Cardiothoracic & vascular anesthesia
Predicting postoperative atrial fibrillation after cardiac surgery using machine learning
Authors' conclusion
A ML model developed with perioperative data demonstrated strong performance in predicting new-onset POAF after cardiac surgery.
Study context
- Population
- Adults undergoing cardiac surgery at Seoul National University Hospital (SNUH) or Seoul National University Bundang Hospital (SNUBH), October 2004–October 2021
- Participants
- 6,859 adults
- Comparison
- Machine-learning (ML) model performance, internal vs external validation
- Internal validationn = 516
- External validationn = 1,701
- Outcome
- Prediction of new-onset POAF (AUROC and AUPRC)
Methodological detailsmachine learning (extreme gradient boosting), internal and external validation, two-centre
- Machine learning (extreme gradient boosting)
- Internal and external validation
- Two-centre
Performance
As reported in the abstract
Model performance for new-onset POAF
| Test or model | AUROC | AUPRC | NPV |
|---|---|---|---|
| Internal validation | 0.891 (95% CI, 0.859-0.920) | 0.861 (95% CI, 0.819-0.898) | 0.867 |
| External validation | 0.726 (95% CI, 0.700-0.751) | 0.646 (95% CI, 0.607-0.684) | 0.773 |
AUROC
Internal validation0.891 (95% CI, 0.859-0.920)
External validation0.726 (95% CI, 0.700-0.751)
AUPRC
Internal validation0.861 (95% CI, 0.819-0.898)
External validation0.646 (95% CI, 0.607-0.684)
NPV
Internal validation0.867
External validation0.773
Further findings
Incidence of new-onset postoperative atrial fibrillation (POAF) after cardiac surgery
- Incidence of new-onset POAF
- 37.2%
Original publication
Predicting postoperative atrial fibrillation after cardiac surgery using machine learning. J Clin Anesth. Published 5 May 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →