Skip to content
Open access

Explainable Machine Learning for Predicting Heavy Metal-Associated Chronic Kidney Disease Risk Using Clinical and Environmental Exposure Biomarkers

Aug 2026 · HEALTH RESEARCH IN AFRICA · Vol 4, pp. 7-16 · 0 citations · 22 references

Abstract

ABSTRACT Background. CKD represents a significant burden on public health globally, with considerable associated illness, death and healthcare costs. A growing body of evidence indicates that heavy metal environmental exposure may lead to renal damage, but these parameters are scarcely considered in predictions. Environmental determinants in addition to clinical risk factors may be identified that facilitate early intervention and prevention of disease. Materials and Methods. This study was based on a sample of 1,845 NHANES 2017-2018 participants. Demographic and socio-economic information as well as biochemical profiles, including urinary heavy metal biomarkers in urine, were studied. The eGFR was used to define CKD status. After preprocessing, median imputation, feature selection, and Synthetic Minority Oversampling Technique (SMOTE) balancing, Random Forest, Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost) models were trained. Model interpretability was examined with SHapley Additive exPlanations (SHAP). Results. Among the evaluated models, XGBoost achieved the highest predictive performance with an accuracy of 85.37%, precision of 65.38%, recall of 48.57%, F1 score of 55.74%, and ROC AUC of 0.7813. SHAP analysis identified serum creatinine, age, serum glucose, gender, and poverty income ratio as the most influential predictors of CKD. Among urinary heavy metals, molybdenum, thallium, manganese, tin, nickel, and cadmium demonstrated the greatest contribution to model predictions. A heavy metal-only validation confirmed that environmental exposure variables contained independent predictive information relevant to CKD risk. Conclusion. Explainable machine learning is able to capture clinically relevant and environmentally mediated risk-factors for chronic kidney disease. The combination of clinical and environmental exposure biomarkers enhances the evaluation of risk for CKD in a manner that is interpretable via SHAP analysis. The proposed GO-based environmental remediation scheme is a new avenue of research and should be substantiated by experimental data prior to environmental application. RESUMEIntroduction. L'insuffisance rénale chronique (IRC) constitue un fardeau majeur de santé publique, qui affecte des centaines de millions de personnes dans le monde. Si les facteurs de risque cliniques (diabète, hypertension) sont bien établis, le rôle des expositions environnementales, notamment aux métaux lourds, reste sous-exploré. Cette étude avait pour but de développer un modèle prédictif de l'IRC intégrant des biomarqueurs urinaires de métaux lourds, et d’identifier les facteurs de risque les plus influents à l'aide de techniques d'apprentissage automatique explicables. Méthodes. Nous avons analysé les données de 1 845 participants de l'enquête NHANES 2017-2018. L'IRC a été définie par un DFG < 60 mL/min/1,73 m² (équation CKD-EPI). Les variables prédictives incluaient les caractéristiques démographiques, socio-économiques, biochimiques et 12 biomarqueurs urinaires de métaux lourds. Après imputation médiane et équilibrage des classes par SMOTE, trois modèles (Random Forest, SVM, XGBoost) ont été entraînés. L'interprétabilité a été assurée par l'analyse SHAP (SHapley Additive exPlanations). Résultats. Parmi les 1 845 participants (âge moyen 50,2 ± 18,6 ans), 351 (19,0 %) présentaient une IRC. Le modèle XGBoost a obtenu les meilleures performances (précision = 85,4 %, AUC-ROC = 0,781). L'analyse SHAP a identifié la créatinine sérique, l'âge, la glycémie, le sexe et le ratio pauvreté/revenu comme prédicteurs majeurs. Parmi les métaux lourds urinaires, le molybdène, le thallium, le manganèse, l'étain, le nickel et le cadmium ont montré les contributions les plus élevées. Un modèle reposant uniquement sur les métaux lourds a atteint une AUC-ROC de 0,624, confirmant leur valeur prédictive indépendante. Conclusion. L'apprentissage automatique explicable identifie efficacement les déterminants cliniques et environnementaux de l'IRC. Les biomarqueurs urinaires de métaux lourds contiennent une information prédictive complémentaire aux facteurs cliniques traditionnels. Ces résultats plaident pour l'intégration de la surveillance environnementale dans l'évaluation du risque rénal, et pour des études longitudinales confirmant le rôle causal de ces expositions.

Read PDF

Similar papers

Open access Aug 2026

Clinical Biomarker-Based Prediction of Chronic Kidney Disease Using Explainable Machine Learning

The results show how a combination of explainable ML and accessible clinical biomarkers can offer a precise, transparent, and clinically interpretable framework for early CKD diagnosis, risk stratification, and informed clinical decision making.

M. Khuntia, Hariballav Mahapatra, N. Lodha · 0 citations
Conference Open access Sep 2026

Predictive Analysis of Chronic Kidney Disease using Explainable Ensemble Learning Models

This work provides an intelligent and comprehensible method of predicting CKD using ML and ensemble methods and identifies six clinically significant items, including haemoglobin, serum creatinine, albumin, high blood pressure, age, and diabetes mellitus.

Unknown authors · 0 citations
#machine learning Review Aug 2026

Population Health-Based Machine Learning Reveals Associations Between Psychosocial Factors and Chronic Kidney Disease

Chronic kidney disease (CKD) progresses silently and severely undermines quality of life, making early detection critical for improving patient outcomes. We present a two-part study that combines large-scale telehealth data with advanced machine learning to both classify self-reported CKD status and identify key driver...

Maryam Shams, D. Eisenberg, Sumaiya Fatema et al. · 0 citations
Open access Aug 2026

Hybrid explainable machine learning for population level cardiovascular disease risk prediction using satellite environmental data

A Hybrid Explainable Artificial Intelligence and Machine Learning (HXAI-ML) framework for area-level CVD risk prediction using health-related and satellite-derived environmental data at the Core-Based Statistical Area (CBSA) level is proposed.

I. Hossain, Kazi Md. Tanvir Anzum, Jarin Tasnim Mumu · 0 citations
Open access Aug 2026

RenalSense: Explainable Prediction of Chronic Kidney Disease.

Chronic kidney disease (CKD) is a progressive disease, and often the symptoms are not detected until later stages, which makes it important to have clear and accurate prediction tools. This study aimed at creating a machine-learning model called RenalSense that explains the model and predicts incident CKD using demogra...

A. U. Green · 0 citations
Sep 2026

Improving 10-year cardiovascular disease risk prediction using automated machine learning.

This study developed an AutoML-based CVD risk prediction model with superior discrimination and calibration, providing clinicians a practical tool for risk stratification, and has the potential to reduce CVD burden at the population level.

Si-Min He, Ju-Ping Wang, Le Zhao et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.