Table 3. Mortality prediction performance of logistic regression as a function of the number of similar patients used in training.
| Number of Similar Patients | AUROC (Mean [95% CI]) | AUPRC (Mean [95% CI]) |
|---|---|---|
| 5000 | 0.830 [0.824, 0.836] | 0.473 [0.460, 0.487] |
| 6000 | 0.830 [0.825, 0.836] | 0.474 [0.460, 0.488] |
| 7000 | 0.829 [0.823, 0.834] | 0.471 [0.457, 0.485] |
| 8000 | 0.828 [0.821, 0.834] | 0.472 [0.457, 0.486] |
| 9000 | 0.827 [0.821, 0.833] | 0.470 [0.456, 0.484] |
| 10000 | 0.826 [0.819, 0.832] | 0.467 [0.453, 0.481] |
| 11000 | 0.824 [0.817, 0.831] | 0.466 [0.452, 0.479] |
| 12000 | 0.822 [0.815, 0.830] | 0.462 [0.448, 0.477] |
| 13000 | 0.820 [0.812, 0.828] | 0.459 [0.444, 0.474] |
| 14000 | 0.816 [0.808, 0.825] | 0.455 [0.441, 0.470] |
| 15000 | 0.814 [0.805, 0.822] | 0.452 [0.437, 0.468] |
| 15649 | 0.810 [0.801, 0.819] | 0.447 [0.432, 0.461] |
The results shown in Fig 2 are tabulated here. AUROC: area under the receiver operating characteristic curve; AUPRC: area under the precision-recall curve; CI: confidence interval.