Table 1 Overview and prognostic performance
Timeline | Type of data | Number of patients | Number of variables | Model performance | Confusion matrix (test set) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Total (n) | Train set (n) | Test set (n) | Before variable selection (n) | After variable selection (n) | Training set (cross validation) AUROC (%) (CI 95%) | Test set AUROC (%) (CI 95%) | Probability threshold | TN (n) | FP (n) | FN (n) | TP (n) | ||
Day 1 (24 h) | Level A | 884 | 702 | 182 | 120 | 22 | 85.7 (83.2–88.6) | 81.9 (75.9–87.9) | 0.981 | 99 | 0 | 82 | 1 |
Level B | 638 | 502 | 136 | 125 | 21 | 89.9 (86.8–92.2) | 81.8 (74.9–88.6) | 0.983 | 70 | 0 | 57 | 9 | |
Level C | 690 | 545 | 145 | 131 | 22 | 96.6 (94.7–97.5) | 95.2 (91.9–98.4) | 0.887 | 76 | 1 | 21 | 47 | |
Day 2 (48 h) | Level A | 812 | 645 | 167 | 174 | 21 | 85.8 (82.9–88.5) | 78.1 (71.1–85.0) | 0.935 | 88 | 2 | 70 | 7 |
Level B | 578 | 460 | 118 | 184 | 17 | 93.7 (91.7–95.9) | 89.7 (84.3–95.0) | 0.963 | 59 | 0 | 44 | 15 | |
Level C | 624 | 495 | 129 | 196 | 21 | 96.6 (95.2–97.9) | 96.1 (93.4–98.8) | 0.986 | 67 | 0 | 31 | 31 | |
Day 3 (72 h) | Level A | 592 | 469 | 107 | 228 | 17 | 84.9 (80.6–87.9) | 86.7 (80.4–93.1) | 0.920 | 60 | 0 | 22 | 25 |
Level B | 415 | 328 | 87 | 243 | 12 | 92.7 (89.8–95.1) | 94.1 (89.4–98.8) | 0.885 | 40 | 0 | 22 | 25 | |
Level C | 442 | 347 | 95 | 261 | 23 | 96.6 (94.9–98.1) | 94.7 (90.6–98.7) | 0.820 | 45 | 0 | 16 | 34 | |