Machine-learning prediction of kidney failure occurrence based on regular health check-up data: a nationwide cohort dataset in South Korea

Objective: While prior research developed numerous machine learning (ML) models for disease prediction and demonstrated their potential, there has yet to be research on models concerning the prediction of kidney failure and its associated risk factors using simple, regular health check-up data. Thus, this study aimed to develop a ML model capable of predicting the risk of kidney failure within five years after receiving regular health check-ups and to identify key contributory factors in the prediction of kidney failure. Methods: We utilized the National Health Insurance Service-National Sample Cohort (NHIS-NSC; n=973,303) to predict kidney failure. The outcome of interest was the risk of kidney failure onset within five years after receiving regular health check-ups. We evaluated the performance of various ML models, including ensemble models, and identified the model with the best predictive capability. Additionally, we conducted a feature importance analysis and ranked the relative significance of each feature. Results: The final dataset used in this study consisted of 1,048,422 cases using multiple records of participants