Abstract
Background/Aims This presentation will focus on data quality issues experienced by new HMO Research Network (HMORN) sites while building Virtual Data Warehouse (VDW) demographics, enrollment, utilization, and provider tables from scratch. Our goal is to benefit new sites joining the HMORN by providing a heads-up on anomalies to look for and what to expect of the initial data quality check.
Methods Some of the quality issues to be discussed:
-
Choosing unique VDW person identifier for the site (Unique ‘Pat_ID’ for the HealthPlan members and unique ‘MRN’ for the hospital inpatients). Discuss issues related to both the patient IDs.
-
Handling missing MRN’s / Pat_ID’s, duplicate MRN’s/ Pat_ID’s, multiple MRN’s for Pat_ID and multiple Pat_ID’s for MRN
-
Conflicting gender, race, birth date information for single MRN in both claims and electronic medical record (EMR) hospital data
-
Rolling up of claims data for inpatient stays
-
Handling multiple diagnoses per single stay and identifying primary diagnosis
-
Interpreting claims data (Claim ID, line number, etc.)
-
How to handle: same MRN, same PROCDATE (procedure date), different provider, different POS (place of service), different PX (procedure code) to determine which is the correct POS, primary provider and PX
-
Variable ëauthorization ID’ to identify unique encounter
-
Building VDW demographics table by merging claims data and EMR data
Results We will thoroughly address and document our progress regarding data quality issues, supporting this presentation with examples and tables.
Conclusions New sites need to develop VDW version 3 specification tables from scratch while other sites are in the process of migrating from VDW version 2 to version 3 specifications. As health system use and research dissemination is only as accurate as the initial data management protocol, dedicated quality efforts are essential throughout the process to ensure consistent validity across HMORN sites.
