Heart failure & cardiomyopathies
Characterizing advanced heart failure risk and hemodynamic phenotypes using interpretable machine learning
Authors' conclusion
Novel interpretable ML models predicted risk categories with a high degree of accuracy.
Study context
- Population
- Advanced heart failure with reduced ejection fraction (HFrEF); trained on ESCAPE trial, validated in 4 advanced heart failure (HF) cohorts from previous trials
- Comparison
- Invasive hemodynamics feature set vs all features feature set
- Outcome
- Prediction of risk category (1–5) based on composite of death, left ventricular assist device (LVAD) implantation or transplantation (DeLvTx) and rehospitalization; AUC per category
- Follow-up
- 6 months
Methodological detailsinterpretable machine learning, unsupervised clustering, external validation in 4 cohorts
- Interpretable machine learning
- Unsupervised clustering
- External validation in 4 cohorts
Performance
As reported in the abstract
Accuracy of 5 separate models predicting risk category
| Test or model | Accuracy range |
|---|---|
| Invasive hemodynamics | 0.896 +/- 0.074 to 0.969 +/- 0.081 |
| All features | 0.858 +/- 0.067 to 0.997 +/- 0.070 |
Accuracy range
Invasive hemodynamics0.896 +/- 0.074 to 0.969 +/- 0.081
All features0.858 +/- 0.067 to 0.997 +/- 0.070
Original publication
Characterizing advanced heart failure risk and hemodynamic phenotypes using interpretable machine learning. Am Heart J. Published 7 February 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →