Predictive algorithm to stratify newborns at-risk for child undernutrition in India: Secondary analysis of the National Family Health Survey-4

BACKGROUND: India is at the epicentre of global child undernutrition. Strategies to identify at-risk populations are needed in the context of limited resources METHODS: Data from children under the age of five surveyed in the 2015-2016 National Family Health Survey were used. Child undernutrition was assessed using anthropometric measurements. Predictor variables were identified from the extant literature and included if they could be measured at the time of delivery. Survey-weighted logistic regression was applied to model the outcome. Internal validation of the model was performed using 200 bootstrapped samples representing half of the total data sets. RESULTS: In 2016, 54.4% (95% CI = 54.0%-54.8%) of Indian children were undernourished, according to a composite index of anthropometric failure. The predictive model for overall undernutrition included maternal (height, education, reproductive history, number of antenatal visits), child (sex, birthweight), and household characteristics (district of residence, caste, rural residence, toilet availability, presence of a separate kitchen). The model demonstrated reasonable discrimination ability (optimism-adjusted c = 0.67). The group of children classified in the lowest decile for risk of undernutrition had a prevalence of 25.9%, while the group classified in the highest decile had a prevalence of 77.4%. CONCLUSIONS: It is possible to stratify newborns at the time of delivery based on their risk for undernutrition in the first five years of life. The model developed by this study represents a first step in adopting a risk-score based approach for the most vulnerable population to receive services in a timely manner.