Table 2.
Characteristics of Study Population
| Characteristic | Model Derivation | Test for Model Comparison | |
|---|---|---|---|
| Hospital A | Hospital A | Hospital B | |
| Study period | Jun 2010–Jan 2017 | Feb–Jul 2017 | Mar–Jul 2017 |
| Total patients, n | 46 725 | 3634 | 1772 |
| Input vectors, n | 2 769 324 | 152 587 | 60 782 |
| Patients with in‐hospital cardiac arrest, n | 396 | 19 | 4 |
| Input vectors, n | 10 772 | 352 | 82 |
| Patients with death without attempted resuscitation, n | 770 | 25 | 19 |
| Input vectors, n | 28 208 | 801 | 446 |
| Age, y, mean±SD | 56.7±23.3 | 58.2±21.7 | 58.1±17.2 |
| Male sex, n (%) | 24 171 (51.7) | 1878 (51.7) | 829 (46.8) |
| Maximum DEWS score, mean±SD | |||
| Nonevent | 16.3±28.4 | 9.2±20.9 | 6.6±17.5 |
| Cardiac arrest | 89.0±21.0 | 60.6±35.7 | 67.2±38.3 |
| Death without attempted resuscitation | 97.9±7.9 | 84.5±27.8 | 81.1±18.5 |
| Hospital type | Cardiovascular teaching hospital | Community general hospital | |
| Number of beds | 310 | 310 | 172 |
DEWS indicates deep learning–based early warning system.