviews, most references are likely to be found in professional journals. • Research journal articles provide brief descriptions of research studies and are designed to communicate the contribution the study has made to knowledge. • Journal articles often consist of an abstract (a brief synopsis of the study) and four major sections: an introduction (explanation of the study problem and its context); method section (the strategies used to address the research problem); results section (the actual study findings); and discussion (the interpretation of the findings). • Research reports are often difficult to read because they are dense, concise, and contain a lot of jargon. Qualitative research reports are written in a more inviting and conversational style than quantitative ones, which are more impersonal and include information on statistical tests. • Statistical tests are procedures for testing research hypotheses and evaluating the believability of the findings. Findings that are statistically significant are ones that have a high probability of being reliable. • In preparing a written review, it is important to organize materials in a logical, coherent fashion. The preparation of an outline is recommended, and the development of summary charts often helps in integrating diverse studies. • The written review should not be a succession of quotes or abstracts. The reviewers’ role is to point out what has been studied, how adequate and dependable the studies are, what gaps exist in the

112

PART 2 Conceptualizing a Research Study

body of research, and (in the context of a new study), what contribution the study would make. STUDY ACTIVITIES Chapter 5 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing the concepts presented in this chapter. In addition, the following study suggestions can be addressed: 1. Read the study by Oermann and her colleagues (2001) entitled, “Teaching by the nurse: How important is it to patients?” Applied Nursing Research, 14, 11–17. Write a summary of the problem, methods, findings, and conclusions of the study. Your summary should be capable of serving as notes for a review of the literature. 2. Suppose that you were planning to study counseling practices and programs for rape trauma victims. Make a list of several key words relating to this topic that could be used for identifying previous work. 3. Below are five sentences from literature reviews that require stylistic improvements. Rewrite these sentences to conform to considerations mentioned in the text. (Feel free to give fictitious references.) a. Children are less distressed during immunization when their parents are present. b. Young adolescents are unprepared to cope with complex issues of sexual morality. c. More structured programs to use part-time nurses are needed. d. Intensive care nurses need so much emotional support themselves that they can provide insufficient support to patients. e. Most nurses have not been adequately educated to understand and cope with the reality of the dying patient. 4. Suppose you were studying factors relating to the discharge of chronic psychiatric patients. Obtain five bibliographic references for this topic. Compare your references and sources with those of other students.

SUGGESTED READINGS Methodologic References Allen, M. (1999). Nursing knowledge: Access via bibliographic databases. In L. Q. These (Ed.), Computers in nursing: Bridges to the future (pp. 149–170). Philadelphia: Lippincott Williams & Wilkins. American Psychological Association. (1994). Publication manual (4th ed.). Washington, DC: Author. Cooper, H. M. (1984). The integrative research review. Beverly Hills, CA: Sage. Fink, A. (1998). Conducting research literature reviews: From paper to the Internet. Thousand Oaks, CA: Sage. Fox, R. N., & Ventura, M. R. (1984). Efficiency of automated literature search mechanisms. Nursing Research, 33, 174–177. Galvan, J. L. (1999). Writing literature reviews. Los Angeles: Pyrczak. Ganong, L. H. (1987). Integrative reviews of nursing research. Research in Nursing & Health, 10, 1–11. Glaser, B. G. (1978). Theoretical sensitivity. Mill Valley, MA: The Sociology Press. Light, R. J., & Pillemer, D. B. (1984). Summing up: The science of reviewing research. Cambridge, MA: Harvard University Press. Martin, P. A. (1997) Writing a useful literature review for a quantitative research project. Applied Nursing Research, 10, 159–162. Munhall, P. L. (2001). Nursing research: A qualitative perspective. Sudbury, MA: Jones & Bartlett. Saba, V. K., Oatway, D. M., & Rieder, K. A. (1989). How to use nursing information sources. Nursing Outlook, 37, 189–195. Smith, L. W. (1988). Microcomputer-based bibliographic searching. Nursing Research, 37, 125–127. Spradley, J. (1979). The ethnographic interview. New York: Holt, Rinehart, & Winston. Van Manen, M. (1990). Researching lived experience. New York: State University of New York Press.

Studies Cited in Chapter 5 Abercrombie, P. D. (2001). Improving adherence to abnormal Pap smear follow-up. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 30, 80–88. Anderson, D. G., & Imle, M. A. (2001). Families of origin of homeless and never-homeless women. Western Journal of Nursing Research, 23, 394–413. Boydell, K. M., Goering, P., & Morrell-Bellai, T. L. (2000). Narratives of identity: Re-presentation of self

CHAPTER 5 Reviewing the Literature in people who are homeless. Qualitative Health Research, 10, 26–38. Hupcey, J. E. (2000). Feeling safe: The psychosocial needs of ICU patients. Journal of Nursing Scholarship, 32, 361–367. McCullagh, M., Lusk, S. L., & Ronis, D. L. (2002). Factors influencing use of hearing protection among farmers. Nursing Research, 51, 33–39. Stein-Parbury, J., & McKinley, S. (2000). Patients’ experiences of being in an intensive care unit: A select literature review. American Journal of Critical Care, 9, 20–27.

113

Teel, C. S., Duncan, P., & Lai, S. M. (2001). Caregiving experiences after stroke. Nursing Research, 50, 53–60. Tolle, S. W., Tilden, V. P., Rosenfeld, A. G., & Hickman, S. E. (2000). Family reports of barriers to optimal care of the dying. Nursing Research, 49, 310–317.

6

Developing a Conceptual Context

G

ood research usually integrates research findings into an orderly, coherent system. Such integration typically involves linking new research and existing knowledge through a thorough review of prior research on a topic (see Chapter 5) and by identifying or developing an appropriate conceptual framework. Both activities provide an important context for a research project. This chapter discusses theoretical and conceptual contexts for nursing research problems. THEORIES, MODELS, AND FRAMEWORKS Many terms have been used in connection with conceptual contexts for research, including theories, models, frameworks, schemes, and maps. There is some overlap in how these terms are used, partly because they are used differently by different writers, and partly because they are interrelated. We offer guidance in distinguishing these terms, but note that our definitions are not universal. Theories The term theory is used in many ways. For example, nursing instructors and students frequently use the term to refer to the content covered in classrooms, as opposed to the actual practice of performing

nursing activities. In both lay and scientific usage, the term theory connotes an abstraction. In research circles, the term theory is used differently by different authors. Classically, scientists have used theory to refer to an abstract generalization that offers a systematic explanation about how phenomena are interrelated. The traditional definition requires a theory to embody at least two concepts that are related in a manner that the theory purports to explain. Others, however, use the term theory less restrictively to refer to a broad characterization of a phenomenon. According to this less restrictive definition, a theory can account for (i.e., thoroughly describe) a single phenomenon. Some authors specifically refer to this type of theory as descriptive theory. For example, Fawcett (1999) defines descriptive theories as empirically driven theories that “describe or classify specific dimensions or characteristics of individuals, groups, situations, or events by summarizing commonalities found in discrete observations” (p. 15). Descriptive theory plays an especially important role in qualitative studies. Qualitative researchers often strive to develop a conceptualization of the phenomena under study that is grounded in the actual observations made by researchers. Components of a Traditional Theory As traditionally defined, scientific theories involve a series of propositions regarding interrelationships

CHAPTER 6 Developing a Conceptual Context

among concepts. The writings on scientific theory include a variety of terms such as proposition, postulate, premise, axiom, law, principle, and so forth, some of which are used interchangeably, and others of which introduce subtleties that are too complex for our discussion. Here, we present a simplified analysis of the components of a theory. Concepts are the basic building blocks of a theory. Examples of nursing concepts are adaptation, health, anxiety, and nurse–client interaction. Classical theories comprise a set of propositions that indicate a relationship among the concepts. Relationships are denoted by such terms as “is associated with,” “varies directly with,” or “is contingent on.” The propositions form a logically interrelated deductive system. This means that the theory provides a mechanism for logically arriving at new statements from the original propositions. Let us consider the following example, which illustrates these points. The Theory of Planned Behavior (TPB; Ajzen, 1988), which is an extension of an earlier theory called the Theory of Reasoned Action (Ajzen & Fishbein, 1980), provides a framework for understanding people’s behavior and its psychological determinants. A greatly simplified construction of the TPB consists of the following propositions: 1. Behavior that is volitional is determined by people’s intention to perform that behavior. 2. Intention to perform or not perform a behavior is determined by three factors: Attitudes toward the behavior (i.e., the overall evaluation of performing the behavior) Subjective norms (i.e., perceived social pressure to perform or not perform the behavior) Perceived behavioral control (i.e., anticipated ease or difficulty of engaging in the behavior) 3. The relative importance of the three factors in influencing intention varies across behaviors and situations. The concepts that form the basis of the TPB include behaviors, intentions, attitudes, subjective norms, and perceived self-control. The theory, which specifies the nature of the relationship

115

among these concepts, provides a framework for generating many hypotheses relating to health behaviors. We might hypothesize on the basis of the TPB, for example, that compliance with a medical regimen could be enhanced by influencing people’s attitudes toward compliance, or by increasing their sense of control. The TPB has been used as the underlying theory in studying a wide range of health decision-making behaviors, including contraceptive choice, AIDS prevention behavior, condom use, vaccination behavior, and preventive health screening. Example of the TPB: Aminzadeh and Edwards (2000) conducted a study, guided by the TPB, in which they examined factors associated with the use of a cane among community-dwelling older adults. Their results provided further evidence of the utility of the TPB in understanding health behaviors and have implications for the design of theory-based fall prevention interventions. Types of Traditional Theories Theories differ in their level of generality. Socalled grand theories or macrotheories purport to describe and explain large segments of the human experience. Some learning theorists, such as Clark Hull, or sociologists, such as Talcott Parsons, developed general theoretical systems to account for broad classes of behavior. Within nursing, theories are more restricted in scope, focusing on a narrow range of experience. Such middle-range theories attempt to explain such phenomena as decision-making, stress, selfcare, health promotion, and infant attachment. Conceptual Models Conceptual models, conceptual frameworks, or conceptual schemes (we use the terms interchangeably) represent a less formal attempt at organizing phenomena than theories. Conceptual models, like theories, deal with abstractions (concepts) that are assembled by virtue of their relevance to a common theme. What is absent from

116

PART 2 Conceptualizing a Research Study

conceptual models is the deductive system of propositions that assert and explain a relationship among concepts. Conceptual models provide a perspective regarding interrelated phenomena, but are more loosely structured than theories. A conceptual model broadly presents an understanding of the phenomenon of interest and reflects the assumptions and philosophic views of the model’s designer. Conceptual models can serve as springboards for generating research hypotheses. Much of the conceptual work that has been done in connection with nursing practice falls into the category we call conceptual models. These models represent world views about the nursing process and the nature of nurse—client relationships. A subsequent section of this chapter describes some conceptual models in nursing and illustrates how they have been used in nursing research. Schematic and Statistical Models The term model is often used in connection with symbolic representations of a conceptualization. There are many references in the research literature to schematic models and mathematic (or statistical) models. These models, like conceptual models, are constructed representations of some aspect of reality; they use concepts as building blocks, with a minimal use of words. A visual or symbolic representation of a theory or conceptual framework often helps to express abstract ideas in a concise and readily understandable form. Schematic models, which are common in both qualitative and quantitative research, represent phenomena graphically. Concepts and the linkages between them are represented through the use of boxes, arrows, or other symbols. An example of a schematic model (also referred to as a conceptual map) is presented in Figure 6-1. This model, known as Pender’s Health Promotion Model, is “a multivariate paradigm for explaining and predicting the health-promotion component of lifestyle” (Pender, Walker, Sechrist, & FrankStromborg, 1990, p. 326). Schematic models of this type can be useful in clarifying associations among concepts.

Statistical models are playing a growing role in quantitative studies. These models use symbols to express quantitatively the nature of relationships among variables. Few relationships in the behavioral sciences can be summarized as elegantly as in the mathematic model F ! ma (force ! mass " acceleration). Because human behavior is complex and subject to many influences, researchers typically are able to model it only in a probabilistic manner. This means that we are not able to develop equations, such as the example of force from mechanics, in which a human behavior can be simply described as the product of two other phenomena. What we can do, however, is describe the probability that a certain behavior or characteristic will exist, given the occurrence of other phenomena. This is the function of statistical models. An example of a statistical model is shown in the following: Y ! #1X1 $ #2X2 $ #3X3 $ #4X4 $ e where Y ! nursing effectiveness, as measured by a supervisor’s evaluation X1 ! nursing knowledge, as measured by a standardized test of knowledge X2 ! past achievement, as measured by grades in nursing school X3 ! decision-making skills, as measured by the Participation in Decision Activities Questionnaire X4 ! empathy, as measured by the Mehrabian Emotional Empathy Scale e ! a residual, unexplained factor #1, #2, #3, and #4 ! weights indicating the importance of X1, X2, X3, and X4, respectively, in determining nursing effectiveness Each term in this model is quantifiable; that is, every symbol can be replaced by a numeric value, such as an individual’s score on a standardized test of knowledge (X1). What does this equation mean and how does it work? This model offers a mechanism for understanding and predicting nursing effectiveness. The model proposes that nurses’ on-the-job effectiveness

CHAPTER 6 Developing a Conceptual Context

117

F I G U R E 6 . 1 The Health Promotion Model (From Pender, N. J., Walker, S. N., Sechrist, K. R., & FrankStromborg, M. [1990]. Predicting health-promoting lifestyles in the workplace. Nursing Research, 39, 331.)

is affected primarily by four factors: the nursing knowledge, past achievement, decision-making skill, and empathy of the nurse. These influences are not presumed to be equally important. The weights (!s) associated with each factor represent a recipe for describing the relative importance of each. If empathy were much more important than past achievement, for example, then the weights

might be 2 to 1, respectively (i.e., two parts empathy to one part past achievement). The e (or error term) at the end of the model represents all those unknown or unmeasurable other attributes that affect nurses’ performance. This e term would be set to a constant value; it would not vary from one nurse to another because it is an unknown element in the equation. Once values for the weights and e

118

PART 2 Conceptualizing a Research Study

have been established (through statistical procedures), the model can be used to predict the nursing effectiveness of any nurse for whom we have gathered information on the four Xs (standardized test scores and so forth). Our prediction of who will make an especially effective nurse is probabilistic and thus will not always be perfectly accurate, in part because of the influence of the unknown factors summarized by e. Perfect forecasting is rarely attainable with probabilistic statistical models. However, such a model makes prediction of nursing effectiveness less haphazard than mere guesswork or intuition. Frameworks A framework is the overall conceptual underpinnings of a study. Not every study is based on a theory or conceptual model, but every study has a framework. In a study based on a theory, the framework is referred to as the theoretical framework; in a study that has its roots in a specified conceptual model, the framework is often called the conceptual framework (although the terms conceptual framework and theoretical framework are frequently used interchangeably). In many cases, the framework for a study is not an explicit theory or conceptual model, but rather is implicit—that is, not formally acknowledged or described. The concepts in which researchers are interested are by definition abstractions of observable phenomena, and our world view (and views on nursing) shape how those concepts are defined and operationalized. What often happens, however, is that researchers fail to clarify the conceptual underpinnings of their research variables, thereby making it more difficult to integrate research findings. As noted in Chapter 2, researchers undertaking a study should make clear the conceptual definition of their key variables, thereby providing information about the study’s framework. Waltz, Strickland, and Lenz (1991) describe a five-step process for developing conceptual definitions. These steps include (1) developing a preliminary definition, (2) reviewing relevant literature, (3) developing or identifying exemplary cases, (4)

mapping the meaning of the concept, and (5) stating the developed conceptual definition. Example of developing a conceptual definition: Beck (1996) provides an example of developing a conceptual definition of the concept panic: 1. Preliminary definition: Panic is a sudden, unpredictable rush of overpowering terror that is associated with a marked physiological uproar along with a loss of reasoning capacity and fears of dying and going crazy. 2. Literature review: Inter- and intra-disciplinary review of panic is undertaken. 3. Developing exemplary cases: Linda is married and the mother of two children, a 3-year-old daughter and 3-month-old son. She is in her late 20s and is a college graduate.... After her first delivery, Linda experienced postpartum panic disorder. With her first child, the panic began 6 weeks postpartum during her daughter’s christening. The panic suddenly came out of nowhere. Her heart started racing, her hands got sweaty, and she could not stop crying. For Linda, it was unbearable trying to stay seated in church. She wanted to run outside very badly. Linda shared that the feelings inside her head were so painful she thought she was going crazy and could not focus on anything. As a result of the panic, Linda altered her lifestyle. She spent a tremendous amount of energy trying to appear normal. She went to great lengths to avoid a panic attack. For example, whenever she went to church, she would make sure she sat in the back and at the end of the aisle so she could quickly exit if she felt panic beginning. While experiencing panic, even if only for 5 or 10 minutes, Linda felt minutes were like hours. When panicking, Linda loses herself. She feels like she is not sitting there (pp. 271–272). 4. Mapping the concept’s meaning: By integrating literature with empirical observations, a conceptual map was developed that organized the various meanings of panic in the literature. 5. Stating the revised theoretical definition: Panic is a sudden, unpredictable rush of overpowering terror that has an all-or-nothing quality and is associated with (1) a marked physiological uproar such as palpitations, faintness and sweating; (2) a distortion of time and loss of reasoning capacity, which engenders fearful cognitions of dying, impending doom, losing control, and/or going crazy; and (3) an intense desire to flee the situation and avoid it in the future (Beck, 1996, pp. 271–272).

CHAPTER 6 Developing a Conceptual Context

Quantitative researchers in general are more guilty of failing to identify their frameworks than qualitative researchers. In most qualitative studies, the frameworks are part of the research tradition within which the study is embedded. For example, ethnographers usually begin their work within a theory of culture. Grounded theory researchers incorporate sociologic principles into their framework and their approach to looking at phenomena. The questions that most qualitative researchers ask and the methods they use to address those questions inherently reflect certain theoretical formulations. T H E N AT U R E O F THEORIES AND CONCEPTUAL MODELS Theories and conceptual models have much in common, including their origin, general nature, purposes, and role in research. In this section, we examine some general characteristics of theories and conceptual models. We use the term theory in its broadest sense, inclusive of conceptual models. Origin of Theories and Models Theories and conceptual models are not discovered; they are created and invented. Theory building depends not only on the observable facts in our environment but also on the originator’s ingenuity in pulling those facts together and making sense of them. Thus, theory construction is a creative and intellectual enterprise that can be engaged in by anyone who is insightful, has a solid knowledge base, and has the ability to knit together observations and evidence into an intelligible pattern. Tentative Nature of Theories and Models Theories and conceptual models cannot be proved. A theory is a scientist’s best effort to describe and explain phenomena; today’s successful theory may be discredited tomorrow. This may happen if new evidence or observations undermine a previously accepted theory. Or, a new theoretical system

119

might integrate new observations with an existing theory to yield a more parsimonious explanation of a phenomenon. Theories and models that are not congruent with a culture’s values and philosophic orientation also may fall into disfavor over time. It is not unusual for a theory to lose supporters because some aspects of it are no longer in vogue. For example, certain psychoanalytic and structural social theories, which had broad support for decades, have come to be challenged as a result of changes in views about women’s roles. This link between theory and values may surprise you if you think of science as being completely objective. Remember, though, that theories are deliberately invented by humans; they are not totally free from human values and ideals, which can change over time. Thus, theories and models are never considered final and verified. There always remains the possibility that a theory will be modified or discarded. Many theories in the physical sciences have received considerable empirical support, and their wellaccepted propositions are often referred to as laws, such as Boyle’s law of gases. Nevertheless, we have no way of knowing the ultimate accuracy and utility of any theory and should, therefore, treat all theories as tentative. This caveat is nowhere more relevant than in emerging sciences such as nursing. Purposes of Theories and Conceptual Models Theoretical and conceptual frameworks play several interrelated roles in the progress of a science. Their overall purpose is to make research findings meaningful and generalizable. Theories allow researchers to knit together observations and facts into an orderly scheme. They are efficient mechanisms for drawing together accumulated facts, sometimes from separate and isolated investigations. The linkage of findings into a coherent structure can make the body of accumulated evidence more accessible and, thus, more useful. In addition to summarizing, theories and models can guide a researcher’s understanding of not only the what of natural phenomena but also the

120

PART 2 Conceptualizing a Research Study

why of their occurrence. Theories often provide a basis for predicting the occurrence of phenomena. Prediction, in turn, has implications for the control of those phenomena. A utilitarian theory has potential to bring about desirable changes in people’s behavior or health. Theories and conceptual models help to stimulate research and the extension of knowledge by providing both direction and impetus. Many nursing studies have been generated explicitly to examine aspects of a conceptual model of nursing. Thus, theories may serve as a springboard for advances in knowledge and the accumulation of evidence for practice. Relationship Between Theory and Research The relationship between theory and research is reciprocal and mutually beneficial. Theories and models are built inductively from observations, and an excellent source for those observations is prior research, including in-depth qualitative studies. Concepts and relationships that are validated empirically through research become the foundation for theory development. The theory, in turn, must be tested by subjecting deductions from it (hypotheses) to further systematic inquiry. Thus, research plays a dual and continuing role in theory building and testing. Theory guides and generates ideas for research; research assesses the worth of the theory and provides a foundation for new theories. It would be unreasonable to assert, however, that research without a formal substantive theory cannot contribute to nursing practice. In nursing research, many facts still need to be accumulated, and purely descriptive inquiries may well form the basis for subsequent theoretical developments. Research that does not test a theory can potentially be linked to one at a later time. Suggestions for linking a study to a conceptual framework are presented later in this chapter. CONCEPTUAL MODELS USED IN NURSING RESEARCH Nurse researchers have used both nursing and nonnursing frameworks to provide a conceptual con-

text for their studies. This section briefly discusses several frameworks that have been found useful by nurse researchers. Conceptual Models of Nursing In the past few decades, several nurses have formulated a number of conceptual models of nursing practice. These models constitute formal explanations of what the nursing discipline is and what the nursing process entails, according to the model developer’s point of view. As Fawcett (1995) has noted, four concepts are central to models of nursing: person, environment, health, and nursing. The various conceptual models, however, define these concepts differently, link them in diverse ways, and give different emphases to relationships among them. Moreover, different models emphasize different processes as being central to nursing. For example, Sister Calista Roy’s Adaptation Model identifies adaptation of patients as a critical phenomenon (Roy & Andrews, 1991). Martha Rogers (1986), by contrast, emphasizes the centrality of the individual as a unified whole, and her model views nursing as a process in which clients are aided in achieving maximum well-being within their potential. The conceptual models were not developed primarily as a base for nursing research. Indeed, these models have had more impact on nursing education and clinical practice than on nursing research. Nevertheless, nurse researchers have turned to these conceptual frameworks for inspiration in formulating research questions and hypotheses. Table 6-1 (p. 122) lists 10 prominent conceptual models in nursing that have been used by researchers. The table briefly describes the model’s key feature and identifies a study that has claimed the model as its framework. Two nursing models that have generated particular enthusiasm among researchers are described in greater detail. Roy’s Adaptation Model In Roy’s Adaptation Model, humans are viewed as biopsychosocial adaptive systems who cope with environmental change through the process of

CHAPTER 6 Developing a Conceptual Context

adaptation. Within the human system, there are four subsystems: physiologic needs, self-concept, role function, and interdependence. These subsystems constitute adaptive modes that provide mechanisms for coping with environmental stimuli and change. The goal of nursing, according to this model, is to promote client adaptation; nursing also regulates stimuli affecting adaptation. Nursing interventions usually take the form of increasing, decreasing, modifying, removing, or maintaining internal and external stimuli that affect adaptation. Example using Roy’s model: Cook, Green, and Topp (2001) explored the incidence and impact of physician verbal abuse on perioperative nurses, using the Roy Adaptation Model as their conceptual framework. The researchers examined how nurses used adaptive coping behaviors and problem-focused skills to deal with the abuse. Orem’s Self-Care Model Orem’s Self-Care Model focuses on each individual’s ability to perform self-care, defined as “the practice of activities that individuals initiate and perform on their own behalf in maintaining life, health, and well-being” (1985, p. 35). Ability to care for oneself is referred to as self-care agency, and the ability to care for others is referred to as dependent-care agency. In Orem’s model, the goal of nursing is to help people meet their own therapeutic self-care demands. Orem identified three types of nursing systems: (1) wholly compensatory, wherein the nurse compensates for the patient’s total inability to perform self-care activities; (2) partially compensatory, wherein the nurse compensates for the patient’s partial inability to perform these activities; and (3) supportive—educative, wherein the nurse assists the patient in making decisions and acquiring skills and knowledge. Example using Orem’s model: McCaleb and Cull (2000) studied the influence of sociocultural characteristics and economic circumstances on the self-care practices of middle adolescents, using Orem’s model as the framework.

121

Other Models Developed by Nurses In addition to conceptual models that are designed to describe and characterize the entire nursing process, nurses have developed other models and theories that focus on more specific phenomena of interest to nurses. Two important examples are Pender’s Health Promotion Model and Mishel’s Uncertainty in Illness Theory. The Health Promotion Model Nola Pender’s (1996) Health Promotion Model (HPM) focuses on explaining health-promoting behaviors, using a wellness orientation. According to the model (see Figure 6-1), health promotion entails activities directed toward developing resources that maintain or enhance a person’s well-being. The HPM encompasses two phases: a decision-making phase and an action phase. In the decision-making phase, the model emphasizes seven cognitive/perceptual factors that compose motivational mechanisms for acquiring and maintaining health-promoting behaviors and five modifying factors that indirectly influence patterns of health behavior. In the action phase, barriers and cues to action trigger activity in health-promoting behavior. Nurse researchers have used the HPM in numerous studies of health promoting behaviors. Example using the HPM: McCullagh, Lusk, and Ronis (2002) used the Pender HPM to identify factors affecting farmer’s use of hearing protection devices. The findings offered further support of the HPM. Uncertainty in Illness Theory Mishel’s Uncertainty in Illness Theory (Mishel, 1988) focuses on the concept of uncertainty—the inability of a person to determine the meaning of illness-related events. According to this theory, people develop subjective appraisals to assist them in interpreting the experience of illness and treatment. Uncertainty occurs when people are unable to recognize and categorize stimuli. Uncertainty results in the inability to obtain a clear conception of the situation, but a situation appraised as uncertain will

122

PART 2 Conceptualizing a Research Study

TABLE 6.1 Conceptual Models of Nursing Used by Nurse Researchers THEORIST AND REFERENCE

NAME OF MODEL/THEORY

KEY THESIS OF THE MODEL

Imogene King, 1981

Open Systems Model

Personal systems, interpersonal systems, and social systems are dynamic and interacting, within which transactions occurs.

Doornbos (2000) based her framework on King’s model; she tested the prediction that family stressors, coping, and other factors affected family health with young adults with serious mental illness.

Madeline Leininger, 1991

Theory of Culture Care Diversity and Universality

Caring is a universal phenomenon but varies transculturally.

Raines and Morgan (2000) studied the culturally grounded meanings of the concept of comfort, presence, and involvement in the context of the childbirth experience of black women and white women.

Myra Levine, 1973

Conservation Model

Conservation of integrity contributes to maintenance of a person’s wholeness.

Deiriggi and Miles (1995) based their study of the effects of waterbeds on heart rate in preterm infants on Levine’s concept of conservation.

Betty Neuman, 1989

Health Care Systems Model

Each person is a complete system; the goal of nursing is to assist in maintaining client system stability.

Brauer (2001) described common patterns of person– environment interaction in adults with rheumatoid arthritis, based on Neuman’s model.

RESEARCH EXAMPLE

Margaret Newman, Health as Expanding Health is viewed as an 1994 Consciousness expansion of consciousness with health and disease parts of the same whole; health is seen in an evolving pattern of the whole in time, space, and movement.

Endo and colleagues (2000) used Newman’s theory to study pattern recognition as a caring partnership between nurses and families of ovarian cancer in Japan.

Dorothea Orem, 1985

Anderson (2001) explored, with a sample of homeless adults, the relationship between selfcare, self-care agency, and well-being.

Self-Care Model

Self-care activities are what people do on their own behalf to maintain health and well-being; the goal of nursing is to help people meet their own therapeutic self-care demands.

CHAPTER 6 Developing a Conceptual Context

TABLE 6.1 Conceptual Models of Nursing Used by Nurse Researchers

123

(continued)

THEORIST AND REFERENCE

NAME OF MODEL/THEORY

KEY THESIS OF THE MODEL

RESEARCH EXAMPLE

Rosemarie Rizzo Parse, 1992, 1995

Theory of Human Becoming

Health and meaning are co-created by indivisible humans and their environment; nursing involves having clients share views about meanings.

Mitchell and Lawton (2000) studied how diabetic patients’ experienced the consequences of personal choices about living with restrictions, and discussed the emerging concepts within Parse’s theory.

Martha Rogers, 1970, 1986

Science of Unitary Human Beings

The individual is a unified whole in constant interaction with the environment; nursing helps individuals achieve maximum well-being within their potential.

Using Rogers’ framework, Bays (2001) explored the phenomenon of hope and associated factors in older patients who had experienced a stroke.

Sr. Callista Roy, 1984, 1991

Adaptation Model

Humans are adaptive systems that cope with change through adaptation; nursing helps to promote client adaptation during health and illness.

Roy’s Adaptation Model provided the framework for John’s (2001) study of whether perceptions of quality of life change over time in adults who receive curative radiation therapy.

Jean Watson, 1999

Theory of Caring

Caring is the moral ideal, and entails mind–body–soul engagement with one another.

Using Watson’s 10 carative factors, Baldursdottir and Jonsdottir (2002) studied the importance of nurse caring behaviors as perceived by patients receiving care at an emergency department.

mobilize individuals to use their resources to adapt to the situation. Mishel’s conceptualization of uncertainty has been used as a framework for both qualitative and quantitative nursing studies. Example using Uncertainty in Illness Theory: Santacroce (2001) studied uncertainty in 25 mothers during their infants’ diagnosis; the infants were HIV seropositive.

Other Models Used by Nurse Researchers Many phenomena in which nurse researchers are interested involve concepts that are not unique to nursing, and therefore their studies are sometimes linked to conceptual models that are not models from the nursing profession. In addition to the previously described Theory of Planned

124

PART 2 Conceptualizing a Research Study

Behavior, three non-nursing models or theories have frequently been used in nursing research investigations: the Health Belief Model, Lazarus and Folkman’s Theory of Stress and Coping, and Bandura’s Social Cognitive Theory.

or her well-being. The model posits that coping strategies are learned, deliberate responses used to adapt to or change stressors. According to this model, a person’s perception of mental and physical health is related to the ways he or she evaluates and copes with the stresses of living.

The Health Belief Model The Health Belief Model (HBM; Becker, 1978) has become a popular conceptual framework in nursing studies focused on patient compliance and preventive health care practices. The model postulates that health-seeking behavior is influenced by a person’s perception of a threat posed by a health problem and the value associated with actions aimed at reducing the threat. The major components of the HBM include perceived susceptibility, perceived severity, perceived benefits and costs, motivation, and enabling or modifying factors. Perceived susceptibility is a person’s perception that a health problem is personally relevant or that a diagnosis is accurate. Even when one recognizes personal susceptibility, action will not occur unless the individual perceives the severity to be high enough to have serious organic or social implications. Perceived benefits are the patients’ beliefs that a given treatment will cure the illness or help prevent it, and perceived costs are the complexity, duration, and accessibility of the treatment. Motivation is the desire to comply with a treatment. Among the modifying factors that have been identified are personality variables, patient satisfaction, and sociodemographic factors.

Example using the Theory of Stress and Coping: Maurier and Northcott (2000) used the Lazarus and Folkman theory as the conceptual framework in a study that examined whether job uncertainty, working conditions, cognitive appraisal, and coping strategies affected the health of nurses during the restructuring of health care in Alberta.

Example using the HBM: Petro-Nustas (2001) used the HBM as the theoretical framework for a study of young Jordanian women’s health beliefs toward mammography as a screening procedure for breast cancer. Lazarus and Folkman’s Theory of Stress and Coping The Theory of Stress and Coping (Folkman & Lazarus, 1988; Lazarus, 1966) is an effort to explain people’s methods of dealing with stress, that is, environmental and internal demands that tax or exceed a person’s resources and endanger his

Bandura’s Social Cognitive Theory Social Cognitive Theory (Bandura, 1986, 1997) offers an explanation of human behavior using the concepts of self-efficacy, outcome expectations, and incentives. Self-efficacy expectations are focused on people’s belief in their own capacity to carry out particular behaviors (e.g., smoking cessation). Selfefficacy expectations, which are context-specific, determine the behaviors a person chooses to perform, their degree of perseverance, and the quality of the performance. Bandura identified four factors that influence a person’s cognitive appraisal of selfefficacy: (1) their own mastery experience; (2) verbal persuasion; (3) vicarious experience; and (4) physiologic and affective cues, such as pain and anxiety. The role of self-efficacy has been studied in relation to numerous health behaviors such as weight control, self-management of chronic illness, phobic reactions, and smoking. Example using Social Cognitive Theory: Using social cognitive constructs, Resnick (2001) tested a model of factors that influence the exercise behavior of older adults. Theoretical Contexts and Nursing Research As previously noted, theory and research have reciprocal, beneficial ties. Fawcett (1978) described the relationship between theory and research as a

CHAPTER 6 Developing a Conceptual Context

double helix, with theory as the impetus of scientific investigations, and with findings from research shaping the development of theory. However, this relationship has not always characterized the progress of nursing science. Many have criticized nurse researchers for producing numerous isolated studies that are not placed in a theoretical context. This criticism was more justified a decade ago than it is today. Many researchers are developing studies on the basis of conceptual models of nursing. Nursing science is still struggling, however, to integrate accumulated knowledge within theoretical systems. This struggle is reflected, in part, in the number of controversies surrounding the issue of theoretical frameworks in nursing. One of these controversies concerns whether there should be one single, unified model of nursing or multiple, competing models. Fawcett (1989) has argued against combining different models, noting that “before all nurses follow the same path, the competition of multiple models is needed to determine the superiority of one or more of them” (p. 9). Research can play a critical role in testing the utility and validity of alternative nursing models. Another controversy involves the desirability and utility of developing theories unique to nursing. Some commentators argue that theories relating to humans developed in other disciplines, such as physiology, psychology, and sociology (so-called borrowed theories), can and should be applied to nursing problems. Others advocate the development of unique nursing theories, claiming that only through such development can knowledge to guide nursing practice be produced. Fawcett (1995) argues that borrowed theories are sometimes used without considering their adequacy for nursing inquiry. When a borrowed theory is tested and found to be empirically adequate in health-relevant situations of interest to nurses, it becomes shared theory. Until these controversies are resolved, nursing research is likely to continue on its current path of conducting studies within a multidisciplinary and multitheoretical perspective. We are inclined to see the use of multiple frameworks as a healthy and unavoidable part of the development of nursing science.

125

TESTING, USING, AND D E V E L O P I N G A T H E O RY OR FRAMEWORK The manner in which theory and conceptual frameworks are used by qualitative and quantitative researchers is elaborated on in the following section. In the discussion, the term theory is used in its broadest sense to include both conceptual models and formal theories. Theories and Qualitative Research Theory is almost always present in studies that are embedded in a qualitative research tradition such as ethnography or phenomenology. As previously noted, these research traditions inherently provide an overarching framework that give qualitative studies a theoretical grounding. However, different traditions involve theory in different ways. Sandelowski (1993) makes a useful distinction between substantive theory (conceptualizations of the target phenomena that are being studied) and theory that reflects a conceptualization of human inquiry. Some qualitative researchers insist on an atheoretical stance vis-à-vis the phenomenon of interest, with the goal of suspending a priori conceptualizations (substantive theories) that might bias their collection and analysis of data. For example, phenomenologists are in general committed to theoretical naiveté, and explicitly try to hold preconceived views of the phenomenon in check. Nevertheless, phenomenologists are guided in their inquiries by a framework or philosophy that focuses their analysis on certain aspects of a person’s lifeworld. That framework is based on the premise that human experience is an inherent property of the experience itself, not constructed by an outside observer. Ethnographers typically bring a strong cultural perspective to their studies, and this perspective shapes their initial fieldwork. Fetterman (1989) has observed that most ethnographers adopt one of two cultural theories: ideational theories, which suggest that cultural conditions and adaptation stem from mental activity and ideas, or materialistic theories, which view material conditions (e.g.,

126

PART 2 Conceptualizing a Research Study

resources, money, production) as the source of cultural developments. The theoretical underpinning of grounded theory studies is symbolic interactionism, which stresses that behavior is developed through human interactions, through ongoing processes of negotiation and renegotiation. Similar to phenomenologists, however, grounded theory researchers attempt to hold prior substantive theory (existing knowledge and conceptualizations about the phenomenon) in abeyance until their substantive theory begins to emerge. Once the theory starts to take shape, grounded theorists do not ignore the literature; rather, previous literature is used for comparison with the emerging and developing categories of the theory. The goal of grounded theory researchers is to develop a conceptualization of a phenomenon that is grounded in actual observations—that is, to explicate an empirically based conceptualization for integrating and making sense of a process or phenomenon. Theory development in a grounded theory study is an inductive process. Grounded theory researchers seek to identify patterns, commonalities, and relationships through the scrutiny of specific instances and events. During the ongoing analysis of data, the researchers move from specific pieces of data to abstract generalizations that synthesize and give structure to the observed phenomenon. The goal is to use the data, grounded in reality, to provide a description or an explanation of events as they occur in reality—not as they have been conceptualized in preexisting theories. Grounded theory methods are designed to facilitate the generation of theory that is conceptually dense, that is, with many conceptual patterns and relationships. Example of a grounded theory study: Schreiber, Stern, and Wilson (2000) developed a grounded theory of how black West IndianCanadian women manage depression and its stigma. These women from a nondominant cultural background used the process they called “being strong” to manage depression. “Being strong” included three subprocesses of “dwelling on it,” “diverting myself,” and “regaining composure.” As illustrated in their graphic model (Figure 6-2), these subprocesses overlap.

In grounded theory studies, theory is produced “from the inside,” but theory can also enter a qualitative study “from the outside.” That is, some qualitative researchers use existing theory as an interpretive framework. For example, a number of qualitative nurse researchers acknowledge that the philosophic roots of their studies lie in conceptual models of nursing such as those developed by Neuman, Parse, and Rogers. Other qualitative researchers use substantive theories about the target phenomenon as a comparative context for interpreting data after the researcher has undertaken a preliminary analysis. Sandelowski (1993) notes that, in this manner, previous substantive theories or conceptualizations are essentially data themselves, and can be taken into consideration, along with study data, as part of an inductively driven new conceptualization. Example of using existing theory as an interpretive framework: In Yeh’s (2001) qualitative study of the adaptation process of 34 Taiwanese children with cancer, she used Roy’s Adaptation Model as a guide for indepth interviews and also for her data analysis. An integrative review of qualitative research studies in a specific topic is another strategy that can lead to theory development. In such integrative reviews, qualitative studies are combined to identify their essential elements. These findings from different sources are then used for theory building. Paterson (2001), for example, used the results of 292 qualitative studies that described the experiences of adults with chronic illness to develop the shifting perspectives model of chronic illness. This model depicts living with chronic illness as an ongoing, constantly shifting process in which individuals’ perspectives change in the amount to which illness is in the foreground or background in their lives. Theories in Quantitative Research Quantitative researchers, like qualitative researchers, link research to theory or models in several ways. The classic approach is to test hypotheses deduced from a previously proposed theory.

gn

Diverting myself

Regaining my composure

es ach pro ap

Tr yi n

127

w ne

Testing a Theory Theories often stimulate new studies. For example, a nurse might read about Orem’s Self-Care Model. As reading progresses, conjectures such as the following might arise: “If Orem’s Self-Care Model is valid, then I would expect that nursing effectiveness can be enhanced in environments more conducive to self-care (e.g., a birthing room versus a delivery room).” Such a conjecture, derived from a theory or conceptual framework, can serve as a starting point for testing the theory’s adequacy. In testing a theory, researchers deduce implications (as in the preceding example) and develop research hypotheses, which are predictions about the manner in which variables would be related if the theory were correct. The hypotheses are then subjected to empirical testing through systematic data collection and analysis.

in g

F I G U R E 6 . 2 Being strong: how black West-Indian Canadian women manage depression and its stigma. (Adapted with permission from Schreiber, R., Stern, P. N., & Wilson, C. [2000]. Being strong: How black West-Indian Canadian women manage depression and its stigma. Journal of Nursing Scholarship, 32, p. 41.)

Cultural stigma of depression

y Tr

Male-female roles and relationships

ew ap pro ac he s

CHAPTER 6 Developing a Conceptual Context

Dwelling on it

Trying new approaches

Belief in Christian doctrine

Visible minority status within a Eurocentric society

TIP: If you are testing a specific theory or conceptual model, be sure to read about it from a primary source. It is important to understand fully the conceptual perspective of the theorist, and the detailed explication of key constructs. The focus of the testing process involves a comparison between observed outcomes with those predicted in the hypotheses. Through this process, a theory is continually subjected to potential disconfirmation. If studies repeatedly fail to disconfirm a theory, it gains support and acceptance (e.g., the Theory of Planned Behavior). The testing process continues until pieces of evidence cannot be interpreted within the context of the theory but can be explained by a new theory that also accounts for previous findings. Theory-testing studies are most useful when researchers devise logically sound deductions from the theory, design a study that reduces the plausibility of alternative

128

PART 2 Conceptualizing a Research Study

explanations for observed relationships, and use methods that assess the theory’s validity under maximally heterogeneous situations so that potentially competing theories can be ruled out. Researchers sometimes base a new study on a theory in an effort to explain earlier descriptive findings. For example, suppose several researchers had found that nursing home patients demonstrate greater levels of anxiety and noncompliance with nursing staff around bedtime than at other times. These findings shed no light on underlying causes of the problem, and consequently suggest no way to improve it. Several explanations, rooted in models such as Lazarus and Folkman’s Stress and Coping Model or Neuman’s Health Care Systems Model, may be relevant in explaining nursing home patients’ behavior. By directly testing the theory in a new study (i.e., deducing hypotheses derived from the theory), a researcher might learn why bedtime is a vulnerable period for the elderly in nursing homes. TIP: It may be useful to read research reports of other studies that were based on a theory in which you are interested—even if the research problem is not similar to your own. By reading other studies, you will be better able to judge how much empirical support the theory has received and perhaps how the theory should be adapted. Tests of a theory sometimes take the form of testing a theory-based intervention. If a theory is correct, it has implications for strategies to influence people’s attitudes or behavior, including health-related ones. The impetus for an intervention may be a theory developed within the context of qualitative studies, as in the example of Swanson’s theory of caring described later in this chapter. The actual tests of the effectiveness of the intervention—which are also indirect tests of the theory— are usually done in structured quantitative research. Example of theory testing in an intervention study: Chang (1999) used Lazarus and Folkman’s Theory of Stress and Coping to develop and test an intervention for homebound caregivers of persons with dementia (PWD). According to the theory, the relationship between stress and a person’s coping ability

is mediated by primary appraisal—the perception of an experience as stressful or nonstressful. Chang reasoned that an intervention that affects primary appraisal could positively affect caregiver anxiety and depression. She developed a cognitive-behavioral intervention designed to provide caregivers with knowledge and skills to improve the PWD’s eating and dressing abilities, and also to increase caregiver knowledge of coping strategies. In a careful study that compared caregivers who received the intervention with those who did not, Chang found that depression decreased more in the intervention group. Researchers sometimes combine elements from more than one theory as a basis for generating hypotheses. In doing this, researchers need to be thoroughly knowledgeable about both theories to see if there is an adequate conceptual and empirical basis for conjoining them. If underlying assumptions or conceptual definitions of key concepts are not compatible, the theories should not be combined (although perhaps elements of the two can be used to create a new conceptual framework with its own assumptions and definitions). Two theories that have given rise to combinatory efforts are the HBM and the Theory of Reasoned Action. Example of testing two combined models: Poss (2001) combined the HBM and the Theory of Reasoned Action to examine the factors associated with the participation by Mexican migrant farm workers in a tuberculosis screening program. Figure 6-3 illustrates how Poss integrated these two models/theories in her study. Testing Two Competing Theories Researchers who directly test two competing theories to explain a phenomenon are in a particularly good position to advance knowledge. Almost all phenomena can be explained in alternative ways, as suggested by the alternative conceptual models of nursing. There are also competing theories for such phenomena as stress, child development, and grieving, all of which are important to nursing. Each competing theory suggests alternative approaches to facilitating a positive outcome or minimizing a negative one. In designing effective

CHAPTER 6 Developing a Conceptual Context

Attitude Cues to action

Intention

Behavior

Severity Subjective norm Normative beliefs

129

a study—an enterprise that may be difficult for beginning researchers. For advanced students, Chinn and Kramer (1999) and Fawcett (1999) present criteria for assessing conceptual frameworks. Box 6-1 presents a few basic questions that can be asked in a preliminary assessment of a theory or model.

Behavioral beliefs

Susceptibility

HBM + TRA HBM TRA

F I G U R E 6 . 3 Combined Health Belief Model and Theory of Reasoned Action. (Adapted with permission from Poss, J. E. [2001]. Developing a new model for cross-cultural research: Synthesizing the Health Belief Model and the Theory of Reasoned Action. Advances in Nursing Science, 23, p. 12.)

nursing interventions, it is important to know which explanation has more validity. Typically, researchers test a single theory (or one combined model) in a study. Then, to evaluate the worth of competing theories, they must compare the results of different studies. Such comparisons are problematic because each study design is unique. For example, one study of stress might use a sample of college students taking an examination, another might use military personnel in a combat situation, and yet another might use terminally ill patients with cancer. Each study might, in addition to having divergent samples, measure stress differently. If the results of these studies support alternative theories of stress to different degrees, it would be difficult to know the extent to which the results reflected differences in study design rather than differences in the validity of the theories. TIP: It is often suggested that theories first be evaluated before being used as a basis for

The researcher who directly tests two (or more) competing theories, using a single sample of subjects and comparable measures of the key research variables, is in a position to make powerful and meaningful comparisons. Such a study requires considerable advance planning and the inclusion of a wider array of measures than would otherwise be the case, but such efforts are important. In recent years, several nurse researchers have used this approach to generate and refine our knowledge base and to provide promising new leads for further research. Example of a test of competing theories: Yarcheski, Mahon, and Yarcheski (1999) tested three alternative theories of anger in early adolescents: one relating anger to stress; another attributing anger to differential emotions; and a third relating anger to personality traits. The findings suggested that all three theories are sound and relevant explanations, but the trait theory provided the most powerful explanation. Using a Model or Theory as an Organizing Structure Many researchers who cite a theory or model as their framework are not directly testing the theory. Silva (1986), in her analysis of 62 studies that used 5 nursing models, found that only 9 were direct and explicit tests of the models cited by the researchers. She found that the most common use of nursing models in empirical studies was to provide an organizing structure. In such an approach, a researcher begins with a broad conceptualization of nursing (or stress, health beliefs, and so on) that is consistent with that of the model developers. These researchers assume that the models they espouse are valid, and then use the model’s constructs or proposed schemas to provide a broad organizational

130

PART 2 Conceptualizing a Research Study

BOX 6.1 Some Questions for a Preliminary Assessment of a Model or Theory ISSUE

QUESTIONS

Theoretical clarity

• Are key concepts defined and are the definitions sufficiently clear? • Do all concepts “fit’’ within the theory? Are concepts used in the theory in a manner compatible with conceptual definitions? • Are basic assumptions consistent with one another? • Are schematic models helpful, and are they compatible with the text? Are schematic models needed but not presented? • Can the theory be followed—is it adequately explained? Are there ambiguities?

Theoretical complexity

• Is the theory sufficiently rich and detailed to explain phenomena of interest? • Is the theory overly complex? • Can the theory be used to explain or predict, or only to describe phenomena?

Theoretical grounding

• • • •

Appropriateness of the theory

• Does the theory suggest possibilities for influencing nursing practice? • Are the tenets of the theory compatible with nursing’s philosophy? • Are key concepts within the domain of nursing?

Importance of the theory

• Will research based on this theory answer critical questions? • How will testing the theory contribute to nursing’s evidence base?

General issues

• Are there theories or models that do a better job of explaining phenomenon of interest? • Is the theory compatible with your world view?

Are the concepts identifiable in reality? Is there an empirical basis for the theory? Is the empirical basis solid? Can the theoretical concepts be adequately operationalized?

or interpretive context. Some use (or develop) data collection instruments that are allied with the model. Silva noted that using models in this fashion can serve a valuable organizing purpose, but such studies offer little evidence about the validity of the theory itself. Example of using a model as organizing structure: Resnick and Jenkins (2000) used Bandura’s social cognitive theory as an organizing structure to revise an unpublished instrument to measure self-efficacy barriers to exercise. Focusing on Bandura’s construct of self-efficacy, Resnick and Jenkins developed the Self-Efficacy for Exercise Scale. They assessed their new measure with 187 older adults living in a continuing care retirement community.

To our knowledge, Silva’s study has not been replicated with a more recent sample of studies. However, we suspect that, even today, most quantitative studies that offer models and theories as their conceptual frameworks are using them primarily as organizational or interpretive tools. Silva (1986) offered seven evaluation criteria for determining whether a study was actually testing a theory, rather than simply identifying an organizational context. Box 6-2, broadly adapted from Silva’s criteria, presents a set of evaluative questions to determine if a study was actually testing a theory. Example of a study meeting seven theorytesting criteria: Woods and Isenberg (2001) tested one aspect of a middle-range theory based on the Roy Adaptation

CHAPTER 6 Developing a Conceptual Context

131

BOX 6.2 Criteria to Determine if a Theory/Model is Being Tested 1. Is the purpose of the study to determine the validity of a theory’s assumptions or propositions? 2. Does the report explicitly note that the theory is the framework for the research? 3. Is the theory discussed in sufficient detail that the relationship between the theory on the one hand and the study hypotheses or research questions on the other is clear? 4. Are study hypotheses directly deduced from the theory? 5. Are study hypotheses empirically tested in an appropriate manner, so as to shed light on the validity of the theory? 6. Is the validity of the theory’s assumptions or propositions supported (or challenged) based on evidence from the empirical tests? 7. Does the report discuss how evidence from empirical tests supports or refutes the theory, or how the theory explains relevant aspects of the findings? Adapted from Silva M. C. (1986). Research testing nursing theory: State of the art. Advances in Nursing Science, 9, 1–11.

Model. Their purpose was to test the efficacy of adaptation as a mediator of intimate abuse and traumatic stress in battered women. They tested the following two relational statements: “(1) physical abuse, emotional abuse, and risk of homicide are focal stimuli that elicit the response of post traumatic stress disorder (PTSD) in battered women and (2) adaptation in the physiologic, selfconcept, role, and interdependence modes acts as a mediator between the focal stimuli of intimate physical abuse, emotional abuse, and risk of homicide and the response of PTSD in women” (p. 215). As a result of empirical testing, Woods and Isenberg reported that adaptation in three of the four modes partially mediated the relationship between intimate abuse, the focal stimuli, and the response of PTSD. Fitting a Problem to a Theory The preceding sections addressed the situation in which a researcher begins with a specific theory or model and uses it either as the basis for developing hypotheses or for an organizational or interpretive purpose. Circumstances sometimes arise in which the problem is formulated before any consideration is given to a conceptual framework. Even in such situations, researchers sometimes try to devise a

theoretical context. Although in some situations such an approach may be appropriate, we nevertheless caution that an after-the-fact linkage of theory to a problem may add little to the study’s worth and, of course, no evidence of the theory’s validity. (An exception is when the researcher is struggling to make sense of findings, and calls on an existing theory to help explain or interpret them.) If it is necessary to find a relevant theory after selecting a problem, the search for such a theory must begin by first conceptualizing the problem on an abstract level. For example, take the research question: “Do daily telephone conversations between a psychiatric nurse and a patient for 2 weeks after discharge from the hospital result in lower rates of readmission by short-term psychiatric patients?” This is a relatively concrete research problem, but it might profitably be viewed within the context of Orem’s Self-Care Model, a theory of reinforcement, a theory of social influence, or a theory of crisis resolution. Part of the difficulty in finding a theory is that a single phenomenon of interest can be conceptualized in a number of ways and, depending on the manner chosen, may refer the researcher to conceptual schemes from a wide range of disciplines.

