Essentials of epidemiology in public health [Fourth edition] 9781284128369, 1284128369, 9781284128352

Citation preview

FOU RTH ED ITION

ESSENTIALS OF

Epidemiology IN PUBLIC HEALTH

Ann Aschengrau, ScD Professor, Department of Epidemiology Boston University School of Public Health Boston, Massachusetts

George R. Seage III, DSc Professor of Epidemiology Harvard T.H. Chan School of Public Health Boston, Massachusetts

World Headquarters Jones & Bartlett Learning 5 Wall Street Burlington, MA 01803 978-443-5000 [email protected] www.jblearning.com Jones & Bartlett Learning books and products are available through most bookstores and online booksellers. To contact Jones & Bartlett Learning directly, call 800-832-0034, fax 978-443-8000, or visit our website, www.jblearning.com. Substantial discounts on bulk quantities of Jones & Bartlett Learning publications are available to corporations, professional associations, and other qualified organizations. For details and specific discount information, contact the special sales department at Jones & Bartlett Learning via the above contact information or send an email to [email protected]. Copyright © 2020 by Jones & Bartlett Learning, LLC, an Ascend Learning Company All rights reserved. No part of the material protected by this copyright may be reproduced or utilized in any form, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without written permission from the copyright owner. The content, statements, views, and opinions herein are the sole expression of the respective authors and not that of Jones & Bartlett Learning, LLC. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not constitute or imply its endorsement or recommendation by Jones & Bartlett Learning, LLC and such reference shall not be used for advertising or product endorsement purposes. All trademarks displayed are the trademarks of the parties noted herein. Essentials of Epidemiology in Public Health, Fourth Edition is an independent publication and has not been authorized, sponsored, or otherwise approved by the owners of the trademarks or service marks referenced in this product. There may be images in this book that feature models; these models do not necessarily endorse, represent, or participate in the activities represented in the images. Any screenshots in this product are for educational and instructive purposes only. Any individuals and scenarios featured in the case studies throughout this product may be real or fictitious, but are used for instructional purposes only. This publication is designed to provide accurate and authoritative information in regard to the Subject Matter covered. It is sold with the understanding that the publisher is not engaged in rendering legal, accounting, or other professional service. If legal advice or other expert assistance is required, the service of a competent professional person should be sought. 12843-7

Production Credits

VP, Product Management: David D. Cella Director of Product Management: Michael Brown Product Specialist: Carter McAlister Production Manager: Carolyn Rogers Pershouse Associate Production Editor, Navigate: Jamie Reynolds Senior Marketing Manager: Sophie Fleck Teague Manufacturing and Inventory Control Supervisor: Amy Bacus

Library of Congress Cataloging-in-Publication Data

Composition: codeMantra U.S. LLC Cover Design: Kristin E. Parker Rights & Media Specialist: John Rusk Media Development Editor: Shannon Sheehan Cover Image (Title Page, Chapter Opener): © Smartboy10/DigitalVision Vectors/Getty Images Printing and Binding: Bang Printing Cover Printing: Bang Printing

Names: Aschengrau, Ann, author. | Seage, George R., author. Title: Essentials of epidemiology in public health / Ann Aschengrau, ScD, Professor of Epidemiology, Boston University School of Public Health, George R. Seage III, ScD, Professor of Epidemiology, Harvard T.H. Chan School of Public Health. Description: Fourth edition. | Burlington, MA : Jones & Bartlett Learning, [2020] | Includes bibliographical references and index. Identifiers: LCCN 2018023772 | ISBN 9781284128352 (paperback) Subjects: LCSH: Epidemiology. | Public health. | Social medicine. | BISAC:  EDUCATION / General. Classification: LCC RA651 .A83 2020 | DDC 614.4–dc23 LC record available at https://lccn.loc.gov/2018023772 6048 Printed in the United States of America 22 21 20 19 18

10 9 8 7 6 5 4 3 2 1

© Smartboy10/DigitalVision Vectors/Getty Images

Contents Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

Chapter 3 Comparing Disease Frequencies. . . . . . . . . . . . . . . . 57 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Chapter 1 The Approach and Evolution of Epidemiology . . . . . . . . . . . . . . 1

Data Organization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Direct Standardization. . . . . . . . . . . . . . . . . . . . . . . . . . . 69

Definition and Goals of Public Health. . . . . . . . . . . . . 2

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Sources of Scientific Knowledge in Public Health. . 3

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Definition and Objectives of Epidemiology. . . . . . . . 5

Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

Measures of Comparison . . . . . . . . . . . . . . . . . . . . . . . . 61

Historical Development of Epidemiology. . . . . . . . . 8 Modern Epidemiology. . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Chapter 2 Measures of Disease Frequency. . . . . . . . . . . . . . . . . . 33 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Chapter 4 Sources of Public Health Data. . . . . . . . . . . . . . . . 77 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 Census of the U.S. Population. . . . . . . . . . . . . . . . . . . . 78 Vital Statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 National Survey of Family Growth. . . . . . . . . . . . . . . . 84 National Health Interview Survey. . . . . . . . . . . . . . . . 84

Definition of a Population. . . . . . . . . . . . . . . . . . . . . . . . 34

National Health and Nutrition Examination Survey. . . . . . . . . . . . . . . . . . . . . . . . . . . 85

Definitions of Health and Disease. . . . . . . . . . . . . . . . 36

Behavioral Risk Factor Surveillance System. . . . . . . 85

Changes in Disease Definitions . . . . . . . . . . . . . . . . . . 37

National Health Care Surveys. . . . . . . . . . . . . . . . . . . . 86

Measuring Disease Occurrence. . . . . . . . . . . . . . . . . . 39

National Notifiable Diseases Surveillance System. . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Types of Calculations: Ratios, Proportions, and Rates. . . . . . . . . . . . . . . . . . . . . . . . 40

Surveillance of HIV Infection. . . . . . . . . . . . . . . . . . . . . 87

Measures of Disease Frequency. . . . . . . . . . . . . . . . . . 41

Reproductive Health Statistics . . . . . . . . . . . . . . . . . . . 88

Commonly Used Measures of Disease Frequency in Public Health. . . . . . . . . . . . . . . . . . . . 51

National Immunization Survey. . . . . . . . . . . . . . . . . . . 89

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Survey of Occupational Injuries and Illnesses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 National Survey on Drug Use and Health. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

iii

iv

Contents

Air Quality System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

Overview of Case–Control Studies. . . . . . . . . . . . . . 163

Surveillance, Epidemiology and End Results Program. . . . . . . . . . . . . . . . . . . . . . . . . . 91

When Is It Desirable to Use a Particular Study Design?. . . . . . . . . . . . . . . . . . . . . . 168

Birth Defects Surveillance and Research Programs. . . . . . . . . . . . . . . . . . . . . . . 91

Other Types of Studies. . . . . . . . . . . . . . . . . . . . . . . . . . 170

Health, United States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

Demographic Yearbook . . . . . . . . . . . . . . . . . . . . . . . . . 92

Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

World Health Statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

Cancer Incidence on Five Continents . . . . . . . . . . . . 93

Chapter 7 Experimental Studies. . . . . . . 181

Other Resources. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

Overview of Experimental Studies. . . . . . . . . . . . . . 182

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

Types of Experimental Studies. . . . . . . . . . . . . . . . . . 185

Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

Study Population. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

Chapter 5 Descriptive Epidemiology . . . . 99

Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 Consent Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

Treatment Assignment . . . . . . . . . . . . . . . . . . . . . . . . . 192

Person. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Use of the Placebo and Masking. . . . . . . . . . . . . . . . 196

Place. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

Maintenance and Assessment of Compliance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

Disease Clusters and Epidemics. . . . . . . . . . . . . . . . . 105

Ascertaining the Outcomes. . . . . . . . . . . . . . . . . . . . . 200

Ebola Outbreak and Its Investigation. . . . . . . . . . . . 110

Data Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Uses of Descriptive Epidemiology. . . . . . . . . . . . . . . 116

Generalizability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Generating Hypotheses About Causal Relationships . . . . . . . . . . . . . . . . . . . . . . . . . 116

Special Issues in Experimental Studies . . . . . . . . . . 205

Public Health Planning and Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

Example: Patterns of Mortality in the United States According to Age. . . . . . . . 118 Overall Pattern of Mortality. . . . . . . . . . . . . . . . . . . . . 121 Examples: Three Important Causes of Morbidity in the United States. . . . . . . . . . . . . 129 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

Chapter 8 Cohort Studies. . . . . . . . . . . . . 211 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 Cohort Study Definitions and Overview. . . . . . . . . 212 Types of Populations Studied. . . . . . . . . . . . . . . . . . . 213 Characterization of Exposure. . . . . . . . . . . . . . . . . . . . 215 Follow-Up and Outcome Assessment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

Chapter 6 Overview of Epidemiological Study Designs . . . . . . . . . . . . . 153

Timing of Cohort Studies. . . . . . . . . . . . . . . . . . . . . . . 216

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

Sources of Information . . . . . . . . . . . . . . . . . . . . . . . . . 224

Overview of Experimental Studies. . . . . . . . . . . . . . 156

Analysis of Cohort Studies. . . . . . . . . . . . . . . . . . . . . . 229

Overview of Cohort Studies. . . . . . . . . . . . . . . . . . . . . 159

Special Types of Cohort Studies. . . . . . . . . . . . . . . . . 231

Issues in the Selection of Cohort Study Populations. . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

Contents

v

Strengths and Limitations of Cohort Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232

Controlling for Confounding in the Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

Residual Confounding. . . . . . . . . . . . . . . . . . . . . . . . . . 309

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234

Assessment of Mediation. . . . . . . . . . . . . . . . . . . . . . . 310

Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311

Chapter 9 Case–Control Studies. . . . . . . 237 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237 The Changing View of Case–Control Studies. . . . 238 When Is It Desirable to Use the Case–Control Method?. . . . . . . . . . . . . . . . . . . . . . . 242 Selection of Cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244 Selection of Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . 247 Sources of Exposure Information. . . . . . . . . . . . . . . . 252 Analysis of Case–Control Studies. . . . . . . . . . . . . . . . 255 The Case–Crossover Study: A New Type of Case–Control Study. . . . . . . . . . . 258 Applications of Case–Control Studies. . . . . . . . . . . 260 Strengths and Limitations of Case–Control Studies. . . . . . . . . . . . . . . . . . . . . . 261 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313

Chapter 12  Random Error . . . . . . . . . . . . 315 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315 History of Biostatistics in Public Health. . . . . . . . . . 316 Precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 Sampling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 Hypothesis Testing and P Values. . . . . . . . . . . . . . . . 320 Confidence Interval Estimation . . . . . . . . . . . . . . . . . 326 P-Value Function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Probability Distributions. . . . . . . . . . . . . . . . . . . . . . . . 330 Hypothesis-Testing Statistics. . . . . . . . . . . . . . . . . . . . 336 Confidence Intervals for Measures of Disease Frequency and Association . . . . . . . . . . 339 Sample Size and Power Calculations. . . . . . . . . . . . 345

Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346

Chapter 10 Bias. . . . . . . . . . . . . . . . . . . . . 267

Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 Overview of Bias. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 Selection Bias. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348

Chapter 13 Effect Measure Modification . . . . . . . . . . . . . 351

Information Bias. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291

Definitions and Terms for Effect Measure Modification. . . . . . . . . . . . . . . . . . . . . . . . 352

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292

Chapter 11  Confounding . . . . . . . . . . . . . 295

Effect Measure Modification Versus Confounding . . . . . . . . . . . . . . . . . . . . . . . . . 353 Evaluation of Effect Measure Modification. . . . . . . 354

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295

Synergy and Antagonism. . . . . . . . . . . . . . . . . . . . . . . 359

Definition and Examples of Confounding . . . . . . . 295

Choice of Measure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360

Confounding by Indication and Severity. . . . . . . . 301

Evaluating Effect Measure Modification and Confounding in Stratified Analyses. . . . . . . . . . . 361

Controlling for Confounding: General Considerations . . . . . . . . . . . . . . . . . . . . . . 302 Controlling for Confounding in the Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302

Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363

vi

Contents

Chapter 14 Critical Review of Epidemiological Studies . . . 367 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367 Guide to Answering the Critique Questions. . . . . 369 Sample Critiques of Epidemiological Studies. . . . 378 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391

Chapter 15 The Epidemiological Approach to Causation. . . . . 393 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 Definitions of a Cause. . . . . . . . . . . . . . . . . . . . . . . . . . . 395 Characteristics of a Cause. . . . . . . . . . . . . . . . . . . . . . . 397 Risk Factors Versus Causes. . . . . . . . . . . . . . . . . . . . . . 398 Historical Development of Disease Causation Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 Hill’s Guidelines for Assessing Causation . . . . . . . . 402 Use of Hill’s Guidelines by Epidemiologists . . . . . 407 Sufficient-Component Cause Model. . . . . . . . . . . . 408 Why Mainstream Scientists Believe That HIV Is the Cause of HIV/AIDS. . . . . . . . . . . . . . . . . . 411 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416

Chapter 16 Screening in Public Health Practice . . . . . . . . . . . 419 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419 Natural History of Disease. . . . . . . . . . . . . . . . . . . . . . . 420 Definition of Primary, Secondary, and Tertiary Prevention. . . . . . . . . . . . . . . . . . . . . . . . . . . 421 Appropriate Diseases for Screening. . . . . . . . . . . . . 423

Predictive Value: A Measure of Screening Program Feasibility. . . . . . . . . . . . . . . . 431 Evaluating a Screening Program. . . . . . . . . . . . . . . . 434 Bias. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434 Selecting an Outcome. . . . . . . . . . . . . . . . . . . . . . . . . . 437 Study Designs to Evaluate Screening Programs. . . . . . . . . . . . . . . . . . . . . . . . . . 438 Examples of the Effect of Screening on Public Health. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445

Chapter 17 Ethics in Research Involving Human Participants. . . . . . . 449 (Contributed by Molly Pretorius Holme) Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449 Historical Perspective. . . . . . . . . . . . . . . . . . . . . . . . . . . 450 International Ethical and Research Practice Guidelines. . . . . . . . . . . . . . . . . . . . . . . . . . . 457 The U.S. Regulatory Framework for Human Subjects Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . 458 Limitations Posed by Ethical Requirements. . . . . . 460 Contemporary Examples . . . . . . . . . . . . . . . . . . . . . . . 460 The Informed Consent Process. . . . . . . . . . . . . . . . . . 461 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466 Chapter Questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467

Chapter 18  Answers to Chapter Questions (Chapters 1–17) . . . 469

Characteristics of a Screening Test . . . . . . . . . . . . . . 426

Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493

Lead Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430

Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503

© Smartboy10/DigitalVision Vectors/Getty Images

Preface What is epidemiology, and how does it contribute to the health of our society? Most people don’t know the answer to this question. This is somewhat paradoxical because epidemiology, one of the basic sciences of public health, affects nearly everyone. It affects both the personal decisions we make about our lives and the ways in which governments, public health agencies, and medical organizations make policy decisions that affect how we live. In recent years, the field of epidemiology has expanded tremendously in size, scope, and influence. The number of epidemiologists has grown rapidly along with the number of epidemiology training programs in schools of public health and medicine. Many subspecialties have arisen to study public health questions, from the molecular to the societal level. Recent years have also witnessed an important evolution in the theory and methods of epidemiological research and analysis, causal inference, and the role of statistics (especially P values) in research. Unfortunately, few of these changes have been taught in introductory epidemiology courses, particularly those for master’s-level students. We believe this has occurred mainly because instructors have mistakenly assumed the new concepts were too difficult or arcane for beginning students. As a consequence, many generations of public health students have received a dated education. Our desire to change this practice was the main impetus for writing this book. For nearly three decades we have successfully taught both traditional and new concepts to our graduate students at Boston University and Harvard

University. Not only have our students successfully mastered the material, but they have also found that the new ideas enhanced their understanding of epidemiology and its application. In addition to providing an up-to-date education, we have taught our students the necessary skills to become knowledgeable consumers of epidemiological literature. Gaining competence in the critical evaluation of this literature is particularly important for public health practitioners because they often need to reconcile confusing and contradictory results. This textbook reflects our educational philosophy of combining theory and practice in our teaching. It is intended for public health students who will be consumers of epidemiological literature and those who will be practicing epidemiologists. The first five chapters cover basic epidemiological concepts and data sources. Chapter 1 describes the approach and evolution of epidemiology, including the definition, goals, and historical development of epidemiology and public health. Chapters 2 and 3 describe how epidemiologists measure and compare disease occurrence in populations. Chapter 4 characterizes the major sources of health data on the U.S. population and describes how to interpret these data appropriately. Chapter 5 describes how epidemiologists analyze disease patterns to understand the health status of a population, formulate and test hypotheses of disease causation, and carry out and evaluate health programs. The next four chapters of the textbook focus on epidemiological study design. vii

viii

Preface

Chapter 6 provides an overview of study designs—including experimental, cohort, case–control, cross-sectional, and ecological studies—and describes the factors that determine when a particular design is indicated. Each of the three following chapters provides a detailed description of the three main analytic designs: experimental, cohort, and case–­ control studies. The next five chapters cover the tools students need to interpret the results of epidemiological studies. Chapter 10 describes bias, including how it influences study results and the ways in which it can be avoided. Chapter 11 explains the concept of confounding, methods for assessing its presence, and methods for controlling its effects. Chapter 12 covers random error, including hypothesis testing, P-value and confidence interval estimation and interpretation, and sample size and power calculations. We believe this chapter provides a balanced view of the appropriate role of statistics in epidemiology. Chapter 13 covers the concept of effect measure modification, an often neglected topic in introductory texts. It explains the difference between confounding and effect measure modification and describes the methods for evaluating effect measure modification. Chapter 14 pulls together the information from Chapters 10 through 13 by providing a framework for evaluating the literature as well as three examples of epidemiological study critiques. Chapter 15 covers the epidemiological approach to causation, including the historical development of causation theories, Hill’s guidelines for assessing causation, and the sufficient-­ component cause model of causation. Chapter 16 explains screening in public health practice, including the natural history of disease, characteristics of diseases appropriate for screening, important features of a screening test, and methods for evaluating a screening program. Finally, Chapter 17 describes the development

and application of guidelines to ensure the ethical conduct of studies involving humans. Up-to-date examples and data from the epidemiological literature on diseases of public health importance are used throughout the book. In addition, nearly 50 new study questions were added to the fourth edition. Our educational background and research interests are also reflected in the textbook’s outlook and examples. Ann Aschengrau received her doctorate in epidemiology from the ­Harvard School of Public Health in 1987 and joined the Department of Epidemiology at the Boston University School of Public Health shortly thereafter. She is currently Professor, Associate Chair for Education, and Co-­Director of the Master of Science Degree Program in Epidemiology. For the past 30 years, she has taught introductory epidemiology to master’s-level students. Her research has focused on the environmental determinants of disease, including cancer, disorders of reproduction and child development, and substance use. George R. Seage III received his doctorate in epidemiology from the Boston University School of Public Health in 1992. For more than a decade, he served as the AIDS epidemiologist for the city of Boston and as a faculty member at the Boston University School of Public Health. He is currently Professor of ­Epidemiology at the Harvard T.H. Chan School of Public Health and Director of the Harvard Chan Program in the Epidemiology of Infectious Diseases. For over 30 years, he has taught courses in HIV epidemiology to master’s and doctoral students. His research focuses on the biological and behavioral determinants of adult and pediatric HIV transmission, natural history, and treatment. Drs. Aschengrau and Seage are happy to connect with instructors and students via email ([email protected] and gseage@hsph .harvard.edu). Also check out Dr. ­Aschengrau’s Twitter feed @AnnfromBoston.

Preface

▸▸ ■■

■■ ■■

New to This Edition Completely updated with new examples and the latest references and public health statistics New section on process of investigating infectious disease outbreaks New section on the Ebola outbreaks and their investigation in Africa

■■ ■■ ■■

■■

ix

Introduction of the latest epidemiological terms and methods New figures depicting epidemiological concepts Expanded ancillary materials, including improved PowerPoint slides, an enlarged glossary, and new in-class exercises and test questions Over 50 new review questions

© Smartboy10/DigitalVision Vectors/Getty Images

Acknowledgments Our ideas about the principles and practice of epidemiology have been greatly influenced by teachers, colleagues, and students. We feel privileged to have been inspired and nurtured by many outstanding teachers and mentors, including Richard Monson, George (Sandy) Lamb, Steve Schoenbaum, Arnold Epstein, Ken Rothman, the late Brian MacMahon, Julie Buring, Fran Cook, Ted Colton, Bob Glynn, Adrienne Cupples, George Hutchison, and the late Alan Morrison. We are pleased to help spread the knowledge they have given us to the next generation of epidemiologists. We are also indebted to the many colleagues who contributed to the numerous editions of this book in various ways, including clarifying our thinking about epidemiology and biostatistics, providing ideas about how to teach epidemiology, reviewing and commenting on drafts and revisions of the text, pilot testing drafts in their classes, and dispensing many doses of encouragement during the time it took to write all four editions of this book. Among these individuals are Bob Horsburgh, Herb Kayne, Dan Brooks, Wayne LaMorte, Michael Shwartz, Dave Ozonoff, Tricia Coogan, Meir Stampfer, Lorelei Mucci, Murray Mittleman, Fran Cook, Charlie Poole, Tom

Fleming, Megan Murray, Marc Lipsitch, Sam Bozeman, Anne Coletti, Michael Gross, Sarah Putney, Sarah Rogers, Kimberly Shea, Kunjal Patel, and Kelly Diringer Getz. We are particularly grateful to Krystal Cantos for her many contributions to this edition, particularly the new sections on disease outbreaks, and Molly Pretorius Holme for contributing the chapter on ethics in human research. Ted Colton also deserves a special acknowledgment for originally recommending us to the publisher. We thank our students for graciously reading drafts and earlier editions of this text in their epidemiology courses and for contributing many valuable suggestions for improvement. We hope that this book will serve as a useful reference as they embark on productive careers in public health. We also recognize Abt Associates, Inc., for providing George Seage with a development and dissemination grant to write the chapter on screening in public health practice. We are very grateful to the staff of Jones & Bartlett Learning for guiding the publication process so competently and quickly. Finally, we thank our son Gregory, an actor, for his patience and for providing many interesting and fun diversions along the way. Break a leg!

xi

© Smartboy10/DigitalVision Vectors/Getty Images

CHAPTER 1

The Approach and Evolution of Epidemiology LEARNING OBJECTIVES By the end of this chapter the reader will be able to: ■■ Define and discuss the goals of public health. ■■ Distinguish between basic, clinical, and public health research. ■■ Define epidemiology and explain its objectives. ■■ Discuss the key components of epidemiology (population and frequency, distribution, determinants, and control of disease). ■■ Discuss important figures in the history of epidemiology, including John Graunt, James Lind, William Farr, and John Snow. ■■ Discuss important modern studies, including the Streptomycin Tuberculosis Trial, Doll and Hill’s studies on smoking and lung cancer, and the Framingham Study. ■■ Discuss the current activities and challenges of modern epidemiologists.

▸▸

Introduction

Most people do not know what epidemiology is or how it contributes to the health of our society. This fact is somewhat paradoxical given that epidemiology pervades our lives. Consider, for example, the following statements involving epidemiological research that have made headline news: ■■ ■■ ■■ ■■ ■■

Ten years of hormone drugs benefits some women with breast cancer. Cellular telephone users who talk or text on the phone while driving cause one in four car accidents. Omega-3 pills, a popular alternative medicine, may not help with depression. Fire retardants in consumer products may pose health risks. Brazil reacts to an epidemic of Zika virus infections. 1

2

Chapter 1 The Approach and Evolution of Epidemiology

The breadth and importance of these topics indicate that epidemiology directly affects the daily lives of most people. It affects the way that individuals make personal decisions about their lives and the way that the government, public health agencies, and medical organizations make policy decisions that affect how we live. For example, the results of epidemiological studies described by the headlines might prompt a person to use a traditional medication for her depression or to replace old furniture likely to contain harmful fire retardants. It might prompt an oncologist to determine which of his breast cancer patients would reap the benefits of hormone therapy, a manufacturer to adopt safer alternatives to fire retardants, public health agencies to monitor and prevent the spread of Zika virus infection, or a state legislature to ban cell phone use by drivers. This chapter helps the reader understand what epidemiology is and how it contributes to important issues affecting the public’s health. In particular, it describes the definition, approach, and goals of epidemiology as well as key aspects of its historical development, current state, and future challenges.

▸▸

Definition and Goals of Public Health

Public health is a multidisciplinary field whose goal is to promote the health of the population through organized community efforts.1(pp3-14) In contrast to medicine, which focuses mainly on treating illness in separate individuals, public health focuses on preventing illness in the community. Key public health activities include assessing the health status of the population, diagnosing its problems, searching for the causes of those problems, and designing solutions for them. The solutions usually involve community-level interventions that control or prevent the cause of the problem. For example, public health interventions include establishing educational programs to discourage teenagers from smoking, implementing screening programs for the early detection of cancer, and passing laws that require automobile drivers and passengers to wear seat belts. Unfortunately, public health achievements are difficult to recognize because it is hard to identify people who have been spared illness.1(pp6-7) For this reason, the field of public health has received less attention and fewer resources than the field of medicine has received. Nevertheless, public health has had a greater effect on the health of populations than medicine has had. For example, since the turn of the 20th century, the average life expectancy of Americans has increased by about 30 years, from 47.3 to 78.8 years.2 Of this increase, 25 years can be attributed to improvements in public health, and only 5 years can be attributed to improvements in the medical care system.3 Public health achievements that account for improvements in health and life expectancy include the routine use of vaccinations for infectious diseases, improvements in motor vehicle and workplace safety, control of infectious diseases through improved sanitation and clean water, modification of risk factors

Sources of Scientific Knowledge in Public Health

for coronary heart disease and stroke (such as smoking cessation and blood pressure control), safer foods from decreased microbial contamination, improved access to family planning and contraceptive services, and the acknowledgment of tobacco as a health hazard and the ensuing antismoking campaigns.4 The public health system’s activities in research, education, and program implementation have made these accomplishments possible. In the United States, this system includes federal agencies, such as the Centers for Disease Control and Prevention; state and local government agencies; nongovernmental organizations, such as Mothers Against Drunk Driving; and academic institutions, such as schools of public health. This complex array of institutions has achieved success through political action and gains in scientific knowledge.1(pp5-7) Politics enters the public health process when agencies advocate for resources, develop policies and plans to improve a community’s health, and work to ensure that services needed for the protection of public health are available to all. Political action is necessary because the government usually has the responsibility for developing the activities required to protect public health.

▸▸

Sources of Scientific Knowledge in Public Health

The scientific basis of public health activities mainly comes from (1) the basic sciences, such as pathology and toxicology; (2) the clinical or medical sciences, such as internal medicine and pediatrics; and (3) the public health sciences, such as epidemiology, environmental health science, health education, and behavioral science. Research in these three areas provides complementary pieces of a puzzle that, when properly assembled, provide the scientific foundation for public health action. Other fields such as engineering and economics also contribute to public health. The three main areas approach research questions from different yet complementary viewpoints, and each field has its own particular strengths and weaknesses. Basic scientists, such as toxicologists, study disease in a laboratory setting by conducting experiments on cells, tissues, and animals. The focus of this research is often on the disease mechanism or process. Because basic scientists conduct their studies in a controlled laboratory environment, they can regulate all important aspects of the experimental conditions. For example, a laboratory experiment testing the toxicity of a chemical is conducted on genetically similar animals that live in the same physical environment, eat the same diet, and follow the same daily schedule.5(pp157-237) Animals are assigned (usually by chance) to either the test group or the control group. Using identical routes of administration, researchers give the chemical under investigation to the test group and an inert chemical to the control group. Thus, the only difference between the two groups is the dissimilar chemical deliberately introduced by

3

4

Chapter 1 The Approach and Evolution of Epidemiology

the investigator. This type of research provides valuable information on the disease process that cannot be obtained in any other way. However, the results are often difficult to extrapolate to real-life situations involving humans because of differences in susceptibility between species and differences in the exposure level between laboratory experiments and reallife settings. In general, humans are exposed to much lower doses than those used in laboratory experiments. Clinical scientists focus their research questions mainly on disease diagnosis, treatment, and prognosis in individual patients. For example, they try to determine whether a diagnostic method is accurate or a treatment is effective. Although clinicians are also involved in disease prevention, this activity has historically taken a backseat to disease diagnosis and treatment. As a consequence, clinical research studies are usually based on people who come to a medical care facility, such as a hospital or clinic. Unfortunately, these people are often unrepresentative of the full spectrum of disease in the population at large because many sick people never come to the attention of healthcare providers. Clinical scientists contribute to scientific knowledge in several important ways. First, they are usually the first to identify new diseases, the adverse effects of new exposures, and new links between an exposure and a disease. This information is typically published in case reports. For example, the epidemic of acquired immune deficiency syndrome (AIDS) (now called HIV for human immunodeficiency virus infection) officially began in the United States in 1981 when clinicians reported several cases of Pneumocystis carinii pneumonia and Kaposi’s sarcoma (a rare cancer of the blood vessels) among previously healthy, young gay men living in New York and California.6,7 These cases were notable because Pneumocystis carinii pneumonia had previously occurred only among individuals with compromised immune systems, and Kaposi’s sarcoma had occurred mainly among elderly men. We now know that these case reports described symptoms of a new disease that would eventually be called HIV/AIDS. Despite their simplicity, case reports provide important clues regarding the causes, prevention, and cures for a disease. In addition, they are often used to justify conducting more sophisticated and expensive studies. Clinical scientists also contribute to scientific knowledge by recording treatment and response information in their patients’ medical records. This information often becomes an indispensable source of research data for clinical and epidemiological studies. For example, it would have been impossible to determine the risk of breast cancer following fluoroscopic X-ray exposure without patient treatment records from the 1930s through the 1950s.8 Investigators used these records to identify the subjects for the study and gather detailed information about subjects’ radiation doses. Public health scientists study ways to prevent disease and promote health in the population at large. Public health research differs from clinical research in two important ways. First, it focuses mainly on disease prevention rather than disease treatment. Second, the units of concern

Definition and Objectives of Epidemiology

TABLE 1-1  Main Differences Among Basic, Clinical, and Public Health Science Research Characteristic

Basic

Clinical

Public health

What/who is studied

Cells, tissues, animals in laboratory settings

Sick patients who come to healthcare facilities

Populations or communities at large

Research goals

Understanding disease mechanisms and the effects of toxic substances

Improving diagnosis and treatment of disease

Prevention of disease, promotion of health

Examples

Toxicology, immunology

Internal medicine, pediatrics

Epidemiology, environmental health science

are groups of people living in the community rather than separate individuals visiting a healthcare facility. For example, a public health research project called the Home Observation and Measures of the Environment (HOME) injury study determined the effect of installing safety devices, such as stair gates and cabinet locks, on the rate of injuries among young children.9 About 350 community-dwelling mothers and their children were enrolled in this home-based project. The main differences between the three branches of scientific inquiry are summarized in TABLE 1-1. Although this is a useful way to classify the branches of scientific research, the distinctions between these areas have become blurred. For example, epidemiological methods are currently being applied to clinical medicine in a field called “clinical epidemiology.” In addition, newly developed areas of epidemiological research, such as molecular and genetic epidemiology, include the basic sciences.

▸▸

Definition and Objectives of Epidemiology

The term epidemiology is derived from the Greek words epi, which means “on or upon”; demos, which means “the common people”; and logy, which means “study.”10(pp484,599,1029) Putting these pieces together yields the following definition of epidemiology: “the study of that which falls upon the common people.” Epidemiology can also be defined as the “branch of medical science which treats epidemics.”11 The latter definition was developed by the London Epidemiological Society, which was formed in 1850 to determine the causes of cholera and other epidemic diseases and methods of preventing them.12 Over the past century, many definitions

5

6

Chapter 1 The Approach and Evolution of Epidemiology

of epidemiology have been set forth. Some early definitions reflect the field’s initial focus on infectious diseases, and later ones reflect a broader scope encompassing all diseases.12 We define epidemiology as follows: The study of the distribution and determinants of disease frequency in human populations and the application of this study to control health problems.13(p1),14(p95) Our definition is a combination of a popular one coined by MacMahon and Pugh in 1970 and another described by Porta in the sixth edition of A Dictionary of Epidemiology.14(p95),15(p1) Note that the term disease refers to a broad array of health-related states and events, including diseases, injuries, disabilities, and death. We prefer this hybrid definition because it describes both the scope and ultimate goal of epidemiology. In particular, the objectives of epidemiology are to (1) study the natural course of disease from onset to resolution, (2) determine the extent of disease in a population, (3) identify patterns and trends in disease occurrence, (4) identify the causes of disease, and (5) evaluate the effectiveness of measures that prevent and treat disease. All of these activities contribute scientific knowledge for making sound policy decisions that protect public health. Our definition of epidemiology has five key words or phrases: (1) population, (2) disease frequency, (3) disease distribution, (4) disease determinants, and (5) disease control. Each term is described in more detail in the following sections.

Population Populations are at the heart of all epidemiological activities because epidemiologists are concerned with disease occurrence in groups of people rather than in individuals. The term population refers to a group of people with a common characteristic, such as place of residence, gender, age, or use of certain medical services. For example, people who reside in the city of Boston are members of a geographically defined population. Determining the size of the population in which disease occurs is as important as counting the cases of the disease because it is only when the number of cases is related to the size of the population that we know the true frequency of disease. The size of the population is often determined by a census—that is, a complete count—of the population. Sources of these data range from the decennial census, in which the federal government attempts to count every person in the United States every 10 years, to computerized records from medical facilities that provide counts of patients who use the facilities.

Disease Frequency Disease frequency refers to quantifying how often a disease arises in a population. Counting, which is a key activity of epidemiologists, includes three steps: (1) developing a definition of disease, (2) instituting

Definition and Objectives of Epidemiology

a mechanism for counting cases of disease within a specified population, and (3) determining the size of that population. Diseases must be clearly defined to determine accurately who should be counted. Usually, disease definitions are based on a combination of physical and pathological examinations, diagnostic test results, and signs and symptoms. For example, a case definition of breast cancer might include findings of a palpable lump during a physical exam and mammographic and pathological evidence of malignant disease. Currently available sources for identifying and counting cases of disease include hospital patient rosters; death certificates; special reporting systems, such as registries of cancer and birth defects; and special surveys. For example, the National Health Interview Survey is a federally funded study that has collected data on the health status of the U.S. population since the 1950s. Its purpose is to “monitor the health of the United States population” by collecting information on a broad range of topics, including health indicators, healthcare utilization and access, and health-related behaviors.16

Disease Distribution Disease distribution refers to the analysis of disease patterns according to the characteristics of person, place, and time, in other words, who is getting the disease, where it is occurring, and how it is changing over time. Variations in disease frequency by these three characteristics provide useful information that helps epidemiologists understand the health status of a population; formulate hypotheses about the determinants of a disease; and plan, implement, and evaluate public health programs to control and prevent adverse health events.

Disease Determinants Disease determinants are factors that bring about a change in a person’s health or make a difference in a person’s health.14(p73) Thus, determinants consist of both causal and preventive factors. Determinants also include individual, environmental, and societal characteristics. Individual determinants consist of a person’s genetic makeup, gender, age, immunity level, diet, behaviors, and existing diseases. For example, the risk of breast cancer is increased among women who carry genetic alterations, such as BRCA1 and BRCA2; are elderly; give birth at a late age; have a history of certain benign breast conditions; or have a history of radiation exposure to the chest.17 Environmental and societal determinants are external to the individual and thereby encompass a wide range of natural, social, and economic events and conditions. For example, the presence of infectious agents, reservoirs in which the organism multiplies, vectors that transport the agent, poor and crowded housing conditions, and political instability are environmental and social factors that cause many communicable diseases around the world.

7

8

Chapter 1 The Approach and Evolution of Epidemiology

Epidemiological research involves generating and testing specific hypotheses about disease determinants. A hypothesis is defined as “a tentative explanation for an observation, phenomenon, or scientific problem that can be tested by further investigation.”10(p866) Generating hypotheses is a process that involves creativity and imagination and usually includes observations on the frequency and distribution of disease in a population. Epidemiologists test hypotheses by making comparisons, usually within the context of a formal epidemiological study. The goal of a study is to harvest valid and precise information about the determinants of disease in a particular population. Epidemiological research encompasses several types of study designs; each type of study merely represents a different way of harvesting the information.

Disease Control Epidemiologists accomplish disease control through epidemiological research, as described previously, and through surveillance. The purpose of surveillance is to monitor aspects of disease occurrence that are pertinent to effective control.18(p704) For example, the Centers for Disease Control and Prevention collects information on the occurrence of HIV infection across the United States.19 For every case of HIV infection, the surveillance system gathers data on the individual’s demographic characteristics, transmission category (such as injection drug use or male-to-male sexual contact), and diagnosis date. These surveillance data are essential for formulating and evaluating programs to reduce the spread of HIV.

▸▸

Historical Development of Epidemiology

The historical development of epidemiology spans almost 400 years and is best described as slow and unsteady. Only since World War II has the field experienced a rapid expansion. The following sections, which are not meant to be a comprehensive history, highlight several historic figures and studies that made significant contributions to the evolution of epidemiological thinking. These people include John Graunt, who summarized the pattern of mortality in 17th-century London; James Lind, who used an experimental study to discover the cause and prevention of scurvy; William Farr, who pioneered a wide range of activities during the mid-19th century that are still used by modern epidemiologists; John Snow, who showed that cholera was transmitted by fecal contamination of drinking water; members of the Streptomycin in Tuberculosis Trials Committee, who conducted one of the first modern controlled clinical trials; Richard Doll and A. Bradford Hill, who conducted early research on smoking and lung cancer; and Thomas Dawber and William Kannel, who began the Framingham Study, one of the most influential and longest-running studies of heart disease in the

Historical Development of Epidemiology

world. It is clear that epidemiology has played an important role in the achievements of public health throughout its history.

John Graunt The logical underpinnings for modern epidemiological thinking evolved from the scientific revolution of the 17th century.20(p23) During this period, scientists believed that the behavior of the physical universe was orderly and could therefore be expressed in terms of mathematical relationships called “laws.” These laws are generalized statements based on observations of the physical universe, such as the time of day that the sun rises and sets. Some scientists believed that this line of thinking could be extended to the biological universe and reasoned that there must be “laws of mortality” that describe the patterns of disease and death. These scientists believed that the “laws of mortality” could be inferred by observing the patterns of disease and death among humans. John Graunt, a London tradesman and founding member of the Royal Society of London, was a pioneer in this regard. He became the first epidemiologist, statistician, and demographer when he summarized the Bills of Mortality for his 1662 publication Natural and Political Observations Mentioned in a Following Index, and Made Upon the Bills of Mortality.21 The Bills of Mortality were a weekly count of people who died that had been conducted by the parish clerks of London since 1592 because of concern about the plague. According to Graunt, the Bills were collected in the following manner: When any one dies, then, either by tolling, or ringing a Bell, or by bespeaking of a Grave of the Sexton, the same is known to the Searchers, corresponding with the said Sexton. The Searchers hereupon (who are ancient matrons, sworn to their office) repair to the place, where the dead Corps lies, and by view of the same, and by other enquiries, they examine by what Disease, or Casualty the Corps died. Hereupon they make their Report to the Parish-Clerk, and he, every Tuesday night, carries in an Accompt of all the Burials, and Christnings, happening that Week, to the Clerk of the Hall. On Wednesday the general Accompt is made up, and Printed, and on Thursdays published and dispersed to the several Families, who will pay four shillings per Annum for them.21(pp25-26) This method of reporting deaths is not very different from the system used today in the United States. Like the “searchers” of John Graunt’s time, modern physicians and medical examiners inspect the body and other evidence, such as medical records, to determine the official cause of death, which is recorded on the death certificate. The physician typically submits the certificate to the funeral director, who files it with the local

9

10

Chapter 1 The Approach and Evolution of Epidemiology

office of vital records. From there, the certificate is transferred to the city, county, state, and federal agencies that compile death statistics. Although 17th-century London families had to pay four shillings for the Bills of Mortality, these U.S. statistics are available free of charge. Graunt drew many inferences about the patterns of fertility, morbidity, and mortality by tabulating the Bills of Mortality.21 For example, he noted new diseases, such as rickets, and he made the following observations: ■■ ■■ ■■ ■■ ■■ ■■

Some diseases affected a similar number of people from year to year, whereas others varied considerably over time. Common causes of death included old age, consumption, smallpox, plague, and diseases of teeth and worms. Many greatly feared causes of death were actually uncommon, including leprosy, suicide, and starvation. Four separate periods of increased mortality caused by the plague occurred from 1592 to 1660. The mortality rate for men was higher than for women. Fall was the “most unhealthful season.”

Graunt was the first to estimate the number of inhabitants, age structure of the population, and rate of population growth in London and the first to construct a life table that summarized patterns of mortality and survival from birth until death (see TABLE 1-2). He found that the mortality rate for children was quite high; only 25 individuals out of 100 survived to age 26 years. Furthermore, even though mortality rates for adults were much lower, very few people reached old age (only 3 of 100 London residents survived to age 66 years). Graunt did not accept the statistics at face value but carefully considered their errors and ambiguities. For example, he noted that it was often difficult for the “antient matron” searchers to determine the exact cause of death. In fact, by cleverly comparing the number of plague deaths and nonplague deaths, Graunt estimated that London officials had overlooked about 20% of deaths resulting from plague.22 Although Graunt modestly stated that he merely “reduced several great confused Volumes into a few perspicuous Tables and abridged such Observations as naturally flowed from them,” historians consider his work much more significant. Statistician Walter Willcox summarized Graunt’s importance: Graunt is memorable mainly because he discovered the numerical regularity of deaths and births, of ratios of the sexes at death and birth, and of the proportion of deaths from certain causes to all causes in successive years and in different areas; or in general terms, the uniformity and predictability of many important biological phenomena taken in the mass. In doing so, he opened the way both for the later discovery of uniformities in many social and volitional phenomena like marriage, suicide and crime, and for a study of these uniformities, their nature and their limits.21(pxiii)

Historical Development of Epidemiology

TABLE 1-2  Life Table of the London Population Constructed by John Graunt in 1662 Age (years)

Number dying

Number surviving

Birth

 0

100

 6

36

  64

16

24

  40

26

15

  25

36

 9

  16

46

 6

  10

56

 4

  6

66

 3

  3

76

 2

  1

86

 1

  0

Data from Graunt J. Natural and Political Observations Made upon the Bills of Mortality. Baltimore, MD: The Johns Hopkins Press; 1932:69.

James Lind Only a few important developments occurred in the field of epidemiology during the 200-year period following the publication of John Graunt’s Bills of Mortality. One notable development was the realization that experimental studies could be used to test hypotheses about the laws of mortality. These studies involve designed experiments that investigate the role of some factor or agent in the causation, improvement, postponement, or prevention of disease.23 Their hallmarks are (1) the comparison of at least two groups of individuals (an experimental group and a control group) and (2) the active manipulation of the factor or agent under study by the investigator (i.e., the investigator assigns individuals either to receive or not to receive a preventive or therapeutic measure). In the mid-1700s, James Lind conducted one of the earliest experimental studies on the treatment of scurvy, a common disease and cause of death at the time.24(pp145-148) Although scurvy affected people living on land, sailors often became sick and died from this disease while at sea. As  a ship’s surgeon, Lind had many opportunities to observe the

11

12

Chapter 1 The Approach and Evolution of Epidemiology

epidemiology of this disease. His astute observations led him to dismiss the popular ideas that scurvy was a hereditary or infectious disease and to propose that “the principal and main predisposing cause” was moist air and that its “occasional cause” was diet.24(pp64-67,85,91) He evaluated his hypothesis about diet with the following experimental study: On the 20th of May, 1747 I took twelve patients in the scurvy, on board the Salisbury at sea. Their cases were as similar as I could have them. They all in general had putrid gums, the spots and lassitude, with weakness of their knees. They lay together in one place, being a proper apartment for the sick in the forehold; and had one diet common to all, viz, water-gruel sweetened with sugar in the morning; fresh mutton-broth often times for dinner; at other times puddings, boiled biscuit with sugar, etc.; and for supper barley and raisins, rice and currents, sago and wine, or the like. Two of these were ordered each a quart of cyder a day. Two others took twenty-five gutts of elixir vitriol three times a-day. . . . Two other[s] took two spoonfuls of vinegar three times a-day. . . . Two of the worst patients, with the tendons in the ham rigid (a symptom none of the rest had), were put under a course of sea-water. . . . Two others had each two oranges and one lemon given them every day. . . . They continued but six days under this course having consumed the quantity that could be spared. . . . The two remaining patients took bigness of a nutmeg three times a day, of an electuary recommended by an hospital-surgeon made of garlic, mustard seed.24(pp145-148) After 4 weeks, Lind reported the following: “The consequence was, that the most sudden and visible good effects were perceived from the use of the oranges and lemons; one of those who had taken them being at the end of six days fit for duty. . . . He became quite healthy before we came into Plymouth which was on the 16th of June. . . . The other was the best recovered of any in his condition; and being now deem pretty well, was appointed nurse to the rest of the sick.”24(p146) Lind concluded, “I shall here only observe, that the result of all my experiments was, that oranges and lemons were the most effectual remedies for this distemper at sea. I am apt to think oranges preferable to lemons though perhaps both given together will be found most serviceable.”24(p148) Although the sample size of Lind’s experiment was quite small by today’s standards (12 men divided into 6 groups of 2), Lind followed one of the most important principles of experimental research—ensuring that important aspects of the experimental conditions remained similar for all study subjects. Lind selected sailors whose disease was similarly severe, who lived in common quarters, and who had a similar diet. Thus, the main difference between the six groups of men was the dietary addition purposefully introduced by Lind. He also exhibited good scientific

Historical Development of Epidemiology

practice by confirming “the efficacy of these fruits by the experience of others.”24(p148) In other words, Lind did not base his final conclusions about the curative powers of citrus fruits on a single experiment, but rather he gathered additional data from other ships and voyages. Lind used the results of this experiment to suggest a method for preventing scurvy at sea. Because fresh fruits were likely to spoil and were difficult to obtain in certain ports and seasons, he proposed that lemon and orange juice extract be carried on board.24(pp155-156) The British Navy took 40 years to adopt Lind’s recommendation; within several years of doing so, it had eradicated scurvy from its ranks.24(pp377-380)

William Farr William Farr made many important advances in the field of epidemiology in the mid-1800s. Now considered one of the founders of modern epidemiology, Farr was the compiler of Statistical Abstracts for the General Registry Office in Great Britain from 1839 through 1880. In this capacity, Farr was in charge of the annual count of births, marriages, and deaths. A trained physician and self-taught mathematician, Farr pioneered a whole range of activities encompassed by modern epidemiology. He described the state of health of the population, he sought to establish the determinants of public health, and he applied the knowledge gained to the prevention and control of disease.25(ppxi-xii) One of Farr’s most important contributions involved calculations that combined registration data on births, marriages, and deaths (as the numerator) with census data on the population size (as the denominator). As he stated, “The simple process of comparing the deaths in a given time out of a given number” was “a modern discovery.”25(p170) His first annual report in 1839 demonstrated the “superior precision of numerical expressions” over literary expressions.25(p214) For example, he quantified and arranged mortality data in a manner strikingly similar to modern practice (see TABLE 1-3). Note that the annual percentage of deaths increased with age for men and women, but for most age groups, the percentage was higher for men than for women. Farr drew numerous inferences about the English population by tabulating vital statistics. For example, he reported the following findings: ■■ ■■ ■■ ■■ ■■ ■■

The average age of the English population remained relatively constant over time at 26.4 years. Widowers had a higher marriage rate than bachelors. The rate of illegitimate births declined over time. People who lived at lower elevations had higher death rates resulting from cholera than did those who lived at higher elevations. People who lived in densely populated areas had higher mortality rates than did people who lived in less populated areas. Decreases in mortality rates followed improvements in sanitation.

13

14

Chapter 1 The Approach and Evolution of Epidemiology

TABLE 1-3  Annual Mortality per Hundred Males and Females in England and Wales, 1838–1871 Age (years)

Males

Female

  0–4

  7.26

  6.27

  5–9

  0.87

  0.85

10–14

  0.49

  0.50

15–24

  0.78

  0.80

25–34

  0.99

  1.01

35–44

  1.30

  1.23

45–54

  1.85

  1.56

55–64

  3.20

  2.80

65–74

  6.71

  5.89

75–84

14.71

13.43

85–94

30.55

27.95

95+

44.11

43.04

Data from Farr W. Vital Statistics: A Memorial Volume of Selections from the Reports and Writings of William Farr. New York, NY: New York Academy of Medicine; 1975:183.

Farr used these data to form hypotheses about the causes and preventions of disease. For example, he used data on smallpox deaths to derive a general law of epidemics that accurately predicted the decline of the rinderpest epidemic in the 1860s.25(px) He used the data on the association between cholera deaths and altitude to support the hypothesis that an unhealthful climate was the disease’s cause, which was a theory that was subsequently disproved. Farr made several practical and methodological contributions to the field of epidemiology. First, he constantly strove to ensure that the collected data were accurate and complete. Second, he devised a categorization system for the causes of death so that these data could be reduced to a usable form. The system that he devised is the antecedent of the modern International Classification of Diseases, which categorizes diseases and

Historical Development of Epidemiology

causes of death. Third, Farr made a number of important contributions to the analysis of data, including the invention of the “standardized mortality rate,” an adjustment method for making fair comparisons between groups with different age structures.

John Snow Another important figure in the development of epidemiological methods during the mid-1800s was John Snow (see FIGURE 1-1). A respected physician who was a successful anesthetist and researcher on anesthetic gases, Snow was also interested in the cause and spread of cholera.26(pxxxiv) Although Farr mistakenly thought that an unhealthful climate accounted for the variation in cholera mortality by altitude, Snow used these data to support an innovative hypothesis that cholera was an infectious disease spread by fecal contamination of drinking water. Snow argued, Cholera must be a poison acting on the alimentary canal by being brought into direct contact with the alimentary mucous surface . . . the symptoms are primarily seated in the alimentary canal and all the after-symptoms of a general kind are the results of the flux from the canal.26(ppxxxiv-xxxv)

FIGURE 1-1  John Snow investigated the cause and spread of cholera in 19th-century London. Courtesy of the National Library of Medicine

15

16

Chapter 1 The Approach and Evolution of Epidemiology

His inference from this was that the poison of cholera is taken directly into the canal by mouth. This view led him to consider the media through which the poison is conveyed and the nature of the poison itself. Several circumstances lent their aid in referring him to water as the chief, though not the only, medium and to the excreted matters from the patient already stricken with cholera as the poison. In 1849, Snow published his views on the causes and transmission of cholera in a short pamphlet titled On the Mode of Communication of Cholera. During the next few years, he continued groundbreaking research testing the hypothesis that cholera was a waterborne infectious disease. The second edition of his pamphlet, published in 1855, describes in greater detail “the whole of his inquiries in regard to cholera.”26(pxxxvi) The cholera investigations for which Snow is best known are described in the following paragraphs. One such investigation focused on the Broad Street epidemic. During August and September of 1854, one of the worst outbreaks of cholera occurred in the London neighborhood surrounding Broad Street. Almost 500 fatalities from cholera occurred within a 10-day period within 250 yards of the junction between Broad and Cambridge Streets (see FIGURE 1-2). According to Snow, “The mortality in this limited area probably equals any that was ever caused in this country, even by the plague; and it was much more sudden, as the greater number of cases terminated in a few hours.”26(p38) Snow continued, As soon as I became acquainted with the situation and extent of this irruption of cholera, I suspected some contamination of the water of the much-frequented street-pump in Broad Street, near the end of Cambridge Street; but on examining the water, on the evening of the 3rd of September, I found so little impurity in it of an organic nature that I hesitated to come to a conclusion. Further inquiry, however, showed me that there was no other circumstance or agent common to the cholera occurred, and not extending beyond it, except the water of the above mentioned pump.26(p39) His subsequent investigations included a detailed study of the drinking habits of 83 individuals who died between August 31 and September 2, 1854.26(pp39-40) He found that 73 of the 83 deaths occurred among individuals living within a short distance of the Broad Street pump and that 10 deaths occurred among individuals who lived in houses that were near other pumps. According to the surviving relatives, 61 of the 73 individuals who lived near the pump drank the pump water and only 6 individuals did not. (No data could be collected for the remaining 6 people because everyone connected with these individuals had either died or departed the city.) The drinking habits of the 10 individuals who lived “decidedly nearer to another street pump” also implicated the Broad Street pump. Surviving relatives reported that 5 of the 10 drank water from the Broad Street pump because they preferred it, and 2 drank its water because they attended a nearby school.

Historical Development of Epidemiology

17

FIGURE 1-2  Distribution of deaths from cholera in the Broad Street neighborhood from August 19 to September 30, 1854. “A black mark or bar for each death is placed in the situation of the house in which the fatal attack took place. The situation of the Broad Street Pump is also indicated, as well as that of all the surrounding Pumps to which the public had access.” Courtesy of The Commonwealth Fund. In: Snow J. Snow on Cholera. New York, NY; 1936.

Snow also investigated pockets of the Broad Street population that had fewer cholera deaths. For example, he found that only 5 cholera deaths occurred among 535 inmates of a workhouse located in the Broad Street neighborhood.26(p42) The workhouse had a pump well on its premises, and “the inmates never sent to Broad Street for water.” Furthermore, no cholera deaths occurred among 70 workers at the Broad Street brewery who never obtained pump water but instead drank a daily ration of malt liquor. Although Snow never found direct evidence of sewage contamination of the Broad Street pump well, he did note that the well was near a major sewer and several cesspools. He concluded, “There had been no particular outbreak or increase of cholera, in this part of London, except among the persons who were in the habit of drinking the water of the above-mentioned pump-well.”26(p40) He presented his findings to the Board of Guardians of St. James’s Parish on September 7, and “the handle of the pump was removed on the following day.”26

18

Chapter 1 The Approach and Evolution of Epidemiology

Snow’s investigation of the Broad Street epidemic is noteworthy for several reasons. First, Snow was able to form a hypothesis implicating the Broad Street pump after he mapped the geographic distribution of the cholera deaths and studied that distribution in relation to the surrounding public water pumps (see Figure 1-2). Second, he collected data on the drinking water habits of unaffected as well as affected individuals, which allowed him to make a comparison that would support or refute his hypothesis. Third, the results of his investigation were so convincing that they led to immediate action to curb the disease, namely, the pump handle was removed. Public health action to prevent disease seldom occurs so quickly. Another series of Snow’s groundbreaking investigations on cholera focused on specific water supply companies. In particular, he found that districts supplied by the Southwark and Vauxhall Company and the Lambeth Company had higher cholera mortality rates than all of the other water companies.26(pp63-64) A few years later, a fortuitous change occurred in the water source of several of the south districts of London. As Snow stated, “The Lambeth Company removed their water works, in 1852, from opposite Hungerforth Market to Thames Ditton; thus obtaining a supply of water quite free from the sewage of London.”26(p68) Following this change, Snow obtained data from William Farr to show that “districts partially supplied with the improved water suffered much less than the others”26(p69) (see TABLE 1-4). Districts with a mixture of the clean and polluted drinking water (Southwark and Vauxhall Company and Lambeth Company combined) had 35% fewer cholera deaths (61 versus 94 deaths per 100,000) than districts with only polluted drinking water (Southwark and Vauxhall Company alone).

TABLE 1-4  Mortality from Cholera in Relation to the Water Supply Companies in the Districts of London, November 1853 Number of cholera deaths

Size of population

Death rate resulting from cholera

Southwark and Vauxhall

111

118,267

94/100,000

Southwark and Vauxhall, Lambeth

211

346,363

61/100,000

Water supply company

Data from Snow J. Snow on Cholera. New York, NY: Hafner Publishers; 1965:69. With permission from the Commonwealth Fund, New York, NY.

Historical Development of Epidemiology

TABLE 1-5  Mortality from Cholera in Relation to the Water Supply Companies in the Subdistricts of London, 1853 Number of cholera deaths

Size of population

Death rate resulting from cholera

Southwark and Vauxhall alone

192

167,654

114/100,000

Southwark and Vauxhall and Lambeth combined

182

301,149

  60/100,000

Lambeth alone

  0

  14,632

  0/100,000

Water supply company

Data from Snow J. Snow on Cholera. New York, NY: Hafner Publishers; 1965: 73. With permission from the Commonwealth Fund, New York, NY.

Snow next analyzed the cholera mortality data in smaller geographic units—London subdistricts—to make an even clearer distinction between the polluted and clean water supplies. In particular, he examined the death rates in the London subdistricts supplied by (1) the Southwark and Vauxhall Company alone (heavily polluted water), (2) the Lambeth Company alone (nonpolluted water), and (3) both companies combined (a mixture of polluted and nonpolluted water). The cholera death rates were highest in subdistricts supplied by the heavily polluted water of the Southwark and Vauxhall Company and were intermediate in subdistricts supplied by the mixed water from the Southwark and Vauxhall Company and Lambeth Company combined. No cholera deaths were observed in subdistricts supplied with the nonpolluted water of the Lambeth Company (see TABLE 1-5). Although Snow thought that these data provided “very strong evidence of the powerful influence which the drinking of water containing the sewage of a town exerts over the spread of cholera, when that disease is present,” he thought that further study of the people living in the subdistricts supplied by both companies would “yield the most incontrovertible proof on one side or another.”26 Snow understood that the differences in cholera death rates between the two companies might not have been caused by the water supply itself but rather by differences between the groups, such as differences in gender, age, and socioeconomic status. Fortunately for Snow, further study revealed that the two groups were strikingly similar. Snow made the following observation: In the subdistricts enumerated in the above table [Table 1-5] as being supplied by both companies, the mixing of the supply is of

19

20

Chapter 1 The Approach and Evolution of Epidemiology

the most intimate kind. The pipes of each Company go down all the street, and into nearly all the courts and alleys. A few houses are supplied by one Company and a few by the other, according to the decision of the owner or occupier at that time when the water companies were in active competition. In many cases a single house has a supply different from that on either side. Each company supplied both rich and poor, both large houses and small; there is no difference either in the condition or occupation of the persons receiving the water of different companies. . . . As there is no difference whatsoever, either in the houses or the people receiving the supply of the two Water Companies, or in any of the physical conditions with which they are surrounded, it is obvious that no experiment could have been devised which would more thoroughly test the effect of water supply on the progress of cholera than this, which circumstances placed ready made before the observer. The experiment, too, was on the grandest scale. No fewer than three hundred thousand people of both sexes, of every age and occupation, and of every rank and station, from gentlefolks down to the very poor, were divided into two groups without their choice, and, in most cases, without their knowledge; one group being supplied with water containing the sewage of London, and amongst it, whatever might have come from the cholera patients, the other group having water quite free from such impurity.26(pp74-75) Snow’s next step was to obtain a listing from the General Register Office of the addresses of persons dying of cholera in the subdistricts that used water from both suppliers. Then, he had the difficult task of going door to door to inquire about the drinking water supplier. According to Snow, The inquiry was necessarily attended with a good deal of trouble. There were very few instances in which I could get the information I required. Even when the water-rates are paid by the residents, they can seldom remember the name of the Water Company till they have looked for the receipt.26(p76) However, Snow found an ingenious solution to this problem: It would, indeed, have been almost impossible for me to complete the inquiry, if I had not found that I could distinguish the water of the two companies with perfect certainty by a chemical test. The test I employed was founded on the great difference in the quantity of chloride of sodium contained in the two kinds of water. On adding solution of nitrate of silver to a gallon of water of the Lambeth Company . . . only 2.28 grains of chloride of silver were obtained. . . . On treating the water of Southwark and Vauxhall Company in the same manner, 91 grains of chloride of silver were obtained.26(pp77-78)

Historical Development of Epidemiology

Thus, Snow identified the drinking water source of each household and was able to link the death rate from cholera to the water supply companies (see TABLE 1-6). He concluded, “The mortality in the houses supplied by the Southwark and Vauxhall Company was therefore between eight and nine times as great as in the houses supplied by the Lambeth Company.”26(p86) Based on his findings, Snow made a series of recommendations for the prevention of cholera. For example, he recommended, Care should be taken that the water employed for drinking and preparing food . . . is not contaminated with the contents of cesspools, house-drains, or sewers; or in the event that water free from suspicion cannot be obtained, is [sic] should be well boiled, and if possible, also filtered.26(pp133-134) Even though his results and recommendations were reported at once to William Farr and others, it took several years for Snow’s theories to be accepted.27 Fortunately, over time we have come to recognize the importance of John Snow’s contributions to our understanding of infectious diseases, in general, and cholera, in particular. For several reasons, Snow’s investigations are considered “a nearly perfect model” for epidemiological research.26(pix) First, Snow organized his observations logically so that meaningful inferences could be derived from them.20(p29) Second, he recognized that “a natural experiment” had occurred in the subdistricts of London that would enable him to gather unquestionable proof either for or against his hypothesis. Third, he conducted a quantitative analysis of the data contrasting the occurrence of cholera deaths in relation to the drinking water company.

TABLE 1-6  Mortality from Cholera in Relation to the Water Supply Companies in the Subdistricts of London, July–August 1854 Number of cholera deaths

Number of houses

Death rate resulting from cholera

Southwark and Vauxhall Company

1,263

  40,046

315/10,000

Lambeth Company

  98

  26,107

  37/10,000

Rest of London

1,422

256,423

  55/10,000

Water supply company

Adapted from Snow J. Snow on Cholera. New York, NY: Hafner Publishers; 1965: 86. With permission from the Commonwealth Fund, New York, NY; and Carvalho FM, Lima F, Kriebel D. Re: on John Snow’s unquestioned long division. Am J Epidemiol. 2004;159:422.

21

22

Chapter 1 The Approach and Evolution of Epidemiology

Modern Experimental Studies The development and application of epidemiological methods advanced slowly during the late 1800s and early 1900s. Only during the 1930s and 1940s did physicians begin to realize that it was necessary to refine the methods used to evaluate the effectiveness of disease treatments.28 Although some physicians still thought that they could assess the usefulness of a treatment merely by observing the patient’s response and comparing it with what they expected on the basis of their education and experience, many realized that “modern” experimental studies with comparable treatment and control groups of patients and comparable methods for assessing the disease changes were needed to yield correct conclusions.29

Streptomycin Tuberculosis Trial In the late 1940s, the Streptomycin in Tuberculosis Trials Committee of the British Medical Research Council conducted one of the first modern experimental studies on the use of streptomycin to treat pulmonary tuberculosis.30 According to the investigators, The natural course of pulmonary tuberculosis is . . . so variable . . . that evidence of improvement or cure following the use of a new drug in a few cases cannot be accepted as proof of the effect of that drug. The history of chemotherapeutic trials in tuberculosis is filled with errors. . . . It had become obvious that . . . conclusions regarding the clinical effect of a new chemotherapeutic agent . . . could be considered valid only if based on . . . controlled clinical trials.30(p4582) Medical Research Council. Streptomycin in Tuberculosis Trials Committee. Streptomycin treatment of pulmonary tuberculosis. Br Med J. 1948;2:769-782.

This controlled clinical trial of streptomycin included 107 patients with acute progressive bilateral pulmonary tuberculosis.30 The investigators decided to include only cases of tuberculosis that were unsuitable for other forms of treatment “to avoid having to make allowances for the effect of forms of therapy other than bed-rest.”30 In addition, they excluded cases in which spontaneous regression was likely and cases in which there was little hope of improvement. One group of 55 patients was treated with bed rest and streptomycin, and a second group of 52 patients was treated with bed rest alone.30 Patients were assigned to these groups by an innovative method known as randomization, which is defined as “an act of assigning or ordering that is the result of a random process.”23(p220) Random assignment methods include flipping a coin or using a sequence of random numbers. The exact process used in the Streptomycin Tuberculosis Trial was as follows: Determination of whether a patient would be treated by streptomycin and bed rest (S case) or by bed rest alone (C case) was made by reference to a statistical series based on random

Historical Development of Epidemiology

sampling numbers drawn up for each sex at each centre by Professor Bradford Hill; the details of the series were unknown to any of the investigators or to the co-ordinator and were contained in a set of sealed envelopes.30(p770) Patients in the streptomycin group received the drug by injection four times a day.30 Although investigators observed toxic effects in many patients, these effects were not so severe as to require the termination of treatment. During the 6-month follow-up period, 7% of the streptomycin patients died, and 27% of the control patients died. Investigators observed X-ray evidence of considerable pulmonary improvement in 51% of the streptomycin patients and only 8% of the control patients. Clinical improvement was also more common in the streptomycin group. The investigators reached the following conclusion: The course of bilateral acute progressive disease can be halted by streptomycin therapy. . . . That streptomycin was the agent responsible for this result is attested by the presence in this trial of the control group of patients, among whom considerable improvement was noted in only four (8%).30(p780) According to Richard Doll, “Few innovations have made such an impact on medicine as the controlled clinical trial that was designed by Sir Austin Bradford Hill for the Medical Research Council’s Streptomycin in Tuberculosis Trials Committee in 1946.”29(p343) Four features of the trial were particularly innovative. First and foremost was its use of randomization to assign patients to the streptomycin and control groups. Although randomization had been used in agriculture and laboratory research, this trial was one of the first instances in which it was used in medical research. The main advantage of randomization is that the order of future assignments cannot be predicted from that of past ones. Lack of predictability is the key to minimizing bias, which is defined as a systematic error in the study that causes a false conclusion. The second innovation was the placement of restrictions on the type of patient eligible for the trial.29 Patients with the type of tuberculosis that was unsuitable for therapies other than bed rest were excluded so that the results would not be obscured by the effects of other treatments. Patients who were likely to get better without any treatment or who were so ill that the streptomycin was unlikely to help were also excluded. Third, the data collection methods helped ensure that the results would be free of bias.29 These methods included using a precise and objective endpoint, such as death, and masking the investigators who were assessing the radiological improvements. Masking means that the investigators who reviewed the X-rays were unaware of the person’s treatment assignment and therefore the chances of their making a biased judgment were reduced.

23

24

Chapter 1 The Approach and Evolution of Epidemiology

Fourth, the investigators considered the ethical issues involved in conducting the trial, including whether it was ethical to withhold the streptomycin treatment from the control group.29 Before the trial was conducted, researchers had already shown that streptomycin inhibited the tubercle bacillus in vitro and reduced experimental infections of guinea pigs. Preliminary results of clinical studies had also been encouraging. However, only a small amount of the drug was available in Britain, and it was impossible to treat all patients with tuberculosis. Thus, the committee reasoned, “It would . . . have been unethical not to have seized the opportunity to design a strictly controlled trial, which could speedily and effectively reveal the value of the treatment.”29(p339)

Doll and Hill’s Studies on Smoking and Lung Cancer Most epidemiologists consider Richard Doll and A. Bradford Hill’s 1950 study on smoking and lung cancer to be one of the major milestones of epidemiology.31 Doll and Hill undertook the study because of the striking increase in lung cancer death rates in England and Wales during the 25-year period following World War I.32 Some scientists argued that the increase was the result of improvements in lung cancer diagnosis. However, Doll and Hill believed that improved diagnosis could not be entirely responsible because the number of lung cancer deaths had increased in areas with and without modern diagnostic facilities. Thus, Doll and Hill thought it was “right and proper” to justify searching for an environmental cause. Their work is emblematic of an important shift in epidemiology following World War II that redirected the focus of epidemiological research from infectious to chronic diseases.31 The shift was fueled by the idea that chronic diseases were not merely degenerative disorders of old age but rather were potentially preventable diseases with environmental origins. Doll and Hill’s first study was a “case–control study,”32 which included 709 subjects who had lung cancer (the cases) and 709 subjects who had diseases other than cancer (the controls). Control patients were purposely selected to be of the same gender, within the same 5-year age group, and in the same hospital at approximately the same time as the lung cancer patients. Patients from each group were interviewed while in the hospital for treatment about their smoking habits. In particular, they were asked, (a) if they had smoked at any period of their lives; (b) the ages at which they had started and stopped; (c) the amount they were in the habit of smoking before the onset of the illness which had brought them to the hospital; (d) the main changes in their smoking history and the maximum they had ever been in the habit of smoking; (e) the varying proportions smoked in pipes and cigarettes; and (f) whether or not they inhaled.32(p741)

Historical Development of Epidemiology

Doll and Hill found that proportionately more lung cancer patients than noncancer patients were smokers.32 In particular, 99.7% of male lung cancer patients and 95.8% of male noncancer patients smoked; 68.3% of female lung cancer patients and only 46.7% of female noncancer patients were smokers. Furthermore, a higher proportion of patients with lung cancer described themselves as heavy smokers. For example, 26.0% of the male lung cancer patients and 13.5% of the male noncancer patients reported that they had smoked 25 or more cigarettes per day before their illness began. Although the authors acknowledged that they did not know what carcinogens in tobacco smoke might be responsible, they concluded that “smoking is an important factor in the cause of lung cancer.” Three other case–control studies published in 1950 also showed an association between smoking and lung cancer. However, modern epidemiologists consider the Doll and Hill study to be “a classic exemplar for the investigation of a given outcome and an array of exposures. . . . No previous research paper lays out the essentials of the case–control method with such understanding and meticulous care.”31(p163) Far in advance of their peers, Doll and Hill considered a wide range of problems in the design and analysis of their study, including errors that may have occurred when they recruited and interviewed their subjects. In the years following the 1950 Doll and Hill study, several more studies were conducted using the case–control approach of comparing the smoking histories of patients with and without lung cancer (such as Wynder and Cornfeld’s 1953 study).33 These studies all found that the proportion of smokers, particularly heavy smokers, was higher among lung cancer patients than among noncancer patients. However, Doll and Hill believed that additional “retrospective” studies were “unlikely to advance our knowledge materially or to throw any new light upon the nature of the association.” (Retrospective studies investigate diseases that have already occurred.) They asserted that if there were “any undetected flaw in the evidence that such studies have produced, it would be exposed only by some entirely new approach.”32 The new approach that they proposed was a “prospective” study—a study that follows participants into the future to observe the occurrence of disease. Doll and Hill initiated a prospective study in 1951 by inviting 59,600 male and female members of the British Medical Association to complete a short questionnaire about their smoking habits.34 The investigators then divided the respondents into four groups on the basis of their answers: nonsmokers, light smokers, moderate smokers, and heavy smokers. The investigators obtained information on the causes of death among those who answered the questionnaire from the General Register Office in the United Kingdom. During the 29-month period following the administration of the questionnaire, 789 deaths were reported among the 24,389 male doctors aged 35 years and older. Of these deaths, 36 were reported to have died

25

26

Chapter 1 The Approach and Evolution of Epidemiology

of lung cancer as either the direct or contributory cause. After accounting for age differences between the smoking groups, the investigators found that death rates caused by lung cancer increased from 0.0 per 1,000 among nonsmokers to 0.48 per 1,000 among light smokers, 0.67 per 1,000 among moderate smokers, and 1.14 per 1,000 among heavy smokers.34 The investigators continued to follow the doctors for the next 50 years.35 During this period, they updated the smoking and mortality data. Of the 34,439 men studied, 25,346 were known to have died from 1951 through 2001. Death rates were about two to three times as high among cigarette smokers as among lifelong nonsmokers. The causes of death related to smoking included not only lung cancer but also heart disease, chronic obstructive lung disease, and a variety of vascular diseases. Because the proportion of doctors who smoked cigarettes declined over the 50-year period, the investigators were also able to examine the death rate among former smokers who had stopped smoking for various lengths of time. They found that, as compared with lifelong nonsmokers, the risk of lung cancer death among ex-smokers steadily declined in relation to the number of years since they had stopped smoking. However, those who had smoked until about age 40 before they stopped still had some excess of lung cancer at older ages. Like their first case–control study, Doll and Hill’s prospective study broke new ground. First, the study included tens of thousands of subjects, and therefore it had adequate “power” to examine numerous health effects of several levels of smoking. Second, the investigators followed the subjects for a long period of time. A long follow-up period is particularly important in the study of diseases such as cancer that take decades to develop. Third, Doll and Hill incorporated changes in smoking habits over time and therefore were able to examine the health benefits of smoking cessation.

The Framingham Study Like the work of Doll and Hill, the Framingham Study is notable for bringing about a shift in focus from infectious to noninfectious diseases following World War II. Considered “the epitome of successful epidemiologic research,” this study “has become the prototype and model of the cohort study.”31(p157) The cohort study is one of the three main study designs used in epidemiological research. According to Susser, the Framingham Study is “undisputedly the foundation stone for current ideas about risk factors in general and the prevention of ischemic heart disease in particular.” In addition, it has provided the impetus for solving difficult design and analysis issues in epidemiological research, including the development of appropriate methods for measuring the major risk factors for coronary heart disease (such as high blood pressure, elevated serum cholesterol levels, physical activity, and life stress) and for solving problems associated with measurements that vary over time.31(pp157-161) The study has also served as a

Modern Epidemiology

stimulus for developing other cohort studies of cardiovascular disease and other topics. When the Framingham Study was started in 1947, its goal was to develop ways of identifying latent cardiovascular disease among healthy volunteers.31 Within a few years, investigators expanded the study’s purpose to include determining the causes of cardiovascular disease. The study now investigates a wide variety of diseases, including stroke, diabetes, Alzheimer’s disease, and cancer, and includes the offspring and grandchildren of the original participants. Initially, the investigators enrolled about 5,000 healthy adult residents in Framingham, Massachusetts, a town located about 18 miles west of Boston.36(pp14-29) In the late 1940s, Framingham was a self-contained community of about 28,000 residents who obtained their medical care from local physicians and two hospitals near the center of town. Framingham residents were considered an excellent population for a community-based prospective study because (1) the town’s population was stable, (2) the investigators could identify a sufficient number of people with and without risk factors for heart disease, and (3) local medical doctors were eager to help recruit study subjects. For more than half a century, Framingham Study participants have undergone interviews, physical exams, laboratory tests, and other tests every 2 years.37 The interviews have gathered information on each subject’s medical history and history of cigarette smoking, alcohol use, physical activity, dietary intake, and emotional stress. The physical exams and laboratory tests have measured characteristics such as height and weight, blood pressure, vital signs and symptoms, cholesterol levels, glucose levels, bone mineral density, and genetic characteristics. These data-gathering efforts have left an immeasurable legacy of research findings on numerous topics. The contributions of the Framingham Study will only multiply in coming years with the addition of offspring, third-generation, and multiethnic cohorts.38

▸▸

Modern Epidemiology

The field of epidemiology has expanded tremendously in size, scope, and influence since the early days of the modern era. The number of epidemiologists has grown rapidly along with the number of epidemiology training programs in schools of public health and medicine. Many subspecialties have been established that are defined either by (1) disease, (2) exposure, or (3) population being studied. Disease-­specific subspecialties include reproductive, cancer, cardiovascular, infectious disease, and psychiatric epidemiology. Exposure-specific subspecialties include environmental, behavioral, and nutritional epidemiology and pharmaco-epidemiology. Population-specific subspecialties include pediatric and geriatric epidemiology.

27

28

Chapter 1 The Approach and Evolution of Epidemiology

In addition, the scope of epidemiological research has expanded in several directions. First, some epidemiologists examine health determinants at the genetic and molecular level and therefore combine the basic and public health sciences. For example, human genome epidemiology uses “epidemiological methods to assess the impact of human genetic variation on disease occurrence” and plays an essential role in the discovery of genes that cause disease and the use of genetic testing for “diagnosing, predicting, treating and preventing disease.”39 Molecular epidemiology involves the use of biolological markers to improve the measurement of exposures, disease susceptibility, and health outcomes.40(pp564-579) For example, biomarkers such as serum micronutrient levels can determine a person’s fruit and vegetable intake more accurately than can personal interviews. The second direction of epidemiological research has involved the study of determinants at the societal level.41 Social epidemiology is the study of exposures and disease susceptibility and resistance at diverse levels, including the individual, household, neighborhood, and region. For example, social epidemiologists investigate how neighborhoods, racial discrimination, and poverty influence a person’s health. The third new direction of epidemiological research has involved the analysis of determinants across the life span. Life course epidemiology, which involves the study of lasting effects of exposures during gestation, childhood, adolescence, and young adulthood on disease risk in later adult life, is based on the notion that exposures throughout life influence health in adult life.14(p74) For example, life course epidemiolgists investigate how undernutrition during gestation increases the risk of chronic diseases among adults. Theories and methods of epidemiological research have also evolved over time. For example, epidemiologists have developed new views on disease causation, and the theoretical framework underlying epidemiological study designs has matured. Finally, the availability of high-powered computer hardware and software has facilitated the analysis of large electronic datasets (now termed “big data”) with substantial numbers of people and many risk factors, enabling epidemiologists to explore new public health questions and assess the effects of multiple risk factors simultaneously. Not surprisingly, epidemiology is currently being used to investigate a wide range of important public health topics. Noteworthy topics that have been examined recently include the risk of adult-onset asthma among Black women experiencing racism,42 social determinants of multidrug-resistant tuberculosis,43 the role of exercise in reducing deaths from breast cancer,44 the numerous genetic determinants of Alzheimer’s disease,45 the effectiveness of a popular diet for diabetes prevention,46 the effect of prenatal exposure to air pollution on the risk of autism,47 and the effectiveness of mindfulness therapy in the treatment of posttraumatic stress disorder.48

Summary

The 21st century poses even more challenging problems for epidemiologists, such as “air, water and soil pollution; global warming; population growth; poverty and social inequality; and civil unrest and violence.”49(p5) Recently, some of these challenges came to the forefront in the United States when the drinking water in Flint, Michigan, became highly contaminated with lead, a potent neurotoxin, and when an alarming string of mass shootings highlighted the country’s inadequate gun control laws. An editorial on epidemiology in the 21st century noted that, like public health achievements of the past, solutions to these problems will occur through “the complementary contributions of different facets of epidemiology: calculating disease trends and probabilities, communicating findings to the public and policy makers, and designing and implementing interventions based on data.”50(p1154) The editorialists went on to observe, Epidemiology’s full value is achieved only when its contributions are placed in the context of public health action, resulting in a healthier populace. . . . Like others in epidemiology’s rich history, we should keep our eyes on the prizes of preventing disease and promoting health.50(p1155) The prospect of preventing disease and death through “analytic prowess” has attracted many great minds to epidemiology throughout its history, and it will undoubtedly continue to attract them in the coming century.

Summary Disease prevention and health promotion are the main goals of public health, a multidisciplinary field that focuses on populations and communities rather than on separate individuals. Epidemiology, one of the basic sciences of public health, is defined as “the study of the distribution and determinants of disease frequency in human populations and the application of this study to control health problems.”13(p1),14(p55) Epidemiology has played an important role in the public health achievements of the past 400 years. Key historic figures and studies have included John Graunt, who summarized the patterns of mortality in 17th-century London; James Lind, who discovered the cause and prevention of scurvy using an experimental study design in the 18th century; William Farr, who originated many modern epidemiological methods in the 19th century, including the combination of numerator and denominator data; John Snow, who demonstrated that contaminated drinking water was the mode of cholera transmission in the 19th century; members of the Streptomycin in Tuberculosis Trials Committee, who conducted one of the first modern controlled clinical trials in the 1940s; Doll and Hill, who conducted case–control studies on smoking and lung cancer in the 1950s;

29

30

Chapter 1 The Approach and Evolution of Epidemiology

and investigators who have worked on the Framingham Study, which was started in 1947 and has become one of the most influential studies of heart disease in the world. In recent years, the field of epidemiology has greatly expanded in size, scope, and influence, and epidemiologists currently investigate a wide range of important public health problems. The 21st century will pose even more challenging problems for epidemiologists. Like past public health achievements, the solutions to these problems will be found by placing the contributions of epidemiology in the context of public health action.

References 1. Schneider MJ. Introduction to Public Health. 5th ed. Burlington, MA: Jones & Bartlett Learning; 2017. 2. National Center for Health Statistics. Health, United States, 2015, With Special Feature on Racial and Ethnic Health Disparities. Hyattsville, MD: Government Printing Office; 2016. 3. Bunker JP, Frazier HS, Mosteller F. Improving health: measuring effects of medical care. Milbank Q. 1994; 72:225-258. 4. Centers for Disease Control and Prevention. Ten great public health achievements—United States, 1900-1999. MMWR. 1999;48:241-243. 5. Loomis TA. Essentials of Toxicology. 3rd ed. Philadelphia, PA: Lea & Febiger; 1978. 6. Gottlieb MS, Schanker HM, Fan PT, Saxon A, Weisman JD, Pozalski I. Pneumocystis pneumonia— Los Angeles. MMWR. 1981;30:250-252. 7. Friedman-Kien A, Laubenstein L, Marmor M, et al. Kaposi’s sarcoma and Pneumocystis pneumonia among homosexual men—New York City and California. MMWR. 1981;30:305-308. 8. Boice JD, Monson RR. Breast cancer in women after repeated fluoroscopic examination of the chest. J Natl Cancer Inst. 1977;59:823-832. 9. Phelan KJ, Khoury K, Xu Y, Liddy S, Hornung R, Lanphear BP. A randomized controlled trial of home injury hazard reduction: the HOME injury study. Arch Pediatr Adolesc Med. 2011;165:339-345. 10. Pickett JP (ed.). The American Heritage Dictionary of the English Language. 4th ed. Boston, MA: Houghton Mifflin; 2000. 11. Oxford English Dictionary (OED) Online. Oxford University Press; 2016. Accessed September 7, 2016. 12. Lilienfeld D. Definitions of epidemiology. Am J Epidemiol. 1978;107:87-90. 13. MacMahon B, Trichopoulos D. Epidemiology: Principles and Methods. 2nd ed. Boston, MA: Little, Brown and Co; 1996. 14. Porta M. A Dictionary of Epidemiology. 6th ed. New York, NY: Oxford University Press; 2014.

