Table 2.
Models’ performances
| Internal testing set | External testing set | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| AUC (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | PPV (95% CI) | NPV (95% CI) | AUC (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | PPV (95% CI) | NPV (95% CI) | |
| Ensemble | 79.4 (77.5–81.3)* | 76.0 (71.3–80.1) | 69.0 (68.0–69.9) | 8.4 (7.5–9.4) | 98.7 (98.4–99.0) | 75.8 (73.0–78.5)* | 78.0 (72.0–83.0) | 58.1 (56.8–59.4) | 6.8 (5.9–7.8) | 98.5 (98.1–98.9) |
| ASP | 75.7 (73.7–77.6)*,** | 65.8 (60.8–70.5) | 70.9 (70.0–71.8) | 7.8 (6.9–8.8) | 98.2 (97.9–98.5) | 72.6 (69.8–75.5)*,** | 68.8 (62.4–74.6) | 68.6 (67.3–69.8) | 7.9 (6.7–9.2) | 98.3 (97.8–98.6) |
| HR-DNN | 74.5 (72.3–76.7)*,** | 92.6 (89.5–94.9) | 41.0 (40.0–42.0) | 5.6 (5.0–6.2) | 99.3 (99.0–99.5) | 74.1 (71.2–77.0)*,** | 92.2 (87.9–95.1) | 37.3 (36.0–38.6) | 5.4 (4.7–6.2) | 99.2 (98.7–99.5) |
| Beat-DNN | 75.4 (73.1–77.8)*,** | 57.7 (52.5–62.6) | 76.7 (75.8–77.5) | 8.5 (7.4–9.6) | 98.0 (97.6–98.3) | 70.0 (66.8–73.1)*,** | 70.2 (63.8–75.9) | 58.6 (57.3–59.8) | 6.2 (5.3–7.2) | 98.1 (97.5–98.5) |
| Age and sex | 70.5 (68.4–72.6)** | 87.4 (83.6–90.4) | 47.2 (46.2–48.2) | 5.8 (5.3–6.5) | 99.0 (98.7–99.3) | 64.9 (62.0–67.7)** | 82.6 (77.0–87.0) | 44.1 (42.8–45.4) | 5.4 (4.7–6.3) | 98.5 (97.9–98.9) |
*
P < 0.05 for comparison of AUC between ensemble model and other models.
**
P < 0.05 for comparison of AUC between age and sex model and other models.