Facilitating health data sharing: The Connecticut Health Information Network's innovative approach to data integration and data dissemination

Background. The Connecticut Health Information Network (CHIN) provides an innovative technological platform for promoting data sharing among state agencies, hospitals, and healthcare providers. CHIN offers researchers, policymakers, and health practitioners the opportunity to access de-identified data from multiple sources for purposes of research, program evaluation, planning, and monitoring.

Objective. This presentation provides an overview of the CHIN infrastructure, featuring the network's innovative approaches to data integration and data dissemination.

Methods. Synthetic and real health datasets were used to assess the adequacy of two features of CHIN: (1) a canopy clustering algorithm used for probabilistic integration of multiple datasets that do not contain unique numeric identifiers; (2) a method for quantifying disclosure risk with ostensibly de-identified data, particularly in datasets that contain outliers or rare elements (e.g., rare diagnoses) that may pose threats to privacy in the absence of personal information.

Results. Results indicate that CHIN's probabilistic matching approach provides both an accurate and efficient solution for the simultaneous integration of 3 or more data sources. Further, use of auxiliary information from databases containing common misspellings and nicknames significantly improves matching accuracy. Second, tests of a probabilistic differential privacy scheme developed by the authors suggest that configurations of data that may compromise individual privacy (even when personal information has been removed) can be identified and used to mitigate the risk of disclosure.

Conclusions. The Connecticut Health Information Network provides innovative solutions to thorny problems in data integration and data dissemination that hinder efforts to promote health data sharing.