BACKGROUND: Body mass index (BMI) derived from self-reported information is widely used and the validity is therefore crucial. We aim at testing the validity of self-reported height and weight, and to test if the accuracy of self-reported information can be improved by calibration by testing if calibration improved the ability to predict diabetes.

METHODS: Data from Danish Health Examination Survey (DANHES) was used. 15 692 participants who had both filled out questionnaire and participated in health examination, and 54 725 participants with questionnaire alone, were included. Data was analyzed using Pearson's R, Cohens Kappa, linear regression and Cox-regression. Self-reported values of height and weight were calibrated using coefficients obtained from linear regression analysis. To evaluate if the calibration improved the ability to predict diabetes, Akaike's information criterion was used.

RESULTS: Self-reported height, weight and BMI were highly correlated with measured values (R ≥ 0.92). BMI was under-reported by 0.32 kg m-2 and 0.38 kg m-2 in women and men. The hazard ratio (HR) (95% confidence interval) for diabetes according to measured BMI was 2.09 (1.89-2.27) and for self-reported BMI was 1.60 (1.50-1.70) per 5 kg m-2. Calibrated values of self-reported BMI improved the predictive value of BMI for the risk of diabetes.

CONCLUSIONS: Self-reported height and weight correlated highly with physical measurement of height and weight. Measured values of BMI were more strongly associated with diabetes risk as compared to self-reported values. Calibration of the self-reported values improved the accuracy of self-reported height and weight.