Machine Learning for Predicting Postoperative Atrial Fibrillation After Cardiac Surgery: A Scoping Review of Current Literature

Machine learning is more accurate and biased than risk scoring tools in the prediction of postoperative atrial fibrillation after cardiac surgery

“…A recent scoping review identified 7 papers that used ML for predicting POAF after cardiac surgery. [9] Of the 7 studies, 3 relied on electrocardiogram data while the remaining 4 used clinical documentation, administrative data, or Holter monitoring. The sample size, ethnicity composition, and model performance of the 4 ML studies using administrative data are summarized in Table 4.…”

“…Machine learning (ML) has been proposed as an alternative to achieve better predictive performance [9]. A recent scoping review found that support vector machines (SVM), gradient boosting machines (GBM), and random forests (RF) using clinical characteristics can predict POAF risk more accurately than existing risk scores with promising specificity, sensitivity, and AUROC scores [9].…”

“…Machine learning (ML) has been proposed as an alternative to achieve better predictive performance [9]. A recent scoping review found that support vector machines (SVM), gradient boosting machines (GBM), and random forests (RF) using clinical characteristics can predict POAF risk more accurately than existing risk scores with promising specificity, sensitivity, and AUROC scores [9]. Three existing works compared multiple ML algorithms with Lu et al [10] and Parise et al [11] concluding that SVM achieved the best performance while GBM performed the best in Karri et al [12].…”

“…We assess the fairness of the predictive models in both race and sex subpopulations. We also restrict our evaluation to common structured data found within electronic health records (EHRs) as such algorithms can provide quicker (and hopefully more accurate) management strategies [9].…”