What does this research mean for the field?

A machine-learning model predicted new-onset postoperative atrial fibrillation with an AUROC of 0.920 in internal validation and 0.804 in external validation. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The study aims to create a machine learning model that predicts postoperative atrial fibrillation using various perioperative factors.

February 8, 2026

Predicting postoperative atrial fibrillation after cardiac surgery using machine learning

Key Result

A machine-learning model using 169 perioperative variables predicted postoperative atrial fibrillation with AUROC 0.920 internally and 0.804 externally.

Key Points

The study aims to create a machine learning model that predicts postoperative atrial fibrillation using various perioperative factors.
Included adults who underwent cardiac surgery across two hospitals in South Korea.
Developed a prediction model using 169 perioperative variables with extreme gradient boosting.
Evaluated model performance using areas under receiver operating characteristic curve and precision recall curve.
6,859 patients were analyzed, showing a 37.2% incidence of new-onset postoperative atrial fibrillation.
Internal validation results: AUROC of 0.920 and AUPRC of 0.884.
External validation results: AUROC of 0.804 and AUPRC of 0.708.

Structured PICO

Can a machine learning model using perioperative variables accurately predict new-onset postoperative atrial fibrillation in adults undergoing cardiac surgery?

Population

6,859 adults who underwent cardiac surgery at two hospitals in the Republic of Korea between October 2004 and October 2021

Intervention

Machine learning prediction model based on extreme gradient boosting using 169 pre-, intra-, and postoperative variables

Outcome

Prediction of new-onset postoperative atrial fibrillation (POAF) assessed by area under the receiver operating characteristic curve (AUROC) and precision recall curve (AUPRC)

A machine learning model utilizing 169 perioperative variables demonstrated strong predictive performance for new-onset postoperative atrial fibrillation following cardiac surgery across both internal and external validation cohorts.

Main Result

Absolute Event Rate: 0% vs 0%

Abstract

Abstract Background Postoperative atrial fibrillation (POAF) is a frequent complication after cardiac surgery that is associated with increased morbidity. Existing risk scores have limited accuracy, highlighting the need for more effective prediction models. Purpose In this study, we aimed to develop a prediction model using various perioperative variables and a machine-learning (ML) algorithm to predict POAF following cardiac surgery. Methods Adults who underwent cardiac surgery at two hospitals in Republic of Korea between October 2004 and October 2021 were included, and 169 perioperative variables were input to develop a prediction model based on extreme gradient boosting. The areas under the receiver operating characteristic curve (AUROC) and precision recall curve (AUPRC), and 95% confidence intervals (CIs) were evaluated to assess model performance. Results The analysis included 6,859 patients: 5,158 at one hospital for internal validation and 1,701 at another hospital for external validation. The incidence of new-onset POAF after cardiac surgery was 37.2%. The AUROC and AUPRC were 0.920 (95% CI, 0.902–0.938) and 0.884 (95% CI, 0.851–0.910), respectively, for internal validation and 0.804 (95% CI, 0.782–0.827) and 0.708 (95% CI, 0.671–0.742) for external validation. Conclusion In this study of cardiac surgery patients, a ML model using pre-, intra-, and postoperative data successfully predicted new-onset POAF, including external validation. This model provides a foundation for future research aimed at preventing and managing POAF through precision prediction models.

Bookmark