Perioperative medicine & outcomes
Machine learning vs. traditional methods for predicting postoperative cardiac complications after non‐cardiac surgery: a systematic review and Bayesian network meta‐analysis
Authors' conclusion
Most machine learning models showed better discrimination than traditional risk scores, with automated machine learning and gradient boosting models ranking highest.
Study context
- Population
- Patients undergoing non-cardiac surgery54 models
- Participants
- 927,113 patients · 13 studies
- Comparison
- Machine learning and traditional prediction models vs Revised Cardiac Risk Index
- Outcome
- Postoperative adverse cardiac events after non-cardiac surgery
Methodological detailsSUCRA ranking, best version and worst version analyses, sensitivity analysis, No included study externally validated its machine learning models; only 6 at low risk of bias
- SUCRA ranking
- Best version and worst version analyses
- Sensitivity analysis
- No included study externally validated its machine learning models; only 6 at low risk of bias
Pooled result
As reported in the abstract
Discrimination for postoperative adverse cardiac events (best version analysis)
Automated machine learning
96.6
SUCRA
Gradient boosting
82.4
SUCRA
Each group vs revised Cardiac Risk IndexMD (95% CrI)
Automated machine learning
0.28(0.16–0.40)
Gradient boosting
0.20(0.14–0.26)
Gupta Perioperative Risk for Myocardial Infarction or Cardiac Arrest score
0.16(0.01–0.32)
Further pooled results
| Outcome | MD (95% CrI) | |
|---|---|---|
| Discrimination, sensitivity analysisAutomated machine learning vs Revised Cardiac Risk Index · CrI | 0.30 (0.14–0.45) | |
| Discrimination, worst version analysisGradient boosting vs Revised Cardiac Risk Index · CrI | 0.18 (0.12–0.25) | |
| −0.200.20.40.6Difference (linear scale) |
Original publication
Machine learning vs. traditional methods for predicting postoperative cardiac complications after non‐cardiac surgery: a systematic review and Bayesian network meta‐analysis. Anaesthesia. Published 22 June 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →