American Journal of Epidemiology Copyright O 1995 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol 142, No 1 Printed In U SA. A BRIEF ORIGINAL CONTRIBUTION Effects of Question Order on Estimates of the Prevalence of Attempted Weight Loss Mary K. Serdula, 1 AJi H. Mokdad, 1 - 2 Elsie R. Pamuk, 3 David F. Williamson, 1 and Tim Byers 1 Although numerous surveys have been conducted to estimate the prevalence of attempted weight loss, little information is available on the possible effects of question order on the prevalence estimates. The authors examined data collected from 231,852 respondents to surveys conducted in the District of Columbia and 20 states that participated in the Behavioral Risk Factor Surveillance System between 1985 and 1992. In surveys conducted from 1985 to 1988, respondents (n = 117,827) were first asked their body weight and then were asked if they were trying to lose weight; 48% of the women and 29% of the men reported that they were trying to lose weight. In 1989, 1991, and 1992 (no questions about weight control were asked in 1990), the order of the questions was reversed so that respondents (n = 114,025) were asked whether they were trying to lose weight before they were asked to report their weight; 41 % of the women and 26% of the men reported that they were trying to lose weight. The authors conclude that survey respondents, especially women, may be more likely to report that they are trying to lose weight when questions about weight control practices immediately follow questions on current weight. This apparent effect of question order points to the need for caution in companng prevalence estimates across surveys in which the questions are not asked in a similar order, even when the questions are worded identically. Am J Epidemiol 1995;142:64-7. data collection; effect modifiers (epidemiology); weight loss Although nearly one in every three American adults is currently attempting to lose weight and over $30 billion is spent annually for weight loss products and services, Americans are becoming heavier (1). Weight control methods and practices are therefore of consid- erable public health importance in the United States (2, 3). Although several large surveys have been con- ducted to estimate the prevalence of voluntary weight loss attempts in the United States (4-6), little infor- mation is available on the effects of questionnaire design on those prevalence estimates. In particular, the ordering of questions on current weight and on weight control practices may be important. The Behavioral Risk Factor Surveillance System, a series of ongoing Received for publication August 25,1994, and in final form March 9, 1995. 1 Division of Nutrition, National Center for Chronic Disease Pre- vention and Health Promotion, Centers for Disease Control and Prevention, Atlanta, GA 2 Epidemiology Division, School of Public Health, Emory Univer- sity, Atlanta, GA. 3 Office of Analysis, Epidemiology, and Health Promotion, Na- tional Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD Reprint requests to Dr. Mary Serdula, Division of Nutrition, Mail- stop K-26, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, 4770 Buford Highway, N.E., Atlanta, GA 30341-3724 state-based risk factor surveys, provided an opportu- nity to examine the effects of altering question order- ing on the estimated prevalence of attempted weight loss. We investigated the possible effects of asking a respondent's weight and height immediately before asking about attempted weight loss versus asking questions on current weight and height at the end of the interview. MATERIALS AND METHODS Behavioral Risk Factor Surveillance System health behavior surveys are conducted by state health depart- ments in collaboration with the Centers for Disease Control and Prevention. A detailed technical descrip- tion of the survey methods is published elsewhere (7). Each year, participating states select independent probability samples of adult residents with telephones in the noninstitutionalized, civilian population. States use a multistage cluster sampling design based on the Waksberg method (8) and an identical core question- naire administered over the telephone by trained inter- viewers. States conduct the Behavioral Risk Factor Surveillance System interviews every month through- out the year. The interviews each take about 25 min- utes to conduct. From 1985 to 1992, interview partic- ipants were from the District of Columbia as well as 64 by guest on December 20, 2016 http://aje.oxfordjournals.org/ Downloaded from