| Bias | Author's judgement | Support for judgement |
| Confounding |
Serious |
"Comment: The following review confounders : Age, Gender, Comorbidities (only COPD) Severity (via SF ratio, WCC, Plts) were included and controlled for in the regression model. Review confounders Ethnicity, Diabetes, Obesity, Immunosuppression and Hypertension were measured but not included in the final model. At baseline these confounders were unbalanced, with higher proportions for Diabetes, Obesity, Immunosuppression (via Cancer) in the control group as well as more Non-Hispanic African Americans in the control group. Further, follow up started either when patients were put on corticosteroids or when they had hypoxia. Authors did not adjust for this time-varying confounding, adjustment for confounding variables at baseline is not sufficient." |
| Selection of participants into the study |
Moderate |
Comment: One of the exclusion criterion was "if had any of primary composite outcome within first 24 hours of admission". Start of follow up and start of treatment is reported to be the same. Quote :"The index date was defined as the date when corticosteroid was commenced in the corticosteroid cohort and the day when patientsâ inclusion criteria were met in the non-corticosteroid cohort to minimize the difference in the severity between two cohorts" |
| Intervention classification |
Low |
Comment : Data related to date of first dose, dosages, duration and all other data was extracted manually from electronic medical records, independantly duplicated and checked by two authors (M.M. and I.H.) (Supplemental file Appendix 1). Two independent residents adjudicated all the outcome data, and any disparity was resolved by consulting the primary investigator. Treatment groups were classified using clearly defined criteria (receipt vs non receipt of systemic corticosteroids) Quote : The decision to give corticosteroids was at the discretion of the treating physician. |
| Deviation from intended intervention |
Moderate |
Comment: Reported co-interventions administered were Tocilizumub (unbalanced per arm, included in the model for adjustment), Anticoagulation (unbalanced per arm, included in the model for adjustment), Hydroxychloroquine (balanced in per arm proportions). No information related to other (experimental) co-interventions or the definition for standard of care was reported. |
| Missing data |
Low |
Comment : 205 included in cohort, 205 analyzed. No participants were excluded due to missing data. Risk assessed to be low for the outcomes: Time to death and Time to clinical improvement. |
| Measurements of outcomes |
Low |
Comment : Outcome assessors were likely aware of the intervention received: the outcome Time to death is an objective measure however the outcome Time to clinical improvement (discharge) is an observer-reported outcome requiring clinical judgement.It could theoretically have been influenced by knowledge of the intervention received by the participants, but this is unlikely in the pandemic context. Risk assessed to be low for the outcome : Time to death, Time to clinical improvement. |
| Selection of the reported results |
Moderate |
Comment: The outcomes and analyses are clearly defined in the Methods section. There is no a-priori registered protocol or statistical analysis plan available. Risk assessed to be moderate for the outcome : Time to death, Time to Clinical improvement |