Predicting End-Stage Renal Disease and Mortality in Chronic Kidney Disease Using Machine Learning: Retrospective...

Area of Science:

• Nephrology
• Medical Informatics
• Machine Learning

Background:

• Chronic kidney disease (CKD) presents a significant global health challenge with varied progression rates.
• Early identification of high-risk CKD patients for end-stage renal disease (ESRD) or mortality is crucial for timely interventions.
• Effective management strategies are essential to mitigate morbidity, mortality, and improve quality of life in CKD patients.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting ESRD and all-cause mortality in CKD patients.
• To assess the predictive performance of various ML algorithms using routinely collected clinical data.
• To evaluate the clinical utility of ML models in stratifying CKD patient risk.

Main Methods:

• Utilized data from 29,677 CKD patients (eGFR <60 mL/min/1.73 m2) treated between 2011-2021.
• Developed and validated ML models (e.g., XGBoost, LightGBM, Random Forest, stacking classifier) using 69 variables.
• Assessed model performance via AUROC, precision-recall curves, calibration, and decision-curve analysis.

Main Results:

• ML models showed high predictive performance for ESRD (AUROCs 0.839-0.894) and modest performance for all-cause mortality (AUROCs 0.752-0.774).
• Precision-recall curves confirmed model utility, especially for ESRD prediction in this cohort.
• Calibration and decision-curve analyses supported the reliability and clinical applicability of the developed models.

Conclusions:

• Machine learning algorithms show promise as effective tools for risk stratification in CKD patients.
• These models can support individualized clinical management strategies for ESRD and mortality risk.
• ML-driven risk assessment can potentially improve patient outcomes in chronic kidney disease.