Abstract

The COVID -19 pandemic led to unprecedented progress in making population-level, linked electronic health record (EHR) data available to approved researchers. OpenSAFELY is a secure, transparent, open-source software platform for the analysis of linked EHR data from over 24 million people for projects relating to the COVID -19 pandemic. This and other platforms enabled work addressing urgent, policy-relevant questions, such as estimating the effectiveness of the COVID -19 vaccines outside clinical trials. Estimating causal effects, such as COVID -19 vaccine effectiveness, from EHR data can be helped by specifying the hypothetical ‘target trial’ whose results the observational data analysis is trying to emulate.

The nature of the COVID -19 vaccine rollout and the dramatic variation in the incidence of infection-related outcomes during the pandemic make estimating vaccine effectiveness from EHR data challenging. I will describe approaches to estimating vaccine effectiveness based on such target trials, and results from analyses using these approaches to estimate the effectiveness of first, second and further doses of COVID -19 vaccines, and compared different vaccine brands. This work was conducted using OpenSAFELY under a collaboration between the University of Bristol, the Bennett Institute at the University of Oxford, the London School of Hygiene and Tropical Medicine, and Harvard University.