15. MacMahon B, Pugh TF. Epidemiology: Principles and Methods. Boston, MA: Little, Brown and Co; 1970. 16. About the National Health Interview Survey. cdc.gov. http://www.cdc.gov/nchs/nhis/about_nhis.htm. Accessed September 6, 2016. 17. Breast Cancer Facts & Figures—2017-2018. cancer.org. https://www.cancer.org/content/dam/cancer-org /research/cancer-facts-and-statistics/breast-cancer -facts-and-figures/breast-cancer-facts-and-figures -2017-2018.pdf. Accessed June 8, 2018. 18. Heymann DL (ed.). Control of Communicable Diseases Manual. 20th ed. Washington, DC: American Public Health Association; 2015. 19. HIV/AIDS: Surveillance Overview. cdc.gov. https:// www.cdc.gov/hiv/statistics/surveillance/index .html?s_cid=cs_2519. Accessed June 20, 2017. 20. Lilienfeld DE, Stolley PD. Foundations of Epidemiology. 3rd ed. New York, NY: Oxford University Press; 1994. 21. Graunt J. Natural and Political Observations Mentioned in a Following Index, and Made Upon the Bills of Mortality. Willcox, WF (ed.). Baltimore, MD: The Johns Hopkins Press; 1939. 22. Rothman KJ. Lessons from John Graunt. Lancet. 1996;347:37-39. 23. Meinert CL. Clinical Trials Dictionary, Terminology and Usage Recommendations. Baltimore, MD: The Johns Hopkins Center for Clinical Trials; 1996. 24. Stewart CP, Guthrie D (eds.). Lind’s Treatise on Scurvy. A Bicentenary Volume Containing a Reprint of the First Edition of a Treatise of the Scurvy by James Lind, MD, with Additional Notes. Edinburgh, Scotland: Edinburgh University Press; 1953. 25. Farr W. Vital Statistics: A Memorial Volume of Selections from the Reports and Writings of William Farr. New York, NY: New York Academy of Medicine; 1975. 26. Snow J. On the Mode of Communication of Cholera. 2nd ed. London, England: Churchill, 1855. Reprinted as Snow J. Snow on Cholera. New York, NY: Hafner Publishing Co; 1965.

Chapter Questions 27. Winkelstein W. A new perspective on John Snow’s communicable disease theory. Am J Epidemiol. 1995;142:S3-S9. 28. Vandenbroucke JP. Second thoughts: a short note on the history of the randomized controlled trial. J Chron Dis. 1987;40:985-987. 29. Doll R. Clinical trials: retrospect and prospect. Stat Med. 1982;1:337-44. 30. Medical Research Council. Streptomycin in Tuberculosis Trials Committee. Streptomycin treatment of pulmonary tuberculosis. Br Med J. 1948;2:769-782. 31. Susser M. Epidemiology in the United States after World War II: the evolution of technique. Epidemiol Rev. 1985;7:147-177. 32. Doll R, Hill AB. Smoking and carcinoma of the lung. Br Med J. 1950;2:739-748. 33. Wynder EL, Cornfeld J. Cancer of the lung in physicians. New Engl J Med. 1953;248:441-444. 34. Doll R, Hill AB. The mortality of doctors in relation to their smoking habits. Br Med J. 1954;1:1451-1455. 35. Doll R, Peto R, Boreham J, Sutherland I. Mortality in relation to smoking: 50 years’ observations on male British doctors. Br Med J. 2004;328(7455):1519. 36. Dawber TR. The Framingham Study: The Epidemiology of Atherosclerotic Disease. Cambridge, MA: Harvard University Press; 1980. 37. Kannel WB. The Framingham Study: its 50-year legacy and future promise. J Atheroscler Thromb. 2000;6:60-66. 38. About the Framingham Heart Study. framinghamheart study.org. https://www.framinghamheartstudy.org /about-fhs/index.php. Accessed June 20, 2017. 39. Khoury MJ, Millikan R, Little J, Gwinn M. The emergence of epidemiology in the genomics age. Intl J Epidemiol. 2004;33:936-944. 40. Khoury MJ, Millikan R, Gwinn M. Genetic and molecular epidemiology. In: Rothman KJ, Greenland S, Lash TL, eds. Modern Epidemiology. 3rd ed. Philadelpia, PA: Wolters Kluwer/Lippincott Williams and Wilkins; 2008.

31

41. Kreiger N. Epidemiology and social sciences: towards a critical reengagement in the 21st century. Epidemiol Rev. 2000;22:155-163. 42. Coogan PF, Yu J, O’Connor GT, Brown TA, Cozier YC, Palmer JR, et al. Experiences of racism and the incidence of adult-onset asthma in the Black Women’s Health Study. Chest. 2014;145:480-485. 43. Odone A, Calderon R, Becerra MC, Zhang Z, Contreras CC, Yataco R, et al. Acquired and transmitted multidrug resistant tuberculosis: the role of socal determinants. PLoS One. 2016;11(1):e0146642. 44. Ammitzboll G, Sogaard K, Karlsen RV, Tjonneland A, Johansen C, Frederiksen K, et al. Physical activity and survival in breast cancer. Eur J Cancer. 2016;66:67-74. 45. Chouraki V, Reitz C, Maury F, Bis JC, Bellenguez C, Yu L, et al. Evaluation of a genetic risk score to improve risk prediction for Alzheimer’s disease. J Alzheimers Dis. 2016;53:921-932. 46. Marrero DG, Palmer KN, Phillips EO, Miller-Kovach K, Foster GD, Saha CK. Comparison of commercial and self-initiated weight loss programs in people with prediabetes: a randomized control trial. Am J Public Health. 2016;106:949-956. 47. Volk HE, Hertz-Picciotto I, Delwiche L, Lumann F, McConnell R. Residential proximity to freeways and autism in the CHARGE study. Environ Health Perspect. 2011;119:873-877. 48. Polusny MA, Erbes CR, Thuras P, Moran A, Lamberty GJ, Collins RC, et al. Mindfulness-based stress reduction for posttraumatic stress disorder among veterans: a randomized clinical trial. JAMA. 2015;314:456-465. 49. Winkelstein W. Interface of epidemiology and history: a commentary on past, present, and future. Epidemiol Rev. 2000;22:2-6. 50. Koplan JP, Thacker SB, Lezin NA. Epidemiology in the 21st century: calculation, communication, and intervention. Am J Public Health. 1999;89:1153-1155.

Chapter Questions 1.

Define each of the following terms: a. Public health b. Epidemiology c. Population d. Disease frequency e. Disease distribution f. Disease determinants g. Disease control h. Hypothesis

2.

What is the primary difference between public health and medicine?

32

3. 4. 5. 6.

7. 8.

Chapter 1 The Approach and Evolution of Epidemiology

What public health achievements have accounted for improved life expectancy in the United States over the past century? What are the main objectives of epidemiology? How do epidemiologists quantify the disease frequency in a population? State the contribution that was made by each of the following historical figures: a. John Graunt b. John Snow c. Richard Doll and Austin Bradford Hill d. James Lind e. William Farr How are the many subspecialities of modern epidemiology typically defined? In which three directions has modern epidemiological research expanded?

© Smartboy10/DigitalVision Vectors/Getty Images

CHAPTER 2

Measures of Disease Frequency LEARNING OBJECTIVES By the end of this chapter the reader will be able to: ■■ Define and provide examples of a population. ■■ Distinguish between a fixed and dynamic (or open) population. ■■ Explain how epidemiologists create a case definition, and discuss how the definition of acquired immunodeficiency syndrome (AIDS) has changed over time. ■■ Describe the key aspects of measuring disease occurrence. ■■ Define and distinguish between cumulative incidence, incidence rate, and prevalence. ■■ Describe the mathematical relationship between the measures of disease frequency. ■■ Provide examples of commonly used measures of disease frequency in public health.

▸▸

Introduction

Epidemiologists study the distribution and determinants of disease frequency in human populations to control health problems.1(p1),2(p95) Thus, the objectives of epidemiology are to determine the extent of disease in a population, identify patterns and trends in disease occurrence, identify the causes of disease, and evaluate the effectiveness of prevention and treatment activities. Measuring how often a disease arises in a population is usually the first step in achieving these goals. This chapter describes how epidemiologists quantify the occurrence of disease in a population. Readers will learn that this quantification process involves developing a definition of a particular disease, counting the number of people who are affected by the disease, determining the size of the population from which the diseased cases arose, and accounting for the passage of time. 33

34

Chapter 2 Measures of Disease Frequency

▸▸

Definition of a Population

Because epidemiology is concerned with the occurrence of disease in groups of people rather than in individuals, populations are at the heart of epidemiologists’ measurements. A population can be defined as a group of people with a common characteristic, such as place of residence, religion, gender, age, use of hospital services, or life event (such as giving birth). Location of residence, such as a country, state, city, or neighborhood, is one of the most common ways to define a population. For example, the people who reside in the Brooklyn borough of New York City, the city of Boston, the state of Oregon, and the country of Sweden are members of distinct populations defined by geopolitical entities ranging in size from a neighborhood to an entire country. Residence near natural geographic features, such as rivers, mountains, lakes, or islands, can also be used to define a population. For example, people who live along the 2,350-mile length of the Mississippi River, around Mount St. Helens in Washington State, and on Nantucket Island off the coast of Massachusetts are members of populations defined by geographic formations. Because epidemiology focuses on disease occurrence, populations are commonly defined in relation to a medical facility, such as a medical professional’s office, clinic, or hospital. The service population of a medical facility (also called catchment population) consists of the people who use the facility’s services. This population is often difficult to define because an individual’s decision to use a facility may depend on how far it is from home, the person’s particular medical condition, his or her type of medical insurance, and so forth. Consider a situation in which a county has only one general hospital that provides the complete range of medical services, including preventive care, birthing services, and diagnostic and therapeutic services for acute and chronic conditions. The catchment population for this general hospital is likely to consist of all people who live in the county where the hospital is located (see FIGURE 2-1). Now, suppose that this hospital enhances its cardiology department, adding many well-trained clinicians and the latest diagnostic equipment. As the cardiology department’s reputation for excellent care grows, patients travel from greater distances to receive care. As a result, the catchment population for the cardiology department expands to the surrounding counties, whereas the catchment population for the other hospital services, particularly those dealing with acute conditions requiring prompt treatment, remains the single county where the hospital is situated (see Figure 2-1). Socioeconomic status is still another determinant of hospital catchment populations. Consider a city in which there are two hospitals—one public and one private—located within a few miles of each other. The private hospital generally treats patients with medical insurance, and the public hospital mainly treats patients without insurance. Even though each catchment population resides roughly within the same geographic

Definition of a Population County hospital specialty clinic with broad catchment population

nts use eside this ty r ho n ou

l ita sp

Al lc

Single county hospital with local catchment population

Other services

Residents from local and surrounding counties

Residents from local county only

yr

pi

u

nt

ta

co

l

A ll

Specialty clinic

e si

d e n ts u s e t h

o is h

s

Public and private hospitals whose catchment populations vary by socioeconomic status

Public

Private

Poor people

Middle class and wealthy people

FIGURE 2-1  Types of hospital catchment areas.

area, the two service groups are distinct in terms of income and probably many other factors (see Figure 2-1). Still another way that a population can be defined is by the occurrence of a life event, such as undergoing a medical procedure, giving birth to a child, entering or graduating from school, or serving in the military. For example, students who graduated from college in 2017 are members of the population known as the “Class of ’17,” and the men and women who served in the U.S. military during the War in Iraq are members of the population known as Iraq War veterans. Populations are often defined by other characteristics such as age, gender, religion, or type of job. A unifying framework for thinking about a population is whether its membership is permanent or transient (see TABLE 2-1). A population whose membership is permanent is called a fixed population. Its membership is always defined by a life event. For example, the people who

35

36

Chapter 2 Measures of Disease Frequency

TABLE 2-1  Types of Populations Type of population

Key element

Example

Fixed

Membership is based on an event and is permanent.

Japanese atomic bomb survivors

Dynamic or open

Membership is based on a condition and is transitory.

Residents of a city, hospital patients

were in Hiroshima, Japan, when the atomic bomb exploded at the end of World War II are members of a fixed population. This population will never gain any new members because only people who were at this historical event can be members. The opposite of a fixed population is a dynamic population, also called an open population. Its membership is defined by a changeable state or condition and therefore is transient. A person is a member of a dynamic population only as long as he or she has the defining state or condition. For example, the population of the city of Boston is dynamic because people are members only while they reside within the city limits. Turnover is always occurring because people enter the city by moving in or by birth, and people leave the city by moving away or by death. The term steady state describes a situation in which the number of people entering the population is equal to the number leaving. Dynamic populations include groups defined by geographic and hospital catchment areas, religious groups, and occupations. Regardless of the way in which it is defined, a population can be divided into subgroups on the basis of any characteristic. For example, men who undergo coronary bypass surgery are a gender subgroup of a fixed population defined by a life event, and all children up to 6 years of age who live along the Mississippi River are an age subgroup of a dynamic population defined by a geographic formation.

▸▸

Definitions of Health and Disease

In 1948, the World Health Organization defined health as “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity.”3 More recently, a national plan for improving the health of the American people stated that health is a key aspect of the quality of life and that it “reflects a personal sense of physical and mental health and the ability to react to factors in the physical and social environments.” The plan also states that the “health-related quality of life is more subjective than life expectancy and therefore can be more difficult to measure.”4

Changes in Disease Definitions

Because measurement is a cornerstone of epidemiology and “health” and “a sense of well-being” are nonspecific and difficult to quantify, epidemiologists have almost entirely focused their activities on the “absence of health,” such as specific diseases, injuries, disabilities, and death. Consequently, epidemiologists must first compose a definition of the “absence of health” or “disease” before they can begin measuring its frequency. The definition of a disease is usually based on a combination of physical and pathological examinations, diagnostic test results, and signs and symptoms. Which and how many criteria are used to define a “case” (a person who meets the disease definition) has important implications for accurately determining who has the disease. Consider the various criteria that can be used to define a heart attack case. One could use the symptoms of chest pain; the results of diagnostic tests, such as electrocardiograms; or blood enzyme tests for cardiac damage. What are the implications of using only chest pain to define heart attack cases? Using only this nonspecific symptom will capture most but not all people who have heart attacks because it will miss people who have “silent” heart attacks, which occur without chest pain. In addition, it will erroneously include many people who have other conditions that produce chest pain, such as indigestion. A definition that includes more specific criteria, such as the results of diagnostic tests, will be more accurate. For example, if positive blood enzyme tests are included, silent heart attacks are likely to be picked up and the other conditions that cause chest pain omitted. In practice, epidemiologists use all available information from physical and pathological examinations and laboratory and other tests to define a case of a disease as accurately as possible.

