Prediction Equations Do Not Eliminate Systematic Error in Self‐Reported Body Mass Index

PLANKEY MICHAEL W, JUNE STEVENS, KATHERINE M FLEGAL, PHILIP F RUST. Prediction equations do not eliminate systematic error in self‐reported body mass index. Epidemiological studies of the risks of obesity often use body mass index (BMI) calculated from self‐reported height and weight. The purpose of this study was to examine the pattern of reporting error associated with self‐reported values of BMI and to evaluate the extent to which linear regression models predict measured BMI from self‐reported data and whether these models could compensate for this reporting error. We examined measured and self‐reported weight and height on 5079 adults aged 30 years to 64 years from the second National Health and Nutrition Examination Survey. Measured and self‐reported BMI (kg/m2) was calculated, and multiple linear regression techniques were used to predict measured BMI from self‐reported BMI. The error in self‐reported BMI (self‐reported BMI minus measured BMI) was not constant but varied systematically with BMI. The correlation between measured BMI and the error in self‐reported BMI was −0.37 for men and −0.38 for women. The pattern of reporting error was only weakly associated with self‐reported BMI, with the correlation being 0.05 for men and −0.001 for women. Error in predicted BMI (predicted BMI minus measured BMI) also varied systematically with measured BMI, but less consistently with self‐reported BMI. More complex models only slightly improved the ability to predict measured BMI compared with self‐reported BMI alone. None of the equations were able to eliminate the systematic reporting error in determining measured BMI values from self‐reported data. The characteristic pattern of error associated with self‐reported BMI is difficult or impossible to correct by the use of linear regression models.