This research complies with all relevant ethical regulations as surveillance of coronavirus disease 2019 (Covid-19) testing and vaccination is undertaken under Regulation 3 of the Health Service (Control of Patient Information) Regulations 2002 to collect confidential patient information (www.legislation.gov.uk/uksi/2002/1438/regulation/3/made. opens in new tab) under Sections 3(i) (a) to (c), 3(i)(d) (i) and (ii), and 3. The study protocol was subject to an internal review by the Public Health England Research Ethics and Governance Group and was found to be fully compliant with all regulatory requirements ref:CAP-2021-07-UPDATE. Given that no regulatory issues were identified, and that ethics review is not a requirement for this type of work, it was decided that a full ethics review would not be necessary.
A test-negative case–control design was used to estimate vaccine effectiveness in those aged 18 years and over against hospitalisation following a PCR test for SARS-CoV-2, as described previously22,23,24,25. Cases were those testing positive and controls those testing negative by PCR. Effectiveness was assessed using a variety of hospitalisation endpoints designed to differentiate between hospitalisations likely to be because of COVID-19 and those that may be hospitalisation with COVID-19 but potentially due to another cause. Effectiveness against Omicron and Delta was assessed using periods in which these variants were circulating and using the information on sequencing, genotyping and PCR s-gene target.
We linked COVID-19 PCR tests from both community Pillar 2 testing (wider population testing including drive/walk in and home testing) and in-hospital Pillar 1 testing (testing for those with a clinical need, health workers and travel) with vaccination data from the national vaccination register and hospitalisation data from either the Emergency Care Dataset (ECDS) or the Secondary Users Service (SUS). The SUS dataset is a national electronic database for all NHS hospital admissions in England with ICD-10 coded discharge coding for completed hospital stays. The ECDS includes hospital admissions through NHS emergency departments but not elective admissions with the reason for attending emergency care SNOMED coded.
The data sources are described in detail in Supplementary Text 1.
A maximum of one negative test per person within each of the following approximate 3-month periods was selected at random: 26 April to 1 August 2021, 2 August 2021 to 21 November 2021, 22 November 2021 to 23 February 2022. For analyses that involved hospitalised controls, any negative tests that led to hospitalisation within 21 days of a previous hospital negative test were excluded.
Analysis was by logistic regression with the PCR test result as the dependent variable where those testing positives were cases and those testing negative controls. Vaccination status was included as an independent variable, and effectiveness defined as 1- odds of vaccination in cases/odds of vaccination in controls. Vaccination status was defined using date of onset, or, if missing or in Pillar 1 where this was not obtained, date of sample. Status was stratified by dose and interval post vaccination at 0–27 and 28+ days post first dose, 0–13, 14–174 and 175+ days post second dose and 0–6, 7–13,14–34, 35–69, 70–104, 105+ post 3rd dose. The analysis was also stratified by the manufacturer (ChAdOx1: Astrazeneca (adenoviral vector vaccine)or BNT162b2: Pfizer/BioNTech (mRNA vaccine) 2 dose priming, and BNT162b2: Pfizer/BioNTech (mRNA vaccine) or mRNA-1273: Moderna (mRNA vaccine)boosting) and by variant (Delta and Omicron). The analyses done to assess effectiveness according to the specificity of the hospitalisation are given in Table 2. The first analysis replicates those previously done for symptomatic infection and the following analyses using different criteria to allow comparison of emergency care and SUS data sources and to assess within SUS how VE changes based on whether the respiratory code is in the primary diagnostic field, the length of stay and the presence of codes for further interventions (oxygen, ventilator, ICU admission).
Vaccine effectiveness was adjusted in logistic regression models for age (5-year bands), sex, index of multiple deprivation (quintile), ethnic group, care home residence status (for age 65+), geographic region (NHS region), period (calendar week of test), health and social care worker status (for age <65), clinical risk group status (for age <65), clinically extremely vulnerable, severely immunosuppressed, and previously testing positive. The logistic model for the log-odds of the probability of being a positive case therefore included all of these explanatory variables as well as the variable for vaccination status. All analyses were stratified by age 18–64 and 65+. For the vaccine manufacturer stratification, only endpoints 2–5, 9 and 12 were considered and only for Omicron. Numbers were too small in those primed with mRNA-1273 to assess this schedule (Supplementary Table S1).
All analyses were conducted in STATA 17™.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.