Deep Learning Model Predicts Atrial Fibrillation From Outpatient ECGs

In HealthDay News
by Healthday

Findings seen for atrial fibrillation in diverse populations

By Lori Solomon HealthDay Reporter

TUESDAY, Oct. 24, 2023 (HealthDay News) — Deep learning of outpatient sinus rhythm electrocardiograms (ECGs) predicts atrial fibrillation (AF) within 31 days, according to a study published online Oct. 18 in JAMA Cardiology.

Neal Yuan, M.D., from University of California in San Francisco, and colleagues assessed whether deep learning models applied to outpatient ECGs in sinus rhythm can predict AF. Outpatient ECGs were performed during 1987 through 2022 at six U.S. Veterans Affairs hospital networks (907,858) and one large non-VA academic medical center (72,483).

The researchers found that a deep learning model predicted the presence of AF within 31 days of a sinus rhythm ECG on held-out test ECGs at VA sites with an area under the receiver operating characteristic curve (AUROC) of 0.86, accuracy of 0.78, and F1 score of 0.30. At the non-VA site, AUROC was 0.93, accuracy was 0.87, and F1 score was 0.46. The model was well calibrated (Brier score of 0.02 across all sites). For individuals identified as high-risk by deep learning, the number needed to screen to detect a positive case of AF was 2.47 individuals, yielding a testing sensitivity of 25 percent, and 11.48 for 75 percent. For patients who were Black, female, younger than 65 years, and those who had CHA2DS2-VASc scores of ≥2, model performance was similar.

“These findings suggest that deep learning applied to ECGs could help identify patients at high risk of AF who could be considered for intensive monitoring programs to help prevent adverse cardiac events,” the authors write.

One author disclosed ties to industry.

Abstract/Full Text (subscription or payment may be required)

Editorial (subscription or payment may be required)

Copyright © 2023 HealthDay. All rights reserved.