Perioperative medicine & outcomes
Machine learning to predict myocardial injury and death after non‐cardiac surgery
Study context
- Population
- Non-cardiac surgery patients in the Vascular Events in Non-cardiac Surgery Patients Cohort Evaluation study
- Participants
- 24,589 patients
- Comparison
- Single-layer vs multiple-layer neural networks
- Outcome
- Discrimination (AUC) for myocardial injury and death
- Follow-up
- 30 postoperative days
Methodological detailssecondary analysis of a cohort study, random-split validation, Validation on a randomly selected subset of the study population
- Secondary analysis of a cohort study
- Random-split validation
- Validation on a randomly selected subset of the study population
Performance
As reported in the abstract
Myocardial injury, AUC · 30 days
AUC, pre-referral variables
0.71 vs 0.70
95% CI 0.70–0.73 vs 0.69–0.72P < 0.001
Multiple-layer
AUC, + admission variables
0.75 vs 0.73
95% CI 0.74–0.76 vs 0.72–0.75P < 0.001
Multiple-layer
AUC, + subsequent variables
0.77 vs 0.76
95% CI 0.76–0.78 vs 0.75–0.77P < 0.001
Multiple-layer
Further findings
| Outcome | Estimate | P value |
|---|---|---|
| Death, AUC · 30 days | AUC, pre-referral variables 0.74 vs 0.71 (95% CI 0.71-0.77 vs 0.66-0.76) | P = 0.04 |
| Death, AUC · 30 days | AUC, + admission variables 0.83 vs 0.78 (95% CI 0.79-0.86 vs 0.73-0.82) | P = 0.01 |
| Death, AUC · 30 days | AUC, + subsequent variables 0.87 vs 0.87 (95% CI 0.85-0.90 vs 0.83-0.89) | P = 0.52 |
- Accuracy of multiple-layer model, all variables
- 70% · accuracy, myocardial injury · accuracy, death 89%
Original publication
Machine learning to predict myocardial injury and death after non‐cardiac surgery. Anaesthesia. Published 18 April 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →