(A-218) Onboarding and characterising NHS health data in the HEalth Data Research UK RWE OMOP Network (HERON-UK)

Background: Real world data (RWD) are a valuable resource for pharmacoepidemiologists, providing insights into the effectiveness and safety of treatments. Characterisation of RWD is essential to ensure the reliability and validity of research findings, helping to identify the strengths and limitations of a data source and enabling researchers to ask the right questions and interpret results.

Objectives: To characterise UK NHS data sources including primary care and hospital records from England and Scotland, recently onboarded to the HEalth Data Research UK Real World Evidence OMOP Network (HERON).

Methods: First, we conducted a benchmark analysis to ensure the database infrastructure was suitable for RWD research. This assessed database connectivity, data type consistency, and ability to perform analytical operations required for computing epidemiological metrics (e.g. incidence) could be performed using existing OMOP analytical pipelines.
Secondly, we utilised the OmopSketch R package to characterise each data source in terms of: population, content of the OMOP tables (source vocabularies and missing data), and record counts.

Results: Seven data partners were onboarded and performed benchmarking analysis, including 6 hospital electronic health records —Barts Health (3M patients), University of Edinburgh - Dataloch (1.8M), Lancashire Teaching Hospitals (1.5M), Great Ormond Street Hospital (GOSH) (204K paediatric), Leeds Cancer Centre (1.5M), and University College London Hospitals (UCLH) (1.1M) -, and a primary care database (49M) linked to hospital records (CPRD AURUM ). All data sources were confirmed to be consistent with the OMOP Common Data Model and successfully completed the database characterisation detailed at https://dpa-pde-oxford.shinyapps.io/HDRUK-Characterisation/. Results, for example, show that the age and sex distribution of the population in observation varied across data partners, reflecting  different care settings and populations. GOSH, as a specialist paediatric provider, had the youngest population in the network, with a median age of 6 years,  DataLoch, Barts, and CPRD Aurum  represented relatively young populations, with median ages between 31 and 33 years. UCLH, Lancashire and Leeds reflected older cohorts, with median ages of 38, 43 and 50 years, respectively .  Across all data partners,  there were slightly more females than males , with the proportion of females ranging from 50.2% to 54.6%.
Regarding follow-up, all data partners—except Lancashire—had a single observation period per person, with median follow-up times ranging from 198 days at UCLH to 2,602 days at DataLoch. In contrast, Lancashire recorded an average of 1.3 observation periods per person, with a median follow-up of 156 days, and gaps between periods separated by a median of 910 days.

Conclusions: We have onboarded 7 UK RWD data partners mapped to OMOP to the HERON network and successfully executed analytics to characterise data and infrastructure of all of them. The results directly informed the development of subsequent study code. This demonstrates the feasibility of federated RWD analyses across the upcoming NHS Secure Data Environments.