Rounding as an indicator of bias in reported body weight in health surveys

Abstract

Due to the higher costs and selection bias of directly measuring weight, the majority of body weight data are based on survey responses. However, these statements are subject to systematic biases of social desirability; therefore, it is important to evaluate the magnitude of bias through indirect indicators such as rounding of weights. Data from seven rounds of the Spanish National Health Survey from 1995 to 2017 were included in the study, with 113,284 subjects. A general rounding index of weights terminating in 0 and 5, and a partial rounding index that estimated the bias direction, were used to estimate the bias distribution in the self-reporting of body weight. All body weights were systematically rounded, although more strongly in the lower weights and even more so in the higher weights. Lower weights were rounded up, and the higher weights rounded down. Regarding gender, men had higher rounding indices than women. The subjects generally reported a weight closer to the socially desirable weight. Rounding allows estimating the historical evolution of this bias in health and nutrition surveys, having more accurate information by population segments and designing public policies against obesity aimed at the more affected social segments.