Skip to main content

Table 3 Results of model output indicators in different age subgroups

From: A machine learning-based model for predicting paroxysmal and persistent atrial fibrillation based on EHR

Model

ACC (95%CI)

Precision

Recall

AUC (95%CI)

F1 Score (95%CI)

SEN (95%CI)

SPE (95%CI)

Age under 60 years old

GradientBoost

0.819 (0.783–0.856)

0.731

0.664

0.891 (0.856–0.892)

0.680 (0.544–0.779)

0.664 (0.417–0.825)

0.885 (0.803–0.957)

AdaBoost

0.819 (0.728–0.857)

0.731

0.717

0.866 (0.828–0.911)

0.709 (0.638–0.763)

0.717 (0.574–0.792)

0.862 (0.700-0.961)

XGBoost

0.816 (0.761–0.878)

0.703

0.700

0.882 (0.836–0.918)

0.695 (0.618–0.802)

0.700 (0.535–0.823)

0.866 (0.785–0.924)

Age from 60 to 64 years old

GradientBoost

0.821 (0.746–0.898)

0.795

0.759

0.892 (0.842–0.929)

0.772 (0.684–0.872)

0.759 (0.625–0.904)

0.863 (0.781–0.934)

AdaBoost

0.776 (0.696–0.813)

0.725

0.722

0.843 (0.724–0.904)

0.721 (0.617–0.774)

0.722 (0.603–0.850)

0.812 (0.750–0.872)

XGBoost

0.825 (0.725–0.905)

0.812

0.749

0.905 (0.860–0.948)

0.776 (0.668–0.884)

0.749 (0.630–0.894)

0.875 (0.731–0.966)

Age 65 years old and above

GradientBoost

0.765 (0.741–0.793)

0.687

0.687

0.837 (0.827–0.862)

0.686 (0.639–0.732)

0.687 (0.610–0.759)

0.812 (0.782–0.837)

AdaBoost

0.760 (0.721–0.811)

0.689

0.654

0.815 (0.770–0.836)

0.669 (0.605–0.758)

0.654 (0.568–0.790)

0.824 (0.798–0.854)

XGBoost

0.765 (0.746–0.810)

0.695

0.668

0.835 (0.820–0.872)

0.680 (0.639-0.745)

0.668 (0.596–0.742)

0.824 (0.789–0.855)

  1. ACC, Accuracy; AUC, Area under curve; CI, Confidence interval; SEN, Sensitivity; SPE, Specificity