Table 1 Demographic data of development and test data set.
From: A scalable physician-level deep learning algorithm detects universal trauma on pelvic radiographs
No. (%) of patients | |||
|---|---|---|---|
Development data set | Test data set | ||
Variables | (n = 5204) | (n = 1888) | p Value |
Year of injury | 2008–2016 | 2017 | |
Age, median (IQR), y | 60.00 [37.00, 78.00] | 55.00 [30.75, 75.25] | <0.001 |
Gender, male | 2752 (52.9) | 908 (48.1) | <0.001 |
Mechanism of injury | <0.001 | ||
Motor vehicle accident | 2193 (42.1) | 893 (47.3) | |
Fall | 2683 (51.6) | 843 (44.7) | |
Mechanical injury | 148 (2.8) | 49 (2.6) | |
Other mechanisms | 112 (2.2) | 42 (2.2) | |
Unavailable | 68 (1.3) | 61 (3.2) | |
Extremity AIS ≥ 3 | 2624 (50.4) | 517 (27.4) | <0.001 |
AIS ≥ 16 | 1087 (20.9) | 185 (9.8) | <0.001 |
Acute trauma finding | 3110 (59.8) | 619 (32.8) | |
Hip fracture | |||
Intracapsular | 1014 (19.5) | 184 (9.7) | <0.001 |
Extracapsular | 1059 (20.3) | 207 (11.0) | <0.001 |
Pelvic area fracture | <0.001 | ||
LC type | 576 (11.1) | 97 (5.1) | |
APC type | 60 (1.2) | 7 (0.4) | |
VS type | 25 (0.5) | 3 (0.2) | |
Other | 259 (5.0) | 42 (2.2) | |
Hip dislocation | 130 (2.5) | 51 (2.7) | 0.693 |
Femoral shaft fracture | 62 (1.2) | 30 (1.6) | 0.234 |
Periprosthetic fracture | 38 (0.7) | 29 (1.5) | 0.003 |
