Predicting and Explaining WASH Inequities in Nigeria: A Machine Learning Approach to Accelerate NTD Elimination

Background

Inadequate water, sanitation, and hygiene (WASH) remain key risk factors for neglected tropical diseases (NTDs) in Nigeria. However, the complex, non-linear determinants of household WASH access are poorly understood, limiting targeted NTD control.

Methods

We conducted a cross-sectional secondary analysis of 30,045 households from the nationally representative 2024 Nigeria Demographic and Health Survey (NDHS). Four supervised machine learning models — Logistic Regression, Decision Tree, Random Forest, and XGBoost — were trained to predict improved household WASH access. Model performance was evaluated using accuracy, precision, recall, F1-score, and AUC. Explainability was achieved using SHAP values and feature importance.

Results

XGBoost achieved the highest performance: 83.3% accuracy, 86.9% precision, 92.3% recall, 89.5% F1-score, and AUC 0.877. Random Forest performed similarly with AUC 0.868. SHAP analysis identified household wealth index, improved sanitation, place of residence, and water availability at handwashing stations as the top predictors. Rural households in northern Nigeria and those in the lowest wealth quintiles had the lowest predicted probability of improved WASH.

Conclusions

Explainable machine learning effectively identifies populations most vulnerable to WASH inequities. Targeting low-wealth, rural households in northern Nigeria with integrated WASH interventions could yield the greatest gains for NTD control. Our open-source code provides a reproducible framework for evidence-based resource allocation.

Author Summary

Neglected tropical diseases thrive where water and sanitation are poor. In Nigeria, millions still lack access to basic WASH, but programs don’t know exactly which households to target first.

Using data from 30,045 households in the 2024 Nigeria Demographic and Health Survey, we used artificial intelligence to predict who is most likely to lack improved WASH. The XGBoost model was 83% accurate and showed that household wealth, type of toilet, rural location, and water at handwashing stations are the biggest factors.

Our results suggest that NTD programs should focus integrated WASH support on low-wealth, rural households, especially in northern Nigeria. By using this open-source AI tool, Nigeria can move from blanket interventions to precision targeting and accelerate progress toward NTD elimination.