Studies performed before vaccines are licensed have limited ability to detect rare adverse events, requiring ongoing assessment after their introduction. With the increasing availability of electronic health records, near real-time vaccine safety surveillance using these data has emerged as an option. This type of surveillance commences shortly after a new vaccine is introduced and examines available data on a regular basis (e.g. monthly), using sequential tests. At every time point a decision is made on whether there is an increased risk (of the outcome studied) or whether surveillance needs to be continued. If an increased risk is detected further analyses are needed to confirm it. In the UK, this type of surveillance relying solely on research databases such as the Clinical Practice Research Datalink (CPRD) has not been considered. The purpose of this study is to assess whether CPRD is a good data source for implementing near real-time surveillance for vaccines. This will be achieved by evaluating different potential adverse events and vaccines and deciding which of these events could be monitored using CPRD. The study will thus contribute to improved knowledge of the implementation of this kind of surveillance, helping to strengthen post-licensure vaccine surveillance systems in the UK.
This study comprises a two-stage feasibility assessment of implementing near real-time vaccine safety surveillance (NRTVSS) methods using CPRD. The first stage will include assessments of validity, volume of data, and data accrual for selected outcomes of interest in CPRD (potential adverse events). To study validity, previous validation studies for the outcomes of interest in CPRD will be considered. For volume of data assessment, power calculations for different sequential tests will be performed, considering the incidence and risk window for the outcome. Analyses to describe the data accrual process will include comparisons of the date of recorded diagnosis in CPRD and Hospital Episode Statistics, and looking at last practice collection dates. This first assessment will enable classification of outcomes as potentially feasible or non-feasible (poor validity, insufficient power or unacceptably delayed recording). For outcomes deemed potentially feasible a more detailed feasibility assessment, mimicking a near-real time system, will compare different statistical tests, appropriate control for confounding strategies, and methods to adjust for data accrual lags. A special focus will be on group sequential methods. Existing methods will be adapted to ensure best use in CPRD. For outcomes considered currently non-feasible, changes needed to enable feasibility will be discussed.
Sara Thomas - Chief Investigator - Not from an Organisation
Andreia Leite - Corresponding Applicant - London School of Hygiene & Tropical Medicine ( LSHTM )
Nick Andrews - Collaborator - Public Health England
Philip Bryan - Collaborator - MHRA
Stephen Evans - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
HES Admitted Patient Care