Using machine learning to predict outcomes following transcarotid artery revascularization.

Area of Science:

• Vascular Surgery
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Transcarotid artery revascularization (TCAR) is a complex procedure with inherent risks.
• Existing risk prediction tools for TCAR have limitations in accuracy.
• Objective risk stratification is crucial for optimizing patient selection and peri-operative management.

Purpose of the Study:

• To develop and validate machine learning (ML) algorithms for predicting 1-year stroke or death after TCAR.
• To assess the performance of ML models using pre-operative, intra-operative, and post-operative data.
• To identify the most effective ML model for TCAR outcome prediction.

Main Methods:

• Utilized the Vascular Quality Initiative (VQI) database (2016-2023) with 38,325 TCAR patients.
• Extracted 115 features across pre-operative, intra-operative, and post-operative phases.
• Trained six ML models, including XGBoost and logistic regression, using a 70/30 train-test split and tenfold cross-validation.

Main Results:

• The XGBoost model demonstrated superior performance, achieving an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.91 for pre-operative prediction.
• ML models showed high predictive accuracy across all stages: pre-operative (AUROC 0.91), intra-operative (AUROC 0.92), and post-operative (AUROC 0.94).
• Logistic regression yielded significantly lower AUROC values, highlighting the advantage of ML approaches.

Conclusions:

• Developed a robust ML algorithm capable of accurately predicting 1-year adverse outcomes following TCAR.
• The ML model shows significant potential for clinical utility in guiding peri-operative risk mitigation strategies.
• This tool can aid clinicians in making informed decisions to prevent complications and improve TCAR patient safety.