Why the study?
Systems-level barriers limit access to advanced heart failure therapies, creating a need for automated systems to help clinicians evaluate patients at the appropriate time.
Does an interpretable machine learning model accurately identify potential candidates for advanced heart failure therapies in patients with heart failure?
Population
4,694 HF patients from the REVIVAL and INTERMACS registries
Comparison
Novel machine learning model vs other commonly used machine learning models
Design
Retrospective registry-based model development and validation study
Key result
A novel interpretable machine learning model identified patients needing advanced heart failure therapies with an F1 score of 43.8%, recall of 51.1%, and precision of 46.9%.
Authors
Loading...
Offers a proof-of-concept tool for predicting advanced HF therapy needs; leaves open whether prospective validation.
Observational (n=4,694)
Yes
Does an interpretable machine learning model accurately identify potential candidates for advanced heart failure therapies in patients with heart failure?
A novel, interpretable machine learning model can identify heart failure patients who may need evaluation for advanced therapies while providing transparent clinical rules for clinicians.
Yao et al. (2022) conducted an observational in Heart failure (n=4,694). Interpretable machine learning model based on tropical geometry and fuzzy logic vs. Other commonly used machine learning models was evaluated on F1 score for identifying need for advanced therapies. A novel interpretable machine learning model identified patients needing advanced heart failure therapies with an F1 score of 43.8%, recall of 51.1%, and precision of 46.9%.