A Machine Learning-Based Prediction Model for Diabetic Kidney Disease in Korean Patients with Type 2 Diabetes...

Area of Science:

• Nephrology
• Endocrinology
• Data Science

Background:

• Diabetic kidney disease (DKD) is a significant complication of type 2 diabetes mellitus (T2DM), leading to end-stage kidney disease.
• Predictive models for DKD onset in Korean T2DM patients are limited, necessitating further research.
• Early identification of DKD risk is crucial for timely intervention and management.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based model for predicting DKD in Korean patients with T2DM.
• To identify key clinical and laboratory factors associated with DKD development.
• To assess the generalizability and clinical utility of the developed ML model.

Main Methods:

• Retrospective analysis of electronic health records from six Korean hospitals.
• Development and internal validation using the Jeonbuk National University Hospital cohort (8:2 train:test split).
• External validation using data from five additional hospitals.
• Comparison of multiple ML algorithms including XGBoost, random forest, and logistic regression.
• Inclusion of demographic data, comorbidities, medications, and laboratory results.

Main Results:

• Out of 5120 T2DM patients, 1361 (26.6%) developed DKD.
• XGBoost demonstrated the highest predictive performance in the development cohort (AUC: 0.8099).
• External validation confirmed robust performance for XGBoost (AUC: 0.8113) and logistic regression models (AUCs: 0.8228-0.8271).
• Key predictors included age, baseline estimated glomerular filtration rate, creatinine, hemoglobin, and HbA1c levels.

Conclusions:

• ML-based approaches, particularly XGBoost and logistic regression, show significant potential for DKD prediction in Korean T2DM patients.
• The validated model offers a valuable tool for early DKD risk assessment.
• The model's generalizability across multiple institutions supports its clinical applicability.