Using machine learning models to predict post-revascularization thrombosis in PAD.

Area of Science:

• Vascular Surgery
• Cardiovascular Medicine
• Biomedical Engineering

Background:

• Graft/stent thrombosis after lower extremity revascularization (LER) is a serious complication for peripheral arterial disease (PAD) patients, potentially leading to amputation.
• Predicting arterial thrombotic events (ATE) within one year post-LER is crucial for patient outcomes.
• High thrombosis rates necessitate novel predictive strategies for LER procedures.

Purpose of the Study:

• To develop a machine learning model (MLM) for predicting ATE following LER.
• To integrate viscoelastic testing (thromboelastography with platelet mapping - TEG-PM) and patient-specific variables into the MLM.
• To improve the identification of patients at high risk for ATE after LER.

Main Methods:

• Prospective enrollment of PAD patients undergoing LER (2020-2024).
• Collection of demographic, clinical, intervention, and perioperative TEG-PM data.
• Development and evaluation of MLMs (logistic regression, XGBoost, decision tree) using SMOTE for class imbalance and cross-validation.

Main Results:

• 18.3% of 308 patients experienced ATE within one year post-LER.
• The logistic regression MLM integrating TEG-PM and baseline characteristics achieved an AUC of 0.76.
• The best performing MLM showed 70% accuracy, 68% sensitivity, and 71% specificity.

Conclusions:

• Combining patient characteristics with TEG-PM values in MLMs effectively predicts ATE after LER.
• This approach enhances the identification of high-risk PAD patients.
• Tailored thromboprophylaxis strategies can be developed based on these predictions.