Accuracy of Self-reported Weight in Hispanic/Latino Adults of the Hispanic Community Health Study/Study of Latinos

Abstract

Background: Previous US population-based studies have found that body weight may be underestimated when self-reported. However, this research may not apply to all US Hispanics/Latinos, many of whom are immigrants with distinct cultural orientations to ideal body size. We assessed the data quality and accuracy of self-reported weight in a diverse, community-based, US sample of primarily foreign-born Hispanic/Latino adults.

Methods: Using baseline data (2008-2011) from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL), we described the difference between contemporaneous self-reported and measured current body weight (n = 16,119) and used multivariate adjusted models to establish whether the observed trends in misreporting in potential predictors of inaccuracy persisted after adjustment for other predictors. Last, we described the weighted percentage agreement in body mass classification using either self-reported or measured weight (n = 16,110).

Results: Self-reported weight was well correlated with (r = 0.95) and on average 0.23 kg greater than measured weight. The range of this misreporting was large and several factors were associated with misreporting: age group, gender, body mass categories, nativity, study site by background, unit of self-report (kg or lb), and end-digit preference. The percentage agreement of body mass classification using self-reported versus measured weight was 86% and varied across prevalent health conditions.

Conclusions: The direction of misreporting in self-reported weight, and thus the anticipated bias in obesity prevalence estimates based on self-reported weights, may differ in US Hispanic/Latinos from that found in prior studies. Future investigations using self-reported body weight in US Hispanic/Latinos should consider this information for bias analyses.See video abstract at, http://links.lww.com/EDE/B276.

Figure 1

A–D. Change in scatterplots of self-reported (SR) weight (Panel A and B, unity shown by gray line) and the difference between SR and measured (M) weight as a function of M weight (Panel C and D, mean difference and ± 4 standard deviations shown by gray lines; 129 versus 137 observations beyond 4 standard deviations from the mean, respectively) in raw (Panels A,C; n=16,203 observations in black) and quality controlled datasets (Panels B,D; n=16,119 observations in black and recoded values in gray).

Figure 2

A–D. Difference between 16,119 SR and measured (M) weight (overall mean difference and ± 4 standard deviations shown by gray lines; recoded values shown in by gray dots) as a function of M weight as shown by categories of measured BMI (underweight, normal, overweight, and obese).

Figure 3

Multivariate estimated differences in misreporting (defined as the difference between self-reported and measured weights) and 95% confidence intervals comparing the seven Hispanic/Latino backgrounds (CA=Central Americans, referent for all sites but San Diego; C=Cubans; D=Dominicans; M=Mexicans, referent for San Diego; PR=Puerto Ricans; SA=South Americans; O=Other) within the study sites (The Bronx, NY; Chicago, IL; Miami, FL; San Diego, CA; ≥100 participants of a given background per site).