132

PART 2 Conceptualizing a Research Study

TIP: If you begin with a research problem and are trying to identify a suitable framework, it is probably wise to confer with others—especially with people who may be familiar with a broad range of theoretical perspectives. By having an open discussion, you are more likely to become aware of your own conceptual perspectives and are thus in a better position to identify an appropriate framework. Textbooks, handbooks, and encyclopedias in the chosen discipline usually are a good starting point for the identification of a framework. These sources usually summarize the status of a theoretical position and document the efforts to confirm and disconfirm it. Journal articles contain more current information but are usually restricted to descriptions of specific studies rather than to broad expositions or evaluations of theories. Perhaps our brief overview of frameworks that have been useful to nurses can serve as a starting point for identifying a suitable model or theory. Fitting a problem to a theory after-the-fact should be done with circumspection. It is true that having a theoretical context can enhance the meaningfulness of a study, but artificially “cramming” a problem into a theory is not the route to scientific utility, nor to enhancing nursing’s evidence base. There are many published studies that purport to have a conceptual framework when, in fact, the tenuous post hoc linkage is all too evident. In Silva’s (1986) previously mentioned analysis of 62 studies that claimed a nursing model as their underpinnings, approximately one third essentially paid only lip service to a model. If a conceptual framework is really linked to a research problem, then the design of the study, the selection of data collection methods, the data analysis, and (especially) the interpretation of the findings flow from that conceptualization. We advocate a balanced and reasoned perspective on this issue: Researchers should not shirk their intellectual duties by ignoring attempts to link their problem to broader theoretical concerns, but there is no point in fabricating such a link when it does not exist. TIP: If you begin with a research question and then subsequently identify a theory or

model, be willing to adapt or augment your original research problem as you gain greater understanding of the theory. The linking of theory and research question may involve an iterative approach. Developing a Framework in a Quantitative Study Novice researchers may think of themselves as unqualified to develop a conceptual scheme of their own. But theory development depends much less on research experience than on powers of observation, grasp of a problem, and knowledge of prior research. There is nothing to prevent an imaginative and sensitive person from formulating an original conceptual framework for a study. The conceptual scheme may not be a full-fledged formal theory, but it should place the issues of the study into some broader perspective. The basic intellectual process underlying theory development is induction—that is, the process of reasoning from particular observations and facts to generalizations. The inductive process involves integrating what one has experienced or learned into some conclusion. The observations used in the inductive process need not be personal observations; they may be (and often are) the findings and conclusions from other studies. When relationships among variables are arrived at this way, one has the makings of a theory that can be put to a more rigorous test. The first step in theory development, then, is to formulate a generalized scheme of relevant concepts, that is, to perform a conceptual analysis. Let us consider the following simple example. Suppose that we were interested in understanding the factors influencing enrollment in a prenatal education program. We might begin by considering two basic sets of forces: those that promote enrollment and those that hinder it. After reviewing the literature, discussing the problem with colleagues, and developing ideas from our own experiences, we might arrive at a conceptual scheme such as the one presented in Figure 6-4. This framework is crude, but it does allow us to study a number of research questions and to place those problems in perspective. For example, the conceptual scheme suggests that as the availability of social supports declines, obstacles to participation in a prenatal

CHAPTER 6 Developing a Conceptual Context

Positive Factors

Negative Factors

Age/maturity

Lack of social supports

Wantedness/intendedness of pregnancy

Enrollment in prenatal educational program

Maternal/paternal educational level

133

Practical impediments (cost, accessibility)

Competing time demands

F I G U R E 6 . 4 Conceptual model—factors that influence enrollment in a prenatal education program.

education program increase. We might then make the following hypothesis: “Single pregnant women are less likely to participate in a prenatal education program than married pregnant women,” on the assumption that husbands are an important source of social support to women in their pregnancy. (Of course, this example is contrived; in reality, several existing theories such as the HBM or TPB could be used to study enrollment in prenatal care education.) Many nursing studies involve conceptual frameworks developed by the researchers. Example of model development: Stuifbergen, Seraphine, and Roberts (2000) conducted a study based on their own conceptual model of quality of life in persons with chronic disabling conditions. The model represented “a synthesis of findings from extant literature and a series of preliminary qualitative and quantitative investigations” (p. 123). RESEARCH EXAMPLES Throughout this chapter, we have described studies that involved various widely used conceptual and theoretical models. This section presents two examples of the linkages between theory and research from

the nursing research literature—one from a quantitative study and the other from a qualitative study. Research Example From a Quantitative Study: Testing Orem’s Self-Care Model Renker (1999) used Orem’s Self-Care Model of nursing to study the relationships between self-care, social support, physical abuse, and pregnancy outcomes of older adolescent mothers and their infants. The study’s research variables included measures of the major constructs in Orem’s model, including basic conditioning factors, self-care agency, and self-care. Orem’s basic conditioning factors (factors that affect people’s ability to engage in self-care) include (1) social-environmental factors and (2) resource availability and adequacy factors. In Renker’s study, physical abuse represented the key social environmental factor, and social support represented the resource availability and adequacy factor. Denyes’ Self-Care Agency Instrument was used to measure pregnancy self-care agency, and Denyes’ Self-Care Practice Instrument measured pregnancy self-care. Based on Orem’s model, Renker hypothesized that the absence of physical abuse and the presence of social support increased self-care agency. Increased levels of self-care agency were expected to enhance

134

PART 2 Conceptualizing a Research Study

self-care practices, which in turn were expected to result in increased infant birth weight and decreased pregnancy complications. Renker tested her hypotheses in a sample of 139 pregnant teenagers. The results lent support to Orem’s model. Abused pregnant teenagers gave birth to infants with significantly lower birth weights than the teenagers who were not abused. Social support, self-care agency, and self-care practice were all significantly related to infant birth weight. A particular strength of this study is that Orem’s Self-Care Deficit Model was interwoven throughout its design. Renker developed hypotheses based on the model and included all major constructs of Orem’s model as research variables. Moreover, several data collection instruments were specifically developed to assess components of Orem’s theory.

Research Example From a Qualitative Study: Development of a Theory of Caring As noted earlier in this chapter, many qualitative studies have theory development as an explicit goal. Here we describe the efforts of a qualitative researcher who developed an empirically derived theory of caring, and has used the theory in the development of a caring-based counseling intervention (Swanson, 1999). Although the qualitative studies were done over a decade ago, they are an excellent illustration of theory development. Using data from three separate qualitative investigations, Swanson (1991) inductively derived and then refined a theory of the caring process. Swanson studied caring in three separate perinatal contexts: as experienced by women who miscarried, as provided by parents and professionals in the newborn intensive care unit, and as recalled by at-risk mothers who had received a long-term public health nursing intervention. Data were gathered through in-depth interviews with study participants and also through observations of care provision. Data from the first study led to the identification and preliminary definition of five caring processes. The outcome of the second study was confirmation of the five processes and refinement of their definitions. In the third study, Swanson confirmed the five processes, redefined one of them, developed subdimensions of each process, and derived a definition of the overall concept of caring: “Caring is a nurtur-

ing way of relating to a valued other toward whom one feels a personal sense of commitment and responsibility” (p. 165). According to Swanson’s theory, the five caring processes are as follows:

• Knowing—striving to understand an event as it has meaning in the life of the other

• Being With—being emotionally present to the other • Doing For—doing for the other as he or she would do for the self if it were at all possible

• Enabling—facilitating the other’s passage through life transitions and unfamiliar events

• Maintaining Belief—sustaining faith in the other’s capacity to get through an event or transition and face a future with meaning In presenting her theory, Swanson described the five processes, supporting each with rich excerpts from her in-depth interviews. Here is an example of the excerpt illustrating the process of knowing: When things weren’t right, I could say that things were fine and it was only a matter of time. I mean the nurse would ask certain questions and there would be no way that I could be consistent without telling the truth. And then we would talk, and pretty soon instead of saying it was fine, I would start out with what was really wrong. (p. 163)

Swanson’s theory of caring, in addition to being used in the development and testing of a caring-based nurse counseling program for women who miscarry (Swanson, 1999), has been used by other researchers, including a qualitative study of the interactions of AIDS family caregivers and professional health care providers (Powell-Cope, 1994) and a study of the involvement of relatives in the care of the dying (Andershed & Ternestedt, 1999).

S U M M A RY P O I N T S • A theory is a broad abstract characterization of phenomena. As classically defined, a theory is an abstract generalization that systematically explains relationships among phenomena. Descriptive theory thoroughly describes a phenomenon. • In a research context, the overall objective of theory is to make findings meaningful, to summarize existing knowledge into coherent systems,

CHAPTER 6 Developing a Conceptual Context

• •

to stimulate new research, and to explain phenomena and relationships among them. The basic components of a theory are concepts; classically defined theories consist of a set of propositions about the interrelationships among concepts, arranged in a logically interrelated system that permits new statements to be derived from them. Grand theories (or macrotheories) attempt to describe large segments of the human experience. Middle-range theories are more specific to certain phenomena. Concepts are also the basic elements of conceptual models, but concepts are not linked in a logically ordered, deductive system. Conceptual models, like theories, provide context for nursing studies. Schematic models (sometimes referred to as conceptual maps) are representations of phenomena using symbols or diagrams. Statistical models use mathematic symbols to express quantitatively the nature and strength of relationships among variables. A framework is the conceptual underpinning of a study. In many studies, the framework is implicit, but ideally researchers clarify the conceptual definitions of key concepts. In qualitative studies, the framework usually springs from distinct research traditions. Several conceptual models of nursing have been developed and have been used in nursing research. The concepts central to models of nursing are person, environment, health, and nursing. Two major conceptual models of nursing used by nurse researchers are Orem’s Self-Care Model and Roy’s Adaptation Model. Non-nursing models used by nurse researchers (e.g., Lazarus and Folkman’s Theory of Stress and Coping) are referred to as borrowed theories; when the appropriateness of borrowed theories for nursing inquiry is confirmed, the theories become shared theories. In some qualitative research traditions (e.g., phenomenology), the researcher strives to suspend previously held substantive conceptualiza-

135

tions of the phenomena under study, but nevertheless there is a rich theoretical underpinning associated with the tradition itself. Some qualitative researchers specifically seek to develop grounded theories, data-driven explanations to account for phenomena under study through inductive processes. In the classical use of theory, researchers test hypotheses deduced from an existing theory. A particularly fruitful approach is the testing of two competing theories in one study. In both qualitative and quantitative studies, researchers sometimes use a theory or model as an organizing framework, or as an interpretive tool. Researchers sometimes develop a problem, design a study, and then look for a conceptual framework; such an after-the-fact selection of a framework is less compelling than the systematic testing of a particular theory.

STUDY ACTIVITIES Chapter 6 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Read the following article: Liken, M. A. (2001). Caregivers in crisis. Clinical Nursing Research, 10, 52–68. What theoretical basis does the author use in this study? Would you classify the theoretical basis as a theory or as a conceptual framework? Did Liken use the framework to test hypotheses formally, or was the framework used as an organizational structure? 2. Select one of the nursing conceptual frameworks or models described in this chapter. Formulate a research question and two hypotheses that could be used empirically to test the utility of the conceptual framework or model in nursing practice. 3. Four researchable problems follow. Abstract a generalized issue or issues for each of these

136

PART 2 Conceptualizing a Research Study

problems. Search for an existing theory that might be applicable and appropriate. a. What is the relationship between angina pain and alcohol intake? b. What effect does rapid weight gain during the second trimester have on the outcome of pregnancy? c. Do multiple hospital readmissions affect the achievement level of children? d. To what extent do coping mechanisms of individuals differ in health and illness? 4. Read the following article: Kelly-Powell, M. L. (1997). Personalizing choices: Patients’ experiences with making treatment decisions. Research in Nursing & Health, 20, 219–227. What evidence does the researcher offer to substantiate that her grounded theory is a good fit with her data? SUGGESTED READINGS Theoretical References Andrews, H. A., & Roy, C. (1986). Essentials of the Roy Adaptation Model. Norwalk, CT: Appleton-CenturyCrofts. Ajzen, I. (1988). Attitudes, personality, and behavior. Milton Keynes, United Kingdom: Open University Press. Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall. Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W. H. Freeman. Becker, M. (1978). The Health Belief Model and sick role behavior. Nursing Digest, 6, 35–40. Chinn, P. L., & Kramer, M. K. (1999). Theory and nursing: Integrated knowledge development (5th ed.). St. Louis: C. V. Mosby. Craig, S. L. (1980). Theory development and its relevance for nursing. Journal of Advanced Nursing, 5, 349–355. Fawcett, J. (1978). The relationship between theory and research: A double helix. Advances in Nursing Science, 1, 49–62.

Fawcett, J. (1989). Analysis and evaluation of conceptual models of nursing (2nd ed.). Philadelphia: F. A. Davis. Fawcett, J. (1995). Analysis and evaluation of conceptual models of nursing (3rd ed.). Philadelphia: F. A. Davis. Fawcett, J. (1999). The relationship between theory and research (3rd ed.). Philadelphia: F. A. Davis. Fetterman, D. M. (1989). Ethnography: Step by step. Newbury Park, CA: Sage. Flaskerud, J. H. (1984). Nursing models as conceptual frameworks for research. Western Journal of Nursing Research, 6, 153–155. Flaskerud, J. H., & Halloran, E. J. (1980). Areas of agreement in nursing theory development. Advances in Nursing Science, 3, 1–7. Folkman, S., & Lazarus, R. S. (1988). Coping as a mediator of emotion. Journal of Personality and Social Psychology, 54, 466–475. Hardy, M. E. (1974). Theories: Components, development, evaluation. Nursing Research, 23, 100–107. King, I. M. (1981). A theory for nursing: Systems, concepts, and process. New York: John Wiley and Sons. Lazarus, R. (1966). Psychological stress and the coping response. New York: McGraw-Hill. Leininger, M. (1991). Culture care diversity and universality: A theory of nursing. New York: National League for Nursing. Levine, M. E. (1973). Introduction to clinical nursing (2nd ed.). Philadelphia: F. A. Davis. Marriner-Tomey, A. (Ed.). (1998). Nursing theorists and their work (4th ed.). St. Louis: C. V. Mosby. Mehrabian, A., & Epstein, N. (1972). A measure of emotional empathy. Journal of Personality, 40, 525–543. Meleis, A. I. (1997). Theoretical nursing: Development and progress (3rd ed.). Philadelphia: LippincottRaven.IQ2 Mishel, M. H. (1988), Uncertainty in illness. Image: Journal of Nursing Scholarship, 20, 225–232. Neuman, B. (1989). The Neuman Systems Model (2nd ed.). Norwalk, CT: Appleton & Lange. Newman, M. (1994). Health as expanding consciousness. New York: NLN. Nicoll, L. H. (1996). Perspectives on nursing theory. Philadelphia: J. B. Lippincott. Orem, D. E. (1985). Concepts of practice (3rd ed.). New York: McGraw-Hill. Parse, R. R. (1992). Human becoming: Parse’s theory. Nursing Science Quarterly, 5, 35–42. Parse, R. R. (1995). Illuminations: The Human Becoming Theory in practice and research. New York: NLN Pub. #15-2670.

CHAPTER 6 Developing a Conceptual Context Pender, N. (1996). Health promotion in nursing practice (3rd ed.). Englewood Cliffs, NJ: Prentice-Hall. Rogers, M. E. (1970). An introduction to the theoretical basis of nursing. Philadelphia: F. A. Davis. Rogers, M. E. (1986). Science of unitary human beings. In V. Malinski (Ed.), Explorations on Martha Rogers’ science of unitary human beings. Norwalk, CT: Appleton-Century-Crofts. Roy, C. (1984). Introduction to nursing: An adaptation model (2nd ed.). Englewood Cliffs, NJ: Prentice-Hall. Roy, C. Sr., & Andrews, H. (1991). The Roy Adaptation Model: The definitive statement. Norwalk, CT: Appleton & Lange. Sandelowski, M. (1993). Theory unmasked: The uses and guises of theory in qualitative research. Research in Nursing & Health, 16, 213–218. Silva, M. C. (1986). Research testing nursing theory: State of the art. Advances in Nursing Science, 9, 1–11. Silva, M. C., & Sorrell, J. M. (1992). Testing of nursing theory: Critique and philosophical expansion. Advances in Nursing Science, 14, 12–23. Stevens, B. J. (1994). Nursing theory: Analysis, application, evaluation (4th ed.). Philadelphia: J. B. Lippincott. Waltz, C., Strickland, O., & Lenz, E. (1991). Measurement in nursing research. Philadelphia: F. A. Davis. Watson, J. (1999). Postmodern nursing and beyond. New York: Churchill Livingston.

Studies Cited in Chapter 6 Aminzadeh, F., & Edwards, N. (2000). Factors associated with cane use among community-dwelling older adults. Public Health Nursing, 17, 474–483. Andershed, B., & Ternestedt, B. (1999). Involvement of relatives in care of the dying in different care cultures. Nursing Science Quarterly, 12, 45–51. Anderson, J. A. (2001). Understanding homeless adults by testing the theory of self-care. Nursing Science Quarterly, 14, 59–67. Baldursdottir, G., & Jonsdottir, H. (2002). The importance of nurse caring behaviors as perceived by patients receiving care at an emergency department. Heart & Lung, 31, 67–75. Bays, C. L. (2001). Older adults’ descriptions of hope after stroke. Rehabilitation Nursing, 26, 18–20. Beck, C. T. (1996). A concept analysis of panic, Archives of Psychiatric Nursing, 10, 265–275.

137

Brauer, D. J. (2001). Common patterns of person-environment interaction in persons with rheumatoid arthritis. Western Journal of Nursing Research, 23, 414–430. Chang, B. L. (1999). Cognitive-behavioral intervention for homebound caregivers of persons with dementia. Nursing Research, 48, 173–182. Cook, J. K., Green, M., & Topp, R. V. (2001). Exploring the impact of physician abuse on perioperative nurses. AORN Journal, 74, 317–327. Deiriggi, P. M., & Miles, K. E. (1995). The effects of waterbeds on heart rate in preterm infants. Scholarly Inquiry for Nursing Practice, 9, 245–262. Doornbos, M. M. (2000). King’s systems framework and family health: The derivation and testing of a theory. Journal of Theory Construction and Testing, 4, 20–26. Endo, E., Nitta., N., Inayoshi, M., Saito, R., Takemura, K., Minegishi, H., Kubo, S., & Kondo, M. (2000). Pattern recognition as a caring partnership in families with cancer. Journal of Advanced Nursing, 32, 603–610. John, L. D. (2001). Quality of life in patients receiving radiation therapy for non-small cell lung cancer. Oncology Nursing Forum, 28, 807–813. Maurier, W. L., & Northcott, H. C. (2000). Job uncertainty and health status for nurses during restructuring of health care in Alberta. Western Journal of Nursing Research, 22, 623–641. McCaleb, A., & Cull, V. V. (2000). Sociocultural influences and self-care practices of middle adolescents. Journal of Pediatric Nursing, 15, 30–35. McCullagh, M., Lusk, S. L., & Ronis, D. L. (2002). Factors influencing use of hearing protection among farmers. Nursing Research, 51, 33–39. Mitchell, G. J., & Lawton, C. (2000). Living with the consequences of personal choices for persons with diabetes. Canadian Journal of Diabetes Care, 24, 23–30. Paterson, B. L. (2001). The shifting perspectives model of chronic illness. Journal of Nursing Scholarship, 33, 21–26. Pender, N. J., Walker, S. N., Sechrist, K. R., & FrankStromborg, M. (1990). Predicting health-promoting lifestyles in the workplace. Nursing Research, 39, 326–332. Petro-Nustas, W. (2001). Young Jordanian women’s health beliefs about mammography. Journal of Community Health Nursing, 18, 177–194. Poss, J. E. (2001). Developing a new model for crosscultural research: Synthesizing the Health Belief

138

PART 2 Conceptualizing a Research Study

Model and the Theory of Reasoned Action. Advances in Nursing Science, 23, 1–15. Powell-Cope, G. M. (1994). Family caregivers of people with AIDS: Negotiating partnerships with professional health care providers. Nursing Research, 43, 324–330. Raines, D. A., & Morgan, Z. (2000). Culturally sensitive care during childbirth. Applied Nursing Research, 13, 167–172. Renker, P. R. (1999). Physical abuse, social support, selfcare, and pregnancy outcomes of older adolescents. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 28, 377–388. Resnick, B. (2001). Testing a model of exercise behavior in older adults. Research in Nursing & Health, 24, 83–92. Resnick, B., & Jenkins, L. S. (2000). Testing the reliability and validity of the Self-Efficacy for Exercise Scale. Nursing Research, 49, 154–159. Santacroce, S. J. (2001). Measuring parental uncertainty during the diagnosis phase of serious illness in a child. Journal of Pediatric Nursing, 16, 3–12. Schreiber, R., Stern, P. N., & Wilson, C. (2000). Being strong: How black West-Indian Canadian women

manage depression and its stigma. Journal of Nursing Scholarship, 32, 39–45. Stuifbergen, A. K., Seraphine, A., & Roberts, G. (2000). An explanatory model of health promotion and quality of life in chronic disabling conditions. Nursing Research, 49, 122–129. Swanson, K. M. (1991). Empirical development of a middle range theory of caring. Nursing Research, 40, 161–166. Swanson, K. M. (1999). Effects of caring, measurement, and time on miscarriage impact and women’s wellbeing. Nursing Research, 48, 288–298. Woods, S. J., & Isenberg, M. A. (2001). Adaptation as a mediator of intimate abuse and traumatic stress in battered women. Nursing Science Quarterly, 14, 215–221. Yarcheski, A., Mahon, N. E., & Yarcheski, T. J. (1999). An empirical test of alternate theories of anger in early adolescents. Nursing Research, 48, 317–323. Yeh, C. H. (2001). Adaptation in children with cancer: Research with Roy’s model. Nursing Science Quarterly, 14, 141–148.

PA R T

3

Designs for Nursing Research

7

Designing Ethical Research

T

his part of the textbook presents materials relating to the planning and design stage of empirical research. Ethical concerns permeate every aspect of the design of a study and the execution of the design. Therefore, before discussing techniques of research design, we present in this chapter major ethical principles that must be considered in developing research plans. The proliferation of research has led to growing concerns about the protection of the rights of study participants. Ethical concerns are especially prominent in the field of nursing because the line of demarcation between what constitutes the expected practice of nursing and the collection of research information has become less distinct as research by nurses increases. Furthermore, ethics can create particular challenges to nurse researchers because ethical requirements sometimes conflict with the need to produce evidence of the highest possible quality for practice. THE NEED FOR ETHICAL GUIDELINES When humans are used as study participants—as they usually are in nursing research—care must be exercised in ensuring that the rights of those humans are protected. The requirement for ethical conduct may strike you as so self-evident as to

require no further comment, but the fact is that ethical considerations have not always been given adequate attention. In this section, we consider some of the reasons that ethical guidelines became imperative. Historical Background As modern, civilized people, we might like to think that systematic violations of moral principles within a research context occurred centuries ago rather than in recent times, but this is not the case. The Nazi medical experiments of the 1930s and 1940s are the most famous example of recent disregard for ethical conduct. The Nazi program of research involved the use of prisoners of war and racial “enemies” in numerous experiments designed to test the limits of human endurance and human reaction to diseases and untested drugs. The studies were unethical not only because they exposed these people to permanent physical harm and even death but because subjects could not refuse participation. Some recent examples of ethical transgressions have also occurred in the United States. For instance, between 1932 and 1972, a study known as the Tuskegee Syphilis Study, sponsored by the U.S. Public Health Service, investigated the effects of syphilis among 400 men from a poor

142

PART 3 Designs for Nursing Research

African-American community. Medical treatment was deliberately withheld to study the course of the untreated disease. Another well-known case of unethical research involved the injection of live cancer cells into elderly patients at the Jewish Chronic Disease Hospital in Brooklyn, without the consent of those patients. Even more recently, it was revealed in 1993 that U.S. federal agencies had sponsored radiation experiments since the 1940s on hundreds of people, many of them prisoners or elderly hospital patients. Many other examples of studies with ethical transgressions— often much more subtle than these examples— have emerged to give ethical concerns the high visibility they have today. Ethical Dilemmas in Conducting Research Research that violates ethical principles is rarely done specifically to be cruel or immoral, but more typically occurs out of a conviction that knowledge is important and potentially life-saving or beneficial to others in the long run. There are research problems in which participants’ rights and study demands are put in direct conflict, posing ethical dilemmas for researchers. Here are examples of research problems in which the desire for rigor conflicts with ethical considerations: 1. Research question: How empathic are nurses in their treatment of patients in the intensive care unit (ICU)? Ethical dilemma: Ethics require that participants be cognizant of their role in a study. Yet if the researcher informs nurses participating in this study that their degree of empathy in treating ICU patients will be scrutinized, will their behavior be “normal?” If the nurses’ usual behavior is altered because of the known presence of research observers, the findings will not be valid. 2. Research question: What are the coping mechanisms of parents whose children have a terminal illness? Ethical dilemma: To answer this question, the researcher may need to probe into the psycho-

logical state of the parents at a vulnerable time in their lives; such probing could be painful and even traumatic. Yet knowledge of the parents’ coping mechanisms might help to design more effective ways of dealing with parents’ grief and anger. 3. Research question: Does a new medication prolong life in patients with cancer? Ethical dilemma: The best way to test the effectiveness of an intervention is to administer the intervention to some participants but withhold it from others to see if differences between the groups emerge. However, if the intervention is untested (e.g., a new drug), the group receiving the intervention may be exposed to potentially hazardous side effects. On the other hand, the group not receiving the drug may be denied a beneficial treatment. 4. Research question: What is the process by which adult children adapt to the day-to-day stresses of caring for a terminally ill parent? Ethical dilemma: In a qualitative study, which would be appropriate for this research question, the researcher sometimes becomes so closely involved with participants that they become willing to share “secrets” and privileged information. Interviews can become confessions—sometimes of unseemly or even illegal or immoral behavior. In this example, suppose a participant admitted to physically abusing an adult parent—how does the researcher respond to that information without undermining a pledge of confidentiality? And, if the researcher divulges the information to appropriate authorities, how can a pledge of confidentiality be given in good faith to other participants? As these examples suggest, researchers involved with human participants are sometimes in a bind. They are obligated to advance knowledge and develop the highest-quality evidence for practice, using the best methods available; however, they must also adhere to the dictates of ethical rules that have been developed to protect human rights. Another type of dilemma arises from the fact that

CHAPTER 7 Designing Ethical Research

nurse researchers may be confronted with conflictof-interest situations, in which their expected behavior as nurses comes into conflict with the expected behavior of researchers (e.g., deviating from a standard research protocol to give needed assistance to a patient). It is precisely because of such conflicts and dilemmas that codes of ethics have been developed to guide the efforts of researchers. Codes of Ethics During the past four decades, largely in response to the human rights violations described earlier, various codes of ethics have been developed. One of the first internationally recognized efforts to establish ethical standards is referred to as the Nuremberg Code, developed after the Nazi atrocities were made public in the Nuremberg trials. Several other international standards have subsequently been developed, the most notable of which is the Declaration of Helsinki, which was adopted in 1964 by the World Medical Association and then later revised, most recently in 2000. Most disciplines have established their own code of ethics. The American Nurses’ Association (ANA) put forth a document in 1995 entitled Ethical Guidelines in the Conduct, Dissemination, and Implementation of Nursing Research (Silva, 1995). Box 7-1 presents the nine ethical principles outlined in that document. The American Sociological Association published a revised Code of Ethics in 1997. Guidelines for psychologists were published by the American Psychological Association (1992) in Ethical Principles of Psychologists and Code of Conduct. Although there is considerable overlap in the basic principles articulated in these documents, each deals with problems of particular concern to their respective disciplines. In the United States, an important code of ethics was adopted by the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research (1978). The commission, established by the National Research Act (Public Law 93–348), issued a report in 1978 that served as the basis for regulations affecting research sponsored by the federal government. The

143

report, sometimes referred to as the Belmont Report, also served as a model for many of the guidelines adopted by specific disciplines. The Belmont Report articulated three primary ethical principles on which standards of ethical conduct in research are based: beneficence, respect for human dignity, and justice. TIP: The following websites offer information about various codes of ethics and ethical requirements for government-sponsored research: U.S. federal policy for the protection of human subjects, from the Office of Human Research Protections (OHRP): http://ohrp.osophs.dhhs.gov Canadian policies, from the Tri-Council Policy Statement of the Natural Sciences and Engineering Research Council of Canada (NSERC): http://www.nserc.ca/programs/ethics/english American Psychological Association: http://www.apa.org/ethics/code.html American Sociological Association: http://www.asanet.org/members/ecoderev.html THE PRINCIPLE OF BENEFICENCE One of the most fundamental ethical principles in research is that of beneficence, which encompasses the maxim: Above all, do no harm. Ethical Principle 2 of the ANA guidelines addresses beneficence. Most researchers consider that this principle contains multiple dimensions. Freedom From Harm Study participants can be harmed in a variety of ways, including harm that is physical (e.g., injury, fatigue), psychological (e.g., stress, fear), social (e.g., loss of friends), and economic (e.g., loss of wages). Researchers should strive to minimize all types of harm and discomfort and to achieve insofar as possible a balance between the potential benefits and risks of being a participant. Clearly, exposing study participants to experiences that result in serious or permanent harm is unacceptable. Research should be conducted only

144

PART 3 Designs for Nursing Research

BOX 7.1 Ethical Principles in Nursing Research THE INVESTIGATOR . . . 1. Respects autonomous research participants’ capacity to consent to participate in research and to determine the degree and duration of that participation without negative consequences. 2. Prevents harm, minimizes harm, and/or promotes good to all research participants, including vulnerable groups and others affected by the research. 3. Respects the personhood of research participants, their families, and significant others, valuing their diversity. 4. Ensures that the benefits and burdens of research are equitably distributed in the selection of research participants. 5. Protects the privacy of research participants to the maximum degree possible. 6. Ensures the ethical integrity of the research process by use of appropriate checks and balances throughout the conduct, dissemination, and implementation of the research. 7. Reports suspected, alleged, or known incidents of scientific misconduct in research to appropriate institutional officials for investigation. 8. Maintains competency in the subject matter and methodologies of his or her research, as well as in other professional and societal issues that affect nursing research and the public good. 9. Involved in animal research maximizes the benefits of the research with the least possible harm or suffering to the animals. From Silva, M. C. (1995). Ethical guidelines in the conduct, dissemination, and implementation of nursing research (pp. v–vi). Washington, DC: American Nurses’ Association.

by qualified people, especially if potentially dangerous technical equipment or specialized procedures are used. Ethical researchers must be prepared to terminate the research if there is reason to suspect that continuation would result in injury, death, disability, or undue distress to study participants. When a new medical procedure or drug is being tested, it is almost always advisable to experiment with animals or tissue cultures before proceeding to tests with humans. (Ethical guidelines relating to the treatment of animal subjects should be consulted for research on animals; see, for example, the American Psychological Association’s Guidelines for ethical conduct in the care and use of animals at http://www.apa.org/science/anguide.html.) Example of risk reduction: Varda and Behnke (2000) studied the effect of the timing of an initial bath (1 hour versus 2 hours after birth) on newborn temperature. To min-

imize risks, the researchers excluded all infants with conditions (e.g., infection, fetal distress, hypoglycemia) that could predispose them to temperature instability. Although protecting human beings from physical harm may be straightforward, the psychological consequences of participating in a study are usually subtle and thus require close attention and sensitivity. For example, participants may be asked questions about their personal views, weaknesses, or fears. Such queries might lead people to reveal sensitive personal information. The point is not that researchers should refrain from asking questions but rather that they need to be aware of the nature of the intrusion on people’s psyches. Researchers can avoid or minimize psychological harm by carefully phrasing questions, by having debriefing sessions that permit participants to ask questions or air complaints after data are collected, and, in some

CHAPTER 7 Designing Ethical Research

situations, by making referrals to appropriate health, social, or psychological services. Example of referrals: In the study by Polit, London, and Martinez (2001) of the health of nearly 4000 poor women in 4 major cities, the 90-minute interviews covered such sensitive topics as substance abuse, depression, parenting stress, and domestic violence. Each interviewer had an information sheet with contact information for local service providers who could assist with any issue about which a participant mentioned a need for help. The need for sensitivity may be greater in qualitative studies, which often involve in-depth exploration into highly personal areas. In-depth probing may actually expose deep-seated fears and anxieties that study participants had previously repressed. Qualitative researchers, regardless of the underlying research tradition, must thus be especially vigilant in anticipating such problems. Example of an issue of risk in a qualitative study: Caelli (2001) conducted a phenomenological study to illuminate nurses’ understandings of health, and how such understandings translated into nursing practice. One participant, having explored her experience of health with the researcher over several interview sessions, resigned from her city hospital job as a result of gaining a new recognition of the role health played in her life. Freedom From Exploitation Involvement in a research study should not place participants at a disadvantage or expose them to situations for which they have not been prepared. Participants need to be assured that their participation, or information they might provide, will not be used against them in any way. For example, a person describing his or her economic circumstances to a researcher should not be exposed to the risk of losing Medicaid benefits; a person reporting drug use should not fear exposure to criminal authorities. Study participants enter into a special relationship with researchers, and it is crucial that this

145

relationship not be exploited. Exploitation may be overt and malicious (e.g., sexual exploitation, use of subjects’ identifying information to create a mailing list, and use of donated blood for the development of a commercial product), but it might also be more subtle. For example, suppose subjects agreed to participate in a study requiring 30 minutes of their time and that the researcher decided 1 year later to go back to them, to follow their progress or circumstances. Unless the researcher had previously explained to participants that there might be a follow-up study, the researcher might be accused of not adhering to the agreement previously reached and of exploiting the researcher—participant relationship. Because nurse researchers may have a nurse— patient (in addition to a researcher—participant) relationship, special care may need to be exercised to avoid exploiting that bond. Patients’ consent to participate in a study may result from their understanding of the researcher’s role as nurse, not as researcher. In qualitative research, the risk of exploitation may become especially acute because the psychological distance between investigators and participants typically declines as the study progresses. The emergence of a pseudotherapeutic relationship is not uncommon, which imposes additional responsibilities on researchers—and additional risks that exploitation could inadvertently occur. On the other hand, qualitative researchers typically are in a better position than quantitative researchers to do good, rather than just to avoid doing harm, because of the close relationships they often develop with participants. Munhall (2001) has argued that qualitative nurse researchers have the responsibility of ensuring that the “therapeutic imperative of nursing (advocacy) takes precedent over the research imperative (advancing knowledge) if conflict develops” (p. 538). Benefits From Research People agree to participate in research investigations for a number of reasons. They may perceive that there are some direct personal benefits. More

146

PART 3 Designs for Nursing Research

often, however, any benefits from the research accrue to society in general or to other individuals. Thus, many individuals may participate in a study out of a desire to be helpful. Researchers should strive insofar as possible to maximize benefits and to communicate potential benefits to participants. The Risk/Benefit Ratio In designing a study, researchers must carefully assess the risks and benefits that would be incurred. The assessment of risks and benefits that individual participants might experience should be shared with them so that they can evaluate whether it is in their best interest to participate. Box 7-2 summarizes the major risks and benefits of research participation. In evaluating the anticipated risk/benefit ratio of a study design, researchers might want to consider how comfortable they would feel if their own family members were participating in the study.

The risk/benefit ratio should also be considered in terms of whether the risks to participants are commensurate with the benefit to society and the nursing profession in terms of the quality of evidence produced. The general guideline is that the degree of risk to be taken by those participating in the research should never exceed the potential humanitarian benefits of the knowledge to be gained. Thus, the selection of a significant topic that has the potential to improve patient care is the first step in ensuring that research is ethical. All research involves some risks, but in many cases, the risk is minimal. Minimal risk is defined as risks anticipated to be no greater than those ordinarily encountered in daily life or during routine physical or psychological tests or procedures. When the risks are not minimal, researchers must proceed with caution, taking every step possible to reduce risks and maximize benefits. If the perceived

BOX 7.2 Potential Benefits and Risks of Research to Participants MAJOR POTENTIAL BENEFITS TO PARTICIPANTS • Access to an intervention that might otherwise be unavailable to them • Comfort in being able to discuss their situation or problem with a friendly, objective person • Increased knowledge about themselves or their conditions, either through opportunity for introspection and self-reflection or through direct interaction with researchers • Escape from normal routine, excitement of being part of a study • Satisfaction that information they provide may help others with similar problems or conditions • Direct monetary or material gains through stipends or other incentives MAJOR POTENTIAL RISKS TO PARTICIPANTS • Physical harm, including unanticipated side effects • Physical discomfort, fatigue, or boredom • Psychological or emotional distress resulting from self-disclosure, introspection, fear of the unknown, discomfort with strangers, fear of eventual repercussions, anger or embarrassment at the type of questions being asked • Social risks, such as the risk of stigma, adverse effects on personal relationships, loss of status • Loss of privacy • Loss of time • Monetary costs (e.g., for transportation, child care, time lost from work)

CHAPTER 7 Designing Ethical Research

risks and costs to participants outweigh the anticipated benefits of the study, the research should be either abandoned or redesigned. In quantitative studies, most of the details of the study are usually spelled out in advance, and therefore a reasonably accurate risk/benefit ratio assessment can be developed. Qualitative studies, however, usually evolve as data are gathered, and it may therefore be more difficult to assess all risks at the outset of a study. Qualitative researchers thus must remain sensitive to potential risks throughout the research process. THE PRINCIPLE OF RESPECT FOR HUMAN DIGNITY Respect for human dignity is the second ethical principle articulated in the Belmont Report. This principle, which includes the right to self-determination and the right to full disclosure, is covered in the ANA guidelines under principles 1 and 3. The Right to Self-Determination Humans should be treated as autonomous agents, capable of controlling their own activities. The principle of self-determination means that prospective participants have the right to decide voluntarily whether to participate in a study, without risking any penalty or prejudicial treatment. It also means that people have the right to ask questions, to refuse to give information, to ask for clarification, or to terminate their participation. A person’s right to self-determination includes freedom from coercion of any type. Coercion involves explicit or implicit threats of penalty from failing to participate in a study or excessive rewards from agreeing to participate. The obligation to protect people from coercion requires careful thought when the researcher is in a position of authority, control, or influence over potential participants, as might often be the case in a nurse—patient relationship. The issue of coercion may require scrutiny even when there is not a preestablished relationship. For example, a generous monetary incentive (or

147

stipend) offered to encourage the participation of an economically disadvantaged group (e.g., the homeless) might be considered mildly coercive because such incentives may place undue pressure on prospective participants; its acceptability might have to be evaluated in terms of the overall risk/benefit ratio. TIP: Stipends used to increase the rate of participation in a study appear to be especially effective when the group under study is difficult to recruit or when the study is time-consuming or tedious. Stipends range from $1 to hundreds of dollars, but most are in the $10 to $25 range. Federal agencies that sponsor research sometimes do not allow the payment of an outright stipend but will allow reimbursement of certain expenses (e.g., for participants’ travel, child care, or lunch money). The Right to Full Disclosure The principle of respect for human dignity encompasses people’s right to make informed, voluntary decisions about study participation, which requires full disclosure. Full disclosure means that the researcher has fully described the nature of the study, the person’s right to refuse participation, the researcher’s responsibilities, and likely risks and benefits. The right to self-determination and the right to full disclosure are the two major elements on which informed consent is based. Procedures for obtaining informed consent from participants are discussed later in this chapter. Although full disclosure is normally provided to participants before they begin a study, there is often a need for further disclosure at a later point, either in debriefing sessions or in written communications. For example, issues that arise during the course of data collection may need to be clarified, or participants may want aspects of the study explained once again. Some researchers offer to send participants summaries of the research findings after the information has been analyzed. In qualitative studies, the consent process may require an ongoing negotiation between researchers and participants.

148

PART 3 Designs for Nursing Research

Issues Relating to the Principle of Respect Although most researchers would, in the abstract, endorse participants’ right to self-determination and full disclosure, these standards are sometimes difficult to adhere to in practice. One issue concerns the inability of some individuals to make well-informed judgments about the risks and benefits of study participation. Children, for example, may be unable to give truly informed consent. The issue of groups that are vulnerable within a research context is discussed later in this chapter. Another issue is that full disclosure can sometimes create two types of bias: first, a bias resulting if subjects provide inaccurate information, and second, a bias resulting if a representative sample is not recruited. Suppose we were studying the relationship between high school students’ substance abuse and their absenteeism from school; we hypothesize that students with a high rate of absenteeism are more likely to be substance abusers than students with a good attendance record. If we approached potential participants and fully explained the purpose of the study, some students might refuse to participate, and nonparticipation would be selective; those least likely to volunteer for such a study might well be students who are substance abusers—the very group of primary interest. Moreover, by knowing the research question, those who do participate might not give candid responses. In such a situation, full disclosure could undermine the study. One technique that researchers sometimes use in such situations is covert data collection or concealment—the collection of information without participants’ knowledge and thus without their consent. This might happen, for example, if a researcher wanted to observe people’s behavior in a real-world setting and was concerned that doing so openly would result in changes in the very behavior of interest. The researcher might choose to obtain the information through concealed methods, such as by observing through a one-way mirror, videotaping with hidden equipment, or observing while pretending to be engaged in other

activities. As another example, hospital patients might unwittingly become participants in a study through researchers’ use of existing hospital records. In general, covert data collection may be acceptable as long as risks are negligible and participants’ right to privacy has not been violated, and if the researcher has arranged to debrief participants about the nature of the study subsequent to data collection. Covert data collection is least likely to be ethically acceptable if the research is focused on sensitive aspects of people’s behavior, such as drug use, sexual conduct, or illegal acts. A more controversial technique is the use of deception. Deception can involve deliberately withholding information about the study, or providing participants with false information. For example, in studying high school students’ use of drugs we might describe the research as a study of students’ health practices, which is a mild form of misinformation. The practice of deception is problematic ethically because it interferes with participants’ right to make a truly informed decision about personal costs and benefits of participation. Some people argue that deception is never justified. Others, however, believe that if the study involves minimal risk to subjects and if there are anticipated benefits to the profession and society, then deception may be justified to enhance the validity of the findings. The ANA guidelines offer this advice about deception and concealment: The investigator understands that concealment or deception in research is controversial, depending on the type of research. Some investigators believe that concealment or deception in research can never be morally justified. The investigator further understands that before concealment or deception is used, certain criteria must be met: (1) The study must be of such small risk to the research participant and of such great significance to the advancement of the public good that concealment or deception can be morally justified. . . . (2) The acceptability of concealment or deception is related to the degree of risks to research participants. . . . (3) Concealment or deception are used only as last resorts, when no other approach can ensure the validity of the study’s findings. . . . (4) The investigator has a moral responsibility to inform research

CHAPTER 7 Designing Ethical Research participants of any concealment or deception as soon as possible and to explain the rationale for its use. (Silva, 1995, p. 10, Section 4.2).

Another issue relating to the principle of respect that has emerged in this new era of electronic communications concerns the collection of data from people over the Internet. For example, some researchers are analyzing the content of messages posted to chat rooms or on listserves. The issue is whether such messages can be used as data without the authors’ permission and their informed consent. Some researchers believe that anything posted electronically is in the public domain and therefore can be used without consent for purposes of research. Others, however, feel that the same ethical standards must apply in cyberspace research and that electronic researchers must carefully protect the rights of individuals who are participants in “virtual” communities. Schrum (1995) has developed some ethical guidelines for use by such researchers. As one example, she advocates that researchers, before collecting electronic data, negotiate their entry into an electronic community (e.g., a chat room) with the list owner. Sixsmith and Murray (2001) also warn researchers that obtaining consent from list moderators does not necessarily mean that every member of the listserve or chat room has provided consent. Researchers should periodically remind members of the on-line group of their presence at the site. THE PRINCIPLE OF JUSTICE The third broad principle articulated in the Belmont Report concerns justice. Justice, which includes participants’ right to fair treatment and their right to privacy, is covered in the ANA guidelines under principles 4 and 5. The Right to Fair Treatment Study participants have the right to fair and equitable treatment before, during, and after their participation in the study. Fair treatment includes the following features:

149

• The fair and nondiscriminatory selection of participants such that any risks and benefits will be equitably shared; participants should be selected based on research requirements and not on the vulnerability or compromised position of certain people • Respect for cultural and other forms of human diversity • The nonprejudicial treatment of those who decline to participate or who withdraw from the study after agreeing to participate • The honoring of all agreements between researchers and participants, including adherence to the procedures described to them and payment of any promised stipends • Participants’ access to research personnel at any point in the study to clarify information • Participants’ access to appropriate professional assistance if there is any physical or psychological damage • Debriefing, if necessary, to divulge information withheld before the study or to clarify issues that arose during the study • Courteous and tactful treatment at all times The Right to Privacy Virtually all research with humans involves intruding into personal lives. Researchers should ensure that their research is not more intrusive than it needs to be and that participants’ privacy is maintained throughout the study. Participants have the right to expect that any data they provide will be kept in strictest confidence. This can occur either through anonymity or through other confidentiality procedures. Anonymity occurs when even the researcher cannot link participants to their data. For example, if questionnaires were distributed to a group of nursing home residents and were returned without any identifying information on them, responses would be anonymous. As another example, if a researcher reviewed hospital records from which all identifying information (e.g., name, address, social security number, and so forth) had been

150

PART 3 Designs for Nursing Research

expunged, anonymity would again protect participants’ right to privacy. Whenever it is possible to achieve anonymity, researchers should strive to do so. Example of anonymity: Thomas, Stamler, Lafrenier, and Dumala (2001) used the Internet to gather data from an international sample of women about their perceptions of breast health education and screening. A website with a questionnaire was established. No identifying information was sought from respondents, and so their anonymity was guaranteed. When anonymity is impossible, appropriate confidentiality procedures need to be implemented. A promise of confidentiality is a pledge that any information participants provide will not be publicly reported in a manner that identifies them and will not be made accessible to others. This means that research information should not be shared with strangers nor with people known to the participants (e.g., family members, physicians, other nurses), unless the researcher has been given explicit permission to share it. Researchers can take a number of steps to ensure that breaches of confidentiality do not occur, including the following: • Obtain identifying information (e.g., name, address) from participants only when essential. • Assign an identification (ID) number to each participant and attach the ID number rather than other identifiers to the actual data. • Maintain identifying information in a locked file. • Restrict access to identifying information to a small number of people on a need-to-know basis. • Enter no identifying information onto computer files. • Destroy identifying information as quickly as practical. • Make research personnel sign confidentiality pledges if they have access to data or identifying information. • Report research information in the aggregate; if information for a specific participant is reported, take steps to disguise the person’s identity, such as through the use of a fictitious name.

TIP: Researchers who plan to collect data from study participants on multiple occasions (or who use multiple data forms that need to be connected) might believe that anonymity is not possible. However, a technique that has been successfully used is to have participants themselves generate an ID number. They might be instructed, for example, to use their birth year and the first three letters of their mother’s maiden names as their ID code (e.g., 1946CRU). This code would be put on every form participants complete, but researchers would not know participants’ identities. Qualitative researchers may need to take extra steps to safeguard the privacy of their participants. Anonymity is almost never possible in qualitative studies because researchers typically become closely involved with participants. Moreover, because of the in-depth nature of qualitative studies, there may be a greater invasion of privacy than is true in quantitative research. Researchers who spend time in the home of a participant may, for example, have difficulty segregating the public behaviors that the participant is willing to share from the private behaviors that unfold unwittingly during the course of data collection. A final issue is adequately disguising participants in research reports. Because the number of respondents is small, qualitative researchers may need to take considerable precautions to safeguard identities. This may mean more than simply using a fictitious name—it may also mean not sharing detailed information about informants’ characteristics, such as their occupation and diagnosis. TIP: Qualitative researchers may have to slightly distort identifying information in their reports, or provide fairly general descriptions. For example, a 49-year-old antique dealer with ovarian cancer might be described as “a middleaged cancer patient who works in retail sales” to avoid identification that could occur with the more detailed description. INFORMED CONSENT Prospective participants who are fully informed about the nature of the research and its potential risks and benefits are in a position to make rational

CHAPTER 7 Designing Ethical Research

decisions about participating in the study. Informed consent means that participants have adequate information regarding the research, are capable of comprehending the information, and have the power of free choice, enabling them to consent to or decline participation voluntarily. This section discusses procedures for obtaining informed consent. The Content of Informed Consent Fully informed consent involves communicating the following pieces of information to participants: 1. Participant status. Prospective participants need to understand clearly the distinction between research and treatment. They should be told which health care activities are routine and which are implemented specifically for the study. They also should be informed that data they provide will be used for research purposes. 2. Study goals. The overall goals of the research should be stated, in lay rather than technical terms. The use to which the data will be put should be described. 3. Type of data. Prospective participants should be told the type of data that will be collected. 4. Procedures. Prospective participants should be given a description of the data collection procedures, and of the procedures to be used in any innovative treatment. 5. Nature of the commitment. Information should be provided regarding participants’ estimated time commitment at each point of contact, and the number of contacts within specified timeframes. 6. Sponsorship. Information on who is sponsoring or funding the study should be noted; if the research is part of an academic requirement, this information should be shared. 7. Participant selection. Researchers should explain how prospective participants were selected for recruitment, and how many people will be participating. 8. Potential risks. Prospective participants should be informed of any foreseeable risks (physical, psychological, social, or economic) or discomforts that might be incurred as a

151

result of participation, and any efforts that will be taken to minimize risks. The possibility of unforeseeable risks should also be discussed, if appropriate. If injury or damage is possible, treatments that will be made available to participants should be described. When risks are more than minimal, prospective participants should be encouraged to seek the advice of others before consenting. 9. Potential benefits. Specific benefits to participants, if any, should be described, as well as information on possible benefits to others. 10. Alternatives. If appropriate, researchers should provide information about alternative procedures or treatments that might be advantageous to participants. 11. Compensation. If stipends or reimbursements are to be paid (or if treatments are offered without fee), these arrangements should be discussed. 12. Confidentiality pledge. Prospective participants should be assured that their privacy will at all times be protected. If anonymity can be guaranteed, this should be noted. 13. Voluntary consent. Researchers should indicate that participation is strictly voluntary and that failure to volunteer will not result in any penalty or loss of benefits. 14. Right to withdraw and withhold information. Prospective participants should be told that even after consenting they have the right to withdraw from the study and to refuse to provide any specific piece of information. Researchers may, in some cases, need to provide participants with a description of circumstances under which researchers would terminate the overall study. 15. Contact information. The researcher should provide information on whom participants could contact in the event of further questions, comments, or complaints. In some qualitative studies, especially those requiring repeated contact with the same participants, it is difficult to obtain a meaningful informed consent at the outset. Qualitative researchers do not always know in advance how the study will evolve.

152

PART 3 Designs for Nursing Research

Because the research design emerges during the data collection and analysis process, researchers may not know the exact nature of the data to be collected, what the risks and benefits to participants will be, or how much of a time commitment they will be expected to make. Thus, in a qualitative study, consent is often viewed as an ongoing, transactional process, referred to as process consent. In process consent, the researcher continually renegotiates the consent, allowing participants to play a collaborative role in the decision-making process regarding ongoing participation. Example of informed consent: Wilde (2002) studied the experience of living with a long-term urinary catheter in a communitydwelling sample of adults. Fourteen men and women were recruited for this phenomenological study. Full informed consent was obtained before each interview, and reaffirmed as interviews continued. Comprehension of Informed Consent Consent information is normally presented to prospective participants while they are being recruited, either orally or in writing. A written notice should not, however, take the place of spoken explanations. Oral presentations provide opportunities for greater elaboration and for participant questioning. Because informed consent is based on a person’s evaluation of the potential risks and benefits of participation, it is important that the critical information not only be communicated but understood. Researchers must assume the role of teacher in communicating consent information. They should be careful to use simple language and to avoid jargon and technical terms whenever possible; they should also avoid biased language that might unduly influence the person’s decision to participate. Written statements should be consistent with the participants’ reading levels and educational attainment. For participants from a general population (e.g., patients in a hospital), the statement should be written at about seventh or eighth grade reading level. For studies involving more than minimal risk, researchers need to make special efforts to ensure

that prospective participants understand what participation will involve. In some cases, this might involve testing participants for their comprehension of the informed consent material before deeming them eligible for participation. Documentation of Informed Consent Researchers usually document the informed consent process by having participants sign a consent form. In the United States, federal regulations covering studies funded by government agencies require written consent of human subjects, except under certain circumstances. In particular, when the study does not involve an intervention and data are collected anonymously (or when existing data from records or specimens are used and identifying information is not linked to the data), regulations requiring written informed consent do not apply. The consent form should contain all the information essential to informed consent, as described earlier. Prospective participants (or their legally authorized representative) should have ample time to review the written document before signing it. The document should also be signed by the researcher, and a copy should be retained by both parties. An example of a written consent form used in a study of one of the authors is presented in Figure 7-1. The numbers in the margins correspond to the types of information for informed consent outlined earlier. (Note that the form does not indicate how subjects were selected, because this is implied in the study purpose, and prospective participants knew they were recruited from a support group for mothers of multiples.) TIP: In developing a consent form, the following guidelines might prove helpful: 1. Organize the form coherently so that prospective participants can follow the logic of what is being communicated. If the form is complex, use headings as an organizational aid. 2. Use a large enough font so that the form can be easily read, and use spacing that avoids making the document appear too dense. Make the form as attractive and inviting as possible.

CHAPTER 7 Designing Ethical Research

F I G U R E 7 . 1 Example of an informed consent form.

153

154

PART 3 Designs for Nursing Research

3. In general, simplify. Use clear and consistent terminology, and avoid technical terms if possible. If technical terms are needed, include definitions. 4. If possible, use a readability formula to estimate the form’s reading level, and make revisions to ensure an appropriate reading level for the group under study. There are several such formulas, the most widely used being the FOG Index (Gunning, 1968), the SMOG index (McLaughlin, 1969), and the Flesch Reading Ease score and Flesch-Kincaid grade level score (Flesch, 1948). Specialized software (e.g., RightWriter) is available, and some wordprocessing software (e.g., Microsoft Word) also provides readability information. 5. Test the form with people similar to those who will be recruited, and ask for feedback. If the informed consent information is lengthy, researchers whose studies are funded by U.S. government agencies have the option of presenting the full information orally and then summarizing essential information in a short form. If a short form is used, however, the oral presentation must be witnessed by a third party, and the signature of the witness must appear on the short consent form. The signature of a third-party witness is also advisable in studies involving more than minimal risk, even when a long and comprehensive consent form is used. For studies that are not government sponsored, researchers should err on the side of being conservative. They should implement consent procedures that fully adhere to the principle that prospective participants can make good decisions about participation only if they are fully informed about the study’s risks and benefits. TIP: When the primary means of data collection is through a self-administered questionnaire, some researchers opt not to obtain written informed consent because they assume implied consent (i.e., that the return of the completed questionnaire reflects voluntary consent to participate). This assumption, however, may not always be warranted (e.g., if patients feel that their treatment might be affected by failure to cooperate with the researcher).

VULNERABLE SUBJECTS Adherence to ethical standards is often straightforward. However, the rights of special vulnerable groups may need to be protected through additional procedures and heightened sensitivity. Vulnerable subjects (the term used in U.S. federal guidelines) may be incapable of giving fully informed consent (e.g., mentally retarded people) or may be at high risk of unintended side effects because of their circumstances (e.g., pregnant women). Researchers interested in studying high-risk groups should become acquainted with guidelines governing informed consent, risk/benefit assessments, and acceptable research procedures for such groups. In general, research with vulnerable subjects should be undertaken only when the risk/benefit ratio is low or when there is no alternative (e.g., studies of childhood development require child participants). Among the groups that nurse researchers should consider as being vulnerable are the following: • Children. Legally and ethically, children do not have the competence to give informed consent. Usually, the informed consent of children’s parents or legal guardians should be obtained. However, it is appropriate—especially if the child is at least 7 years of age—to obtain the child’s assent as well. Assent refers to the child’s affirmative agreement to participate. If the child is developmentally mature enough to understand the basic information involved in informed consent (e.g., a 13-yearold), it is advisable to obtain written consent from the child as well, as evidence of respect for the child’s right to self-determination. Lindeke, Hauck, and Tanner (2000) and Broome (1999) provide excellent guidance regarding children’s assent and consent to participate in research. The U.S. government has issued special regulations for the additional protection of children as study participants (see Code of Federal Regulations, 1991, Subpart D). • Mentally or emotionally disabled people. Individuals whose disability makes it impossible for them to weigh the risks and benefits of participation and make an informed decision (e.g.,

CHAPTER 7 Designing Ethical Research

people affected by mental retardation, senility, mental illness, or unconsciousness) also cannot legally or ethically provide informed consent. In such cases, researchers should obtain the written consent of a legal guardian. Researchers should, however, be aware of the fact that a legal guardian may not necessarily have the person’s best interests in mind. In such cases, informed consent should also be obtained from someone whose primary interest is the person’s welfare. As in the case of children, informed consent or assent from prospective participants themselves should be sought to the extent possible, in addition to guardians’ consent. • Severely ill or physically disabled people. For patients who are very ill or undergoing certain treatments, it might be necessary to assess their ability to make reasoned decisions about study participation. For example, Higgins and Daly (1999) described a process they used to assess the decisional capacity of mechanically ventilated patients. Another issue is that for certain disabilities, special procedures for obtaining consent may be required. For example, with deaf participants, the entire consent process may need to be in writing. For people who have a physical impairment preventing them from writing or for participants who cannot read and write, alternative procedures for documenting informed consent (such as audiotaping or videotaping consent proceedings) should be used. • The terminally ill. Terminally ill people who participate in the study can seldom expect to benefit personally from the research, and thus the risk/benefit ratio needs to be carefully assessed. Researchers must also take steps to ensure that if the terminally ill participate in the study, the health care and comfort of these individuals are not compromised. Special procedures may be required for obtaining informed consent if they are physically or mentally incapacitated. • Institutionalized people. Nurses often conduct studies with hospitalized or institutionalized people. Particular care may be required in recruiting such people because they often depend on health care personnel and may feel

155

pressured into participating or may feel that their treatment would be jeopardized by their failure to cooperate. Inmates of prisons and other correctional facilities, who have lost their autonomy in many spheres of activity, may similarly feel constrained in their ability to give free consent. The U.S. government has issued specific regulations for the protection of prisoners as study participants (see Code of Federal Regulations, 1991, Subpart C). Researchers studying institutionalized groups need to emphasize the voluntary nature of participation. • Pregnant women. The U.S. government has issued stringent additional requirements governing research with pregnant women and fetuses (Code of Federal Regulations, 1991, Subpart B). These requirements reflect a desire to safeguard both the pregnant woman, who may be at heightened physical and psychological risk, and the fetus, who cannot give informed consent. The regulations stipulate that a pregnant woman cannot be involved in a study unless the purpose of the research is to meet the health needs of the pregnant woman and risks to her and the fetus are minimized or there is only a minimal risk to the fetus. Example of research with a vulnerable group: Anderson, Nyamathi, McAvoy, Conde, and Casey (2001) conducted a study to explore perceptions of risk for human immunodeficiency virus infection/acquired immunodeficiency syndrome among adolescents in juvenile detention. The researchers obtained approval to conduct the study from the presiding judge, the detention facility, and a human subjects committee at their own institution. They structured their protocols to assure teens that their participation would be voluntary and would influence neither the duration of their detention nor their adjudication process. The data were collected in spaces that provided privacy for sound and afforded visual surveillance by probation staff. It should go without saying that researchers need to proceed with extreme caution in conducting research with people who might fall into two or more vulnerable categories, as was the case in this example.

156

PART 3 Designs for Nursing Research

EXTERNAL REVIEWS AND THE PROTECTION OF HUMAN RIGHTS Researchers may not be objective in assessing risk/benefit ratios or in developing procedures to protect participants’ rights. Biases may arise as a result of the researchers’ commitment to an area of knowledge and their desire to conduct a study with as much rigor as possible. Because of the risk of a biased evaluation, the ethical dimensions of a study should normally be subjected to external review. Most hospitals, universities, and other institutions where research is conducted have established formal committees and protocols for reviewing proposed research plans before they are implemented. These committees are sometimes called human subjects committees, ethical advisory boards, or research ethics committees. If the institution receives funds from the U.S. government to help pay for the costs of research, the committee likely will be called an Institutional Review Board (IRB). TIP: If the research is being conducted within an institution or with its help (e.g., assistance in recruiting subjects), you should find out early what the institution’s requirements are regarding ethical issues, in terms of its forms, procedures, and review schedules. Federally sponsored studies (including fellowships) are subject to strict guidelines for evaluating the treatment of human participants. Before undertaking such a study, researchers must submit research plans to the IRB, and must also go through a formal IRB training process. The duty of the IRB is to ensure that the proposed plans meet the federal requirements for ethical research. An IRB can approve the proposed plans, require modifications, or disapprove the plans. The main requirements governing IRB decisions may be summarized as follows (Code of Federal Regulations, 1991, §46.111): • Risks to participants are minimized. • Risks to participants are reasonable in relation to anticipated benefits, if any, and the impor-

tance of the knowledge that may reasonably be expected to result. • Selection of participants is equitable. • Informed consent will be sought, as required. • Informed consent will be appropriately documented. • Adequate provision is made for monitoring the research to ensure participants’ safety. • Appropriate provisions are made to protect participants’ privacy and confidentiality of the data. • When vulnerable subjects are involved, appropriate additional safeguards are included to protect their rights and welfare. Example of IRB approval: Jones, Bond, Gardner, and Hernandez (2002) studied the family planning patterns of immigrant Hispanic women in relation to their acculturation to American culture. The researchers sought and obtained approval for the study from the IRB of both a university and a medical center. Many research projects require a full IRB review. For a full review, the IRB convenes meetings at which most IRB members are present. An IRB must have five or more members, at least one of whom is not a researcher (e.g., a member of the clergy or a lawyer may be appropriate). One IRB member must be a person who is not affiliated with the institution and is not a family member of a person who is affiliated. To protect against potential biases, the IRB cannot comprise entirely men, women, or members from a single profession. For certain kinds of research involving no more than minimal risk, the IRB can use expedited review procedures, which do not require a meeting. In an expedited review, a single IRB member (usually the IRB chairperson or a member designated by the chairperson) carries out the review. Examples of research activities that qualify for an expedited IRB review, if they are deemed to be minimal-risk, include (1) the collection of blood samples in amounts not exceeding 550 ml in an 8-week period, from healthy, nonpregnant adults weighing at least 110 pounds; and (2) research on individual or group characteristics or behavior or “research employing survey, interview, focus group, program evaluation,

CHAPTER 7 Designing Ethical Research

human factors evaluation, or quality assurance methodologies” (Federal Register notice cited in Code of Federal Regulations, 1991, §46.110). The federal regulations also allow certain types of research to be totally exempt from IRB review. These are studies in which there are no apparent risks to human participants. The website of the Office of Human Research Protections, in its policy guidance section, includes decision charts designed to clarify whether a study is exempt from the federal regulations. TIP: Not all research is subject to federal guidelines, and so not all studies are reviewed by formal committees. Nevertheless, researchers must ensure that their research plans are ethically sound and are encouraged to seek outside advice on the ethical dimensions of a study before it gets underway. Advisers might include faculty members, the clergy, representatives from the group being asked to participate, or advocates for that group.

Research design:

Intervention:

Sample:

BUILDING ETHICS INTO THE DESIGN OF THE STUDY Researchers need to give careful thought to ethical requirements during the planning of a research project and to ask themselves continually whether planned safeguards for protecting humans are sufficient. They must persist in being vigilant throughout the implementation of the research plans as well, because unforeseen ethical dilemmas may arise during the conduct of the study. Of course, a first step in doing ethical research is to scrutinize the research questions to determine whether they are clinically significant and whether it is feasible to undertake the study in a manner that conforms to ethical guidelines. The remaining chapters of the book offer advice on how to design studies that yield highquality evidence for practice. Methodologic decisions about rigor, however, must factor in ethical considerations. Here are some examples of the kinds of questions that might be posed in thinking about various aspects of study design:

Data collection:

Reporting:

157

• Will participants get allocated to different treatment groups fairly? • Will research controls add to the risks participants will incur? • Will the setting for the study be selected to protect against participant discomfort? • Is the intervention designed to maximize good and minimize harm? • Under what conditions might a treatment be withdrawn or altered? • Is the population defined so as to unwittingly and unnecessarily exclude important segments of people (e.g., women, minorities)? • Is the population defined in such a way that especially high-risk people (e.g., unstable patients) could be excused from the study? • Will potential participants be recruited into the study equitably? • Will data be collected in such a way as to minimize respondent burden? • Will procedures for ensuring confidentiality of data be adequate? • Will data collection staff be appropriately trained to be sensitive and courteous? • Will participants’ identities be adequately protected?

TIP: As a means of enhancing both individual and institutional privacy, research reports frequently avoid giving explicit information about the locale of the study. For example, the report might say that data were collected in a

158

PART 3 Designs for Nursing Research

200-bed, private, for-profit nursing home, without mentioning its name or location. Once the study procedures have been developed, researchers should undertake a self-evaluation of those procedures to determine if they meet ethical requirements. Box 26-15 in Chapter 26 provides some guidelines for such a self-evaluation. RESEARCH EXAMPLES Because researchers usually attempt to report research results as succinctly as possible, they rarely describe in much detail the efforts they have made to safeguard participants’ rights. (The absence of any mention of such safeguards does not, of course, imply that no precautions were taken.) Researchers are especially likely to discuss their adherence to ethical guidelines for studies that involve more than minimal risk or when the people being studied are a vulnerable group. Two research examples that highlight ethical issues are presented in the following sections. Research Example from a Quantitative Study Willson, McFarlane, Lemmy, and Malecha (2001) conducted a study to evaluate whether abused women’s use of the police reduced further violence. The study sought to describe the extent of violence and homicide danger experienced by women before and after filing assault charges against an intimate through the police department. After obtaining approval for conducting the study from the agency (a special family violence unit in a large metropolitan police department) and the researchers’ IRB, the researchers sought to interview a consecutive sample of women who met study criteria (18 years of age or older, English speaking) and who attempted to file assault charges during a 1month period in 1998. Investigators approached prospective participants and explained the study purpose, research protocols, administration time, and follow-up schedules. Women were paid a $20 stipend for each completed interview. Both verbal and written consent was obtained from a sample of 90 women.

The researchers took care to protect the women’s rights during data collection. Data were obtained confidentially in private interview rooms. The women were assigned an ID number to maintain confidentiality. The subjects’ safety was ensured for follow-up interviews by establishing a convenient, private, and safe time for the 3- and 6-month follow-up interviews. A total of 83 women completed all three rounds of interviews. The researchers found that women seeking police help had significantly reduced threats of abuse, actual experiences of abuse, and perceived danger of being killed than before.

Research Example From a Qualitative Study Wackerbarth (1999) undertook an in-depth study designed to describe the dynamics of caretaker decision making. Wackerbarth’s study focused on understanding the decision process among family caregivers of persons with dementia. A local chapter of the Alzheimer’s Association mailed out 100 preinterview questionnaires with an introductory letter from the director of the chapter. Caregivers interested in participating in the study mailed back a completed consent form and the preinterview questionnaire. From the pool of 80 caregivers who returned the questionnaire, 28 were selected to be interviewed. The sample was carefully selected to represent a broad viewpoint for developing a decision-making model. Wackerbarth’s article carefully explained the attention that was paid to participants’ rights in this study: (1) the study objectives and methods were described orally and in writing to ensure that they were understood; (2) an informed consent form, which highlighted the voluntary nature of participation and indicated the safeguards that would be taken to protect their confidentiality, was signed before data collection began; (3) all preinterview questionnaires, tape recordings, and interview transcripts were kept in a locked file cabinet; (4) no identifying information was appended to study materials; and (5) participants were asked to review written materials and to give permission before publication of quotes and study findings. On the basis of the interviews, Wackerbarth developed a model charting the caregiving experience over time, and documented decisions made to maintain tolerable situations. The model captured the intrapersonal

CHAPTER 7 Designing Ethical Research

struggle driving the decision-making efforts of caregivers who care for family members with dementia.

S U M M A RY P O I N T S • Because research has not always been conducted ethically, and because of the genuine ethical dilemmas researchers often face in designing studies that are both ethical and methodologically rigorous, codes of ethics have been developed to guide researchers. • The three major ethical principles incorporated into most guidelines are beneficence, respect for human dignity, and justice. • Beneficence involves the protection of participants from physical and psychological harm, protection of participants from exploitation, and the performance of some good. • In deciding to conduct a study, researchers must carefully weigh the risk/benefit ratio of participation to individuals and also the risks to participants against potential benefits to society. • Respect for human dignity involves the participants’ right to self-determination, which means participants have the freedom to control their own activities, including study participation. • Respect for human dignity also encompasses the right to full disclosure, which means that researchers have fully described to prospective participants their rights and the full nature of the study. When full disclosure poses the risk of biased results, researchers sometimes use covert data collection or concealment (the collection of information without the participants’ knowledge or consent) or deception (either withholding information from participants or providing false information). If deception or concealment is deemed necessary, extra precautions should be used to minimize risks and protect other rights. • Justice includes the right to fair treatment (both in the selection of participants and during the course of the study) and the right to privacy. Privacy can be maintained through anonymity (wherein not even researchers know participants’ identities) or through formal confidentiality pro-

159

cedures that safeguard the information participants provide. • Informed consent procedures, which provide prospective participants with information needed to make a reasoned decision about participation, normally involve signing a consent form to document voluntary and informed participation. In qualitative studies, consent may need to be continually renegotiated with participants as the study evolves, through process consent. • Vulnerable subjects require additional protection. These people may be vulnerable because they are not able to make a truly informed decision about study participation (e.g., children); because their circumstances make them believe free choice is constrained (e.g., prisoners); or because their circumstances heighten the risk of physical or psychological harm (e.g., pregnant women, the terminally ill). • External review of the ethical aspects of a study by a human subjects committee or Institutional Review Board (IRB) is highly desirable and may be required by either the agency funding the research or the organization from which participants are recruited. • In studies in which risks to participants are minimal, an expedited review (review by a single member of the IRB) may be substituted for a full board review; in cases in which there are no anticipated risks, the research may be exempted from review. • Researchers are always advised, even in the absence of an IRB review, to consult with at least one external adviser whose perspective allows an objective evaluation of the ethics of a proposed study. • Researchers need to give careful thought to ethical requirements throughout the study’s planning and implementation and to ask themselves continually whether safeguards for protecting humans are sufficient. STUDY ACTIVITIES Chapter 7 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers

160

PART 3 Designs for Nursing Research

various exercises and study suggestions for reinforcing concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Point out the ethical dilemmas that might emerge in the following studies: a. A study of the relationship between sleeping patterns and acting-out behaviors in hospitalized psychiatric patients b. A study of the effects of a new drug treatment for diabetic patients c. An investigation of an individual’s psychological state after an abortion d. An investigation of the contraceptive decisions of high school students at a schoolbased clinic 2. For each of the studies described in question 1, indicate whether you think the study would require a full IRB review or an expedited review, or whether it would be totally exempt from review. 3. For the study described in the research example section (Willson et al., 2001), prepare an informed consent form that includes required information, as described in the section on informed consent. SUGGESTED READINGS References on Research Ethics American Nurses’ Association. (1975). Human rights guidelines for nurses in clinical and other research. Kansas City, MO: Author. American Nurses’ Association. (1985). Code for nurses with interpretive statements. Kansas City, MO: Author. American Psychological Association. (1992). Ethical principles of psychologists and code of conduct. Washington, DC: Author. American Sociological Association. (1997). Code of ethics. Washington, DC: Author. Broome, M. E. (1999). Consent (assent) for research with pediatric patients. Seminars in Oncology Nursing, 15, 96–103. Code of Federal Regulations. (1991). Protection of human subjects: 45CFR46 (revised as of June 18, 1991). Washington, DC: Department of Health and Human Services.

Cowles, K. V. (1988). Issues in qualitative research on sensitive topics. Western Journal of Nursing Research, 10, 163–179. Damrosch, S. P. (1986). Ensuring anonymity by use of subject-generated identification codes. Research in Nursing & Health, 9, 61–63. Davis, A. J. (1989a). Clinical nurses’ ethical decisionmaking in situations of informed consent. Advances in Nursing Science, 11, 63–69. Davis, A. J. (1989b). Informed consent process in research protocols: Dilemmas for clinical nurses. Western Journal of Nursing Research, 11, 448–457. Flesch, R. (1948). New readability yardstick. Journal of Applied Psychology, 32, 221–223. Gunning, R. (1968). The technique of clear writing (Rev. ed.). New York: McGraw-Hill. Higgins, P. A., & Daly, B. J. (1999). Research methodology issues related to interviewing the mechanically ventilated patient. Western Journal of Nursing Research, 21, 773–784. McLaughlin, G. H. (1969). SMOG grading: A new readability formula. Journal of Reading, 12, 639–646. Meade, C. D. (1999). Improving understanding of the informed consent process and document. Seminars in Oncology Nursing, 15, 124–137. Munhall, P. L. (2001). Ethical considerations in qualitative research. In P. L. Munhall (Ed.), Nursing research: A qualitative perspective (pp. 537–549). Sudbury, MA: Jones & Bartlett. National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research. (1978). Belmont report: Ethical principles and guidelines for research involving human subjects. Washington, DC: U.S. Government Printing Office. Ramos, M. C. (1989). Some ethical implications of qualitative research. Research in Nursing & Health, 12, 57–64. Rempusheski, V. F. (1991a). Elements, perceptions, and issues of informed consent. Applied Nursing Research, 4, 201–204. Rempusheski, V. F. (1991b). Research data management: Piles into files—locked and secured. Applied Nursing Research, 4, 147–149. Sales, B. D., & Folkman, S. (Eds.). (2000). Ethics in research with human participants. Washington, DC: American Psychological Corporation. Schrum, L. (1995). Framing the debate: Ethical research in the information age. Qualitative Inquiry, 1, 311–326. Silva, M. C. (1995). Ethical guidelines in the conduct, dissemination, and implementation of nursing research. Washington, DC: American Nurses’ Association.

CHAPTER 7 Designing Ethical Research Silva, M. C., & Sorrell, J. M. (1984). Factors influencing comprehension of information for informed consent. International Journal of Nursing Studies, 21, 233–240. Sixsmith, J., & Murray, C. D. (2001). Ethical issues in the documentary data analysis of Internet posts and archives. Qualitative Health Research, 11, 423–432. Lindeke, L., Hauck, M. R., & Tanner, M. (2000). Practical issues in obtaining child assent for research. Journal of Pediatric Nursing, 15, 99–104. Thurber, F. W., Deatrick, J. A., & Grey, M. (1992). Children’s participation in research: Their right to consent. Journal of Pediatric Nursing, 7, 165–170. Watson, A. B. (1982). Informed consent of special subjects. Nursing Research, 31, 43–47.

Studies Cited in Chapter 7 Anderson, N. L. R., Nyamathi, A., McAvoy, J. A., Conde, F., & Casey, C. (2001) Perceptions about risk for HIV/AIDS among adolescents in juvenile detention. Western Journal of Nursing Research, 23, 336–359. Caelli, K. (2001). Engaging with phenomenology: Is it more of a challenge than it needs to be? Qualitative Health Research, 11, 273–281. Jones, M. E., Bond, M. L., Gardner, S., & Hernandez, M. (2002). Acculturation level and family planning

161

patterns of Hispanic immigrant women. MCN: The American Journal of Maternal/Child Nursing, 27, 26–32. Polit, D. F., London, A. S., & Martinez, J. M. (2001). The health of poor urban women. New York: MDRC. Thomas, B., Stamler, L. L., Lafrenier, K., & Dumala, R. (2001). The Internet: An effective tool for nursing research with women. Computers in Nursing, 18, 13–18. Varda, K. E., & Behnke, R. S. (2000). The effect of timing of initial bath on newborn’s temperature. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 29, 27–32. Wackerbarth, S. (1999). Modeling a dynamic decision process: Supporting the decisions of caregivers of family members with dementia. Qualitative Health Research, 9, 294–314. Wilde, M. H. (2002). Urine flowing: A phenomenological study of living with a urinary catheter. Research in Nursing & Health, 25, 14–24. Willson, P., McFarlane, J., Lemmey, D., & Malecha, A. (2001). Referring abused women: Does police assistance decrease abuse? Clinical Nursing Research, 10, 69–81.

8

Designing Quantitative Studies

T

he research design of a study spells out the basic strategies that researchers adopt to develop evidence that is accurate and interpretable. The research design incorporates some of the most important methodologic decisions that researchers make, particularly in quantitative studies. Thus, it is important to understand design options when embarking on a research project. This chapter and the two that follow focus on design issues for quantitative research, and Chapter 11 discusses designs for qualitative research. TIP: If you are doing a study, you will need to make many important decisions about the study’s design. These decisions will affect the overall believability of your findings. In some cases, the decisions will affect whether you receive funding (if you are seeking financial support for your study) or whether you are able to publish your research report (if you plan to submit it to a journal). Therefore, a great deal of care and thought should go into these decisions. ASPECTS OF Q U A N T I TAT I V E RESEARCH DESIGN The overall plan for addressing a research problem encompasses multiple issues, all of which have implications for the quality of evidence the study yields.

Intervention A fundamental design decision concerns the researcher’s role vis-à-vis study participants. In some studies, nurse researchers want to test the effects of a specific intervention (e.g., an innovative program to promote breast self-examination). In such experimental studies, researchers play an active role by introducing the intervention. In other studies, referred to as nonexperimental studies, the researcher observes phenomena as they naturally occur without intervening. There are numerous specific experimental and nonexperimental designs from which to choose. Comparisons In most studies, researchers develop comparisons to provide a context for interpreting results. The most common of types of comparison are as follows: 1. Comparison between two or more groups. For example, suppose we wanted to study the emotional consequences of having an abortion. To do this, we might compare the emotional status of women who had an abortion with that of women with an unintended pregnancy who delivered the baby. 2. Comparison of one group’s status at two or more points in time. For example, we might

CHAPTER 8 Designing Quantitative Studies

want to assess patients’ levels of stress before and after introducing a new procedure to reduce preoperative stress. Or we might want to compare coping processes among caregivers of patients with AIDS early and later in the caregiving experience. 3. Comparison of one group’s status under different circumstances. For example, we might compare people’s heart rates during two different types of exercise. 4. Comparison based on relative rankings. If, for example, we hypothesized a relationship between level of pain of cancer patients and their degree of hopefulness, we would be asking whether patients with high levels of pain feel less hopeful than patients with low levels of pain. This research question involves a comparison of those with different rankings—high versus low—on both variables. 5. Comparison with other studies. Researchers may directly compare their results with results from other studies, sometimes using statistical procedures. This type of comparison typically supplements rather than replaces other types of comparisons. In quantitative studies, this approach is useful primarily when the dependent variable is measured with a widely accepted approach (e.g., blood pressure measures or scores on a standard measure of depression). Example of using comparative data from other studies: Beckie, Beckstead, and Webb (2001) studied quality of life and health of women who had suffered a cardiac event. Women in their sample were administered standard scales for which there were national comparison data, enabling the researchers to evaluate their sample’s outcomes relative to national norms in the United States. Comparisons are often the central focus of a study, but even when they are not, they provide a context for understanding the findings. In the example of studying the emotional status of women who had an abortion, it would be difficult to know whether their emotional status was of concern without comparing it with that of others.

163

In some studies, a natural comparison group suggests itself. For example, if we were testing the effectiveness of a new nursing procedure for a group of nursing home residents, an obvious comparison group would be nursing home residents who were exposed to the standard procedure rather than to the innovation. In other cases, however, the choice of a comparison group is less clearcut, and the researcher’s decision about a comparison group can affect the interpretability of the findings. In the example about the emotional consequences of an abortion, we opted to use women who had delivered a baby as the comparison group. This reflects a comparison focusing on pregnancy outcome (i.e., pregnancy termination versus live birth). An alternative comparison group might be women who had a miscarriage. In this case, the comparison focuses not on the outcome (in both groups, the outcome is pregnancy loss) but rather on the determinant of the outcome. Thus, in designing a study, researchers must choose comparisons that will best illuminate the central issue under investigation. Controls for Extraneous Variables As noted in Chapter 2, the complexity of relationships among human characteristics often makes it difficult to answer research questions unambiguously unless efforts are made to isolate the key research variables and to control other factors extraneous to the research question. Thus, an important feature of the research design of quantitative studies is the steps that will be taken to control extraneous variables. Familiarity with the research literature often helps to identify especially important variables to control. Methods for enhancing research control are discussed in Chapter 9. Timing of Data Collection In most studies, data are collected from participants at a single point in time. For example, patients might be asked on a single occasion to describe their health-promoting behaviors. Some designs, however, call for multiple contacts with participants, usually to determine how things have

164

PART 3 Designs for Nursing Research

changed over time. Thus, in designing a study, the researcher must decide on the number of data collection points needed to address the research question properly. The research design also designates when, relative to other events, data will be collected. For example, the design might call for interviews with pregnant women in the sixteenth and thirtieth weeks of gestation, or for blood samples to be drawn after 10 hours of fasting. Research Sites and Settings Research designs also specify the site and setting for the study. As discussed in Chapter 2, sites are the overall locations for the research, and settings are the more specific places where data collection will occur. Sites and settings should be selected so as to maximize the validity and reliability of the data. In designing a study, it may be important to consider whether participants are influenced by being in settings that may be anxiety-provoking or foreign to their usual experiences. Communication With the Subjects In designing the study, the researcher must decide how much information to provide to study participants. As discussed in the previous chapter, full disclosure to subjects before obtaining their consent is ethically correct, but can sometimes undermine the value of the research. The researcher should also consider the costs and benefits of alternative means of communicating information to study participants. Among the issues that should be addressed are the following: • How much information about the study aims will be provided to (and withheld from) prospective subjects while they are being recruited and during the informed consent process? • How will information be provided—orally or in writing? • What is the reading and comprehension level of the least skilled participants? • Who will provide the information, and what will that person be expected to say in response to additional questions participants might ask?

• Will there be debriefing sessions after data are collected to explain more fully the study purpose or to answer questions? The nature of the communication with participants can affect their cooperation and the data they provide, and so these issues should be given careful consideration in designing the study. TIP: In making design decisions, you will often need to balance various considerations, such as time, cost, ethical issues, and study integrity. Try to get a firm understanding of your “upper limits” before making final design decisions. That is, what is the most money that can be spent on the project? What is the maximal amount of time available for conducting the study? What is the limit of acceptability with regard to ethical issues, given the risk/benefit ratio of the study? These limits often eliminate some design options. With these constraints in mind, the central focus should be on designing a study that maximizes the validity of the data. OVERVIEW OF RESEARCH DESIGN TYPES Quantitative research designs vary along a number of dimensions (some of which relate to factors discussed in the preceding section), as shown in Table 8-1. Some dimensions are independent of the others. For example, an experimental design can be cross-sectional or longitudinal. Quantitative designs in general share one thing in common: they tend to be fairly structured. Typically, quantitative researchers specify the nature of any intervention, comparisons to be made, methods to be used to control extraneous variables, timing of data collection, the study site and setting, and information to be given to participants—all before a single piece of data is gathered. Once data collection is underway, modifications to the research design are rarely instituted. As discussed in Chapter 11, research design in a qualitative study is more fluid: qualitative researchers often make deliberate modifications that are sensitive to what is being learned as data are gathered.

CHAPTER 8 Designing Quantitative Studies

165

TABLE 8.1 Dimensions of Research Designs DIMENSION

DESIGN

MAJOR FEATURES

Degree of structure

Structured Flexible

Design is specified before data are collected Design evolves during data collection

Type of group comparisons

Between-subjects

Subjects in groups being compared are different people Subjects in groups being compared are the same people at different times or in different conditions

Within-subjects Time frame

Cross-sectional Longitudinal

Data are collected at one point in time Data are collected at two or more points in time over an extended period

Control over independent variable

Experimental

Manipulation of independent variable, control group, randomization Manipulation of independent variable, but no randomization or no control group Manipulation of independent variable, no randomization or control group, limited control over extraneous variables No manipulation of independent variable

Quasi-experimental Preexperimental Nonexperimental Measurement of independent and dependent variables

Retrospective Prospective

This section describes several of the dimensions along which quantitative research designs vary. Other dimensions are discussed later in this chapter. Between-Subjects and Within-Subjects Designs As previously noted, most quantitative studies involve making comparisons, which are often between separate groups of people. For example, the hypothesis that the drug tamoxifen reduces the rate of breast cancer in high-risk women could be tested by comparing women who received tamoxifen and those who did not. In this example, those getting

Study begins with dependent variable and looks backward for cause or antecedent Study begins with independent variable and looks forward for the effect

the drug are not the same people as those not getting it. In another example, if we were interested in comparing the pain tolerance of men and women, the groups being compared would obviously involve different people. This class of design is referred to as between-subjects designs. Example of a study with a between-subjects design: Nantais-Smith and her colleagues (2001) examined differences in plasma and nipple aspirate carotenoid 12 months postpartum between women who had and women who had not breastfed their infants. It is sometimes desirable to make comparisons for the same study participants. For example, we

166

PART 3 Designs for Nursing Research

might be interested in studying patients’ heart rates before and after a nursing intervention, or we might want to compare lower back pain for patients lying in two different positions. These examples both call for a within-subjects design, involving comparisons of the same people under two conditions or at two points in time. The nature of the comparison has implications for the type of statistical test used. Example of a study with a within-subjects design: Hill, Kurkowski, and Garcia (2000) examined the effect of oral support (cheek and jaw support) on nutritive sucking patterns of preterm infants during feeding. Twenty preterm infants were observed under two conditions: with the support and without it. The Time Dimension Although most studies collect data at a single point in time, there are four situations in which it is appropriate to design a study with multiple points of data collection: 1. Studying time-related processes. Certain research problems specifically focus on phenomena that evolve over time (e.g., healing, learning, recidivism, and physical growth). 2. Determining time sequences. It is sometimes important to determine the sequencing of phenomena. For example, if it is hypothesized that infertility results in depression, then it would be important to determine that the depression did not precede the fertility problem. 3. Developing comparisons over time. Some studies are undertaken to determine if changes have occurred over time. For example, a study might be concerned with documenting trends in the smoking behavior of teenagers over a 10-year period. As another example, an experimental study might examine whether an intervention led to both short-term and long-term effects. 4. Enhancing research control. Some research designs for quantitative studies involve the collection of data at multiple points to enhance the interpretability of the results. For example,

when two groups are being compared with regard to the effects of alternative interventions, the collection of data before any intervention occurs allows the researcher to detect—and control—any initial differences between groups. Studies are often categorized in terms of how they deal with time. The major distinction is between cross-sectional and longitudinal designs. Cross-Sectional Designs Cross-sectional designs involve the collection of data at one point in time: the phenomena under study are captured during one period of data collection. Cross-sectional studies are appropriate for describing the status of phenomena or for describing relationships among phenomena at a fixed point in time. For example, we might be interested in determining whether psychological symptoms in menopausal women are correlated contemporaneously with physiologic symptoms. Cross-sectional designs are sometimes used for time-related purposes, but the results may be ambiguous. For example, we might test the hypothesis, using cross-sectional data, that a determinant of excessive alcohol consumption is low impulse control, as measured by a psychological test. When both alcohol consumption and impulse control are measured concurrently, however, it is difficult to know which variable influenced the other, if either. Cross-sectional data can most appropriately be used to infer time sequence under two circumstances: (1) when there is evidence or logical reasoning indicating that one variable preceded the other (e.g., in a study of the effects of low birth weight on morbidity in school-aged children, there would be no confusion over whether birth weight came first); and (2) when a strong theoretical framework guides the analysis. Cross-sectional studies can also be designed to permit inferences about processes evolving over time, such as when measurements capture a process at different points in its evolution with different people. As an example, suppose we wanted to study changes in professionalism as nursing

CHAPTER 8 Designing Quantitative Studies

students progress through a 4-year baccalaureate program. One way to investigate this would be to gather data from students every year until they graduate; this would be a longitudinal design. On the other hand, we could use a cross-sectional design by gathering data at a single point from members of the four classes, and then comparing the responses of the four classes. If seniors had higher scores on a measure of professionalism than freshmen, it might be inferred that nursing students become increasingly socialized professionally by their educational experiences. To make this kind of inference, we must assume that the seniors would have responded as the freshmen responded had they been questioned 3 years earlier, or, conversely, that freshmen students would demonstrate increased professionalism if they were questioned 3 years later. Such a design, which involves a comparison of multiple age cohorts, is sometimes referred to as a cohort comparison design. The main advantage of cross-sectional designs in such situations is that they are practical: they are easy to do and are relatively economical. There are, however, problems in inferring changes over time using a cross-sectional design. In our example, seniors and freshmen may have different attitudes toward the nursing profession, independent of any experiences during their 4 years of education. The overwhelming number of social and technologic changes in our society frequently makes it questionable to assume that differences in the behaviors, attitudes, or characteristics of different age groups are the result of time passing rather than a reflection of cohort or generational differences. In cross-sectional studies designed to study change, there are frequently several alternative explanations for the research findings—and that is precisely what good research design tries to avoid. Example of a cross-sectional study: Mindell and Jacobson (2000) assessed sleep patterns and the prevalence of sleep disorders during pregnancy. With a cross-sectional design, they compared women who were at four points in pregnancy: 8 to 12 weeks, 18 to 22 weeks; 25 to 28 weeks; and 35 to 38 weeks. They concluded that

167

sleep disturbances were especially common in late pregnancy. Longitudinal Designs A study in which data are collected at more than one point in time over an extended period uses a longitudinal design. (A study involving the collection of postoperative patient data on vital signs over a 2-day period would not be described as longitudinal.) There are several types of longitudinal designs. Trend studies are investigations in which samples from a population are studied over time with respect to some phenomenon. Different samples are selected at repeated intervals, but the samples are always drawn from the same population. Trend studies permit researchers to examine patterns and rates of change over time and to predict future developments. Trend studies typically are based on surveys, which are described in further detail in Chapter 10. Example of a trend study: Greenfield, Midanik, and Rogers (2000) studied trends in alcohol consumption in the United States over a 10-year period, using data from the 1984, 1990, and 1995 National Alcohol Surveys. They found that rates of heavy drinking had fallen between 1984 and 1990, but had remained unchanged between 1990 and 1995. Cohort studies are a particular kind of trend study in which specific subpopulations are examined over time. The samples are usually drawn from specific age-related subgroups. For example, the cohort of women born from 1946 to 1950 may be studied at regular intervals with respect to health care utilization. In a design known as a cross-sequential design,* two or more age cohorts are studied longitudinally so that both changes over time and generational (cohort) differences can be detected. In panel studies, the same people are used to supply data at two or more points in time. The term panel refers to the sample of subjects providing data. Because the same people are studied over time, design is sometimes referred to as a cohort-sequential design, or as a longitudinal cohort comparison design. *This

168

PART 3 Designs for Nursing Research

researchers can identify individuals who did and did not change and then examine characteristics that differentiate the two groups. As an example, a panel study could be designed to explore over time the antecedent characteristics of smokers who were later able to quit. Panel studies also allow researchers to examine how conditions and characteristics at time 1 influence characteristics and conditions at time 2. For example, health outcomes at time 2 could be studied among individuals with different healthrelated behaviors at time 1. Panel studies are intuitively appealing as an approach to studying change but are expensive to manage and can run into difficulties. The most serious challenge is the loss of participants over time—a problem known as attrition. Attrition is problematic because those who drop out of the study often differ in important ways from those who continue to participate, resulting in potential biases and lack of generalizability. Example of a panel study: Wilson, White, Cobb, Curry, Greene, and Popovich (2000) explored relationships between paternal—and maternal—fetal attachment and infant temperament. They first gathered data from pregnant women and their partners during the third trimester of pregnancy. These parental data were then linked to the infants’ temperament 1 year later when they were 8 to 9 months of age. Follow-up studies are similar to panel studies, but are usually undertaken to determine the subsequent development of individuals who have a specified condition or who have received a specified intervention—unlike panel studies, which have samples drawn from more general populations. For example, patients who have received a particular nursing intervention or clinical treatment may be followed to ascertain the long-term effects of the treatment. As another example, samples of premature infants may be followed to assess their later perceptual and motor development. Example of a follow-up study: McFarlane, Soeken, and Wiist (2000) tested three alternative interventions designed to decrease intimate partner violence to pregnant women. A sample of over 300 pregnant, physically abused

women were followed up through interviews at 2, 6, 12, and 18 months after delivery. In sum, longitudinal designs are appropriate for studying the dynamics of a phenomenon over time. Researchers must make decisions about the number of data collection points and the intervals between them based on the nature of study and available resources. When change or development is rapid, numerous time points at short intervals may be needed to document it. Researchers interested in outcomes that may occur years after the original data collection must use longer-term follow-up. However, the longer the interval, the greater the risk of attrition—and, usually, the costlier the study. TIP: Try not to make design decisions single-handedly. Seek the advice of professors, colleagues, or research consultants. Once you have made design decisions, it may be useful to write out a rationale for your choices, and share it with those you have consulted to see if they can find any flaws in your reasoning or if they can make suggestions for further improvements. EXPERIMENTS A basic distinction in quantitative research design is that between experimental and nonexperimental research. In an experiment, researchers are active agents, not passive observers. Early physical scientists learned that although pure observation of phenomena is valuable, complexities occurring in nature often made it difficult to understand important relationships. This problem was handled by isolating phenomena in a laboratory and controlling the conditions under which they occurred. The procedures developed by physical scientists were profitably adopted by biologists during the 19th century, resulting in many achievements in physiology and medicine. The 20th century has witnessed the use of experimental methods by researchers interested in human behavior. Characteristics of True Experiments The controlled experiment is considered by many to be an ideal—the gold standard for yielding

CHAPTER 8 Designing Quantitative Studies

reliable evidence about causes and effects. Except for purely descriptive research, the aim of many research studies is to understand relationships among phenomena. For example, does a certain drug result in the cure of a disease? Does a nursing intervention produce a decrease in patient anxiety? The strength of true experiments lies in the fact that experimenters can achieve greater confidence in the genuineness of causal relationships because they are observed under controlled conditions. As we pointed out in Chapter 4, hypotheses are never proved or disproved by scientific methods, but true experiments offer the most convincing evidence about the effect one variable has on another. A true experimental design is characterized by the following properties: • Manipulation—the experimenter does something to at least some subjects • Control—the experimenter introduces controls over the experimental situation, including the use of a control group • Randomization—the experimenter assigns subjects to a control or experimental group on a random basis Each of these features is discussed more fully in the following sections. Manipulation Manipulation involves doing something to study participants. The introduction of that “something” (i.e., the experimental treatment or intervention) constitutes the independent variable. The experimenter manipulates the independent variable by administering a treatment to some subjects and withholding it from others (or by administering some other treatment). The experimenter thus consciously varies the independent variable and observes the effect on the dependent variable. For example, suppose we hypothesized that gentle massage is effective as a pain relief measure for elderly nursing home residents. The independent variable (the presumed cause) in this example is receipt of gentle massage, which could be manipulated by giving some patients the massage intervention and withholding it from others. We would then

169

compare patients’ pain level (the dependent variable) in the two groups to see if differences in receipt of the intervention resulted in differences in average pain levels. Control Control is achieved in an experimental study by manipulating, by randomizing, by carefully preparing the experimental protocols, and by using a control group. This section focuses on the function of the control group in experiments. Obtaining evidence about relationships requires making at least one comparison. If we were to supplement the diet of premature infants with a particular nutrient for 2 weeks, their weight at the end of 2 weeks would tell us nothing about treatment effectiveness. At a bare minimum, we would need to compare their posttreatment weight with their pretreatment weight to determine if, at least, their weight had increased. But let us assume that we find an average weight gain of 1 pound. Does this gain support the conclusion that the nutritional supplement (the independent variable) caused weight gain (the dependent variable)? No, it does not. Babies normally gain weight as they mature. Without a control group—a group that does not receive the nutritional supplements—it is impossible to separate the effects of maturation from those of the treatment. The term control group refers to a group of subjects whose performance on a dependent variable is used to evaluate the performance of the experimental group or treatment group (the group that receives the intervention) on the same dependent variable. Randomization Randomization (also called random assignment) involves placing subjects in groups at random. Random essentially means that every subject has an equal chance of being assigned to any group. If subjects are placed in groups randomly, there is no systematic bias in the groups with respect to attributes that could affect the dependent variable. Let us consider the purpose of random assignment. Suppose we wanted to study the effectiveness of a contraceptive counseling program for

170

PART 3 Designs for Nursing Research

multiparous women who have just given birth. Two groups of subjects are included—one will be counseled and the other will not. The women in the sample are likely to differ from one another in many ways, such as age, marital status, financial situation, attitudes toward child-rearing, and the like. Any of these characteristics could affect a woman’s diligence in practicing contraception, independent of whether she receives counseling. We need to have the “counsel” and “no counsel” groups equal with respect to these extraneous characteristics to assess the impact of the experimental counseling program on subsequent pregnancies. The random assignment of subjects to one group or the other is designed to perform this equalization function. One method might be to flip a coin for each woman (more elaborate procedures are discussed later). If the coin comes up “heads,” the woman would be assigned to one group; if the coin comes up “tails,” she would be assigned to the other group. Although randomization is the preferred scientific method for equalizing groups, there is no guarantee that the groups will, in fact, be equal. As an extreme example, suppose the study sample involves 10 women who have given birth to 4 or more children. Five of the 10 women are aged 35 years or older, and the remaining 5 are younger than age 35. We would expect random assignment to result in two or three women from the two age ranges in each group. But suppose that, by chance, the older five women all ended up in the experimental group. Because these women are nearing the end of their childbearing years, the likelihood of their conceiving is diminished. Thus, follow-up of their subsequent childbearing (the dependent variable) might suggest that the counseling program was effective in reducing subsequent pregnancies; however, a higher birth rate for the control group may reflect only age and fecundity differences, not lack of exposure to counseling. Despite this possibility, randomization remains the most trustworthy and acceptable method of equalizing groups. Unusual or deviant assignments such as this one are rare, and the likelihood of obtaining markedly unequal groups is reduced as the number of subjects increases.

You may wonder why we do not consciously control those subject characteristics that are likely to affect the outcome. The procedure that is sometimes used to accomplish this is known as matching. For example, if matching were used in the contraceptive counseling study, we might want to ensure that if there were a married, 38-year-old woman with six children in the experimental group, there would be a married, 38-year-old woman with six children in the control group as well. There are two serious problems with matching, however. First, to match effectively, we must know (and measure) the characteristics that are likely to affect the dependent variable, but this information is not always known. Second, even if we knew the relevant traits, the complications of matching on more than two or three characteristics simultaneously are prohibitive. With random assignment, on the other hand, all possible distinguishing characteristics—age, gender, intelligence, blood type, religious affiliation, and so on—are likely to be equally distributed in all groups. Over the long run, the groups tend to be counterbalanced with respect to an infinite number of biologic, psychological, economic, and social traits. To demonstrate how random assignment is performed, we turn to another example. Suppose we were testing two alternative interventions to lower the preoperative anxiety of children who are about to undergo tonsillectomy. One intervention involves giving structured information about the surgical team’s activities (procedural information); the other involves structured information about what the child will feel (sensation information). A third control group receives no special intervention. With a sample of 15 subjects, 5 children will be in each of the 3 groups. With three groups, we cannot use a coin flip to determine group assignments. We could, however, write the children’s names on slips of paper, put the slips into a hat, and then draw names. The first five individuals whose names were drawn would be assigned to group I, the second five would be assigned to group II, and the remaining five would be assigned to group III. Pulling names from a hat involves a lot of work if the sample is big. Researchers typically use

CHAPTER 8 Designing Quantitative Studies

a table of random numbers in the randomization process. A portion of such a table is reproduced in Table 8-2. In a table of random numbers any digit from 0 to 9 is equally likely to follow any other digit. Going in any direction from any point in the table produces a random sequence. In our example, we would number the 15 subjects from 1 to 15, as shown in the second column of Table 8-3, and then draw numbers between 01

171

and 15 from the random number table. A simple procedure for finding a starting point is to close your eyes and let your finger fall at some point on the table. For the sake of following the example, let us assume that our starting point is at number 52, circled on Table 8-2. We can move in any direction in the table from that point, selecting numbers that fall between 01 and 15. Let us move to the right, looking at two-digit combinations (to get numbers

TABLE 8.2 Small Table of Random Digits 46 69 14 56 81

85 24 01 30 30

05 89 33 38 44

23 34 17 73 85

26 60 92 15 85

34 67 75 45 30 50 59 74 76 16 52 06 68 65 22

83 75 72 96 73

00 21 77 76 76

74 91 06 43 45 61 31 83 18 55 76 50 33 45 13 11 65 49 98 93 92 85 25 58 66

70 90 39 88 45

28 41 90 15 13

42 59 40 20 46

43 36 21 00 35

26 14 15 80 45

79 33 59 20 59

37 52 58 55 40

59 12 94 49 47

52 66 90 14 20

20 65 67 09 59

01 15 96 55 82 34 66 82 14 96 27 74 43 94 75

32 76 15 82 16

67 41 75 57 80

70 37 18 05 95

01 23 63 32 09

41 93 73 78 66

50 32 75 21 79

21 95 09 62 46

41 05 82 20 48

29 87 44 24 46

06 00 49 78 08

73 11 90 17 55

12 19 05 59 58

71 92 04 45 15

85 78 92 19 19

71 42 17 72 11

59 63 37 53 87

57 40 01 32 82

43 80 80 80 93

25 85 08 89 12

38 40 87 07 81

41 92 70 80 84

45 79 74 02 64

60 43 88 94 74

83 52 72 81 45

32 90 25 33 79

59 63 67 19 05

83 18 36 00 61

01 38 66 54 72

29 38 16 15 84

14 47 44 58 81

13 47 94 34 18

49 61 31 36 34

82 53 82 13 29

47 34 64 57 59

42 24 12 41 38

55 42 28 72 86

93 76 20 00 27

48 75 92 69 94

54 12 90 90 97

53 21 41 26 21

52 17 31 37 15

47 24 41 42 98

18 74 32 78 62

61 62 39 46 09

91 77 21 42 53

36 37 97 25 67

74 07 63 01 87

86 44 93 52 04

88 98 39 16 73

75 91 94 29 72

50 68 55 02 10

87 22 47 86 31

19 36 94 54 75

15 02 45 15 05

20 40 87 83 19

00 08 42 42 30

23 67 84 43 29

12 76 05 46 47

30 37 04 97 66

28 84 14 83 56

07 16 98 54 43

83 05 07 82 82

Reprinted from A Million Random Digits with 100,000 Normal Deviates. New York: The Free Press, 1955. Used with permission of the Rand Corporation, Santa Monica, CA.

172

PART 3 Designs for Nursing Research

TABLE 8.3 Example of Random Assignment Procedure

TABLE 8.4 Breakdown of the Gender Composition of the Three Groups

NAME OF SUBJECT

NUMBER

GENDER GROUP I GROUP II GROUP III

Kristina N.

1

I

Trevor S.

2

III

Adeline B.

3

III

Lauren C.

4

II

Rebecca C.

5

II

Nathan O.

6

I

GROUP ASSIGNMENT

Lindsey S.

7

III

Thomas N.

8

III

Sean S.

9

II

Amy D.

10

III

Alana M.

11

I

Emily B.

12

II

Gabriel B

13

II

Taylor M.

14

I

Christopher R.

15

I

greater than 9, i.e., from 10 to 15). The number to the right of 52 is 06. The person whose number is 06, Nathan O., is assigned to group I. Moving along in the table, the next number within the range of 01 to 15 is 11. (To find numbers in the required range, we have to bypass numbers between 16 and 99.) Alana M., whose number is 11, is also assigned to group I. When we get to the end of the row, we move down to the next row, and so forth. The next three numbers are 01, 15, and 14. Thus, Kristina N., Christopher R., and Taylor M. are all put into group I. The next five numbers between 01 and 15 that emerge in the random number table are used to assign five individuals to group II in the same fashion, as shown in the third column of Table 8-3. The remaining five people in the sample are put into group III. Note that numbers that have already been used often reappear in the table before the task is completed. For example, the number 15

Boys

3

2

2

Girls

2

3

3

appeared four times during the randomization procedure. This is perfectly normal because the numbers are random. After the first time a number appears and is used, subsequent appearances can be ignored. It might be useful to look at the three groups to see if they are about equal with respect to one readily discernible characteristic, that is, the subjects’ gender. We started out with eight girls and seven boys in all. As Table 8-4 shows, randomization did a good job of allocating boys and girls about equally across the three research groups. We must accept on faith the probability that other characteristics (e.g., family income, health status, preoperative anxiety) are fairly well distributed in the randomized groups as well. TIP: Researchers usually do not have the full sample assembled when the study begins, but rather take subjects into the sample on a “rolling enrollment” basis. However, the same system as just described can be used without knowing any names if sample size is predetermined. Thus, using the assignments shown in Table 8-3, in a study of 15 subjects and three groups, the first person to enter the study would be assigned to Group I, the second and third person would be assigned to Group III, and so on. This method ensures that the sample sizes for the groups being compared are equal. Note that in the previous discussion we did not say that the five subjects in group I would be assigned to the procedural information group. This is because it is a good strategy to randomly assign groups to treatments, as well as individuals to groups. Let us give the procedural information, sensation informa-

CHAPTER 8 Designing Quantitative Studies

tion, and control conditions the numbers 1, 2, and 3, respectively. Finding a new starting point in the random number table, we look for the numbers 1, 2, or 3. This time we can look at one digit at a time because 3 is the biggest number. We will start at number 8 in the ninth row of the table, indicated by a rectangle. Reading down this time, we find the number 1. We therefore assign group I to the procedural information condition. Further along in the same column we come to the number 3. Group III, therefore, is assigned to the second condition, sensation information, and the remaining group, group II, is assigned to the control condition. In most cases, as just discussed, randomization involves the random assignment of individual subjects to different groups. However, an alternative is cluster randomization, which involves randomly assigning clusters of individuals to different treatment groups (Hauck, Gilliss, Donner, & Gortner, 1991). Cluster randomization may sometimes enhance the feasibility of conducting a true experiment. Groups of patients who enter a hospital unit at the same time, or groups of patients from different medical practices, can be randomly assigned to a treatment condition as a unit—thus ruling out, in some situations, some practical impediments to randomization. This approach also reduces the possibility of contamination between two different treatments, that is, the mingling of subjects in the groups, which could reduce the effectiveness of the manipulation. The main disadvantages of cluster randomization are that the statistical analysis of data obtained through this approach is more complex, and sample size requirements are usually greater for a given level of accuracy. Researchers usually assign subjects to groups in proportion to the number of groups being compared. For example, a sample of 300 subjects in a 2-group design would generally have 150 people in the experimental group and 150 in the control group. If there were 3 groups being compared, there would be 100 per group. However, it is also possible (and sometimes desirable ethically) to have a different allocation. For example, if an especially promising treatment for a serious illness were developed, we could assign 200 to the treatment group

173

and 100 to the control group. Such an allocation does, however, make it more difficult to detect treatment effects at statistically significant levels. Experimental Designs There are numerous experimental designs; the most widely used designs are described in this section and summarized in Table 8-5. Basic Experimental Designs At the beginning of this chapter, we described a study that tested the effect of gentle massage on the pain levels of elderly nursing home residents. This example illustrates a simple design that is sometimes referred to as an after-only design or a posttest-only design because data on the dependent variable are collected only once—after random assignment is completed and the experimental treatment has been introduced. Example of a posttest-only experimental design: Milne (2000) used a posttest-only design to study the effect of an educational intervention relating to urinary incontinence on the subsequent help-seeking behavior of older adults. One group received individualized instruction and written information, and the other received written information alone. Two months later, Milne determined how many subjects in each group sought professional help for urinary incontinence. A second basic design is the most widely used experimental design by nurse researchers. Suppose we hypothesized that convective airflow blankets are more effective than conductive water-flow blankets in cooling critically ill patients with fever. We decide to use a design that involves assigning patients to the two different types of blankets (the independent variable) and measuring the dependent variable (body temperature) twice, before and after the intervention. This scheme permits us to examine whether one blanket type is more effective that the other in reducing fever—that is, with this design researchers can examine change. This design, because of its two measurement points, is referred to as a

174

PART 3 Designs for Nursing Research

TABLE 8.5 Experimental Designs NAME OF DESIGN

PREINTERVENTION DATA?

