Predicting Cardiovascular Rehabilitation of Patients with Coronary Artery Disease Using Transfer Feature Learning

Area of Science:

• Cardiology
• Machine Learning
• Health Informatics

Background:

• Cardiovascular diseases are a leading global cause of mortality.
• Cardiovascular rehabilitation programs are essential for patient recovery, particularly for those with coronary artery disease.
• The COVID-19 pandemic highlighted the need for flexible remote and hybrid rehabilitation models to overcome geographical and time barriers.

Purpose of the Study:

• To develop a predictive model for cardiovascular rehabilitation outcomes using stacked machine learning and transfer learning.
• To address challenges in remote and hybrid cardiovascular rehabilitation programs by identifying patients requiring additional support.
• To improve patient retention and success rates in cardiovascular rehabilitation.

Main Methods:

• Utilized retrospective and prospective data with diverse features.
• Employed stacked machine learning, transfer feature learning, and joint distribution adaptation.
• Validated the model using 10-fold cross-validation at a Chilean rehabilitation center.

Main Results:

• Achieved a Normalized Mean Squared Error (NMSE) of 0.03±0.013 and R2 of 63±19% in 10-fold cross-validation.
• The best performance reached an NMSE of 0.008 and an R2 of 92%.
• Demonstrated the model's potential to predict patient success in rehabilitation settings.

Conclusions:

• The developed machine learning model shows promising results for predicting cardiovascular rehabilitation outcomes.
• These predictive models can aid in prioritizing remote patients who need enhanced support for successful rehabilitation.
• The findings encourage the adoption and refinement of remote and hybrid cardiovascular rehabilitation programs.