Maternal dietary practices during pregnancy and obesity of neonates: a machine learning approach towards hierarchical and nested relationships in a Tibet Plateau cohort study

Abstract Studies on obesity and risk factors from a life-course perspective among residents in the Tibet Plateau with recent economic growth and increasing obesity are important and urgently needed. The birth cohort in this area provides a unique opportunity to examine the association between maternal dietary practice and neonatal obesity. The study aims to detect the prevalence of obesity among neonates, associated with maternal diet and other factors, supporting life-course strategies for obesity control. A cohort of pregnant women was enrolled in Tibet Plateau and followed till childbirth. Dietary practice during pregnancy was assessed using the Chinese FFQ – Tibet Plateau version, food items and other variables were associated with the risk for obesity of neonates followed by logistic regression, classification and regression trees (CART) and random forest. Of the total 1226 mother–neonate pairs, 40·5 % were Tibetan and 5·4 % of neonates with obesity. Consuming fruits as a protective factor for obesity of neonates with OR (95 % CI) = 0·61 (0·43, 0·87) from logistic regression; as well as OR = 0·20 (0·12, 0·35) for consuming fruits (≥ weekly) from CART. Removing fruit consumption to avoid overshadowing effects of other factors, the following were influential from CART: maternal education (more than middle school, OR = 0·22 (0·13, 0·37)) and consumption of Tibetan food (daily, OR = 3·44 (2·08, 5·69). Obesity among neonates is prevalent in the study population. Promoting healthy diets during pregnancy and strengthening maternal education should be part of the life-course strategies for obesity control.