Define the clinical event and repeated-record inclusion criteria in the BCG analysis

Determine which event terms and repeated patient records should be included when defining COVID-19 severity in the BCG analysis, so that the outcome corresponds to the stated clinical endpoint rather than unrelated adverse events.

Background

The paper reports that clinical regression estimates depend on the case definition, treatment of repeated records from the same participant, analysis population, category coding, and reference groups. In the BCG analysis, the question concerns COVID-19 severity, whereas the reference notebook defines each participant’s severity using the highest value across all adverse-event records, including injection-site events unrelated to COVID-19.

The unresolved issue is which event terms and repeated records should define the clinical outcome. Different inclusion rules can change both the magnitude and direction of the estimated association, making the result non-unique until the endpoint and record-handling criteria are specified.