A novel algorithm for describing population level trends in body weight

Abstract

Modeling population trends and predicting the impact of interventions to address obesity requires algorithms for predicting body weight status in the future. Predictions can be based on statistical consideration of different risk factors, or be an extrapolation of past and current trends. Despite the well known correlation between previous and future weight, individual weight history has not been used to predict future trends. We developed a novel population-level model to examine trends of different classes of body weight considering individual body weight histories from the National Longitudinal Survey of Youth (NLSY79). A subset of data used to as sess the predictive ability of our proposed model with actual data. Our results confirm the importance of weight history in determining future weight status. Over 80% of individuals in a specific weight category (normal, overweight, obese) will stay in the same weight category after two years (except overweight females). The length of body weight stability was also found to be important. The probability of remaining normal weight increased with longer prior periods of being at a normal weight over 18 years (0.834 to 0.893). We demonstrate that an individual’s most probable weight class in the future is consistent with their maximal historical weight class.