A Risk Prediction Model (CMC-AKIX) for Postoperative Acute Kidney Injury Using Machine Learning: Algorithm...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Nephrology Research

Background:

• Postoperative acute kidney injury (AKI) is a major complication of general anesthesia, increasing mortality and morbidity.
• Existing AKI prediction models often lack generalizability and require external validation.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting postoperative AKI risk.
• To identify robust preoperative predictors for AKI development.

Main Methods:

• Retrospective cohort analysis of 239,267 noncardiac surgeries (2009-2019) across seven South Korean university hospitals.
• Evaluation of six machine learning models: deep neural network, logistic regression, decision tree, random forest, light gradient boosting machine, and naïve Bayes.
• Performance assessment using AUC, accuracy, precision, sensitivity, specificity, and F1-score.

Main Results:

• Postoperative AKI occurred in 7,935 cases (3.3%).
• Deep neural network (AUC=0.832), light gradient boosting machine (AUC=0.836), and logistic regression (AUC=0.825) showed superior predictive performance.
• A user-friendly website was developed based on the deep neural network model.

Conclusions:

• A robust, high-performance AKI risk prediction system using preoperative data is feasible for clinical application.
• The developed model and its web integration improve clinical utility for personalized patient care and risk management.