▸▸

Changes in Disease Definitions

Even when clear-cut criteria are used, disease definitions often change over time as more is learned about a disease and its various manifestations. For example, the official definition of HIV/AIDS (human immunodeficiency virus infection/acquired immune deficiency syndrome) has been changed several times as researchers have gained knowledge about the disease’s cause and natural course (see TABLE 2-2). The story began in the summer of 1981 when the Centers for Disease Control and Prevention (CDC) reported that Kaposi’s sarcoma and Pneumocystis carinii pneumonia had been observed among previously healthy gay men.5 Previously, Kaposi’s sarcoma, a malignant neoplasm of the blood vessels, had been seen primarily among elderly males, and opportunistic infections, such as Pneumocystis carinii pneumonia, had occurred almost exclusively in people with compromised immune systems. In 1982, the CDC composed the first definition of AIDS for national reporting as “a disease, at least moderately predictive of a defect in cell-mediated immunity, occurring in a person with no known

37

38

Chapter 2 Measures of Disease Frequency

TABLE 2-2  Changes in the Definition of HIV/AIDS over Time Year

State of knowledge and practices

1982

Knowledge very limited

Only a few conditions, including Kaposi’s sarcoma, Pneumocystis carinii pneumonia, and other severe opportunistic infections

1985

HIV virus discovered as cause of AIDS; antibody test developed

23 clinical conditions with laboratory evidence of infection

1993

Discovered importance of CD4 T lymphocytes in monitoring immunodeficiency and disease progression

26 clinical conditions and asymptomatic HIV infected cases with low CD4 T lymphocyte counts

2008

HIV testing widely available and diagnostic testing improved

Single case definition for HIV infection that includes AIDS and requires laboratory evidence of infection; 26 clinical conditions retained

2014

Further improvements in diagnotic testing, including recognition of early HIV infection

Single case definition ranging from early HIV infection to AIDS diagnosis; 27 AIDS-defining opportunistic illnesses included

Criteria for case definition

Data from Centers for Disease Control and Prevention. Update on acquired immune deficiency syndrome (AIDS)—United States. MMWR. 1982;31:507-514; Centers for Disease Control and Prevention. Revision of the case definition of acquired immune deficiency syndrome for national reporting—United States. MMWR. 1985;34:373-375; Centers for Disease Control and Prevention. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR. 1992;44:1-19; Centers for Disease Control and Prevention. Revised surveillance case definitions for HIV infection among adults, adolescents, and children aged < 18 months and for HIV infection and AIDS among children aged 18 months to < 13 years—United States, 2008. MMWR. 2008;57:1-12; and Centers for Disease Control and Prevention. Revised surveillance case definition for HIV infection–United States, 2014. MMWR. 2014; 63:1-10.

cause for diminished resistance to that disease … including Kaposi’s sarcoma, Pneumocystis carinii pneumonia, and serious other opportunistic infections.”6 Even then, the CDC acknowledged that the case definition was imperfect because it could possibly miss the full range of AIDS ­manifestations and falsely include individuals who were not truly immunodeficient. However, because the cause of AIDS was unknown at that time, the definition at least served as a consistent tool for monitoring the epidemic. The CDC made the first major change in the official AIDS case definition in 1985 after HIV was determined to be the cause of AIDS and a highly accurate laboratory test was developed to detect the antibody to the HIV virus.7 Because epidemiologists now knew what to look for and how to find its trail, it was possible to expand the AIDS case

Measuring Disease Occurrence

definition from a few severe clinical conditions to a wide range of conditions accompanied by laboratory evidence of HIV infection. Conversely, it also became possible to exclude patients who had AIDS-like symptoms but had negative HIV antibody test results. At the end of 1992, the CDC changed the AIDS case definition again after natural history studies revealed the importance of lymphocytes in the HIV infection.8 The new definition was expanded to include individuals who had no disease symptoms but had low levels of CD4 T lymphocytes, a type of white blood cell that is responsible for fighting off infections. (These cells are the primary target of HIV and an excellent marker of disease progression.) In 2008, the CDC further refined the disease definition by combining HIV infection and AIDS into a single case definition and requiring laboratory confirmation of infection.9 In 2014, further refinements were made primarily to recognize early HIV infection.10 In summary, as epidemiologists discovered the cause of AIDS, learned more about its manifestations, and developed a test to detect HIV, the case definition was expanded from a few severe diseases to many varied conditions and the results of sophisticated laboratory tests (see Table 2-2). What effect did these changes in case definition have on the number of reported HIV/AIDS cases? In general, any expansion in the case definition will increase the number of reportable cases, and any contraction will decrease that number. With HIV/AIDS, the definition changes increased the number of people who were officially counted as cases in national statistics from the 1980s until 1993.8 However, incidence subsequently declined despite the 1993 expanded case definition because of HIV testing and prevention programs, which reduced the rate of transmission, and treatment programs that included highly effective antiretroviral therapy.11

▸▸

Measuring Disease Occurrence

Epidemiologists must always consider three factors when they measure how commonly a disease occurs in a group of people: (1) the number of people who are affected by the disease, (2) the size of the population from which the cases of disease arise, and (3) the length of time that the population is followed. Failure to consider all three components will give a false impression about the effect of the disease on a population. Consider the following hypothetical data on the frequency of breast cancer in two counties. In County A, with a population of 50,000, a total of 100 new cases of breast cancer occurred over a 1-year period. In County B, with a population of 5,000, 75 new cases occurred over a 3-year period. Which county has a higher frequency of new breast cancer cases? If one considers only the number of new cases, it appears that County A has a higher frequency (100 versus 75). However, simply comparing the number of cases in each county does not provide a full picture because the cases occurred over different lengths of time (1 versus 3 years) and among populations of different sizes (50,000 versus 5,000) (see TABLE 2-3).

39

40

Chapter 2 Measures of Disease Frequency

TABLE 2-3  Hypothetical Data on the Frequency of Breast Cancer in Two Counties Type of data

County A

County B

Number of cases

100

75

Population size

50,000

5,000

Follow-up period

1 year

3 years

Comparable disease frequency

200/100,000/year

500/100,000/year

To make a meaningful comparison between the two counties, it is necessary to convert the data into the same population size and time period. Let us estimate the frequency of breast cancer in the two counties over a 1-year period and as if each population consisted of 100,000 people. The frequency of breast cancer in County A is 100 cases/50,000 population/1  year; if the county’s population were 100,000, the frequency would double to become 200 cases/100,000 population/1 year. Two steps are needed to make a similar conversion for County B: (1) divide the numerator by 3 to convert the frequency of breast cancer from a 3- to a 1-year period: 25 cases/5,000 population/1 year; (2) multiply both the numerator and denominator by 20 to estimate the frequency for a population of 100,000. Thus, the frequency of new cases in County B is 500 cases/100,000 population/1 year. Now, it is clear that the “rate” at which new cases are occurring is much higher in County B than County A (500 cases/100,000 population/1 year versus 200 cases/100,000 population/1 year, respectively) and that examining only the number of new cases gives a false impression. Note that the decision to convert the frequencies to a population size of 100,000 and a 1-year time period, although commonly done, is arbitrary. Other population sizes (such as 1,000 or 10,000) and time periods (such as 1 month or 5 years) could be used. The guiding principle is that the same population size and time period should be used for the compared groups.

▸▸

Types of Calculations: Ratios, Proportions, and Rates

Three types of calculations are used to describe and compare measures of disease occurrence: ratios, proportions, and rates (see TABLE 2-4). A ratio is simply one number divided by another. The entities represented by the two numbers are not required to be related to one another. In other words, the individuals in the numerator can be different from those in

Measures of Disease Frequency

TABLE 2-4  Types of Calculations Type of calculation

Characteristics

Ratio

Division of two unrelated numbers

Proportion

Division of two related numbers; numerator is a subset of denominator

Rate

Division of two numbers; time is always in denominator

the denominator. For example, the “gender ratio” is a ratio of two unrelated numbers: the number of males divided by the number of females, usually expressed as the number of males per 100 females. According to 2016 U.S. Census estimates, the gender ratio among U.S. residents in Florida was 95.5 males per 100 females.12 A proportion is also one number divided by another, but the entities represented by these numbers are related to one another. In fact, the numerator of a proportion is always a subset of the denominator. Proportions, also known as fractions, are often expressed as percentages and range from 0 to 1 or 0% to 100%. For example, the proportion of U.S. residents who are Black is the number of Black residents divided by the total number of U.S. residents of all races. According to 2016 U.S. Census estimates, the proportion of Black U.S. residents was 0.141, or 14.1%.12 A rate is also one number divided by another, but time is an integral part of the denominator. We are familiar with rates in our daily travels because a rate is a measure of how fast we travel. For example, in many areas of the United States, the maximum speed or rate at which cars are permitted to travel is 55 miles per hour. This rate can also be written as 55 miles/1 hour. The measure of time in the denominator is what makes this number a rate. The measures of disease occurrence calculated previously for Counties A and B are also rates (200 cases/100,000 population/1 year and 500 cases/100,000 population/1  year, respectively). Unfortunately, the term rate is often incorrectly used to describe ratios and proportions.13

▸▸

Measures of Disease Frequency

The two basic measures of disease frequency in epidemiology are incidence and prevalence. Incidence measures the occurrence of new disease, and prevalence measures the existence of current disease. Each measure describes an important part of the natural course of a disease. Incidence deals with the transition from health to disease, and prevalence focuses on the period of time that a person lives with a disease.

41

42

Chapter 2 Measures of Disease Frequency

Incidence Incidence is defined as the occurrence of new cases of disease that develop in a candidate population over a specified time period. There are three key ideas in this definition. First, incidence measures new disease events. For diseases that can occur more than once, it usually measures the first occurrence of the disease. Second, new cases of disease are measured in a candidate population, which is a population of people who are “at risk” of getting the disease. Someone is at risk because he or she has the appropriate body organ, is not immune, and so forth. For example, a woman who still has an intact uterus (i.e., she has not undergone a hysterectomy) is a candidate for getting uterine cancer, and a child who has not been fully immunized against the measles virus is a candidate for contracting a measles infection. Although it is possible to define and measure the incidence of disease in a population not at risk (e.g., the incidence of uterine cancer in women who have undergone hysterectomies is, by definition, zero), it is not a particularly interesting pursuit. Third, incidence takes into account the specific amount of time that the members of the population are followed until they develop the disease. Because incidence measures a person’s transition from a healthy to a diseased state, time must pass for this change to occur and be observed. As described in the following sections, there are two types of incidence measures: cumulative incidence and incidence rate. Although closely related, each measure has different strengths and weaknesses and is used in different settings.

Cumulative Incidence Cumulative incidence is defined as the proportion of a candidate population that becomes diseased over a specified period of time. Mathematically, it is expressed as follows: Number of new cases of disease Over a specified time period Number in candidate population Note that the numerator (new cases of disease) is a subset of the denominator (candidate population) and therefore the possible value of cumulative incidence ranges from 0 to 1, or if expressed as a percentage, from 0% to 100%. Time is not an integral part of this proportion but rather is expressed by the words that accompany the numbers of the cumulative incidence measure. Thus, cumulative incidence is dimensionless (see TABLE 2-5). Cumulative incidence can be thought of as the average risk of getting a disease over a certain period of time. (A risk is the probability of getting a disease.) A commonly cited measure of cumulative incidence is the “lifetime risk of breast cancer” among women. Currently estimated at “one in eight” among U.S. women, it means that about 12% of women will develop breast cancer sometime during the course of their lives.14

Measures of Disease Frequency

TABLE 2-5  Distinguishing Characteristics of Incidence and Prevalence Type of number

Units

Range

Numerator

Denominator

Major uses

Cumulative incidence

Proportion

None

0 to 1

New cases

Population at risk

Research on causes, prevention, and treatment of disease

Incidence rate

True rate

1/time, or t−1

0 to infinity

New cases

Person-time at risk

Research on causes, prevention, and treatment of disease

Prevalence

Proportion

None

0 to 1

Existing cases

Total population

Resource planning

Measure

Cumulative incidence is influenced by the length of time to which it applies. Generally, the cumulative incidence over a long period of time (such as a lifetime) will be higher than that over a few years. Cumulative incidence is mainly used in fixed populations when there are no or small losses to follow-up. Consider, for example, the estimated 255,000 residents of Hiroshima, Japan, who were present in the city when the atomic bomb was dropped on August 6, 1945 (see FIGURE 2-2). During the blast and in the weeks and months immediately thereafter, an estimated 65,000 people died from physical trauma, burns, and acute radiation sickness, resulting in a 25% cumulative incidence of mortality over a 4-month period.15 Officials estimate that another 21% of the population died during the year following that initial 4-month period. During the subsequent years and decades, the cumulative incidence of death was much lower, and different causes of death, such as cancer and circulatory disease, predominated.16 The cumulative incidence of death was 58% over the 53-year period from 1950 through 2003 among Hiroshima and Nagasaki survivors who were enrolled in the Life Span Cohort Study of the effects of high-dose radiation on the incidence of disease and death.16 Note that the cumulative incidence of death during the first 16 months after the bomb was dropped (25% + 21% = 46%) was close to that during the 53-year follow-up period of the Life Span Cohort Study (58%). In addition, the size of the candidate population dwindles as members die. For example, all 255,000 residents were at risk of dying when the bomb was dropped in August 1945, whereas approximately 190,000 survivors of the initial period were at risk of dying during the subsequent years (see Figure 2-2).

43

Chapter 2 Measures of Disease Frequency Number of Living Individuals in the Atomic Bomb Population of Hiroshima, Japan

44

300,000 270,000 240,000 210,000 180,000 150,000 120,000

August 1945 November 1945 November 1946

90,000 60,000

1950 2003

30,000 0 1945

1952

1959

1966

1973

1980

1987

1994

2001

2009

Year

FIGURE 2-2  Dwindling size of the Hiroshima atomic bomb population over time: 1945–2003. Data from Liebow, AA. Encounter with Disaster: A Medical Diary of Hiroshima. New York, NY: WW Norton and Co; 1970:175-176; Committee for the Compilation of Materials on Damage Caused by the Atomic Bombs in Hiroshima and Nagasaki. Hiroshima and Nagasaki: The Physical, Medical, and Social Effects of the Atomic Bombings. New York, NY: Basic Books; 1981:113; and Ozasa K, Shimizu Y, Suyama A, Kasagi F, Soda M, Grant EJ, et al. Studies of the mortality of atomic bomb survivors, Report 14, 1950-2003: an overview of cancer and noncancer diseases. Radiat Res. 2012;177:229-243.

The following hypothetical example further highlights the importance of selecting the appropriate time period for measuring cumulative incidence measure. Suppose that a pharmaceutical company has developed a new drug for the treatment of acute migraine headaches. Before the U.S. Food and Drug Administration (FDA) will approve its use, the company is required to conduct a small study (N = 20 patients) to determine whether the new drug is more effective than the most popular drug currently approved for migraine treatment. The FDA requests that the study answer two key questions: First, what proportion of patients experience symptom relief within 10  hours of taking the new drug as compared to the current drug? Second, does the new drug relieve symptoms faster than the current drug? As the study results indicate in ­FIGURE  2-3, the proportion of patients who experience symptom relief within 10 hours is identical for the two drugs. Stated in epidemiologic terms, the 10-hour cumulative incidence of relief is 60% for both drugs. However, all patients who took the new drug had symptom relief within 3 hours, whereas those who took the current drug had symptom relief several hours later. Note that the 10-hour measure of cumulative incidence misses this important finding and makes the drugs appear equally effective. Only when you examine the timing of relief can you see this important difference. Furthermore, if a 5-hour cumulative incidence had been selected as the measure of symptom relief instead of the 10-hour measure, this notable difference would have been captured. A final critical assumption that underlies the cumulative incidence measure is that everyone in the candidate population has been followed for the specified time period. Thus, in the examples above, the cumulative incidence measures assume that everyone in the migraine study population was followed for 10 hours and that everyone in the Hiroshima Life Span

Measures of Disease Frequency

X

0

X New Drug (n = 10)

0 0

X X X

0 X

X

0

X Current Drug (n = 10)

0 0

X

0

X X

X 1

45

2

3

4

5

6

7

8

9

10

Number of Hours After Taking Drug X = Symptom relief 0 = No relief

FIGURE 2-3  Importance of selecting appropriate time period for cumulative incidence.

Cohort Study was followed for 53 years. Clearly, the former is much easier to accomplish than the latter. In fact, complete follow-up is difficult to attain when there is a long follow-up period, particularly in a dynamic population in which members are continually entering and exiting. Thus, cumulative incidence is usually reserved for fixed populations, particularly when the follow-up period is short and there are no or few losses to follow-up.

