Identifying Optimal level of characteristics of risk group subsets minimizing adverse nutritional outcome in India: A CART based Regression Tree Approach

Child stunting remains a major public health challenge in India, reflecting persistent inequities in nutrition, health, and living conditions. This study applies a Classification and Regression Tree (CART) methodology to identify risk group subsets of children under five years, guided by the Sustainable Development Goal (SDG) principle of “Leave No One Behind (LNOB)” Using nationally representative NFHS-5 data (2019–21, n=208,657), we examine three sets of “circumstances” beyond a child’s control—standard (wealth, maternal education, sex, residence), extended (adding caste and religion), and reduced (excluding wealth)—to uncover intersectional vulnerabilities. CART regression trees reveal that maternal education is the most consistent protective factor, reducing stunting by 10–15 percentage points across contexts. Poverty, caste disadvantage, and rural residence compound risk, with Scheduled Castes, Scheduled Tribes, and Muslims disproportionately affected. At the national level, the Dissimilarity Index (D=0.2747) shows that nearly 27% of stunting inequality is attributable to unequal circumstances, with states such as Bihar, Uttar Pradesh, Gujarat, and Meghalaya recording the highest inequities. By contrast, Kerala, Punjab, and Sikkim show lower risks and more equitable outcomes. These findings demonstrate that stunting in India is shaped not only by economic deprivation but also by structural inequalities, underscoring the need for equity-focused, state-specific interventions.