Heart failure & cardiomyopathies
Heart failure risk stratification using artificial intelligence applied to electrocardiogram images: a multinational study
Authors' conclusion
An AI model applied to a single ECG image defined the risk of future HF, representing a digital biomarker for stratifying HF risk.
Study context
- Population
- Individuals without baseline heart failure (HF) in Yale New Haven Health System (YNHHS), UK Biobank and ELSA-Brasil231,285 YNHHS, 42,141 UK Biobank (UKB), 13,454 ELSA-Brasil
- Exposure
- Positive vs negative AI-ECG screen; AI-ECG with PCP-HF vs PCP-HF alone
- Outcome
- First HF hospitalization (incident HF)
Methodological detailsmultinational cohorts (YNHHS, UKB, ELSA-Brasil)
- Multinational cohorts (YNHHS, UKB, ELSA-Brasil)
Main finding
As reported in the abstract
New-onset HF, positive AI-ECG screen
Positive AI-ECG screen vs Negative AI-ECG screenHazard ratio (95% CI)
YNHHS; age-, sex-adjusted
3.88(3.63–4.14)
UKB; age-, sex-adjusted
12.85(6.87–24.02)
ELSA-Brasil; age-, sex-adjusted
23.50(11.09–49.81)
Observational data: associations, not effects.
Further findings
Incident HF events per cohort
- YNHHS: primary HF hospitalizations over 4.5 years (IQR 2.5-6.6)
- 4472
- UKB: developed HF over 3.1 (2.1-4.5) years
- 46
- ELSA-Brasil: developed HF over 4.2 (3.7-4.5) years
- 31
Model discrimination (Harrell's C-statistic)
- YNHHS
- 0.718
- UKB
- 0.769
- ELSA-Brasil
- 0.810
- AI-ECG with PCP-HF significantly improved discrimination over PCP-HF alone in YNHHS and ELSA-Brasil
Original publication
Heart failure risk stratification using artificial intelligence applied to electrocardiogram images: a multinational study. Eur Heart J. Published 13 January 2025. doi:10.
Summarised from the published abstract; every number is checked against it