Methodological challenges and recommendations for identifying childhood immunisations using routine electronic health records in the United Kingdom

Routinely collected electronic health records (EHR) offer a valuable opportunity to carry out research on immunisation uptake, effectiveness and safety, using large and representative samples of the population. However, using EHR presents challenges for identifying vaccinated and unvaccinated cohorts. Some vaccinations are delivered in different care settings, so may not be fully recorded in primary care EHR. In contrast to other drugs, they do not require electronic prescription in many settings, which may lead to ambiguous coding of vaccination status and timing. Additionally, for childhood vaccination, there may be other challenges of identifying the study population eligible for vaccination due to changes in immunisation schedules over time, different vaccine indications depending on the context (e.g., tetanus vaccination after exposure) and the lack of full dates of birth in many databases of data confidentiality restrictions. In this paper, we described our approach to tackling methodological issues related to identifying childhood immunisations in the Clinical Practice Research Datalink (CPRD) Aurum, a UK primary care dataset of EHR, as an example, and we introduce a comprehensive algorithm to support high-quality studies of childhood vaccination. We showed that a broad variety of considerations is important to identify vaccines in EHR and offer guidance on decisions to ascertain the vaccination status, such as considering data source and delivery systems (e.g., primary or secondary care), using a wide range of medical codes in combination to identify vaccination events, and using appropriate wash-out periods and quality checks to deal with issues of over-recording and back dating in EHR. Our algorithm reproduced estimates of vaccination coverage which are comparable to official national estimates in England. This paper aims to improve transparency, quality, comparability and reproducibility of studies on immunisations.