Same question, different data source, different answers? Data source agreement for surgical procedures on women with breast cancer.

This study assessed the accuracy of the Manitoba Cancer Registry (MCR) and two administrative data sources, the Manitoba Health hospital discharge file and the Manitoba Health medical claims file, for capturing surgical procedures related to the treatment of breast cancer. The study cohort included all women diagnosed in Manitoba with invasive or in situ breast cancer between 1995 and 1999. The surgical procedures of interest were mastectomy, breast conserving surgery and axillary node dissection. Analysis focused on assessing concordance between data sources following record linkage. Agreement was measured using the kappa statistic, and chart reviews of discordant information were completed to identify the more reliable data source and to validate data files. The effect of using each data set alone to calculate procedure rates was determined to identify any clinically important differences arising from the choice of data source. Results indicate that capture of breast cancer patients using administrative data sets alone can be quite good and that the population-based cancer registry is superior to other administrative data sets for capturing surgical treatment information on cancer cases.