Why the study?
The generalizability of machine learning models across institutions remains uncertain, and head-to-head comparisons between machine learning algorithms and established clinical risk scores for predicting 30-day MACE after noncardiac surgery are scarce.
Do machine learning models improve the prediction of 30-day major adverse cardiac events in patients undergoing noncardiac surgery compared to established clinical risk scores?
Population
Patients in a site-separated two-center cohort (derivation n = 707, validation n = 378)
Comparison
Ten machine learning algorithms vs AUB-HAS2, ASA, and RCRI clinical risk scores
Design
Site-separated two-center external validation study
Follow-up
30-day
Key result
Machine learning models demonstrated metric-dependent performance but no statistically significant discrimination advantage over the AUB-HAS2 clinical risk score for predicting 30-day MACE (all p>0.05).
Authors
Loading...
ML models for preoperative MACE prediction need prospective validation across centers; leaves open superiority over clinical scores in routine practice.
Cohort (n=1,085)
Yes
Do machine learning models improve the prediction of 30-day major adverse cardiac events in patients undergoing noncardiac surgery compared to established clinical risk scores?
Effect estimate: AUROC 0.738 (95% CI 0.668-0.804)
p-value: p=>0.05
Machine learning models did not outperform the established AUB-HAS2 clinical risk score for predicting 30-day MACE after noncardiac surgery, highlighting the need for local recalibration and prospective evaluation before clinical deployment.
Erdoğan et al. (2026) conducted a cohort in Noncardiac surgery (n=1,085). Machine learning models vs. Established clinical risk scores (AUB-HAS2, ASA, RCRI) was evaluated on 30-day major adverse cardiac events (MACE) (AUROC 0.738, 95% CI 0.668-0.804, p=>0.05). Machine learning models demonstrated metric-dependent performance but no statistically significant discrimination advantage over the AUB-HAS2 clinical risk score for predicting 30-day MACE (all p>0.05).