Machine Learning in Risk Prediction of Continuous Renal Replacement Therapy After Surgical Repair of Acute Type A...

Area of Science:

• Cardiovascular Surgery
• Nephrology
• Machine Learning in Medicine

Background:

• Acute type A aortic dissection (ATAAD) repair is a complex procedure with a significant risk of acute kidney injury.
• Continuous renal replacement therapy (CRRT) is often required post-operatively, necessitating early identification of at-risk patients.

Purpose of the Study:

• To develop and validate machine learning models for predicting CRRT requirement after ATAAD repair.
• To facilitate timely interventions by identifying patients at high risk for CRRT.

Main Methods:

• Retrospective observational cohort study of 588 ATAAD patients undergoing total arch replacement with frozen elephant trunk.
• Lasso regression for feature selection, followed by training and validation of seven machine learning models using fivefold cross-validation.
• Performance evaluation using AUC, accuracy, sensitivity, specificity, F1 score, and AUPRC; SHapley Additive exPlanations for feature importance.

Main Results:

• Key predictors identified: peak intraoperative lactate, blood transfusion volume, renal artery involvement, myoglobin, cystatin C, and creatine kinase MB.
• The XGBoost model achieved the highest performance (AUC = 0.96, accuracy = 0.96, sensitivity = 0.93, specificity = 0.96, F1 = 0.79, AUPRC = 0.83).
• Peak intraoperative lactate was identified as the most significant predictor via SHapley Additive exPlanations.

Conclusions:

• An XGBoost-based machine learning model effectively predicts CRRT necessity post-ATAAD surgery.
• This predictive model supports early risk stratification and timely clinical intervention for patients.
• The findings highlight the potential of AI in optimizing perioperative care for complex cardiovascular surgeries.