WITHIN- OR BETWEEN-GROUPS

Posttest-only (after-only)

No

Between

One data collection point after the intervention; not appropriate for measuring change

Pretest–posttest (before–after)

Yes

Between

Data collection both before and after the intervention; appropriate for measuring change; can determine differences between groups (experimental) and change within groups (quasi- experimental)

Solomon four-group

For some subjects

Between

Data collection before and after the intervention for one experimental and one control group, but after only for a second experimental and control group, to assess pretest effects

Factorial

Optional

Between

Experimental manipulation of more than one independent variable; permits a test of main effects for each manipulated variable and interaction effects for combinations of manipulated variables

Randomized block

Optional

Between

Random assignment to groups within different levels of a blocking variable that is not under experimental control (e.g., gender)

Crossover/repeated measures

Optional

Within

Subjects are exposed to all treatments but are randomly assigned to different orderings of treatments; subjects serve as their own controls

before—after design or a pretest—posttest design. In such designs, the initial measure of the dependent variable is often referred to as the baseline measure, and the posttest measure of the dependent variable may be referred to as the outcome

FEATURES

measure—that is, the measure that captures the outcome of the experimental intervention. TIP: When using an experimental design that involves the collection of data both

CHAPTER 8 Designing Quantitative Studies

before and after the intervention, it is considered good practice to collect the pretest data before randomization to groups. This ensures that subjects (and researchers) will not be biased in any way by knowledge of the group assignments. Example of a pretest—posttest experimental design: Sandgren, McCaul, King, O’Donnell, and Foreman (2000) conducted an experiment to test the effectiveness of a cognitive-behavioral telephone therapy intervention for patients with breast cancer. Women in the study were randomly assigned to the intervention or to a control group. Measures of the dependent variables (e.g., psychological distress, coping, and quality of life) were measured at baseline and at follow-up, and changes over time were determined. Solomon Four-Group Design When data are collected both before and after an intervention, as in a pretest—posttest design, the posttest measure of the dependent variable may be affected not only by the treatment but also by exposure to the pretest. For example, if the intervention was a workshop to improve nurses’ attitudes toward patients with AIDS, a pretest attitudinal measure may in itself constitute a sensitizing treatment that could mask the workshop’s effectiveness. Such a situation calls for the Solomon four-group design, which involves two experimental groups and two control groups. One experimental group and one control group are administered the pretest and the other groups are not, thereby allowing the effects of the pretest measure and intervention to be segregated. Figure 8-1 illustrates this design. Example of a Solomon four-group design: Swanson (1999) used a Solomon four-group design in her study of the effects of a caring-based counseling intervention on the emotional wellbeing of women who had had a miscarriage. Swanson adopted this design because of a concern that “the potential existed that participating in a longitudinal control group with early focused attention on loss might, in itself, serve as a form of recognition, support, and validation” (p. 290).

175

Data Collection Group

Before

After

X

X

Experimental—with pretest Experimental—without pretest

X

Control—with pretest

X

X

Control—without pretest

X

F I G U R E 8 . 1 Solomon four-group experimental design.

Factorial Design The three designs described thus far are ones in which the experimenter manipulates only one independent variable. It is possible, however, to manipulate two or more variables simultaneously. Suppose we were interested in comparing two therapeutic strategies for premature infants: tactile stimulation versus auditory stimulation. At the same time, we are interested in learning if the daily amount of stimulation (15, 30, or 45 minutes) affects infants’ progress. The dependent variables for the study are measures of infant development (e.g., weight gain and cardiac responsiveness). Figure 8-2 illustrates the structure of this experiment. This factorial design permits the testing of multiple hypotheses in a single experiment. In this example, the three research questions being addressed are as follows: 1. Does auditory stimulation have a more beneficial effect on the development of premature infants than tactile stimulation, or vice versa? Type of stimulation Auditory A1 Daily exposure

Tactile A2

15 Min. B1

A1

B1

A2

B1

30 Min. B2

A1

B2

A2

B2

45 Min. B3

A1

B3

A2

B3

F I G U R E 8 . 2 Example of a factorial design.

176

PART 3 Designs for Nursing Research

2. Is the duration of stimulation (independent of type) related to infant development? 3. Is auditory stimulation most effective when linked to a certain dose and tactile stimulation most effective when coupled with a different dose? The third question demonstrates the strength of factorial designs: they permit us to evaluate not only main effects (effects resulting from experimentally manipulated variables, as exemplified in questions 1 and 2) but also interaction effects (effects resulting from combining treatments). We may feel that it is insufficient to say that auditory stimulation is preferable to tactile stimulation (or vice versa) and that 45 minutes of stimulation per day is more effective than 15 or 30 minutes per day. How these two variables interact (how they behave in combination) is also of interest. Our results may indicate that 15 minutes of tactile stimulation and 45 minutes of auditory stimulation are the most beneficial treatments. We could not have learned this by conducting two separate experiments that manipulated only one independent variable and held the second one constant. In factorial experiments, subjects are assigned at random to a specific combination of conditions. In the example that Figure 8-2 illustrates, premature infants would be assigned randomly to one of six cells. A cell in experimental research refers to a treatment condition; it is represented in a schematic diagram as a box (cell) in the design. Figure 8-2 can also be used to define some design terminology. The two independent variables in a factorial design are the factors. Type of stimulation is factor A and amount of daily exposure is factor B. Each factor must have two or more levels (if there were only one level, the factor would not be a variable). Level 1 of factor A is auditory and level 2 of factor A is tactile. When describing the dimensions of the design, researchers refer to the number of levels. The design in Figure 8-2 is 2 ! 3 design: two levels in factor A times three levels in factor B. If a third source of stimulation, such as visual stimulation, were added, and if a daily dosage of 60 minutes were also added, the design would be a 3 ! 4 design. Factorial experiments can be performed with three or more independent variables (factors), but designs with more than three factors are rare.

Example of a factorial design: Schultz, Ashby-Hughes, Taylor, Gillis, and Wilkins (2000) used a 2 ! 2 factorial design to study treatments to reduce diarrhea among critically ill tube-fed patients receiving antibiotics. One factor was fiber-containing versus fiber-free tube feedings. The second was administration of pectin versus a placebo. The researchers found a trend toward less diarrhea in the fiber/pectin group. Randomized Block Design A design that looks similar to a factorial design in structure is the randomized block design.* In such a design, there are two factors (independent variables), but one factor is not experimentally manipulated. Suppose that we were interested in comparing the effects of tactile versus auditory stimulation for male versus female infants. We could structure this as a 2 ! 2 experiment, with type of stimulation as one factor and gender as the other. The variable gender, which we cannot manipulate, is known as a blocking variable. In an experiment to test the effectiveness of alternative stimulation therapies, we obviously could not randomly assign subjects to one of four cells, as in a factorial experiment, because infants’ gender is a given. We can, however, randomly assign male and female subjects separately to the two stimulation methods. Suppose there are 40 male infants and 40 female infants available for the study. We would not randomly assign half the 80 infants to tactile stimulation and the other half to auditory stimulation. Rather, we would randomize boys and girls separately to the two treatments, thereby guaranteeing 20 subjects in each cell of this 4-cell design. The inclusion of a blocking variable in a study design enhances the researcher’s control over sample composition (i.e., to ensure that sufficient numbers of subjects with specific characteristics are included) and over extraneous variables. That is, if we consider gender a confounding variable because we believe that male and female infants will respond differently to the two therapies, then a randomized block design is *The terminology for this design varies from text to text. Some authors refer to this as a factorial design; others call it a levelsby-treatment design.

CHAPTER 8 Designing Quantitative Studies

needed. In randomized block designs, as in factorial designs, interaction effects can be examined. Example of a randomized block design: Harrison, Williams, Berbaum, Stem, and Leeper (2000) used a randomized block design in their study of the effects of gentle human touch on such outcomes as behavioral distress, sleep, and motor activity in preterm infants. The infants were randomly assigned to an experimental or control group within blocks based on gestational age. The blocking variable divided the sample into three gestational age groups: 27 to 28 weeks; 29 to 31 weeks, and 32 to 33 weeks. The design can be extended to include more than one blocking variable. For example, we could add infant birth weight as a blocking variable in our study of alternative stimulation therapies. It is also possible to include more than one manipulated variable, thereby creating a design that is both a randomized block and a factorial design. In theory, the number of blocking and manipulated variables is unlimited, but practical concerns usually dictate a relatively small number of each. Expansion of the design usually requires that more subjects be used. As a general rule of thumb, a minimum of 20 subjects per cell is recommended to achieve stability within cells. This means that, whereas a minimum of 80 subjects would be needed for a 2 ! 2 design, 160 subjects would be needed for a 2 ! 2 ! 2 design. Example of a combined randomized block and factorial design: Metzger, Jarosz, and Noureddine (2000) studied the effects of two experimentally manipulated factors (high-fat versus low-fat diet and forced exercise versus sedentary condition) on obesity in rats. The blocking variable in this study was the genetic obesity of the rats: both genetically obese and lean rats were randomly assigned to the four treatment conditions, yielding a 2 ! 2 ! 2 design. Crossover Design Thus far, we have described experimental studies in which subjects who are randomly assigned to different treatments are different people. For instance, in the previous example, the infants

177

exposed to auditory stimulation were not the same infants as those exposed to tactile stimulation. A crossover design (also known as a repeated measures design) involves the exposure of the same subjects to more than one experimental treatment. This type of within-subjects design has the advantage of ensuring the highest possible equivalence among subjects exposed to different conditions— the groups being compared are equal with respect to age, weight, health, and so on because they are composed of the same people. In a crossover experimental design, subjects are randomly assigned to different orderings of treatments. For example, if a crossover design were used to compare the effects of auditory and tactile stimulation on infant development, some infants would be randomly assigned to receive auditory stimulation first, and others would be assigned to receive tactile stimulation first. In such a study, the three conditions for an experiment have been met: there is manipulation, randomization, and a control group, with subjects serving as their own controls. Although crossover designs are extremely powerful, they are inappropriate for certain research questions because of the problem of carry-over effects. When subjects are exposed to two different treatments or conditions, they may be influenced in the second condition by their experience in the first condition. As one example, drug studies rarely use a crossover design because drug B administered after drug A is not necessarily the same treatment as drug B administered before drug A. Example of a crossover design: Winkelman (2000) used a randomized crossover design to examine the effect of two alternative backrest positions (flat/horizontal versus 30-degree elevation) on intracranial and cerebral perfusion pressures in brain-injured adults. The elevated position resulted in significant and clinically important improvements. Experimental and Control Conditions In designing experimental studies, researchers make many decisions about what the experimental and control conditions entail, and these decisions can affect the researchers’ conclusions.

178

PART 3 Designs for Nursing Research

The Experimental Condition To give an experimental intervention a fair test, researchers need to carefully design an intervention that is appropriate to the problem and of sufficient intensity and duration that effects might reasonably be expected. The full nature of the intervention must be clearly delineated in formal protocols that spell out exactly what the treatment is. Among the questions researchers need to address are the following: • What is the intervention, and how does it differ from usual methods of care? • If there are two alternative interventions, how exactly do they differ? • What are the specific procedures to be used with those receiving the intervention? • What is the dosage or intensity of the intervention? • Over how long a period will the intervention be administered, how frequently will it be administered, and when will the treatment begin (e.g., 2 hours after surgery)? • Who will administer the intervention? What are their credentials, and what type of special training will they receive? • Under what conditions will the intervention be withdrawn or altered? The goal in most experimental studies is to have a comparable intervention for all subjects in the treatment group. This goal is difficult to achieve without careful advance planning and clear written protocols. TIP: Qualitative studies can provide valuable information in helping researchers develop interventions to be tested in experimental studies. Gamel, Grypdonck, Hengeveld, and Davis (2001) examined research findings from qualitative studies to develop a nursing intervention for women with gynecologic cancer. Their intervention focused on sexual adaptation and qualitative data helped bring the patient perspective into this treatment. The Control Condition The control group condition used as a basis of comparison in a study is referred to as the counterfactual. Researchers have choices about what to use as the counterfactual. Their decision is sometimes

based on theoretical or substantive grounds, but may also be driven by practical or ethical concerns. In some research, control group subjects receive no treatment at all—they are merely observed with respect to performance on the dependent variable. This kind of situation is not always feasible for nursing research projects; if we wanted to evaluate the effectiveness of a nursing intervention on hospital patients, we would not devise an experiment in which patients in the control group received no nursing care at all. Among the possibilities for the counterfactual are the following: 1. An alternative intervention; for example, in our example of infant stimulation, subjects were exposed to alternative therapies. 2. A placebo or pseudointervention presumed to have no therapeutic value; for example, in studies of the effectiveness of drugs, some patients get the experimental drug and others get an innocuous substance, as in the previously described study by Schultz et al. (2000) that compared pectin with a placebo. Placebos are used to control for the nonpharmaceutical effects of drugs, such as the attention being paid to subjects (although there can be placebo effects—changes in the dependent variable attributable to the placebo condition—because of subjects’ expectations). 3. Standard methods of care—the normal procedures used to treat patients; for example, in Parent and Fortin’s (2000) study of an intervention for cardiac surgery patients, described later in this chapter, control group subjects were treated according to the usual hospital procedures for such patients. 4. Different doses or intensities of treatment wherein all subjects get some type of intervention, but the experimental group gets an intervention that is richer, more intense, or longer; for example, in Milne’s (2000) study of an educational intervention regarding urinary incontinence for community-dwelling elders, one group got written instruction and the other got written instruction plus individual support.

CHAPTER 8 Designing Quantitative Studies

5. Delayed treatment; that is, the control group eventually receives the full experimental treatment, but treatment is deferred. Example of delayed treatment: Garcia de Lucio, Gracia Lopez, Marin Lopez, Mas Hesse, and Camana Vaz (2000) tested the effectiveness of an intervention designed to improve the communication skills of nurses when interacting with relatives of seriously ill patients. The experimental group received immediate training; control group training was delayed 6 months. The communication skills of both groups were compared after the initial training, but before control subjects were trained. Methodologically, the best possible test is between two conditions that are as different as possible, as when the experimental group receives a strong treatment and the control group gets no treatment at all. Ethically, however, the most appealing counterfactual is probably the “delay of treatment” approach (number 5), which may be difficult to do pragmatically. Testing two competing interventions (number 1) also has ethical appeal, but the risk is that the results will be inconclusive because it is difficult to detect differential effects, especially if both interventions are at least moderately effective. Whatever decision is made about a control group strategy, researchers need to be as careful in spelling out the counterfactual as in delineating the intervention. In research reports, researchers sometimes say that the control group got the “usual methods of care” without explaining what that condition was and how different it was from the intervention being tested. In drawing on an evidence base for practice, nurses need to understand exactly what happened to study participants in different conditions. TIP: Some researchers elect to combine two or more comparison strategies. For example, they might test two alternative treatments (option 1) against usual methods of care (option 3). Such an approach is attractive but, of course, adds to the cost and complexity of the study.

179

Strengths and Limitations of Experiments Controlled experiments are often considered the ideal of science. In this section, we explore the reasons why experimental methods are held in high esteem and examine some of their limitations. Experimental Strengths True experiments are the most powerful method available for testing hypotheses of cause-and-effect relationships between variables. Studies using an experimental design are in general thought to yield the highest-quality evidence regarding the effects of specific interventions and nursing actions. Because of their special controlling properties, experiments offer greater corroboration than any other research approach that, if the independent variable (e.g., diet, drug dosage, or teaching approach) is manipulated in a specified way, then certain consequences in the dependent variable (e.g., weight loss, recovery of health, or learning) may be expected to ensue. This “if... then” type of relationship is important because of its implications for prediction and control. The great strength of experiments, then, lies in the confidence with which causal relationships can be inferred. Lazarsfeld (1955), whose thinking reflects the ideas of John Stuart Mill, identified three criteria for causality. The first criterion is temporal: a cause must precede an effect in time. If we were testing the hypothesis that saccharin causes bladder cancer, it would be necessary to demonstrate that subjects had not developed cancer before exposure to saccharin. In experiments, researchers manipulate the independent variable and then measure subsequent outcomes, and so the sequence is under their control. The second requirement is that there be an empirical relationship between the presumed cause and the presumed effect. In the saccharin and cancer example, we would have to demonstrate an association between saccharin consumption and the presence of a carcinoma, that is, that a higher percentage of saccharin users than nonusers developed cancer. In an experiment, this empirical relationship between the independent and dependent

180

PART 3 Designs for Nursing Research

variables is put to a direct test. The final criterion for establishing a causal relationship is that the relationship cannot be explained as being caused by a third variable. Suppose, for instance, that people who elect to use saccharin tend also to drink more coffee than nonusers of saccharin. There would then be a possibility that any relationship between saccharin use and bladder cancer reflects an underlying causal relationship between a substance in coffee and bladder cancer. It is particularly because of this third criterion that the experimental approach is so strong. Through the controls imposed by manipulation, comparison, and randomization, alternative explanations to a causal interpretation can often be ruled out or discredited. Experimental Limitations Despite the benefits of experimental research, this type of design also has limitations. First of all, there are often constraints that make an experimental approach impractical or impossible. These constraints are discussed later in this chapter. Experiments are sometimes criticized for their artificiality. Part of the difficulty lies in the requirements for randomization and then equal treatment within groups. In ordinary life, the way we interact with people is not random. For example, certain aspects of a patient (e.g., age, physical appearance, or severity of illness) may cause us to modify our behavior and our care. The differences may be subtle, but they undoubtedly are not random. Another aspect of experiments that is sometimes considered artificial is the focus on only a handful of variables while holding all else constant. This requirement has been criticized as being reductionist and as artificially constraining human experience. Experiments that are undertaken without a guiding theoretical framework are sometimes criticized for being able to establish a causal connection between an independent and dependent variable without providing any causal explanation for why the intervention resulted in the observed outcomes. A problem with experiments conducted in clinical settings is that it is often clinical staff, rather than researchers, who administer an intervention,

and therefore it can sometimes be difficult to determine if subjects in the experimental group actually received the treatment, and if those in the control group did not. It may be especially difficult to maintain the integrity of the intervention and control conditions if the study period extends over time. Moreover, clinical studies are usually conducted in environments over which researchers have little control—and control is a critical factor in experimental research. McGuire and her colleagues (2000) describe some issues relating to the challenges of testing interventions in clinical settings. Sometimes a problem emerges if subjects themselves have discretion about participation in the treatment. Suppose, for example, that we randomly assigned patients with HIV infection to a special support group intervention or to a control group. Experimental subjects who elect not to participate in the support groups, or who participate infrequently, actually are in a “condition” that looks more like the control condition than the experimental one. The treatment is diluted through nonparticipation, and it may become difficult to detect any effects of the treatment, no matter how effective it might otherwise have been. Such issues need to be taken into consideration in designing the study. Another potential problem is the Hawthorne effect, which is a placebo effect. The term is derived from a series of experiments conducted at the Hawthorne plant of the Western Electric Corporation in which various environmental conditions, such as light and working hours, were varied to determine their effects on worker productivity. Regardless of what change was introduced, that is, whether the light was made better or worse, productivity increased. Knowledge of being included in the study appears to have affected people’s behavior, thereby obscuring the effect of the variable of interest. In a hospital situation, researchers might have to contend with a double Hawthorne effect. For example, if an experiment to investigate the effect of a new postoperative patient routine were conducted, nurses and hospital staff, as well as patients, might be aware of their participation in a study, and both groups might alter their actions accordingly. It is precisely for this reason that double-blind experiments, in which

CHAPTER 8 Designing Quantitative Studies

neither subjects nor those who administer the treatment know who is in the experimental or control group, are so powerful. Unfortunately, the doubleblind approach is not feasible for most nursing studies because nursing interventions are more difficult to disguise than drugs. Example of a double-blind experiment: McCormick, Buchman, and Maki (2000) used a double-blind approach (with a before—after design) to test the effectiveness of two alternative hand-care treatments (an oil-containing lotion and a novel barrier skin cream) for health care workers with severe hand irritation. Subjects in both groups experienced marked improvements. In sum, despite the clearcut superiority of experiments in terms of their ability to test causal hypotheses, they are subject to a number of limitations, some of which may make them difficult to apply to real-world problems. Nevertheless, with the growing demand for evidence-based practice, true experimental designs are increasingly being used to test the effects of nursing practices and procedures. Research Example of an Experimental Study Parent and Fortin (2000) used an experimental design to test an intervention for reducing anxiety and increasing postoperative activity in male cardiac surgery patients. The intervention involved “vicarious experience,” in which patients about to undergo elective coronary artery bypass graft (CABG) surgery were linked to volunteers who had recovered from similar surgery. The linkage, which involved three supportive visits during the hospitalization and recovery period, was designed to demonstrate the active lives former patients were leading. A sample of 56 first-time male patients undergoing CABG surgery were randomly assigned to the experimental vicarious experience group or to a control group not receiving the special intervention. Patients in both groups were given routine information on surgery and recovery by health professionals. A coin flip was used to assign subjects to groups. The two groups were found to be comparable before the intervention in terms of a wide range of background characteristics (age, occupa-

181

tional category, smoking status, number of bypass grafts). However, the average baseline anxiety level of those in the experimental group was significantly higher than that of control group members. Thus, randomization equalized the two groups on most, but not all, characteristics. By chance alone, more high-anxiety patients were assigned to the treatment group. A before—after design was used to assess declines in patient anxiety. Anxiety was measured at 48 hours and 24 hours before surgery, and again at 5 days and 4 weeks after surgery. Data collection after surgery also involved questions about activities and about self-efficacy expectations. Despite the initially higher anxiety levels of the men who received the vicarious experience treatment, the experimental group had significantly lower anxiety levels after the intervention than the control group. Only men in the experimental group showed a significant decline in anxiety during hospitalization; moreover, they reported higher levels of self-efficacy and self-reported activity after surgery than men in the control group. The researchers concluded that vicarious experience provided through dyadic support is an effective strategy to help cardiac patients cope with surgical anxiety.

TIP: In general, it is wise to randomize whenever possible, to reduce the possibility of biases. In an experiment, this means randomly assigning subjects to groups, groups to treatments, and conditions to subjects (in a repeated measures design). It also means, in general, looking for other opportunities to randomize whenever conditions vary across subjects, such as randomly assigning patients to rooms or nursing staff to patients. QUASI-EXPERIMENTS Quasi-experiments, like true experiments, involve the manipulation of an independent variable, that is, an intervention. However, quasi-experimental designs lack randomization to treatment groups, which characterizes true experiments, as shown in Figure 8-3. Quasi-Experimental Designs Quasi-experiments are not as powerful as experiments in establishing causal connections between interventions and outcomes. Before showing why

182

PART 3 Designs for Nursing Research

Is there an intervention (control over the independent variable)?

YES

NO

Nonexperimental Research

Is there random assignment to treatment groups?

NO

YES

Experimental Research

Are there efforts to compensate for the lack of random assignment?

NO

Preexperimental Research

YES

Quasi-experimental Research

this is so, it is useful to introduce some notation based on Campbell and Stanley’s (1963) classic monograph. Figure 8-4 presents a symbolic representation of a pretest—posttest experimental design. In this figure, R means random assignment to groups; O represents an observation (i.e., the collection of data on the dependent variable); and X stands for exposure to an intervention. Thus, the top line in this figure represents the experimental group that had subjects randomly assigned to it (R), had both a pretest (O1) and a posttest (O2), and has been exposed to an experimental intervention (X). The second row represents the control group, which differs from the experimental group only by absence of the treatment (X). We are now equipped to examine some quasi-experimental designs and to better understand their limitations relative to experimental designs.

F I G U R E 8 . 3 Characteristics of different quantitative research designs.

Nonequivalent Control Group Designs The most frequently used quasi-experimental design is the nonequivalent control group pretest— posttest design, which involves an experimental treatment and two groups of subjects observed before and after its implementation. Suppose, for example, we wished to study the effect of introducing primary nursing on staff morale in a large metropolitan hospital. Because the new system of nursing

F I G U R E 8 . 4 Symbolic representation of a pretest–posttest (before–after) experimental design.

CHAPTER 8 Designing Quantitative Studies

F I G U R E 8 . 5 Nonequivalent control group pretest–posttest design (quasiexperimental).

care delivery is being implemented throughout the hospital, randomization is not possible. Therefore, we decide to collect comparison data from nurses in another similar hospital that is not instituting primary nursing. Data on staff morale is collected in both hospitals before the change is made (the pretest) and again after the new system is implemented in the first hospital (the posttest). Figure 8-5 depicts this study symbolically. The top row is our experimental (primary nursing) hospital; the second row is the hospital using traditional nursing. A comparison of this diagram with the one in Figure 8-4 shows that they are identical, except that subjects have not been randomly assigned to treatment groups in the second diagram. The design in Figure 8-5 is the weaker of the two because it can no longer be assumed that the experimental and comparison groups are equivalent at the outset. Because there was no randomization, this study is quasi-experimental rather than experimental. The design is nevertheless strong, because the pretest data allow us to determine whether the groups had similar morale initially. If the comparison and experimental groups are similar on the pretest, we could be relatively confident that any posttest difference in self-reported morale was the result of the new system of nursing care. If the morale of the two groups is very different initially, however, it will be difficult to interpret any posttest differences, although there are statistical procedures that can help. Note that in quasi-experiments, the term comparison group is usually used in lieu of control group to refer to the group against which outcomes in the treatment group are evaluated. Now suppose we had been unable to collect pretest data. This design, diagramed in Figure 8-6, has a flaw that is difficult to remedy. We no longer have information about the initial equivalence of the two nursing staffs. If we find that staff morale in the experimental hospital is lower than that in the con-

183

F I G U R E 8 . 6 Nonequivalent control group only posttest design (preexperimental).

trol hospital at the posttest, can we conclude that the new method of delivering care caused a decline in staff morale? There could be alternative explanations for the posttest differences. In particular, it might be that the morale of the employees in the two hospitals differed even at the outset. Campbell and Stanley (1963) call the nonequivalent control group posttest-only design in Figure 8-6 preexperimental rather than quasi-experimental because of its fundamental weakness. Thus, although quasi-experiments lack the controlling properties of true experiments, the hallmark of quasi-experiments is the effort to introduce strategies to compensate for the absence of either randomization or control groups. Example of a nonequivalent control group pretest—posttest design: Johnson, Budz, Mackay, and Miller (1999) evaluated the effect of a nurse-delivered smoking cessation intervention on smoking status and smoking selfefficacy among patients hospitalized with cardiac disease. Experimental subjects were admitted to one cardiac unit, and comparison subjects were admitted to another. The researchers preferred this approach to randomization within units because information sharing among patients in the same unit could have contaminated treatment conditions. By collecting pretest data, the researchers learned that the two groups were comparable with regard to demographic characteristics and preintervention smoking histories. Time Series Designs In the designs just described, a control group was used but randomization was not, but some studies involving an intervention have neither. Let us suppose that a hospital is adopting a requirement that all its nurses accrue a certain number of continuing education units before being eligible for a promotion or raise. The nursing administrators want to assess the consequences of this mandate on turnover rate,

184

PART 3 Designs for Nursing Research

F I G U R E 8 . 7 One group pretest– posttest design (preexperimental).

absentee rate, number of raises and promotions awarded, and so on. For the purposes of this example, assume there is no other hospital that can serve as a good comparison for this study. In such a case, the only kind of comparison that can be made is a before—after contrast. If the requirement were inaugurated in January, one could compare the turnover rate, for example, for the 3-month period before the new rule with the turnover rate for the subsequent 3-month period. The schematic representation of such a study is shown in Figure 8-7. This one-group pretest—posttest design seems straightforward, but it has weaknesses. What if either of the 3-month periods is atypical, apart from the new regulation? What about the effects of any other hospital rules inaugurated during the same period? What about the effects of external factors that influence employment decisions, such as changes in the local economy? This preexperimental design cannot control these factors.* This one-group pretest—posttest design could be modified so that at least some alternative explanations for change in nurses’ turnover rate could be ruled out. One such design is the time series design, (sometimes referred to as the interrupted time series design) and is diagramed in Figure 8-8. In a time series design, information is collected over an extended period and an intervention is introduced during that period. In the figure, O1 through O4 represent four separate instances of data collection on a dependent variable before treatment; X represents the treatment (the introduction of the independent variable); and O5 through O8 represent four post-

*One-group before—after designs are not always unproductive. For example, if the intervention involved a brief teaching intervention, with baseline knowledge data obtained immediately before the intervention and posttest knowledge data collected immediately after it, it may be reasonable to conclude that the intervention caused gains in knowledge. This is because the intervention is the most plausible—and perhaps the only— explanation for knowledge gains.

treatment observations. In our present example, O1 might be the number of nurses who left the hospital in January through March in the year before the new continuing education rule, O2 the number of resignations in April through June, and so forth. After the rule is implemented, data on turnover are similarly collected for four consecutive 3-month periods, giving us observations O5 through O8. Even though the time series design does not eliminate all problems of interpreting changes in turnover rate, the extended time period strengthens the ability to attribute change to the intervention. Figure 8-9 demonstrates why this is so. The two diagrams (A and B) in the figure show two possible outcome patterns for eight turnover observations. The vertical dotted line in the center represents the timing of the continuing education rule. Patterns A and B both reflect a feature common to most time series studies—fluctuation from one data point to another. These fluctuations are normal. One would not expect that, if 48 nurses resigned from a hospital in a year, the resignations would be spaced evenly with 4 resignations per month. It is precisely because of these fluctuations that the design shown in Figure 8-7, with only one observation before and after the experimental treatment, is so weak. Let us compare the kind of interpretations that can be made for the outcomes reflected in Figure 89, patterns A and B. In both cases, the number of resignations increased between O4 and O5, that is, immediately after the new continuing education requirement. In B, however, the number of resignations fell at O6 and continued to fall at O7. The increase at O5 looks similar to other apparently haphazard fluctuations in the turnover rate at other periods. Therefore, it probably would be erroneous to conclude that the new rule affected resignations. In A, on the other hand, the number of resignations increases at O5 and remains relatively high for all subsequent observations. Of course, there may be other explanations for a change in turnover rate from one year to the next. The time series design, however, does permit us to rule out the possibility that the data reflect unstable measurements of resignations at only two points in time. If we had used the design in Figure 8-7, it would have been

CHAPTER 8 Designing Quantitative Studies

185

F I G U R E 8 . 8 Time series design (quasi-experimental).

analogous to obtaining the measurements at O4 and O5 of Figure 8-9 only. The outcomes in both A and B look similar at these two points in time. Yet the broader time perspective leads us to draw different conclusions about two patterns of outcomes.

A particularly powerful quasi-experimental design results when the time series and nonequivalent control group designs are combined, as diagramed in Figure 8-10. In the example just described, a time series nonequivalent control group design would involve collecting data over an extended period from both the hospital introducing the continuing education mandate and another hospital not imposing the mandate. This use of information from another hospital with similar characteristics would make any inferences regarding the effects of the mandate more convincing because trends influenced by external factors would presumably be observed in both groups.

Example of a time series design: Nahm and Poston (2000) used a time series design to assess the effect of an integrated point-ofcase computer system on the quality of nurses’ documentation. Measurements of the quality of documentation were made before the intervention was implemented, and again at 6-, 12-, and 18-months after implementation. The researchers found that quality of nursing documentation increased, and variability in charting decreased.

(Treatment)

Turnover rate

X

A

B O1

O2

O3

O4

O5

Data collection periods

F I G U R E 8 . 9 Two possible time series outcome patterns.

O6

O7

O8

186

PART 3 Designs for Nursing Research

F I G U R E 8 . 1 0 Time series nonequivalent control group design (quasi-experimental).

Example of a time series nonequivalent control group design: Song, Daly, Rudy, Douglas, and Dyer (1997) examined rates of absenteeism and turnover among nurses working in a nurse-managed special care unit compared with nurses working in a traditional intensive care unit. The two units were compared over a 4-year period. Numerous variations on the simple time series design are possible and are being used by nurse researchers. For example, additional evidence regarding the effects of a treatment can be achieved by instituting the treatment at several different points in time, strengthening the treatment over time, or instituting the treatment at one point in time and then withdrawing the treatment at a later point, sometimes with reinstitution of treatment. These three designs are diagramed in Figures 8-11 through 8-13. Clinical nurse researchers are often in a good position to use such time series designs because measures of patient functioning are usually routinely made at multiple points over an extended period. A particular application of a time series approach sometimes used in clinical studies is called single-subject experiments (sometimes referred to as N-of-1 studies). Single-subject studies use time series designs to gather information about an intervention based on the responses of a single patient (or a small number of patients) under controlled conditions. In the literature on single-subject methods, the most basic design involves a baseline phase of data gathering (A) and an intervention phase (B), yielding what is referred to as an AB design (the design diagrammed in Fig. 8-8). If the treatment is withdrawn, it would be an ABA design; and if a withdrawn treatment is reinstituted (as diagramed in

Fig. 8-13) it would be an ABAB design. Portney and Watkins (2000) offer valuable guidance about single-subject studies in clinical settings. Example of a single-subject design: Landolt, Marti, Widner, and Meuli (2002) used a single-subject experiment (with multiple patients) to test whether cartoon movie viewing is effective in decreasing burned children’s pain behavior. Experimental and Comparison Conditions Researchers using a quasi-experimental approach, like those adopting an experimental design, should strive to develop strong interventions that provide an opportunity for a fair test, and must develop protocols documenting what the interventions entail. Researchers need to be especially careful in understanding and documenting the counterfactual in quasi-experiments. In the case of nonequivalent control group designs, this means understanding the conditions to which the comparison group is exposed. In our example of using a hospital with traditional nursing systems as a comparison for the new primary nursing system, the nature of that traditional system should be fully understood. In time series designs, the counterfactual is the condition existing before implementing the intervention. Strengths and Limitations of Quasi-Experiments A great strength of quasi-experiments is that they are practical. In the real world, it may be difficult, if not impossible, to conduct true experiments.

F I G U R E 8 . 1 1 Time series with multiple institutions of treatment (quasi-experimental).

CHAPTER 8 Designing Quantitative Studies

187

F I G U R E 8 . 1 2 Time series with intensified treatment (quasi-experimental).

Nursing research usually occurs in real-life settings, where it is difficult to deliver an innovative treatment randomly to some people but not to others. Quasi-experimental designs introduce some research control when full experimental rigor is not possible. Researchers using quasi-experimental designs need, however, to be acquainted with their weaknesses, and take these weaknesses into account in interpreting results. When a quasi-experimental design is used, there may be several rival hypotheses competing with the experimental manipulation as explanations for the results. (This is discussed further in Chapter 9.) Take as an example the case in which we administer certain medications to a group of babies whose mothers are heroin addicts to assess whether this treatment results in a weight gain in these typically lowbirth-weight babies. If we use no comparison group or if we use a nonequivalent control group and then observe a weight gain, we must ask the questions: Is it plausible that some other factor caused or influenced the gain? Is it plausible that pretreatment differences between the experimental and comparison groups of babies resulted in differential gain? Is it plausible that the babies gained the weight simply as a result of maturation? If the answer is “yes” to any of these questions, then the inferences that can be made about the effect of the experimental treatment are weakened considerably. The plausibility of any one threat cannot, of course, be answered unequivocally. It is usually a situation in which judgment must be exercised. Because the conclusions from quasi-experiments ultimately depend in part on human judgment, rather than on more objective criteria, cause-and-effect inferences may be less convincing.

Research Example of a Quasi-Experimental Study Bull, Hansen, and Gross (2000) used a strong quasiexperimental design to evaluate the effects of implementing a professional–patient partnership model of discharge planning with elders hospitalized with heart failure. The intervention was designed to facilitate identification of the elderly patients’ needs for followup care and to identify those requiring more in-depth assessments. The discharge planning model was implemented in one hospital, and a nonequivalent control group was secured through a different hospital that did not adopt the new model. The two hospitals were matched in terms of size, type, and the discharge planning practices used in the hospitals’ cardiac units. Data on a range of measures (including health status, client satisfaction, and health locus of control) were collected from patients and caregivers before the intervention, and then again at 2 weeks and 2 months postdischarge. Moreover, data were gathered in both hospitals both before the intervention and then again after it was implemented, thus combining features of a nonequivalent control group design with that of a mini—time series. Unfortunately, however, data from the comparison group in the second hospital were contaminated by the fact that this hospital introduced an innovation that weakened comparisons. Thus, postintervention data from the experimental group were compared with two rather than three comparison groups, as originally planned: comparison groups from both hospitals before the new model being implemented. The three groups of elders were found to be comparable in terms of demographics and predischarge measures of health. The two comparison groups were combined for the analysis because of the absence of any significant differences. The findings indicated that elders in the treatment group felt better prepared to manage care, reported more continuity of information about care management,

F I G U R E 8 . 1 3 Time series with withdrawn and reinstituted treatment (quasi-experimental).

188

PART 3 Designs for Nursing Research

and felt they were in better health than elders in the comparison group. Moreover, elders in the treatment group who were readmitted to the hospital spent fewer days in the hospital than their comparison group counterparts. The strength of the design makes it possible to draw inferences about the effectiveness of the innovative discharge planning model, even in the absence of randomization.

N O N E X P E R I M E N TA L RESEARCH Many research problems cannot be addressed with an experimental or quasi-experimental design. For example, suppose we were interested in studying the effect of widowhood on health status. Our independent variable is widowhood versus nonwidowhood. Clearly, we cannot manipulate widowhood; people lose their spouses by a process that is neither random nor subject to research control. Thus, we would have to proceed by taking two groups (widows and nonwidows) as they naturally occur and comparing them in terms of health status. Reasons for Undertaking Nonexperimental Research Most studies involving human subjects, including nursing studies, are nonexperimental. One reason for using a nonexperimental design is that a vast number of human characteristics are inherently not subject to experimental manipulation (e.g., blood type, personality, health beliefs, medical diagnosis); the effects of these characteristics on other phenomena cannot be studied experimentally. A second issue is that in nursing research, as in other fields, there are many variables that could technically be manipulated but could not be manipulated ethically. If manipulating the independent variable could cause physical or mental harm to subjects, then the variable should not be controlled experimentally. For example, if we were studying the effect of prenatal care on infant mortality, it would be unethical to provide such care to one group of pregnant women while deliberately depriving a second group. We would need to locate a naturally occurring group of pregnant women who

had not received prenatal care. Their birth outcomes could then be compared with those of women who had received appropriate care. The problem, however, is that the two groups of women are likely to differ in terms of many other characteristics, such as age, education, nutrition, and health, any of which individually or in combination could affect infant mortality, independent of the absence or presence of prenatal care. This is precisely why experimental designs are so strong in demonstrating cause-and-effect relationships. Third, there are many research situations in which it is simply not practical to conduct a true experiment. Constraints might involve insufficient time, lack of administrative approval, excessive inconvenience to patients or staff, or lack of adequate funds. Fourth, there are some research questions for which an experimental design is not appropriate. This is especially true for descriptive studies, which seek to document the characteristics, prevalence, intensity, or full nature of phenomena. As we discuss in Chapter 11, qualitative studies are nonexperimental. Manipulation is neither attempted nor considered desirable; the emphasis is on the normal experiences of humans. Finally, nonexperimental research is usually needed before an experimental study can be planned. Experimental interventions are developed on the basis of nonexperimental research documenting the scope of a problem and describing critical relationships between relevant variables. Ex Post Facto/Correlational Research There are two broad classes of nonexperimental research, the first of which has been called ex post facto research. The literal translation of the Latin term ex post facto is “from after the fact.” This means that the study has been conducted after variations in the independent variable have occurred. Ex post facto research attempts to understand relationships among phenomena as they naturally occur, without any intervention. Ex post facto research is more often referred to as correlational research. Basically, a correlation is an interrelationship or

CHAPTER 8 Designing Quantitative Studies

189

F I G U R E 8 . 1 4 Schematic diagram comparing nonequivalent control group and ex post facto designs.

association between two variables, that is, a tendency for variation in one variable to be related to variation in another. For example, in human adults, height and weight are correlated because there is a tendency for taller people to weigh more than shorter people. Correlational studies often share some structural characteristics with experimental, quasiexperimental, and preexperimental research. If we use the notation scheme described in the previous section to represent symbolically the hypothetical study of the effects of widowhood, we find that it bears a strong resemblance to the nonequivalent control group posttest-only design. Both designs are presented in Figure 8-14. As these diagrams show, the preexperimental design is distinguished from the correlational study only by the presence of an X, the introduction of an intervention. The purpose of correlational research, like experimental research, is to understand relationships among variables. It is, however, riskier to infer causal relationships in correlational research because of the lack of control over the independent variable. In experiments, investigators make a prediction that deliberate variation of X, the independent variable, will result in a change to Y, the dependent variable. For example, they might predict that if a new medication is administered, patient improvement will result. Experimenters have direct control over the X; the experimental treatment can be administered to some and withheld from others, and the two groups can be equalized with respect to everything except the independent variable through randomization. In correlational research, on the other hand, investigators do not control the independent variable, which has already occurred. The examination of the independent variable—the presumed causative factor—is done after the fact. As a result, attempts to draw any cause-and-effect conclusions

are problematic. For example, we might hypothesize that there is a correlation between smoking and lung cancer, and empirical data would likely corroborate this expectation: smokers are more likely than nonsmokers to develop lung cancer. The inference we would like to make is that cigarette smoking causes cancer. This kind of inference, however, is subject to a fallacy called post hoc ergo propter (“after this, therefore caused by this”). The fallacy lies in the assumption that one thing has caused another merely because it occurred before the other. To illustrate why a cause-and-effect conclusion might not be warranted, let us assume (strictly for the sake of an example) that there is a preponderance of cigarette smokers in urban areas, and people in rural areas are largely nonsmokers. Let us further assume that lung cancer is actually caused by poor environmental conditions in cities. Therefore, we would be incorrect to conclude that cigarette smoking causes lung cancer, despite the strong relationship shown to exist between the two variables. This is because there is also a strong relationship between cigarette smoking and the “real” causative agent, living in a polluted environment. Of course, cigarette smoking/lung cancer studies in reality have been replicated in so many different places with so many different groups of people that causal inferences are justified. This hypothetical example illustrates a famous research dictum: Correlation does not prove causation. The mere existence of a relationship— even a strong one—between variables is not enough to warrant the conclusion that one variable caused the other. Although correlational studies are inherently weaker than experimental studies in elucidating cause-and-effect relationships, different designs offer different degrees of supportive evidence.

190

PART 3 Designs for Nursing Research

Retrospective Designs Studies with a retrospective design are ones in which a phenomenon existing in the present is linked to phenomena that occurred in the past, before the study was initiated. That is, the researcher is interested in a present outcome and attempts to determine antecedent factors that caused it. Most of the early epidemiologic studies of the link between cigarette smoking and lung cancer were retrospective. In such a study, the researcher begins with groups of people with and without lung cancer (the dependent variable). The researcher then looks for differences between the two groups in antecedent behaviors or conditions. Retrospective studies are often cross-sectional, with data on both the dependent and independent variables collected once, simultaneously. Researchers can sometimes strengthen a retrospective design by taking certain steps. For example, one type of retrospective design, referred to as a case—control design, involves the comparison of cases (subjects with a certain illness or condition, such as lung cancer victims) with controls (e.g., people without lung cancer). In conducting a strong case—control study, researchers find the cases and obtain from them (or about them, if records are available) information about the history of the presumed cause. Then the researchers must find controls without the disease or condition who are as similar as possible to the cases with regard to key extraneous variables (e.g., age, gender) and also obtain historical information about the presumed cause. If controls are well chosen, the only difference between them and the cases is exposure to the presumed cause. Researchers sometimes use matching or other techniques (described in Chapter 9) to control for extraneous variables. To the degree that researchers can demonstrate comparability between cases and controls with regard to extraneous traits, inferences regarding the presumed cause of the disease are enhanced. Example of a retrospective study: Heitkemper, Jarrett, Taylor, Walker, Landenburger, and Bond (2001) used a retrospective design in their study of factors contributing to

the onset of irritable bowel syndrome (IBS). They compared samples of women with and without IBS in terms of their history of sexual and physical abuse, and found that abusive experiences were more prevalent among women with IBS. Prospective Nonexperimental Designs A nonexperimental study with a prospective design (sometimes called a prospective cohort design) starts with a presumed cause and then goes forward in time to the presumed effect. For example, we might want to test the hypothesis that the incidence of rubella during pregnancy (the independent variable) is related to infant abnormalities (the dependent variable). To test this hypothesis prospectively, we would begin with a sample of pregnant women, including some who contracted rubella during their pregnancy and others who did not. The subsequent occurrence of congenital anomalies would be assessed for all subjects, and we would examine whether women with rubella were more likely than other women to bear malformed infants. Prospective designs are often longitudinal, but may also be cross-sectional (from the subjects’ point of view) if reliable information about the independent variable is available in records or existing data sources. TIP: Not all longitudinal studies are prospective, because sometimes the independent variable has occurred long before the initial wave of data collection. And not all prospective studies are longitudinal in the classic sense. For example, an experimental study that collects data at 2, 4, and 6 hours after an intervention would be considered prospective but not longitudinal (i.e., data are not collected over an extended period of time.) Prospective studies are more costly than retrospective studies. For one thing, a substantial follow-up period may be necessary before the dependent variable manifests itself, as is the case in prospective studies of cigarette smoking and lung cancer. Also, prospective designs may require large samples, particularly if the dependent variable of interest is rare, as in the example of malformations associated with maternal rubella. Another issue is

CHAPTER 8 Designing Quantitative Studies

that in a good prospective study, researchers take steps to confirm that all subjects are free from the effect (e.g., the disease) at the time the independent variable is measured, and this may in some cases be difficult or expensive to do. For example, in prospective smoking/lung cancer studies, lung cancer may be present initially but not yet diagnosed. Despite these issues, prospective studies are considerably stronger than retrospective studies. In particular, any ambiguity about whether the presumed cause occurred before the effect is resolved in prospective research if the researcher has confirmed the initial absence of the effect. In addition, samples are more likely to be representative, and investigators may be in a position to impose controls to rule out competing explanations for the results. Some prospective studies are exploratory. That is, the researcher measures a wide range of possible “causes” at one point in time, and then examines an outcome of interest at a later point (e.g., length of stay in hospital). Such studies are usually stronger than retrospective studies if it can be determined that the outcome was not present initially because time sequences are clear. However, they are not as powerful as prospective studies that involve specific a priori hypotheses and the comparison of cohorts known to differ on a presumed cause. Researchers doing exploratory retrospective or prospective studies are sometimes accused of going on “fishing expeditions” that can lead to erroneous conclusions because of spurious or idiosyncratic relationships in a particular sample of subjects. Example of a prospective nonexperimental study: Brook, Sherman, Malen, and Kollef (2000) conducted a prospective cohort study to examine clinical and cost outcomes of early versus late tracheostomy in patients who require prolonged mechanical ventilation. Early tracheostomy was found to be associated with shorter lengths of hospital stay and lower hospital costs. Natural Experiments Researchers are sometimes able to study the outcomes of a “natural experiment” in which a

191

group exposed to natural or other phenomena that have important health consequences are compared with a nonexposed group. Such natural experiments are nonexperimental because the researcher does not intervene but simply observes the outcome of an external event or circumstance. However, they are called “natural experiments” when people are affected essentially at random. For example, the psychological well-being of people living in a community struck with a natural disaster (e.g., a hurricane) could be compared with the well-being of people living in a similar but unaffected community to determine the toll exacted by the disaster (the independent variable). Note that the independent variable does not need to be a “natural” phenomenon. It could, for example, be an act of terrorism. Moreover, the groups being compared do not need to be different people—if pre-event measures had been obtained, before—after comparisons might be profitable. Natural experiments can offer strong evidence about the effect of an independent variable on outcomes of interest if the comparison is carefully selected to achieve equivalence of groups being compared with regard to everything but the event. Example of a natural experiment: Keane, Jepson, Pickett, Robinson, and McCorkle (1996) studied the experiences of fire survivors, and attributed high levels of distress to those experiences 14 weeks after the fires, even among victims who did not sustain physical injury. Although there were no prefire measures of distress, it seems reasonable to attribute much of the stress to the “intervention” (the fires). Path Analytic Studies Researchers interested in testing theories of causation based on nonexperimental data are increasingly using a technique known as path analysis (or similar techniques). Using sophisticated statistical procedures, the researcher tests a hypothesized causal chain among a set of independent variables, mediating variables, and a dependent variable. Path analytic procedures, which are described more fully in Chapter 21, allow researchers to test whether

192

PART 3 Designs for Nursing Research

nonexperimental data conform sufficiently to the underlying model to justify causal inferences.

category, women usually considered themselves larger than they were.

Example of a path analytic study: Horsburgh, Beansland, Locking-Cusolito, Howe and Watson (2000) used path analysis to test a model to predict self-care in adults awaiting renal transplantation in Ontario. Their analyses tested hypothesized causal pathways between personality traits, health status, self-care abilities, and self-care behavior.

Univariate Descriptive Studies Some descriptive studies are undertaken to describe the frequency of occurrence of a behavior or condition rather than to study relationships. For example, an investigator may wish to describe the health care and nutritional practices of pregnant teenagers. Univariate descriptive studies are not necessarily focused on only one variable. For example, a researcher might be interested in women’s experiences during menopause. The study might describe the frequency of various symptoms, the average age at menopause, the percentage of women seeking formal health care, and the percentage of women using medications to alleviate symptoms. There are multiple variables in this study, but the primary purpose is to describe the status of each and not to relate them to one another. Two types of descriptive study from the field of epidemiology are especially worth noting. Prevalence studies are done to determine the prevalence rate of some condition (e.g., a disease or a behavior, such as smoking) at a particular point in time. Prevalence studies rely on crosssectional designs in which data are obtained from the population at risk of the condition. The researcher takes a “snapshot” of the population at risk to determine the extent to which the condition of interest is present. The formula for a point prevalence rate (PR) is:

Descriptive Research The second broad class of nonexperimental studies is descriptive research. The purpose of descriptive studies is to observe, describe, and document aspects of a situation as it naturally occurs and sometimes to serve as a starting point for hypothesis generation or theory development. Descriptive Correlational Studies Although researchers often focus on understanding the causes of behaviors, conditions, and situations, sometimes they can do little more than describe relationships without comprehending causal pathways. Many research problems are cast in noncausal terms. We ask, for example, whether men are less likely than women to bond with their newborn infants, not whether a particular configuration of sex chromosomes caused differences in parental attachment. Unlike other types of correlational research—such as the cigarette smoking and lung cancer investigations—the aim of descriptive correlational research is to describe the relationship among variables rather than to infer cause-and-effect relationships. Descriptive correlational studies are usually cross-sectional. Example of a descriptive correlational study: Morin, Brogan, and Flavin (2002) described the relationship between body image perceptions of postpartum African-American women on the one hand, and their weight (based on the body mass index) on the other. Irrespective of body mass

Number of cases with the condition or disease at a given point in time !K Number in the population at risk of being a case K is the number of people for whom we want to have the rate established (e.g., per 100 or per 1000 population). When data are obtained from a sample (as would usually be the case), the denominator is the size of the sample, and the numerator is the number of cases with the condition, as identified in the study.

CHAPTER 8 Designing Quantitative Studies

193

If we sampled 500 adults aged 21 years and older living in a community, administered a measure of depression, and found that 80 people met the criteria for clinical depression, then the estimated point prevalence rate of clinical depression per 100 adults in that community would be 16 per 100. Incidence studies are used to measure the frequency of developing new cases. Longitudinal designs are needed to determine incidence because the researcher must first establish who is at risk of becoming a new case—that is, who is free of the condition at the outset. The formula for an incidence rate (IR) is:

Example of an incidence and prevalence study: Whittington, Patrick, and Roberts (2000) conducted a national study involving 116 acute care facilities in 34 states to determine the incidence and prevalence of pressure ulcers. Prevalence for 17,650 patients in medical-surgical or intensive care units was measured during a 24-hour period at each facility. Incidence was measured over the average length of hospital stay for each facility.

Number of new cases with the condition or disease over a given time period !K Number at risk of becoming a new case (free of the condition at the outset)

The quality of a study is not necessarily related to its approach; there are many excellent nonexperimental studies as well as flawed experiments. Nevertheless, nonexperimental studies have several drawbacks, and we focus here on the weaknesses of correlational studies.

If we continued with our previous example, suppose in October, 2001 we found that 80 in a sample of 500 people were clinically depressed (PR " 16 per 100). To determine the 1-year incidence rate, we would reassess the sample in October, 2002. Suppose that, of the 420 previously deemed not to be clinically depressed in 2001, 21 were now found to meet the criteria for depression. In this case, the estimated 1-year incidence rate would be 5 per 100 ((21 # 420) ! 100 " 5). Prevalence and incidence rates can be calculated for subgroups of the population (e.g., for men versus women). When this is done, it is possible to calculate another important descriptive index. Relative risk is an estimate of risk of “caseness” in one group compared with another. Relative risk is computed by dividing the rate for one group by the rate for another. Suppose we found that the 1-year incidence rate for depression was 6 per 100 women and 4 per 100 men. Women’s relative risk for developing depression over the 1-year period would be 1.5, that is, women would be estimated to be 1.5 times more likely to develop depression than men. Relative risk is an important index in determining the contribution of risk factors to a disease or condition (e.g., by comparing the relative risk for lung cancer for smokers versus nonsmokers).

Strengths and Limitations of Correlational Research

Limitations of Correlational Research As already noted, the major disadvantage of nonexperimental studies is that, relative to experimental and quasi-experimental research, they are weak in their ability to reveal causal relationships. Correlational studies are susceptible to faulty interpretations. This situation arises because the researcher works with preexisting groups that were not formed at random, but rather by a self-selecting process. Kerlinger and Lee (2000) offer the following description of self-selection: Self-selection occurs when the members of the groups being studied are in the groups, in part, because they differentially possess traits or characteristics extraneous to the research problem, characteristics that possibly influence or are otherwise related to the variables of the research problem (p. 560).

