IGF::OT::IGF STATISTICAL RECORD LINKAGE OF HEALTH DATA

Investigators

Abstract

The cancer surveillance data collected and released by the NCI?s SEER program has served as a cornerstone in describing national trends in cancer incidence and survival in the United States. The recent advances in medicine call for collecting more detailed information on patient and tumor characteristics, as well as treatment for evaluating cancer control interventions. However, because SEER relies heavily on data abstracted from hospitals? medical records, information about medical tests or treatments that are likely to be given at doctor?s office, is often not fully captured by SEER. Linking records of individual patients in SEER and health databases, for example medical claims data that have more detailed and complete therapy information provides a unique opportunity to augment the SEER data. Because the quality of the linked data heavily relies upon the methods that one uses to establish the linkage, it is important to evaluate record linkage methods and practices to ensure the linked data can be used for providing valid answers to scientific research. The efficiency and quality can be improved if such additional data can be acquired and linked to existing cancer registry records through the SEER Data Management System (SEER*DMS). SEER*DMS is an NCI centrally-managed system deployed to SEER registries to import, edit, link, consolidate and report data. Record linkage is an integral component of SEER*DMS. However, the current deterministic linkage algorithm as implemented in SEER*DMS may be inferior to the probabilistic algorithm as commonly implemented in stand-alone software, such as Census? BigMatch and CDC?s Linkplus. True matches may not be identified especially when the linking variables are not accurate. The purpose of this project is to provide recommendations to the NCI about methods and approaches to link medical claims data with traditional cancer registry data using SEER*DMS, and design evaluation studies to identify areas for improvement on SEER*DMS linkage algorithm.