Coronary disease & ACSCohort study
Electrocardiogram-based machine learning for risk stratification of patients with suspected acute coronary syndrome
- Prospective observational cohort study
- external validation cohort
Authors' conclusion
The externally validated machine learning-based model, exclusively utilizing features from the 12-lead electrocardiogram, outperformed the HEART score in stratifying the mortality risk of patients with acute chest pain.
Study context
- Population
- Consecutive, non-traumatic patients with chest pain (suspected ACS)external testing cohort N = 3095
- Participants
- 4,015 participants
- Exposure
- ECG feature-based machine learning model vs HEART score
- Outcome
- All-cause mortality
- Follow-up
- Median 3.05 years (IQR 1.75-5.32)
Main finding
As reported in the abstract
All-cause mortality risk stratification
| Group | NPV | Sensitivity |
|---|---|---|
| ML model | 93.4% | 85.9% |
| HEART score | 89.0% | 75.0% |
- Mortality rate (derivation cohort)
- 20.3%
- ML risk groups separated low-, moderate- and high-risk (log-rank test statistic = 121.14)
- P < 0.001
- ML model reduced rate of missed events vs HEART score
- >90%
Observational data: associations, not effects.
Further findings
| Outcome | OR (CI) | |
|---|---|---|
| Mortality by ML risk group, external testing cohort · 30 days — Moderate risk vs Low risk | 3.62 (1.35–9.74) | |
| Mortality by ML risk group, external testing cohort · 30 days — High risk vs Low risk | 6.12 (2.38–15.75) | |
| 0.050.21520Odds ratio (log scale) |
- Mortality by ML risk group, external testing cohort — 30-day mortality, external cohort
- 3.5%
- Mortality by ML risk group, external testing cohort: independent external testing cohort (N = 3095)
Original publication
Electrocardiogram-based machine learning for risk stratification of patients with suspected acute coronary syndrome. Eur Heart J. Published 13 January 2025. doi:10.1093/eurheartj/ehae880
Summarised from the published abstract; every number is checked against it