Predicting mortality and hospitalization in heart failure using machine learning: A systematic literature review

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Machine learning (ML) offers potential to enhance clinical decision-making and patient outcomes in heart failure management.
• Despite potential benefits, ML applications in cardiology, particularly for heart failure, are underutilized in clinical practice.

Purpose of the Study:

• To systematically review factors hindering the clinical integration of ML-derived risk scores for adult patients with acute and chronic heart failure.

Main Methods:

• A systematic search of four academic databases and Google Scholar was conducted.
• Studies focused on developing predictive models for all-cause mortality, cardiac death, and hospitalizations using heart failure patient data.

Main Results:

• Thirty studies were included, with sample sizes ranging from 71 to 716,790 patients.
• Model performance varied, with Area Under the Curve (AUC) for mortality prediction ranging from 0.48 to 0.92.
• Logistic regression, random forests, and decision trees were common modeling techniques; no models used data from Africa or the Middle East.

Conclusions:

• Widespread adoption of ML risk calculators is limited by diverse heart failure etiologies, poor access to structured health data, and clinician skepticism.
• Modest accuracy of current predictive models and lack of external validation also impede clinical integration.