Table 1.
Demographic characteristics of the entire cohort and the parsed training, validation, and final test datasets at the development site and the entire cohort, training, and validation datasets, as well as the generalizable model validation dataset at the external validation site.
Development Site | Validation Site | ||||
---|---|---|---|---|---|
Characteristic | Entire Cohort | Training Dataset | Validation Dataset | Final Test Dataset | Generalizable Model Validation Dataset |
N = 18,568 | n = 10,744 | n = 3,582 | n = 4,242 | N = 6,825 | |
Age (months), median (IQR) | 70 (18, 161) | 67 (18, 158) | 68.5 (18, 160) | 77 (19, 167) | 96 (18,171) |
Female, n (%) | 8,325 (45) | 4,748 (44) | 1,593 (44) | 1,984 (47) | 3,159 (46) |
Glasgow Coma Scale Score, median (IQR)* | 15 (12, 15) | 15 (11, 15) | 15 (11, 15) | 15 (14, 15) | 14 (14,15) |
Mechanical Ventilation, n (%) | 5,352 (29) | 3,322 (31) | 1,143 (32) | 887 (21) | 1,948(29) |
Endotracheal Tube, n (%) | 1,759 (9) | 1,142 (11) | 370 (10) | 247 (6) | 1,283 (19) |
Vasoactive Medication, n (%)** | 720 (4) | 393 (4) | 148 (4) | 179 (4) | 221 (3) |
Sedative-Analgesic Medication, n (%)*** | 4,660 (25) | 2,948 (27) | 987 (27) | 727 (17) | 1,930 (28) |
The last recorded Glasgow coma scale score for the encounter prior to the censored time horizon
Vasoactive medications include dobutamine, dopamine, epinephrine, norepinephrine, or milrinone
Sedative-analgesic medications include fentanyl, hydromorphone, midazolam, or morphine
Abbreviations: IQR, interquartile range