Geographic Partitioning Techniques for the Anonymization of Health Care Data

As the demand for the availability of detailed health care data sets continues to increase, organizations are faced with the conflicting interests of releasing this important information while protecting the confidentiality of the individuals to whom the data pertains. A major concern when releasing health care data is the geographic information which has a large influence on the re-identifiability of the data and yet is essential for many research applications. In this work, a novel system for data anonymization is presented. At the core of the system is the aggregation of an initial regionalization guided by the use of a Voronoi diagram. Testing is conducted via an implementation designed to run and analyze the results of the various combinations of approaches. In addition, a comparative test is conducted with another system, GeoLeader. It is shown that our system is capable of producing comparable results with a much faster running time.