An Interpretable Model-Based Prediction of Severity and Crucial Factors in Patients with COVID-19
Table 3
The AUC, sensitivity, and specificity comparisons.
AUC (95% CI)
Sensitivity (95% CI)
Specificity (95% CI)
LR
0.891 (0.783, 0.956)
90.91 (58.7, 99.8)
93.88 (83.1, 98.7)
0.1306
KNN
0.857 (0.743, 0.934)
100.00 (71.5, 100.0)
61.22 (46.2, 74.8)
0.2844
DT
0.707 (0.575, 0.817)
45.45 (16.7, 76.6)
95.92 (86.0, 99.5)
0.0095
RF
0.907 (0.804, 0.967)
90.91 (58.7, 99.8)
95.92 (86.0, 99.5)
0.1915
SVM
0.892 (0.785, 0.958)
90.91 (58.7, 99.8)
91.84 (80.4, 97.7)
0.2006
XGBoost
0.924 (0.826, 0.976)
90.91 (58.7, 99.8)
97.96 (89.1, 99.9)
—
Two-sided values were calculated by comparing AUC for the XGBoost model with the other models. AUC comparisons were evaluated using the DeLong test; LR: logistic regression; KNN: -nearest neighbor; DT: decision tree; RF: random forest; SVM: support vector machine; XGBoost: eXtreme gradient boosting.
