Table 2 Verification results for the six models in the test set
From: Performance of machine learning-based models to screen obstructive sleep apnea in pregnancy
Accuracy | AUC | Sensitivity | Specificity | Precision | F1 | |
|---|---|---|---|---|---|---|
Berlin questionnaire | 0.687 | 0.684 (0.579–0.788) | 0.554 (0.423–0.684) | 0.814 (0.714–0.913) | 0.738 (0.605–0.871) | 0.6327 |
Epworth questionnaire | 0.548 | 0.542 (0.423–0.661) | 0.321 (0.199–0.444) | 0.763 (0.654–0.871) | 0.563 (0.391–0.734) | 0.4091 |
STOP questionnaire | 0.617 | 0.613 (0.501–0.725) | 0.446 (0.316–0.577) | 0.780 (0.674–0.885) | 0.658 (0.507–0.809) | 0.5319 |
STOP-Bang questionnaire | 0.643 | 0.635 (0.490–0.780) | 0.303 (0.183–0.424) | 0.966 (0.920–1.000) | 0.895 (0.757–1.000) | 0.4533 |
Improved model for BQ (a) | 0.739 | 0.808 (0.730–0.886) | 0.803 (0.700–0.908) | 0.678 (0.559–0.797) | 0.703 (0.591–0.815) | 0.7500 |
Improved model for EES (b) | 0.713 | 0.785 (0.703–0.867) | 0.803 (0.700–0.908) | 0.627 (0.504–0.750) | 0.672 (0.559–0.784) | 0.7317 |
Improved model for SQ (c) | 0.739 | 0.818 (0.742–0.894) | 0.821 (0.721–0.922) | 0.661 (0.540–0.782) | 0.697 (0.586–0.808) | 0.7541 |
Improved model for SBQ (d) | 0.730 | 0.813 (0.737–0.890) | 0.821 (0.721–0.921) | 0.644 (0.522–0.766) | 0.687 (0.576–0.798) | 0.7480 |
Integrated model (e) | 0.739 | 0.795 (0.713–0.876) | 0.786 (0.678–0.893) | 0.695 (0.577–0.812) | 0.710 (0.597–0.823) | 0.7458 |
Momosa (f) | 0.739 | 0.823 (0.748–0.898) | 0.821 (0.721–0.922) | 0.661 (0.540–0.781) | 0.697 (0.586–0.808) | 0.7541 |
Random Forest | 0.756 | 0.830 (0.755–0.904) | 0.786 (0.678–0.893) | 0.729 (0.615–0.842) | 0.733 (0.621–0.845) | 0.7586 |
XG-Boost | 0.722 | 0.816 (0.740–0.893) | 0.804 (0.699–0.908) | 0.644 (0.522–0.766) | 0.682 (0.569–0.794) | 0.7377 |
