Echocardiography-based machine learning model for atrial fibrillation risk assessment in hypertension.

BACKGROUND

Early identification of atrial fibrillation (AF) allows for timely interventions to reduce cardiovascular complications. Risk scores including C2HEST and CHA2DS2-VASc have limitations in capturing cardiovascular remodeling. Given the association among hypertension, cardiovascular remodeling, and AF, integrating echocardiographic and demographic parameters may improve AF risk assessment in hypertension.

OBJECTIVE

We aimed to develop and validate a machine learning model integrating echocardiographic and demographic parameters to assess AF risk in hypertension.

METHODS

This study included 48,571 hypertensive patients from 8 hospitals. AF was diagnosed based on electrocardiograms or medical records. Patients from The Second Affiliated Hospital of Chongqing Medical University (n = 40,811) were randomly divided into training and internal test sets (7:3), whereas patients from other hospitals (n = 7760) were split into early (2014-2019; n = 3648) and late temporal test sets (2020-2021; n = 4112). 8 machine learning models integrating echocardiographic and demographic parameters were developed and optimized via 10-fold cross-validation in the training set. The best-performing model was selected in the internal test set and further evaluated across all test sets using area under the receiver operating characteristic curve, calibration, and clinical utility.

RESULTS

An extreme gradient boosting-based model achieved an area under the receiver operating characteristic curve of 0.875 in both internal test (95% confidence interval [CI], 0.861-0.888) and early temporal test sets (95% CI, 0.850-0.899) and 0.886 (95% CI, 0.865-0.904) in the late temporal test set, significantly outperforming C2HEST and CHA2DS2-VASc scores while maintaining calibration and clinical utility.

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.