Machine Learning for Predicting Heart Failure Progression in Hypertrophic Cardiomyopathy
Key result
Machine learning model predicts progressive heart failure risk in HCM with an AUC of 0.81.
- AUC 0.81
- 95% CI 0.76-0.86
- n=1,427
Why the study?
Advanced heart failure symptoms are the most common adverse pathway in HCM patients, but ability to identify patients at risk remains limited.
Can a machine learning-based model accurately predict the progression to advanced heart failure in patients with hypertrophic cardiomyopathy?
Population
1,427 HCM patients with NYHA class I/II and LVEF >35%
Design
Consecutive cohort study with development, model selection, and independent validation subsets
Follow-up
Mean 4.7 ± 3.7 years
Authors
Discussion
Loading...
Member takes
Implication
May support ML risk stratification in HCM; leaves open prospective validation before clinical use.
Study Design
Type
Observational (n=1,427)
Multicenter
No
Structured PICO
Can a machine learning-based model accurately predict the progression to advanced heart failure in patients with hypertrophic cardiomyopathy?
P
Population
1,427 patients with hypertrophic cardiomyopathy (HCM) evaluated at the Tufts HCM Institute from 2001 to 2018. Key inclusion: New York Heart Association (NYHA) functional class I/II and LV ejection fraction (LVEF) by echocardiography >35%.
I
Intervention
Machine learning (ML)-based risk stratification model using 17 clinical and imaging variables
O
Outcome
Composite of progression to NYHA class III/IV, drop in LVEF below 35%, septal reduction procedure, and/or heart transplantationcomposite
Main Result
Effect estimate: AUC 0.81 (95% CI 0.76-0.86)
A machine learning model utilizing 17 clinical and imaging variables can accurately predict the 5-year risk of progressive heart failure in patients with hypertrophic cardiomyopathy.
Limitations
- Model is designed to accommodate a typical clinical protocol implemented by a single medical center and is not tested using data acquired using different protocols.
- Novel potential risk markers, such as serum biomarkers or mechanical deformation parameters like global longitudinal strain, were not available.
- Not every patient was followed for the full 5-year term.
- Patients who did not develop HF symptoms during the follow-up period were treated as not having the outcome of interest, which could bias the model.
Cite This Study
Fahmy et al. (2021) conducted an observational in Hypertrophic Cardiomyopathy (n=1,427). Machine learning risk stratification model was evaluated on Progression to advanced heart failure (NYHA class III/IV, LVEF <35%, septal reduction procedure, or heart transplantation) (AUC 0.81, 95% CI 0.76-0.86). An ensemble machine learning model using 17 clinical and imaging variables accurately predicted the 5-year risk of progressive heart failure in hypertrophic cardiomyopathy patients with an AUC of 0.81.