- Email: [email protected]
The effect of Medicaid expansions on demand for care from the Veterans Health Administration
Healthcare 3 (2015) 123–128
Contents lists available at ScienceDirect
Healthcare journal homepage: www.elsevier.com/locate/hjdsi
The effect of Medicaid expansions on demand for care from the Veterans Health Administration Austin B. Frakt a,b,n, Amresh Hanchate b, Steven D. Pizer c a
VA Boston Healthcare System, Department of Veterans Affairs, 150 S. Huntington Ave (152H), Boston, MA 02130, USA Boston University, USA c Northeastern University, USA b
art ic l e i nf o
a b s t r a c t
Article history: Received 15 October 2014 Received in revised form 21 January 2015 Accepted 18 February 2015 Available online 12 March 2015
Background: Adequate access to care at Veterans Health Administration (VA) medical centers has become a high-profile policy issue. The Affordable Care Act (ACA) could improve access to care for veterans, particularly if its Medicaid expansion is implemented in all states. The relationship between Medicaid expansion on the one hand and VA enrollment and utilization on the other has not previously been explored for all states. Methods: Using VA and other public data from 2002 to 2008, we calculated a measure of Medicaid eligibility sensitive to state-year varying policy change but not changes in demographics or economic conditions. Next, controlling for potential confounding factors, we estimated fixed effects Poisson models of VA enrollment and inpatient and outpatient utilization. We then used these estimates to simulate the effect of the ACA's Medicaid expansion on demand for VA care. Results: If the ACA's Medicaid expansion had been implemented in all states, enrollment for VA health coverage, acute inpatient care (days), and outpatient visits would have been 9%, 6%, and 12% lower, respectively. In states that did not expand Medicaid in 2014, VA enrollment, inpatient days, and outpatient visits were, respectively, 10, 6, and 13 percentage points higher than they would have been otherwise. Conclusion: VA medical centers in states that did not expand Medicaid in 2014 are likely to have experienced a higher demand, and commensurately longer wait times. As policymakers continue to address VA capacity issues, they should be mindful of the potential role of Medicaid, and that it will change over time as more states adopt the expansion. Published by Elsevier Inc.
Keywords: Medicaid Veterans Health Administration Health insurance Affordable Care Act Demand
1. Introduction Just after the Affordable Care Act's (ACA's; Public Law 111–148) first open enrollment period came to a close in the spring of 2014, access to care for US military veterans became a high-profile policy issue. In the wake of revelations about long waiting times at Veterans Health Administration (VA) medical centers1,15, Congress passed and the President signed the Veterans’ Access to Care through Choice, Accountability, and Transparency Act of 2014 (VACCA; Public Law 113–146), intended to increase access to care for veterans. However,
n Corresponding author at: Department of Veterans Affairs, VA Boston Healthcare System, Health Care Financing & Economics and School of Medicine, Boston University, 150 S. Huntington Ave. (152H), Boston, MA, USA. Tel.: þ1 857 364 6064; fax: þ1 857 364 4511. E-mail address: [email protected] (A.B. Frakt). 1 Throughout, for simplicity we use the term “VA medical center” to refer to any VA facility, whether a VA hospital or any of its affiliated, community-based outpatient clinics.
http://dx.doi.org/10.1016/j.hjdsi.2015.02.004 2213-0764/Published by Elsevier Inc.
the ACA could improve access to care for veterans as well, particularly if its Medicaid expansion is implemented in all states. Medicaid expansion has not proceeded as originally intended, with large effects on veterans, among other Americans. Architects of the ACA expected all states to expand Medicaid to individuals with incomes below 138% of the federal poverty level (FPL). An estimated four in ten uninsured veterans would qualify for Medicaid if all states did so.11 However, a 2012 US Supreme Court decision gave states the option to decline the Medicaid expansion without forgoing all federal funding for Medicaid.8 About half of states committed to expanding their Medicaid programs in 2014,2 and over half of uninsured veterans who could qualify for Medicaid did not live in states that expanded their programs in that year.12 Because Medicaid-financed care serves as an alternative to VA care for low-income veterans—both satisfy the individual mandate to have coverage for health care—the choice by states to decline expansion could have the effect of increasing reliance on the VA for care among veterans in those states, potentially increasing waiting times. How much does Medicaid expansion affect demand for and utilization of VA care?
124
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
This question has not been directly investigated at a national level, though one study has examined both Medicaid and VA enrollment in Massachusetts. Using 2004–2013 Current Population Survey data, Wong et al.17 estimated changes in Medicaid, VA, and private health insurance enrollment due to Massachusetts’ implementation of health reform in 2006, relative to neighboring New England states. With their preferred specification, the authors found no statistically significant change in VA enrollment but a statistically significant 2.5 percentage point increase in Medicaid enrollment in Massachusetts, relative to comparator states. Excluding years 2009–2010 yielded a statistically significant decrease in VA enrollment of 2.4 percentage points. The results are in contrast to official VA forecasts of an increase in VA enrollment following health care reform, and the authors recommend further analysis that also considers effects on health care utilization, which our study addresses. Other studies have examined the extent to which VA enrollees have non-VA coverage. The 2011 Survey of Veteran Enrollees’ Health and Reliance Upon VA16 found that among nonelderly VA enrollees, 34% had private health insurance, 4% had Medicaid coverage, 17% were enrolled in Medicare, and 37% reported having no non-VA coverage; the share of all outpatient visits financed by VA was 54% among nonelderly VA enrollees and 77% among those with no non-VA coverage. A different study, using data from 1999, indicated that 10.2% of VA's patient load consisted of VA–Medicaid dual enrollees, over half of whom were nonelderly.14 In this paper, we report our analysis of the historical relationships between eligibility for Medicaid and VA enrollment and utilization of inpatient and outpatient care. This work was based on area-level VA enrollment and utilization data over time, a measure of Medicaid eligibility that is sensitive to state-year varying policy but not variations in economic or demographic conditions, and other, public data sources from 2002 to 2008. We also report our simulation, based on estimated, historical relationships, of the potential impact of states’ Medicaid expansion decisions in 2014 on demand for VA care. (Though other aspects of the ACA, like the individual mandate, could also affect demand for VA care, national data are insufficient to analyze them at this time.) Our qualitative expectation was that greater access to Medicaid reduces VA enrollment and utilization of VA services. To our knowledge, prior to our work a VA enrollment response to coverage expansion had not been quantified nationally and no prior work has investigated a VA utilization response.
2. Material and methods To examine the historical relationship between policy-driven Medicaid eligibility and VA enrollment and utilization, we constructed an analytic file from VA and public sources. Because our intent was to estimate relationships that could inform us about the likely effects of the ACA, we focused attention on veterans less than 65 years old, and hence not age-eligible for Medicare, a subset of the target population (non-elderly adults) for the ACA's coverage expansion in 2014. Our unit of analysis was year-“sector,” with years spanning 2002–2008. Sectors are groups of counties used by the VA to track and predict enrollment and utilization. There were 566 sectors in our analytic file, spanning 7 years, for a total of 3962 observations. Dependent variables for analysis were VA enrollment, inpatient care utilization, and outpatient care utilization, obtained from VA administrative data sources. Enrollment is a cumulative indicator of the number of veterans enrolled for VA health care at each point in time; for each year 2002–2008, we examined end-of-year enrollment cumulative since April 1, 1999. We emphasize that not all VA enrollees use VA services in a given year. So, enrollment
alone may overstate demand. In part for this reason, we also directly measured inpatient utilization in inpatient days and outpatient utilization in clinic stops. A VA clinic stop is a record of the receipt of care from a distinct clinic or provider type. In a single visit to a VA medical center, a veteran might visit multiple clinics and see a variety of clinicians (e.g., primary care, orthopedics, urology, etc.). Each such visit counts as a clinic stop, even when they occur on the same day at the same medical center. Our key independent variable was a state-level index of Medicaid eligibility generosity that is sensitive only to crosssectional and temporal policy variation, not demographic or economic variation. This Medicaid eligibility generosity index is based on Medicaid eligibility income and asset thresholds for each state and year, which we obtained from documentation of the Urban Institute's TRIM3 microsimulation model (http://trim3. urban.org) for years 2002–2008 (the latest year available). We applied these state-year Medicaid thresholds to a nationally representative sample from the Medical Expenditure Panel Survey (MEPS) from 1998. For a given state and year 2002–2008, the resulting variable measures the proportion of the national population that would have been Medicaid eligible had the entire national population lived in that state in that year. (Thresholds vary within state and year by category of eligibility in ways that only TRIM3 has captured in a research-ready format. Additional details are in the appendix Section A.1.) This variable is similar to one built from the Current Population Survey (CPS) and used in a series of papers on the effects of Medicaid eligibility expansions for women and children in the 1990s.6,7 We chose MEPS instead of CPS because it includes detailed information on medical spending that enabled us to include thresholds for medically needy programs in our calculations. In contrast to actual proportions of state populations enrolled, this measure reflects policy differences without the potentially confounding effects of state differences in income, employment, and medical costs. We also constructed, for each state, a second, simulated Medicaid eligibility variable using the same technique above, but under the assumption that the Medicaid expansion provision of the ACA had been implemented. Factors other than Medicaid eligibility can affect VA enrollment and utilization. Those that are constant over time within sector or are secular trends were absorbed by sector- and year-fixed effects included in our empirical models. However, economic factors that likely affect whether to enroll in and use the VA, like employment and wealth, can vary by sector and year. We captured those with two variables: (1) the US Federal Housing Administration's housing price index, which measures housing price change from a baseline set to 1 in 1991; (2) the employment-to-population ratio from the US Bureau of Labor Statistics. Both are state-level measures that vary by year. Our hypothesis is that in states and years with greater housing wealth or higher employment, working-age veterans are more likely to have access to substitutes for VA care (e.g., individual market or employer-based plans), apart from the nature of the state-year's Medicaid eligibility policy. Because our dependent variables are integer counts of enrollment, inpatient days, and outpatient clinic stops, we estimated fixed effects Poisson models3 (p. 668, 805) with robust standard errors to address potential overdispersion4 (p. 561). We used the year-sector-level, non-elderly veteran population, obtained from VA and American Community Survey data, as the exposure variable, which adjusted for the fact that some sectors have greater enrollment and utilization because more veterans reside in them.2
2 Total veteran population counts (elderly and non-elderly combined) were available from the VA at the sector level for all years. We estimated the rate of
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
After we estimated models of VA demand, we restricted the data to year 2008 and predicted enrollment, inpatient utilization, and outpatient utilization with the original Medicaid variable (our baseline) and, separately, with a version that simulated the ACA Medicaid expansion. The ratios of the ACA-based predicted values of enrollment, inpatient utilization, and outpatient utilization to the baseline 2008 predictions provided us with estimates of the proportional changes we might have expected from the ACA's Medicaid expansion provision alone, had it been implemented in 2008. We report state-level aggregates of these proportional changes for all states (i.e., simulating Medicaid expansion for all states) and indicate which states had committed to the optional Medicaid expansion in 2014. From these results, we estimated how much VA demand would change if states that did not expand Medicaid in 2014 did so. All statistical work was conducted using Stata version 10.
