Machine learning vs. traditional methods for predicting postoperative cardiac complications after non-cardiac surgery: a systematic review and Bayesian network meta-analysis.
Saavan Dhaliwal, Shichao Chen, Chris Papas et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and Bayesian network meta-analysis evaluated the predictive performance of machine learning models versus traditional risk scores for postoperative cardiac complications in non-cardiac surgery. Analyzing 54 models across 927,113 patients, automated machine learning demonstrated the highest performance (SUCRA 96.6) and a mean difference of 0.28 compared to the Revised Cardiac Risk Index, indicating superior predictive capability.
Key findings
- Automated machine learning ranked highest with a SUCRA of 96.6 and a mean difference of 0.28.
- Gradient boosting models outperformed the Revised Cardiac Risk Index with a mean difference of 0.20.
- The Gupta Perioperative Risk score showed a mean difference of 0.16 over the Revised Cardiac Risk Index.
- Only six studies were at low risk of bias, and none validated their models externally.
Why it matters
Accurate prediction of cardiac complications is essential for optimizing pre-operative decision-making. Improved predictive models can enhance patient safety and surgical outcomes, but their clinical integration requires further validation.
What to keep in mind
None of the included studies externally validated their machine learning models, and many were at risk of bias.