Death recording in the Clinical Practice Research Datalink

Abstract

For observational research studies, the ability to accurately determine follow‐up time is pivotal in epidemiological evaluations and therefore records on vital status (dead versus alive) and the associated date are crucial. The recording of deaths in UK primary care records is subject to question since there is no direct link between the general practice and the national registry of deaths, and there may be limited incentive for correct recording of vital status and the date of death in the record. Individual level linkage to other data sources offers the opportunity to add supplementary information. However, linked data are only available for a subset of the population, based on consenting GP practices in England. This restriction reduces the potential cohort size for a study and adds complexity to the generation of the appropriate denominator. The overall thesis objective was to evaluate the use of UK GP data for analysing death information in pharmacoepidemiological research by reviewing the CPRD GOLD as a data source, conducting case studies, assessing the impact of using linked data, and evaluating the agreement with national death registration data in capturing vital status and the date of death. This research found that information on deaths may not be accurately recorded in primary care records. Concordance with the official death certificate is good, however the primary care data will often have a date of death recorded that is later than the date on the death certificate. This will have a larger impact on studies of person time.