Methods for Examining Data Quality in Healthcare Integrated Data Repositories
Co-organizers: Vojtech Huser, Michael Kahn, and Jeffrey Brown
Large Integrated Data Repositories (IDRs) have become indispensable for clinical research. Recent emergence of Common Data Models (CDMs) facilitated creation of tools that provide syntactic integration (shared information model) and in some cases also semantic integration (shared set of target terminologies used by structured data). Retrospective data analyses are increasingly being executed on multiple datasets, and distributed research networks are creating reusable tools that streamline data wrangling, data repository maintenance, and data analytics. Examples of large, well-coordinated IDRs developed using a CDM and and distributed network approach include the Vaccine Safety Datalink (9 sites focused on vaccine safety), the Health Care Systems Research Network (multi-purpose research network of 18 sites), the FDA Sentinel Initiative (18 sites representing billions of medical encounters to support medical product safety surveillance), MDPHnet (public health surveillance network in Massachusetts), and PCORnet (over 70 sites with millions of encounters to support clinical research). Each of these distributed networks has a unique approach to addressing data quality, including some shared approaches, and each has developed tools to facilitate data quality querying. However, the various data quality approaches, and tools, of these networks have not all been well-documented and\or are not readily available or easily usable by others. Development of well-documented and readily-available data quality software tools and methods is an emerging need to support use in the new data environments being developed to support clinical research. For example, the Achilles tool created by the OHDSI (Observational Health Data Sciences and Informatics) consortium allows sites that converted their data to the Observational Medical Outcomes Partnership (OMOP) CDM to readily execute a set of data quality rules and data characterization pre-computations. OHDSI supports active open community engagement in developing new tools or adding new functionality to existing tools which has enabled the Achilles tool to expand in data quality assessment capabilities based on community needs and interests.
Data Quality Topics in Scope
The workshop goal is to exchange novel approaches for evaluating data quality (DQ) and innovative ways of reporting DQ findings in a standardized, readily accessible format across multiple data partners. Specific topics are:
- Standardized data quality algorithms
- Data visualisation approaches that facilitate rapid data quality assessment (DQA)
- Creation of novel user interfaces for data custodians and data users to examine data quality
- Development of an international community focused on standardized data quality assessment for meaningful comparisons across partners
- Description of existing DQA practices across multiple national data networks
- Data quality approaches within the context of multi-purpose multi-site continually refreshed data sources
- Trade-offs between database specific versus study specific data quality assessments
- Development of data quality approaches and metrics that account for variation in data source characteristics such as longitudinality, population, and data types.
Email: vojtech dot huser at nih dot gov