Background Electronic health records from primary care are increasingly utilised as a resource for epidemiological research in cancer. The present study aimed to evaluate the validity of cancer diagnoses recorded in a primary care database compared with cancer registry data in England.

Methods Eligible cancer codes were reviewed and agreed by two epidemiologists with experience in cancer and primary care. The eligible cohort comprised 42,556 participants, registered with English general practices in the General Practice Research Database (GPRD) that consented to cancer registry linkage, who presented with haematuria, haemoptysis, dysphagia or rectal bleeding or were diagnosed with cancer of the lung, urinary tract, oesophagus/stomach, or colon/rectum between 2002 and 2006. Cancer registry (CR) records were linked through a unique identifier by a third party. Cancer registry and primary care records were compared for cancer diagnosis, date of cancer diagnosis and death.

Results There were 5,429 cancer diagnoses in GPRD and 5,710 in the CR, with 5,216 (91% of CR total) diagnosed in both sources. There were 494 (9%) diagnosed in CR but not in GPRD and 213 (4%) that were diagnosed in GPRD but not CR. The predictive value of a GPRD cancer diagnosis was 96% for lung cancer, 92% for urinary tract cancer, 96% for gastro-oesophageal cancer and 98% for colorectal cancer. ‘False negative’ primary care records were sometimes accounted for by registration end dates being shortly before cancer diagnosis dates. The median (interquartile range) difference in date of cancer diagnosis (CR minus GPRD) was -11 (-30 to 6) days. Death records were consistent for the two sources for 3,337/3,397 (99%) of cases.

Conclusion Recording of cancer diagnosis and mortality in primary care electronic records is generally consistent with cancer registrations in England. Linkage studies must pay careful attention to selection of codes to define eligibility and timing of diagnoses in relation to beginning and end of record.