Validity of self-reported BMI in older adults and an adjustment model

To determine reporting bias of self-reported vs. measured anthropometrics (body weight, height, body mass index, BMI), and to adjust self-reported BMI regarding the bias.

We compared self-reported with measured anthropometrics utilizing 659 control persons (age mean 68; range 37 to 80 years) from a case-control stroke study. The Bland-Altman approach examined the agreement between self-reported and measured values. A linear model was applied to correct the bias dependent on sex, age and self-reported BMI.

Under-reporting of weight and over-reporting of height was found. On average, this resulted in lower self-reported BMIs by 1.0 kg/m² in men, 1.2 kg/m² in women (p < 0.001). Bias correction of self-reported BMIs was derived from self-reported BMI (p < 0.001), age (p < 0.001), age-BMI interaction (p < 0.001) and sex (p < 0.05). Under-estimation of correct BMI resulted in the under-estimation of an overweight prevalence, with relatively low sensitivity regarding self-reported values (88 %). Our estimates should be recalibrated, if applied to other studies.

Self-reported anthropometric measures are systematically biased despite high correlations with measured values. A correction removes the average bias and improves accuracy.