Anaemia Burden and Women’s Empowerment in India: A Multi-Modal Socioeconomic Analysis With Advanced Machine Learning Predictive Modeling

Anaemia remains a critical public health challenge in India, affecting more than half of women of reproductive age. While biological determinants are well-documented, the relationship between socioeconomic empowerment indicators and anaemia burden across Indian states remains incompletely characterized.

This cross-sectional study analyzed NFHS-5 (2019–21) data from 29 Indian states and union territories, examining four anaemia outcome variables against five women’s empowerment indicators. Six analytical approaches were employed: descriptive statistics, principal component analysis (PCA), Pearson correlation, signal complexity analysis (Shannon and approximate entropy), scatter matrix visualization, and a multi-algorithm machine learning pipeline comprising six classifiers — including novel application of XGBoost and Gradient Boosting — with Leave-One-Out and stratified 5-fold cross-validation.

The mean anaemia prevalence was 52.85% (SD: 11.68%) among women aged 15–49. PCA identified a dominant socio-developmental gradient explaining 58.52% of total variance (PC1), along which states with higher women’s literacy (r = –0.548, p = 0.002), internet access (r = –0.590, p = 0.001), and years of schooling (r = –0.497, p = 0.006) demonstrated significantly lower anaemia prevalence. In socioeconomic-only machine learning models, XGBoost achieved the best performance (5-fold CV accuracy: 78.7%, AUC-ROC: 0.911), followed by Random Forest (CV: 75.3%, AUC-ROC: 0.978), confirming that women’s empowerment indicators alone carry substantial predictive power. Leave-One-Out validation confirmed Random Forest as the most stable classifier (LOO accuracy: 75.9%).

This study establishes a robust socio-developmental gradient that structures the anaemia burden across India. Internet access and literacy emerge as the strongest modifiable socioeconomic predictors. Asset ownership and financial inclusion show minimal independent association. These findings support investment in women’s digital inclusion and education as anaemia prevention strategies, alongside targeted interventions in high-burden eastern and central Indian states.