Capturing Social Prescribing in Primary Care Health Records: A Study of Coding Practices in England

Background: Social prescribing has expanded rapidly in England and internationally. However, the collection of consistent data to monitor and evaluate its reach and impact remains to be challenging. Methods: This study used routine data from the Clinical Practice Research Datalink (CPRD) Aurum in England. UpSet plots were used to describe intersections among different social prescribing codes, and between social prescribing and other non-clinical interventions. After identifying potential problematic intersections, we fitted multilevel regression models to examine variation between general practices and to understand factors associated with potentially problematic coding. Findings: Between 2019 and 2025, approximately 3.8 million consultations involving social prescribing codes were recorded in CPRD Aurum. Social prescribing codes were most often recorded individually (38.1%), but intersections existed both among different social prescribing codes, and between social prescribing and related non-clinical support. Many intersections reflect plausible stages and scenarios within the social prescribing pathway. However, some coding patterns appeared ambiguous or potentially inconsistent, particularly those related to social prescribing being offered and referrals being made. Results from the multilevel regression models suggested that 28.4-34.6% of variation in these codes was attributable to differences between practices, and problematic coding varied by region and year of recording. Interpretation: Our findings highlight opportunities to improve the quality and consistency of social prescribing coding in primary care, thereby strengthening service monitoring and evaluation both in England and internationally. They also provide important insights for researchers using CPRD data to investigate social prescribing.