Abstract
Chronic kidney disease of unknown etiology (CKDu) has emerged as an important public health challenge, particularly in agricultural communities across Southern Asia and Central America. Our research aims to explore the role of environmental factors in contributing to CKDu prevalence in these regions. Using an Extreme Gradient Boosting Machine Learning (XGBoost) model, we analyzed an environmental dataset from the CKDu endemic region of Sri Lanka. The XGBoost model achieved 85% accuracy in predicting CKDu prevalence in a total of 100 locales. Significant predictor variables included fluoride concentration in water, electrical conductivity of drinking water (EC), pH, and soil type. Soil type was the most influential factor, followed by pH and EC, which influence the solubility and bioavailability of nephrotoxic chemicals in water sources, with fluoride concentration as an additional contributing variable. The study findings highlight the need for targeted water analysis programs and interventions in water quality management, agrochemical usage, and soil treatment in CKDu-endemic regions. These insights also provide a framework for future research to identify causative agents and develop strategies for reducing CKDu prevalence.