A researcher doing a correlational study, unlike an experimental study, cannot assume that the groups being compared are similar before the occurrence of the independent variable. Thus, preexisting differences may be a plausible alternative explanation for any group differences on the dependent variable. To illustrate this problem, let us consider a hypothetical study in which a researcher is examining

194

PART 3 Designs for Nursing Research

the relationship between type of nursing program a person attended (the independent variable) and job satisfaction 1 year after graduation. If the investigator finds that diploma school graduates are more satisfied with their work than baccalaureate graduates, the conclusion that the diploma school program leads to increased job satisfaction may or may not be accurate. The students in the two programs undoubtedly differed to begin with in terms of a number of important characteristics, such as personality, career goals, personal values, and so forth. Students selected themselves into one of the two programs and selection traits may have caused different job expectations and satisfactions. The difficulty of interpreting correlational findings stems from the fact that, in the real world, behaviors, states, attitudes, and characteristics are interrelated (correlated) in complex ways. Another example may help to clarify the problems of interpreting results from correlational studies. Suppose we conducted a cross-sectional study that examined the relationship between level of depression in cancer patients and their social support (i.e., assistance and emotional sustenance through a social network). We hypothesize that social support (the independent variable) affects levels of depression (the dependent variable). Suppose we find that the patients without social support are significantly more depressed than the patients with adequate social support. We could interpret this finding to mean that people’s emotional states are influenced by the adequacy of their social supports. This relationship is diagrammed in Figure 8-15A. There are, however,

A

alternative explanations. Perhaps there is a third variable that influences both social support and depression, such as the patients’ family configuration (e.g., whether they are married or have children). It may be that the availability or quantity of significant others is a powerful influence on how depressed cancer patients feel and on the quality of their social support. This set of relationships is diagramed in Figure 8-15B. A third possibility may be reversed causality, as shown in Figure 8-15C. Depressed cancer patients may find it more difficult to elicit needed social support from others than patients who are more cheerful or sociable. In this interpretation, the person’s depression causes the amount of received social support, and not the other way around. You may be able to invent other alternatives. The point is that interpretations of most correlational results should be considered tentative, particularly if the research has no theoretical basis. Strengths of Correlational Research Earlier, we discussed constraints that limit the possibility of applying experimental designs to some research problems. Correlational research will continue to play a crucial role in nursing, medical, and social science research precisely because many interesting problems in those fields are not amenable to experimentation. Despite our emphasis on causal relationships, it has already been noted that some kinds of research, such as descriptive research, do not focus on understanding causal networks. Furthermore, if the study is testing a causal hypothesis that has

X

Y

Social Support

Depression

Depression

B

Family Configuration Social Support

C

Depression

Social Support

F I G U R E 8 . 1 5 Alternative explanations for relationship between depression and social support in cancer patients.

CHAPTER 8 Designing Quantitative Studies

been deduced from an established theory, causal inferences may be possible, especially if strong designs (e.g., a prospective design) have been used. Correlational research is often an efficient means of collecting a large amount of data about a problem. For example, it would be possible to collect extensive information about the health histories and eating habits of a large number of individuals. Researchers could then examine which health problems were associated with which diets, and could thus discover a large number of interrelationships in a relatively short amount of time. By contrast, an experimenter looks at only a few variables at a time. One experiment might manipulate foods high in cholesterol, whereas another might manipulate protein consumption, for example. Finally, correlational research is often strong in realism and therefore has an intrinsic appeal for solving practical problems. Unlike many experimental studies, correlational research is seldom criticized for its artificiality. TIP: It is usually advantageous to design a study with as many relevant comparisons as possible. Preexperimental designs are weak in part because the comparative information they yield is limited. In nonexperimental studies, multiple comparison groups can be effective in dealing with self-selection, especially if the comparison groups are selected to address competing biases. For example, in case—control studies of patients with lung cancer, one comparison group could comprise people with a respiratory disease other than lung cancer and a second could comprise those with no respiratory disorder. Research Example of a Nonexperimental Study Faulkner, Hathaway, Milstead, and Burghen (2001) noted that there was little evidence about the onset or trajectory of cardiovascular autonomic deterioration in people with type 1 diabetes. To help fill this void, they conducted a cross-sectional nonexperimental study to examine whether age (adolescent versus adult) and diabetes status were associated with differences in measures of heart rate variability. Their descriptive correlational design involved four groups of subjects: adults

195

with type 1 diabetes and renal failure, healthy similaraged adult control subjects, adolescents with type 1 diabetes, and healthy adolescent control subjects. Most of the adult patients with diabetes had been diagnosed during childhood or adolescence. The researchers conducted reflex tests to measure subjects’ short-term R-R (beat-to-beat) heart rate variability, with deep breathing and with the Valsalva maneuver, in a temperature-controlled and noisecontrolled laboratory environment. In addition, 24hour ambulatory heart rate monitoring with power spectral analysis was obtained. The researchers then examined differences among subjects in the four groups in evoked, frequency, and time domain measures for heart rate variability. The results indicated that adult patients with type 1 diabetes had significantly poorer heart rate variability measures than subjects in the other three groups. Adult control subjects had significantly lower average values than either of the two adolescent groups. Although most long-term R-R variability measures were lower in adolescents with diabetes than those without the disease, only one measure was significantly lower. Nevertheless, the researchers noted that the trends were important because they provide evidence that autonomic neuropathy begins relatively early in the course of diabetes. In this example, the researchers were able to describe differences in heart rate variability in relation to two independent variables, age and diabetes status. Neither of these variables could have been experimentally manipulated and so a nonexperimental study was required.

DESIGNS AND RESEARCH EVIDENCE Evidence for nursing practice depends on descriptive, correlational, and experimental research. There is often a logical progression to knowledge expansion that begins with rich description, including description from qualitative research. Descriptive studies can be invaluable in documenting the prevalence, nature, and intensity of health-related conditions and behaviors, and are critical in the development of effective interventions. Descriptive studies that contribute to the development of descriptive theories can make a particularly valuable contribution.

196

PART 3 Designs for Nursing Research

Strongest design

True experimental design

QuasiPreexperimental experimental design design

Weakest design

Natural experiment

Path analytic design

Prospective correlational design

Retrospective correlational design

Descriptive designs

F I G U R E 8 . 1 6 Continuum of designs for inferring causality.

Correlational studies are often undertaken in the next phase of developing a knowledge base. Exploratory retrospective studies may pave the way for more rigorous case—control studies, and for prospective studies. As the knowledge base builds, conceptual models may be developed and tested using path analytic designs and other nonexperimental theory-testing strategies. These studies can provide hints about how to structure an intervention, who can most profit from it, and when it can best be instituted. Thus, the next important phase is to develop interventions to improve health outcomes. During the design and early testing of an intervention, it is often appropriate to conduct a pilot study (sometimes called a feasibility study). Rigorous experimental tests of interventions are expensive, and so it is often useful to begin with a small-scale test to determine the feasibility of a larger study and to ascertain whether a proposed approach shows promise. Feasibility can be assessed in terms of various considerations, including acceptability of the intervention to clients and to staff administering it; costs of the intervention; and ease of integrating it into clinical settings. Pilot studies can provide clues about the likely success of the intervention, and about ways in which the intervention can be strengthened or modified. Pilot studies also provide methodologic guidance (e.g., in determining sample size requirements for a full test, or strategies for recruiting subjects). TIP: A pilot study is not the same as a small-scale study. The term pilot study has been misused by some researchers who appear to use it as an excuse for not using a bigger sample (King, 2001). The purpose of a pilot study is not so much to test research hypotheses, but rather to test

protocols, data collection instruments, sample recruitment strategies, and other aspects of a study in preparation for a larger study. Example of a pilot study: Zust (2000) described a pilot study that explored the feasibility of a future large-scale experiment to test the effectiveness of a 20-week cognitive therapy intervention on depressed battered women. The pilot test used an experimental before—after design with a sample of 18 rural women. The results supported an expanded test of the intervention. The progression of evidence-building from descriptive studies to rigorous experimental ones is related to the ability of designs to reveal causal relationships, which may viewed on a continuum. True experimental designs are at one end of that continuum, and descriptive studies are at the other end, as shown in Figure 8-16. S U M M A RY P O I N T S • The research design is the researcher’s overall plan for answering the research question. In quantitative studies, the design indicates whether there is an intervention; the nature of any comparisons; methods used to control extraneous variables; timing and location of data collection; and information to be provided to subjects. • Between-subjects designs, in which different groups of people are compared, contrast with within-subjects designs that involve comparisons of the same subjects. • Cross-sectional designs involve the collection of data at one point in time, whereas longitudinal

CHAPTER 8 Designing Quantitative Studies

designs involve data collection at two or more points over an extended period. • Longitudinal studies, which are used to study trends, changes, or development over time, include trend studies (multiple points of data collection with different samples from the same population), and panel studies and follow-up studies (which gather data from the same subjects more than once). • Longitudinal studies are typically expensive, time-consuming, and subject to the risk of attrition (loss of participants over time), but can produce extremely valuable information. • Experiments involve manipulation (the researcher manipulates the independent variable by introducing a treatment or intervention); control (including the use of a control group that is not given the intervention and is compared to the experimental group); and randomization or random assignment (with subjects allocated to experimental and control groups at random to make the groups comparable at the outset). • Random assignment is done by methods that give every subject an equal chance of being included in any group, such as by flipping a coin or using a table of random numbers. Randomization is the most reliable method for equating groups on all possible characteristics that could affect study outcomes. • A posttest-only (or after-only) design involves collecting data only once—after the introduction of the treatment. • In a pretest—posttest (or before—after) design, data are collected both before and after the experimental manipulation, thereby permitting an analysis of change. • Factorial designs, in which two or more variables are manipulated simultaneously, allow researchers to test both main effects (effects from the experimentally manipulated variables) and interaction effects (effects resulting from combining the treatments). • In a randomized block design, subjects are randomly assigned to groups in different levels of a blocking variable that is not manipulated (e.g., gender).

197

• In a crossover (or repeated measures) design, research subjects are exposed to more than one experimental condition and serve as their own controls. • Researchers can expose the control group to various conditions, including no treatment; an alternative treatment; a placebo or pseudointervention; standard treatment; different doses of the treatment; and delayed treatment. • True experiments are considered by many to be the ideal of science because they come closer than any other type of research approach to meeting the criteria for inferring causal relationships. • Quasi-experimental designs involve manipulation but lack a comparison group or randomization. In quasi-experiments, control strategies are introduced to compensate for these missing components. By contrast, preexperimental designs have no such safeguards. • The nonequivalent control group before—after design involves the use of a comparison group that was not created through random assignment and the collection of pretreatment data that permit an assessment of initial group equivalence. • In a time series design, there is no comparison group; information on the dependent variable is collected over a period of time before and after the treatment. • In evaluating the results of quasi-experiments, it is important to ask whether it is plausible that factors other than the intervention caused or affected the outcomes (i.e., whether there are rival hypotheses for explaining the results). • Nonexperimental research includes descriptive research—studies that summarize the status of phenomena—and ex post facto (or correlational) studies that examine relationships among variables but involve no manipulation of the independent variable. • Nonexperimental research is undertaken because (1) a number of independent variables, such as height and gender, are not amenable to randomization; (2) some variables are technically manipulable but cannot ethically be manipulated; (3) there are sometimes practical constraints to manipulating variables; and (4) researchers

198

PART 3 Designs for Nursing Research

sometimes deliberately choose not to manipulate variables, to achieve a more realistic understanding of phenomena as they exist in naturalistic settings. • There are various designs for correlational studies, including retrospective designs (which begin with the outcome and look back in time for antecedent causes); case—control studies (retrospective studies that test hypotheses about antecedent causes by comparing cases that have a disease or condition with controls that do not); prospective designs (longitudinal studies that begin with a presumed cause and look forward in time for its effect); natural experiments (comparisons of groups in which one group is affected by a seemingly random event, such as a disaster, and the other is not); and path analytic studies (which test causal models developed on the basis of theory). • Descriptive studies include both descriptive correlational studies (which describe how phenomena are interrelated without inferring causality) and univariate descriptive studies (which examine the occurrence, frequency, or average value of variables without examining interrelationships). • Descriptive studies include prevalence studies that document the prevalence rate of some condition at a particular point in time, and incidence studies that document the frequency of new cases, over a given time period. When the incidence rates for two groups are determined, it is possible to compute the relative risk of “caseness” for the two. • The primary weakness of ex post facto or correlational studies is that they can harbor biases due to self-selection into groups being compared. STUDY ACTIVITIES Chapter 8 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing the concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Suppose you wanted to study how nurses’ attitudes toward death change in relation to years

of nursing experience. Design a cross-sectional study to research this question, specifying the samples that you would want to include. Now design a longitudinal study to research the same problem. Identify the problems and strengths of each approach. 2. A researcher is interested in studying the effect of sensitivity training for nurses on their behavior in crisis intervention situations. Describe how you would set up an experiment to study this. Now describe two quasiexperimental or preexperimental designs that could be used to study the same problem. Discuss what the weaknesses of each would be. 3. Assume that you have 10 individuals—Z, Y, X, W, V, U, T, S, R, and Q—who are going to participate in an experiment you are conducting. Using a table of random numbers, assign five individuals to group I and five to group II. Then randomly assign the groups to an experimental or control treatment. 4. Using the notation presented in Figures 8-5 to 8-13, diagram a few of the research examples described in the text that are not already shown. 5. A nurse researcher is interested in studying the success of several different approaches to feeding patients with dysphagia. Can the researcher use a correlational design to examine this problem? Why or why not? Could an experimental or quasi-experimental approach be used? How? 6. A nurse researcher is planning to investigate the relationship between the level of economic disadvantage of hospitalized children and the frequency and content of child-initiated communications with the nursing staff. Which is the independent variable, and which is the dependent variable? Would you classify this research as basically experimental or correlational, or could both approaches be used? SUGGESTED READINGS Methodologic References Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Chicago: Rand McNally.

CHAPTER 8 Designing Quantitative Studies Christensen, L. M. (1991). Experimental methodology (5th ed.). Boston: Allyn and Bacon. Clinton, J., Beck, R., Radjenovic, D., Taylor, L., Westlake, S., & Wilson, S. E. (1986). Time series designs in clinical nursing research. Nursing Research, 35, 188–191. Conn, V. S., Rantz., M. J., Wipke-Tevis, D. D., & Maas, M. L. (2001). Designing effective nursing interventions. Research in Nursing & Health, 24, 433–442. Cook, T., & Campbell, D. T. (1979). Quasi-experimental design and analysis issues for field settings. Chicago: Rand McNally. Given, B. A., Keilman, L. J., Collins, C., & Given, C. W. (1990). Strategies to minimize attrition in longitudinal studies. Nursing Research, 39, 184–186. Hauck, W. W., Gilliss, C. L., Donner, A., & Gortner, S. (1991). Randomization by cluster. Nursing Research, 40, 356–358. Kerlinger, F. N., & Lee, H. B. (2000). Foundations of behavioral research (4th ed.). Orlando, FL: Harcourt College Publishers. King, K. M. (2001). The problem of under-powering in nursing research. Western Journal of Nursing Research, 23, 334–335. Kirchoff, K. T., & Dille, C. A. (1994). Issues in intervention research: Maintaining integrity. Applied Nursing Research, 7, 32–37. Lazarsfeld, P. (1955). Foreword. In H. Hyman (Ed.), Survey design and analysis. New York: The Free Press. Lipsey, M. W. (1990). Design sensitivity: Statistical power for experimental research. Newbury Park, CA: Sage. McGuire, D.B., DeLoney, V., Yeager, K., Owen, D., Peterson, D., Lin, L., & Webster, J. (2000). Maintaining study validity in a changing clinical environment. Nursing Research, 49, 231–235. Montgomery, D. C. (2000). Design and analysis of experiments (5th ed.). New York: John Wiley & Sons. Motzer, S. A., Moseley, J. R., & Lewis, F. M. (1997). Recruitment and retention of families in clinical trials with longitudinal designs. Western Journal of Nursing Research, 19, 314–333. Page, R. M., Cole, G. E., & Timmreck, T. C. (1995). Basic epidemiological methods and biostatistics. Boston: Jones and Bartlett. Sidani, S., & Stevens, B. Alternative therapies and placebos: Conceptual clarification and methodologic implications. Canadian Journal of Nursing Research, 31, 73–86.

199

Studies Cited in Chapter 8 Beckie, T. M., Beckstead, J. W., & Webb, M. S. (2001). Modeling women’s quality of life after cardiac events. Western Journal of Nursing Research, 23, 179–194. Brook, A.D., Sherman, G., Malen, J., & Kollef, M. H. (2000). Early versus late tracheostomy in patients who require prolonged mechanical ventilation. American Journal of Critical Care, 9, 352–359. Bull, M. J., Hansen, H. E., & Gross, C. R. (2000). A professional–patient partnership model of discharge planning with elders hospitalized with heart failure. Applied Nursing Research, 13, 19–28. Faulkner, M. S., Hathaway, D. K., Milstead, E. J., & Burghen, G. A. (2001). Heart rate variability in adolescents and adults with type I diabetes. Nursing Research, 50, 95–104. Gamel, C., Grypdonck, M., Hengeveld, M., & Davis, B. (2001). A method to develop a nursing intervention: The contribution of qualitative studies to the process. Journal of Advanced Nursing, 33, 806–819. Garcia de Lucio, L., Gracia Lopez, F., Marin Lopez, M., Mas Hesse, M., & Camana Vaz, M. (2000). Training programme in techniques of self-control and communication skills to improve nurses’ relationships with relatives of seriously ill patients. Journal of Advanced Nursing, 32, 425–431. Greenfield, T. K., Midanik, L. T., and Rogers, J. D. (2000). A 10-year national trend study of alcohol consumption, 1984–1995. American Journal of Public Health, 90, 47–52. Harrison, L. L., Williams, A. K., Berbaum, M. L., Stem, J. T., & Leeper, J. (2000). Physiologic and behavioral effects of gentle human touch on preterm infants. Research in Nursing & Health, 23, 435–446. Heitkemper, M., Jarrett, M., Taylor, P., Walker, E., Landenburger, K., & Bond, E.F. (2001). Effect of sexual and physical abuse on symptom experiences in women with irritable bowel syndrome. Nursing Research, 50, 15–23. Hill, A. S., Kurkowski, T. B., & Garcia, J. (2000). Oral support measures used in feeding the preterm infant. Nursing Research, 49, 2–10. Horsburgh, M. E., Beansland, H., Locking-Cusolito, H., Howe, A., & Watson, D. (2000). Personality traits and self-care in adults awaiting renal transplant. Western Journal of Nursing Research, 22, 407–437. Johnson, J. L., Budz, B., Mackay, M., & Miller, C. (1999). Evaluation of a nurse-delivered smoking

200

PART 3 Designs for Nursing Research

cessation intervention for hospitalized patients with cardiac disease. Heart & Lung, 28, 55–64. Keane, A., Jepson, C., Pickett, M., Robinson, L., & McCorkle, R. (1996). Demographic characteristics, fire experiences, and distress of residential fire survivors. Issues in Mental Health Nursing, 17, 487–501. Landolt, M. A., Marti, D., Widner, J., & Meuli, M. (2002). Does cartoon movie distraction decrease burned children’s pain behavior. Journal of Burn Care & Rehabilitation, 23, 61–65. McCormick, R. D., Buchman, T. L., & Maki, D. G. (2000). Double-blind, randomized trial of scheduled use of a novel barrier cream and an oil-containing lotion for protecting the hands of health care workers. American Journal of Infection Control, 28, 302–310. McFarlane, J., Soeken, K., & Wiist, W. (2000). An evaluation of interventions to decrease intimate partner violence to pregnant women. Public Health Nursing, 17, 443–451. Metzger, B. L., Jarosz, P. A., & Noureddine, S. (2000). The effect of high-fat diet and exercise on the expression of genetic obesity. Western Journal of Nursing Research, 22, 736–748. Milne, J. (2000). The impact of information on health behaviors of older adults with urinary incontinence. Clinical Nursing Research, 9, 161–176. Mindell, J. A., & Jacobson, B. J. (2000). Sleep disturbances in pregnancy. Journal of Obstetric, Gynecologic, and Neonatal Nursing, 29, 590–597. Morin, K. H., Brogan, S., & Flavin, S. K. (2002). Attitudes and perceptions of body image in postpartum African American women. MCN: The Journal of Maternal Child/Nursing, 27, 20–25. Nahm, R., & Poston, I. (2000). Measurement of the effects of an integrated, point-of-care computer system on quality of nursing documentation and patient satisfaction. Computers in Nursing, 18, 220–229. Nantais-Smith, L. M., Covington, C., Nordstrom-Klee, B., Grubbs, C. J., Eto, I., Lawson, D., Pieper, B., &

Northouse, L. (2001). Differences in plasma and nipple aspirate carotenoid by lactation status. Nursing Research, 50, 172–177. Parent, N., & Fortin, F. (2000). A randomized, controlled trial of vicarious experience through peer support for male first-time cardiac surgery patients. Heart & Lung, 29, 389–400. Sandgren, A. K., McCaul, K. D., King, B., O’Donnell, S., & Foreman, G. (2000). Telephone therapy for patients with breast cancer. Oncology Nursing Forum, 27, 683–688. Schultz, A. A., Ashby-Hughes, B., Taylor, R., Gillis, D., & Wilkins, M. (2000). Effects of pectin on diarrhea in critically ill tube-fed patients receiving antibiotics. American Journal of Critical Care, 9, 403–411. Song, R., Daly, B. J., Rudy, E. B., Douglas, S., & Dyer, M. A. (1997). Nurses’ job satisfaction, absenteeism, and turnover after implementing a special care unit practice model. Research in Nursing & Health, 20, 443–452. Swanson, K. M. (1999). Effects of caring, measurement, and time on miscarriage impact and women’s wellbeing. Nursing Research, 48, 288–298. Whittington, K., Patrick, M., & Roberts, J. L. (2000). A national study of pressure ulcer prevalence and incidence in acute care hospitals. Journal of Wound Ostomy and Continence Nursing, 27, 209–215. Wilson, M. E., White, M. A., Cobb, B., Curry, R., Greene, D., & Popovich, D. (2000). Family dynamics, parental-fetal attachment, and infant temperament. Journal of Advanced Nursing, 31, 204–210. Winkelman, C., (2000). Effect of backrest position on intracranial and cerebral perfusion pressures in traumatically brain-injured adults. American Journal of Critical Care, 9, 373–382. Zust, B. L. (2000). Effect of cognitive therapy on depression in rural, battered women. Archives of Psychiatric Nursing, 14, 51–63.

9

Enhancing Rigor in Quantitative Research

T

his chapter describes methods that can be used to strengthen a wide array of quantitative research designs, including ways to enhance rigor through control over extraneous variables. There are two basic types of extraneous variables: those that are intrinsic to subjects and those that are external, stemming from the research situation. We begin by discussing methods to control situational factors. CONTROLLING THE R E S E A R C H S I T U AT I O N In quantitative studies, researchers often take steps to minimize situational contaminants to make the conditions under which data are collected as similar as possible for all subjects. The control that researchers impose by attempting to maintain constancy of conditions probably represents one of the earliest forms of scientific control. The environment has been found to exert a powerful influence on people’s emotions and behavior, and so, in designing quantitative studies, researchers need to pay attention to environmental context. Control over the environment is most easily achieved in laboratory experiments in which subjects are brought into environments arranged by the experimenter. Researchers have less control over the environment in studies that occur in natural settings. This does not mean that researchers should

forego efforts to make environments similar. For example, in conducting a nonexperimental study in which data are gathered through an interview, researchers ideally should conduct all interviews in basically the same kind of environment. That is, it is not considered a good practice to interview some respondents in their own homes, some in their places of work, and some in the researcher’s office. In each of these settings, participants assume different roles (e.g., wife, husband, parent; employee; client), and responses to questions may be influenced to some degree by those roles. In real-life settings, even when subjects are randomly assigned to groups, differentiation between groups may be difficult to control. As an example, suppose we were planning to teach nursing students a unit on dyspnea, and we have used a lecture-type approach in the past. If we were interested in trying a computerized autotutorial approach to cover the same material and wanted to evaluate its effectiveness before adopting it for all students, we might randomly assign students to one of the two methods. But now, suppose students in the two groups talk to one another about their experiences. Some lecture-group students might go through parts of the computer program. Some students in the autotutorial group might sit in on some lectures. In short, field experiments are often subject to the problem of contamination of treatments. In the same study, it would also be

202

PART 3 Designs for Nursing Research

difficult to control other variables, such as the place where learning occurs for the individualized group. Another external factor to consider is time. Depending on the study topic, the dependent variable may be influenced by the time of day or time of year in which the data are collected. In such cases, it would be desirable to strive for constancy of times across subjects. If an investigator were studying fatigue or perceptions of physical wellbeing, it would matter a great deal whether the data were gathered in the morning, afternoon, or evening, or in the summer as opposed to the winter. Although time constancy is not always critical, it should be considered in designing the study because it is often relatively easy to control. TIP: If constancy of research conditions cannot be achieved, you should consider controlling external factors by another method. For example, if you suspect that time of day may influence measurement of the dependent variable but cannot collect all the data at the same time of day, perhaps you could assign subjects randomly to morning versus afternoon sessions. Another issue concerns constancy of communications to subjects. In most studies, researchers inform participants about the study purpose, the use that will be made of the data, under whose auspices the study is being conducted, and so forth. This information should be prepared ahead of time, and the same message should be delivered to all subjects. In general, there should be as little ad-libbing as possible in a quantitative study. In studies involving an intervention, care should be taken to adhere to intervention protocols. For example, in experiments to test the effectiveness of a new drug to cure a medical problem, researchers would have to ensure that subjects in the experimental group received the same chemical substance and the same dosage, that the substance was administered in the same way, and so forth. When treatments are “fuzzier” than in the case of administering a drug (as is the case for most nursing interventions), researchers should spell out in detail the exact behaviors required of those responsible for administering the treatment.

TIP: Achieving constancy of conditions is not always easy, especially in clinical studies, but various steps can be taken. For example, in addition to having standard protocols, it is important to thoroughly train the people who will be collecting the data and, in the case of an experiment or quasi-experiment, the personnel responsible for implementing the intervention. The extent to which the protocols are followed should be monitored. In nonexperimental research, researchers do not manipulate the independent variable, so there is no means of ensuring constancy of conditions. Let us take as an example a correlational study that explores whether there is a relationship between people’s knowledge of nutrition and their eating habits. Suppose we find no relationship between nutritional knowledge and eating patterns. That is, the results indicate that people who are well informed about nutrition are just as likely as uninformed people to maintain inadequate diets. In this case, however, we had no control over the source of a person’s nutritional knowledge (the independent variable). This knowledge was measured after the fact, and the conditions under which the information was obtained cannot be assumed to be constant or even similar. We might conclude from the study that it is not important to teach nutrition to people because knowledge has no impact on their eating behavior. It may be, however, that different methods of providing nutritional information vary in their ability to motivate people to alter their eating habits. Thus, the ability to control or manipulate the independent variable may be critical in understanding relationships between variables, or the absence of relationships. Example of controlling external factors: Wipke-Tevis, Stotts, Williams, Froelicher, and Hunt (2001) conducted a quasi-experimental study in which great care was taken to ensure constancy of conditions. The study purpose was to compare tissue oxygenation in four body positions among people with venous ulcers. As an example of how the researchers controlled environmental factors, all measurements were made in the early morning; subjects had been instructed to fast so

CHAPTER 9 Enhancing Rigor in Quantitative Research

that a fasting blood sample could be drawn; subjects were then provided the same breakfast. After breakfast all subjects rested in bed supine for 30 minutes before testing began. CONTROLLING INTRINSIC SUBJECT FA C T O R S Participant characteristics almost always need to be controlled for quantitative findings to be interpretable. This section describes six ways of controlling extraneous subject characteristics. Randomization We have already discussed the most effective method of controlling individual extraneous variables— randomization. The primary function of randomization is to secure comparable groups, that is, to equalize groups with respect to extraneous variables. A distinct advantage of random assignment, compared with other control methods, is that randomization controls all possible sources of extraneous variation, without any conscious decision on the researcher’s part about which variables need to be controlled. Suppose we were assessing the effect of a physical training program on cardiovascular functioning among nursing home residents. Characteristics such as age, gender, smoking history, diet, and length of stay in the nursing home could all affect a patient’s cardiovascular system, independently of the special program. The effects of these other variables are extraneous to the research problem and should be controlled to understand the intervention’s effectiveness. Through randomization, we could expect that an experimental group (receiving the training program) and control group (not receiving the program) would be comparable in terms of these as well as any other factors that influence cardiovascular functioning. Example of randomization: Kelleher (2002) studied whether the timing of removing indwelling urinary catheters after surgery affected outcomes. Half of the 160 subjects

203

were randomly assigned to have their catheters removed at midnight and the other half had them removed at 6:00 AM, the time traditionally used in England, where the study was conducted. Patients in the midnight removal group passed a greater volume of urine with both their first and second voids, which permitted earlier discharge from the hospital. Repeated Measures Randomization in the context of a crossover design is an especially powerful method of ensuring equivalence between groups being compared. However, such a design is not appropriate for all studies because of the problem of carry-over effects. When subjects are exposed to two different conditions, they may be influenced in the second condition by their experience in the first. In our example of the physical training program, a crossover design is unsuitable because the “no-program-followed-by-program” condition would not be the same as the “programfollowed-by-no-program” condition: Subjects who were in the program in the first condition might decide to exercise more during the time they are not in the program, for example. Because treatments are not applied simultaneously in repeated measures designs, the order of the treatment may be important in affecting subjects’ performance. The best approach is to use randomized ordering. When there are only two conditions in a repeated measures design, the researcher simply designates that half the subjects, at random, will receive treatment A first and that the other half will receive treatment B first. When there are three or more conditions to which each subject will be exposed, the procedure of counterbalancing can be used to rule out ordering effects. For example, if there were three conditions (A, B, C), subjects would be randomly assigned to six different orderings in a counterbalanced scheme: A, B, C B, C, A C, A, B

A, C, B B, A, C C, B, A

Note that, in addition to their great potential for control over extraneous subject traits, crossover

204

PART 3 Designs for Nursing Research

designs offer another advantage: fewer subjects are needed. Fifty subjects exposed to two treatments in random order yield 100 pieces of data (50 ! 2); 50 subjects randomly assigned to two different treatment groups yield only 50 pieces of data (25 ! 2). In sum, a crossover design can be powerful and efficient for eliminating extraneous variables. When carry-over effects from one condition to another are anticipated, however, as might be the case in many nursing interventions, the researcher will need to seek other designs. Example of counterbalancing: Stevens, Johnston, Franck, Petryshen, Jack, and Foster (1999) exposed 122 low-birth-weight infants to four different pain-relieving interventions during a heel lance procedure. The interventions included (a) prone positioning; (b) receipt of a pacifier with sterile water; (c) receipt of a pacifier with sucrose; and (d) the standard treatment. The order in which the infants received the four treatments was completely randomized. Homogeneity When randomization and repeated measures are not feasible, alternative methods of controlling extraneous characteristics should be used. One such method is to use only subjects who are homogeneous with respect to confounding variables. The extraneous variables, in this case, are not allowed to vary. In our example of the physical training program, suppose our subjects were in two different nursing homes; those in one nursing home will receive the physical training program and those in the other nursing home will not receive it. If gender were considered to be an important confounding variable (and if the two nursing homes had different proportions of men and women), we could control gender by using only men (or only women) as subjects. Similarly, if we were concerned about the confounding effects of subjects’ age on cardiovascular functioning, participation could be limited to those within a specified age range. Using a homogeneous sample is easy and offers considerable control. The limitation of this approach lies in the fact that research findings can be

generalized only to the type of subjects who participated in the study. If the physical training program were found to have beneficial effects on the cardiovascular status of a sample of men 65 to 75 years of age, its usefulness for improving the cardiovascular status of women in their 80s would be strictly a matter of conjecture. Indeed, one noteworthy criticism of this approach is that researchers sometimes exclude subjects who are extremely ill or incapacitated, which means that the findings cannot be generalized to the very people who perhaps are most in need of scientific breakthroughs. Example of control through homogeneity: Zauszniewski and Chung (2001) studied the effects of depressive symptoms and learned resourcefulness on the health practices of diabetic patients. They restricted their sample to adult women with type 2 diabetes who had no other medical disorder and no history of a mental disorder. Blocking A fourth approach to controlling extraneous variables is to include them in the research design as independent variables. To pursue our example of the physical training program, if gender were thought to be a confounding variable, we could build it into the study in a randomized block design. In such a design, elderly men and women would be randomly assigned separately to the treatment group or control group. This approach has the advantage of enhancing the likelihood of detecting differences between our experimental and control groups because we can eliminate the effect of the blocking variable (gender) on the dependent variable. In addition, if the blocking variable is of interest substantively, this approach gives researchers the opportunity to study differences in groups created by the blocking variable (e.g., men versus women). The design can be extended to include more than one blocking variable, as shown in Figure 9-1. In this design, subjects’ age has been included to control for this second extraneous variable. Once again, we would randomly assign subjects from each block to either the experimental or control

CHAPTER 9 Enhancing Rigor in Quantitative Research

205

Program status Gender Male

Age group

Physical training program (Experimental)

No physical training program (Control)

66–70

71–75

76–80 Female 66–70

71–75

76–80

FIGURE 9.1

Schematic diagram of a 2 ! 2 ! 3 randomized block design.

conditions. In other words, half the men 66 to 70 years of age would randomly be assigned to the program, as would half the men 71 to 75 years of age, and so forth. Although in theory the number of blocks that could be added is unlimited, practical concerns usually dictate a relatively small number of blocks (and, hence, a small number of extraneous variables that can be controlled). Strictly speaking, a blocking design is appropriate only in experiments, but in reality it is used commonly in quasi-experimental and correlational studies as well. If we were studying the effects of a physical training program on cardiovascular functioning after the fact (i.e., subjects self-selected themselves into one of the two groups, and we had no control over who was included in each group), we could set up the analysis in such a way that differential program effects for men and women would be analyzed. The design structure would look similar to the randomized block design, but the conclusions that could be drawn would be

different than if we had been able to randomly assign subjects to groups. Example of control through blocking: Jones, Jaceldo, Lee, Zhang, and Meleis (2001) studied role integration and perceived health in Asian-American women who were caregivers of aging parents. Relationships among the research variables were examined separately for Chinese and Filipino women. Matching A fifth method of dealing with extraneous variables is matching. Matching (also known as pair matching) involves using knowledge of subject characteristics to form comparison groups. If matching were to be used in our physical training program example, and age and gender were the extraneous variables, we would need to match each subject in the physical training group with one in the comparison group with respect to age and gender.

206

PART 3 Designs for Nursing Research

Despite the intuitive appeal of such a design, there are reasons why matching is problematic. First, to match effectively, researchers must know in advance what the relevant extraneous variables are. Second, after two or three variables, it often becomes impossible to pair match adequately. Suppose we were interested in controlling for subjects’ age, gender, race, and length of nursing home stay. Thus, if subject 1 in the physical training program were an 80-year-old African-American woman whose length of stay was 5 years, we would need to seek another woman with these same or similar characteristics as her comparison group counterpart. With more than three variables, matching becomes cumbersome, if not impossible. Yet there are usually far more than three extraneous variables that could affect researchers’ dependent variables. For these reasons, matching as a technique for controlling extraneous variables should, in general, be used only when other, more powerful procedures are not feasible, as might be the case for some correlational studies (e.g., case—control designs). Sometimes, as an alternative to pair matching, in which subjects are matched on a one-to-one basis for each matching variable, researchers use a balanced design with regard to key extraneous variables. In such situations, researchers attempt only to ensure that the composition of the groups being compared have proportional representation with regard to extraneous variables. For example, if gender and race were the two extraneous variables of concern in our example of the physical training program, in adopting a balanced design we would strive to ensure that the same percentage of men and women (and the same percentage of white and African-American subjects) were in the physical training and comparison groups. Such an approach is much less cumbersome than pair matching, but it has similar limitations. Nevertheless, both pair matching and balancing are preferable to failing to control subject characteristics at all. Example of control through matching: Bliss, McLaughlin, Jung, Lowry, Savik, and Jensen (2000) compared the dietary intake of 39 people with fecal incontinence to that of 39 com-

parison group members who had normal bowel function. The groups were matched in terms of age and gender. Statistical Control A sixth method of controlling extraneous variables is through statistical analysis. Some of you likely are unfamiliar with basic statistical procedures, let alone the sophisticated techniques referred to here. Therefore, a detailed description of powerful statistical control mechanisms will not be attempted. If you have a background in statistics, you should consult Chapter 21 or a textbook on advanced statistics for fuller coverage of this topic. Because the notion of statistical control may mystify readers, however, we explain underlying principles with a simple illustration of a procedure called analysis of covariance. Returning to the physical training program example, suppose we had one group participating in the program and a comparison group not participating (e.g., residents of two different nursing homes, only one of which is offering the program). Suppose we used resting heart rate as one of our measures of cardiovascular functioning in this quasi-experimental study. There undoubtedly will be individual differences in heart rate—that is, it will vary from one person to the next. The research question is, Can some of the individual differences be attributed to a person’s participation in the physical training program? We know that differences in cardiovascular functioning are also related to other, extraneous characteristics, such as subjects’ age. In Figure 9-2, the large circles may be taken to represent total variability (extent of individual differences) for subjects’ resting heart rate. A certain amount of variability can be explained by virtue of the subjects’ age, which is shown in the figure as the small circle on the left in Figure 9-2A. Another part of the variability can be explained by subjects’ participation or nonparticipation in the training program, represented as the small circle on the right in A. The fact that the two small circles (age and program participation) overlap indicates that there is a relationship between those two variables. In other words,

CHAPTER 9 Enhancing Rigor in Quantitative Research

Participation in program

Age

207

Participation in program

Resting heart rate (dependent variable)

Resting heart rate (dependent variable)

A

B

FIGURE 9.2

Schematic diagram illustrating the principle of analysis of covariance.

subjects in the group receiving the physical training program are, on average, either older or younger than members of the comparison group. Therefore, age should be controlled. Otherwise, it will be impossible to determine whether differences in resting heart rate should be attributed to differences in age or program participation. Analysis of covariance accomplishes this by statistically removing the effect of extraneous variables on the dependent variable. In the illustration, the portion of heart rate variability attributable to age (the hatched area of the large circle in A) is removed through analysis of covariance. Figure 9-2B illustrates that the final analysis examines the effect of program participation on heart rate after removing the effect of age. By controlling heart rate variability resulting from age, we get a more accurate estimate of the effect of the training program on heart rate. Note that even after removing variability due to age, there is still individual variation not associated with the program treatment—the bottom half of the large circle in B. This means that the study can probably be further enhanced by controlling additional extraneous variables, such as gender, smoking history, and so forth. Analysis of covariance

and other sophisticated procedures can control multiple extraneous variables. Pretest measures of the dependent variables, when available, are an excellent choice for control variables; controlling them statistically greatly improves estimates of the effect of the intervention or independent variable. In our example, controlling preprogram measures of cardiovascular functioning through analysis of covariance would be especially powerful because it would remove the effect of individual variation stemming from many other extraneous factors. Example of statistical control: Moore and Dolansky (2001) used a pretest—posttest experimental design to test the effectiveness of an audiotaped intervention for patients who had coronary artery bypass graft (CABG) surgery. A taped message describing the typical recovery experiences of CABG patients was given to treatment group patients, both in the hospital and to take home. Control group members were given usual discharge instructions. In the analysis of treatment effectiveness on physical functioning and psychological distress, baseline

208

PART 3 Designs for Nursing Research

TABLE 9.1 Methods of Research Control METHOD

BENEFITS

LIMITATIONS

Randomization

• Controls all extraneous variables • Does not require advance

• Ethical and practical constraints on

Repeated measures

• If done with randomization,

• Cannot be used if there are

Homogeneity

• Easy to achieve in all types

• Limits the generalizability

knowledge of which variables to control

Blocking

Matching

strongest possible approach • Reduces sample size requirements of research • Enhances interpretability of relationships

• Enhances the ability to detect

and interpret relationships • Offers opportunity to examine blocking variable as an independent variable

• Enhances ability to detect

and interpret relationships • May be easy if there is a large “pool” of potential available controls

variables that can be manipulated.

• Possible artificiality of conditions

possible carry-over effects from one condition to the next

of the results

• Requires knowledge of which variables to control

• Usually restricted to a few blocking variables

• Requires knowledge of which variables to control

• Usually restricted to a few matching variables

• Requires knowledge of which variables to match

• May be difficult to find

comparison group matches, especially if there are more than two matching variables

Statistical control

• Enhances ability to detect

and interpret relationships • Relatively economical means of controlling numerous extraneous variables

• Requires knowledge of

which variables to control, and requires measurement of those variables • Requires statistical sophistication

CHAPTER 9 Enhancing Rigor in Quantitative Research

measures of the outcomes were statistically controlled, as were age, comorbidity, and presurgical cardiac functional status. Moreover, the researchers did separate analyses for men and women. Thus, they used randomization, statistical control, and blocking to control extraneous variables. TIP: The extraneous variables that need to be controlled vary from one study to another, but we can nevertheless offer some guidance. The best variable is the dependent variable itself, measured before the introduction of the independent variable. Major demographic variables—age, race/ethnicity, gender, education, income, marital status—are good candidates to measure and control because they correlate with many nursing outcomes (as well as with willingness to participate and remain in a study). When the dependent variable is biophysiologic, measures of health status, medication, hospitalization history, and so on are likely to be important. Extraneous variables that are particular to the research problem should be identified through a literature review.

Evaluation of Control Methods Table 9-1 summarizes the benefits and drawbacks of the six control mechanisms. Overall, random assignment of subjects to groups is the most effective approach to managing extraneous variables because randomization tends to cancel out individual variation on all possible extraneous variables. Crossover designs, although an extremely useful supplement to randomization, cannot be applied to most nursing research problems. The remaining alternatives— homogeneity, blocking, matching, and analysis of covariance—have one disadvantage in common: researchers must know or predict in advance the relevant extraneous variables. To select homogeneous samples, develop a blocking design, match, or perform analysis of covariance, researchers must know which variables need to be measured and controlled. This constraint may limit the degree of control possible, particularly because researchers can seldom deal explicitly with more than two or three

209

extraneous variables at one time (except in the case of statistical control). Although we have repeatedly hailed randomization as the ideal mechanism for controlling extraneous subject characteristics, it is clear that randomization is not always possible. For example, if the independent variable cannot be manipulated, then other techniques must be used. In correlational and quasi-experimental studies, the control options available to researchers include homogeneity, blocking, matching, and analysis of covariance. In quantitative research, the use of any of the control procedures discussed here is preferable to the absence of any attempt to control intrinsic extraneous variables. TIP: This section offered various strategies for controlling a key source of extraneous variation: study participants themselves. These alternative strategies are not mutually exclusive; whenever possible, multiple methods should be used (as was the case in Moore and Dolansky’s study). For example, statistical methods of control can be used in conjunction with blocking or matching. Even when randomization has been used, analysis of covariance increases the precision of the design. CHARACTERISTICS OF GOOD DESIGN In selecting a research design, researchers should be guided by one overarching consideration: whether the design does the best possible job of providing trustworthy answers to the research questions. Usually, a given research question can be addressed with a number of different designs, and researchers have flexibility in selecting one. Yet many designs are completely unsuitable for dealing with certain research problems. For example, a loosely structured design, such as those used in qualitative studies, would be inappropriate to address the question of whether non-nutritive sucking opportunities among premature infants facilitate early oral feedings. On the other hand, a tightly controlled study may unnecessarily restrict researchers interested in

210

PART 3 Designs for Nursing Research

understanding the processes by which nurses make diagnoses. There are many research questions of interest to nurses for which highly structured designs are unsuitable. TIP: Although techniques of research control are mechanisms for controlling bias, there are situations in which too much control can introduce bias. For example, if researchers tightly control the ways in which key study variables can manifest themselves, it is possible that the true nature of those variables will be obscured. When the key concepts are phenomena that are poorly understood or dimensions of which have not been clarified, then a design that allows some flexibility is better suited to the study aims. Cook and Campbell (1979), in their classic book on research design, describe four considerations that are important in evaluating research design for studies that focus on relationships among variables. The questions that must be addressed by researchers (and evaluated by research consumers) regarding research design are as follows: 1. What is the strength of the evidence that a relationship exists between two variables? 2. If a relationship exists, what is the strength of the evidence that the independent variable of interest (e.g., an intervention), rather than extraneous factors, caused the outcome? 3. If the relationship is plausibly causal, what are the theoretical constructs underlying the related variables? 4. If the relationship is plausibly causal, what is the strength of evidence that the relationship is generalizable across people, settings, and time? These questions, respectively, correspond to four aspects of a study’s validity: (1) statistical conclusion validity, (2) internal validity, (3) construct validity, and (4) external validity. In this section we discuss certain aspects of statistical conclusion validity, internal validity, and external validity, and factors that can undermine validity. Construct validity, which concerns the measurement of variables, is discussed in Chapter 18.

Statistical Conclusion Validity As noted in Chapter 8, the first criterion for establishing causality is demonstrating that there is, in fact, an empirical relationship between the independent and dependent variable. Statistical methods are used to determine if such a relationship exists. Design decisions can influence whether statistical tests will actually detect true relationships, and so researchers need to make decisions that protect against reaching false statistical conclusions. Although we cannot at this point in the text discuss all aspects of statistical conclusion validity, we can describe a few design issues that can be threats to making valid statistical inferences. Low Statistical Power Statistical power refers to the ability of the design to detect true relationships among variables. Adequate statistical power can be achieved in various ways, the most straightforward of which is to use a sufficiently large sample. When small samples are used, statistical power tends to be low, and the analyses may fail to show that the independent and dependent variables are related, even when they are. Power and sample size are discussed in Chapters 13 and 20. Another aspect of a powerful design concerns the construction or definition of the independent variable, and the counterfactual. Both statistically and substantively, results are clearer when differences between groups and treatments being compared are large. Researchers should usually aim to maximize group differences on the dependent variables by maximizing differences on the independent variable. In other words, the results are likely to be more clearcut if the groups are as different as possible. Conn, Rantz, Wipke-Tevis, and Maas (2001) offer excellent suggestions for strengthening the power and effectiveness of nursing intervention. Advice about strengthening group differences is more easily followed in experimental than in nonexperimental research. In experiments, investigators can devise treatment conditions that are distinct and

CHAPTER 9 Enhancing Rigor in Quantitative Research

as strong as time, money, ethics, and practicality permit. Even in nonexperimental research, however, there are frequently opportunities to operationalize independent variables in such a way that power to detect differences is enhanced. Inadequate Precision Quantitative researchers usually try to design a study to achieve the highest possible precision, which is achieved through accurate measuring tools, controls over extraneous variables, and powerful statistical methods. Precision can best be understood through a specific example. Suppose we were studying the effect of admission into a nursing home on depression by comparing elders who were or were not admitted. Depression varies from one elderly person to another for a various reasons. In the present study, we are interested in isolating—as precisely as possible—the portion of variation in depression attributable to nursing home admission. Mechanisms of research control that reduce variability attributable to extraneous factors can be built into the research design, thereby enhancing precision. In a quantitative study, the following ratio expresses what researchers wish to assess: Variation in depression due to nursing home admission Variation in depression due to other factors (e.g., age, pain, medical prognosis, social support) This ratio, although greatly simplified here, captures the essence of many statistical tests. We want to make variability in the numerator (the upper half) as large as possible relative to variability in the denominator (the lower half), to evaluate clearly the relationship between nursing home admission and levels of depression. The smaller the variability in depression due to extraneous variables (e.g., age, prognosis), the easier it will be to detect differences in depression between elders who were or were not admitted to a nursing home. Designs that enable researchers to reduce variability caused by extraneous variables increase the precision of the research. As a purely hypothetical

211

illustration of why this is so, we will attach some numeric values* to the ratio as follows: Variability due to nursing home admission 10 " Variability due to extraneous variables 4 If we can make the bottom number smaller, say by changing it from 4 to 2, then we will have a purer and more precise estimate of the effect of nursing home admission on depression, relative to other influences. All of the control mechanisms described in the previous section help to reduce variability caused by extraneous variables, and so should be considered in designing studies. We illustrate this by continuing our example. The total variability in levels of depression can be conceptualized as having three components: Total variability in depression " Variability due to nursing home admission #Variability due to age #Variability due to other extraneous variables. This equation can be taken to mean that part of the reason why some elderly individuals are depressed and others are not is that some were admitted to a nursing home and others were not; some were older and some were younger; and other factors, such as level of pain, medical prognosis, availability of social supports, also had an effect on depression. One way to increase the precision in this study would be to control age, thereby removing the variability in depression that results from age differences. We could do this, for example, through homogeneity (i.e., by including in our sample only elderly people within a fairly narrow age range), by using age as a blocking variable, or by statistically controlling age. With any of these methods, the variability in depression due to age would be reduced or eliminated. As a result, the effect of nursing home admission on depression becomes greater, relative to the remaining extraneous variability. Thus, we can say that these designs enabled

*At this point, you should not be concerned with how these numbers can be obtained. The procedure is explained in Chapter 20.

212

PART 3 Designs for Nursing Research

us to get a more precise estimate of the effect of nursing home admission on level of depression. Research designs differ considerably in the sensitivity with which effects under study can be detected with statistical tools. Lipsey (1990) has prepared an excellent guide to assist researchers in enhancing the sensitivity of research designs. Unreliable Implementation of a Treatment The strength of an intervention (and hence statistical power) can be undermined if the intervention is not as powerful in reality as it is “on paper.” An intervention can be weakened by a number of factors, most of which can be influenced to some degree by the researcher. One issue concerns the extent to which the intervention is similar from one subject to the next. Lack of standardization (constancy of conditions) adds extraneous variation and can diminish the full force of the intervention. Using the notions just described, when standard protocols are not used or not followed, variability due to the intervention (i.e., in the numerator) can be suppressed and variability due to extraneous factors (i.e., in the denominator) can be inflated, possibly leading to the erroneous conclusion that the intervention was ineffective. This suggests the need for careful standardization, adequate training of personnel, and vigilant monitoring to ensure that the intervention is being implemented as planned. Training and monitoring are especially important when doubleblind procedures are not used, because the staff administering (or withholding) the treatment may inadvertently send out inappropriate signals or blur the treatment conditions. Of course, in clinical settings there may be pressures on researchers to administer the treatment to some controls, which can threaten the design. Example of problems with standardization: Winterburn and Fraser (2000) tested the effect of postnatal hospital stay on breastfeeding rates in a teaching hospital in northern England. Women in their third trimester were randomly assigned to a short postnatal stay (6 to 48 hours) or a longer stay (more than 48 hours). The study design

was compromised by the fact that some women in the long-stay group were reluctant to stay in the hospital, resulting in relatively small group differences in hospital stay. Subjects may be exposed to different conditions than were planned if, for example, those in the experimental group elect not to participate fully in the treatment (e.g., they stop going to treatment sessions), or if those in the control group seek and gain access to the treatment. Researchers should design the study to enhance the integrity of the treatment conditions, taking steps in particular to encourage participation among members of the experimental group. As an example, researchers using an experimental design can sometimes affect participation by the timing of random assignment. If subjects are randomly assigned before treatment conditions are explained to them, they may drop out immediately after learning what the conditions will entail; random assignment after the explanation may result in less subject loss among those randomly assigned. Nonparticipation in an intervention is rarely random, so researchers should document which subjects got what amount of treatment so that individual differences in “dose” can be taken into account in the analysis or interpretation of results. Example of monitoring participation: Cowan, Pike, and Budzynski (2001) conducted a rigorous experimental study of the effectiveness of a psychosocial nursing therapy in reducing mortality for patients who experienced sudden cardiac arrest. Subjects were randomly assigned to the intervention (11 individual therapeutic sessions) or to a control group. Those administering the intervention protocol were trained extensively. Subject adherence to the intervention was carefully monitored (e.g., therapists completed a checklist for each subject), and deemed to be excellent. When subjects withdraw from a study, both statistical conclusion validity and internal validity (discussed next) can be compromised. In analyzing data from such studies, researchers are in a dilemma about whom to “count” as being “in” a condition. A procedure that is sometimes used is an

CHAPTER 9 Enhancing Rigor in Quantitative Research

on-protocol analysis, which includes members in a treatment group only if they actually received the treatment. Such an analysis is problematic, however, because the sample is no longer representative of the entire group of interest, and self-selection into a nonintervention condition nullifies the comparability of groups. This type of analysis will almost always be biased toward finding positive treatment effects. A more conservative approach is to use a principle known as intention to treat, which involves an analysis that assumes that each person received the treatment to which he or she was assigned. This, however, may yield an underestimate of the effects of a treatment if many subjects did not actually get the assigned treatment. If data are analyzed both ways and the outcomes are the same, researchers can have more confidence in the results. Another alternative is to include measures of the “dose” of treatment received into the analysis (Sidani, 1998). Example of intention to treat: Dougherty and her colleagues (2002) used an experimental design to test an intervention to manage symptoms of urinary incontinence among older rural women. They used the intention-to-treat approach: women in the experimental group remained in that group even if they failed to comply with the study protocol; and control group members remained in the control group even if they learned a technique that was part of the intervention and practiced it on their own. Internal Validity Internal validity refers to the extent to which it is possible to make an inference that the independent variable is truly causing or influencing the dependent variable and that the relationship between the two is not the spurious effect of an extraneous variable. The control mechanisms reviewed earlier in this chapter are all strategies for improving internal validity. If researchers are not careful in managing extraneous variables and in other ways controlling the design of the study, there may be reason to challenge the conclusion that the subjects’ perfor-

213

mance on the dependent measure was caused by the independent variable. Threats to Internal Validity True experiments possess a high degree of internal validity because the use of manipulation, randomization, and a control group usually enables the researcher to rule out most alternative explanations for the results. Researchers who use quasi-experimental, preexperimental, or correlational designs must always contend with competing explanations for obtained results. A few of these competing explanations (threats to internal validity) are examined here. History. The threat of history refers to the occurrence of external events that take place concurrently with the independent variable that can affect the dependent variables. For example, suppose we were studying the effectiveness of a county-wide nurse outreach program to encourage pregnant women in rural areas to improve their health-related practices before delivery (e.g., better nutritional practices, cessation of smoking, earlier prenatal care). The program might be evaluated by comparing the average birth weight of infants born in the 12 months before the outreach program with the average birth weight of those born in the 12 months after the program was introduced, using a time series design. However, suppose that 1 month after the new program was launched, a highly publicized docudrama regarding the inadequacies of prenatal care for poor women was aired on national television. Infants’ birth weight might now be affected by both the intervention and the messages in the docudrama, and it becomes impossible to disentangle the two effects. In a true experiment, history usually is not a threat to a study’s internal validity because we can often assume that external events are as likely to affect the experimental as the control group. When this is the case, group differences on the dependent variables represent effects over and above those created by external factors. There are, however, exceptions. For example, when a crossover design is used, an external event may occur during the first half (or second half) of the experiment, and so

214

PART 3 Designs for Nursing Research

treatments would be contaminated by the effect of that event. That is, some people would receive treatment A with the event and others would receive treatment A without it, and the same would be true for treatment B. Selection. Selection encompasses biases resulting from preexisting differences between groups. When individuals are not assigned randomly to groups, there is always a possibility that the groups are nonequivalent. They may differ in ways that are subtle and difficult to detect. If the groups are nonequivalent, differences on outcomes may result from initial differences rather than from the effect of the independent variable. For example, if we found that women with a fertility problem were more likely to be depressed than women who were mothers, it would be impossible to conclude that the two groups differed in depression because of differences in reproductive status; women in the two groups might have been different in terms of psychological adjustment from the start. The problem of selection is reduced if researchers can collect data on subject characteristics before the occurrence of the independent variable. In our example of infertility, the best design would be to collect data on women’s depression before they attempted to become pregnant. Selection bias is one of the most problematic and frequently encountered threats to the internal validity of studies not using an experimental design. But selection can also enter into experimental designs if some subjects elect not to receive the treatment; these subjects would select themselves into the control condition. Selection biases also often interact with other biases to compound the threat to the internal validity. For example, if the comparison group is different from the treatment group or main group of interest, then the characteristics of the members of the comparison group could lead them to have different intervening experiences, thereby introducing both history and selection biases into the design. Maturation. In a research context, maturation refers to processes occurring within subjects during the course of the study as a result of the passage of time rather than as a result of a treatment or indepen-

dent variable. Examples of such processes include physical growth, emotional maturity, fatigue, and the like. For instance, if we wanted to evaluate the effects of a special sensorimotor development program for developmentally delayed children, we would have to consider that progress does occur in these children even without special assistance. A design such as a one-group pretest—posttest design (see Figure 8-7 in Chapter 8), for example, would be highly susceptible to this threat to internal validity. Maturation is a relevant consideration in many areas of nursing research. Remember that maturation here does not refer to aging or development exclusively but rather to any change that occurs as a function of time. Thus, wound healing, postoperative recovery, and many other bodily changes that can occur with little or no nursing or medical intervention must be considered as an explanation for outcomes that rivals an explanation based on the effects of the independent variable. Testing. Testing refers to the effects of taking a pretest on subjects’ performance on a posttest. It has been documented in several studies, particularly in those dealing with opinions and attitudes, that the mere act of collecting data from people changes them. Suppose we administered to a group of nursing students a questionnaire about their attitudes toward assisted suicide. We then acquaint them with various arguments that have been made for and against assisted suicide, outcomes of court cases, and the like. At the end of instruction, we give them the same attitude measure and observe whether their attitudes have changed. The problem is that the first administration of the questionnaire might sensitize students, resulting in attitude changes regardless of whether instruction follows. If a comparison group is not used in the study, it becomes impossible to segregate the effects of the instruction from the effects of taking the pretest. In true experiments, testing may not be a problem because its effects would be expected to be about equal in all groups, but the Solomon four-group design (discussed in Chapter 8) could be used if researchers wanted to isolate intervention effects from pretest effects.

