Importance of health system context for evaluating utilization patterns across systems.
Measuring health services provided to patients can be difficult when patients see providers across multiple health systems and all visits are rarely captured in a single data source covering all systems where patients receive care. Studies that account for only one system will omit the out-of-system health-care use at the patient level. Combining data across systems and comparing utilization patterns across health systems creates complications for both aggregation and accuracy because data-generating processes (DGPs) tend to vary across systems. We develop a hybrid methodology for aggregation across systems, drawing on the strengths of the DGP in each system, and demonstrate its validity for answering research questions requiring cross-system assessments of health-care utilization. Positive and negative predictive probabilities can be useful to assess the impact of the hybrid methodology. We illustrate these issues comparing public sector (administrative records from the US Department of Veterans Affairs system) and private sector (billing records from the US Medicare system) patient level data to identify primary-care utilization. Understanding the context of a particular health system and its effect on the DGP is important in conducting effective valid evaluations.
Duke Scholars
Published In
DOI
EISSN
Publication Date
Volume
Issue
Start / End Page
Location
Related Subject Headings
- United States Department of Veterans Affairs
- United States
- Public Sector
- Private Sector
- Primary Health Care
- Medical Record Linkage
- Humans
- Health Services
- Health Policy & Services
- Data Interpretation, Statistical
Citation
Published In
DOI
EISSN
Publication Date
Volume
Issue
Start / End Page
Location
Related Subject Headings
- United States Department of Veterans Affairs
- United States
- Public Sector
- Private Sector
- Primary Health Care
- Medical Record Linkage
- Humans
- Health Services
- Health Policy & Services
- Data Interpretation, Statistical