Machine Learning Prediction for Prognosis of Patients With Aortic Stenosis.

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Aortic valve stenosis (AS) is a serious condition linked to adverse patient outcomes.
• Predicting prognosis in AS is crucial for timely intervention and management.

Purpose of the Study:

• To develop and validate a machine learning-based risk prediction model for aortic stenosis (AS) prognosis.
• To identify key clinical and echocardiographic predictors of AS outcomes.

Main Methods:

• A time-to-event model, random survival forest (RSF), was trained on a large AS registry (10,407 patients).
• Clinical, echocardiographic, laboratory, and medication data were utilized.
• The SHapley Additive exPlanations method was employed for variable importance and personalized risk assessment.
• The model was validated in two independent external cohorts.

Main Results:

• The RSF model demonstrated strong predictive performance with an AUC of 0.83 at 1 and 5 years in the primary cohort.
• External validation showed similar performance (AUCs of 0.73 and 0.74).
• Significant predictors included AS severity, age, serum albumin, pulmonary artery pressure, and chronic kidney disease.

Conclusions:

• A machine learning algorithm effectively predicts outcomes in patients with AS.
• The model provides accurate prognostic insights and identifies critical risk factors.
• This tool can support clinical decision-making and guide risk factor modification strategies to enhance patient prognosis.