Table 2.
Ablation study using a subset of the training data.
| Training dataset size | AUROC | Sensitivity | Specificity | PPV |
|---|---|---|---|---|
| 25% | 86.325 (± 2.9) | 9.663 (± 5.6) | 99.616 (± 0.3) | 63.483 (± 26.8) |
| 50% | 87.965 (± 3.3) | 14.962 (± 5.6) | 99.374 (± 0.4) | 62.781 (± 14.4) |
| 75% | 88.365 (± 3.3) | 17.015 (± 5.7) | 99.315 (± 0.4) | 62.221 (± 13.262) |
| 100% | 88.761 (± 3.0) | 18.514 (± 5.5) | 99.266 (± 0.4) | 63.431 (± 12.7) |
In each row, we report the AUROC, sensitivity, specificity, and positive predictive value (PPV) for the EPPT trained using a random subset of the training data. Each value denotes the mean (in percent) computed across 20 cross-validation folds and in parentheses we show the standard deviation.