Testing Health-Care Integrated Systems with Anonymized Test-Data Extracted from Production Systems

Abstract

Testing of data-centric health-care integrated systems involve numerous non-traditional testing challenges, particularly in the areas of input validation, functional testing, regression testing, and load testing. For these and other types of testing, the test-data suites typically need to be relatively large and demonstrate characteristics that are similar to real-data. Generating test-data for integrated system is problematic because records from different systems need to be inter-related in realistic and less-than perfect ways. Using real-data is also not a feasible choice, because health-care data contains sensitive personal identifying information (PII). As a foundation, this paper provides a classification of testing challenges for health-care integrated systems and a comparison of anonymization techniques. It also narrates our experiences with a test-data creation tool [13] that extracts and anonymizes loosely correlated slices of data from multiple operational health-care systems while preserving those real-data characteristics, discussed under the classification scheme.