Integrity Testing: Biomedical Data Sources and Data Federation
Errors due to human mistakes and software faults are unavoidable in complex, dynamic federated environments. The Biomedical Informatics Program (BIP) of the Atlanta Clinical and Translational Science Institute (ACTSI) is addressing this issue by developing a framework of tools and techniques designed to detect errors by combining domain knowledge, modeling, and software testing techniques. The framework implements support for test models, which represent constraints and dependencies associated with federated databases, and a family of testing techniques, which leverage these models to evaluate data integrity and correctness in the federated environment. A test model is a set of rules derived from (1) constraints within and among databases, (2) business processes (e.g., study protocols), (3) user-defined rules, and (4) rules based on domain knowledge. The testing techniques are model-based and will identify relevant test scenarios for the environment. The framework is presently under development using a testbed consisting of several of the ACTSI data repositories. This project is funded by BIP with additional funds from GT and Emory and led by GT (Drs. Orso and Harrold) in collaboration with Emory (Drs. Saltz and Kurc).