[Underestimation of obesity and overweight based on self-report measures in the general population: prevalence and a...

The study's objectives: compare self-report measures of weight and height with direct measures; assess the impact of these discrepancies on body mass index (BMI) and prevalence of overweight and obesity; and apply correction models to the self-report measures and assess the degree of improvement in the corrected measures produced with the use of these models. A cross-sectional study was performed, assessing 4,151 adults (18 to 60 years) participating in the São Paulo Megacity Epidemiological Study. Linear regression models stratified by sex were proposed for correction of self-reported measures. Agreement was assessed with the intraclass correlation coefficient for the direct measures, self-report measures, and corrected measures, and kappa coefficient for BMI classification categories. Self-reported weight and the resulting BMI were underestimated, while height was overestimated, compared to direct measures. With all the correction models, the corrected measures were closer to the direct measures. Prevalence rates for excess weight, calculated by self-report measures, were underestimated by 24% in men and by 28% in women; with corrections, the underestimation decreased to 8% and 10%, respectively. The results showed moderate agreement for self-report measures and substantial agreement for corrected measures compared to direct measures. The use of correction equations for self-report data proved to be a useful method for producing more trustworthy estimates of prevalence of overweight and obesity in the general population, usually estimated from self-report measures of weight and height in population surveys.