Predicting postoperative atrial fibrillation after cardiac surgery using machine learning.

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Postoperative atrial fibrillation (POAF) is a common complication after cardiac surgery, increasing patient morbidity.
• Current risk prediction scores for POAF lack sufficient accuracy.
• There is a critical need for improved models to predict new-onset POAF.

Purpose of the Study:

• To develop and validate a machine-learning (ML) model for predicting new-onset POAF in patients undergoing cardiac surgery.
• To utilize perioperative variables for enhanced prediction accuracy.
• To provide a tool for risk stratification and improved perioperative decision-making.

Main Methods:

• A cohort of 6859 adult patients undergoing cardiac surgery was analyzed.
• An extreme gradient boosting ML model was developed using the top 20 perioperative variables.
• The model underwent internal validation (516 patients) and external validation (1701 patients).

Main Results:

• The incidence of new-onset POAF was 37.2%.
• The ML model achieved high performance with an AUROC of 0.891 (internal) and 0.726 (external).
• Strong negative predictive values (0.867 internal, 0.773 external) indicate utility in identifying low-risk patients.

Conclusions:

• A machine-learning model using perioperative data effectively predicts new-onset POAF after cardiac surgery.
• This predictive model shows potential for risk-stratified postoperative monitoring.
• The findings support the use of ML in enhancing perioperative decision-making for cardiac surgery patients.