3. Results Table 1 provides descriptive statistics for the dependent, independent, and exposure variables at the year–sector level. In our data, an average year-sector has 4457 enrollees that use 1564 inpatient days and 17,660 outpatient clinic stops, though there is considerable variation in all three variables. The Medicaid eligibility variable is 0.089 at baseline, on average, but twice as large (0.18) when based on the ACA Medicaid expansion, if adopted in all states. Not shown in Table 1 are trends in our key variables. Over the 2002–2008 study window, VA enrollment, inpatient days, and outpatient clinic stops grew by 59%, 87%, and 141%, respectively. (In the appendix Section A.2, we provide yearly totals for these variables.) The vast majority of states experienced growth in Medicaid eligibility generosity as well. In the average state, it grew 23% between 2002 and 2008. Also not shown are year–state level statistics for our outcome variables, which are as follows: On average, across the 357 year-states in our sample, there are 49,465 enrollees (std. dev.: 49,157, min: 3857, max: 269,057) using 17,362 inpatient days (std. dev.: 18,579, min: 969, max: 106,970) and 195,992 outpatient clinic stops (std. dev.: 208,618, min: 13,172, max: 1,208,335). Table 2 presents Poisson regression coefficients for the three dependent variables: enrollment, inpatient days, and outpatient clinic stops. All three are statistically significantly and negatively associated with Medicaid eligibility. Where and when policy-driven Medicaid eligibility is higher, VA enrollment and utilization are lower. Enrollment and outpatient clinic stops are also statistically significantly and negatively associated with housing price index. Where and when home values are higher, VA enrollment and outpatient clinic stops are lower. The same is true of employmentto-population ratio. Where and when employment is higher, relative to the population, VA enrollment and outpatient clinic stops are lower. Table 2 shows a lack of statistical significance in association between inpatient days and housing price index or the employment-to-population ratio. This could occur if level and setting (VA or non-VA) of inpatient care is relatively inelastic with respect to changes in wealth or income that housing price and employment changes represent. These results are consistent with our expectations. Medicaid is a substitute for VA care, so when more veterans are eligible for it, fewer would be expected to use the VA. Similarly, when the (footnote continued) growth of the non-elderly VA population from 2006 to 2010 American Community Survey 5-year estimates (not available for earlier years) and extrapolated the proportion of the veteran population that is non-elderly back to 2002. We applied this proportion to the VA population data to obtain annual counts of non-elderly veterans only.
125
economy or housing markets are stronger, veterans are more likely to be wealthier and employed, providing a greater opportunity for take up of individual- or group-market insurance plans— which also substitute for VA care—or a greater ability to afford cost-sharing such plans require. Concerned that year-sectors with small enrollment might be exerting a disproportionate influence on our findings, we conducted a robustness check on a sample without the 5% smallest enrollment year-sectors. Results were nearly identical to those from the entire sample. Because the effects of the results of Table 2 are hard to interpret, we also estimated elasticities for the three dependent variables with respect to Medicaid eligibility and evaluated at the mean of the data. The elasticity of enrollment is 0.11 [95% CI: 0.12, 0.089, p o0.001], of inpatient days is 0.065 [95% CI: 0.12, 0.0098, p ¼0.02], and of outpatient clinic stops is 0.14 [95: CI: 0.18, 0.10, p o0.001]. A more policy-relevant way to understand the results of Table 2 is to use them to simulate what they imply for the ACA's Medicaid expansion. Combining the mean, ACA-simulated Medicaid eligibility value with the elasticity estimates above provides summary predictions. Table 1 shows that the ACA-simulated Medicaid eligibility variable is about double the baseline value at the mean (0.18 vs. 0.09). According to our estimated elasticities, a doubling of Medicaid eligibility corresponds to an 11% reduction in VA enrollment, a 6.5% reduction in VA inpatient days, and a 14% reduction in VA outpatient clinic stops. In absolute numbers, if applied to data for 2008, these would represent a decrease of 0.34M enrollees (from a baseline of 3.06M to 2.72M), of 0.07M inpatient days (1.14M to 1.07M), and 1.97M outpatient clinic stops (14.1M to 12.13M). Table 3 provides state-level results of our simulation. For each state, it lists whether it expanded Medicaid in 2014 and by what proportion VA enrollment, inpatient days, and outpatient clinic stops would fall if it did so, per our simulation. (Though we show predicted changes for all states, for those that do not actually expand, there would be no actual change.) With the exception of Vermont, in which ACA-like expansion had already occurred, VA enrollment, inpatient days, and outpatient clinic stops would fall in each state under the ACA's Medicaid expansion. Depending on state, enrollment would fall to between 0.85 and 0.98 of its baseline level; inpatient days would fall to between 0.90 and 0.99 of its baseline level; and outpatient clinic stops would fall to between 0.80 and 0.97 of its baseline level (all excluding Vermont). Appendix Section A.3 includes the results of Table 3 in figures, stratified by expansion status. The bottom three rows of Table 3 show the average, predicted effects for states that did and did not expand Medicaid in 2014, as well as for all states. On average, in expanding states, VA enrollment, inpatient days, and outpatient clinic stops are predicted to fall to 0.92, 0.95, and 0.89 of their baseline values, respectively. If non-expanding states did expand, VA enrollment, inpatient days, and outpatient clinic stops would fall more, to 0.90, 0.94, and 0.87 of their baseline values, respectively. Put another way, in states that did not expand Medicaid, demand on VA is 10 percentage points higher in terms of enrollment, 6 percentage points higher in terms of inpatient visits, and 13 percentage points higher in terms of outpatient clinic stops than it would otherwise be. If all states expanded their programs, VA demand would be 9%, 6%, and 12% lower by these three measures, respectively.
4. Discussion We used VA and publicly available data to estimate the historical relationships between policy-driven Medicaid eligibility
126
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
levels and VA enrollment and utilization. Consistent with expectations, we found that higher levels of Medicaid eligibility are associated with lower VA enrollment, inpatient and outpatient utilization. Using our estimates, we found that if the ACA's Medicaid expansion had been implemented in all states in 2008, and holding all else constant, VA enrollment, inpatient days, and outpatient clinic stops would have been 9%, 6%, and 12% lower, respectively. Our results suggest that states in which Medicaid did not expand in 2014, VA enrollment, inpatient days, and outpatient clinic stops were 10, 6, and 13 percentage points higher than they would have otherwise been, respectively. These results suggest that Medicaid expansion could considerably reduce the burden of demand placed on VA medical centers. Though not directly comparable, our estimated impact of Medicaid expansion in Massachusetts on the VA in that state is similar in magnitude to the findings of Wong et al.17 They examined the impact of Massachusetts’ 2006 health reform on the VA while we simulated the additional effect of the ACA's Medicaid expansion in that state. In one of their specifications, they found that that state's reform was associated with a 2.4 percentage point decrease in VA enrollment, relative to neighboring New England states, similar to our finding of a 3 percentage point decrease for Massachusetts after the ACA's Medicaid expansion. However, our national results are larger in magnitude, reflecting the ACA's larger pre-post change in Medicaid eligibility in much of the rest of the nation relative to Massachusetts. Our findings are accompanied by several limitations. First, our work is based on historical data available only through 2008. Compared to past Medicaid expansions, the ACA's is larger. Extrapolations warrant caution in general. In the case of the ACA's Medicaid expansion in particular, some have expressed concern
Table 1 Descriptive statistics. Variable
Mean (std. dev.)
[Min–Max]
Dependent variables Enrollmenta Inpatient daysa Outpatient clinic stopsa
4457 (3848) 1564 (1631) 17,660 (16,701)
[36–49,337] [9–19,331] [85–235,711]
Independent variables Medicaid eligibility: baselineb Medicaid eligibility: ACA-simulatedb Housing price indexb,c Employment-to-pop ratiob Exposure variable Veteran 18–64 population
5. Conclusion 0.089 0.18 1.97 62.72
(0.055) (0.030) (0.40) (3.52)
26,705 (22,631)
[0.018–0.36] [0.17–0.36] [1.07–3.48] [52.28–71.58] [132–287,567]
N ¼3962 year-sector observations. a b c
that the large increase in demand for care it motivates might outstrip available supply.1 If that is the case, then Medicaid may be a less viable substitute for VA care than the historical record suggests. Second, we are unable to simulate the effects of the ACA's individual coverage or employer mandate and other features such as the establishment of subsidized exchange coverage. Because VA enrollment satisfies the law's individual mandate, that could encourage veterans to enroll. Also, increased awareness about the mandate and exposure to outreach efforts could increase veterans’ interest in the VA.11 How firms respond to the law's employer mandate could also affect availability of employer-based coverage for veterans, thereby changing the demand for VA coverage among employed veterans.17 In principle, availability of subsidized, exchange-based coverage (for those who qualify for it) could be a substitute for VA care as well. Over a half-million nonelderly veterans (about 40% of uninsured, nonelderly veterans) are potentially eligible for exchange-based subsidies.11 However, exchanges could attempt to inform veterans of their VA eligibility, a potentially countervailing factor.13 Notably, Wong et al.17 did not find a statistically significant increase in the rate of private coverage among veterans after Massachusetts’ coverage expansion, relative to those in neighboring New England states. Finally, our study years (2002–2008) coincide with a period of rapid VA enrollment growth that has abated somewhat. It is possible that our results are driven by a reduction of that growth where and when Medicaid eligibility was relatively more expansive (extensive margin), but that there was little effect on existing VA enrollees (intensive margin). Data available for this study are not amenable to separately identifying the impact of Medicaid expansion on VA utilization by existing vs. new VA enrollees. If our results are substantially driven by a reduction in the rate of VA enrollment, not by changes in utilization by existing enrollees, then we would expect a more modest effect of the ACA's Medicaid expansion if VA enrollment growth continues to moderate. Therefore, future research should estimate these relationships during years of slower VA enrollment growth.
Year–sector level variables. Year–state level variables. Indexed to 1.00 in 1991.
The ACA's Medicaid expansion could play a large role in reducing the burden of demand on VA providers, though some of its effect could be offset by other provisions of the law. VA medical centers in states that did not expand Medicaid in 2014 are likely to have experienced a higher demand, and commensurately longer wait times, than they would have otherwise under Medicaid expansion. As policymakers continue to address VA capacity issues, they should be mindful of the potential role of Medicaid, and that it will change over time as more states adopt the expansion.
Table 2 Poisson regression coefficients. Variable
Medicaid eligibility Housing price index Employment-to-pop ratio
Coefficient (std. dev.) Enrollment
Inpatient days
Outpatient clinic stops
1.20 (0.10)nnn 0.30 (0.0056)nnn 0.019 (0.0018)nnn
0.73 (0.32)n 0.28 (0.016) 0.0065 (0.0056)
1.60 (0.23)nnn 0.068 (0.0082)nnn 0.0077 (0.0029)nn
N ¼3962 year-sector observations. Exposure variable: Veteran 18–64 population. Year and sector fixed effects not shown. n
po 0.05. p o0.01. nnn p o0.001. nn
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
127
Table 3 Simulation results. State
Expanded Medicaid in 2014a
AK No AL No AR Yes AZ Yes CA Yes CO Yes CT Yes DC Yes DE Yes FL No GA No HI Yes IA Yes ID No IL Yes IN No KS No KY Yes LA No MA Yes MD Yes ME No MI Yes MN Yes MO No MS No MT No NC No ND Yes NE No NH No NJ Yes NM Yes NV Yes NY Yes OH Yes OK No OR Yes PA No RI Yes SC No SD No TN No TX No UT No VA No VT Yes WA Yes WI No WV Yes WY No Ave for all Medicaid-expanding states (baseline total) Ave for all non-expanding states (baseline total)b Ave for all states (baseline total)
Assuming expansion, relative proportional change in (and baseline values of) … Enrollmentb
Inpatient daysb
Outpatient clinic stopsb
0.92 (15,129) 0.90 (63,947) 0.91 (43,830) 0.93 (77,661) 0.92 (262,479) 0.87 (54,047) 0.86 (22,598) 0.92 (5813) 0.96 (9017) 0.93 (227,040) 0.86 (109,911) 0.96 (12,074) 0.92 (32,828) 0.92 (19,327) 0.88 (97,547) 0.85 (63,733) 0.86 (29,939) 0.88 (54,081) 0.90 (49,578) 0.97 (36,468) 0.88 (53,956) 0.94 (19,600) 0.90 (74,005) 0.96 (52,304) 0.92 (68,721) 0.87 (38,804) 0.89 (17,593) 0.89 (114,708) 0.86 (10,291) 0.86 (21,366) 0.92 (14,361) 0.93 (43,121) 0.86 (27,938) 0.91 (39,151) 0.98 (150,647) 0.92 (111,550) 0.91 (47,994) 0.92 (46,094) 0.90 (106,399) 0.89 (8168) 0.89 (66,559) 0.87 (14,989) 0.89 (74,409) 0.91 (258,593) 0.89 (18,864) 0.86 (90,543) 1.00 (7623) 0.92 (76,070) 0.9 (56,784) 0.92 (32,913) 0.90 (9485) 0.92 (1.44M) 0.90 (1.62M) 0.91 (3.06M)
0.95 0.94 0.94 0.95 0.95 0.92 0.91 0.95 0.98 0.95 0.91 0.98 0.95 0.95 0.93 0.90 0.91 0.92 0.94 0.98 0.93 0.96 0.93 0.98 0.95 0.92 0.93 0.93 0.92 0.91 0.95 0.95 0.91 0.95 0.99 0.95 0.94 0.95 0.94 0.93 0.93 0.92 0.93 0.95 0.93 0.91 1.00 0.95 0.94 0.95 0.94 0.95 0.94 0.94
0.90 (45,491) 0.87 (288,633) 0.88 (227,522) 0.90 (373,539) 0.89 (1,178,442) 0.83 (231,889) 0.82 (108,549) 0.89 (43,461) 0.95 (33,290) 0.90 (1,216,681) 0.82 (442,031) 0.95 (35,594) 0.90 (146,826) 0.89 (81,741) 0.85 (457,956) 0.80 (287,069) 0.82 (139,512) 0.84 (256,588) 0.87 (230,560) 0.96 (167,982) 0.84 (248,157) 0.92 (92,990) 0.86 (333,467) 0.95 (253,784) 0.90 (317,633) 0.84 (186708) 0.85 (76,303) 0.85 (499,915) 0.82 (34,888) 0.82 (108,171) 0.89 (61,314) 0.90 (168,542) 0.82 (141,894) 0.89 (179,323) 0.97 (620,958) 0.89 (615,658) 0.88 (233,522) 0.90 (218,748) 0.87 (481,676) 0.86 (45,462) 0.85 (280,025) 0.83 (70,355) 0.86 (345,951) 0.88 (1,224,139) 0.86 (78,130) 0.81 (350,784) 1.00 (29,114) 0.90 (308,929) 0.87 (279,329) 0.90 (179,908) 0.87 (44,224) 0.89 (6.64M) 0.87 (7.46M) 0.88 (14.10M)
(3437) (18,869) (25,035) (27,930) (95,675) (17,074) (9162) (3770) (2643) (95,330) (37,727) (2978) (10,331) (7095) (37,705) (21,112) (13,606) (20,976) (22,935) (16,601) (16,323) (5806) (27,028) (14,570) (28,536) (15,705) (6809) (42,338) (2585) (6927) (4045) (11,773) (10,704) (13,680) (43,207) (42,895) (22,833) (18,096) (36,482) (4005) (25,776) (4455) (32,211) (112,962) (8247) (31,446) (1735) (24,951) (19,400) (13,641) (3757) (0.52M) (0.63M) (1.14M)
