Predicting Outcomes Following Carotid Artery Stenting Using Machine Learning.

Area of Science:

• Vascular Surgery
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Carotid artery stenting (CAS) involves significant perioperative risks.
• Current outcome prediction tools for CAS are limited in their effectiveness.
• There is a need for improved methods to assess and manage risks associated with CAS.

Purpose of the Study:

• To develop and evaluate machine learning (ML) algorithms for predicting 30-day outcomes after transfemoral CAS.
• To compare the performance of ML models against traditional methods like logistic regression.
• To identify key preoperative predictors of adverse events following CAS.

Main Methods:

• Utilized the National Surgical Quality Improvement Program (NSQIP) targeted vascular database (2011-2021).
• Trained six ML models and a logistic regression comparator using 36 preoperative variables.
• Evaluated models based on area under the receiver operating characteristic curve (AUROC), calibration plots, and Brier scores.

Main Results:

• The XGBoost ML model achieved a high AUROC of 0.93 for predicting 30-day major adverse cardiovascular events (MACE).
• This performance significantly surpassed logistic regression (AUROC 0.67) and existing literature tools (AUROC 0.58-0.74).
• Key predictors included symptomatic carotid stenosis, age, and American Society of Anesthesiologists classification; models showed robust performance across subgroups.

Conclusions:

• Developed ML models accurately predict 30-day outcomes after transfemoral CAS using preoperative data.
• These models demonstrate superior performance compared to logistic regression and current prediction tools.
• The ML algorithms offer significant potential for guiding risk-mitigation strategies and improving patient outcomes in CAS procedures.