Modeling the Determinants of Anemia Among School Children in Mathura District Using Logistic Regression and LASSO Regularization

Abstract

Anemia remains a significant public health concern among school-aged children in India, affecting cognitive development and academic performance. Understanding the determinants of anemia is crucial for designing targeted interventions. 

Objective:

To identify and model the key determinants of anemia among school children (6-14 years) in Mathura District, Uttar Pradesh, using conventional logistic regression and LASSO (Least Absolute Shrinkage and Selection Operator) regularization technique. 

Methods:

A cross-sectional study was conducted among 2000 school children from 28 randomly selected schools in Mathura District. Hemoglobin levels were measured using the HemoCue method, and anemia was classified according to WHO criteria. Data on sociodemographic, dietary, health, and environmental factors were collected through structured questionnaires. Binary logistic regression and LASSO regularization with 10-fold cross-validation were employed to identify significant predictors. Model performance was evaluated using area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, and specificity. 

Results:The overall prevalence of anemia was 77%. Significant predictors of anemia included low socioeconomic status, poor parental education, lack of toilet facility, low dietary diversity, rare intake of green leafy vegetables, irregular Mid-Day Meal participation, non-receipt of iron-folic acid (IFA) supplementation, and no deworming. Higher SES, improved sanitation, iodized salt use, clean fuel, diversified diet, regular Mid-Day Meal, IFA supplementation, and deworming were found protective against anemia. In the LASSO model, a parsimonious set of predictors had moderate discriminative ability with an AUC of approximately 0.67 and sensitivity of greater than 95%.

Conclusions: Anemia among school children in Mathura District is influenced by multiple modifiable factors spanning nutritional, health, and socioeconomic domains. LASSO regularization provided a better model without sacrificing predictive accuracy, which facilitates practical implementation. Multi-sectoral interventions targeting dietary diversification, regular deworming, improved maternal education, and reduction of indoor air pollution are recommended.