a
Source: The Advisory Board.2 For non-expanding states, values shown are those simulated as if those states did expand. Proportional changes would be, by construction, 1.00 (no change) for a simulation of non-expansion. b
Conflict of interest This statement accompanies the article "The Effect of Medicaid Expansions on Demand for Care from the Veterans Health Administration," authored (co-authored) by Austin B Frakt, Amresh Hanchate, Steven D Pizer and submitted to Healthcare as an original article. Below all authors have disclosed relevant commercial associations that might pose a conflict of interest:
Consultant arrangements: None Stock/other equity ownership: None Patent licensing arrangements: None Grants/research support: None Employment: None Speakers' bureau: Austin Frakt was under contract with Speakers on Healthcare during preparation of this manuscript. Expert witness: None Other: None
128
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
Acknowledgments This work was funded by the VA HSR&D (Project MRC 13-427) and supported by the VHA Office of the Assistant Deputy Under Secretary for Health for Policy and Planning (10P1A). The views expressed are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs, Boston University, or Northeastern University. The authors have no conflicts of interest to disclose. This work has been approved by the VA Boston Healthcare System's Institutional Review Board.
Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at: http://dx.doi.org/10.1016/j.hjdsi.2015.02.004. References 1. Abraham J. How might the Affordable Care Act's coverage expansion provisions influence demand for medical care? Milbank Q. 2014;92(1):63–86. 2. The Advisory Board. Where the states stand on Medicaid expansion. February. 〈http://www.advisory.com/daily-briefing/resources/primers/medicaidmap〉; 2014 19.05.14.
3. Cameron AC, Trivedi PK. Microeconometrics: Methods and Applications. New York: Cambridge University Press; 2005. 4. Cameron AC, Trivedi PK. Microeconometrics Using Stata, vol. 5. College Station, TX: Stata Press; 2009. 6. Currie J, Gruber J. Health insurance eligibility, utilization of medical care, and child health. Q J Econ. 1996;111(2):431–466. 7. Cutler DM, Gruber J. Does public insurance crowd out private insurance? Q J Econ. 1996;111(2):391–430. 8. Frakt AB, Carroll AE. Sound policy trumps politics: states should expand medicaid. J Health Polit Policy Law. 2013;38(1):165–178. 11. Haley J, Kenney GM. Uninsured Veterans and Family Members: State and National Estimates of Expanded Medicaid Eligibility Under the ACA. Urban Institute. March. 〈http://www.urban.org/publications/412775.html〉; 2013. 14.05.14. 12. Haley J, Kenney GM. Uninsured veterans and family members: national and state estimates and new coverage options under the ACA. US Department of Veterans Affairs Cyber-Seminar. 21 November. 〈http://www.hsrd.research.va. gov/for_researchers/cyber_seminars/archives/773-notes.pdf〉; 2013 14.05.14. 13. Heiss C, McMahon SM. Veterans and the ACA: How Health Reform Boosts Eligibility for VA Health Care. Hamilton, NJ: Center for Health Care Strategies; 2012. 14. Hendricks A, Gardner J, Frakt A, et al. What can Medicaid data add to research on VA patients? J Rehabil Res Dev. 2010;47(8):773–780. 15. Kizer KW, Jha AK. Restoring trust in VA health care. New Engl J Med. 2014. 16. Office of the Assistant Deputy Under Secretary for Health for Policy and Planning. 2011 Survey of veteran enrollees’ health and reliance upon VA. Department of Veterans Affairs. 17. Wong ES, Maciejewski ML, Herbert PL, Bryson CL, Liu CF. Massachusetts health reform and veterans affairs health system enrollment. Am J Manag Care. 2014;20(8):629–636.