CHAPTER 9 Enhancing Rigor in Quantitative Research

Sensitization, or testing, problems are more likely to occur when pretest data come from selfreports (e.g., through a questionnaire), especially if subjects are exposed to controversial or novel material in the pretest. For some nursing studies (e.g., those that involve biophysiologic data), testing effects are not a major concern. Instrumentation. Another threat related to measurements is the threat of instrumentation. This bias reflects changes in measuring instruments or methods of measurement between two points of data collection. For example, if we used one measure of stress at baseline and a revised measure at follow-up, any differences might reflect changes in the measuring tool rather than the effect of an independent variable. Instrumentation effects can occur even if the same measure is used. For example, if the measuring tool yields more accurate measures on the second administration (e.g., if the people collecting the data are more experienced) or less accurate measures the second time (e.g., if subjects become bored or fatigued), then these differences could bias the results. Mortality. Mortality is the threat that arises from differential attrition in groups being compared. The loss of subjects during the course of a study may differ from one group to another because of a priori differences in interest, motivation, health, and so on. For example, suppose we used a nonequivalent control group design to assess the morale of nurses from two different hospitals, one of which was initiating primary nursing. The dependent variable, nursing staff morale, is measured in both hospitals before and after the intervention. Comparison group members, who may have no particular commitment to the study, may decline to complete a posttest questionnaire because of lack of incentive. Those who do fill it out may be unrepresentative of the group as a whole—they may be those who are especially critical of their work environment, for example. It might thus appear that nurses’ morale in the comparison hospital had declined over time, but the decline might only be an artifact of biased attrition. The risk of attrition is especially great when the length of time between points of data collection

215

is long. A 12-month follow-up of subjects, for example, tends to produce higher rates of attrition than a 1-month follow-up. In clinical studies, the problem of attrition may be especially acute because of patient death or disability. If attrition is random (i.e., those dropping out of a study are similar to those remaining in it with respect to extraneous characteristics), then there would not be bias. However, attrition is rarely totally random. In general, the higher the rate of attrition, the greater the likelihood of bias. Although there is no absolute standard for acceptable attrition rates, biases are usually of concern if the rate exceeds 20%. TIP: In longitudinal studies, a good method to reduce attrition is to use procedures to help relocate subjects. Attrition often occurs because researchers cannot find participants, rather than because of their refusal to continue in the study. There are many sophisticated (and costly) methods of tracing subjects, but a simple and effective strategy is to obtain contact information from participants at each point of data collection. Contact information includes, at a minimum, the names, addresses, and telephone numbers of two or three people with whom the subject is close (e.g., parents, siblings, or good friends)—people who would be likely to know how to contact subjects if they moved. Internal Validity and Research Design Quasi-experimental, preexperimental, and correlational studies are especially susceptible to threats to internal validity. Table 9-2 lists the specific types of designs that are most vulnerable to the threats just described (although it should not be assumed that the threats do not emerge in designs not listed). These threats represent alternative explanations (rival hypotheses) that compete with the independent variable as a cause of the dependent variable. The aim of a strong research design is to rule out these competing explanations. A good experimental design normally rules out rival hypotheses, but even in true experiments researchers may need to attend to them. For example,

216

PART 3 Designs for Nursing Research

TABLE 9.2 Research Designs and Threats to Internal Validity THREAT

DESIGNS MOST LIKELY TO BE AFFECTED

History

One-group pretest–posttest Time series Prospective cohort Crossover/repeated measures

Selection

Nonequivalent control group (especially, posttest-only) Case–control “Natural’’ experiments

Maturation

One-group pretest–posttest

Testing

All pretest–posttest designs

Instrumentation

All pretest–posttest designs

Mortality

Prospective cohort Longitudinal experiments and quasi-experiments

if constancy of conditions is not maintained for experimental and control groups, then history might be a rival explanation for any group differences. Mortality might also be a salient threat in true experiments. Because the experimenter does things differently with the experimental and control groups, subjects in the groups may drop out of the study differentially. This is particularly apt to happen if the experimental treatment is painful, inconvenient, or time-consuming, or if the control condition is boring or bothersome. When this happens, subjects remaining in the study may differ from those who left in important ways, thereby nullifying the initial equivalence of the groups. In short, in designing a study, researchers should consider how best to guard against and detect all possible threats to internal validity, no matter what design is used. TIP: In designing a study, try to anticipate negative findings and consider whether design adjustments might affect the results. For example, suppose we hypothesized that environ-

mental factors such as light and noise affect the incidence of acute confusion among the hospitalized elderly. With a preliminary design in mind, try to imagine findings that fail to support the hypothesis. Then ask yourself what could be done to decrease the possibility of these negative results. Can power be increased by making differences in environmental conditions sharper? Can precision be increased by controlling additional extraneous variables? Can bias be eliminated by better training of research personnel? Internal Validity and Data Analysis The best strategy for enhancing internal validity is to use a strong research design that includes the use of control mechanisms discussed earlier in this chapter. Even when this is possible (and, certainly, when this is not possible), it is highly advisable to analyze the data to determine the nature and extent of biases that arose. When biases are detected, the information can be used to interpret substantive results; in some cases, biases can be statistically controlled.

CHAPTER 9 Enhancing Rigor in Quantitative Research

Researchers need to be self-critics. They need to consider fully and objectively the types of biases that could have arisen within their chosen design— and then systematically search for evidence of their existence (while hoping, of course, that no evidence can be found). A few examples should illustrate how to proceed. Selection biases are the most prevalent threat to internal validity and should be examined whenever possible. Typically, this involves comparing subjects on pretest measures, when pretest data have been collected. For example, if we were studying depression in women who delivered a baby by cesarean delivery versus those who delivered vaginally, an ideal way to evaluate selection bias would be to compare depression in these two groups during or before the pregnancy. If there are significant predelivery differences, then postdelivery differences would have to be interpreted with initial differences in mind (or with differences controlled). In posttest-only designs or in cross-sectional correlational studies in which there is no pretest measure of the dependent variable, researchers should nevertheless search for selection biases by comparing groups with respect to important background variables, such as age, gender, ethnicity, social class, health status, and so on. Selection biases should be analyzed even when random assignment has been used to form groups because there is no absolute guarantee that randomization will yield perfectly comparable groups. Whenever the research design involves multiple points of data collection, researchers should analyze attrition biases. This is typically achieved through a comparison of those who did and did not complete the study with regard to baseline measures of the dependent variable or other characteristics measured at the first point of data collection. Example of an examination of attrition and selection bias: Moser and Dracup (2000) studied the effect of two alternative cardiopulmonary resuscitation (CPR) training interventions against a control condition among 219 spouses of cardiac patients recovering from an acute cardiac event. At the 1-month

217

follow-up, 196 subjects provided outcome data, for a response rate of 89.5%. The researchers noted that there were no significant differences in background characteristics (e.g., race, education) between subjects completing or not completing the study, nor between subjects in the three treatment groups. In a repeated measures design, history is a potential threat both because an external event could differentially affect subjects in different treatment orderings and because the different orderings are in themselves a kind of differential history. The substantive analysis of the data involves comparing the dependent variable under treatment A versus treatment B. The analysis for evidence of bias, by contrast, involves a comparison of subjects in the different orderings (e.g., A then B versus B then A). If there are significant differences between the two orderings, then this is evidence of an ordering bias. In summary, efforts to enhance the internal validity of a study should not end once the design strategy has been put in place. Researchers should seek additional opportunities to understand (and possibly to correct) the various threats to internal validity that can arise. TIP: It is important to build in opportunities to analyze bias. This means giving careful consideration to variables that should be measured. For example, information on characteristics that help identify a selection problem should be collected. In a longitudinal study, variables that are likely to be related to attrition should be measured. External Validity The term external validity refers to the generalizability of the research findings to other settings or samples. Research is almost never conducted to discover relationships among variables for a specific group of people at one point in time. The aim of research typically is to reveal enduring relationships, the understanding of which can be used to improve human health and well-being. If a nursing intervention under investigation is found to be successful, others will want to adopt it. Therefore, an important question is whether the intervention will

218

PART 3 Designs for Nursing Research

work in another setting and with different patients. A study is externally valid to the extent that the sample is representative of the broader population, and the study setting and experimental arrangements are representative of other environments. External Validity and Sampling One aspect of a study’s external validity concerns the adequacy of the sampling design. If the research sample is representative of the population, then generalization is straightforward. Sampling designs are described in Chapter 13. Strictly speaking, study findings can be safely generalized only to the population from which a sample has been selected at random. If we were studying the effects of a newly developed therapeutic treatment for heroin addicts, we might begin with a population of addicts in a particular drug treatment center in Detroit. From this population, a random sample of drug users could be selected as subjects, who would then be randomly assigned to the treatment or control condition. If the results revealed that the treatment was effective in reducing recidivism in this sample of addicts, could it be concluded that all addicts in North America would benefit from the treatment? Unfortunately, no. The population of heroin addicts undergoing treatment in one particular facility may not be representative of all addicts. For example, drug users from certain ethnic, socioeconomic, or age groups might use the facility in question. Perhaps the new treatment is effective only with individuals from such groups. Of relevance here is Kempthorne’s (1961) distinction between accessible and target populations. The accessible population is the population available for a particular study. In our example, heroin addicts enrolled at that Detroit treatment center would be the accessible population. When random procedures have been used to select a sample from an accessible population, there is no difficulty generalizing the results to that group. The target population is the total group of subjects about whom a researcher is interested and to whom results could reasonably be generalized. This second type of generalization is more risky and cannot be done with as much confidence as

when generalizing to the accessible population. The appropriateness of such an inference hinges on the similarity of characteristics in the two populations. Thus, researchers must be aware of the characteristics of the accessible population and, in turn, define the target population to be like it. In the drug treatment example, the accessible population might predominantly comprise voluntarily admitted white men in their twenties living in Detroit. Although we might ideally like to generalize our results to all drug addicts, we would be on much safer ground if we defined our target population as young, urban, white men who present themselves for treatment. Threats to External Validity In addition to characteristics of the sample that limit the generalizability of research findings, there are various aspects of the research situation that affect the study’s representativeness and, hence, its external validity. These characteristics should be taken into consideration in designing a study and in interpreting results. Among the most noteworthy threats to the external validity of studies—particularly those involving an intervention—are the following five effects: 1. Expectancy effects. As discussed in Chapter 8, subjects may behave in a particular manner largely because they are aware of their participation in a study (i.e., the Hawthorne effect). If a certain type of behavior is elicited specifically because of the research context, then the results cannot be generalized to more natural settings. Similarly, a placebo effect occurs when subjects administered a pseudointervention show changes or improvements. That same placebo might not, however, have any benefits when not administered in the context of a study. (There are also examples of a so-called nocebo effect, which involves adverse side effects experienced by those getting the placebo.) Example of a study with possible Hawthorne effect: Hundley, Milne, Leighton-Beck, Graham, and Fitmaurice (2000) designed an intervention aimed

CHAPTER 9 Enhancing Rigor in Quantitative Research

at raising research awareness among midwives and nurses. The treatment group got an educational program, whereas the control group did not. Both groups showed a knowledge gain from pretest to posttest, which the researchers interpreted as a Hawthorne effect (it might also have been the results of a testing effect). 2. Novelty effects. When a treatment is new, subjects and research agents alike might alter their behavior in various ways. People may be either enthusiastic or skeptical about new methods of doing things. Results may reflect reactions to the novelty rather than to the intrinsic nature of an intervention; once the treatment is more familiar, results might be different. 3. Interaction of history and treatment effect. The results may reflect the impact of the treatment and some other events external to the study. When the treatment is implemented again in the absence of the other events, different results may be obtained. For example, if a dietary intervention for people with high cholesterol levels was being evaluated shortly after extensive media coverage of research demonstrating a link between oat bran consumption and reduced cholesterol levels, it would be difficult to know whether any observed effects would be found again if the intervention were implemented several months later with a new group of people. 4. Experimenter effects. Subjects’ behavior may be affected by characteristics of the researchers. The investigators often have an emotional or intellectual investment in demonstrating that their hypotheses are correct and may unconsciously communicate their expectations to subjects. If this is the case, the results in the original study might be difficult to replicate in a more neutral situation. 5. Measurement effects. Researchers collect a considerable amount of data in most studies, such as pretest information, background data, and so forth. The results may not apply to another group of people who are not also exposed to the same data collection (and attention-giving) procedures.

219

Issues in Achieving Study Validity Researcher strive to design studies that are strong with respect to all four types of study validity. In some instances, however, the requirements for ensuring one type of validity interfere with the possibility of achieving others. As one example, consider researchers who use homogeneity to enhance the internal validity of a study. By controlling extraneous variables through selection of a homogeneous sample, researchers strengthen internal validity but limit external validity (i.e., the ability to generalize the study results to an entire population of interest). As another example, if researchers exert a high degree of control over a study through constancy of conditions in an effort to maximize internal validity, the setting may become highly artificial and pose a threat to the generalizability of the findings to more naturalistic environments. Thus, it is often necessary to reach a compromise by introducing sufficient controls while maintaining some semblance of realism. When there is a conflict between internal and external validity, it is often preferable to opt for stronger internal validity. Indeed, it can be argued that if findings are not internally valid, they cannot possibly be externally valid. That is, it makes little sense to generalize findings if the findings are themselves ambiguous. Whenever a compromise is necessary, the concept of replication, or the repetition of a study in a new setting with new subjects, is critical. Much greater confidence can be placed in study findings if it can be demonstrated that the results can be replicated in other settings and with new subjects. RESEARCH EXAMPLE We conclude this chapter with an example of a study that was especially careful in exercising research control and establishing internal validity. Knebel, Bentz, and Barnes (2000) conducted a study to determine whether short-term oxygen administration might decrease dyspnea and improve exercise performance among middle-aged adults with a

220

PART 3 Designs for Nursing Research

deficiency of the circulating protein alpha1-antitrypsin, a condition that can lead to pulmonary disease. They used an experimental crossover design in which 33 subjects were exposed, in randomized sequence, to an experimental condition (6-minute walks with administration of oxygen) and a control condition (walks with compressed air). A table of random numbers was used to determine order of administration. Both gases were delivered by nasal cannula, and both subjects and research agents were blind to which gas was being used, ruling out any experimenter effect or a placebo effect. The research protocols were standardized to ensure constancy of conditions. For example, a standardized message of encouragement was given. Subjects were instructed not to talk and to cover as much distance on a 140-foot circular track as possible. All walks occurred at least 2 hours after meals or waking. To control for learning effects, subjects completed three practice walks (without any treatment) before the double-blind tests. The dependent variables in this study included walk distance (to measure activity tolerance), pulse oximetry saturation (to measure oxygen saturation), heart rate and breathing frequency (to measure response to exercise), and a selfassessment of dyspnea intensity. In addition to testing the research hypotheses, the researchers did some analyses to rule out possible threats to validity. For example, they examined the threat of history by testing for ordering effects (i.e., whether getting oxygen first resulted in different outcomes than getting oxygen last) on walk distance and dyspnea intensity. No ordering effects were found. They also did meticulous tests to rule out learning effects. In brief, they determined that most of the changes in the dependent variables over time occurred during the three practice walks and not during the experimental tests. There was no attrition in this study; both tests were done on the same day and all patients were hospitalized. With regard to the primary research questions, statistical tests revealed that subjects in the two conditions performed significantly differently with regard to oxygen saturation, but not with regard to other dependent variables. Oxygen saturation was significantly higher with oxygen compared with compressed air. In addition to using randomization and a crossover design to control for extraneous variables, the researchers also used blocking to increase the precision of their analyses and to test possible interaction effects. Specifically, data for men and women were

analyzed separately. The results indicated that for men, dyspnea was not different in the two conditions, but women had a significantly lower dyspnea score with oxygen than with compressed air.

S U M M A RY P O I N T S • Research control is used to remove the effect of situational factors that could affect the study outcomes (e.g., the environment) and intrinsic subject characteristics extraneous to the research question. • Quantitative researchers strive to achieve constancy of conditions under which a study is performed, as a means of controlling situational factors. • The ideal method of controlling intrinsic subject characteristics is random assignment of subjects to groups, which effectively controls for all possible extraneous variables. • In some studies, subjects can be exposed to more than one level of a treatment and thus serve as their own controls, although such crossover (repeated measures) designs may be unsuitable because of potential carry-over effects. • A third control technique is homogeneity—the use of a homogeneous sample of subjects to eliminate variability on characteristics that could affect study outcomes. • Extraneous variables can also be built into the design of a study as independent (blocking) variables, as in the case of a randomized block design. • Matching involves efforts to make groups comparable by matching subjects (either through pair-matching or balancing groups) on the basis of one or more extraneous variables. • Another technique is to control extraneous variables through statistical control. One such procedure is known as analysis of covariance. • Homogeneity, blocking, matching, and statistical control share one disadvantage: Researchers must know in advance which variables to control. • Four types of validity that affect the rigor of a quantitative study include statistical conclusion validity, construct validity, internal validity, and external validity.

CHAPTER 9 Enhancing Rigor in Quantitative Research

• Statistical conclusion validity concerns the strength of evidence that a relationship exists between two variables. Threats to statistical conclusion validity include low statistical power (the ability to detect true relationships among variables); low precision (the exactness of the relationships revealed after controlling extraneous variables); and factors that undermine a strong treatment. • Internal validity concerns the degree to which the results of a study can be attributed to the independent variable. Threats to internal validity include history (the occurrence of events external to an independent variable that can affect outcomes); selection (preexisting group differences; maturation (changes resulting from the passage of time); testing (effects of a pretest on outcomes); instrumentation (changes in the way data are gathered over time); and mortality (effects attributable to subject attrition). • External validity refers to the generalizability of study findings to other samples and settings. External validity is increased to the extent that the sample is representative of the population, and the study setting and experimental arrangements are representative of other environments. • The accessible population is the population from which a sample is drawn, and the target population represents a larger group of interest. Researchers should define the target population in terms of characteristics that are present in the accessible population. • Threats to external validity include expectancy effects (Hawthorne effect, placebo effect, nocebo effect); novelty effects; interaction of treatment and history effects; experimenter effects; and measurement effects. • A research design must balance the need for various types of validity, which sometimes compete with each other. STUDY ACTIVITIES Chapter 9 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforc-

221

ing the concepts presented in this chapter. In addition, the following study questions can be addressed: 1. How do you suppose the use of identical twins in a study could enhance control? 2. Read a research report suggested under the Studies Cited references in Chapter 8. Assess the adequacy of the control mechanisms used by the investigator, and recommend additional controls if appropriate. 3. For each of the following examples, indicate the types of design that could be used to study the problem (experimental, quasi-experimental, and so forth), the design you would recommend using, and how you would go about controlling extraneous variables. a. What effect does the presence of the newborn’s father in the delivery room have on the mother’s subjective report of pain? b. What is the effect of different types of bowel evacuation regimes on quadriplegic patients? c. Does the reinforcement of intensive care unit nonsmoking behavior in smokers affect postintensive care unit behaviors? d. Is the degree of change in body image of surgical patients related to their need for touch? SUGGESTED READINGS Methodologic References Beck, S. L. (1989). The crossover design in clinical nursing research. Nursing Research, 38, 291–293. Braucht, G. H., & Glass, G. V. (1968). The external validity of experiments. American Educational Research Journal, 5, 437–473. Clark, A. J. (1996). Optimizing the intervention in research studies. Advanced Practice Nursing Quarterly, 2, 1–4. Conlon, M., & Anderson, G. C. (1991). Three methods of random assignment: Comparison of balance achieved on potentially confounding variables. Nursing Research, 39, 376–378. Conn, V. S., Rantz, M. J., Wipke-Tevis, D. D., & Maas, M. L. (2001). Designing effective nursing interventions. Research in Nursing & Health, 24, 433–442. Cook, T. D., & Campbell, D. T. (1979). Quasi-experimental: Design and analysis issues for field settings. Boston: Houghton, Mifflin.

222

PART 3 Designs for Nursing Research

Fogg, L., & Gross, D. (2000). Threats to validity in randomized clinical trials. Research in Nursing & Health, 23, 79–87. Gilliss, C. L., & Kulkin, I. L. (1991). Technical notes: Monitoring nursing interventions and data collection in a randomized clinical trial. Western Journal of Nursing Research, 13, 416–422. Kempthorne, O. (1961). The design and analysis of experiments with some reference to educational research. In R. O. Collier & S. M. Elan (Eds.), Research design and analysis (pp. 97–126). Bloomington, IN: Phi Delta Kappa. Kerlinger, F. N., & Lee, H. B. (2000). Foundations of behavioral research (4th ed.). Orlando, FL: Harcourt College Publishers. Kwekkeboom, K. L. (1997). The placebo effect in symptom management. Oncology Nursing Forum, 24, 1393–1399. Lipsey, M. W. (1990). Design sensitivity: Statistical power for experimental research. Newbury Park, CA: Sage. McGuire, D. B., DeLoney, V., Yeager, K., Owen, D., Peterson, D., Lin, L., & Webster, J. (2000). Maintaining study validity in a changing clinical environment. Nursing Research, 49, 231–235. Motzer, S. A., Moseley, J. R., & Lewis, F. M. (1997). Recruitment and retention of families in clinical trials with longitudinal designs. Western Journal of Nursing Research, 19, 314–333. Rosenthal, R. (1976). Experimenter effects in behavioral research. New York: Halsted Press. Sidani, S. (1998). Measuring the intervention in effectiveness research. Western Journal of Nursing Research, 20, 621–635.

Studies Cited in Chapter 9 Bliss, D. Z., McLaughlin, J., Jung, H., Lowry, A., Savik, K., & Jensen, L. (2000). Comparison of the nutritional composition of diets of persons with fecal incontinence and that of age- and gender-matched controls. Journal of Wound, Ostomy, and Continence Nursing, 27, 90–97. Cowan, M. J., Pike, K. C., & Budzynski, H. K. (2001). Psychosocial nursing therapy following sudden cardiac arrest: Impact on two-year survival. Nursing Research, 50, 68–76.

Dougherty, M., Dwyer, J., Pendergast, J., Boyington, A., Tomlinson, B., Coward, R., Duncan, R. P., Vogel, B., & Rooks, L. (2002). A randomized trial of behavioral management for continence with older rural women. Research in Nursing & Health, 25, 3–13. Hundley, V., Milne, J., Leighton-Beck, L., Graham, W., & Fitmaurice, A. (2000). Raising research awareness among midwives and nurses. Journal of Advanced Nursing, 31, 78–88. Jones, P. S., Jaceldo, K. B., Lee, J. R., Zhang, X. E., & Meleis, A. I. (2001). Role integration and perceived health in Asian American women caregivers. Research in Nursing & Health, 24, 133–144. Kelleher, M. M. (2002). Removal of urinary catheters: Midnight vs. 0600 hours. British Journal of Nursing, 11, 84–90. Knebel, A. R., Bentz, E., & Barnes, P. (2000). Dyspnea management in alpha-1 antitrypsin deficiency: Effect of oxygen administration. Nursing Research, 49, 333–338. Moore, S. M., & Dolansky, M. A. (2001). Randomized trial of a home recovery intervention following coronary artery bypass surgery. Research in Nursing & Health, 24, 94–104. Moser, D. K., & Dracup, K. (2000). Impact of cardiopulmonary resuscitation training on perceived control in spouses of recovering cardiac patients. Research in Nursing & Health, 23, 270–278. Stevens, B., Johnston, C., Franck, L., Petryshen, P., Jack, A., & Foster, G. (1999). The efficacy of developmentally sensitive interventions and sucrose for relieving procedural pain in very low birth weight neonates. Nursing Research, 48, 35–43. Winterburn, S., & Fraser, R. (2000). Does the duration of postnatal stay influence breast-feeding rates at one month in women giving birth for the first time? Journal of Advanced Nursing, 32, 1152–1157. Wipke-Tevis, D. D., Stotts, N. A., Williams, D. A., Froelicher, E. S., & Hunt, T. K. (2001). Tissue oxygenation, perfusion, and position in patients with venous leg ulcers. Nursing Research, 50, 24–32. Zauszniewski, J. A., & Chung, C. W. (2001). Resourcefulness and health practices of diabetic women. Research in Nursing & Health, 24, 113–121.

10

Quantitative Research for Various Purposes

A

ll studies can be categorized as either experimental, quasi-experimental/preexperimental, or nonexperimental in design. This chapter describes types of quantitative research that vary according to the study’s purpose rather than along the intervention/control dimensions discussed in Chapter 8. The research described here is usually quantitative, but it is important to note that for certain types of research (e.g., evaluation research), qualitative methods may also be used. S T U D I E S T H AT A R E T Y P I C A L LY E X P E R I M E N TA L O R Q U A S I - E X P E R I M E N TA L In this section we describe types of research that usually involve an experimental or quasi-experimental design. In other words, the studies—or certain components of them—involve testing an intervention to determine its effects. Clinical Trials Clinical trials are studies designed to assess the effectiveness of clinical interventions. Methods associated with clinical trials were developed for medical and epidemiologic research, but nurse researchers are increasingly adopting these methods to test nursing interventions.

Phases of a Full Clinical Trial Clinical trials undertaken to test a new drug or an innovative therapy often are designed in a series of phases. • Phase I of the trial occurs after the initial development of the drug or therapy, and is designed primarily to determine things like drug dose (or strength of the therapy) and safety. This phase typically uses preexperimental designs (e.g., before—after without a control group). The focus is not on efficacy, but on developing the best possible (and safest) treatment. • Phase II of the trial involves seeking preliminary evidence of the effectiveness of the treatment as it has been designed in phase I, typically using preexperimental or quasiexperimental designs. During this phase, researchers ascertain the feasibility of launching a more rigorous test, seek evidence that the treatment holds promise, and look for signs of possible side effects. This phase is sometimes considered a pilot test of the treatment. There have been clinical trials of drug therapies that have shown such powerful effects during this phase that further phases were considered unnecessary (and even unethical), but this would rarely be the case in nursing studies.

224

PART 3 Designs for Nursing Research

Example of an early phase clinical trial: In preparation for a phase II trial for a cancer treatment, Schutta and Burnett (2000) gathered data that were useful for assessing the trial’s feasibility. The researchers focused on factors influencing patients’ decision not to participate in the phase I portion of the study. • Phase III is a full experimental test of the treatment, involving random assignment to an experimental or control group (or to orderings of treatment conditions). The objective of this phase is to arrive at a decision about whether the innovation is more effective than the standard treatment (or an alternative counterfactual). In addition to data about treatment effectiveness, however, researchers may collect data about safety and side effects. Any of the experimental designs discussed in Chapter 8 can be used in this phase of a trial. When the term clinical trial is used in the nursing literature, it most often is referring to a phase III trial, which may also be referred to as a randomized clinical trial or RCT. Phase III clinical trials often involve the use of a large and heterogeneous sample of subjects, frequently selected from multiple, geographically dispersed sites to ensure that findings are not unique to a single setting, and to increase the sample size and hence the power of the statistical tests. Multisite clinical trials are challenging administratively, requiring strong oversight and good systems of communication, staff supervision, and data management. • Phase IV of the trial occurs after the decision to adopt an innovative treatment has been made. In this phase, researchers focus primarily on longterm consequences of the intervention, including both benefits and side effects. This phase might use a nonexperimental, preexperimental, or quasiexperimental design (less often a true experimental design). In nursing, phase IV studies may be part of a utilization project (see Chapter 27). Example of a multisite randomized clinical trial: A nurse-managed intervention called the Women’s Initiative for Nonsmoking (WINS), developed on

the basis of a well-tested smoking-cessation intervention, was tailored specifically to meet the needs of women (Martin, Froelicher, & Miller, 2000). Ten hospitals in the San Francisco area participated in the trial. In each hospital, 50% of the subjects were assigned to the 3-month experimental condition or to a “usual care” group. Follow-up data are being collected at 6, 12, 24, and 30 months after baseline (Froelicher & Christopherson, 2000). Sequential Clinical Trials Traditional phase III clinical trials have important drawbacks in certain situations. In particular, it may take many months to recruit and randomize a sufficiently large sample; this is especially problematic if the population being treated is relatively small (e.g., people with a rare disease). Relatedly, in a standard clinical trial it may take months or years to draw conclusions about the intervention’s effectiveness (i.e., until all data have been collected and analyzed). An alternative is the sequential clinical trial in which experimental data are continuously analyzed as they become available. Results accumulate over time, so that the experiment can be stopped as soon as the evidence is strong enough to support a conclusion about the intervention’s efficacy. The design for this approach involves a series of “mini-experiments.” When the first patient becomes available for the study, he or she is randomly assigned (e.g., by a coin toss) to either the experimental (E) or control (C) condition. The next patient is then automatically assigned to the alternative condition, thereby creating a series of randomized paired comparisons. Most sequential trials use measures indicating preference for either the E or C condition. Preference can be defined qualitatively or quantitatively on the basis of clinically meaningful outcomes. Preference measures can include such indicators as survived/did not survive; showed improvement/showed no improvement; resulted in an increase of 20 degrees or more in range of motion/showed smaller or no increase in range of motion. Preference measures are dichotomous (i.e., have two possible outcomes). Using such preference measures, each pair is compared,

CHAPTER 10 Quantitative Research for Various Purposes

225

(U) is crossed, we would make a terminal decision to conclude that the experimental treatment is more effective. When the lower boundary (L) is crossed, the conclusion is that the control condition is more effective. Finally, when the middle boundary (M) is crossed, the decision is reached that the two treatments are equally effective (or ineffective). In this example, we tested a total of 18 non-tied pairs (for a total sample size of 36) and were then able to conclude that the experimental treatment was significantly superior to the control condition. Sequential trials have considerable appeal for clinical studies, because decisions typically can be reached much earlier than with traditional designs. However, these trials are not always appropriate (e.g., when three conditions are being compared), or are ambiguous if there are many ties. They may also be complicated if there are multiple outcomes of interest for which preferences have to be plotted separately. For more information about sequential trials, consult Portney and Watkins (2000) or Armitage (1975).

10

15

and there are three possibilities: E is preferred; C is preferred; or the two are tied. Ties are usually thrown out, and all remaining paired comparisons are plotted on graphs for which there are pre-established boundaries with decision rules. An example of such a graph is presented in Figure 10-1. The horizontal axis in the middle of this graph represents the number of randomized pairs (here, from 1 to 30 pairs, or 60 subjects). The vertical axis is used to indicate which way the “preference” comparison turned out. When the preference for a given pair favors E, the plotted line goes up; when it favors C, the plotted line goes down. The curved butterfly-shaped lines designate decision boundaries. In this example, the first comparison resulted in a preference for the experimental intervention, and so the graph plots a line from the origin to one unit up. The second comparison favored the control condition, and so the plot goes down where the number of pairs equals two. This procedure continues until the plot crosses one of the boundaries, which designate three stopping rules. When the upper boundary

Experimental condition is preferable U

Preference

5

E (+)

M No difference between conditions

10

15

20

25

(-) C -5

5

F I G U R E 1 0 . 1 Example of a sequential clinical trial graph.

-15

-10

M

L Control condition is preferable

Number of untied pairs 30

226

PART 3 Designs for Nursing Research

Evaluation Research Evaluation research is an applied form of research that involves finding out how well a specific program, practice, procedure, or policy is working. In evaluations, the research objective is utilitarian— the purpose is to answer the practical questions of people who must make decisions: Should a new program be adopted or an existing one discontinued? Do current practices need to be modified, or should they be abandoned altogether? Do the costs of implementing a new program outweigh the benefits? Clinical trials are sometimes evaluations. The multisite clinical trial of the WINS program used earlier as an example is also an evaluation of that program. The clinical trial is being used to determine if the WINS program is meeting the objective of reducing smoking. In general, the term evaluation research is used when researchers are trying to determine the effectiveness of a rather complex program, rather than when they are evaluating a specific entity (e.g., alternative drugs or sterilizing solutions). Thus, not all clinical trials would be called evaluations, and not all evaluations use methods associated with clinical trials. Moreover, evaluations often try to answer broader questions than simply whether an intervention is more effective clinically than care as usual. Evaluations often involve determining whether the intervention is cost-effective, for example. Evaluation research plays an important role both locally and nationally. Evaluations are often the cornerstone of an area of research known as policy research. Nurses have become increasingly aware of the potential contribution their research can make to the formulation of national and local health policies and thus are undertaking evaluations that have implications for policies that affect the allocation of funds for health services (Wood, 2000). In doing an evaluation, researchers often confront problems that are organizational, interpersonal, or political. Evaluation research can be threatening. Even when the focus of an evaluation is on a nontangible entity, such as a program, it is people who are implementing it. People tend to think

that they, or their work, are being evaluated and may feel that their jobs or reputation are at stake. Thus, evaluation researchers need to have more than methodologic skills—they need to be diplomats, adept in interpersonal relations with people. Evaluation Research Models Various schools of thought have developed concerning the conduct of evaluation research. The traditional strategy for an evaluation consists of four broad phases: determining program objectives, developing a means to measure the attainment of those objectives, collecting the data, and interpreting the data in terms of the objectives. It is often not easy to spell out program objectives. There may be many objectives, some of which are vague. The classic evaluation model stresses the importance of developing behavioral objectives. A behavioral objective is an intended program outcome stated in terms of the behavior of the people at whom the program is aimed, that is, the behavior of the beneficiaries, rather than the agents, of the program. Thus, if the goal is to have patients ambulate after surgery, the behavioral objective might be stated as, “The patient will walk the length of the corridor within 3 days after surgery.” The objective should not be stated as, “The nurse will teach the patient to walk the length of the corridor within 3 days after surgery.” An emphasis on behavioral objectives can be taken to extremes, however. An evaluation may be concerned with psychological dimensions such as morale or an emotion (e.g., anxiety) that do not always manifest themselves in behavioral terms. An alternative evaluation model is the goalfree approach. Proponents of this model argue that programs may have a number of consequences besides accomplishing their official objectives and that the classic model is handicapped by its inability to investigate these other effects. Goal-free evaluation represents an attempt to evaluate the outcomes of a program in the absence of information about intended outcomes. The job of the evaluator—a demanding one—is basically that of describing the repercussions of a program or practice

CHAPTER 10 Quantitative Research for Various Purposes

on various components of the overall system. The goal-free model can often be a profitable approach but, in many cases, the model may not be practical because there are seldom unlimited resources (personnel, time, or money) for an evaluation. Decision-makers may need to know whether objectives are being met so that immediate decisions can be made. Types of Evaluation Evaluations are undertaken to answer a variety of questions about a program or policy. Some questions involve the use of an experimental (or quasi-experimental) design, but others do not. In evaluations of large-scale interventions (sometimes called demonstrations if they are implemented on a trial basis), evaluators may well undertake all the evaluation activities discussed here. Process or Implementation Analysis. A process or implementation analysis is undertaken when there is a need for descriptive information about the process by which a program gets implemented and how it actually functions. A process analysis is typically designed to address such questions as the following: Does the program operate the way its designers intended? What are the strongest and weakest aspects of the program? What exactly is the treatment, and how does it differ (if at all) from traditional practices? What were the barriers to implementing the program successfully? How do staff and clients feel about the intervention? A process analysis may be undertaken with the aim of improving a new or ongoing program; in such a situation, it might be referred to as a formative evaluation. In other situations, the purpose of the process analysis is primarily to describe a program carefully so that it can be replicated by others—or so that people can better understand why the program was or was not effective in meeting its objectives. In either case, a process analysis involves an in-depth examination of the operation of a program, often involving the collection of both qualitative and quantitative data. This type of evaluation is descriptive and therefore nonexperimental.

227

Example of a process analysis: Root (2000) described the process of implementing a shared governance model in a California hospital’s surgical services department. The goal of the new model was to improve efficiency and morale by transferring decision-making to the staff level. Outcome Analysis. Evaluations typically focus on whether a program or policy is meeting its objectives. Evaluations that assess the worth of a program are sometimes referred to as summative evaluations, in contrast to formative evaluations. The intent of such evaluations is to help people decide whether the program should be discarded, replaced, modified, continued, or replicated. Many evaluation researchers distinguish between an outcome analysis and an impact analysis. An outcome analysis tends to be descriptive and does not use a rigorous experimental design. Such an analysis simply documents the extent to which the goals of the program are attained, that is, the extent to which positive outcomes occur. For example, a program may be designed to encourage women in a poor rural community to obtain prenatal care. An outcome analysis would document outcomes without rigorous comparisons. For example, the researchers might document the percentage of pregnant women in the community who had obtained prenatal care, the average month in which prenatal care was begun, and so on, and perhaps compare this information to existing preintervention community data. Example of an outcome analysis: Prozialeck and Pesole (2000) evaluated clinical outcomes for clients of a Family Case Management (FCM) program. They compared the birth outcomes (e.g., birth weight, gestational age, and public health nurse contacts) for women who had previously had a low-birth-weight infants and then became pregnant again and used FCM family services. Impact Analysis. An impact analysis attempts to identify the net impacts of a program, that is, the impacts that can be attributed exclusively to the program over and above the effects of the

228

PART 3 Designs for Nursing Research

counterfactual (e.g., standard treatment). It might be said that whereas outcomes analysis can describe effectiveness, impact analysis can demonstrate relative efficiency. Impact analyses use an experimental or quasi-experimental design because the aim of such evaluations is to attribute a causal influence to the special program. In the example cited earlier, let us suppose that the program to encourage prenatal care involved having nurses make home visits to women in the rural community to explain the benefits of early care during pregnancy. If the visits could be made to pregnant women on a random basis, the labor and delivery outcomes of the group of women receiving the home visits and of those not receiving them could be compared to determine the net impacts of the intervention, that is, the percentage increase in receipt of prenatal care among the experimental group relative to the control group. Many nursing evaluations are impact analyses, although they are not necessarily labeled as such. Impact analyses often involve subgroup analyses to determine the types of people for whom a program is most (and least) effective. For example, in our example of the rural outreach program, the researcher might compare program impacts for teenage mothers and older mothers, for multiparas and nulliparas, and so on. This would be done by comparing experimental and control group members for each subgroup. Example of an impact analysis: Ritz and co-researchers (2000) used a pretest–posttest experimental design to evaluate the effects of advanced practice nursing (APN) care on the quality of life and well-being of women diagnosed with breast cancer. The control group received standard medical care and the intervention group received standard care plus APN interventions. Impacts were assessed both for the overall sample and for subgroups (e.g., married and unmarried women). Cost Analysis. New programs or policies are often expensive to implement, but existing programs also may be expensive to operate. In our current situation of spiraling health care costs, program evaluations increasingly include a cost analysis to

determine whether the benefits of the program outweigh the monetary costs. Administrators and public policy officials make decisions about resource allocations for health services not only on the basis of whether something “works,” but also on the basis of whether it is economically viable. Cost—benefit analyses are typically done in connection with impact analyses and phase III clinical trials, that is, when researchers establish solid evidence regarding program effectiveness. As described by Chang and Henry (1999), there are several different types of cost analyses, the two most common of which are the following: • Cost–benefit analysis, in which monetary estimates are established for both costs and benefits. One difficulty with such an analysis is that it is sometimes difficult to quantify benefits of health services in monetary terms. There is also controversy about methods of assigning dollar amounts to the value of human life. Cost– benefit analyses are, however, the most widely used approach to cost analysis. • Cost-effectiveness analysis, which is used to compare health outcomes and resource costs of alternative interventions. Costs are measured in monetary terms, but outcome effectiveness is not. The point of such analyses is to estimate what it costs to produce impacts on outcomes that cannot easily be valued in dollars. This approach avoids the pitfalls of assigning dollar values to such outcomes as quality of life. However, without information on monetary benefits, such research faces more challenges in persuading decision-makers to make changes. Researchers doing such cost analyses need to document what it costs to operate both the new intervention and its alternative. For complex programs, cost analyses may need to show costs for individual program components. It may also be useful to identify costs for different subgroups whose resource requirements are expected to vary. In doing cost–benefit analyses, researchers need to think carefully about an array of possible short-term benefits (e.g., clients’ days of work missed in 6 months after the intervention) and

CHAPTER 10 Quantitative Research for Various Purposes

long-term benefits (e.g., years of productive work life). Often the cost–benefit analyst examines economic gains and losses from several different accounting perspectives—for example, for the target group; the hospital or facility implementing the program; third-party payers; employers; taxpayers; and society as a whole (i.e., the target group and taxpayers combined). Distinguishing these different perspectives is crucial if a particular program effect is a loss for one group (e.g., taxpayers) but a gain for another (e.g., the target group). Nurse researchers increasingly will be called on to become involved in such economic analyses. Duren-Winfield and her colleagues (2000) provide an excellent description of the methods used in a cost-effectiveness analysis of an exercise intervention for patients with chronic obstructive pulmonary disease. Example of a cost analysis: Wilson (2000) did a cost–benefit analysis of a city-wide, school-based hepatitis B vaccination program. The percentage of fully immunized sixth-grade students rose from 8% to over 80%. School-based administration of the vaccine was estimated as costing $1.46 less per dose than traditional methods. Wilson also estimated that over $20 million of health care costs could potentially be avoided through such a program. Intervention Research Both clinical trials and evaluations usually involve interventions. However, the term intervention research* is increasingly being used to describe a research approach distinguished not so much by a particular research methodology as by a distinctive process of planning, developing, implementing, testing, and disseminating interventions. The approach is being espoused by researchers and planners in different disciplines, including nursing (Rothman & Thomas, 1994; Sidani & Braden, 1998).

*The term intervention research is not used uniformly. It is often use to refer to any study involving an intervention, or to any study using an experimental design.

229

Proponents of the process are critical of the rather simplistic and atheoretical approach that is often used to design and evaluate nursing interventions. The recommended process for intervention research involves careful, collaborative planning at all steps, and the development of an intervention theory to guide the inquiry. More specifically, the process includes the following: 1. Project planning begins by putting together a project team with diverse clinical, research, and dissemination skills. The team may also include members of the target population or the affected community, resulting in what is sometimes called participatory research. The team’s initial job is to clearly define the problem to be solved, gather relevant information about the problem and prior solutions and interventions, and then develop an intervention theory that clearly articulates what must be done to achieve desired outcomes. The theory indicates, based on the best available knowledge, the nature of the clinical intervention, factors that would mediate the effects of clinical procedures on expected outcomes, and extraneous variables that would need to be controlled or considered as part of a test. 2. Intervention design flows from the intervention theory. The design of the intervention is done incrementally, building on early tests and refinements. The intervention design specifies not only what the clinical inputs would be but also such aspects as duration and intensity of the intervention. 3. Implementation of a data collection system begins before the intervention is introduced. Such advance data collection might detect aspects of the community or population of relevance to the intervention and possibly lead to further refinements of the intervention and its test. 4. Testing the intervention occurs in stages that are not dissimilar to the four phases of a clinical trial. An intervention prototype is developed, pilot tested, and then formally evaluated,

230

PART 3 Designs for Nursing Research

most often using experimental designs. If the intervention’s effectiveness is established, advanced testing focuses on identifying the subgroups for whom and settings in which effectiveness is strongest (and weakest). The final phase involves field tests in clinical settings. 5. Dissemination is a built-in feature of this model of research, which involves such activities as establishing standards for using the intervention, identifying possible markets, creating demand for the intervention, and making provisions to offer technical assistance. This model of intervention research is, at this point, more of an ideal than an actual practice. A few research teams have begun to implement portions of the model, and efforts are likely to expand. However, undertaking such a long-term, ambitious research agenda is clearly expensive. The ultimate effectiveness (both in terms of cost and in terms of health outcomes) of the full process—as opposed to more traditional approaches to designing and evaluating interventions—has yet to be established. Example of intervention research: Riesch and her colleagues (Riesch, Tosi, & Thurston, 1999; Riesch, Tosi, Thurston, Forsythe, Kuenning, & Kestly, 1993) undertook an intervention project that involved years of careful advance planning and collaboration with members of the community in which the intervention was implemented. The intervention involved communication skills training for adolescents and their parents. S T U D I E S T H AT C A N B E E I T H E R E X P E R I M E N TA L O R N O N E X P E R I M E N TA L The studies described in the previous section sometimes have a nonexperimental component, but, because they involve an intervention, almost always involve an experimental or quasi-experimental design as well. In this section we look at three types of research that can be experimental, but just as often are not.

Outcomes Research Outcomes research, designed to document the effectiveness of health care services, is gaining momentum as a research enterprise in nursing and health care fields. Outcomes research overlaps in some instances with evaluation research, but evaluation research more typically focuses on an appraisal of a specific new intervention, whereas outcomes research represents a more global assessment of nursing and health care services. The impetus for outcomes research comes from the quality assessment and quality assurance functions that grew out of the professional standards review organizations in the 1970s. Outcomes research represents a response to the increasing demand from policy makers, insurers, and the public to justify care practices and systems in terms of both improved patient outcomes and costs. The focus of outcomes research in the 1980s was predominantly on patient health status and costs associated with medical care, but there is a growing interest in studying broader patient outcomes in relation to nursing care. Although many nursing studies are concerned with examining patient outcomes and patient satisfaction, specific efforts to appraise and document the quality of nursing care—as distinct from the care provided by the overall health care system— are not numerous. A major obstacle is attribution— that is, linking patient outcomes to specific nursing actions or interventions, distinct from the actions of other members of the health care team. It is also difficult in some cases to determine a causal connection between outcomes and health care interventions because factors outside the health care system (e.g., patient characteristics) affect outcomes in complex ways. Nevertheless, outcomes research will likely gain momentum in this new century. Outcomes research has used a variety of traditional designs, sampling strategies, and data collection and analysis approaches, but is also developing a rich array of methods that are not within the traditional research framework. The complex and multidisciplinary nature of outcomes research

CHAPTER 10 Quantitative Research for Various Purposes

suggests that this evolving area will offer opportunities for methodologic creativity in the years ahead. Models of Health Care Quality In appraising quality in health care and nursing services, various factors need to be considered. Donabedian (1987), whose pioneering efforts created a framework for outcomes research, emphasized three factors: structure, process, and outcomes. The structure of care refers to broad organizational and administrative features. Structure can be appraised in terms of such attributes as size, location, range of services, type of facilities, technology, organization structure, and organizational climate. Nursing skill mix and nursing autonomy in decision-making are two structural variables that have been found to be related to patient outcomes. Processes involve aspects of clinical management, decision making, and clinical interventions. Outcomes refer to the specific clinical end results of patient care. Mitchell, Ferketich, and Jennings (1998) note that “the emphasis on evaluating quality of care has shifted from structures (having the right things) to processes (doing the right things) to outcomes (having the right things happen)” (p. 43). There have been several suggested modifications to Donabedian’s framework for appraising health care quality (e.g., Holzemer, 1994; Mitchell, Ferketich, & Jennings, 1998). Mitchell and her colleagues (1998), for example, have offered a model that is less linear and more dynamic than the original framework, and that takes client characteristics into account. Their model does not link interventions and processes to outcomes, but rather the effects of interventions are seen as mediated by client and system characteristics. Outcomes studies usually concentrate on various linkages within such models, rather than on testing an overall model. For example, researchers have studied the effect of health care structures on various health care processes and outcomes, although this has not been a major focus among nurses. Efforts have also begun on ways to accurately measure aspects of organizational structures from a nursing perspective (Aiken & Patrician,

231

2000; Brennan and Anthony, 2000). Most outcomes research in nursing, however, has focused on the process–patient–outcomes nexus. Example of research on structure: Lichtig, Knauf, and Milholland (1999) used data from California and New York to examine the effect of different nursing staffing patterns on patients’ length of stay in hospitals. Nursing Processes and Interventions To demonstrate nurses’ effects on health outcomes, researchers need to carefully describe and document (quantitatively and qualitatively) nurses’ clinical actions and behaviors. Examples of nursing process variables include macrolevel and microlevel nursing actions such as the following: • Nurses’ problem-solving skills • Clinical decision-making • Clinical competence • Nurses’ autonomy • Nursing intensity • Clinical leadership • Specific actions or interventions (e.g., communication, touch, clinical actions) There is increasing interest in describing the work that nurses do in terms of established classification systems and taxonomies, and there is also interest in maintaining complete, accurate, and systematic records of nursing actions in computerized data sets (often referred to as nursing minimal data sets or NMDS). A number of research-based classification systems of nursing interventions are being developed, refined, and tested, including the following: • Nursing Diagnoses Taxonomy of the North American Nursing Diagnosis Association or NANDA (North American Nursing Diagnosis Association, 1992); • Omaha System, a classification originally designed by the Omaha Visiting Nurse Association for community health nursing (Martin & Norris, 1996); • Nursing Intervention Classification (NIC), developed at the University of Iowa (Iowa Intervention Project, 1993);

232

PART 3 Designs for Nursing Research

• Home Health Care Classification or HHCC (Saba, 1992); • Ozbolt’s Patient Care Data Set (Ozbolt, Fruchtnight, & Hayden, 1994); • Perioperative Nursing Data Set (Kleinbeck, 1999); and • Nursing Management Minimum Data Set or NMMDS (Huber, Schumacher, & Delaney, 1997) We expect that many studies in the future will link processes from these classification systems to health outcomes. Studies with these classification systems have thus far focused on descriptions of patient problems and nursing interventions, and assessments of the utility of these systems. Example of classification system research: Bowles (2000) used two parts of the Omaha System (the Problem Classification Scheme and Intervention Scheme) to describe the types and frequency of problems in hospitalized elders and the interventions used by nurses. The study described the most frequently experienced problems and linked them to the most common nursing interventions. A major focus of outcomes research involves studies of the effects of nursing interventions on patient outcomes. When an intervention is new or has not been formally tested, outcomes studies can adopt a quasi-experimental or experimental design within an evaluation research framework. Patient Risk Assessment Variations in patient outcomes depend not only on the care patients receive, but also on differences in patient conditions and comorbidities. Adverse outcomes can occur no matter what nursing intervention is used. Thus, in evaluating the effects of nursing interventions on outcomes, there needs to be some way of controlling or taking into account patients’ risks for poor outcomes, or the mix of risks in a caseload. Risk adjustments have been used in a number of studies of the outcomes of medical care, and are only beginning to emerge in nursing outcomes studies. These studies typically involve the use of global measures of patient risks, such as the Case Mix Index

(Anderson, Su, Hsieh, Allred, Owensby, & JoinerRogers, 1999), or the APACHE III Acute Physiology Scale (Bakken, Dolter, & Holzemer, 1999). Outcomes Measuring outcomes and linking them to nursing actions is critical in developing an evidence-based practice and in launching high-quality improvement efforts. Outcomes of relevance to nursing can be defined in terms of physical or physiologic function (e.g., heart rate, blood pressure, complications), psychological function (e.g., comfort, life quality, satisfaction), or social function (e.g., relations with family members). Outcomes of interest to nurses may be either short term and temporary (e.g., postoperative body temperature) or more long term and permanent (e.g., return to regular employment). Furthermore, outcomes may be defined in terms of the end-results to individual patients receiving care, or to broader units such as a community or our entire society, and this would include cost factors. Just as there have been efforts to develop classifications of nursing interventions, work has begun on developing outcome classification systems. Of particular note is the Nursing-Sensitive Outcomes Classification (NOC), which has been developed by nurses at the University of Iowa College of Nursing to complement the Nursing Intervention Classification (Maas, Johnson, & Moorhead, 1996). Example of outcomes research: Greenberg (2000) studied the use of telephone nursing and telephone triage on such outcomes as client satisfaction, reduction in drop-in clinic visits, and unnecessary emergency department and urgent care visits associated with an outpatient pediatric clinic population. Replication Studies Replication studies are direct attempts to determine if findings obtained in an original piece of research can be duplicated in another independent study. Replications are appropriate for both

CHAPTER 10 Quantitative Research for Various Purposes

experimental/quasi-experimental and nonexperimental research. A strong evidence-based practice requires replications. Practice cannot be altered on the basis of a single isolated study, but must rely instead on an accumulation of evidence. Evidence can accumulate through a series of “close-enough-tocompare” studies, but deliberate replications offer special advantages in both establishing the credibility of research findings and extending their generalizability. There are, however, relatively few published replication studies in the nursing literature, perhaps reflecting a bias for original research on the part of both researchers and editors (and perhaps research funders).* As discussed by Beck (1994), there have been several attempts to classify replication strategies. One strategy is known as identical replication (or literal replication), which is an exact duplication of the original methods (e.g., sampling, measurement, analysis). Such exact duplication is rare, except in the case of a subsequent study by the original researcher. More common is virtual replication (or operational replication), which involves attempts to approximate the methods used in the reference study as closely as possible, but precise duplication is not sought. A third strategy is systematic extension replication (or constructive replication), in which methods are not duplicated, but there are deliberate attempts to test the implications of the original research. Many nursing studies that build on earlier research could be described as extension replications, but they usually are not so labeled and are not necessarily conceptualized as systematic extensions. Beck’s (1994) analysis of the nursing literature for the years 1983 through 1992 revealed very few examples of replications, and in the examples she found, there was considerable fuzziness about how the studies replicated the original ones. In addition to pointing out the need for more replication studies, she made several important recommendations: *Interestingly,

a textword search for the word “replication” in the CINAHL database yielded far more studies calling for replications in their conclusions than actual replication studies.

