Ensemble Machine Learning for Predicting Renal Function Decline in Chronic Kidney Disease: Development and External Validation

Hong Chen^*Yuping HuangLizhen Chen

• the 95th Hospital of Putian in China RongTong Medical Health Corporation, Putian, China

Chronic kidney disease (CKD) is recognized as one of the most significant global health challenges to day. Its effective management relies on timely intervention which is needed when there are indication s of renal function deterioration. Current predictive approaches tend to be inaccurate and superficial. We engineered and verified a machine learning model to predict the risk of renal function deterioratio n using data from 1,200 patients with CKD from different clinics. The study population was carefully chosen through strict and standardized inclusion and exclusion criteria, extensive collection of clinic al information, and precise laboratory evaluation. We developed an ensemble model based on the alg orithms of Random Forest, XGBoost, and Light GBM, employing advanced feature selection and hyp erparameter tuning. The model achieved an AUC of 0.89 (95% CI: 0.87-0.91) which is outstanding, o utperforming traditional Cox models (AUC: 0.82, 95% CI: 0.79-0.85) and average machine learning t echniques (AUC: 0.85, 95% CI: 0.83-0.87). In a SHAP analysis, key predictive factors were eGFR, a ge, and urinary protein-creatinine ratio. The model showed impressive calibration (slope: 0.96, 95% CI: 0.94-0.98) and strong performance across diverse patient subgroups, especially in high-risk popul ations where prediction was strongest. The model was found to be generalizable by internal validation using five-fold cross-validation and external validation in three medical centers. Analysis of resource utilization shows a 60.6% reduction in computational needs relative to traditional methods with unm atched prediction accuracy. Our model serves as an effective solution toward early risk stratification i n CKD management for timely and tailored intervention while optimizing resources. Such a novel ap proach provides a substantial improvement to predictive modelling of kidney function decline and eq uips clinicians with an effective tool for data-driven decision-making. Our code is publicly available at: https://gitee.com/forest-AI/CDK-Model.