Abstract
Background/Aims KPCO continuously checks our electronic health data for unexpected values. These values may be errors or legitimate reflections of clinical procedures and data collection rules. Rigorous data analysis requires an understanding of these unexpected values as well as methods to address them. We do not yet have an organized way to document unexpected data values.
Methods We followed a multi-phase plan to develop a database that describes data issues from HMORN data partners. In Phase 1 (Requirements), we met with stakeholders at KPCO to determine needs. For Phase 2 (Design), we drafted a data collection survey and built an initial prototype of the database. In Phase 3 (Development), we improved upon multiple versions of the database and documented procedures to enter new data items, create reports, and review the accuracy of existing data items. When the database framework was sufficiently developed, members of the Analytic Team at KPCO entered issues based on their experience with HMORN studies. In Phase 4 (Testing), we evaluated our documentation and ran sample reports against the new data content. We also assessed the degree of overlap with the VDW Issue Tracker. During Phase 5 (Implementation), we presented the database to key stakeholders at KPCO and began discussions to build similar databases at other HMORN sites and/or enhance the Issue Tracker as a multi-site centralized database.
Results The Analytic Team at KPCO recorded a wide range of data issues occurring in projects from 2010–2013. Data items came from general VDW quality assessments as well as study-specific data checks. Database content included data errors as well as real phenomena attributable to clinical practice or data collection.
Conclusions Analysts can better anticipate problems and nuances in electronic health data with a database that describes the past data anomalies and methods for using the data.
