Clinical Practice Research Datalink (CPRD) collects anonymised patient data from a network of general practices across the UK.
This study is designed to describe the representativeness of CPRD primary care data in relation to the general population of the UK. We will compare the distribution of characteristics in the populations of CPRD to those of the UK using data from the UK Censuses 2011 and 2021. We will compare these populations in terms of geography, age, sex, and ethnicity.
We will assess the completeness of data in CPRD related to other factors that influence health and care (e.g., national identity, language, religion, sexual orientation, gender identity, housing, health/disability, unpaid care, etc.), meaning the proportion of patients in CPRD that have data pertaining to the relevant characteristic. If there is sufficient data pertaining to a given characteristic, the representativeness of CPRD compared to the UK Censuses will be assessed.
This project will provide a better understanding of the utility of the CPRD databases in representing the UK population and will serve to improve interpretation of research results and translation of those results into practice for patients.
This study will be a retrospective cohort study of the representativeness of the Clinical Practice Research Datalink (CPRD) GOLD and Aurum databases in relation to the UK populations at the time of the 2011 (27 March 2011) and 2021 UK Censuses (21 March 2021 for England, Wales, and Northern Ireland; 20 March 2022 for Scotland).
The following characteristics will be assessed for representativeness: geographic distribution, age, sex, and ethnicity. The distribution of these characteristics in each database and in relevant UK geographies will be compared using Pearson’s Chi-squared. The characteristic distribution of each database and population will be compared to the characteristic distributions from the Censuses.
Additionally, the availability of data pertaining to wider social determinants of health will be assessed in CPRD, including data on national identity, language, religion, sexual orientation, gender identity, housing, health/disability, unpaid care, etc. Completeness will be assessed as the count and proportion of acceptable, currently registered patients with a usable characteristic recording. Usability will be assessed as having at least one characteristic recording, other than ‘not stated’ or ‘not known’. If there is ≥60% completeness of these additional characteristics, the representativeness of that characteristic will be assessed, as described above.
Completeness and representativeness will be stratified by age, sex, and geographic regions.
This project will benefit patients and researchers by providing up-to-date information on the representativeness of the CPRD GOLD and CPRD Aurum databases in terms of the UK population to inform choice of data sources, interpretation of results, and translation of those results into practice for patients.
Distribution of and/or proportion of complete data for the following characteristics in the acceptable, currently registered populations of CPRD GOLD and CPRD Aurum:
- Geography (UK, GB, England, Scotland, Wales, Northern Ireland, English regions)
- Age
- Sex
- Ethnicity
- National identity
- Immigration status
- Language
- Religion
- Sexual orientation
- Gender identity
- Housing
- Health/disability
- Unpaid care
Eleanor Axson - Chief Investigator - CPRD
Eleanor Axson - Corresponding Applicant - CPRD
Abbie Hunt - Collaborator - CPRD
Alice Hinchliffe - Collaborator - CPRD
Chisomo Mutafya - Collaborator - CPRD
Justin Chan - Collaborator - CPRD
Kathleen White - Collaborator - CPRD
Mia Harley - Collaborator - CPRD
Rachael Williams - Collaborator - CPRD
Sonia Coton - Collaborator - CPRD
CPRD Aurum Ethnicity Record;CPRD GOLD Ethnicity Record