Utilisation-based model, specialist cohort, risk threshold | Diagnosis-based model, specialist cohort, risk threshold | Utilisation-based model, event cohort, risk threshold | Diagnosis-based model, event cohort, risk threshold | |||||||||||||||||
0.2 | 0.3 | 0.4 | 0.5 | 0.6 | 0.2 | 0.3 | 0.4 | 0.5 | 0.6 | 0.1 | 0.2 | 0.3 | 0.4 | 0.5 | 0.1 | 0.2 | 0.3 | 0.4 | 0.5 | |
n# | 962 | 709 | 454 | 251 | 148 | 1016 | 768 | 517 | 300 | 108 | 3438 | 1397 | 570 | 284 | 101 | 4926 | 1427 | 555 | 122 | 29 |
Sensitivity % | 0.960 | 0.820 | 0.607 | 0.371 | 0.231 | 0.969 | 0.820 | 0.646 | 0.437 | 0.163 | 0.729 | 0.460 | 0.267 | 0.158 | 0.065 | 0.812 | 0.371 | 0.198 | 0.056 | 0.014 |
Specificity % | 0.219 | 0.500 | 0.735 | 0.878 | 0.936 | 0.144 | 0.412 | 0.668 | 0.850 | 0.949 | 0.628 | 0.869 | 0.956 | 0.981 | 0.994 | 0.439 | 0.854 | 0.950 | 0.990 | 0.998 |
PPV % | 0.454 | 0.526 | 0.608 | 0.673 | 0.709 | 0.434 | 0.486 | 0.569 | 0.663 | 0.685 | 0.194 | 0.301 | 0.428 | 0.507 | 0.584 | 0.150 | 0.238 | 0.326 | 0.418 | 0.448 |
NPV % | 0.891 | 0.804 | 0.734 | 0.674 | 0.642 | 0.874 | 0.772 | 0.736 | 0.690 | 0.626 | 0.950 | 0.929 | 0.914 | 0.905 | 0.897 | 0.950 | 0.917 | 0.906 | 0.895 | 0.892 |
F-1 score | 0.617 | 0.641 | 0.607 | 0.479 | 0.348 | 0.600 | 0.610 | 0.605 | 0.527 | 0.263 | 0.306 | 0.364 | 0.329 | 0.241 | 0.116 | 0.254 | 0.290 | 0.247 | 0.099 | 0.028 |
Assessment was based on testing dataset only. #: number of eligible patients whose risk was above each risk threshold.