Arrhythmias & electrophysiologyDiagnostic or prognostic study
Artificial Intelligence–Enhanced Electrocardiography for Complete Heart Block Risk Stratification
- Prognostic cohort study (development and external validation)
Authors' conclusion
In this study, a first-of-its-kind deep learning model identified the risk of incident CHB.
Study context
- Population
- Patients with electrocardiography (ECGs) at Beth Israel Deaconess Medical Center; external validation in UK Biobank volunteer cohort
- Participants
- 189,539 patients · 1,163,401 ECGs
- Comparison
- AI-ECG risk estimator vs bifascicular block for predicting incident complete heart block (CHB)
- Outcome
- New diagnosis of complete heart block (CHB) >31 days after the ECG
Performance
Incident CHB (development cohort)
Adjusted hazard ratio
11.6
95% CI 7.62–17.7
P < 0.001
By group
AIRE-CHB
0.836 (95% CI, 0.819-0.534)
C index · AUROC for incident CHB within 1 year 0.889 (95% CI, 0.863-0.916)
Bifascicular block
0.594 (95% CI, 0.567-0.620)
AUROC
High-risk quartile vs Low-risk group
aHR 11.6 (95% CI 7.62–17.7); P < 0.001
As reported in the abstract
Further findings
External validation, UK Biobank cohort (50 641 ECGs): incident CHB prediction
- C index for incident CHB prediction
- 0.936 (95% CI, 0.900-0.972)
- aHR for incident CHB
- 7.17 (95% CI, 1.67-30.81; P < .001)
Original publication
Artificial Intelligence–Enhanced Electrocardiography for Complete Heart Block Risk Stratification. JAMA Cardiol. Published 1 November 2025. doi:10.1001/jamacardio.2025.2522
Summarised from the published abstract; every number is checked against it