Estimating the effectiveness of COVID-19 vaccination against COVID-19 hospitalisation and death: a cohort study based on the 2021 Census, England.

ObjectiveTo estimate the effectiveness of COVID-19 vaccination against hospitalisation for COVID-19 and death involving COVID-19 in England using linked population level data sources including the 2021 Census. DesignRetrospective cohort study. SettingEngland, 21 March 2021 to 20 March 2022. ParticipantsIndividuals alive and aged 16+ on 21 March 2021, resident in England, enumerated in the 2021 Census as a usual resident, and able to link to an NHS number. A sample of 583,840 individuals was used for the analysis. ExposuresCOVID-19 vaccination: first dose, second dose and third dose/first booster dose, with categories for time since each dose. Main outcome measuresHospitalisation for COVID-19 or death involving COVID-19. An adjusted Cox proportional hazard model was used to estimate the hazard ratio for the outcomes for vaccinated participants for different doses and time since dose compared to unvaccinated individuals. Vaccine effectiveness was estimated as (1-hazard ratio)x100%. A control outcome of non-COVID-19 death was also assessed. ResultsVaccine effectiveness against hospitalisation for COVID-19 was 52.1% (95% confidence interval 51.3% to 52.8%) for a first dose, 55.6% (55.2% to 56.1%) for a second dose and 77.6% (77.3% to 78.0%) for a third dose, with a decrease in vaccine effectiveness 3+ months after the third dose. Vaccine effectiveness against COVID-19 mortality was 58.7% (52.7% to 63.9%) for a first dose, 88.5% (87.5% to 89.5%) for a second dose and 93.2% (92.9% to 93.5%) for a third dose, with evidence of waning 3+ months after the second and third doses. For the second dose, which is the most comparable across the different time-periods, vaccine effectiveness was higher against COVID-19 hospitalisation but slightly lower against COVID-19 mortality in the Omicron dominant period than the period before the Omicron variant became dominant. Vaccine effectiveness against both COVID-19 hospitalisation and mortality was higher in general for mRNA vaccines than non mRNA vaccines, however this could be influenced by the different populations given each vaccine vector. Non-zero VE against non-COVID-19 mortality indicates that residual confounding may impact the results, despite the inclusion of up-to-date socio-demographic adjustments and various sources of health data, with possible frailty bias, confounding by indication and a healthy vaccinee effect observed. ConclusionsThe vaccine effectiveness estimates show increased protection with number of doses and a high level of protection against both COVID-19 hospitalisation and mortality for the third/booster dose, as would be expected from previous research. However, despite the various sources of health data used to adjust the models, the estimates for different breakdowns and for non-COVID-19 mortality expose residual confounding by health status, which should be considered when interpreting estimates of vaccine effectiveness.