Predictive Approach For Liberation From Acute Dialysis In Icu Patients Using Interpretable Machine Learning

A machine-learning algorithm was developed to predict the likelihood of dialysis liberation in critically ill patients with AKI using dynamic parameters from routine data within the first 72 hours of dialysis. The model showed good discrimination in predicting renal recovery and has potential for integration into clinical decision-support systems. Critical features influencing the predictions were identified using SHAP value and PDP.