Contents lists available at ScienceDirect
Healthcare journal homepage: www.elsevier.com/locate/hjdsi
The effect of Medicaid expansions on demand for care from the Veterans Health Administration Austin B. Frakt a,b,n, Amresh Hanchate b, Steven D. Pizer c a
VA Boston Healthcare System, Department of Veterans Affairs, 150 S. Huntington Ave (152H), Boston, MA 02130, USA Boston University, USA c Northeastern University, USA b
art ic l e i nf o
a b s t r a c t
Article history: Received 15 October 2014 Received in revised form 21 January 2015 Accepted 18 February 2015 Available online 12 March 2015
Background: Adequate access to care at Veterans Health Administration (VA) medical centers has become a high-profile policy issue. The Affordable Care Act (ACA) could improve access to care for veterans, particularly if its Medicaid expansion is implemented in all states. The relationship between Medicaid expansion on the one hand and VA enrollment and utilization on the other has not previously been explored for all states. Methods: Using VA and other public data from 2002 to 2008, we calculated a measure of Medicaid eligibility sensitive to state-year varying policy change but not changes in demographics or economic conditions. Next, controlling for potential confounding factors, we estimated fixed effects Poisson models of VA enrollment and inpatient and outpatient utilization. We then used these estimates to simulate the effect of the ACA's Medicaid expansion on demand for VA care. Results: If the ACA's Medicaid expansion had been implemented in all states, enrollment for VA health coverage, acute inpatient care (days), and outpatient visits would have been 9%, 6%, and 12% lower, respectively. In states that did not expand Medicaid in 2014, VA enrollment, inpatient days, and outpatient visits were, respectively, 10, 6, and 13 percentage points higher than they would have been otherwise. Conclusion: VA medical centers in states that did not expand Medicaid in 2014 are likely to have experienced a higher demand, and commensurately longer wait times. As policymakers continue to address VA capacity issues, they should be mindful of the potential role of Medicaid, and that it will change over time as more states adopt the expansion. Published by Elsevier Inc.
Keywords: Medicaid Veterans Health Administration Health insurance Affordable Care Act Demand
1. Introduction Just after the Affordable Care Act's (ACA's; Public Law 111–148) first open enrollment period came to a close in the spring of 2014, access to care for US military veterans became a high-profile policy issue. In the wake of revelations about long waiting times at Veterans Health Administration (VA) medical centers1,15, Congress passed and the President signed the Veterans’ Access to Care through Choice, Accountability, and Transparency Act of 2014 (VACCA; Public Law 113–146), intended to increase access to care for veterans. However,
n Corresponding author at: Department of Veterans Affairs, VA Boston Healthcare System, Health Care Financing & Economics and School of Medicine, Boston University, 150 S. Huntington Ave. (152H), Boston, MA, USA. Tel.: þ1 857 364 6064; fax: þ1 857 364 4511. E-mail address: [email protected] (A.B. Frakt). 1 Throughout, for simplicity we use the term “VA medical center” to refer to any VA facility, whether a VA hospital or any of its affiliated, community-based outpatient clinics.
http://dx.doi.org/10.1016/j.hjdsi.2015.02.004 2213-0764/Published by Elsevier Inc.
the ACA could improve access to care for veterans as well, particularly if its Medicaid expansion is implemented in all states. Medicaid expansion has not proceeded as originally intended, with large effects on veterans, among other Americans. Architects of the ACA expected all states to expand Medicaid to individuals with incomes below 138% of the federal poverty level (FPL). An estimated four in ten uninsured veterans would qualify for Medicaid if all states did so.11 However, a 2012 US Supreme Court decision gave states the option to decline the Medicaid expansion without forgoing all federal funding for Medicaid.8 About half of states committed to expanding their Medicaid programs in 2014,2 and over half of uninsured veterans who could qualify for Medicaid did not live in states that expanded their programs in that year.12 Because Medicaid-financed care serves as an alternative to VA care for low-income veterans—both satisfy the individual mandate to have coverage for health care—the choice by states to decline expansion could have the effect of increasing reliance on the VA for care among veterans in those states, potentially increasing waiting times. How much does Medicaid expansion affect demand for and utilization of VA care?
124
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
This question has not been directly investigated at a national level, though one study has examined both Medicaid and VA enrollment in Massachusetts. Using 2004–2013 Current Population Survey data, Wong et al.17 estimated changes in Medicaid, VA, and private health insurance enrollment due to Massachusetts’ implementation of health reform in 2006, relative to neighboring New England states. With their preferred specification, the authors found no statistically significant change in VA enrollment but a statistically significant 2.5 percentage point increase in Medicaid enrollment in Massachusetts, relative to comparator states. Excluding years 2009–2010 yielded a statistically significant decrease in VA enrollment of 2.4 percentage points. The results are in contrast to official VA forecasts of an increase in VA enrollment following health care reform, and the authors recommend further analysis that also considers effects on health care utilization, which our study addresses. Other studies have examined the extent to which VA enrollees have non-VA coverage. The 2011 Survey of Veteran Enrollees’ Health and Reliance Upon VA16 found that among nonelderly VA enrollees, 34% had private health insurance, 4% had Medicaid coverage, 17% were enrolled in Medicare, and 37% reported having no non-VA coverage; the share of all outpatient visits financed by VA was 54% among nonelderly VA enrollees and 77% among those with no non-VA coverage. A different study, using data from 1999, indicated that 10.2% of VA's patient load consisted of VA–Medicaid dual enrollees, over half of whom were nonelderly.14 In this paper, we report our analysis of the historical relationships between eligibility for Medicaid and VA enrollment and utilization of inpatient and outpatient care. This work was based on area-level VA enrollment and utilization data over time, a measure of Medicaid eligibility that is sensitive to state-year varying policy but not variations in economic or demographic conditions, and other, public data sources from 2002 to 2008. We also report our simulation, based on estimated, historical relationships, of the potential impact of states’ Medicaid expansion decisions in 2014 on demand for VA care. (Though other aspects of the ACA, like the individual mandate, could also affect demand for VA care, national data are insufficient to analyze them at this time.) Our qualitative expectation was that greater access to Medicaid reduces VA enrollment and utilization of VA services. To our knowledge, prior to our work a VA enrollment response to coverage expansion had not been quantified nationally and no prior work has investigated a VA utilization response.
