Using hospital discharge files to enhance cancer surveillance

Abstract

Use of the traditional mechanism for cancer surveillance, hospital-based registries, may limit ascertainment of incident cases. In this study, the authors evaluated the ability of a statewide hospital discharge file (HDF) to enhance central cancer registry reporting. Incident cancers from a Virginia cancer registry were linked with an HDF for 1995. Medical record abstractions for over 2,000 cancers verified HDF and registry data. There were 19,740 unique cases ascertained from the two combined data sources. The registry captured approximately 83% of cases, while the HDF captured 62%. Although the HDF missed a substantial number of registry cases, the HDF positive predictive value for identifying the correct cancer site was 94%. Logistic regression was used to identify significant characteristics of cases likely to be captured only by the HDF; these characteristics included hospital cancer program certification, the position of the cancer diagnosis on the claim, and cancer surgery. This study represents the evaluation of a novel approach to enhancing registry completeness and accuracy using statewide HDFs. The results strongly suggest that neither a central cancer registry nor an HDF is a sufficient source for complete capture of cases. Using HDFs to supplement a central cancer registry may be a valuable and efficient method for improving registry completeness of reporting.