Table 2.
Model | Metric | Overall | Normal | Mild | Moderate | Severe |
---|---|---|---|---|---|---|
Binary | AUROC | 0.846 [0.842, 0.851] |
0.819 [0.793, 0.846] |
0.796 [0.780, 0.812] |
0.798 [0.789, 0.807] |
0.859 [0.852, 0.866] |
Sensitivity | 0.744 (8289/11135) [0.736, 0.753] |
0.625 (185/296) [0.570, 0.680] |
0.657 (595/906) [0.626, 0.688] |
0.674 (2641/3917) [0.660, 0.689] |
0.809 (4868/6016) [0.799, 0.819] |
|
Specificity | 0.803 (19065/23757) [0.797, 0.808] |
0.868 (5464/6298) [0.859, 0.876] |
0.767 (3916/4911) [0.786, 0.809] |
0.781 (6293/8063) [0.771, 0.790] |
0.756 (3392/4485) [0.744, 0.769] |
|
PPV | 0.639 (8289/12981) [0.630, 0.648] |
0.182 (185/1019) [0.138, 0.226] |
0.374 (595/1590) [0.343, 0.406] |
0.599 (2641/4411) [0.583, 0.614] |
0.817 (4868/5961) [0.807, 0.826] |
|
NPV | 0.870 (19065/21911) [0.866, 0.874] |
0.980 (5464/5575) [0.977, 0.984] |
0.926 (3916/4227) [0.919, 0.934] |
0.831 (6293/7569) [0.823, 0.840] |
0.747 (3392/4540) [0.734, 0.760] |
|
Accuracy | 0.784 (27354/34892) [0.780, 0.788] |
0.857 (5649/6594) [0.848, 0.865] |
0.776 (4511/5817) [0.765, 0.786] |
0.746 (8934/11980) [0.738, 0.754] |
0.787 (8260/10501) [0.779, 0.794] |
|
Multiclass | AUROC | 0.844 [0.840, 0.849] |
0.816 [0.788, 0.843] |
0.807 [0.791, 0.822] |
0.795 [0.787, 0.804] |
0.858 [0.850, 0.865] |
Sensitivity | 0.721 (8024/11135) [0.712, 0.729] |
0.628 (186/296) [0.573, 0.683] |
0.605 (548/906) [0.573, 0.637] |
0.638 (2498/3917) [0.623, 0.653] |
0.797 (4792/6016) [0.786, 0.807] |
|
Specificity | 0.813 (19305/23757) [0.808, 0.818] |
0.850 (5352/6298) [0.841, 0.859] |
0.829 (4070/4911) [0.818, 0.839] |
0.800 (6454/8063) [0.792, 0.809] |
0.765 (3429/4485) [0.752, 0.777] |
|
PPV | 0.643 (8024/12476) [0.634, 0.652] |
0.164 (186/1132) [0.122, 0.207] |
0.395 (548/1389) [0.363, 0.426] |
0.608 (2498/4107) [0.593, 0.624] |
0.819 (4792/5848) [0.810, 0.829] |
|
NPV | 0.861 (19305/22416) [0.857, 0.866] |
0.980 (5352/5462) [0.976, 0.983] |
0.919 (4070/4428) [0.912, 0.927] |
0.820 (6454/7873) [0.811, 0.828] |
0.737 (3429/4653) [0.724, 0.750] |
|
Accuracy | 0.783 (27329/34892) [0.779, 0.788] |
0.840 (5538/6594) [0.831, 0.849] |
0.794 (4618/5817) [0.783, 0.804] |
0.747 (8952/11980) [0.739, 0.755] |
0.783 (8221/10501) [0.775, 0.791] |
Data are presented as values (numerator/denominator, if applicable) [95% confidence intervals]. AUROC, area under the receiver operating characteristic curve; PPV, positive predictive value; NPV, negative predictive value.