Key result
Support Vector Machine models outperform CHA2DS2-VASc predicting ischemic stroke in non-AF HFrEF with ~0.87 AUC.
Why the study?
Predicting ischaemic stroke in patients with HFrEF without AF remains challenging, prompting evaluation of machine learning models to identify incident stroke.
Do machine learning models improve the prediction of incident ischaemic stroke compared to the CHA2DS2-VASc score in patients with HFrEF without AF?
Population
2213 patients with HFrEF without AF from the WARCEF trial
Comparison
9 machine learning models vs CHA2DS2-VASc score
Design
Post-hoc analysis of a randomized trial
Follow-up
Mean 3.3 ± 1.8 years
Authors
Loading...
May aid stroke risk stratification in HFrEF without AF; hypothesis-generating and requires prospective validation before clinical use.
Observational (n=2,213)
Do machine learning models improve the prediction of incident ischaemic stroke compared to the CHA2DS2-VASc score in patients with HFrEF without AF?
Effect estimate: AUC 0.874 (95% CI 0.769-0.959)
Machine learning models, particularly SVM and XGBoost, significantly outperform the CHA2DS2-VASc score in predicting incident ischaemic stroke in patients with HFrEF without atrial fibrillation.
Ishiguchi et al. (2024) conducted an observational in Heart failure with reduced ejection fraction (HFrEF) without atrial fibrillation (n=2,213). Machine learning models (Support Vector Machine and XGBoost) vs. CHA2DS2-VASc score was evaluated on Incident ischaemic stroke (AUC 0.874, 95% CI 0.769-0.959). Support Vector Machine models (AUC 0.874; 95% CI 0.769-0.959) outperformed the CHA2DS2-VASc score (AUC 0.643) in predicting incident ischaemic stroke in HFrEF patients without atrial fibrillation.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: