Individual-Level Double Burden of Malnutrition among Ethiopian Women: Prevalence, Machine-Learning Prediction and Explainability Using the 2024/25 Ethiopian Demographic and Health Survey

ABSTRACT Background: The double burden of malnutrition, combining overweight or obesity with anemia is vital in nutritional transitions. Evidence regarding its national distribution and prediction using routine survey data in Ethiopia remains limited. Objective: To estimate the national prevalence of individual-level double burden of malnutrition among non-pregnant Ethiopian women aged 15-49 years and evaluate machine-learning models using data from the 2024/25 Ethiopian Demographic and Health Survey. Methods: we analyzed 14250 non-pregnant women. Double burden of malnutrition was defined as a body mass index equal to or greater than 25.0 kilograms per square meter with an altitude- and smoking-adjusted hemoglobin concentration below 12.0 grams per deciliter. Data were split into an 80% training set and a 20% holdout test set. Six algorithms were evaluated using cross-validation, while SHapley Additive exPlanations described feature contributions. Results: the weighted national double burden of malnutrition prevalence was 8.4% (95% confidence interval 7.9%-8.9%), higher in urban residents (20.0% versus 4.0%) and the richest wealth quintile (26.7% versus 2.7%). Extreme Gradient Boosting achieved an area under the receiver operating characteristic curve of 0.892 (95% confidence interval 0.875-0.909) and an area under the precision-recall curve of 0.764 outperforming logistic regression (area under the receiver operating characteristic curve of 0.781). Extreme Gradient Boosting sensitivity was 82.1%, specificity was 81.4% and the Brier score was 0.068. Household wealth, residence and respondent age were top predictors. Conclusions; the double burden of malnutrition affected one in twelve non-pregnant Ethiopian women unevenly across groups. Routine characteristics effectively discriminated double burden of malnutrition cases though external validation is needed prior to clinical use.