Clinical data extracted from disparate systems for CER studies have data quality challenges caused by an infinite variety of "unique circumstances." Although assessing data quality (DQ) is a first step in data analysis, there are no widely accepted frameworks for assessing or reporting data quality findings. We will present recent work on developing a community-based DQ assessment and reporting framework and illustrate its use in sample CER studies as a proposed CER methodology best practice.
We will also provide an understanding of the various procedures and tools used by multiple distributed networks for assessing data quality, share examples of data anomalies and their resolution that data quality assessment methods uncovered in real network settings, and highlight the use of visual analytics to enable rapid detection of data anomalies or inconsistencies for focusing more detailed analytics.Next steps for researchers to complete the framework and ways for the community to get involved will also be discussed.
Learning Objectives: By the end of this session, participants will:
- Next steps for researchers to complete the framework
- Ways for the community to get involved
Course Level: 101 (Introductory)
Faculty: Michael Kahn, University of Colorado Denver, Jeffrey Brown, Harvard Pilgrim Health Care; Daniella Meeker, RAND; Patrick Ryan, Observational Medical Outcomes Partnership; Lisa Schilling, University of Colorado Denver
Kahn, Michael G.; Brown, Jeffrey; Meeker, Daniella; Ryan, Patrick; and Schilling, Lisa M., "Addressing Variations in Data Quality to Facilitate Multi-Institution Comparative Effectiveness Research" (2013). Webinars. 11.