Predicting outcomes following lower extremity open revascularization using machine learning.

Area of Science:

• Vascular Surgery
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Peripheral artery disease necessitates lower extremity open revascularization, a procedure with substantial peri-operative risks.
• Current tools for predicting outcomes after these procedures are limited, hindering effective risk stratification.

Purpose of the Study:

• To develop and validate machine learning algorithms for predicting 30-day adverse outcomes following lower extremity open revascularization.
• To identify key pre-operative variables influencing patient outcomes.

Main Methods:

• Utilized the National Surgical Quality Improvement Program vascular database (2011-2021) for patient identification.
• Trained six machine learning models using 37 pre-operative variables, with data split into training (70%) and testing (30%) sets.
• Employed tenfold cross-validation to evaluate model performance, focusing on 30-day major adverse limb events (MALE) or death.

Main Results:

• Included 24,309 patients; 9.3% experienced 30-day MALE or death.
• The XGBoost model demonstrated superior performance with an area under the receiver operating characteristic curve of 0.93 (0.92-0.94).
• The model exhibited good calibration, with a Brier score of 0.08, indicating strong agreement between predicted and observed probabilities.

Conclusions:

• Developed a high-performing machine learning algorithm for predicting 30-day MALE or death after lower extremity open revascularization.
• The algorithm shows significant potential for clinical utility in guiding risk mitigation strategies.
• Implementation could lead to improved patient outcomes in vascular surgery.