» Back to Table of Contents
The Veterans Administration (VA) garners praise as a quality leader, despite tremendous complexity associated with being the largest health care system in the United States, with a reach from Manila to Maine. Rich national databases have contributed to VA's remarkable
transformation. These centralized data sources provide ample opportunities for clinical leaders, policymakers, and researchers
to peer into the workings of the VA, to understand where it excels and where it can improve. Compared to primary data collection
methodologies, these data sources offer an opportunity to study the universe of VA patients at relatively low cost and without selection bias issues inherent in other approaches.
Thus, VA data sources contribute greatly to system transparency.
However, if data is the glass through which we examine VA, it is important to understand
imperfections that can distort the view. This article describes a few queries one might make of these databases, to illustrate
problems that can arise.
This sounds simple. However, it turns out that there are many non-Veterans (e.g., employees,
CHAMPVA, Tricare, etc.) included in National Patient Care Database (NPCD) outpatient files. If the focus is upon Veterans,
then failure to exclude these people would lead to over-estimation of numbers of patients, and under-estimation of health care needs. This data issue impacts women much more than men: half of women in VA outpatient files are non-Veterans.1
VA data includes a spectrum of processes of care. However, any work on processes of care should account for uncaptured care occurring outside VA. For example, some diabetic VA patients who appear to have failed to receive glycemic monitoring (based on VA data alone) may actually have been tested outside VA. One strategy is to link VA utilization data to other data sources, e.g., fee basis data, Medicaid data, Medicare data.2 Another strategy is to link NPCD to survey data (e.g., Office of Quality and Performance Survey of Health Experiences of Patients), using the latter to select patients who identify VA as their exclusive source of care.
As one example, data quality control checks for the HSR&D funded project, IIR 04-248 revealed that, at many facilities, Decision Support System (DSS) lab data indicated that zero warfarin-treated patients with atrial fibrillation had received
International Normalized Ratio (INR) testing in 2003. This represented a problem with data transmission to the centralized database in the first year of roll-out of this new data element, and improved in subsequent years. In cases like this where irreconcilable data issues arise at some facilities, those analyzing the data may have to settle for conducting a "multi-site" query (i.e., at sites with reliable data), rather than a study of patients
using every facility in the country.
Since every encounter generates a record, NPCD reliably describes many types of utilization.
However, some data elements are populated inconsistently, precluding secondary data manipulations to resolve data limitations.
For example, the Women's Health Evaluation Initiative (WHEI) conducts database analyses for the Women Veterans Health Strategic Health Care Group (WVHSHG)
to inform national women's health policy development. WHEI data quality checks on the clinic "stop code"(clinic type) 322, labeled "Women's Clinic,"revealed erratic application of the code across facilities:
sometimes the code reflected care in a comprehensive Women's Health Center, sometimes care in a preventive health 'pap"clinic, and sometimes care in a general medical
clinic.3 Therefore, it was not possible to draw meaningful conclusions about how women were using gender-specific primary care services. However, in response to these findings, the instructions for coding of 322 have been modified to permit systematic capture
of women's health model of care. This points to VA's commitment to continuously improving robustness of its databases.
Clinical leaders, policymakers, and researchers
need to attend to caveats like these when using national VA databases to reduce the hazard of drawing erroneous conclusions. However, they should not shy away from taking advantage of these rich sources of information. With a little polishing, these databases provide a valuable window into the care that we provide to Veterans.
- Frayne S, et al. "Gender Disparities in Veterans Health Administration Care: Importance of Accounting for
Veteran Status," Medical Care, 2008; 46:549.
- Halanych J, et al. "Racial/Ethnic Differences in Diabetes Care for Older Veterans: Accounting for Dual Health System Use Changes Conclusions," Medical Care 2006; 44:439.
- Herrera L, et al. Can Stop Codes Identify Women's Health Care? Poster Presentation, VA Women's Health