233

• Reports on replication studies should provide specific detail about what was replicated, and how. They should also make clear how the replication was similar to or different from the original. • The original research being replicated should be thoroughly critiqued, especially if modifications were made on the basis of any shortcoming. • Benchmarking—comparing the results of the original and replicated study—is essential. The comparison should be accompanied by conclusions about both the internal and external validity of the study findings. Many nurse researchers have called for more deliberate replication studies; the push for an evidence-based practice may strengthen their legitimacy as important scientific endeavors. Example of a replication study: Gaffney, Barndt-Maglio, Myers, and Kollar (2002) conducted a three-wave longitudinal study of the relationship between mothers’ experiences of discipline as children and their own disciplinary intentions with their own children. Their study replicated a study conducted in 1996 and extended the study by examining maternal behaviors as well as intentions. Methodologic Research Methodologic research refers to investigations of the ways of obtaining, organizing, and analyzing data. Methodologic studies address the development, validation, and evaluation of research tools or methods. Nurse researchers in recent years have become increasingly interested in methodologic research. This is not surprising in light of growing demands for sound and reliable outcome measures and for sophisticated procedures for obtaining and analyzing data. Most methodologic studies are descriptive and nonexperimental, often focusing on instrument development and testing. Suppose, for example, we developed and evaluated an instrument to accurately measure patients’ satisfaction with nursing care. In such a study, we would not examine levels of patient satisfaction or how satisfaction relates to

234

PART 3 Designs for Nursing Research

characteristics of nurses, hospitals, or patients. Our goals are to develop an effective and trustworthy instrument that can be used by others, and to determine our success in accomplishing this. Instrument development research is becoming increasingly important, and it often involves complex and sophisticated research designs and analyses. Those interested in more information on instrument development can consult a book such as that by Gable and Wolf (1993). Occasionally researchers use an experimental or quasi-experimental design to test competing methodologic strategies. For example, a researcher might test whether a financial incentive increases the number of volunteers willing to participate in a study. Prospective subjects could be randomly assigned to an incentive or no-incentive condition. The dependent variable in this case is whether people agree to participate. Methodologic research may appear less exciting than substantive research, but it is virtually impossible to conduct high-quality and useful research on a substantive topic with inadequate research methods. Studies of a methodologic nature are indispensable in any scientific discipline, and perhaps especially so in fields that deal with highly complex, intangible phenomena such as human behavior, as is the case with nursing. Example of a methodologic study: Mahon, Yarcheski, and Yarcheski (2002) administered the Personal Lifestyle Questionnaire (PLQ), a widely used measure of positive health practices for adults, to a sample of 222 adolescents to evaluate its reliability and validity with young people. Their study concluded that a subscale of the PLQ yielded an adequate measure of general health practices with adolescents. S T U D I E S T H AT A R E T Y P I C A L LY N O N E X P E R I M E N TA L In the types of study described in the following sections, researchers typically do not have an option of controlling independent variables, and so the studies are nonexperimental.

Survey Research A survey is designed to obtain information about the prevalence, distribution, and interrelations of variables within a population. The decennial census of the U.S. population is one example of a survey. Political opinion polls, such as those conducted by Gallup or Harris, are other examples. When surveys use samples of individuals, as they usually do, they may be referred to as sample surveys (as opposed to a census, which covers the entire population). Surveys obtain information from a sample of people by means of self-report—that is, study participants respond to a series of questions posed by investigators. Surveys, which tend to yield quantitative data primarily, may be cross-sectional or longitudinal (e.g., panel studies). The greatest advantage of survey research is its flexibility and broadness of scope. It can be applied to many populations, it can focus on a wide range of topics, and its information can be used for many purposes. The information obtained in most surveys, however, tends to be relatively superficial: surveys rarely probe deeply into such complexities as contradictions of human behavior and feelings. Survey research is better suited to extensive rather than intensive analysis. Although surveys can be conducted within the context of large-scale experiments, surveys are usually done as nonexperimental studies. Survey Content The content of a self-report survey is essentially limited only by the extent to which respondents are able and willing to report on the topic. Any information that can reliably be obtained by direct questioning can be gathered in a survey, although surveys include mostly questions that require brief responses (e.g., yes/no, always/sometimes/never). Often, surveys focus on what people do or how they feel: what they eat, how they care for their health, their compliance in taking medications, how anxious they are, and so forth. In some instances, the emphasis is on what people plan to do—how they plan to vote, for example. Surveys also collect information on people’s knowledge, opinions, attitudes, and values.

CHAPTER 10 Quantitative Research for Various Purposes

Survey Administration Survey data can be collected in a number of ways, but the most respected method is through personal interviews (or face-to-face interviews), in which interviewers meet in person with respondents to ask them questions. In general, personal interviews are rather costly: They require considerable planning and interviewer training and tend to involve a lot of personnel time. Nevertheless, personal interviews are regarded as the best method of collecting survey data because of the quality of information they yield. A further advantage of personal interviews is that relatively few people refuse to be interviewed in person. Example of a survey with personal interviews: Polivka, Nickel, Salsberry, Kuthy, Shapiro, and Slack (2000) conducted in-person interviews with a sample of 474 low-income women with preschool-age children to explore factors associated with the children’s hospitalizations and emergency department use. The interviews, conducted in clinics or human service agency offices, lasted about 20 to 25 minutes. Telephone interviews are a less costly, but often less effective, method of gathering survey information. When the interviewer is unknown, respondents may be uncooperative in a telephone situation. Telephoning can, however, be a convenient method of collecting information if the interview is short, specific, and not too personal, or if researchers have had prior personal contact with respondents. As the following example shows, telephone interviews may be difficult for certain groups of respondents, including low-income people (who do not always have a telephone) or the elderly (who may have hearing problems). Example of a telephone survey: Pesata, Pallija, and Webb (1999) conducted a telephone survey to determine why some children missed their clinic appointments, and to explore barriers to health care. A sample of 200 parents with a history of missed appointments were selected for the survey. Of these 200, nearly half (95) did not have a telephone.

235

Because of the expense of collecting in-person data, researchers sometimes adopt what is referred to as a mixed-mode strategy to collecting survey data. In this approach, an interviewer first attempts to interview a sample member by telephone. If attempts at conducting a telephone interview fail (either because there is no telephone or the person refuses to be interviewed by phone), an interviewer then attempts a personal interview to collect the data (or goes to the subject’s home with a cellular telephone and personally asks the person to complete the interview by phone). Questionnaires differ from interviews in that they are self-administered. (They are sometimes referred to as SAQs, that is, self-administered questionnaires.) Respondents read the questions on a written form and give their answers in writing. Because respondents differ in their reading levels and in their ability to communicate in writing, questionnaires are not merely a printed form of an interview schedule. Great care must be taken in developing questionnaires to word questions clearly, simply, and unambiguously. Self-administered questionnaires are economical but are not appropriate for surveying certain populations (e.g., the elderly, children). In survey research, questionnaires are often distributed through the mail, but may also be distributed in other ways (e.g., through the Internet). Example of a mailed survey: Havens (2000) studied factors associated with the execution or nonexecution of advance directives for health care (i.e., living wills or durable powers of attorney). She collected her data from a sample of community-dwelling adults through mailed questionnaires. Needs Assessments As the name implies, a needs assessment is a study in which researchers collect data to estimate the needs of a group, community, or organization. The aim of such a study is to determine if there is a need for a special intervention or outreach effort, or if a program is meeting the needs of those who are

236

PART 3 Designs for Nursing Research

supposed to benefit from it. Nursing educators may wish to assess the needs of their clients (students); hospital nurses may wish to learn the needs of those they serve (patients); public health nurses may wish to gather information on the needs of some target population (e.g., adolescents in the community). Because resources are seldom limitless, information that can help in establishing priorities can be valuable. There are various methods of doing a needs assessment, and these methods are not mutually exclusive. The key informant approach collects information about a group’s needs from people who are in a key position to know those needs. These key informants could be community leaders, prominent health care workers, agency directors, or other knowledgeable individuals. Questionnaires or interviews are usually used to collect the data. (In some cases, key informant interviews are used to collect narrative, qualitative information rather than quantitative data.) Needs assessments most often use a survey approach, which involves collecting data from a sample of the group whose needs are being assessed. In a survey, questioning would not be restricted to people who have special expertise. A representative sample from the group or community would be asked about their needs. Another alternative is to use an indicators approach, which relies on facts and statistics available in existing reports or records. For example, a nurse-managed clinic that is interested in analyzing the needs of its clients could examine over a 5-year period the number of appointments that were kept, the employment rate of its clients, the changes in risk appraisal status, methods of payment, and so forth. The indicators approach is cost-effective because the data are available but need organization and interpretation. Needs assessments almost always involve the development of recommendations. Researchers conducting a needs assessment usually offer judgments about priorities based on their results (taking costs and feasibility into consideration), and may also offer advice about the means by which the most highly prioritized needs can be addressed.

Example of a needs assessment: Patterson, Moylan, Bannon, and Salih (2000) used a survey approach to investigate the need for and level of interest in five types of cancer-related information (medical, psychological, and so on). A questionnaire was distributed to a cancer population and their families in South Western Sydney. The needs assessment resulted in the development of a cancer education program. Secondary Analysis Secondary analysis involves the use of data gathered in a previous study to test new hypotheses or explore new relationships. In a typical study, researchers collect far more data than are actually analyzed. Secondary analysis of existing data is efficient and economical because data collection is typically the most time-consuming and expensive part of a research project. Nurse researchers have used a secondary analysis approach with both large national data sets and smaller, more localized sets. A number of avenues are available for making use of an existing set of quantitative data: 1. Variables and relationships among variables that were previously unanalyzed can be examined (e.g., an dependent variable in the original study could become the independent variable in the secondary analysis). 2. Data that were collected for nonresearch purposes can be used to answer research questions. 3. The secondary analysis can focus on a particular subgroup rather than on the full original sample (e.g., survey data about health habits from a national sample could be analyzed to study smoking among urban teenagers). 4. The unit of analysis can be changed. A unit of analysis is the basic unit that yields data for an analysis; in nursing studies, each individual subject is typically the unit of analysis. However, data are sometimes aggregated to yield information about larger units (e.g., a study of individual nurses from 25 hospitals could be converted to aggregated data about the hospitals).

CHAPTER 10 Quantitative Research for Various Purposes

Researchers interested in performing secondary analyses must undertake several preparatory activities. After determining the research question and identifying data needs, researchers must identify, locate, and gain access to appropriate databases. They should then do a thorough assessment of the identified data sets in terms of their appropriateness for the research question, adequacy of data quality, and technical usability of the data. An important source of data for secondary analysis are the various clinical nursing databases that are available as management and policy tools. Nail and Lange (1996) note that such databases have great potential for research on the processes and outcomes of nursing care. A number of groups, such as university institutes and federal agencies, have made efforts to make survey data available to researchers for secondary analysis. The policies regulating public use of data vary from one organization to another, but it is not unusual for a researcher to obtain a data set at roughly the cost of duplicating data files and documentation. Thus, in some cases in which data collection originally cost hundreds of thousands of dollars, reproduced materials may be purchased for less than 1% of the initial costs. Some universities and research institutes in universities maintain libraries of data sets from large national surveys. Surveys sponsored by the National Center for Health Statistics (NCHS) and other government agencies are an important resource for secondary analysis. For example, NCHS periodically conducts such national surveys as the National Health Interview Survey, the Health Promotion and Disease Prevention Survey, and the National Comorbidity Survey, which gather health-related information from thousands of people all over the United States. The use of available data makes it possible to bypass time-consuming and costly steps in the research process, but there are some noteworthy disadvantages in working with existing data. In particular, if researchers do not play a role in collecting the data, the chances are fairly high that the data set will be deficient in one or more ways, such as in the sample used, the variables measured, and

237

so forth. Researchers may continuously face “if only” problems: if only they had asked questions on a certain topic or had measured a particular variable differently. Nevertheless, opportunities for secondary analysis are worth exploring. Example of a secondary analysis: Clarke, Frasure-Smith, Lespérance, and Bourassa (2000) used existing data from clinical trials sponsored by the U.S. National Heart, Lung, and Blood Institute (Studies of Left Ventricular Dysfunction Prevention and Treatment). Their analysis, which combined members of the original experimental and control groups, focused on psychosocial and other factors that were predictors of 1-year functional status. Meta-Analysis Chapter 5 described the function of a literature review as a preliminary step in a research project. However, careful and systematic integration of research findings constitutes an important scholarly endeavor that can contribute new knowledge— knowledge that can play a key role in developing an evidence-based practice. The procedure known as meta-analysis represents an application of statistical procedures to findings from research reports. In essence, meta-analysis treats the findings from one study as a single piece of data: The study is itself the unit of analysis. The findings from multiple studies on the same topic can be combined to yield a data set that can be analyzed in a manner similar to that obtained from individual subjects. Traditional narrative reviews of the literature are handicapped by several factors. The first is that if the number of studies on a topic is large and if the results are inconsistent, then it is difficult to draw conclusions. Moreover, narrative reviews are often subject to potential biases. Researchers may unwittingly give more weight to findings that are congruent with their own viewpoints. Meta-analytic procedures provide an objective method of integrating a large body of findings and of observing patterns and relationships that might otherwise have gone undetected. Furthermore, meta-analysis provides information about the magnitude of

238

PART 3 Designs for Nursing Research

differences and relationships. Meta-analysis can thus serve as an important scholarly tool in theory development as well as in research utilization. Because of the importance of meta-analyses, we present more information on how to do them in Chapter 27. Meta-analytic techniques can be used in contexts other than integrative literature reviews. For example, if a clinical trial was implemented in 15 sites, a meta-analysis could be conducted, using results from each site as one piece of data. This is particularly useful if pooling of the raw data across all sites is not possible (e.g., if the intervention was not implemented the same way in the different sites). Example of a meta-analysis: Evans (2002) conducted a meta-analysis of the effectiveness of music as an intervention for hospital patients. All 19 studies included in the meta-analysis were clinical trials. The results indicated that music played through headphones reduces patient anxiety during normal care delivery. Delphi Surveys Delphi surveys were developed as a tool for shortterm forecasting. The technique involves a panel of experts who are asked to complete a series of questionnaires focusing on their opinions, predictions, or judgment about a topic of interest. The Delphi technique differs from other surveys in several respects. In a Delphi survey, each expert is asked to complete several rounds of questionnaires. Multiple iterations are used to achieve consensus, without the necessity of face-to-face discussion. A second feature is the use of feedback to panel members. Responses to each round of questionnaires are analyzed, summarized, and returned to the experts with a new questionnaire. The experts can then reformulate their opinions with the group’s viewpoint in mind. The process of response—analysis—feedback—response is usually repeated at least three times until a general consensus is obtained. The Delphi technique is an efficient means of combining the expertise of a large, geographically

dispersed group for planning and prediction purposes. The experts are spared the necessity of being brought together for a formal meeting, thus saving time and expense. Another advantage is that a persuasive or prestigious expert cannot have an undue influence on the opinions of others, as could happen in a face-to-face situation. All panel members are on an equal footing. Anonymity probably encourages greater frankness than might be expressed in a formal meeting. The feedback—response loops allow for multichannel communication without any risk of the members being sidetracked from their mission. However, the Delphi technique is timeconsuming for researchers. Experts must be solicited, questionnaires prepared and mailed, responses analyzed, results summarized, new questionnaires prepared, and so forth. The cooperation of the panel members may wane in later rounds of the questionnaire mailings. The problem of bias through attrition is a constant concern. Another concern is how to define consensus (i.e., how many participants have to agree before researchers conclude that consensus has been achieved). Recommendations range from a liberal 51% to a more cautious 70%. On the whole, the Delphi technique represents a significant methodologic tool for problem-solving, planning, and forecasting. Example of a Delphi survey: Scheffer and Rubenfeld (2000) conducted a five-round Delphi survey, involving an international panel of experts from nine countries, to define critical thinking in nursing. A consensus definition of critical thinking in nursing was achieved; the panel also identified 7 skills and 10 habits of the mind that contribute to critical thinking. RESEARCH EXAMPLES O F VA R I O U S T Y P E S O F Q U A N T I TAT I V E S T U D I E S Studies of each type of research presented in this chapter have already been noted. Two studies are described in greater detail in the following sections.

CHAPTER 10 Quantitative Research for Various Purposes

Research Example of a Clinical Trial Simpson, Parsons, Greenwood, and Wade (2001) noted that many pregnant women consume raspberry leaf herb because of the widespread belief that it shortens labor and makes labor easier. Although the use of the herb is promoted by some midwives and doctors, there is little scientific evidence about its effectiveness or safety. The researchers undertook a randomized clinical trial to address questions about whether regular intake of raspberry leaf tablets has adverse effects on the mother or baby, and whether it is effective in shortening labor. The researchers were careful in estimating how many subjects would be needed to achieve adequate statistical power in their analyses. Based on their calculations, a total of 240 low-risk nulliparous women were recruited from a hospital in Sydney, Australia. Women who agreed to participate were randomly assigned to either the treatment group or a control group, which received a placebo. Because this was the first study of its kind, a conservative dose of the raspberry herb (2.4 grams per day) was tested. Raspberry herb was added to the mixture in the placebo tablets (calcium phosphate and other constituents). Women in both groups were given tablets, which they were instructed to ingest twice daily (with breakfast and the evening meal) beginning in their thirty-second week of gestation. Tablets were distributed in a double-blind manner, with neither subjects nor those giving the pills aware of which women were in the experimental group. Some 48 subjects withdrew from the study for various reasons, leaving 192 subjects (96 per group). The two treatment groups were compared with regard to demographic characteristics and found to be comparable in terms of age, weight, and ethnicity. The investigators also examined compliance with tablet consumption and found that the average rate of compliance was 89% (i.e., 89% of the prescribed tablets were consumed). The researchers began their analyses by examining the safety of ingesting the raspberry leaf tablets. The two groups were compared in terms of such factors as maternal blood loss, maternal diastolic blood pressure prebirth and postbirth, length of gestation, infant birth weight, and infant Apgar score at 5 minutes. The researchers also explored whether there might be side effects of the herb, and so asked subjects to report various symptoms (e.g., vomiting, nausea, dizziness) at each antenatal visit. The researchers found no adverse effects of the herb for mother or baby.

239

The main outcome variable was length of time in labor. Contrary to popular belief, the raspberry leaf tablets did not shorten the first stage of labor, but the experimental group did have a significantly shorter second stage of labor than the control group (a difference of about 10 minutes). There was also a lower rate of forceps delivery among those in the treatment group (19% versus 30%), perhaps the result of the shorter second stage. The researchers concluded that although the tablets did not result in anticipated effects on the length of first-stage labor, their effects on second-stage labor duration and on the need for artificial rupture of membranes were clinically significant. They recommend further clinical trials, with an emphasis on finding the optimal dose that will maintain safety while producing beneficial effects.

Research Example of a Methodologic Study Within a Survey Neumark, Stommel, Given, and Given (2001) were involved in the conduct of a longitudinal survey of older families with cancer and their caregivers, the Family Care Study. Survey respondents were interviewed by telephone four times over the first year after a cancer diagnosis, and also completed four rounds of selfadministered questionnaires. Subjects for the study were recruited over a 3-year period in multiple hospitals and cancer treatment centers in two states. The researchers undertook a methodologic study within the context of this survey research, to identify factors that could account for loss of subjects in the earliest phases of sample accrual. They compared three groups: eligible patients who declined to participate (nonconsenters), patients who originally consented to participate but then later declined (early dropouts), and subjects who actually took part in the study (participants). The researchers examined two broad types of factors that might help to explain nonparticipation in the study: subject characteristics and research design characteristics. The aim was to obtain information that would benefit others in designing studies and recruiting subjects. The researchers had recruited subjects in facilities that were fortunately able to provide basic demographic and illness-related characteristics of all subjects eligible for the study (e.g., age, race, gender, cancer diagnosis). In addition, the researchers

240

PART 3 Designs for Nursing Research

carefully maintained records about aspects of the research design for each eligible person who was recruited. For example, they had information about the type of recruiter used, whether recruiters were reimbursed for obtaining consent, whether a family caregiver was involved, and in what phase of data collection the recruitment occurred. The researchers found that both subject and research design factors contributed to nonparticipation. Older patients, for example, were significantly less likely to consent (but not more likely to be an early dropout once they had consented). In terms of design features, the most powerful factor was whether a family caregiver was approached to participate. Patients were more likely to give consent and less likely to drop out early when caregivers were also approached. Also, paid recruiters were notably more successful in getting subjects’ consent than unpaid recruiters—but these subjects were more likely to drop out early, neutralizing the effect of recruiter payment. The researchers concluded that predicting nonparticipation “may help target recruitment and retention efforts, particularly in reducing the extent to which study-related factors contribute to attrition” (p. 368).

S U M M A RY P O I N T S • Quantitative studies vary according to purpose as well as design. Studies that almost always involve an experimental or quasi-experimental design include clinical trials, evaluations, and intervention research. • Clinical trials, which are studies designed to assess the effectiveness of clinical interventions, are often designed in a series of phases. • Phase I of a clinical trial is designed to finalize the features of the intervention. Phase II involves seeking preliminary evidence of the effectiveness of the treatment as designed in phase I. Phase III is a full experimental test of the treatment, often referred to as a randomized clinical trial. In phase IV, the researcher focuses primarily on long-term consequences of the intervention, including both benefits and side effects. • In a sequential clinical trial, experimental data from paired “mini-experiments” are continu-

ously analyzed; the experiment is stopped as soon as the evidence supports a conclusion about the efficacy of the intervention. • Most sequential trials use measures indicating preference for either the experimental or control condition for each pair of observations. Preferences are plotted on special graphs until the plot crosses one of the boundaries, which designate stopping rules for the experiment. • Evaluation research assesses the effectiveness of a program, policy, or procedure to assist decision-makers in choosing a course of action. • The classic evaluation model assesses the congruence between the goals of the program and actual outcomes; the goal-free model attempts to understand all the effects of a program, regardless of whether they were intended. • Evaluations can answer a variety of questions. Process or implementation analyses describe the process by which a program gets implemented and how it functions in practice. Outcome analyses describe the status of some condition after the introduction of an intervention. Impact analyses test whether an intervention caused any net impacts relative to the counterfactual. Cost analyses seek to determine whether the monetary costs of a program are outweighed by benefits and include both cost–benefit analyses and cost-effectiveness analyses. • Intervention research is a term sometimes used to refer to a distinctive process of planning, developing, implementing, testing, and disseminating interventions. A key feature of this process is the development of an intervention theory from which the design and evaluation of an intervention flow. • Outcomes research is undertaken to document the quality and effectiveness of health care and nursing services. A model of health care quality encompasses several broad concepts, including: structure (factors such as accessibility, range of services, facilities, and organizational climate); process (nursing interventions and actions); client risk factors (e.g., severity of illness and case mix of the caseload); and outcomes (the

CHAPTER 10 Quantitative Research for Various Purposes

specific end-results of patient care in terms of patient functioning). • Replication studies include identical replications (exact duplication of methods of an earlier study in a new study); virtual replication (close approximation but not exact duplication of methods); and systematic extension replication (deliberate attempts to test the implications of the original research). • In methodologic research, the investigator is concerned with the development, validation, and assessment of methodologic tools or strategies. • Survey research examines people’s characteristics, behaviors, attitudes, and intentions by asking them to answer a series of questions. • The preferred survey method is through personal interviews, in which interviewers meet respondents face-to-face and question them. Telephone interviews are more economical, but are not recommended if the interview is long or detailed or if the questions are sensitive or personal. Questionnaires are self-administered (i.e., questions are read by respondents, who then give written responses). • Needs assessments are studies to document the needs of a group or community. The three main techniques used to conduct needs assessments include the key informant, survey, or indicator approach. • Secondary analysis refers to studies in which researchers analyze previously collected data. The secondary analyst may examine unanalyzed variables, test unexplored relationships, focus on a particular subsample, or change the unit of analysis. • Meta-analysis is a method of integrating the findings of prior research using statistical procedures, counting each study as one unit of analysis. • The Delphi technique is a method of problem solving in which several rounds of questionnaires are mailed to a panel of experts. Feedback from previous questionnaires is provided with each new questionnaire so that the experts can converge on a consensus opinion in subsequent rounds.

241

STUDY ACTIVITIES Chapter 10 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Suppose you were interested in doing a survey of nurses’ attitudes toward caring for cancer patients. Would you use a personal interview, telephone interview, or questionnaire to collect your data? Defend your decision. 2. A psychiatric nurse therapist working with emotionally disturbed children is interested in evaluating a program of play therapy. Explain how you might proceed if you were to use (a) the classic evaluation model and (b) a goalfree approach. Which approach do you think would be more useful, and why? 3. Explain how you would use the key informant, survey, and indicator approaches to assess the need to teach Spanish to nurses in a given community. 4. In the research example by Neumark and colleagues (2001), what were the dependent and independent variables? How might other researchers benefit from this research? SUGGESTED READINGS Methodologic References Armitage, P. (1975). Sequential medical trials (2nd ed.). New York: John Wiley. Beck, C. T. (1994). Replication strategies for nursing research. Image: The Journal of Nursing Scholarship, 26, 191–194. Brown, J. S., & Semradek, J. (1992). Secondary data on health-related subjects: Major sources, uses, and limitations. Public Health Nursing, 3, 162–171. Brown, S. A. (1991). Measurement of quality of primary studies for meta-analysis. Nursing Research, 40, 352–355. Chang, W., & Henry, B. M. (2000). Methodologic principles of cost analyses in the nursing, medical, and health services literature, 1990–1996. Nursing Research, 48, 94–104.

242

PART 3 Designs for Nursing Research

Crisp, J., Pelletier, D., Duffield, C., Adams, A., & Nagy, S. (1997). The Delphi method? Nursing Research, 46, 116–118. Dillman, D. (1978). Mail and telephone surveys: The Total Design Method. New York: John Wiley and Sons. Donabedian, A. (1987). Some basic issues in evaluating the quality of health care. In L. T. Rinke (Ed.), Outcome measures in home care (Vol. I, pp. 3–28). New York: National League for Nursing. Duren-Winfield, V., Berry, M. J., Jones, S. A., Clark, D. H., & Sevick, M. A. (2000). Cost-effectiveness analysis methods for the REACT study. Western Journal of Nursing Research, 22, 460–474. Fetter, M. S., Feetham, S. L., D’Apolito, K., Chaze, B. A., Fink, A., Frink, B., Hougart, M., & Rushton, C. (1989). Randomized clinical trials: Issues for researchers. Nursing Research, 38, 117–120. Fowler, F. J. (1993). Survey research methods (2nd ed.). Beverly Hills, CA: Sage. Gable, R., & Wolf, M. (1993). Instrument development in the affective domain (2nd ed.). Hingham, MA: Kluwer Academic Publishers. Glass, G. V., McGaw, B., & Smith, M. L. (1981). Metaanalysis of social research. Beverly Hills, CA: Sage. Hasson, F., Keeney, S., & McKenna, H. (2000). Research guidelines for the Delphi survey technique. Journal of Advanced Nursing, 32, 1008–1015. Holzemer, W. L. (1994). The impact of nursing care in Latin America and the Caribbean: A focus on outcomes. Journal of Advanced Nursing, 20, 5–12. Huber, D., Schumacher, L., & Delaney, C. (1997). Nursing Management Minimum Data Set. Journal of Nursing Administration, 27, 42–48. Iowa Intervention Project. (1993). The NIC taxonomy. Image: The Journal of Nursing Scholarship, 25, 187–192. Jacobson, A. F., Hamilton, P., & Galloway, J. (1993). Obtaining and evaluating data sets for secondary analysis in nursing research. Western Journal of Nursing Research, 15, 483–494. Kirchoff, K. T., & Dille, C. A. (1994). Issues in intervention research: Maintaining integrity. Applied Nursing Research, 7, 32–37. Kleinbeck, S. V. M. (1999). Development of the perioperative nursing data set. AORN Journal, 70, 15–28. Maas, M. L., Johnson, M., & Moorhead, S. (1996). Classifying nursing-sensitive patient outcomes. Image: The Journal of Nursing Scholarship, 28, 295–301.

Martin, K. S., & Norris, J. (1996). The Omaha System: A model for describing practice. Holistic Nursing Practice, 11, 75–83. McCain, N. L., Smith, M. C., & Abraham, I. L. (1986). Meta-analysis of nursing interventions. Western Journal of Nursing Research, 8, 155–167. McKillip, J. (1986). Needs analysis: Tools for the human services and education. Beverly Hills, CA: Sage. Mitchell, P. H., Ferketich, S., & Jennings, B. M. (1998). Quality health outcomes model. Image: The Journal of Nursing Scholarship, 30, 43–46. Nail, L. M., & Lange, L. L. (1996). Using computerized clinical nursing data bases for nursing research. Journal of Professional Nursing, 12, 197–206. Neuliep, J. W. (Ed.). (1990). Handbook of replication research in the behavioral and social sciences [Special issue]. Journal of Social Behavior and Personality, 15 (4). North American Nursing Diagnosis Association. (1992). NANDA nursing diagnoses: Definitions and classification, 1992. Philadelphia: Author. Ozbolt, J., Fruchtnight, J. N., & Hayden, J. (1994). Toward clinical standards for clinical nursing information. Journal of the American Medical Informatics Association, 1, 175–185. Pocock, S. J. (1996). Clinical trials: A practical approach. New York: John Wiley. Portnoy, L. G., & Watkins, M. P. (2000). Foundations of clinical research: Applications to practice (2nd ed.). Upper Saddle River, NJ: Prentice-Hall Heath. Reynolds, N. R., Timmerman, G., Anderson, J., & Stevenson, J. S. (1992). Meta-analysis for descriptive research. Research in Nursing & Health, 15, 467–475. Rossi, P. H., & Freeman, H. E. (1993). Evaluation: A systematic approach (5th ed.). Beverly Hills, CA: Sage. Rothman, J. & Thomas, E. J. (Eds.). (1994). Intervention research: Design and development for human service. New York: Haworth Press. Saba, V. (1992). The classification of home health care nursing: Diagnoses and interventions. Caring, 11, 50–57. Sidani, S. (1996). Methodological issues in outcomes research. Canadian Journal of Nursing Research, 28, 87–94. Sidani, S., & Braden, C. J. (1998). Evaluating nursing interventions: A theory driven approach. Thousand Oaks, CA: Sage. Stewart, D. W., & Kamins, M. A. (1993). Secondary research: Information sources and methods. (2nd ed.). Thousand Oaks, CA: Sage.

CHAPTER 10 Quantitative Research for Various Purposes Warheit, G. J., Bell, R. A., & Schwab, J. J. (1975). Planning for change: Needs assessment approaches. Washington, DC: National Institute of Mental Health. Witkin, B. R., & Altschuld, J. W. (1995). Planning and conducting needs assessments. Thousand Oaks, CA: Sage. Wood, M. J. (2000). Influencing health policy through research. Clinical Nursing Research, 9, 2113–216.

Studies Cited in Chapter 10 Aiken, L. H., & Patrician, P. A. (2000). Measuring organizational traits of hospitals: The Revised Nursing Work Index. Nursing Research, 49, 146–153. Anderson, R. A., Su, H., Hsieh, P., Allred, C. A., Owensby, S., & Joiner-Rogers, G. (1999). Case mix adjustment in nursing systems research. Research in Nursing & Health, 22, 271–283. Bakken, S., Dolter, K. J., & Holzemer, W. L. (1999). A comparison of three strategies for risk-adjustment of outcomes for AIDS patients hospitalized for Pneumocystis carinii pneumonia. Journal of Advanced Nursing, 30, 1424–1431. Bowles, K. H. (2000). Patient problems and nurse interventions during acute care and discharge planning. Journal of Cardiovascular Nursing, 14, 29–41. Brennan, P. F., & Anthony, M. K. (2000). Measuring Nursing Practice Models using multi-attribute utility theory. Research in Nursing & Health, 23, 372–382. Clarke, S. P., Frasure-Smith, N., Lespérance, F., & Bourassa, M. G. (2000). Psychosocial factors as predictors of functional status at 1 year in patients with left ventricular dysfunction. Research in Nursing & Health, 23, 290–300. Evans, D. (2002). The effectiveness of music as an intervention for hospital patients: A systematic review. Journal of Advanced Nursing, 37, 8–18. Froelicher, E. S., & Christopherson, D. J. (2000). Women’s Initiative for Nonsmoking (WINS) I: Design and methods. Heart & Lung, 29, 429–437. Gaffney, K. F., Barndt-Maglio, B., Myers, S., & Kollar, S. J. (2002). Early clinical assessment for harsh child discipline strategies. MCN: American Journal of Maternal-Child Nursing, 27, 34–40. Greenberg, M. E. (2000). Telephone nursing: Evidence of client and organizational benefits. Nursing Economics, 18, 117–123. Havens, G. A. D. (2000). Differences in the execution/ nonexecution of advance directives by communitydwelling adults. Research in Nursing & Health, 23, 319–333.

243

Lichtig, L. K., Knauf, R. A., & Milholland, D. K. (1999). Some impacts of nursing on acute care hospital outcomes. Journal of Nursing Administration, 29, 25–53. Mahon, N. E., Yarcheski, A., & Yarchesky, T. J. (2002). Psychometric evaluation of the Personal Lifestyle Questionnaire for adolescents. Research in Nursing & Health, 25, 68–75. Martin, K., Froelicher, E. S., & Miller, N. H. (2000). Women’s Initiative for Nonsmoking (WINS) I: The intervention. Heart & Lung, 29, 438–445. Neumark, D. E., Stommel, M., Given, C. W., & Given, B. A. (2001). Research design and subject characteristics predicting nonparticipation in a panel survey of older families with cancer. Nursing Research, 50, 363–368. Patterson, P., Moylan, E., Bannon, S., & Salih, F. (2000). Needs analysis of a cancer education program in South Western Sydney. Cancer Nursing, 23, 186–192. Pesata, V., Pallija, G., & Webb, A. A. (1999). A descriptive study of missed appointments: Families’ perceptions of barriers to care. Journal of Pediatric Health Care, 13, 178–182. Polivka, B. J., Nickel, J. T., Salsberry, P. J., Kuthy, R., Shapiro, N., & Slack, C. (2000). Hospital and emergency department use by young low-income children. Nursing Research, 49, 253–261. Prozialeck, L. L., & Pesole, L. (2000). Performing a program evaluation in a family case management program: Determining outcomes for low birthweight deliveries. Public Health Nursing, 17, 195–201. Riesch, S. K., Tosi, C. B., & Thurston, C. A. (1999). Accessing young adolescents and their families for research. Image: The Journal of Nursing Scholarship, 31, 323–326. Riesch, S. K., Tosi, C. B., Thurston, C. A., Forsythe, D. M., Kuenning, T. S., & Kestly, J. (1993). Effects of communication training on parents and young adolescents. Nursing Research, 42, 10–16. Ritz, L. J., Nissen, M. J., Swenson, K. K., Farrell, J. B., Sperduto, P. W., Sladek, P. W., Lally, R. M., & Schroeder, L. M. (2000). Effects of advanced nursing care on quality of life and cost outcomes of women diagnosed with breast cancer. Oncology Nursing Forum, 27, 923–932. Root, S. D. (2000). Implementing a shared governance model in the perioperative setting. AORN Journal, 72, 95–102. Scheffer, B. K., & Rubenfeld, M. G. (2000). A consensus statement on critical thinking in nursing. Journal of Nursing Education, 39, 352–359.

244

PART 3 Designs for Nursing Research

Schutta, K. M., & Burnett, C. B. (2000). Factors that influence a patient’s decision to participate in a phase I cancer clinical trial. Oncology Nursing Forum, 27, 1435–1438. Simpson, M., Parsons, M., Greenwood, J., & Wade, K. (2001). Raspberry leaf in pregnancy: Its safety and

efficacy in labor. Journal of Midwifery & Women’s Health, 46, 51–59. Wilson, T. (2000). Economic evaluation of a metropolitanwide, school-based hepatitis B vaccination program. Public Health Nursing, 17, 222–227.

11

Qualitative Research Design and Approaches

THE DESIGN OF Q U A L I TAT I V E S T U D I E S As we have seen, quantitative researchers carefully specify a research design before collecting even one piece of data, and rarely depart from that design once the study is underway. In qualitative research, by contrast, the study design typically evolves over the course of the project. Decisions about how best to obtain data, from whom to obtain data, how to schedule data collection, and how long each data collection session should last are made in the field as the study unfolds. Qualitative studies use an emergent design—a design that emerges as researchers make ongoing decisions reflecting what has already been learned. As noted by Lincoln and Guba (1985), an emergent design in qualitative studies is not the result of sloppiness or laziness on the part of researchers, but rather a reflection of their desire to have the inquiry based on the realities and viewpoints of those under study—realities and viewpoints that are not known or understood at the outset. Characteristics of Qualitative Research Design Qualitative inquiry has been guided by different disciplines, and each has developed methods for ad-

dressing questions of particular interest. However, some general characteristics of qualitative research design tend to apply across disciplines. In general, qualitative design: • Often involves a merging together of various data collection strategies; • Is flexible and elastic, capable of adjusting to what is being learned during the course of data collection; • Tends to be holistic, striving for an understanding of the whole; • Requires researchers to become intensely involved, often remaining in the field for lengthy periods of time; • Requires the researcher to become the research instrument; and • Requires ongoing analysis of the data to formulate subsequent strategies and to determine when field work is done. With regard to the first characteristic, qualitative researchers tend to put together a complex array of data, derived from a variety of sources and using a variety of methods. This tendency has sometimes been described as bricolage, and the qualitative researcher has been referred to as a bricoleur, a person who “is adept at performing a large number of diverse tasks, ranging from interviewing to observing, to interpreting personal and

246

PART 3 Designs for Nursing Research

historical documents, to intensive reflection and introspection” (Denzin & Lincoln, 1994, p. 2). Qualitative Design and Planning Although design decisions are not specified in advance, qualitative researchers typically do considerable advance planning that supports their flexibility in developing an emergent design. In the total absence of planning, design choices might actually be constrained. For example, researchers initially might anticipate a 6-month period for data collection, but may need to be prepared (financially and emotionally) to spend even longer periods of time in the field to pursue data collection opportunities that could not have been foreseen. In other words, qualitative researchers plan for broad contingencies that may be expected to pose decision opportunities once the study has begun. Examples of the areas in which advanced planning is especially useful include the following: • Selecting a broad framework or tradition (described in the next section) to guide certain design decisions • Identifying potential study collaborators and reviewers of the research plans • Developing a broad data collection strategy (e.g., will interviews be conducted?), and identifying opportunities for enhancing credibility (e.g., through triangulation) • Selecting the site where the study will take place and identifying the types of settings • Identifying any “gatekeepers” who can provide (or deny) access to important sources of data, and can make arrangements for gaining entrée • Collecting relevant written or photographic materials about the site (e.g., maps, organizational charts, resource directories.) • Determining the maximum amount of time available for the study, given costs and other constraints • Identifying the types of equipment that could aid in the collection and analysis of data in the field (e.g., audio and video recording equipment, laptop computers) • Determining the number and type of assistants needed (if any) to complete the project

• Training any assistants—and self-training • Identifying appropriate informed consent procedures, including contingencies for dealing with ethical issues as they present themselves during data collection • Developing plans for assessing the trustworthiness of the data and the overall inquiry Thus, a qualitative researcher needs to plan for a variety of potential circumstances, but decisions about how he or she will deal with them must be resolved when the social context of time, place, and human interactions are better understood. By both allowing for and anticipating an evolution of strategies, qualitative researchers seek to make their research design responsive to the situation and to the phenomenon under study. One further task that qualitative researchers typically undertake before collecting data is an analysis of their own biases and ideology. Qualitative researchers tend to accept that research is subjective and may be ideologically driven. Decisions about research design and research approaches are not valuefree. Qualitative researchers, then, are more inclined to take on as an early research challenge the identification of their own biases and presuppositions. Such an identification is particularly important in qualitative inquiry because of the intensely personal nature of the data collection and data analysis experience. Example illustrating self-disclosure of possible bias: Rashid (2001) studied women’s views about the use of the Norplant contraceptive implant in rural Bangladesh. She wrote: “My writing on this subject is influenced by my position as a native (born in Bangladesh) and as an outsider (I grew up overseas from 1979 to 1993). The kind of fieldwork I carried out was influenced by my cultural background, Muslim identity, status as an unmarried Bengali woman, and mixed cultural upbringing...” (p. 89). Phases in Qualitative Design Although the exact form of a qualitative study cannot be known and specified in advance, Lincoln

CHAPTER 11 Qualitative Research Design and Approaches

247

and Guba (1985) have noted that a naturalistic inquiry typically progresses through three broad phases while in the field:

research design, we refer the reader to the design elements identified in Table 8-1.

1. Orientation and overview. Quantitative researchers usually believe that they know what they do not know—that is, they know exactly what type of knowledge they expect to obtain by doing a study, and then strive to obtain it. Qualitative researchers, by contrast, enter the study not knowing what is not known—that is, not knowing what it is about the phenomenon that will drive the inquiry forward. Therefore, the first phase of many qualitative studies is to get a handle on what is salient about the phenomenon of interest. 2. Focused exploration. The second phase of the study is a more focused scrutiny and in-depth exploration of those aspects of the phenomenon that are judged to be salient. The questions asked and the types of people invited to participate in the study are shaped by the understandings developed in the first phase. 3. Confirmation and closure. In the final phase, qualitative researchers undertake efforts to establish that their findings are trustworthy, often by going back and discussing their understanding with study participants.

Control Over the Independent Variable Qualitative research is almost always nonexperimental (although, as discussed in the next chapter, a qualitative study sometimes is embedded in an experimental project). Researchers conducting a study within the naturalistic paradigm do not normally conceptualize their studies as having independent and dependent variables, and they rarely control or manipulate any aspect of the people or environment under study. The goal of most qualitative studies is to develop a rich understanding of a phenomenon as it exists in the real world and as it is constructed by individuals in the context of that world.

The three phases are not discrete events. Rather, they overlap to a greater or lesser degree in different projects. For example, even the first few interviews or observations are typically used as a basis for selecting subsequent informants, even though the researcher is still striving to understand the full scope of the phenomenon and to identify its major dimensions. The various phases may take many months or even years to complete. Qualitative Design Features Some of the design features of quantitative studies (see Chapter 8) also apply to qualitative ones. However, qualitative design features are often post hoc characterizations of what happened in the field rather than features specifically planned in advance. To further contrast qualitative and quantitative

Type of Group Comparisons Qualitative researchers typically do not plan in advance to make group comparisons because the intent of most qualitative studies is to thoroughly describe and explain a phenomenon. Nevertheless, patterns emerging in the data sometimes suggest that certain comparisons are relevant and illuminating. Sometimes, of course, comparisons are planned in qualitative studies (e.g., comparisons of two different cultures). Example of qualitative comparisons: Draucker and Stern (2000) conducted a grounded theory study to describe women’s responses to sexual violence by male intimates. They labeled the central process for these women as forging ahead, but discovered that there were variants to this process for three subgroups of women who experienced different types of sexual violence. Number of Data Collection Points Qualitative research, like quantitative research, can be either cross-sectional, with one data collection point, or longitudinal, with multiple data collection points over an extended time period, to observe the evolution of some phenomenon. Sometimes qualitative researchers plan in advance for a longitudinal design, but, in other cases, the decision to study a phenomenon longitudinally may be made in the field after preliminary data have been collected and analyzed.

248

PART 3 Designs for Nursing Research

Examples of the time dimension in qualitative studies: • Cross-sectional: Dewar and Lee (2000) examined how people with a catastrophic illness or injury managed their circumstances. In a single interview, the researchers asked participants (who had endured their condition for between 3 and 25 years) to describe their coping processes over time. • Longitudinal: Reising (2002) conducted a longitudinal study of the early socialization processes of critical care nurses. New critical care nurses were interviewed multiple times and kept a journal over a 5-month period. Occurrence of the Independent and Dependent Variables Qualitative researchers typically would not apply the terms retrospective or prospective to their studies. Nevertheless, in trying to elucidate the full nature of a phenomenon, they may look back retrospectively (with the assistance of study participants) for antecedent events leading to the occurrence of a phenomenon. Qualitative researchers may also study the effects of a phenomenon prospectively. Examples of exploring influences on phenomena in qualitative designs: • Retrospective exploration: Hawley (2000) studied hospitalized patients’accounts of their physical and emotional well-being in relation to their descriptions of comforting strategies their nurses had used. • Prospective exploration: Williams, Schutte, Evers, and Holkup (2000) conducted a prospective study to explore the short- and longer-term effects of getting normal test results from predictive gene testing for neurodegenerative disorders such as Huntington disease. Research Setting Qualitative researchers collect their data in realworld, naturalistic settings. And, whereas a quantitative researcher usually strives to collect data in one type of setting to maintain constancy of condi-

tions (e.g., conducting all interviews in study participants’ homes), qualitative researchers may deliberately strive to study their phenomena in a variety of natural contexts. Example of variation in settings: Long, Kneafsey, Ryan, and Berry (2002) conducted a 2-year qualitative study to examine the nurse’s role within multiprofessional rehabilitation teams in the United Kingdom. Forty-nine clients were recruited. Their pathways through rehabilitation services were observed for 6 months in a variety of settings, including homes, outpatient clinics, hospital wards, and nursing homes. Q U A L I TAT I V E R E S E A R C H TRADITIONS Despite the fact that there are some features common to many qualitative research designs, there is nevertheless a wide variety of overall approaches. Unfortunately, there is no readily agreed-upon classification system or taxonomy for the various approaches. Some authors have categorized qualitative studies in terms of analysis styles, others have classified them according to their broad focus. One useful system is to describe various types of qualitative research according to disciplinary traditions. These traditions vary in their conceptualization of what types of questions are important to ask in understanding the world in which we live. The section that follows provides an overview of several qualitative research traditions (some of which we have previously introduced), and subsequent sections describe in greater detail four traditions that have been especially useful for nurse researchers. Overview of Qualitative Research Traditions The research traditions that have provided a theoretical underpinning for qualitative studies come primarily from the disciplines of anthropology, psychology, and sociology. As shown in Table 11-1, each discipline has tended to focus on one or two broad domains of inquiry.

CHAPTER 11 Qualitative Research Design and Approaches

249

TABLE 11.1 Overview of Qualitative Research Traditions DISCIPLINE

DOMAIN

RESEARCH TRADITION

AREA OF INQUIRY

Anthropology

Culture

Ethnography Ethnoscience (cognitive anthropology)

Holistic view of a culture Mapping of the cognitive world of a culture; a culture’s shared meanings, semantic rules

Psychology/ philosophy

Lived experience

Phenomenology

Experiences of individuals within their lifeworld Interpretations and meanings of individuals’ experiences

Psychology

Behavior and events

Sociology

Social settings

Hermeneutics Ethology Ecologic psychology Grounded theory

Behavior observed over time in natural context Behavior as influenced by the environment

Symbolic interaction (semiotics)

Social structural processes within a social setting Manner by which shared agreement is achieved in social settings Manner by which people make sense of social interactions

Ethnomethodology

Sociolinguistics

Human communication

Discourse analysis

Forms and rules of conversation

History

Past behavior, events, and conditions

Historical analysis

Description and interpretation of historical events

The discipline of anthropology is concerned with human cultures. Ethnography (discussed more fully later in this chapter) is the primary research tradition in anthropology. Ethnographers study cultural patterns and experiences in a holistic fashion. Ethnoscience (sometimes referred to as cognitive anthropology) focuses on the cognitive world of a culture, with particular emphasis on the semantic rules and shared meanings that shape behavior. Phenomenology has its disciplinary roots in both philosophy and psychology. As noted in

Chapter 3, phenomenology focuses on the meaning of lived experiences of humans. A closely related research tradition is hermeneutics, which uses lived experiences as a tool for better understanding the social, cultural, political, or historical context in which those experiences occur. Hermeneutic inquiry almost always focuses on meaning and interpretation—how socially and historically conditioned individuals interpret their world within their given context. The discipline of psychology has several other qualitative research traditions that focus on behavior.

250

PART 3 Designs for Nursing Research

Human ethology, which is sometimes described as the biology of human behavior, studies behavior as it evolves in its natural context. Human ethologists use primarily observational methods in an attempt to discover universal behavioral structures. Ecological psychology focuses more specifically on the influence of the environment on human behavior, and attempts to identify principles that explain the interdependence of humans and their environmental context. Example of an ethologic study: Morse, Penrod, Kassab, and Dellasega (2000) studied comforting behaviors of trauma team practitioners and their effect on successful completion of the uncomfortable procedure of nasogastric tube insertion. Thirty-two attempts at nasogastric insertion from 193 videotaped trauma cases were analyzed in detail. Sociologists study the social world in which we live and have developed several research traditions of importance to qualitative researchers. The grounded theory tradition (described briefly earlier and elaborated upon later in this chapter) seeks to describe and understand key social psychological and structural processes in social settings. Ethnomethodology seeks to discover how people make sense of their everyday activities and interpret their social worlds so as to behave in socially acceptable ways. Within this tradition, researchers attempt to understand a social group’s norms and assumptions that are so deeply ingrained that the members no longer think about the underlying reasons for their behaviors. Example of an ethnomethodologic study: Wakefield (2000) conducted an ethnomethodologic study to examine the practicalities of organizing surgical nurses’ work in the United Kingdom health service. Wakefield made detailed observations of nurses working on a 32-bed general surgical unit. Unlike most other qualitative researchers, ethnomethodologists occasionally engage in ethnomethodologic experiments. During such experiments, researchers disrupt ordinary activity by doing something that violates the group’s norms

and assumptions. They then observe what group members do and how they try to make sense of what is happening. An example is observing what happens when a nurse violates expectations of appropriate behavior at nurses’ change of shift report on a medical intensive care unit. For example, a nurse might be enlisted to start her day shift by deliberately taking out a novel and reading it while the charge nurse from nights is reporting on the status of critically ill patients. The researcher would then observe how the rest of the unit staff deals with this departure from expected behavior. Symbolic interaction is a sociologic and social-psychological tradition with roots in American pragmatism. Like other qualitative frameworks, symbolic interaction has been defined and specified in various ways and is therefore difficult to describe briefly. Basically, symbolic interaction focuses on the manner in which people make sense of social interactions and the interpretations they attach to social symbols, such as language. There are three basic premises underlying this tradition: first, that people act and react on the basis of the meanings that objects and other people in their environment have for them; second, that these meanings are based on social interaction and communication; and third, that these meanings are established through an interpretive process undertaken by each individual. Symbolic interactionists sometimes use semiotics, which refers to the study of signs and their meanings. A sign is any entity or object that carries information (e.g., a diagram, map, or picture). Example of an symbolic interaction study: Rehm (2000) studied Mexican-American parents’ perceptions of family relationships influenced by their children’s chronic physical conditions. The researcher believed that symbolic interaction, which focuses on the interaction between family members and the meaning shared through this social interaction, was a useful framework for studying families’ coping. The domain of inquiry for sociolinguists is human communication. The tradition often referred to as discourse analysis (sometimes called

CHAPTER 11 Qualitative Research Design and Approaches

conversation analysis) seeks to understand the rules, mechanisms, and structure of conversations. Discourse analysts are interested in understanding the action that a given kind of talk “performs.” The data for discourse analysis typically are transcripts from naturally occurring conversations, such as those between nurses and their patients. In discourse analysis, the texts are situated in their social, cultural, political, and historical context. Parker (1992) and Potter and Wetherall (1994) offer suggestions on how to approach a discourse analysis. Example of a discourse analysis: Hallett, Austin, Caress, and Luker (2000) used discourse analysis to study community nurses’ perceptions of patient compliance in wound care. Sixty-two nurses in the United Kingdom were interviewed. The texts of their interview transcripts were analyzed to identify hidden meanings in the nurses’ narratives. Finally, historical research—the systematic collection and critical evaluation of data relating to past occurrences—is also a tradition that relies primarily on qualitative data. Nurses have used historical research methods to examine a wide range of phenomena in both the recent and more distant past. Researchers in each of these traditions have developed methodologic guidelines for the design and conduct of relevant studies. Thus, once a researcher has identified what aspect of the human experience is of greatest interest, there is typically a wealth of advice available about methods likely to be productive and design issues that need to be handled in the field. TIP: Sometimes a research report identifies more than one tradition as having provided the framework for a qualitative inquiry (e.g., a phenomenological study using the grounded theory method). However, such “method slurring” (Baker, Wuest, & Stern, 1992) has been criticized because each research tradition has different intellectual assumptions and methodologic prescriptions.

