Machine Learning-Based Predictive Model of Aortic Valve Replacement Modality Selection in Severe Aortic Stenosis...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Severe aortic stenosis (AS) necessitates bioprosthetic valve replacement, with current options including surgical aortic valve replacement (SAVR) and transcatheter aortic valve replacement (TAVR).
• Selecting the optimal valve replacement modality is crucial for patient outcomes.

Purpose of the Study:

• To evaluate a machine learning-based predictive model for selecting between SAVR and TAVR.
• To identify key clinical variables influencing this selection process.

Main Methods:

• Analysis of 415 adult patients with AS undergoing aortic valve replacement.
• Utilized LASSO and decision tree models with 72 clinical variables.
• Assessed model performance using prediction accuracy on a confusion matrix.

Main Results:

• The LASSO model achieved 98% accuracy, identifying frailty score as the most predictive variable.
• Key predictors for SAVR included low frailty score and complex coronary artery disease.
• Key predictors for TAVR included high frailty score, prior coronary artery bypass surgery, calcified aorta, and chronic kidney disease.

Conclusions:

• Machine learning models demonstrate high accuracy (93-98%) in predicting SAVR vs. TAVR selection.
• Frailty score is a critical factor in determining the appropriate aortic valve replacement strategy.