Table.
Characteristics of the three-site cohort used for model training and testing
|
Attended a long COVID clinic |
Did not attend a long COVID clinic |
||||
|---|---|---|---|---|---|
| Hospitalised (n=428) | Not hospitalised (n=169) | Hospitalised (n=15 193) | Not hospitalised (n=58 182) | ||
| Sex | |||||
| Female | 237 (55·4%) | 127 (75·1%) | 8465* (55·7%) | 34 771 (59·8%) | |
| Male | 191 (44·6%) | 42 (24·9%) | 6716* (44·2%) | 23 258 (40·0%) | |
| Unknown | 0 | 0 | <20 | 153 (0·3%) | |
| Race | |||||
| Asian | <20 | <20 | 571 (3·8%) | 1361 (2·3%) | |
| Black | 190 (44·4%) | 31 (18·3%) | 3207 (21·1%) | 6370 (10·9%) | |
| Native Hawaiian or Pacific Islander | <20 | <20 | 43 (0·3%) | 138 (0·2%) | |
| Other | <20 | <20 | 85 (0·6%) | 254 (0·4%) | |
| Unknown | 81 (18·9%) | 27 (16·0%) | 2695 (17·7%) | 9842 (16·9%) | |
| White | 142 (33·2%) | 107 (63·3%) | 8592 (56·6%) | 40 217 (69·1%) | |
| Ethnicity | |||||
| Hispanic or Latino | 69* (16·1%) | 26* (15·4%) | 3064 (20·2%) | 11 416 (19·6%) | |
| Not Hispanic or Latino | 354* (82·7%) | 128* (75·7%) | 11 869 (78·1%) | 45 119 (77·5%) | |
| Unknown | <20 | <20 | 260 (1·7%) | 1647 (2·8%) | |
| Age, years | |||||
| 18–25 | <20 | <20 | 790 (5·2%) | 7573 (13·0%) | |
| 26–45 | 86* (20·1%) | 75 (44·4%) | 3824 (25·2%) | 22 732 (39·1%) | |
| 46–65 | 188* (43·9%) | 69 (40·8%) | 5249 (34·5%) | 19 015 (32·7%) | |
| ≥66 | 147* (34·3%) | <20 | 5330 (35·1%) | 8862 (15·2%) | |
| Mean age (SD) | 58·29 (15·03) | 48·14 (14·02) | 56·50 (18·60) | 45·92 (17·32) | |
| Pre-COVID-19 comorbidities | |||||
| Diabetes | 86 (20·1%) | <20 | 2412 (15·9%) | 4842 (8·3%) | |
| Chronic kidney disease | 70 (16·4%) | <20 | 1721 (11·3%) | 2272 (3·9%) | |
| Congestive heart failure | 48 (11·2%) | <20 | 960 (6·3%) | 1133 (1·9%) | |
| Chronic pulmonary disease | 45 (10·5%) | 29 (17·2%) | 1415 (9·3%) | 3698 (6·4%) | |
Data are n (%) unless otherwise stated. All patients shown had acute COVID-19. Diabetes was not separated by type.
*
In accordance with the N3C download policy,21 for demographics where small cell sizes (<20 patients) could be derived from context, we have shifted the counts by a random number between 1 and 5. The accompanying percentages reflect the shifted number.