251

Ethnography Ethnography is a type of qualitative inquiry that involves the description and interpretation of cultural behavior. Ethnographies are a blend of a process and a product, field work, and a written text. Field work is the process by which the ethnographer inevitably comes to understand a culture, and the ethnographic text is how that culture is communicated and portrayed. Because culture is, in itself, not visible or tangible, it must be constructed through ethnographic writing. Culture is inferred from the words, actions, and products of members of a group. Ethnographic research is in some cases concerned with broadly defined cultures (e.g., a Samoan village culture), in what is sometimes referred to as a macroethnography. However, ethnographies sometimes focus on more narrowly defined cultures in a microethnography. Microethnographies are exhaustive, fine-grained studies of either small units in a group or culture (e.g., the culture of homeless shelters), or of specific activities in an organizational unit (e.g., how nurses communicate with children in an emergency department). An underlying assumption of the ethnographer is that every human group eventually evolves a culture that guides the members’ view of the world and the way they structure their experiences. Ethnographers seek to learn from (rather than to study) members of a cultural group—to understand their world view. Ethnographic researchers sometimes refer to “emic” and “etic” perspectives (terms that originate in linguistics, i.e., phonemic versus phonetic). An emic perspective refers to the way the members of the culture envision their world—it is the insiders’ view. The emic is the local language, concepts, or means of expression that are used by the members of the group under study to name and characterize their experiences. The etic perspective, by contrast, is the outsiders’ interpretation of the experiences of that culture; it is the language used by those doing the research to refer to the same phenomena. Ethnographers strive to acquire an emic perspective of a culture under study. Moreover, they strive to reveal what has

252

PART 3 Designs for Nursing Research

been referred to as tacit knowledge, information about the culture that is so deeply embedded in cultural experiences that members do not talk about it or may not even be consciously aware of it. Although it is important to grasp the insider’s perspective, it is also important for the ethnographer to illuminate the connection between the emic and the second-order, integrative and interpretational concepts that advance the aims of knowledge. Ethnographers almost invariably undertake extensive field work to learn about the cultural group in which they are interested. Ethnographic research typically is a labor-intensive endeavor that requires long periods of time in the field—years of field work may be required. In most cases, researchers strive to participate actively in cultural events and activities. The study of a culture requires a certain level of intimacy with members of the cultural group, and such intimacy can be developed only over time and by working directly with those members as active participants. The concept of researcher as instrument is frequently used by anthropologists to describe the significant role ethnographers play in analyzing and interpreting a culture. Three broad types of information are usually sought by ethnographers: cultural behavior (what members of the culture do), cultural artifacts (what members of the culture make and use), and cultural speech (what people say). This implies that ethnographers rely on a wide variety of data sources, including observations, in-depth interviews, records, charts, and other types of physical evidence (e.g., photographs, diaries, letters). The product of ethnographic research usually is a rich and holistic description of the culture under study. Ethnographers also make interpretations of the culture, describing normative behavioral and social patterns. Among health care researchers, ethnography provides access to the health beliefs and health practices of a culture or subculture. Ethnographic inquiry can thus help to facilitate understanding of behaviors affecting health and illness. A rich array of ethnographic methods have been developed and cannot be fully explicated in this general textbook, but more information may be found in Hammersley and Atkinson (1983),

Spradley and McCurdy (1972), Fetterman (1989), and Goetz and LeCompte (1984). Two variants of ethnographic research (ethnonursing research and ethnoscience) are described here, and a third (critical ethnography) is described later in this chapter. Example of an ethnographic study: Lipson (2001) conducted an ethnographic study about the experiences of people with multiple chemical sensitivity. She gathered her data (which included in-depth interviews and observations) in two U.S. and two Canadian settings. Her report includes a particularly valuable discussion of issues relating to the conduct of autoethnography (or insider research), in which ethnographers study their own culture or group. Ethnonursing Research Many nurse researchers have undertaken ethnographic studies. Indeed, Leininger has coined the phrase ethnonursing research, which she defines as “the study and analysis of the local or indigenous people’s viewpoints, beliefs, and practices about nursing care behavior and processes of designated cultures” (1985, p. 38). In conducting an ethnonursing study, the investigator uses a broad theoretical framework to guide the research, such as Leininger’s theory of culture care. Leininger (1991) developed a number of enablers to help guide researchers in conducting ethnonursing research. Enablers are ways to help discover complex phenomena like human care. Some of her enablers include her Stranger—Friend Model, Observation—Participation—Reflection Model, and Acculturation Enabler Guide. Example of an ethnonursing study: Wittig (2001) conducted an ethnonursing study focusing on organ donation beliefs of African-American women living in rural Mississippi. Wittig made numerous visits to the site and conducted in-depth interviews with 10 African-American women. Ethnoscience In cognitive anthropology, culture is defined in purely mentalistic terms. This type of ethnography

CHAPTER 11 Qualitative Research Design and Approaches

concentrates on understanding cultural knowledge through an emphasis on relationships between words. Cognitive anthropologists assume that a group’s cultural knowledge is reflected in its language. One of the purposes of cognitive anthropology is to produce a map of the cognitive world of a culture that addresses its semantic rules. Ethnoscience often relies on quantitative as well as qualitative data. Findings of ethnoscience are often displayed in taxonomic trees. Example of an ethnoscience study: Banister (1999) investigated midlife women’s perceptions of their changing bodies in Western culture using an ethnoscience approach. Eleven women 40 to 53 years of age were interviewed and their transcripts were analyzed using Spradley and McCurdy’s (1972) method. Phenomenology Phenomenology, rooted in a philosophical tradition developed by Husserl and Heidegger, is an approach to discovering the meaning of people’s life experiences. Phenomenological researchers ask: What is the essence of this phenomenon as experienced by these people and what does it mean? Phenomenologists assume there is an essence—an essential invariant structure—that can be understood, in much the same way that ethnographers assume that cultures exist. Phenomenologists investigate subjective phenomena in the belief that critical truths about reality are grounded in people’s lived experiences. There are two “schools” of phenomenology: descriptive phenomenology and interpretive phenomenology (hermeneutics). Descriptive phenomenology was developed first by Husserl (1962), who was primarily interested in the question: What do we know as persons? His philosophy emphasized descriptions of the meaning of human experience. Heidegger, a student of Husserl, moved away from his professor’s philosophy into interpretive phenomenology. To Heidegger (1962), the critical question is: What is Being? He stressed interpreting and understanding—not just describing—

253

human experience. The focus of phenomenological inquiry, then, is the meaning of people’s experience in regard to a phenomenon (descriptive phenomenology), and how those experiences are interpreted (hermeneutics). Phenomenologists believe that lived experience gives meaning to each person’s perception of a particular phenomenon. The goal of phenomenological inquiry is to fully describe lived experience and the perceptions to which it gives rise. Four aspects of lived experience that are of interest to phenomenologists are lived space or spatiality; lived body or corporeality; lived time or temporality; and lived human relation or relationality. Phenomenologists believe that human existence is meaningful and interesting because of people’s consciousness of that existence. The phrase being-in-the-world (or embodiment) is a concept that acknowledges people’s physical ties to their world—they think, see, hear, feel, and are conscious through their bodies’ interaction with the world. In a phenomenological study, the main data source is in-depth conversations, with researchers and informants as full coparticipants. The researcher helps the informant to describe lived experiences without leading the discussion. Through in-depth conversations, the researcher strives to gain entrance into the informants’ world, to have full access to their experiences as lived. Sometimes two separate interviews or conversations may be needed. For some phenomenological researchers, the inquiry includes not only gathering information from informants, but also efforts to experience the phenomenon in the same way, typically through participation, observation, and introspective reflection. Although there are a number of methodologic interpretations of phenomenology, a descriptive phenomenological study often involves the following four steps: bracketing, intuiting, analyzing, and describing. Bracketing refers to the process of identifying and holding in abeyance preconceived beliefs and opinions about the phenomenon under study. Although bracketing can never be achieved totally, researchers bracket out the world and any presuppositions, to the extent possible, so as to confront the data in pure form. Bracketing is an

254

PART 3 Designs for Nursing Research

iterative process that involves preparing, evaluating, and providing systematic, ongoing feedback about the effectiveness of the bracketing. Porter (1993) believes that bracketing can result in more productive use of researchers’ time if they attempt to understand the effects of their experiences rather than expending energy trying to eliminate them. Ahern (1999) provides 10 tips to help qualitative researchers with bracketing through notes in a reflexive journal: 1. Make note of interests that, as a researcher, you may take for granted (i.e., gaining access). 2. Clarify your personal values and identify areas in which you know you are biased. 3. Identify areas of possible role conflict. 4. Recognize gatekeepers’ interest and make note of the degree to which they are favorably or unfavorably disposed toward your research. 5. Identify any feelings you have that may indicate a lack of neutrality. 6. Describe new or surprising findings in collecting and analyzing data. 7. Reflect on and profit from methodologic problems that occur during your research. 8. After data analysis is complete, reflect on how you write up your findings. 9. Reflect on whether the literature review is truly supporting your findings, or whether it is expressing the similar cultural background that you have. 10. Consider whether you can address any bias in your data collection or analysis by interviewing a participant a second time or reanalyzing the transcript in question. Intuiting, the second step in descriptive phenomenology, occurs when researchers remain open to the meanings attributed to the phenomenon by those who have experienced it. Phenomenological researchers then proceed to the analysis phase (i.e., extracting significant statements, categorizing, and making sense of the essential meanings of the phenomenon). (Chapter 23 provides further information regarding the analysis of data collected in phenomenological studies.) Finally, the descriptive

phase occurs when the researcher comes to understand and define the phenomenon. Note that an important distinction between descriptive and interpretive phenomenology is that in an interpretive phenomenological study, bracketing does not occur. For Heidegger, it was not possible to bracket one’s being-in-the-world. Hermeneutics presupposes prior understanding on the researcher’s part. The phenomenological approach is especially useful when a phenomenon of interest has been poorly conceptualized. The topics appropriate to phenomenology are ones that are fundamental to the life experiences of humans; for health researchers, these include such topics as the meaning of stress, the experience of bereavement, and quality of life with a chronic illness. A wealth of resources are available on phenomenological methods. Interested readers may wish to consult Spiegelberg (1975), Giorgi (1985), Colaizzi (1978), or Van Manen (1990). Example of a phenomenological study: Rungreangkulkij and Chesla (2001) conducted a phenomenological study of Thai mothers’ experiences caring for a child with schizophrenia. In-depth interviews were conducted with 12 Thai mothers who had adult schizophrenic children. Findings centered on the mothers’ attempts to smooth their hearts with lots of water. In Thai culture, the metaphor of water and fire is used to help people calm down when experiencing negative emotions such as anger or frustration. Grounded Theory Grounded theory has become an important research method for the study of nursing phenomena, and has contributed to the development of many middlerange theories of phenomena relevant to nurses. Grounded theory began more as a systematic method of qualitative research than as a philosophy. Grounded theory was developed in the 1960s by two sociologists, Glaser and Strauss (1967), whose own theoretical links were in symbolic interactionism. One of their earliest studies (Glaser & Strauss, 1965) was a grounded theory study on dying in hospitals, in which the “prime controllable” variable

CHAPTER 11 Qualitative Research Design and Approaches

was characterized as awareness context (i.e., who knows what about the patient’s dying). Grounded theory is an approach to the study of social processes and social structures. The focus of most grounded theory studies is the development and evolution of a social experience— the social and psychological stages and phases that characterize a particular event or episode. As noted in Chapter 6, the primary purpose of the grounded theory approach is to generate comprehensive explanations of phenomena that are grounded in reality. In-depth interviews and observation are the most common data source in grounded theory studies, but existing documents and other data sources may also be used. Grounded theory research can involve the analysis of quantitative as well as qualitative data (Glaser & Strauss, 1967), but this rarely happens in practice. Grounded theory methods constitute an entire approach to the conduct of field research. For example, a study that truly follows Glaser and Strauss’s precepts does not begin with a highly focused research problem; the problem emerges from the data. One of the fundamental features of the grounded theory approach is that data collection, data analysis, and sampling of study participants occur simultaneously. Grounded theory methods are inherently nonlinear in nature, and therefore difficult to characterize. The process is recursive: researchers must systematically collect data, categorize them, describe the central phenomenon, and recycle earlier steps. A procedure referred to as constant comparison is used to develop and refine theoretically relevant categories. Categories elicited from the data are constantly compared with data obtained earlier in the data collection process so that commonalities and variations can be determined. As data collection proceeds, the inquiry becomes increasingly focused on emerging theoretical concerns. Data analysis in a grounded theory framework is described in greater depth in Chapter 23. TIP: Beginning qualitative researchers should be aware that if they have constraints

255

in the amount of time they can devote to conducting a study, a grounded theory study is a much lengthier process than a phenomenological study. Example of a grounded theory study: Knobf (2002) sought to develop a substantive theory to explain women’s responses to chemotherapy-induced premature menopause within the context of breast cancer. Vulnerability was identified as the women’s basic social problem, and carrying on is the basic social process explaining how women respond to vulnerability. Formal Grounded Theory Glaser and Strauss (1967) distinguished two types of grounded theory: substantive and formal. Substantive theory is grounded in data on a specific substantive area, such as postpartum depression. It can serve as a springboard for formal grounded theory, which involves developing a higher, more abstract level of theory from a compilation of substantive grounded theory studies regarding a particular phenomenon. Glaser and Strauss’ (1971) theory of status passage is an example of a formal grounded theory. Kearney (1998) used an interesting analogy to differentiate substantive theories (custom-tailored clothing) and formal theory (ready-to-wear clothing). Formal grounded theories were likened to clothing sold in department stores that can fit a wider variety of users. Formal grounded theory is not personally tailored like substantive theory, but rather provides a conceptualization that applies to a broader population experiencing a common phenomenon. Formal grounded theories are not situation specific. The best data for constructing formal grounded theories are substantive grounded theories. Example of a formal grounded theory: Kearney (2001a) developed a formal grounded theory of women’s response to violent relationships, based on 13 studies of women’s experiences with domestic violence. The formal grounded theory described women’s basic process of enduring love.

256

PART 3 Designs for Nursing Research

Alternate Views of Grounded Theory In 1990, Strauss and Corbin published what was to become a controversial book, Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Strauss and Corbin stated that the purpose of the book was to provide beginning grounded theory students with a more concrete description of the procedures involved in building theory at the substantive level. Glaser, however, disagreed with some of the procedures advocated by Strauss (his original coauthor) and Corbin (a nurse researcher). Glaser published a rebuttal in 1992, Emergence Versus Forcing: Basics of Grounded Theory Analysis. Glaser believed that Strauss and Corbin developed a method that is not grounded theory but rather what he calls “full conceptual description.” According to Glaser, the purpose of grounded theory is to generate concepts and theories about their relationships that explain, account for, and interpret variation in behavior in the substantive area under study. Conceptual description, in contrast, is aimed at describing the full range of behavior of what is occurring in the substantive area, “irrespective of relevance and accounting for variation in behavior” (Glaser, 1992, p. 19). Nurse researchers have conducted grounded theory studies using both the original Glaser and Strauss and the Strauss and Corbin approaches. Example of grounded theory alternatives: Kendall (1999) provided an excellent comparison of the two approaches to grounded theory from her own research on families with a child with attention deficit—hyperactivity disorder. She described study results two ways—first, results obtained using Strauss and Corbin’s approach, and second, the findings that emerged using Glaser and Strauss’ original grounded theory approach. Kendall felt that Strauss and Corbin’s coding procedure was a distraction that hindered her ability to reach the higher level of abstract thinking needed in grounded theory analysis. Historical Research Historical research is the systematic collection, critical evaluation, and interpretation of historical evidence (i.e., data relating to past occurrences). In gen-

eral, historical research is undertaken to answer questions about causes, effects, or trends relating to past events that may shed light on present behaviors or practices. An understanding of contemporary nursing theories, practices, or issues can often be enhanced by an investigation of a specified segment of the past. Historical data are usually qualitative, but quantitative data are sometimes used (e.g., historical census data). Historical research can take many forms. For example, many nurse researchers have undertaken biographical histories that study the experiences or contributions of individuals, such as nursing leaders. Currently, some historians are focusing on the history and experience of the ordinary person, often studying such issues as gender, race, and class. Other historical researchers undertake social histories that focus on a particular period in attempts to understand prevailing values and beliefs that may have helped to shape subsequent developments. Still others undertake what might be called intellectual histories, where historical ideas or ways of thinking are scrutinized. Historical research should not be confused with a review of the literature about historical events. Like other types of research, historical inquiry has as its goal the discovery of new knowledge, not the summary of existing knowledge. One important difference between historical research and a literature review is that historical researchers are often guided by specific hypotheses or questions, or by a theoretical orientation or ideology (e.g., feminism). Hypotheses in historical research represent attempts to explain and interpret the conditions, events, or phenomena under investigation. Such hypotheses are not tested statistically; rather, they are broadly stated conjectures about relationships among historical events, trends, and phenomena. Example of a hypothesis in historical research: Lusk (2000) analyzed images of nurses in advertisements from 1930 to 1950. She hypothesized that nurses’ relative status in advertisements would be higher in 1940 (when women were encouraged to enter nursing as a patriotic duty) than in 1930 or 1950. Her hypothesis was supported.

CHAPTER 11 Qualitative Research Design and Approaches

After research questions or hypotheses are developed, researchers must determine what types of data are available. Historical researchers typically need to devote considerable effort to identifying and evaluating data sources on events and situations that occurred in the past. Collecting Historical Data Data for historical research are usually in the form of written records: diaries, letters, notes, newspapers, minutes of meetings, medical or legal documents, and so forth. However, nonwritten materials may also be of interest. For example, physical remains and objects are potential sources of information. Visual materials, such as photographs and films, are forms of data, as are audio materials, such as records and tapes. In some cases, it is possible to conduct interviews with people who participated in historical events (e.g., nurses who served in Vietnam). Many historical materials may be difficult to obtain and, in many cases, have been discarded. Historically significant materials are not always conveniently indexed by subject, author, or title. The identification of appropriate historical materials usually requires a considerable amount of time, effort, and detective work. Fortunately, there are several archives of historical nursing documents, such as the collections at several universities (e.g., Archives of Nursing Leadership, University of Connecticut; The Nursing Archives, Mugar Memorial Library, Boston University; Teachers College Milbank Memorial Library, Columbia University; Sterling Memorial Library, Yale University; South West Center for Nursing History, University of Texas, Austin; Center for the Study of the History of Nursing, University of Pennsylvania), as well as collections at National Library of Medicine, the American Journal of Nursing Company, the National League for Nursing, and the New York Public Library. Useful sources for identifying archives in the United States include the National Inventory of Documentary Sources in the United States and the Directory of Archives and Manuscript Repositories in the United States. Finally, the American

257

Association for the History of Nursing, which publishes the Nursing History Review, is a potential source for locating public and private manuscript collections. TIP: Archives are different from libraries. Archives contain unpublished materials that are accessed through finding aids rather than card catalogs. A finding aid is a resource that tells researchers what is in the archive. Archival materials do not circulate; researchers are almost always required to use the material on site (although sometimes microfiches are available). Typically, because of the fragile nature of the material, it cannot be photocopied, so researchers must take detailed notes (laptop computers are invaluable). Sometimes gloves are required when touching original materials. Access to archives may be limited to researchers who present a description of a proposed project to archivists. Historical materials usually are classified as either primary or secondary sources. A primary source is first-hand information, such as original documents, relics, or artifacts. Examples are the diaries and writings of Sophronia Bucklin (1869), minutes of American Nurses Association meetings, hospital records, and so forth. Written primary sources are authored by people directly involved in a described event. Primary sources represent the most direct link with historical events or situations: Only the narrator (in the case of written materials) intrudes between original events and the historical researcher. Secondary sources are second- or third-hand accounts of historical events or experiences. For example, textbooks, other reference books, and newspaper articles are secondary sources. Secondary sources, in other words, are discussions of events written by individuals who did not participate in them, but are often summarizing or interpreting primary source materials. Secondary sources may be historical (e.g., newspaper accounts contemporaneous with the events under study), or more modern interpretations of past events. Primary sources should be used whenever possible in historical research. The further removed from the historical

258

PART 3 Designs for Nursing Research

event the information is, the less reliable, objective, and comprehensive the data are likely to be. However, secondary sources can be useful in identifying primary sources. It is particularly important in reading secondary source material to pay careful attention to footnotes, which often provide important clues about primary sources. Evaluating Historical Data Historical evidence is subjected to two types of evaluation, which historians refer to as external and internal criticism. External criticism is concerned with the data’s authenticity. For example, a nursing historian might have a diary presumed to be written by Dorothea Dix. External criticism would involve asking such questions as: Is this the handwriting of Ms. Dix? Is the diary’s paper of the right age? Are the writing style and ideas expressed consistent with her other writings? There are various scientific techniques available to determine the age of materials, such as x-ray and radioactive procedures. Other flaws, however, may be less easy to detect. For example, there is the possibility that material of interest may have been written by a ghost writer, that is, by someone other than the person of interest. There are also potential problems of mechanical errors associated with transcriptions, translations, or typed versions of historical materials. Internal criticism of historical data refers to an evaluation of the worth of the evidence. The focus of internal criticism is not on the physical aspects of the materials but on their content. The key issue is the accuracy or truth of the data. For example, researchers must question whether a writer’s representations of historical events are unbiased. It may also be appropriate to ask if the author of a document was in a position to make a valid report of an event or occurrence, or whether the writer was competent as a recorder of fact. Evidence bearing on the accuracy of historical data often includes comparisons with other people’s accounts of the same event to determine the degree of agreement, knowledge of the time at which the document was produced (reports of events or situations tend to be more accurate if

they are written immediately after the event), and knowledge of the writers’ point of view or biases and their competence to record events authoritatively and accurately. TIP: Tuchman (1994) offers this useful advice: “Ask questions of all data, primary and secondary sources. Do not assume anything about the data is ‘natural,’ inevitable, or even true. To be sure, a datum has a physical presence: One may touch the page ... one has located. But that physical truth may be radically different from the interpretive truth...” (p. 321). Analyzing and Interpreting Historical Data In historical research, data analysis and data collection are usually ongoing, concurrent activities. The analysis of historical data is broadly similar to other approaches to qualitative analysis (see Chapter 23), in that researchers search for themes. In historical research, however, the thematic analysis is often guided by underlying hypotheses or theoretical frameworks. Within the selected framework, researchers concentrate on particular issues present in the data. Historical research is usually interpretive. Historical researchers try to describe what happened, and also how and why it happened. Relationships between events and ideas, between people and organizations, are explored and interpreted within both their historical context and within the context of new viewpoints about what is historically significant. There are many resources available for those interested in historical nursing research, including Lewenson (2003), Fitzpatrick (2001), and Lusk (1997). Example of a historical study: Dunphy (2001a, b) examined the stories of nurses and patients about the iron lung (the socalled “steel cocoon”) between 1929 and 1955. She used a variety of sources such as photographs, drawings, procedural pamphlets from the March of Dimes and American Red Cross, first-person accounts, and a website called “Virtual Museum of the Iron Lung.”

CHAPTER 11 Qualitative Research Design and Approaches

TIP: Avoid jumping to conclusions about the qualitative research tradition of a study based on the report’s title. For example, the study “Hardships and personal strategies of Vietnam War nurses” (Scannell-Desch, 2000) is not a historical study, but rather a phenomenological study. As another example, despite the title’s reference to a cultural group, “Hearing and caring for Haitian adolescents” (Colin, 2001) is a phenomenological study, not an ethnography. OTHER TYPES OF Q U A L I TAT I V E RESEARCH Qualitative studies typically can be characterized and described in terms of the disciplinary research traditions discussed in the previous section. However, several other important types of qualitative research also deserve mention. This section discusses qualitative research that is not associated with any particular discipline. Case Studies Case studies are in-depth investigations of a single entity or a small number of entities. The entity may be an individual, family, group, institution, community, or other social unit. In a case study, researchers obtain a wealth of descriptive information and may examine relationships among different phenomena, or may examine trends over time. Case study researchers attempt to analyze and understand issues that are important to the history, development, or circumstances of the entity under study. One way to think of a case study is to consider what is center stage. In most studies, whether qualitative or quantitative, a certain phenomenon or variable (or set of variables) is the core of the inquiry. In a case study, the case itself is central. As befits an intensive analysis, the focus of case studies is typically on determining the dynamics of why an individual thinks, behaves, or develops in a particular manner rather than on what his or her status, progress, or actions are. It is not unusual for probing research of this type to require detailed study

259

over a considerable period. Data are often collected that relate not only to the person’s present state but also to past experiences and situational factors relevant to the problem being examined. The data for in-depth case studies are usually qualitative, but may also be quantitative. A distinction is sometimes drawn between an intrinsic and instrumental case study. In an intrinsic case study, researchers do not have to select the case. For instance, a process evaluation is often a case study of the implementation of a particular program; the “case” is a given. In an instrumental case study, researchers begin with a research question or perplexity, and seek out a case that offers illumination. The aim of such a case study is to use the case to understand something else, some phenomenon of interest. In such a situation, a case is usually selected not because it is typical, but rather because it can maximize what can be learned about the phenomenon. Although the foremost concern of a case study is to understand the particular case, case studies are sometimes a useful way to explore phenomena that have not been rigorously researched. The information obtained in case studies can be used to develop hypotheses to be tested more rigorously in subsequent research. The intensive probing that characterizes case studies often leads to insights concerning previously unsuspected relationships. Furthermore, in-depth case studies may serve the important role of clarifying concepts or of elucidating ways to capture them. The greatest strength of case studies is the depth that is possible when a limited number of individuals, institutions, or groups is being investigated. Case studies provide researchers with opportunities of having an intimate knowledge of a person’s condition, thoughts, feelings, actions (past and present), intentions, and environment. On the other hand, this same strength is a potential weakness because researchers’ familiarity with the person or group may make objectivity more difficult— especially if the data are collected by observational techniques for which the researchers are the main (or only) observers. Perhaps the biggest criticism of case studies concerns generalizability: If

260

PART 3 Designs for Nursing Research

researchers discover important relationships, it is difficult to know whether the same relationships would occur with others. However, case studies can often play a critical role in challenging generalizations based on other types of research. It is important to recognize that case study research is not simply anecdotal descriptions of a particular incident or patient. Case study research is a disciplined process and typically requires an extended period of systematic data collection. Two excellent resources for further reading on case study methods are the books by Yin (1994) and Stake (1995). Example of a case study: Porter, Ganong, and Armer (2000) presented a case study of an older rural African-American woman’s support network and preferences for care providers. Based on their analysis, the researchers discussed implications for appraising the appropriateness of rural elders’ in-home services. Narrative Analyses Narrative analysis focuses on story in studies in which the purpose is to determine how individuals make sense of events in their lives (Riessman, 1993). What distinguishes narrative analysis from other types of qualitative research designs, such as ethology, is its focus on the broad contours of a narrative. Stories are not fractured and dissected. The broad underlying premise of narrative research in the social sciences (as opposed to literary analysis) is the belief that people most effectively make sense of their world—and communicate these meanings—by constructing, reconstructing, and narrating stories. Individuals construct stories when they wish to understand specific events and situations that require linking an inner world of desire and motive to an external world of observable actions. Muller (1999) has delineated five primary dimensions of the narrative approach: 1. People organize significant events in terms of stories and, through the telling of these stories, they make meaning of these experiences in their lives.

2. Time and plot are structural properties of narratives. Events in a story follow a sequence. 3. Narratives have a cultural contextual; they do not occur by themselves. 4. Narratives are relational. Stories are told to other people. 5. Narratives have power to shape human behavior. Narratives can be used to produce a moral story of how people are supposed to behave. There is no standard approach to narrative analysis. Mishler (1995) has proposed a typology of models of narrative analysis. This threecategory typology centers on the type of problem that is the central concern of the narrative analysis. Models in the first category focus on the temporal order of events in the discourse. The second set of models is concerned with the structure and coherence of how narratives are organized. Models in the third set of the typology focus on the cultural, social, and psychological contexts and functions of narratives. Example of a narrative analysis: Using narrative analysis, Ashida (2000) investigated women’s experiences of smoking relapse after giving birth to a child. Twenty-seven women in Canada told their stories of smoking relapse. Analysis of the transcribed interviews involved reading the narratives closely, preparing brief summaries of each narrative, and identifying the central story lines. Qualitative Outcome Analysis Qualitative researchers have made considerable contributions to understanding health phenomena and people’s experiences of health, illness, injury, and caretaking. However, qualitative researchers have not focused much attention on the development, implementation, and evaluation of interventions derived from qualitative research. Morse, Penrod, and Hupcey (2000) believe that qualitative researchers have lacked procedural guidelines for moving from theory development to the identification and evaluation of specific clinical interventions. They developed a procedure that they call

CHAPTER 11 Qualitative Research Design and Approaches

qualitative outcome analysis (QOA) to address the theory—practice gap in qualitative research. Qualitative researchers typically focus on a comprehensive understanding of a phenomenon in a specific context. Nursing strategies that served as interventions during the experience may not be documented or incorporated into the researchers’ conceptual scheme because the interventions were not the project’s focus. Additional data relevant to interventions can, however, be collected in subsequent QOA projects. QOA builds on an alreadycompleted qualitative study in which a clinical problem has been examined. Table 11-2 presents a comparison of the purpose, focus, and outcome of the original qualitative study and the QOA project. Morse and her colleagues outlined a series of steps that researchers can follow in planning and undertaking a QOA project. Two preliminary steps are critical, namely, assessing clinical settings for possible implementation of an intervention, and building an appropriate research team. After completing these steps and receiving administrative and institutional review board approval, the following procedures are undertaken: Step 1: Outline the program of intervention strategies. The original theory is used to conceptualize the dynamics of the clinical problem and to identify appropriate interventions. Step 2: Identify types of data to be collected. All types of data that could assist in interpreting and

261

evaluating the intervention should be considered, but observation is likely to be especially important. Step 3: Devise data collecting protocols. Data recording forms for recording strategies and insights of those implementing them are developed. Step 4: Analyze data. Data are analyzed qualitatively according to both process and outcome. Step 5: Disseminate findings. QOA project results must be published if they are to be significant for nursing practice. Example of a QOA project: In their initial qualitative project, Morse and Doberneck (1995) discovered seven stages in the process of hoping in four groups of patients: breast cancer, heart transplantation, spinal cord injury, and breastfeeding after returning back to work. Next, a Hope Assessment Guide was developed to assist nurses in identifying the stage of hope that a patient was in (Morse, Hutchinson, & Penrod, 1998; Penrod & Morse, 1997). The QOA project was designed to evaluate the feasibility of using the Hope Assessment Guide in clinical settings. Secondary Analysis In the preceding chapter, we noted that original quantitative studies sometimes involve an analysis of previously collected data, in what is referred to as a secondary analysis. Qualitative researchers,

TABLE 11.2 Comparison of an Initial Clinical Qualitative Study and a Qualitative Outcome Analysis (QOA) Project TYPE OF PROJECT

PURPOSE

FOCUS

OUTCOME

Initial study

To understand a patient’s experience

A clinical phenomenon or problem

A theory explaining patient experiences

QOA project

To identify strategies and evaluate their implementation

Intervention strategies

Efficacy of intervention strategies

Adapted from Morse, J. M., Penrod, J., & Hupcey, J. E. (2000). Qualitative outcome analysis: Evaluating nursing interventions for complex clinical phenomena. Image: The Journal of Nursing Scholarship, 32, 125–130.

262

PART 3 Designs for Nursing Research

like quantitative researchers, typically collect far more data than are analyzed originally. Secondary analyses of qualitative data provide opportunities to exploit rich data sets. There are, however, some impediments to secondary analysis with qualitative data. The first is difficulty in identifying a suitable database. The availability of large quantitative data sets (especially from government-sponsored surveys) is widely publicized. There are a few repositories of qualitative data sets (e.g., the Murray Research Center at Radcliffe College, which has especially rich resources in the area of mental health). Identifying a suitable dataset usually means doing a lot of investigation. Probably the most typical method is to approach a researcher or team of researchers whose qualitative data appear to be of interest, based on reports they have written. (Not all qualitative researchers, however, are willing to share their data, because they themselves may intend to do further analyses.) Another issue is that data from quantitative studies are easy to maintain in computer files, and there are standard protocols for documenting what is in the data set. Qualitative data are more voluminous, and there are few established conventions for maintaining, documenting, and encoding the data in computer files. This situation is beginning to change, however, as researchers have come to recognize the value of facilitating use of rich qualitative data sets for varied purposes by multiple users (Manderson, Kelaher, & Woelz-Stirling, 2001). Thorne (1994) has identified five types of qualitative secondary analysis: 1. Analytic expansion. Original researchers use their own data to answer new questions as the theory base increases or to ask questions at a higher level of analysis. 2. Retrospective interpretation. The original database is used to examine new questions that were not thoroughly assessed in the original study 3. Armchair induction. This type is used for theory development where inductive approaches to the existing data sets are used. 4. Amplified sampling. Broader theories are developed by comparing different databases.

5. Cross-validation. Existing data sets are used in an effort to confirm new results and identify patterns beyond the ability of the original samples. Thorne warned of potential hazards in secondary analysis of qualitative studies such as the possibility of exaggerating the effects of the original researcher’s biases, and ethical questions regarding the use of the original data sets. Another challenge concerns evaluating the quality of the data set from the primary study. Hinds, Vogel, and Clarke-Steffen (1997) developed an assessment tool for just such a purpose. This assessment tool addresses such criteria as determining the fit of secondary research questions, completeness of the data set, and training of the primary team. Example of a secondary analysis of qualitative data: Butcher, Holkup, and Buckwalter’s (2001) study involved secondary analysis of 103 transcribed interviews describing the experience of caring for a family member with Alzheimer’s disease. These interviews had been obtained from a larger, 4-year longitudinal study that tested a community-based psychoeducational nursing intervention. Butcher and colleagues conducted a secondary analysis of these interviews, conducted before the nursing intervention took place, using a phenomenological method. Metasynthesis As discussed in Chapter 10, the need to integrate knowledge across studies is growing—partly in response to the evidence-based movement—and this is true for both qualitative and quantitative research. Efforts are underway to develop techniques for qualitative metasynthesis (the qualitative analogue of meta-analysis). An interpretive metasynthesis is more than just a narrative integration and summary of qualitative findings (i.e., a traditional literature review). Metasynthesis has been defined as “the theories, grand narratives, generalizations, or interpretive translations produced from the integration or comparison of findings from qualitative studies” (Sandelowski, Docherty, and Emden, 1997, p. 366).

CHAPTER 11 Qualitative Research Design and Approaches

Metasynthesis is described in greater detail in Chapter 27. It might be noted that the terminology for qualitative synthesis approaches can be confusing. Researchers use different terms to label qualitative synthesis, such as qualitative meta-analysis, metaethnography, metainterpretation, or aggregating qualitative findings. Kearney’s (2001b) “meta family” helps to clear up some of the confusion by placing current approaches to qualitative synthesis on a continuum from most theorizing to most interpretive. Starting with the theorizing end of the continuum and going to the interpretive end, Kearney orders these approaches as follows: grounded theory, metainterpretation, aggregated analysis, metastudy, metasynthesis, and metaethnography. Example of a metasynthesis: Beck (2001) conducted a metasynthesis of 14 qualitative studies on the meaning of caring in nursing education. The metasynthesis revealed five themes that permeated caring in nursing education, whether it was faculty caring for each other or their students, or nursing students caring for each other or their clients. These themes centered on reciprocal connecting, which consisted of presencing, sharing, supporting, competence, and the uplifting effects of caring. Descriptive Qualitative Studies Most qualitative researchers acknowledge a link to one of the research traditions or types of studies discussed in this chapter. Some qualitative studies, however, claim no particular disciplinary or methodologic roots. The researchers may simply indicate that they have conducted a qualitative study or a naturalistic inquiry, or they may say that they have done a content analysis of their qualitative data (i.e., an analysis of themes and patterns that emerge). Thus, some qualitative studies do not have a formal name or do not fit into the typology we have presented in this chapter. We refer to these as descriptive qualitative studies. Sandelowski (2000) notes that in doing such descriptive qualitative studies, researchers tend not to penetrate their data in any interpretive depth.

263

These studies present comprehensive summaries of a phenomenon or of events in everyday language. Qualitative descriptive designs tend to be eclectic and are based on the general premises of naturalistic inquiry. Sandelowski stresses that researchers should not be ashamed to “just” use qualitative description as their research design. It is the method of choice if what they want is straight description of an event or phenomenon. Example of a descriptive qualitative study: Stubblefield and Murray (2001) interviewed 15 parents whose children had undergone lung transplantation to study the effect of the procedure on the parents’ interpersonal relationships. The authors wrote that they “conducted a content analysis to formulate a narrative description of the parents’ relationships with others” (p. 58). RESEARCH WITH IDEOLOGICAL PERSPECTIVES Some researchers, who have relied predominantly on qualitative data, conduct inquiries within an ideological framework, typically to draw attention to certain social problems or the needs of certain groups. These approaches represent important investigative avenues and are briefly described in this section. Critical Theory Critical theory originated with a group of Marxistoriented German scholars in the 1920s, collectively referred to as the Frankfurt School. Variants of critical theory abound in the social sciences. Essentially, a critical researcher is concerned with a critique of society and with envisioning new possibilities. Critical social science is typically actionoriented. Its broad aim is to integrate theory and practice such that people become aware of contradictions and disparities in their beliefs and social practices, and become inspired to change them. Critical researchers reject the idea of an objective and disinterested inquirer and are oriented toward a

264

PART 3 Designs for Nursing Research

transformation process. An important feature of critical theory is that it calls for inquiries that foster enlightened self-knowledge and sociopolitical action. Moreover, critical theory involves a selfreflexive aspect. To prevent a critical theory of society from becoming yet another self-serving ideology, critical theorists must account for their own transformative effects. The design of research in critical theory often begins with a thorough analysis of certain aspects of the problem. For example, critical researchers might analyze and critique taken-for-granted assumptions that underlie the problem, the language used to depict the situation, and the biases of prior researchers investigating the problem. Critical researchers often triangulate multiple methodologies, and emphasize multiple perspectives (e.g., alternative racial or social class perspectives) on problems. Critical researchers typically interact with study participants in ways that emphasize participants’ expertise. Critical theory has been applied in a number of disciplines, and has played an especially important role in ethnography. Critical ethnography focuses

on raising consciousness and aiding emancipatory goals in the hope of effecting social change. Critical ethnographers address the historical, social, political and economic dimensions of cultures and their value-laden agendas. An assumption in critical ethnographic research is that actions and thoughts are mediated by power relationships (Hammersley, 1992). Critical ethnographers attempt to increase the political dimensions of cultural research and undermine oppressive systems (Giroux, 1992). Some of the features that distinguish more traditional qualitative research and critical research are summarized in Table 11-3. Morrow and Brown (1994) and Carspecken and Apple (1992) provide useful accounts of critical theory methodology. Example of a critical ethnography: Herdman (2000) described a critical ethnography that focused on institutional discrimination in a hospital regarding health care to people with HIV/AIDS. The study sought to identify structures in the hospital that contributed to discrimination, so that these structures could be transformed. The

TABLE 11.3 Comparison of Traditional Qualitative Research and Critical Research TRADITIONAL QUALITATIVE RESEARCH

CRITICAL RESEARCH

Research aims

Understanding; reconstruction of multiple constructions

Critique; transformation; consciousnessraising; advocacy

View of knowledge

Transactional/subjective; knowledge is created in interaction between investigator and participants

Transactional/subjective; value-mediated and value-dependent; importance of historical insights

Methods

Dialectic: truth is arrived at logically through conversations

Dialectic and didactic: dialogue designed to transform naivety and misinformation

Evaluative criteria for inquiry quality

Authenticity; trustworthiness

Historical situatedness of the inquiry; erosion of ignorance; stimulus for change

Researcher’s role

Facilitator of multivoice reconstruction

Transformative agent; advocate; activist

ISSUE

CHAPTER 11 Qualitative Research Design and Approaches

publication of the research report gave rise to heated debates, including objections by the hospital’s institutional ethics committee and a defensive reaction by some health care professionals. Feminist Research Feminist research approaches are similar to critical theory research, but the focus is sharply on gender domination and discrimination within patriarchal societies. Similar to critical researchers, feminist researchers seek to establish collaborative and nonexploitative relationships with their informants, to place themselves within the study to avoid objectification, and to conduct research that is transformative. Gender is the organizing principle in feminist research, and investigators seek to understand how gender and a gendered social order have shaped women’s lives and their consciousness. The aim is to ameliorate the “invisibility and distortion of female experience in ways relevant to ending women’s unequal social position” (Lather, 1991, p. 71). Although feminist researchers generally agree that it is important to focus on women’s diverse situations and the institutions and relationships that frame those situations, there are many variants of feminist inquiry. Three broad models (within each of which there is diversity) have been identified: (1) feminist empiricism, whose adherents usually work within fairly standard norms of qualitative inquiry but who seek to portray more accurate pictures of the social realities of women’s lives; (2) feminist standpoint research, which holds that inquiry ought to begin in and be tested against the lived everyday sociopolitical experiences of women, and that women’s views are particular and privileged; and (3) feminist postmodernism, which stresses that “truth” is a destructive illusion, and views the world as endless stories, texts, and narratives. In nursing and health care, feminist empiricism and feminist standpoint research have been most prevalent. The scope of feminist research ranges from studies of the particular and subjective views of individual women, to studies of social movements, structures, and broad policies that affect (and often

265

exclude) women. Olesen (1994), a sociologist who studied nurses’ career patterns and definitions of success, has noted that some of the best feminist research on women’s subjective experiences has been done in the area of women’s health. Feminist research methods typically include indepth, interactive, and collaborative individual interviews or group interviews that offer the possibility of reciprocally educational encounters. Feminists usually seek to negotiate the meanings of the results with those participating in the study, and to be self-reflexive about what they themselves are experiencing and learning. Feminist research, like other research that has an ideological perspective, has raised the bar for the conduct of ethical research. With the emphasis on trust, empathy, and nonexploitative relationships, proponents of these newer modes of inquiry view any type of deception or manipulation as abhorrent. As Punch (1994) has noted in speaking about ethics and feminist research, “you do not rip off your sisters” (p. 89). Those interested in feminist methodologies may wish to consult such writers as Lather (1991) or Reinharz (1992). Example of feminist research: Gustafson (2000) collaborated with 10 other nurses in a study of nurses’ job displacement in Canadian hospitals during the 1990s. She adopted a method that was “intended to be political in standpoint, gendered in focus, reflexive in process, and transformative in outcome” (p. 717). Gustafson described in some detail the benefits and limitations of using a feminist, collaborative research method, and how her “best laid plans” regarding the research were not always easy to sustain. Participatory Action Research A type of research known as participatory action research is closely allied to both critical research and feminist research. Participatory action research (PAR), one of several types of action research that originated in the 1940s with social psychologist Kurt Lewin, is based on a recognition that the

266

PART 3 Designs for Nursing Research

production of knowledge can be political and can be used to exert power. Researchers in this approach typically work with groups or communities that are vulnerable to the control or oppression of a dominant group or culture. Participatory action research is, as the name implies, participatory. There is collaboration between researchers and study participants in the definition of the problem, the selection of an approach and research methods, the analysis of the data, and the use to which findings are put. The aim of PAR is to produce not only knowledge, but action and consciousness-raising as well. Researchers specifically seek to empower people through the process of constructing and using knowledge. The PAR tradition has as its starting point a concern for the powerlessness of the group under study. Thus, a key objective is to produce an impetus that is directly used to make improvements through education and sociopolitical action. In PAR, the research methods take second place to emergent processes of collaboration and dialogue that can motivate, increase self-esteem, and generate community solidarity. Thus, the “data-gathering” strategies used are not only the traditional methods of interview and observation (including both qualitative and quantitative approaches), but may include storytelling, sociodrama, drawing and painting, plays and skits, and other activities designed to encourage people to find creative ways to explore their lives, tell their stories, and recognize their own strengths. Useful resources for learning more about PAR include Whyte (1990), Elden and Chisholm (1993), Morrison and Lilford (2001), and Holter and Schwartz-Barcott (1993). The multidisciplinary journal Collaborative Inquiry is specifically devoted to PAR and related modes of research. Example of PAR: Anderson, Nyamathi, McAvoy, Conde, and Casey (2001) used PAR to examine the perceptions of adolescents in juvenile detention regarding risks for HIV and other health problems. The researchers chose PAR because it involved the adolescents in a shared partnership in the research endeavor. PAR

facilitated the empowerment of the detained adolescents by providing them with knowledge and skills generated from the research findings. RESEARCH EXAMPLES Nurse researchers have conducted studies in all of the qualitative research traditions described in this chapter, and several examples have been described. In the following sections we present more detailed descriptions of three qualitative nursing studies. Research Example of a Critical/Feminist Ethnography Brown and Fiske (2001) conducted a critical ethnography, drawing on feminist approaches, to explore First Nations (Aboriginal) women’s experiences with mainstream health care services. The study addressed two central questions: How do First Nations women describe their encounters with local mainstream health care services, and how are these encounters shaped by social, political, and economic factors? The research was conducted in partnership with a First Nations reserve community with a population of 600 in a rural area of northwest Canada. The community believed that it would benefit from a thorough description of the women’s encounters as it developed plans for improving health and health care for its members. Women from the community (including one of the researchers) were renowned locally as First Nations leaders in health, and yet First Nations people were rarely invited to join nearby mainstream health boards or decision-making bodies. Using input from community leaders and elders, the researchers selected 10 women to participate in two rounds of interviews. Each woman was interviewed separately for 1 to 2 hours. The second interviews were used to clarify and verify information from the first interviews. Participants were asked to describe both positive encounters (model cases) and negative encounters (contrary cases) with the health care establishment. An interpretive thematic analysis was conducted with transcripts from these interviews. The initial analysis was subjected to critical questioning, reflection, and discussions with participants.

CHAPTER 11 Qualitative Research Design and Approaches

The narratives revealed that the “women’s encounters were shaped by racism, discrimination, and structural inequalities that continue to marginalize and disadvantage First Nations women” (p. 126). Participants described situations in which their health concerns or reported symptoms were not taken seriously or were trivialized. Encounters that revealed health care workers’ discriminatory attitudes and behaviors were found to be pervasive.

Research Example of a Phenomenological Study McInnis and White (2001) conducted a phenomenological study to explore the meaning of loneliness for 20 older adults living in the community. Before data collection, the researchers made every effort to bracket their presuppositions. For example, one of the researchers bracketed her extensive experience as a geropsychiatric nurse. In-depth interviews were conducted with each older adult. The two main questions asked were, “Tell me about your loneliness” and “Describe the circumstances around this experience” (p. 132). The interviews were recorded, transcribed, and then analyzed using Giorgi’s phenomenological method. This method entailed (1) reading and rereading the transcripts to dwell with the data, (2) identifying meaning units from each transcript, (3) expressing the psychological insight contained in each meaning unit, and (4) synthesizing all the formulated meanings into the essence of the experience. The essence of the experience of loneliness for these elders consisted of five themes. One such theme was that loneliness is a state of anxiety or fear, influenced by dependency or the fear of it, and the decreased level of functioning. An excerpt illustrating this theme is as follows: You are so alone and you don’t have anybody, you’re almost afraid, don’t want the night to come ... you see, during the day you might go out to the store or things like that ... but at night you sit and a woman is not going to be running around at night. (p. 134)

McInnis and White took great care to enhance and document the rigor of their study. For instance, a field journal was kept throughout the study to help with bracketing. The researchers also provided a piece of the documentation they maintained to give readers a concrete example of their research procedures.

267

Research Example of a Grounded Theory Study Sayre (2000) conducted a grounded theory study of adults’ perceptions of being psychiatric patients. She conducted biweekly interviews, from admission to discharge, with 35 patients diagnosed with schizophrenia who had been hospitalized in psychiatric units of an urban public facility. The central question of the inquiry was: How do individuals explain their admission to a psychiatric facility? Subquestions included: What contextual and intervening experiences influenced their explanation? What strategies resulted? and What were the consequences of those strategies? Sayre strategically combined data from multiple sources to develop a thorough and comprehensive understanding of the patients’ perceptions and strategies. Data were collected through in-depth interviews with the patients, observations, review of medical records, and informal conversations with staff. Two hundred hours of interview data were obtained. The constant comparative method of data analysis revealed that managing self-worth was the process inpatients used to cope with the stigma of being a psychiatric patient. Factors that affected individual responses to psychiatric hospitalization included substance abuse, lack of social capital (especially lack of stable housing), and medication noncompliance. Six attribution styles, rated on a continuum from acceptance to rejection of the psychiatric treatment model, emerged from the analysis: problem, disease, crisis, punishment, ordination, and violation. These attribution styles helped the patients to maintain their sense of self-worth during the stigmatizing process of psychiatric hospitalization.

S U M M A RY P O I N T S • Qualitative research involves an emergent design—a design that emerges in the field as the study unfolds. • As bricoleurs, qualitative researchers tend to be creative and intuitive, putting together an array of data drawn from many sources to arrive at a holistic understanding of a phenomenon. • Although qualitative design is elastic and flexible, qualitative researchers nevertheless can plan for broad contingencies that can be expected to pose decision opportunities for study design in the field.

268

PART 3 Designs for Nursing Research

• A naturalistic inquiry typically progresses through three broad phases: an orientation and overview phase to determine salient aspects of the phenomenon under study; a focused exploration phase that closely examines important aspects of the phenomenon; and a confirmation and closure phase to confirm findings. • Qualitative research traditions have their roots in anthropology (e.g., ethnography and ethnoscience); philosophy (phenomenology and hermeneutics); psychology (ethology and ecological psychology); sociology (grounded theory, ethnomethodology, and symbolic interaction); sociolinguistics (discourse analysis); and history (historical research). • Ethnography focuses on the culture of a group of people and relies on extensive field work. Ethnographers strive to acquire an emic (insider’s) perspective of a culture rather than an etic (outsider’s) perspective. Nurses sometimes refer to their ethnographic studies as ethnonursing research. • The concept of researcher as instrument is frequently used by ethnographers to describe the significant role researchers play in analyzing and interpreting a culture. • Phenomenology seeks to discover the essence and meaning of a phenomenon as it is experienced by people. In descriptive phenomenology, researchers strive to bracket out preconceived views and to intuit the essence of the phenomenon by remaining open to meanings attributed to it by those who have experienced it. Bracketing is not a feature of interpretive (hermeneutical) phenomenology. • Grounded theory, an approach to studying social psychological processes and social structures, aims to discover theoretical precepts grounded in the data. This approach uses constant comparison: categories elicited from the data are constantly compared with data obtained earlier so that shared themes and variations can be determined. • There are two types of grounded theory: substantive theory, which is grounded in data on a specific substantive area, and formal grounded

theory (often using data from substantive theory studies), which is at a higher level of abstraction. • A major controversy among grounded theory researchers concerns whether to follow the original Glaser and Strauss procedures or to use the adapted procedures of Strauss and Corbin; Glaser has argued that the latter approach does not result in grounded theories but rather in conceptual descriptions. • Historical research is the systematic attempt to establish facts and relationships about past events. Historical data are normally subjected to external criticism, which is concerned with the authenticity of the source, and internal criticism, which assesses the worth of the evidence. • Case studies are intensive investigations of a single entity or a small number of entities, such as individuals, groups, organizations, families, or communities; such studies usually involve collecting data over an extended period. • Narrative analysis focuses on story in studies in which the purpose is to determine how individuals make sense of events in their lives. • Qualitative outcome analysis (QOA) is a systematic means of confirming the applicability of clinical strategies suggested by a qualitative study and to evaluate clinical outcomes. • Secondary analyses of qualitative data offer special opportunities, but researchers interested in such studies face several challenges. • Qualitative metasyntheses are interpretive translations produced from the integration of findings from qualitative studies. • A number of descriptive qualitative studies have no formal name or do not fit into any disciplinary tradition. Such studies may simply be referred to as qualitative studies, naturalistic inquiries, or as qualitative content analyses. • Research is sometimes conducted within an ideological perspective, and such research tends to rely primarily on qualitative research. • Critical theory is concerned with a critique of existing social structures; critical researchers strive to conduct inquiries that involve collaboration

CHAPTER 11 Qualitative Research Design and Approaches

with participants and foster enlightened selfknowledge and transformation. Critical ethnography uses the principles of critical theory in the study of cultures. • Feminist research, like critical research, is designed to be transformative, but the focus is sharply on how gender domination and discrimination shape women’s lives and their consciousness. • Participatory action research (PAR) produces knowledge through close collaboration with groups or communities that are vulnerable to control or oppression by a dominant culture; in PAR research, methods take second place to emergent processes that can motivate people and generate community solidarity. STUDY ACTIVITIES Chapter 11 of the Study Guide to Accompany Nursing Research: Principles and Methods, 7th edition, offers various exercises and study suggestions for reinforcing concepts presented in this chapter. In addition, the following study questions can be addressed: 1. Develop a research question for a nursing research macroethnography. Then develop a research question that would be appropriate for a microethnography. 2. Suppose a researcher goes to a hospital waiting room and deliberately sits next to other people, even though there are many empty seats, and observes the behavior of those people. What type of study would this be? 3. Which of the following topics is best suited to a phenomenological inquiry? To an ethnography? To a grounded theory study? Provide a rationale for each response. a. The passage through menarche among Haitian refugees. b. The process of coping among AIDS patients. c.