Table 2.
Prediction performance of internal validation in CMUH
| Model | AUC | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|
| CatBoost | 0.755 (0.743–0.767) | 48.70% (47.52–49.89) | 83.12% (81.43–84.72) | 90.64% (89.77–91.45) | 32.56% (31.91–33.23) |
| XGBoost | 0.749 (0.736–0.761) | 51.50% (50.31–52.69) | 81.08% (79.32–82.75) | 90.13% (89.28–90.92) | 33.25% (32.55–33.97) |
| Random Forest | 0.733 (0.720–0.745) | 33.80% (34.67–36.95) | 88.18% (86.71–89.55) | 91.04% (90.00–91.99) | 29.05% (28.56–29.54) |
| Decision tree | 0.704 (0.691–0.717) | 44.05% (42.87–45.23) | 81.76% (80.02–83.41) | 89.02% (88.05–89.91) | 30.34% (29.72–30.96) |
| Logistic regression | 0.694 (0.681–0.707) | 28.71% (27.65–29.80) | 89.80% (88.46–91.03) | 89.80% (88.55–90.93) | 27.13% (28.30–29.14) |
Abbreviation: AUC Area under the receiver operating characteristic curve, PPV Positive predictive value, NPV Negative predictive value, CMUH China Medical University Hospital