Colorectal cancer incidence on the General Practice Research Database

Rachel Charlton, Julia Snowball, K Bloomfield, Corinne De Vries

Research output: Contribution to journalArticlepeer-review

6 Citations (Scopus)

Abstract

Purpose: The General Practice Research Database (GPRD) contains longitudinal patient medical records collected within UK primary care. This study aimed to identify incident cases of colorectal cancer on the GPRD and to compare incidence rates for 2007 with those reported by the UK cancer registries. Methods: Algorithms were created to identify incident cases of colorectal cancer on the GPRD and cases were required to have additional medical codes to support the diagnosis. Age-specific and sex-specific incidence rates for 2007 were calculated using the GPRD data and compared with those reported by the cancer registries. Results: Trends in colorectal cancer by age and sex were similar for the two data sources; however, the incidence of colorectal cancer on the GPRD was lower than that of the registries, particularly when supporting evidence was required: 57.0 compared with 70.2 per 100000 per year for men and 42.0 compared with 56.6 per 100000 per year for women. Inclusion of cases without supporting evidence still resulted in lower rates but increased the GPRD rates to 63.7 and 48.4 for men and women, respectively. The largest discrepancy was observed in the older age groups. Conclusion: Colorectal cancer rates on the GPRD were lower than those reported by UK cancer registries, especially when requiring supporting evidence in addition to a diagnosis code. It appears that the requirement of supporting evidence on the GPRD for colorectal cancer identification may result in some true cases being excluded, particularly in the very elderly.
Original languageEnglish
Pages (from-to)775-783
JournalPharmacoepidemiology and Drug Safety
Volume21
Issue number7
DOIs
Publication statusPublished - 2012

Fingerprint Dive into the research topics of 'Colorectal cancer incidence on the General Practice Research Database'. Together they form a unique fingerprint.

Cite this