Incidence Rate Incidence rate is defined as the occurrence of new cases of disease that arise during person-time of observation. Mathematically, the incidence rate is expressed as follows: Number of new cases of disease Person-time of observation in candidate population Note that the numerator for incidence rate is identical to that of cumulative incidence. The difference between the two measures lies in the denominator. The incidence rate’s denominator integrates time (t) and therefore is a true rate.13 Thus, its dimension is 1/t, or t−1, and its possible values range from zero to infinity (Table 2-5). An incidence rate of infinity is possible if all members of a population die instantaneously. The concept of person-time can be difficult to understand. Person-time is accrued only among candidates for the disease. Thus, a person contributes time to the denominator of an incidence rate only up until he or she is diagnosed with the disease of interest. However, unlike cumulative incidence, the incidence rate is not based upon the assumption

Cumulative incidence of symptom relief = 6/10 over 10 hours

Cumulative incidence of symptom relief = 6/10 over 10 hours

46

Chapter 2 Measures of Disease Frequency

that everyone in the candidate population has been followed for a specified time period. Person-time is accrued only while the candidate is being followed. Accrual of person-time stops when the person dies or is lost to follow-up (such as when a person moves to an unknown community). The incidence rate can be calculated for either a fixed or dynamic population. However, because it directly takes into account population changes (such as migration, birth, and death), it is especially useful as a measure of the transition between health and disease in dynamic populations. Consider the following population of a hypothetical town (FIGURE  2-4). This dynamic population of five individuals is followed for the 10-year period from 2006 to 2016. Person A moved into town in 2007, was diagnosed with disease in 2011, and therefore accrued 4 years of person-time. Person B was a resident of the town at the start of the observation period in 2006, died in 2012, and therefore accrued 6 person-years. Person C moved to town in 2008 and remained healthy until he moved away in 2011. The investigator could not determine where he moved and therefore could not learn whether he became diseased later. Person C was considered lost to follow-up as of the time he moved and therefore accrued 3 person-years. Person D was a resident of the town at the start of the observation period, remained healthy for the entire period, and therefore accrued 10 years of person-time. Person E was born to Person A in 2009, remained healthy for the entire observation period, and therefore accrued 7 person-years. The incidence rate in this hypothetical population is 1/30 person-years. Only one person became diseased, and 30 person-years of observation were accrued by the population. The denominator of the incidence rate is the sum of person-time accrued by each member of the population at risk. Now, consider an actual population of 59,000 U.S. Black women who were studied to quantify the effect of racism and segregation on No. Person-years accrued (Diagnosed with disease)

A

4

B

(Dies)

6

C

(Moves away, lost)

3

D

10

E

7

Total population (A, B, C, D, E) 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Observation period

FIGURE 2-4  Measurement of person-time in a hypothetical population.

∑ 30

Measures of Disease Frequency

the incidence of obesity.17 The women were enrolled in 1995 and followed through 2009 to determine their weight over time. Each woman’s follow-up time began with the date of return of her questionnaire in 1997 when questions of racism were first asked and continued until one of the following events occurred: the occurrence of obesity (defined as a body mass index ≥ 30 kg/m2), loss to follow-up, death, or the end of follow-up for the study, whichever came first. Investigators identified 1,105 incident cases of obesity among 33,235 person-years of follow-up (33.2/1,000 person-years) among women who experienced the lowest levels of daily racism in contrast to 1,043 incident cases of obesity among 23,403 person-years of follow-up (44.6/1,000 person-years) among women who experienced the highest levels of daily racism. Although it is the most accurate to determine person-time on an individual basis, as was done in this study, this method is very time consuming and labor intensive. Thus, researchers sometimes use shortcuts, such as multiplying the number of people under observation by an estimate of the average follow-up time. In the hypothetical example portrayed in Figure 2-4, five individuals who were followed for varying amounts of time accrued 30 person-years of follow-up. However, depending on the size of the population and length of follow-up, the 30 person-years could have been accrued in many ways, such as 5 people each contributing 6 years, 3 people each contributing 10 years, and so forth. Regardless of how the person-time is accrued (e.g., from 5 or 50 people), the person-time units are assumed to be equivalent. This assumption is usually reasonable, except in extreme situations in which a small number of people are followed for a long period of time. The particular time unit used to measure person-time can vary, but decisions are guided by how long it takes for the disease to develop. For example, person-years are commonly used for diseases that take many years to develop (such as cancer), and person-months or person-days are used for diseases that develop rapidly (such as infection). The number of person-time units in the denominator is arbitrary. For example, the same incidence rate can be expressed in terms of 1 person-year, 10 person-years, or 100 person-years. Epidemiologists generally use 100,000 person-years for rare diseases and those that take a long time to develop.

Relationship Between Cumulative Incidence and Incidence Rate It is possible to obtain cumulative incidence from an incidence rate. The simplest situation to demonstrate this relationship is in a fixed population with a constant incidence rate and small cumulative incidence (less than 10%). Here, the mathematical relationship is as follows: CI = IR i × ti where CI is cumulative incidence, IRi is incidence rate, and ti is the specified period of time.

47

48

Chapter 2 Measures of Disease Frequency

When the incidence rate is not constant, it is necessary to take into account the different rates that prevail during each time period: CI = ∑ (IR i × ti ) For example, the mortality rate (a type of incidence rate) among Hiroshima residents was much higher shortly after the atomic bomb explosion than during subsequent years (see Figure 2-2).

Incidence Summary In summary, two measures of disease frequency—cumulative incidence and incidence rate—focus on measuring the transition from health to disease. These measures have complementary strengths and weaknesses. The cumulative incidence is easy to calculate and understand, although it is less accurate when its assumptions are not met. Although the incidence rate has greater accuracy, its person-time denominator is more difficult to calculate and understand. Finally, the incidence rate is more useful for dynamic populations, and cumulative incidence is usually reserved for fixed populations.

Prevalence Whereas incidence measures the frequency with which new disease develops, prevalence measures the frequency of existing disease. It is simply defined as the proportion of the total population that is diseased. There are two types of prevalence measures—point prevalence and period prevalence—that relate prevalence to different amounts of time (see FIGURE 2-5). Point prevalence refers to the proportion of the population that is diseased at a single point in time and can be thought of as a single snapshot of the population. The point can be either a particular calendar date such as July 1, 2017, or a point in someone’s life, such as college graduation. Period prevalence refers to the proportion of the population that is diseased during a specified duration of time, such as Point in time

x

x

xx x

x

x

x xx

x x

x

x

x

x

x x xxx

x x TIME

January 1

July 1

Period of Time

FIGURE 2-5  Time frame for point and period prevalence.

December 31

x = case of disease

Measures of Disease Frequency

during the year 2017. The period prevalence includes the number of cases that were present at any time over the course of the year. Mathematically, point prevalence is expressed as follows: Number of existing cases of disease At a point in time Number in total population Period prevalence can be expressed as follows: Number of existing cases of disease During a period of time Number in total population Let’s use these formulas to calculate the point and period prevalence of pneumonia in a nursing home population. The point and period of interest are July 1, 2017, and January 1 through December 31, 2017, respectively. On July 1, 2017, there were 5 cases of pneumonia among the 500 nursing home residents. Thus, the point prevalence of pneumonia was 5/500, or 1%, on that date. During the period January 1 through December 31, 2017, there were 45 cases of pneumonia among the 500 nursing home residents; therefore, the period prevalence was 45/500, or 9%, during the year. Note that, in this example, the size of the nursing home population remained stable over the year, but if it had gained or lost members, the average size of the nursing home population during 2017 would have been the appropriate denominator for the period prevalence measure. Note that the numerator (existing cases) is a subset of the denominator (total population). Unlike the numerator for the two incidence measures, the prevalence numerator includes all currently living cases regardless of when they first developed. The denominator includes everyone in the population—sick, healthy, at risk, and not at risk. Because prevalence is a proportion, it is dimensionless, and its possible values range from 0 to 1, or 0% to 100% (see Table 2-5).

Relationship Between Prevalence and Incidence Prevalence depends on the rate at which new cases of disease develop (the incidence rate) as well as the duration or length of time that individuals have the disease. The duration of a disease starts at the time of diagnosis and ends when the person either is cured or dies. Mathematically, the relationship between prevalence and incidence is as follows: P = IR × D (1 − P) where P is prevalence (the proportion of the total population with the disease), (1 - P) is the proportion of the total population without the disease, IR is incidence rate, and D is the average duration (or length of time) that an individual has the disease.

49

50

Chapter 2 Measures of Disease Frequency

Incidence Prevalence

Cure and death

FIGURE 2-6  Relationship among incidence, prevalence, mortality, and cure.

This equation assumes that the population is in steady state (i.e., inflow equals outflow) and that the incidence rate and duration do not change over time. If the frequency of disease is rare (i.e., less than 10%), the equation simplifies to P = IR × D To better understand this relationship, think of the variables that influence the level of water in a sink (see FIGURE 2-6). The water level is influenced by both the inflow from the faucet and the outflow down the drain. The water level will be high if the inflow is large, if the outflow is low, or if both occur. The water level will be low if inflow is low, outflow is high, or both. Now, consider the water level in the sink as prevalence, the incoming water as incidence, and the outflowing water as diseased people who either are cured or die. The number of cases of people currently living with a disease (prevalence) will be influenced by the rate at which new cases develop as well as by the rate at which they are eliminated through cure or death.

Uses of Incidence and Prevalence Epidemiologists and other public health professionals use each measure of disease frequency for specific purposes (see Table 2-5). Incidence is most useful for evaluating the effectiveness of programs that try to prevent disease from occurring in the first place. In addition, researchers who study the causes of disease prefer to study new cases (incidence) over existing ones (prevalence) because they are usually interested in exposures that lead to developing the disease. Prevalence obscures causal relationships because it combines incidence and survival. In addition, many researchers prefer to use incidence because the timing of exposures in relation to disease occurrence can be determined more accurately. On the other hand, prevalence is useful for estimating the needs of medical facilities and allocating resources for treating people who already have a disease. In addition, researchers who study diseases such as birth defects (wherein it is difficult to gather information on defects present in miscarried

Commonly Used Measures of Disease Frequency in Public Health

and aborted fetuses) and chronic conditions such as arthritis (whose beginnings are difficult to pinpoint) have no choice but to use prevalence. Unfortunately, results of such studies are difficult to interpret because it is unclear how much the association is influenced by using a group of survivors.

▸▸

Commonly Used Measures of Disease Frequency in Public Health

There are many measures of disease frequency that are commonly used in the public health disciplines. Some are incidence measures, some are prevalence measures, some are ratios. Descriptions and examples of the major measures follow. Note that the word rate is often used incorrectly to describe a proportion or ratio. Crude mortality (or death) rate: Total number of deaths from all causes per 100,000 population per year. The term crude means that the rate is based on raw data. In 2015 the crude mortality rate in the United States was 844.0/100,000 population/year.18 Cause-specific mortality rate: Number of deaths from a specific cause per 100,000 population per year. In 2015, the cause-specific mortality rate from heart disease in the United States was 197.2/100,000/year.18 Age-specific mortality rate: Total number of deaths from all causes among individuals in a specific age category per 100,000 population per year in the age category. In 2015, the age-specific death rate was 589.6/100,000/year among U.S. children under the age of 1 year.18 Years of potential life lost: The number of years that an individual was expected to live beyond his or her death. In 2015, a total of 957 years were lost from heart disease, 1,283 years were lost from cancer, and 1,172 were lost from unintentional injuries before age 75 per 100,000 population younger than 75 years of age in the United States.18 The number of years of potential life lost reflects both the number of individuals who died of a particular cause and the age at which the death occurred. For example, a cause of death that is more common among children and young adults (such as unintentional injuries) will result in more years of life lost per individual than a cause of death that is common among the elderly (such as heart disease). Livebirth rate: Total number of livebirths per 1,000 population per year. A livebirth is a pregnancy that results in a child who, after separation, breathes or shows any other evidence of life. Sometimes, the denominator includes only women of childbearing age. In 2015, the crude livebirth rate among women who were residents of the United States was 12.4/1,000/year.18 Infant mortality rate: Number of deaths of infants less than 1 year of age per 1,000 livebirths per year. This statistic is often divided into neonatal deaths (those occurring during the first 27 days following birth) and postneonatal deaths (those occurring from 28 days through 12 months). In 2014, the infant mortality rate in the United States was 5.8/1,000 livebirths/year, the neonatal mortality rate was 3.9/1,000 livebirths/year, and the postneonatal death rate was 1.9/1,000 livebirths/year.17

51

52

Chapter 2 Measures of Disease Frequency

Birth defect rate (also called congenital anomaly or malformation rate): Number of children born with defects, usually per 10,000 births. The numerator and denominator often include both livebirths and stillbirths. In 2016– 2017, the prevalence of brain malformations, including microcephaly, was 5% among women with evidence of recent possible Zika virus infection.19 Morbidity rate: Number of existing or new cases of a particular disease or condition per 100 population. The time period that is covered and the population size in the denominator vary. Morbidity is a general word that can apply to a disease, condition, or event. For example, from 2011 to 2014, the prevalence of physician-diagnosed diabetes among U.S. adults aged 65 years and over was 20.6%.18 Attack rate: Number of new cases of disease that develop (usually during a defined and short time period) per the number in a healthy population at risk at the start of the period. This cumulative incidence measure is usually reserved for infectious disease outbreaks. For example, the 24-hour attack rate for food poisoning was 50% among people who ate chicken salad at the banquet. Case fatality rate: Number of deaths per number of cases of disease. Note that this measure is a type of cumulative incidence and therefore it is necessary to specify the length of time to which it applies. For example, in 2014 in the Democratic Republic of Congo, the 5-month case fatality rate among individuals with Ebola virus disease was 74.2%.20 Survival rate: Number of living cases per number of cases of disease. This rate is the complement of the case fatality rate and is also a cumulative incidence measure. Five-year relative survival rates for cancer compare people with a particular cancer to similar people in the general population. For example, from 2007 to 2013, 5-year relative survival rates for prostate cancer were 100% among men diagnosed while the tumor was still confined to the prostate or had spread only to the regional lymph nodes and 29.8% among men whose tumor had metastasized to distant sites.14

Summary A population is defined as a group of people with a common characteristic, such as place of residence, age, or the occurrence of an event. There are two main types of populations, fixed and dynamic (or open). The membership of a fixed population is defined by a life event and is permanent, whereas the membership of a dynamic population is defined by a changeable characteristic and is transient. Three factors should always be considered when measuring how commonly a disease occurs in a population: (1) the number of affected individuals or cases, (2) the size of the population from which the cases arise, and (3) the amount of time that this population is followed. Before epidemiologists can count the number of affected cases, they must compose a disease definition that is usually based on physical and pathological examinations, diagnostic tests, and signs and symptoms. Disease definitions often

References

53

change over time as more is learned about a disease and its manifestations. For example, the official case definition of HIV/AIDS expanded when its cause was discovered and improvements in detection were made. Incidence and prevalence are the two basic measures of disease frequency. Incidence measures the occurrence of new disease and therefore captures the transition from health to disease. Cumulative incidence and incidence rate are the two main types of incidence measures. Cumulative incidence is defined as the proportion of a candidate population that becomes diseased over a specified time period. It is a dimensionless proportion that measures the average risk of contracting a disease over a certain time period. Incidence rate is the occurrence of new cases of disease that arise during person-time of observation; therefore, it is a true rate. It is important to remember that person-time accumulates only among candidates for disease. Cumulative incidence and incidence rate are related mathematically. Both measures are most useful for evaluating the effectiveness of disease-prevention activities and for etiologic studies of disease. Prevalence measures existing disease and therefore focuses on the period when a person is ill. Prevalence measures the proportion of the total population that is diseased at a point in time or during a period of time. Its numerator consists of the number of existing cases, and its denominator includes the total population, including sick, healthy, at-risk, and immune individuals. Point prevalence refers to a single point in time and is like a snapshot. Period prevalence refers to a specific duration of time that may be derived from a series of snapshots. Prevalence is typically used for estimating the needs of medical facilities and allocating resources for treating diseased individuals. The incidence rate and prevalence are mathematically related. Many measures of disease frequency are commonly used in public health, including the crude, cause-specific, and age-specific mortality rates; morbidity rate; livebirth rate; infant mortality rate; attack rate; case fatality rate; and survival rate. Note that the term rate is often incorrectly used to refer to proportions and ratios.

References 1. MacMahon B, Trichopoulos D. Epidemiology: Principles and Methods. 2nd ed. Boston, MA: Little, Brown and Co; 1996. 2. Porta M. A Dictionary of Epidemiology. 6th ed. New York, NY: Oxford University Press; 2014. 3. Constitution of the World Health Organization. who.int. http://www.who.int/governance/eb/who _constitution_en.pdf. Accessed June 8, 2018. 4. U.S. Department of Health and Human Services. Healthy People 2010: Understanding and Improving Health. 2nd ed. Washington, DC: U.S. Government Printing Office; 2000. 5. Centers for Disease Control and Prevention. Kaposi’s sarcoma and Pneumocystis pneumonia among homosexual men—New York City and California. MMWR. 1981;30:305-308.

6. Centers for Disease Control and Prevention. Update on acquired immune deficiency syndrome (AIDS)— United States. MMWR. 1982;31:507-514. 7. Centers for Disease Control and Prevention. Revision of the case definition of acquired immune deficiency syndrome for national reporting—United States. MMWR. 1985;34:373-375. 8. Centers for Disease Control and Prevention. HIV/ AIDS Surveillance Report. 1997;9(2): 1-43. https:// www.cdc.gov/hiv/pdf/library/reports/surveillance /cdc-hiv-surveillance-report-1997-vol-9-2.pdf. Accessed June 8, 2018. 9. Centers for Disease Control and Prevention. Revised surveillance case definitions for HIV infection among adults, adolescents, and children aged