Advanced machine learning approaches for predicting Neglected Tropical Disease co-endemicity in Kenya: A focus on soil-transmitted helminths, schistosomiasis, and lymphatic filariasis

Comprehensive Summary

This study, presented by Nyere and Mulwa, investigates the use of machine learning models to predict and map the co-endemicity of neglected tropical diseases (NTDs)—specifically soil-transmitted helminths (STH), schistosomiasis (SCH), and lymphatic filariasis (LF)—in Kenya. The researchers conducted a secondary data analysis using 2022 nationwide data from the Expanded Special Project for Elimination of Neglected Tropical Diseases (ESPEN), incorporating demographic and environmental variables such as water, sanitation, and hygiene (WASH) indicators. Three machine learning models—Random Forest (RF), Gradient Boosting Machine (GBM), and XGBoost—were trained and evaluated using metrics such as area under the curve (AUC), accuracy, sensitivity, and variable importance. The results show that the Random Forest model achieved the best predictive performance (AUC ≈ 0.70), outperforming the other models. Key predictors of disease co-endemicity included sanitation access, population density, and overlap with other infections. Spatial analysis identified Eastern and North-Eastern Kenya as persistent high-risk hotspots, strongly associated with low WASH coverage. In the discussion, the authors emphasize that machine learning provides valuable tools for identifying disease hotspots and guiding targeted interventions, though model performance was moderate and limited by available data and variable scope.

Outcomes and Implications

Neglected tropical diseases affect over a billion people globally and disproportionately burden low-resource regions, where limited infrastructure and surveillance systems hinder effective control. Identifying high-risk areas is critical for optimizing interventions such as mass drug administration and improving public health outcomes. Clinically and from a public health perspective, the findings suggest that machine learning–based risk prediction tools can enhance disease surveillance by enabling more precise targeting of interventions to high-burden regions. The identification of WASH-related factors as major predictors reinforces the importance of integrating infrastructure improvements—such as access to clean water and sanitation—with traditional treatment programs. These tools could support policymakers in allocating resources more efficiently and implementing preventive strategies tailored to local risk profiles. However, the study highlights that model outputs should be used as decision-support tools rather than definitive predictors, given moderate accuracy and potential misclassification. The authors suggest that near-term implementation is feasible through real-time dashboards for dynamic risk mapping, with future improvements involving more comprehensive datasets and integration of environmental and climate variables to strengthen predictive performance and guide long-term disease elimination strategies.