The amount, the kind, or the location of care that clinicians provide to their patients can range from aggressive to parsimonious. We refer to this as the style of care. Certain characteristics of the VA system give reason to believe that it may employ a more parsimonious style of care compared to fee-for-service (FFS) Medicare. VA HSR&D-funded researchers have conducted numerous studies that compare the outcomes, quality of care, and patient satisfaction at the VA compared to Medicare. However, VA HSR&D has not comprehensively assessed the use and cost of care at the VA versus Medicare. Without information on use, the VA cannot address the issue of the value of the healthcare it provides to Veterans.
The objective of this study was to compare practice styles, and geographic variation in use of health care services in the VA and Medicare. Specific aims were to characterize differences in the style of care between Medicare-eligible Veterans treated at the VA and those receiving care in Fee-for-Service (FFS) Medicare (Aim 1); and to compare the inter-facility variation in health care use provided in the VA to that provided by Medicare (Aim 2).
The study was a retrospective cohort analysis using VA administrative databases, Medicare claims, and existing VA surveys. For Aim 1, the study sample consisted of Veterans who responded to the 1999 Large Health Survey of Enrolled Veterans (LHSEV) and were newly age-eligible (age 65 to 69) for Medicare in 1999 (n=35,269). We followed their healthcare use in the VA, fee basis, and Medicare through 2012. The key exposure was reliance on the VA, which was measured by the proportion of all (VA+Medicare) Evaluation & Management (E&M) procedure codes that were provided in the VA. We used novel longitudinal clustering techniques to group veterans based on the degree to which they chose to rely on VA providers to manage their care from 1999-2012. We also estimated panel data negative-binomial models to test whether a greater reliance on the VA in the current year was correlated with reduced healthcare use (provided by the VA or Medicare) in the subsequent year. For Aim 2 we conducted a cross sectional analysis of Medicare-eligible Veteran respondents to the FY2009-11 VA Survey of Health Experiences of Patients (SHEP) (n=215,290) to rank VA medical centers by risk-adjusted per-Veteran VA outpatient costs. We assessed whether VAMCs that had low VA cost achieved this because Veterans at those facilities received more care from Medicare providers.
For Aim 1, the clustering identified three clusters of veterans: Veterans (67% of the cohort) who from 1999-2012 had few E&M visits in either Medicare or VA; Veterans (17%) who had heavy VA reliance and light Medicare; and Veterans (16%) who had heavy Medicare reliance and light VA reliance. Comparing heavy-VA reliance to heavy-Medicare reliance, heavy-VA users were more likely to be non-white (27.9% v. 17.0%; p<0.001); had worse baseline health (SF12 physical 29.6 VA v. 32.1 Medicare; p<0.001); poorer health behaviors including smoking daily (15.9% vs 12.5%; p<0.001); and substantially greater social risk factors, including being unmarried (40.0% vs 23.2%; p<0.001), and twice as likely to say they never feel loved and wanted (14.5% vs 7.7%; p<0.001).
Healthcare managed by VA clinicians was modestly less resource intensive than care managed by Medicare providers. VA reliance in a given year was associated with modestly fewer visits in the subsequent year for. For example, among patients who used specialty care in either the VA or Medicare, Veterans who received 75% or more of E&M visits in the VA in a given year had 1.9 fewer total (VA+Medicare) specialty visits in the subsequent year (p<0.001) than did Veterans who received 75% or more of E&M visits in Medicare. Similar results were found for primary care, surgical care, mental health, and rehab care. However VA-reliance was also associated with an elevated probability of emergency department use (+2.1 percentage points; p=0.008).
For Aim 2, we found substantial inter-facility variation in risk-adjusted per veteran VA outpatient costs across VAMCs. Per veteran costs at lowest and highest tertiles of VA costs where $5311 and $8162, respectively (p<0.001). However, this variation was not due to low-cost VA facilities off-loading costs to Medicare. The mean Medicare costs at the lowest-tertile of VA-costs ($2057) was not significantly different from those at high-VA-cost facilities $2367; p=0.355). The relative parsimony of care at a VA facility was unrelated to Veterans' satisfaction with care or health-related quality of life at those facilities.
A major contribution of this study is that the algorithm for combining VA and Medicare data has been adopted by the VA Office of Analytics and Business Intelligence (OABI). They have applied it to every VA encounter, Fee Basis claim, and Medicare claim generated by Medicare-eligible Veterans enrolled in primary care from 2003 to the present, and will update it each year. They are using it as an omnibus measure of whether specific VA facilities, services, clinical specialties, and VA initiatives (e.g., PACT) are succeeding in the market place for Veterans' care.
We anticipate the study will also make contributions to HSR methods. The compelling findings of the longitudinal clustering analysis suggest that these novel methods should be used more extensively in HSR. A manuscript that is in process that tests the assumptions underlying various instrumental variables will likely be highly referenced by VA researchers who are interested in using distance as an instrument.
External Links for this Project
Grant Number: I01HX000494-01
- Liu CF, Batten A, Wong ES, Fihn SD, Hebert PL. Fee-for-Service Medicare-Enrolled Elderly Veterans Are Increasingly Voting with Their Feet to Use More VA and Less Medicare, 2003-2014. Health services research. 2018 Dec 1; 53 Suppl 3:5140-5158. [view]