Assessing the validity of diagnostic information in administrative health care utilization data: experience in Saskatchewan.

Administrative health care utilization databases have been promoted as resources for pharmacoepidemiologic research for several years, but the validity of the diagnostic data in many systems has not been thoroughly investigated. This work outlines issues in assessing the validity of diagnostic information from a broad perspective, using results from the datafiles of the province of Saskatchewan to illustrate them and, where possible, contrasting them with findings from other databases. An evaluation of the validity of health care utilization data should be more than a simple check of coding and data entry errors. Comparisons of data records with information external to the computer system, an examination of the internal consistency of the data (especially in systems consisting of several data sources) and, where practicable, an assessment of the cohesiveness of the data within the context of current medical knowledge should be performed on a broad range of diagnoses and procedures. Careful consideration should also be given to the generalizability of results from the database. Summaries of such evaluations for all databases used for population-health research purposes should be widely disseminated. We caution researchers to continue to seek ways to validate their results in each use of these databases.