. 2022 Nov 24;4(3):158–168. doi: 10.1016/j.hroo.2022.11.004

Table 2.

Summary of model performances for predicting oral anticoagulation prescription in training (5-fold cross-validation) and test sets

		Regular model: CHA₂DS₂-VASc components					Enhanced ML model: CHA₂DS₂-VASc components + new features
		Accuracy	AUROC	PRAUC	Precision	Recall	Accuracy	AUROC	PRAUC	Precision	Recall
XGBOOST	Test set	0.69	0.62	0.76	0.70	0.96	0.77	0.81	0.89	0.79	0.89
XGBOOST	Cross-validation	0.70	0.64	0.77	0.70	0.95	0.78	0.83	0.90	0.80	0.89
Logistic regression	Test set	0.69	0.60	0.73	0.70	0.96	0.73	0.75	0.86	0.77	0.87
Logistic regression	Cross-validation	0.69	0.60	0.73	0.70	0.96	0.74	0.76	0.86	0.76	0.88
Random forest	Test set	0.68	0.59	0.73	0.68	0.99	0.76	0.79	0.85	0.79	0.88
Random forest	Cross-validation	0.75	0.78	0.88	0.76	0.93	0.99	0.99	0.85	0.99	0.99
LASSO-penalized logistic regression	Test set	0.69	0.60	0.73	0.70	0.96	0.74	0.76	0.88	0.76	0.89
LASSO-penalized logistic regression	Cross-validation	0.69	0.60	0.73	0.70	0.96	0.74	0.76	0.99	0.76	0.89

AUROC = area under the receiver-operating characteristic curve; CHA₂DS₂-VASc = congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, prior stroke or transient ischemic attack or thromboembolism, vascular disease, age 65–74 years, sex category; ML = machine learning; PRAUC = area under the precision-recall curve.