Machine learning-based risk stratification for mortality in patients with severe aortic regurgitation

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Current guidelines for severe aortic regurgitation (AR) recommend intervention based on symptoms or ventricular dysfunction.
• Emerging evidence suggests current guidelines may delay intervention, potentially missing optimal treatment windows.
• Early intervention in AR is crucial for improving patient prognosis and outcomes.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting mortality risk in patients with severe AR.
• To identify high-risk patients who may benefit from early intervention, irrespective of aortic valve replacement (AVR) status.
• To assess the utility of ML algorithms in optimizing AR patient management.

Main Methods:

• Trained ML models, including a conditional random survival forest, on a dataset of 1035 patients with severe AR.
• Validated model performance on an independent dataset of 207 patients.
• Utilized five-fold cross-validation and selected 19 key variables (e.g., age, ejection fraction, NYHA class) for the final predictive model.

Main Results:

• The optimal ML model demonstrated strong predictive performance for survival.
• Concordance indices for survival prediction were 0.84 at 1 year, 0.86 at 2 years, and 0.87 overall.
• Key predictors included age, body mass index, blood pressure, NYHA class, AVR status, comorbidities, and echocardiographic parameters.

Conclusions:

• Machine learning models effectively predict survival in severe AR patients using common clinical and echocardiographic data.
• This ML-based approach can identify high-risk individuals who may benefit from earlier intervention.
• Implementing ML tools could lead to improved patient outcomes in severe AR management.