Table 3.
Characteristics of the Included Scores
| Characteristics | All | Level | |
|---|---|---|---|
| N = 242 n (%)e |
Level 1d, N = 193 n (%)e |
Level 2, N = 49 n (%)e |
|
| Category | |||
| 1 First/early contact with healthcare facility ➝ Fatal outcome | 100 (41.3) | 79 (40.9) | 21 (42.9) |
| 2 First/early contact with healthcare facility ➝ Deterioration | 112 (46.3) | 94 (48.7) | 18 (36.7) |
| 3 Severe disease or ICU admission ➝ Deterioration or fatal outcome | 13 (5.4) | 12 (6.2) | 1 (2.0) |
| 4 First diagnosis and contact with outpatient healthcare facility ➝ Hospitalization | 14 (5.8) | 5 (2.6) | 9 (18.4) |
| 5 Acute infection ➝ PCC | 3 (1.2) | 3 (1.6) | 0 (0.0) |
| Study design | |||
| Prospective | 33 (13.6) | 30 (15.5) | 3 (6.1) |
| Retro- and prospective | 12 (5.0) | 3 (1.6) | 9 (18.4) |
| Retrospective | 182 (75.2) | 150 (77.7) | 32 (65.3) |
| Unknown | 15 (6.2) | 10 (5.2) | 5 (10.2) |
| Multicenter design | |||
| ≥ 2 centers | 103 (42.9) | 54 (28.3) | 49 (100.0) |
| Samples size | |||
| Cumulative number of participants ≥1000 | 87 (36.0) | 47 (24.4) | 40 (81.6) |
| Estimated events per variablea (median, IQR) | … | … | 15.6 (IQR = [6.6, 267.3]) |
| Health sector | |||
| Hospitals/emergency department | 216 (89.6) | 182 (94.8) | 34 (69.4) |
| In- or outpatient sites | 16 (6.6) | 3 (1.6) | 13 (26.5) |
| Outpatient sites | 7 (2.9) | 5 (2.6) | 2 (4.1) |
| Other | 2 (0.8) | 2 (1.0) | 0 (0.0) |
| Population | |||
| Patients in the emergency department | 25 (10.3) | 18 (9.3) | 7 (14.3) |
| Inpatients with severe disease | 37 (15.3) | 35 (18.1) | 2 (4.1) |
| Inpatients without restriction to specific conditionsb | 158 (65.3) | 132 (68.4) | 26 (53.1) |
| Inhabitants of one region | 1 (0.4) | 1 (0.5) | 0 (0.0) |
| Out- and inpatients | 11 (4.5) | 2 (1.0) | 9 (18.4) |
| Outpatients | 10 (4.1) | 5 (2.6) | 5 (10.2) |
| Study/recruitment time | |||
| 2020 | … | … | 38 (77.6) |
| 2020–2021 | … | … | 4 (8.2) |
| 2020–2022 | … | … | 7 (14.3) |
| Country | |||
| China | 45 (18.6) | 39 (20.2) | 6 (12.2) |
| Italy | 25 (10.3) | 24 (12.4) | 1 (2.0) |
| United States | 33 (13.6) | 18 (9.3) | 15 (30.6) |
| Other | 139 (57.4) | 112 (58.0) | 27 (55.1) |
| Timing of predictor measurement | |||
| Admission to hospital or emergency department | 190 (79.8) | 159 (84.1) | 31 (63.3) |
| Admission to ICU | 7 (2.9) | 6 (3.2) | 1 (2.0) |
| SARS-CoV2 testing/diagnosis | 13 (5.5) | 12 (6.3) | 1 (2.0) |
| Other | 28 (11.8) | 12 (6.3) | 16 (32.7) |
| Outcomes | |||
| Deterioration (composite, with fatal outcomes) | 109 (45.0) | 91 (47.2) | 18 (36.7) |
| Fatal outcomes (single endpoint) | 116 (47.9) | 94 (48.7) | 22 (44.9) |
| Hospitalization | 14 (5.8) | 5 (2.6) | 9 (18.4) |
| Post-acute COVID syndrome | 3 (1.2) | 3 (1.6) | 0 (0.0) |
| Handling of missing values | |||
| Any imputation method applied | … | … | 19 (38.8) |
| Multiple imputation | … | … | 11 (22.4) |
| Modeling technique | |||
| (Cox, (Bayesian) Logistic, LASSO) Regression | … | … | 41 (83.7) |
| Machine learning | … | … | 2 (4.1) |
| Mixed methods or other | … | … | 6 (12.2) |
| Validationc | |||
| Separate cohort present | 138 (57.0) | 89 (46.1) | 49 (100) |
| Geographical validation | … | … | 10 (20.4) |
| Temporal validation | … | … | 17 (34.7) |
| Temporal and geographical validation | … | … | 7 (14.3) |
| Random split | … | … | 13 (26.5) |
| Validation with different population characteristics | … | … | 1 (2.0) |
| Independent external validation | … | … | 2 (4.1) |
| Discrimination | |||
| AUC of the strongest validation ≥ 0.75 | 190 (78.5) | 141 (73.1) | 49 (100.0) |
| AUC (median, IQR) | 0.83 (IQR = [0.77, 0.90]) | 0.84 (IQR = [0.77, 0.91]) | 0.81 (IQR = [0.80, 0.85]) |
| Calibrationc | |||
| Any method applied | … | … | 30 (61.2) |
| Calibration plot or table | … | … | 23 (46.9) |
| Hosmer-Lemeshow | … | … | 12 (24.5) |
| Application | |||
| Formula | 65 (26.9) | 65 (33.7) | 0 (0.0) |
| Points-based and formula | 172 (71.1) | 123 (63.7) | 49 (100.0) |
| Formula | 3 (1.2) | 3 (1.6) | 0 (0.0) |
| Other | 2 (0.8) | 2 (1.0) | 0 (0.0) |
We present n (%) for categorical information and the median (IQR) for continuous information. The column “All” includes all scores fulfilling the a priori inclusion criteria. In contrast, Level 1 merely includes scores that did not fulfill the selection criteria and Level 2 only includes the scores fulfilling the criteria (see Methods section). As a result of two granularity levels of data extraction, some information is only available for Level 2 scores.
Abbreviations: AUC, area under the curve; COVID, coronavirus disease; ICU, intensive care unit; IQR, interquartile range; PCC, post-COVID-19 condition; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.
aEvents per variable (EPV) were estimated using the absolute number of candidate predictors. Some studies did not precisely name the number of candidate predictors. To generate assumptions regarding the sample size, we counted predictors indicated as candidates in tables or texts (signed by “∼” in Supplementary Table 2), even though we acknowledge that using the number of regression coefficients instead is more precise [15].
bRegarding population characteristics, “severe disease” includes ICU patients and patients with respiratory complications, pneumonia, intubation, or other severe conditions.
cMultiple options possible.
dLevel 1 (L1) includes those scores among “all” scores that did not fulfill the Level 2 selection criteria.
eOr median with IQR.