List of sensitivity analyses
| Sensitivity analysis | Justification |
| 1. Additionally adjusted for ethnicity in multivariable-adjusted models. | In the main analysis, we did not adjust for ethnicity as it was not anticipated to be a strong confounder and due to a sizeable proportion of individuals with missing ethnicity (~23%). We undertook complete case analysis to address missing data. |
| 2. Additionally adjusted for the number of primary care consultations and A&E attendance in the past year in multivariable-adjusted models. | To explore the impact of healthcare-seeking behaviours. |
| 3. For covariate of diabetes severity, we separated people with diabetes diagnosis and HbA1c measures ≥58 mmol/mol and those with diabetes diagnosis but without HbA1c measures in the past year into two different categories. | People with a diabetes diagnosis but not having HbA1c measures in the past year are likely to have uncontrolled diabetes due to their potential lack of monitoring and management of diabetes. Therefore, we classified these people as uncontrolled diabetes in the main analysis. This is an analysis to test the sensitivity of the results. |
| 4. Repeated main analysis with a choice of covariates selected by a DAG approach (post hoc analysis). | To test the robustness of the results by choosing a set of covariates that are confounders with the use of a structured visual presentation (online supplemental figure S2). |
| 5. Repeated main analysis varying the definition of currently prescribed an NSAID to within 2 months of 1 March 2020. | To assess the sensitivity of exposure definition. |
| 6. Repeated main analysis excluding indometacin from all NSAIDs as the exposure of interest. | Indometacin was the only NSAID that was suggested to have antiviral activity against SARS virus.13 |
| 7. Repeated main analysis without censoring people who were prescribed NSAIDs after study start date in the non-use group. | To examine data as an intention-to-treat analysis, in order to limit potential bias due to informative censoring. |
| 8. Repeated main analysis excluding people ever prescribed aspirin before study start date. | To assess the sensitivity of exclusion criteria. |
DAG, directed acyclic graph; NSAID, non-steroidal anti-inflammatory drug.