Abstract

Obesity is one of the most important risk factors for non-communicable diseases. Nutritional status is generally measured by the body mass index (BMI) and its estimation is especially relevant to analyse long-term trends of overweight and obesity at the population level. Nevertheless, in most context nationally representative data on BMI is scarce and the probability of individuals to progress to obese status is not observed longitudinally. In the literature, several authors have addressed the problem to obtain this estimation using mathematical/computational models under a scenario where the parameters and transition probabilities between nutritional states are possible to compute from regular official data. In contrast, the developing countries exhibit poor data quality and then, the approaches provided from the literature could not be extended to them. In this paper, we deal with the problem of estimating nutritional status transition probabilities in settings with scarce data such as most developing countries, formulating a non-linear programming (NLP) model for a disaggregated characterization of population assuming the transition probabilities depend on sex and age. In particular, we study the case of Chile, one of the countries with the highest prevalence of malnutrition in Latin America, using three available National Health Surveys between the years 2003 and 2017. The obtained results show a total absolute error equal to 5.11% and 10.27% for sex male and female, respectively. Finally, other model applications and extensions are discussed and future works are proposed.