Table 3 Model performance on the validation and test datasets (maximizing sensitivity and specificity)
|  | Validation dataset (n = 1806) | Test dataset (n = 1815) | |
|---|---|---|---|
Actual | Bootstrap (95% CI) | ||
True positive (TP) | 307 | 345 | Â |
True negative (TN) | 1000 | 1001 | Â |
False positive (FP) | 399 | 385 | Â |
False negative (FN) | 100 | 84 | Â |
AUC | 79.1% | 82.9% | (80.6%–84.9%) |
Accuracy | 72.4% | 74.2% | (74.1%–74.3%) |
Sensitivity | 75.4% | 80.4% | (80.4%–80.6%) |
Specificity | 71.5% | 72.2% | (72.2%–72.3%) |
Positive Predictive Value (PPV) | 43.5% | 47.3% | (47.2%–47.4%) |
Negative Predictive Value (NPV) | 90.9% | 92.3% | (92.3%–92.4%) |
Positive likelihood ratio (LR+) | 2.64 | 2.90 | (2.90–2.91) |
Negative likelihood ratio (LR−) | 0.34 | 0.27 | (0.26–0.27) |
F1 score | 0.55 | 0.60 | (0.59–0.60) |
Mathews correlation coefficient (MCC) | 0.40 | 0.46 | (0.45–0.46) |