Accuracy of self-reported body weight: Stepped Approach Model component assessment

This study was conducted to investigate innovative solutions to a measurement problem pertaining to self-reported body weight data as a key component of the Stepped Approach Model (SAM) of service delivery. Subjects (n = 223) were randomly assigned to one of two conditions: Informed Group (of self-report and weight measurement) + six body weighing habit items (IG, n = 113) and Uninformed Group (of self-report and weight measurement) + one body weight item (UG, n = 110). A t-test indicated that IG subjects reported significantly more accurately, t(194) = 2.99, P = 0.002, and with significantly less variability than UG subjects, F(109,112) = 1.95, P < 0.0005. A multiple regression of absolute difference weight (observed--self-reported weight) on observed weight revealed consistent accuracy across the weight range for IG subjects, whereas UG subjects' accuracy decreased as body weight increased. The slope of the IG did not significantly differ from 0, t(218) = 1.44, P = 0.150, but did significantly differ from the slope of the UG, t(218) = 2.78, P = 0.006. The following conclusions are noted when IG conditions are used: (1) a three-component strategy designed for maximum effect size results in accurate reporting across the entire weight range, (2) self-reported body weights under prescribed conditions can be used as valid 'proxies' for observed measurements, and (3) SAM proponents can rely on the validity of self-report body weight as a credible basis for decisions about changing intervention steps and evaluating intervention efficacy.