Key result
Perioperative ECG-clinical neural network predicts absence of AF at discharge with ~99% negative predictive value.
Why the study?
Atrial fibrillation at discharge is an infrequent but clinically relevant complication following cardiac surgery linked to adverse long-term outcomes, prompting the development of artificial intelligence-based predictive models.
Does an artificial intelligence-based model using perioperative electrocardiographic and clinical variables accurately predict atrial fibrillation at discharge in patients undergoing cardiac surgery?
Observational (n=1,905)
No
Does an artificial intelligence-based model using perioperative electrocardiographic and clinical variables accurately predict atrial fibrillation at discharge in patients undergoing cardiac surgery?
Effect estimate: NPV 99.2% (95% CI 98.6-99.6)
An artificial intelligence-based model incorporating perioperative ECG and clinical variables can accurately predict the absence of atrial fibrillation at discharge following cardiac surgery, potentially enabling safe identification of low-risk patients and guiding targeted prophylaxis.
AI models may aid post-cardiac surgery AF prediction; hypothesis-generating and requires prospective validation before use.
AbstractBackgroundAtrial fibrillation at discharge is an infrequent but clinically relevant complication after cardiac surgery, linked to adverse long-term outcomes. This study aimed to develop and evaluate artificial intelligence-based models for predicting atrial fibrillation using electrocardiographic conduction abnormalities and other clinical risk factors.MethodsA total of 3,405 patients undergoing coronary artery bypass grafting, valvular, aortic, or combined procedures were included. Perioperative 12-lead electrocardiograms and clinical variables were preselected via univariable logistic regression (p<0.1). Predictive models—neural networks, support vector machines, k-nearest neighbours, random forests, and Bayes classifiers—were trained with 10-fold cross-validation and tested independently. A genetic variable selection algorithm optimized predictor selection based on neural network performance and a ‘reject option’ withheld predictions in uncertain cases.ResultsAmong 1,905 patients included in the final model, 2.2% developed-atrial fibrillation at discharge. The model provided predictions in 69.9% of cases, with a negative predictive value of 99.2% (95% CI: 98.6–99.6). Key predictors were age, CHA2DS2-VASc score, left ventricular ejection fraction, EuroSCORE II, cardiopulmonary bypass time, prior cardiogenic shock, perioperative atrioventricular block, and mitral valve surgery.ConclusionsAn artificial intelligence-based model incorporating perioperative electrocardiographic parameters accurately predicted atrial fibrillation at discharge. This tool enables efficient risk stratification, potentially guiding targeted prophylaxis and reducing overtreatment.
No takes yet. Share an insight, caveat, or question.
Leiler et al. (2026) conducted an observational in Atrial fibrillation following cardiac surgery (n=1,905). Neural network-based prediction model was evaluated on Correct prediction of absence of atrial fibrillation at discharge (Negative Predictive Value) (NPV 99.2%, 95% CI 98.6-99.6). A neural network model integrating perioperative electrocardiographic and clinical variables predicted the absence of atrial fibrillation at discharge with a negative predictive value of 99.2% in 69.9% of patients.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: