Table 1 Characteristics of hospitalizations in training and test sets
From: Scalable and accurate deep learning with electronic health records
Training data (n = 194,470) | Test data (n = 21,751) | |||
|---|---|---|---|---|
Hospital A (n = 85,522) | Hospital B (n = 108,948) | Hospital A (n = 9624) | Hospital B (n = 12,127) | |
Demographics | ||||
Age, median (IQR) y | 56 (29) | 57 (29) | 55 (29) | 57 (30) |
Female sex, no. (%) | 46,848 (54.8%) | 62,004 (56.9%) | 5364 (55.7%) | 6935 (57.2%) |
Disease cohort, no. (%) | ||||
Medical | 46,579 (54.5%) | 55,087 (50.6%) | 5263 (54.7%) | 6112 (50.4%) |
Cardiovascular | 4616 (5.4%) | 6903 (6.3%) | 528 (5.5%) | 749 (6.2%) |
Cardiopulmonary | 3498 (4.1%) | 9028 (8.3%) | 388 (4.0%) | 1102 (9.1%) |
Neurology | 6247 (7.3%) | 6653 (6.1%) | 697 (7.2%) | 736 (6.1%) |
Cancer | 14,544 (17.0%) | 19,328 (17.7%) | 1617 (16.8%) | 2087 (17.2%) |
Psychiatry | 788 (0.9%) | 339 (0.3%) | 64 (0.7%) | 35 (0.3%) |
Obstetrics and newborn | 8997 (10.5%) | 10,462 (9.6%) | 1036 (10.8%) | 1184 (9.8%) |
Other | 253 (0.3%) | 1148 (1.1%) | 31 (0.3%) | 122 (1.0%) |
Previous hospitalizations, no. (%) | ||||
0 hospitalizations | 54,954 (64.3%) | 56,197 (51.6%) | 6123 (63.6%) | 6194 (51.1%) |
≥1 and <2 hospitalizations | 14,522 (17.0%) | 19,807 (18.2%) | 1620 (16.8%) | 2175 (17.9%) |
≥2 and <6 hospitalizations | 12,591 (14.7%) | 24,009 (22.0%) | 1412 (14.7%) | 2638 (21.8%) |
≥6 hospitalizations | 3455 (4.0%) | 8935 (8.2%) | 469 (4.9%) | 1120 (9.2%) |
Discharge location no. (%) | ||||
Home | 70,040 (81.9%) | 91,273 (83.8%) | 7938 (82.5%) | 10,109 (83.4%) |
Skilled nursing facility | 6601 (7.7%) | 5594 (5.1%) | 720 (7.5%) | 622 (5.1%) |
Rehabilitation | 2666 (3.1%) | 5136 (4.7%) | 312 (3.2%) | 649 (5.4%) |
Another healthcare facility | 2189 (2.6%) | 2052 (1.9%) | 243 (2.5%) | 220 (1.8%) |
Expired | 1816 (2.1%) | 2679 (2.5%) | 170 (1.8%) | 265 (2.2%) |
Other | 2210 (2.6%) | 2214 (2.0%) | 241 (2.5%) | 262 (2.2%) |
Primary outcomes | ||||
In-hospital deaths, no. (%) | 1816 (2.1%) | 2679 (2.5%) | 170 (1.8%) | 265 (2.2%) |
30-day readmissions, no. (%) | 9136 (10.7%) | 15,932 (14.6%) | 1013 (10.5%) | 1837 (15.1%) |
Hospital stays at least 7 days, no. (%) | 20,411 (23.9%) | 26,109 (24.0%) | 2145 (22.3%) | 2931 (24.2%) |
No. of ICD-9 diagnoses, median (IQR) | 12 (16) | 10 (10) | 12 (16) | 10 (10) |
