Background: The Breast and Prostate Cancer Data Quality and Patterns of Care (POC-BP) Study enabled a reabstraction study of the quality of population-based, central cancer registry data on the characteristics and initial treatment of breast cancer in females and prostate cancer in the United States.
Methods: Stratified random samples of 9,103 female breast cancers and 8,995 prostate cancers were available for the analysis, using the independently reabstracted data as the gold standard to compute measurements of agreement.
Results: A slight majority (53% [8/15]) of the cancer site and treatment combinations showed kappa statistics > or = 0.60 and percent agreements, sensitivities, and predictive values positive > or = 80%: surgery and radiation for the 2 cancers, radiation completed and chemotherapy for breast cancer, and radiation modality and hormone therapy for prostate cancer. The qualities of the Collaborative Stage (CS) site-specific factors and derived variables for the 2 cancers were inconsistent, which confirmed the need to evaluate the recently-implemented CS algorithm.
Conclusion: The data quality analysis from POC-BP underscores the importance of examining the quality of specific data variables by cancer site, thereby highlighting those variables for which data collection procedures could be improved.