IJPDS International Journal of Population Data Science IJPDS 2399-4908 Swansea University 10.23889/ijpds.v10i3.3123 10:3:104 Developing a Research-Ready-Data-Asset (RRDA) for Welsh primary care data within the SAIL Databank: enhancing data quality and reproducible research. Abbasizanjani Hoda 1 Bedston Stuart 1 Akbari Ashley 1 Swansea University, Swansea, United Kingdom 01 06 2025 2025 8 4 3123 This work is licenced under a Creative Commons Attribution 4.0 International License. This article is available from the IJPDS website at: https://ijpds.org/article/view/3123 Objectives

We aimed to develop a high-performance RRDA for the Welsh Longitudinal General Practice (WLGP) data to standardise curation, enhance reproducibility, improve query performance and add additional value/features for research. The RRDA provides a curated normalised asset with a comprehensive clinical code look-up and assigned activities type.

 Methods

WLGP data has a long-format event-list structure with potential data quality issues, including duplicates, re-inserted GP-to-GP-transferred records, and missing/invalid entries. To address these, the RRDA involves three steps: data cleaning, data curation using patient's GP registration history from demographic data, and transforming data into a structured, normalised format to eliminate redundancy and support faster, flexible large-scale queries.

The WLGP-RRDA includes a look-up of primary care official/local codes (Read-V2/SNOMED/EMIS/Vision). Additionally, we implemented a four-layer approach for identifying healthcare providers, patient access mode, interaction type, and details of individual codes to capture the complexity of activities, enabling patient-practice interaction analysis.

 Results

Curating WLGP data (1990-2024, 4,565m records, 5m people) revealed significant improvements in data quality and completeness over time, with data retaining rates after cleaning/curation increased from 38% to 94%. Similarly, patient inclusion in WLGP-RRDA improved from 43% to 98% during the same period, indicating improved data accuracy and Welsh residents coverage.

The normalisation process resulted in an efficient three-table structure with unique integer keys for clinical codes and events, optimising database performance/scalability. The extensive clinical code look-up improved coverage of events with known descriptions, showing increased local/SNOMED code use since the pandemic.

Additionally, implementing a multi-layered approach to identify interaction types (e.g., face-to-face/remote consultations) using official/local code hierarchies enabled analysis of national trends in GP activities and impact of the pandemic.

 Conclusion

The WLGP-RRDA development enhanced data quality and streamlines the research processes through a reproducible, maintainable, standardised curation and a multi-layered approach to extract activity types. This methodology/RRDA benefits SAIL users and wider across other environments with similar data, through our shared resources to promote transparency and collaboration.