Data quality trust: a provenance-based data quality assessment method and its integration with interoperable electronic medical record systems




Stoldt, Jean-Philippe

Journal Title

Journal ISSN

Volume Title



Data quality is a critical requirement for data-driven clinical decision making in modern healthcare. It is a key prerequisite to many clinical analytics applications, yet much of the research to date focuses on assessing data quality of electronic health records for secondary use. New approaches to data quality assessment are needed that allow clinical data users to quickly assess whether the quality of diagnostic data can be trusted for a clinical decision or not. This dissertation proposes a data quality trust model and provenance-based assessment method for considering contextual data quality during clinical decision making at the point of care. Our software engineering approach uses fuzzy logic to infer relative data quality trust from a data user’s trust preferences with respect to agents, data production methods, verification activities, and the certification of agents and data production methods. Extensions to the FHIR interface standard for data quality trust allow for platform interoperability across system contexts. We consider dual process theories to propose user interface extensions that visualize data quality trust for clinical users in heuristic and systematic cognitive processing modes. A visual prototype with an existing SMART on FHIR app for a pediatric hypertension clinical example demonstrates the feasibility of the assessment method with electronic medical record systems and clinical workflows. We show how application of data quality trust with more complex clinical examples, such as diagnosis of food or medication allergies, may enable predictive and prescriptive system functionality to guide diagnostic workflows. Trust model and data quality assessment method may be adapted for other application domains that rely on data quality for data-driven decision making.