Coagulation Risk Predicting in Anticoagulant-Free Continuous Renal Replacement Therapy

Area of Science:

• Nephrology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Continuous renal replacement therapy (CRRT) is vital for acute kidney injury.
• Anticoagulant-free CRRT is necessary for bleeding-risk patients but increases circuit coagulation.
• Predicting and preventing CRRT circuit coagulation is crucial for treatment efficacy and resource management.

Purpose of the Study:

• To develop and validate an AI-driven machine learning model for predicting circuit coagulation risk during anticoagulant-free CRRT.
• To identify key clinical metrics that signify a high risk of CRRT circuit coagulation.
• To facilitate personalized treatment strategies for patients undergoing anticoagulant-free CRRT.

Main Methods:

• Retrospective analysis of 212 patients undergoing anticoagulant-free CRRT.
• Development and validation of eight machine learning models to predict circuit coagulation within 24 hours.
• Performance evaluation using Area Under the Curve (AUC) and 5-fold cross-validation; feature importance analysis using SHAP plots.

Main Results:

• Machine learning models demonstrated excellent predictive performance, with ensemble learning achieving an AUC of 0.863.
• Random forest was the best single-algorithm model with an AUC of 0.819.
• Key predictors of circuit coagulation included platelet count, filtration fraction (FF), and triglycerides.

Conclusions:

• An AI-based model effectively predicts circuit coagulation risk in anticoagulant-free CRRT with high accuracy (AUC 0.863).
• Platelets, filtration fraction, and triglycerides are significant indicators of coagulation risk.
• The model supports personalized treatment strategies to mitigate CRRT circuit coagulation, improving patient outcomes and reducing costs.