Interpretable machine learning models for stroke risk prediction in patients with newly diagnosed atrial fibrillation - PubMed
4 hours ago
- #stroke prediction
- #atrial fibrillation
- #machine learning
- Developed interpretable machine learning models (logistic regression and XGB) for 1-year stroke risk prediction in newly diagnosed atrial fibrillation patients.
- Models used age, comorbidities, and medication data, achieving higher AUCs (0.877-0.915) than CHA₂DS₂-VASc (0.614-0.621) in internal and external validation.
- Models showed strong calibration and clinical utility, with superior long-term risk stratification and treatment responsiveness to guide DOAC initiation.