•  
  •  
 

Document Type

Original Article

Abstract

Chronic Kidney Disease (CKD) affects medication dose safety due to altered pharmacokinetics. Current clinical decision support systems lack stage-specific personalization and explainability. Therefore, this study aimed to identify the appropriate medications and doses for chronic kidney disease using machine learning algorithms by comparing the methods. This study is divided into two phases. Phase 1 involves a literature review that compares four classification techniques for predicting drug prescriptions for individuals with kidney disease. After a screening and evidence evaluation process, 12 papers were selected for analysis. From the 12 studies, the types of analysis methods obtained were Decision Tree, K-Nearest Neighbour (kNN), Naive Bayes, and deep learning. Phase 2 was to test each method with data in the hospital to identify the accuracy. Novel framework development with explainable AI (SHAP), multi-stage classification (CKD Stages 1-5), and adverse event prediction module. After analysing the four methods, we concluded that they were suitable for testing using hospital data to assess their level of accuracy. However, a study comparing four classification algorithms showed that Decision Tree achieved 99% accuracy, kNN 75%, Naive Bayes 95%, and Deep Learning 96%. Meanwhile, based on AUC, Decision Tree achieved 0.99, kNN achieved 93%, Naive Bayes achieved 97%, and Deep Learning achieved 98%. Decision Tree achieved 99% accuracy (AUC=0.99). Novelty contributions: (1) SHAP-based feature importance identified cholesterol level, blood pressure, and Na-to-potassium ratio as top predictors; (2) Stage-specific recommendations reduced inappropriate prescriptions by 34%; (3) ADE prediction module achieved AUC=0.87 for nephrotoxicity risk. Our explainable, stage-aware framework provides clinically actionable drug recommendations with integrated safety monitoring, addressing critical gaps in CKD pharmacotherapy

Receive Date

30 Nov 2025

Revise Date

14 June 2026

Accept Date

18 June 2026

Publication Date

2026

Download

Share

COinS