Use this URL to cite or link to this record in EThOS:
Title: Near real-time vaccine safety surveillance using United Kingdom electronic health records
Author: Leite, A.
ISNI:       0000 0004 7428 8713
Awarding Body: London School of Hygiene & Tropical Medicine
Current Institution: London School of Hygiene and Tropical Medicine (University of London)
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis describes the feasibility of implementing a near real-time vaccine safety surveillance system (NRTVSS) using data from the Clinical Practice Research Datalink (CPRD), a United Kingdom (UK) research-level primary care database. NRTVSS is one method in the vaccine safety post-licensure toolkit, used since 2005. To understand how NRTVSS has been applied I conducted a systematic review of studies using NRTVSS. I identified 31 systems, mainly in the USA. Several sequential tests were in use, most commonly the Poisson-based maximized sequential probability ratio test (PMaxSPRT, 44%) and its binomial version (BMaxSPRT, 24%). Only 75% of studies addressed confounding, mainly by adjusting the expected rate. Delays in data availability may hinder the feasibility of implementing a system; some studies delayed the analysis, whilst others adjusted for delays and partially accrued periods. In CPRD, delays in recording outcomes are particularly relevant. Hence, I assessed those delays for selected outcomes of interest for vaccine safety (Bell’s palsy, Guillain-Barré syndrome (GBS), optic neuritis, and febrile seizures (FS)) by comparing the deemed date of diagnosis to the date the event was recorded in the system. Three-quarters of the records accrued during the first month, considered as sufficient to implement NRTVSS. I thus trialled the implementation of a system using previously collected CPRD data, for seasonal influenza/GBS and measles-mumps-rubella/FS. This included power calculations for detecting a signal. I used PMaxSPRT for both vaccine/outcome pairs and BMaxSPRT for measles-mumps-rubella/FS. Both tests were adjusted for delays in recording outcomes, based on the previous analysis. It was possible to implement a system, but power was < 80% to detect less than a four-fold increase in the risk of GBS following influenza vaccine. For this pair, I re-evaluated power after removing delays in data availability, with no significant improvement. This work establishes the foundation of a NRTVSS using CPRD for potential application in the UK. Future research could assess further vaccine/outcome pairs and explore the use of other statistical tests. Overall, this project contributes to UK vaccine pharmacovigilance.
Supervisor: Thomas, S. L. ; Andrews, N. Sponsor: National Institute for Health Research
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral