Abstract
Although routine healthcare data are not collected for research, they are increasingly used in epidemiology and are key real-world evidence for improving healthcare. This study presents a method to identify prostate cancer cases from a large English primary care database. 19,619 (1.3%) men had a code for prostate cancer diagnosis. Codes for medium and high Gleason grading enabled identification of additional 94 (0.5%) cases. Many studies do not report codes used to identify patients, and if published, the lists of codes differ from study to study. This can lead to poor research reproducibility and hinder validation. This work demonstrates that carefully developed comprehensive lists of clinical codes can be used to identify prostate cancer; and that approaches that do not solely rely on clinical codes such as ontologies or data linkage should also be considered.