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

OBJECTIVE

To develop machine learning models that predict the need for continuous renal replacement therapy (CRRT) following acute type A aortic dissection (ATAAD) repair. This approach aims to facilitate early identification of at-risk patients and enable timely intervention.

DESIGN

A retrospective observational cohort study.

SETTING

A large single-center specialty hospital dedicated to cardiovascular diseases.

PARTICIPANTS

A total of 588 ATAAD patients who underwent total arch replacement with frozen elephant trunk between October 2020 and October 2024.

INTERVENTIONS: The dataset was divided into training (70%, n = 412) and validation (30%, n = 176) sets. Lasso regression was applied for feature selection, and seven machine learning models were trained and validated using fivefold cross-validation. Model performance was evaluated based on the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, F1 score, and area under the precision-recall curve (AUPRC). The best-performing model was selected, and SHapley Additive exPlanations analysis was conducted to identify key predictive features.

MEASUREMENTS AND MAIN

RESULTS

Lasso regression identified several key predictors, including peak intraoperative lactate, blood transfusion volume, renal artery involvement, myoglobin, cystatin C, and creatine kinase MB. The XGBoost model demonstrated the best performance (AUC = 0.96, accuracy = 0.96, sensitivity = 0.93, specificity = 0.96, F1 = 0.79, AUPRC = 0.83). SHapley Additive exPlanations analysis revealed that peak intraoperative lactate was the most significant predictor.

CONCLUSION

The XGBoost model effectively predicts the requirement for CRRT after ATAAD surgery, enabling early risk identification and intervention.