Table 3 Performance comparison of the predictive models in the development and testing datasets
Dataset | Model | AUC (95% CI) | p-value | threshold | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|---|---|---|
Development | Clinical | 0.734 (0.674–0.788) | 0.068 | > 0.4487 | 52% | 90% | 79% | 73% |
Single-DL | 0.710 (0.649–0.766) | < 0.001 | > 0.5433 | 56% | 73% | 60% | 70% | |
Multi-DL | 0.767 (0.710–0.819) | < 0.001 | > 0.4651 | 57% | 82% | 69% | 73% | |
Hybrid-DL | 0.857 (0.807–0.898) | - | > 0.3086 | 85% | 76% | 72% | 87% | |
Testing | Clinical | 0.726 (0.615–0.819) | 0.123 | > 0.4280 | 56% | 79% | 66% | 71% |
Single-DL | 0.676 (0.563–0.776) | 0.028 | > 0.7583 | 65% | 70% | 61% | 73% | |
Multi-DL | 0.738 (0.628–0.829) | 0.066 | > 0.4690 | 56% | 83% | 70% | 72% | |
Hybrid-DL | 0.839 (0.741–0.911) | - | > 0.2210 | 77% | 85% | 79% | 83% |