Table 2

Performance for 6-month incident AF with optimal threshold determined by Youden Index

Algorithm
FIND-AFMLRCHA2DS2-VAScC2HEST
AUROC (95% CI)0.824 (0.814 to 0.834)0.765 (0.755 to 0.769)0.784 (0.773 to 0.794)0.757 (0.744 to 0.770)
Sensitivity (95% CI)0.781 (0.731 to 0.829)0.760 (0.653 to 0.814)0.847 (0.829 to 0.866)0.642 (0.619 to 0.791)
Specificity (95% CI)0.731 (0.693 to 0.771)0.679 (0.635 to 0.776)0.611 (0.608 to 0.612)0.790 (0.622 to 0.792)
PPV (%(95% CI))2.5% (2.3 to 2.7)2.0% (1.8 to 2.6)2.2% (2.1 to 2.3)2.0% (1.5 to 2.2)
NPV (%(95% CI))99.8% (99.8 to 99.8)99.7% (99.6 to 99.7)99.8% (99.8 to 99.8)99.7% (99.7 to 99.8)
Calibration slope* (95% CI)0.782 (0.743 to 0.824)0.698 (0.654 to 0.735)0.621 (0.589 to 0.652)0.608 (0.576 to 0.648)
Brier score0.0690.0970.0930.102
  • *Calibration slope was derived from linear regression models by forcing the intercept through origin (0, 0).

  • AF, atrial fibrillation; AUROC, area under the receiver operating characteristic; CHA2DS2-VASc, Congestive heart failure, Hypertension, Age >75 (2 points), Stroke/transient ischaemic attack/thromboembolism (2 points), Vascular disease, Age 65–74, Sex category; C2HEST, Coronary artery disease/Chronic obstructive pulmonary disease (1 point each), Hypertension, Elderly (age ≥75, 2 points), Systolic heart failure, Thyroid disease (hyperthyroidism); FIND-AF, Future Innovations in Novel Detection of Atrial Fibrillation; MLR, multivariable logistic regression; NPV, negative predictive value; PPV, positive predictive value.