4. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (all‐cause mortality).

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Serious risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Serious risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. To minimise the impact of the absence of randomisation, an adjusted analysis with propensity scores was performed considering age, sex, disease severity, admission to ICU and COVID‐19 treatment. However, the essential confounding factors: 'participants already using anticoagulants', 'participants who underwent surgery during the hospitalisation', 'active cancer treatment', 'concomitant antiplatelet use' and 'history of venous thromboembolism' were not considered.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used. It is not clear how prevalent use of anticoagulation was handled.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data possibly could not cause an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study has some important problems.
Rentsch 2020	Moderate risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Moderate risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and history of venous thromboembolism, were not considered. However, an appropriate analysis method to control for measured confounders was used (inverse probability of treatment weighting), and all the important confounding domains for this study were probably controlled.	Participants included in both groups were selected from a nationwide cohort of patients receiving care in the Department of Veterans Affairs in the USA, and selection may have not been related to intervention and outcome. The start of follow‐up and start of intervention coincided for most participants (the first 24 h of hospitalisation).	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is sound for a non‐randomised study with regard to this domain but cannot be considered comparable to a well‐ performed randomised trial.
Santoro 2020	Critical risk	No information	Serious risk	Low risk	Low risk	Low risk	Serious risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and antiplatelet use were not considered. The Cox's multivariable regression analysis was performed to define independent risk factors for the mortality outcome, but only for participants with respiratory failure.	Insufficient information to judge. There was insufficient information if the start of follow‐up and the start of intervention coincided for most participants.	The intervention groups were not clearly defined and recorded at the start of the intervention. Information about frequency and dose was not provided.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was available, but it is not possible to exclude bias in selection of reported effect estimate, based on the results, from multiple outcome measurements within the outcome domain (mortality), and multiple effect estimates for different subgroups were provided, omitting varying proportions of the original cohort.	The study is too problematic to provide useful evidence.
COVID‐19: coronavirus disease 2019; ICU: intensive care unit