QResearch primary care records linked to national hospital and cancer registry data: a validation study.

BACKGROUND: Population-based linkage of primary care electronic health records to oncology-relevant datasets enables detailed analysis of cancer care and outcomes. QResearch is a large primary care database containing longitudinal health records from a subset of general practices in England, linked to national hospital, cancer registry, treatment and mortality data. However, the extent to which this primary care-hospital-registry linkage accurately reflects the national cancer population and cancer treatment patterns has not been validated. MATERIALS AND METHODS: This validation study included adults aged ≥18 years registered at QResearch-contributing practices with an incident diagnosis of breast, prostate, colorectal, melanoma, kidney, pancreatic or gastro-oesophageal cancer between 2013 and 2020. Demographic characteristics and proportions treated with systemic anticancer therapy (SACT) or radiotherapy (RT) were compared with national statistics. We evaluated, in the absence of the cancer registry, the ascertainment and representativeness of patients with cancer identifiable in other sources. RESULTS: Between 2013 and 2020, the QResearch-linked dataset captured 14.6% (n = 196 565/1 348 243) of national cancer registrations across seven cancer sites. Distributions of sex, age and stage closely matched national data, with regional variation reflecting participating practices. Proportions treated with SACT and RT aligned with national estimates, with differences typically within 2%. The 93.8% of cases identifiable without registry data remained demographically and clinically representative, with minimal impact on overall cohort characteristics. CONCLUSION: The QResearch-linked dataset offers a nationally representative and comprehensive platform for population oncology research. Its integration of detailed primary care information, including comorbidities, medications and lifestyle factors, should enable evaluation of treatment predictors and outcomes across diverse populations.