Perioperative medicine & outcomes
Preoperative risk prediction of major cardiovascular events in noncardiac surgery using the 12-lead electrocardiogram: an explainable deep learning approach
Authors' conclusion
A multimodal deep-learning model that integrates preoperative ECG waveforms with routinely collected clinical data improves prediction of major adverse cardiovascular events, compared with the Revised Cardiac Risk Index.
Study context
- Population
- Adults undergoing major noncardiac surgery, MIMIC-IV database, 2008–2019
- Participants
- 37,081 adults
- Exposure
- Fusion model vs ECG-only models vs Revised Cardiac Risk Index (RCRI)
- Outcome
- Discrimination (AUROC) for in-hospital MI, in-hospital mortality, and composite of in-hospital MI, in-hospital stroke and 30-day mortality
Methodological detailsdeep learning prediction model, 10-fold cross-validation, explainable (generative counterfactual framework)
- Deep learning prediction model
- 10-fold cross-validation
- Explainable (generative counterfactual framework)
Main finding
As reported in the abstract
AUROC of fusion model
| Group | AUROC, MI (95% CI 0.845-0.872) | AUROC, in-hospital mortality (95% CI 0.889-0.908) | AUROC, composite outcome (95% CI 0.827-0.843) |
|---|---|---|---|
| Fusion model | 0.858 | 0.899 | 0.835 |
Fusion model
AUROC, MI (95% CI 0.845-0.872)0.858
AUROC, in-hospital mortality (95% CI 0.889-0.908)0.899
AUROC, composite outcome (95% CI 0.827-0.843)0.835
- Fusion model vs ECG-only: P ≤ 0.002 — fusion AUROC values exceeded ECG-only
- Fusion model vs RCRI: P = 0.001 — fusion exceeded RCRI
- Fusion model vs RCRI: P < 0.001 — fusion exceeded RCRI
Observational data: associations, not effects.
Further findings
Counterfactual waveform explanations
- Prolonged QRS duration, low-voltage complexes and ST-segment depression consistently correlated with higher predicted risk
Original publication
Preoperative risk prediction of major cardiovascular events in noncardiac surgery using the 12-lead electrocardiogram: an explainable deep learning approach. Br J Anaesth. Published 17 September 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →