Arrhythmias & electrophysiologyDiagnostic or prognostic study
Echocardiography-based machine learning model for atrial fibrillation risk assessment in hypertension
- development and validation of a machine learning model
- 8 hospitals
- 10-fold cross-validation
- internal and temporal test sets
Authors' conclusion
An extreme gradient boosting-based model integrating echocardiographic and demographic parameters outperformed existing scores, offering a reliable tool for AF risk assessment in hypertension.
Study context
- Population
- Hypertensive patients from 8 hospitalshypertensive patients from 8 hospitals · The Second Affiliated Hospital of Chongqing Medical University (training and internal test sets, 7:3) 40,811 · Other hospitals 7,760 · Early temporal test set (2014-2019) 3,648 · Late temporal test set (2020-2021) 4,112
- Participants
- 48,571 patients
- Comparison
- Extreme gradient boosting-based model vs C2HEST and CHA2DS2-VASc scores
Performance
As reported in the abstract
Atrial fibrillation (AF) risk discrimination (area under the receiver operating characteristic curve)
| Test or model | Internal test set AUC | Early temporal test set AUC | Late temporal test set AUC |
|---|---|---|---|
| Extreme gradient boosting-based model | 0.875 (95% CI, 0.861-0.888) | 0.875 (95% CI, 0.850-0.899) | 0.886 (95% CI, 0.865-0.904) |
Original publication
Echocardiography-based machine learning model for atrial fibrillation risk assessment in hypertension. Heart Rhythm. Published 25 May 2026. doi:10.1016/j.hrthm.2026.05.036
Summarised from the published abstract; every number is checked against it