2. Material and methods To examine the historical relationship between policy-driven Medicaid eligibility and VA enrollment and utilization, we constructed an analytic file from VA and public sources. Because our intent was to estimate relationships that could inform us about the likely effects of the ACA, we focused attention on veterans less than 65 years old, and hence not age-eligible for Medicare, a subset of the target population (non-elderly adults) for the ACA's coverage expansion in 2014. Our unit of analysis was year-“sector,” with years spanning 2002–2008. Sectors are groups of counties used by the VA to track and predict enrollment and utilization. There were 566 sectors in our analytic file, spanning 7 years, for a total of 3962 observations. Dependent variables for analysis were VA enrollment, inpatient care utilization, and outpatient care utilization, obtained from VA administrative data sources. Enrollment is a cumulative indicator of the number of veterans enrolled for VA health care at each point in time; for each year 2002–2008, we examined end-of-year enrollment cumulative since April 1, 1999. We emphasize that not all VA enrollees use VA services in a given year. So, enrollment
alone may overstate demand. In part for this reason, we also directly measured inpatient utilization in inpatient days and outpatient utilization in clinic stops. A VA clinic stop is a record of the receipt of care from a distinct clinic or provider type. In a single visit to a VA medical center, a veteran might visit multiple clinics and see a variety of clinicians (e.g., primary care, orthopedics, urology, etc.). Each such visit counts as a clinic stop, even when they occur on the same day at the same medical center. Our key independent variable was a state-level index of Medicaid eligibility generosity that is sensitive only to crosssectional and temporal policy variation, not demographic or economic variation. This Medicaid eligibility generosity index is based on Medicaid eligibility income and asset thresholds for each state and year, which we obtained from documentation of the Urban Institute's TRIM3 microsimulation model (http://trim3. urban.org) for years 2002–2008 (the latest year available). We applied these state-year Medicaid thresholds to a nationally representative sample from the Medical Expenditure Panel Survey (MEPS) from 1998. For a given state and year 2002–2008, the resulting variable measures the proportion of the national population that would have been Medicaid eligible had the entire national population lived in that state in that year. (Thresholds vary within state and year by category of eligibility in ways that only TRIM3 has captured in a research-ready format. Additional details are in the appendix Section A.1.) This variable is similar to one built from the Current Population Survey (CPS) and used in a series of papers on the effects of Medicaid eligibility expansions for women and children in the 1990s.6,7 We chose MEPS instead of CPS because it includes detailed information on medical spending that enabled us to include thresholds for medically needy programs in our calculations. In contrast to actual proportions of state populations enrolled, this measure reflects policy differences without the potentially confounding effects of state differences in income, employment, and medical costs. We also constructed, for each state, a second, simulated Medicaid eligibility variable using the same technique above, but under the assumption that the Medicaid expansion provision of the ACA had been implemented. Factors other than Medicaid eligibility can affect VA enrollment and utilization. Those that are constant over time within sector or are secular trends were absorbed by sector- and year-fixed effects included in our empirical models. However, economic factors that likely affect whether to enroll in and use the VA, like employment and wealth, can vary by sector and year. We captured those with two variables: (1) the US Federal Housing Administration's housing price index, which measures housing price change from a baseline set to 1 in 1991; (2) the employment-to-population ratio from the US Bureau of Labor Statistics. Both are state-level measures that vary by year. Our hypothesis is that in states and years with greater housing wealth or higher employment, working-age veterans are more likely to have access to substitutes for VA care (e.g., individual market or employer-based plans), apart from the nature of the state-year's Medicaid eligibility policy. Because our dependent variables are integer counts of enrollment, inpatient days, and outpatient clinic stops, we estimated fixed effects Poisson models3 (p. 668, 805) with robust standard errors to address potential overdispersion4 (p. 561). We used the year-sector-level, non-elderly veteran population, obtained from VA and American Community Survey data, as the exposure variable, which adjusted for the fact that some sectors have greater enrollment and utilization because more veterans reside in them.2
2 Total veteran population counts (elderly and non-elderly combined) were available from the VA at the sector level for all years. We estimated the rate of
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
After we estimated models of VA demand, we restricted the data to year 2008 and predicted enrollment, inpatient utilization, and outpatient utilization with the original Medicaid variable (our baseline) and, separately, with a version that simulated the ACA Medicaid expansion. The ratios of the ACA-based predicted values of enrollment, inpatient utilization, and outpatient utilization to the baseline 2008 predictions provided us with estimates of the proportional changes we might have expected from the ACA's Medicaid expansion provision alone, had it been implemented in 2008. We report state-level aggregates of these proportional changes for all states (i.e., simulating Medicaid expansion for all states) and indicate which states had committed to the optional Medicaid expansion in 2014. From these results, we estimated how much VA demand would change if states that did not expand Medicaid in 2014 did so. All statistical work was conducted using Stata version 10.
3. Results Table 1 provides descriptive statistics for the dependent, independent, and exposure variables at the year–sector level. In our data, an average year-sector has 4457 enrollees that use 1564 inpatient days and 17,660 outpatient clinic stops, though there is considerable variation in all three variables. The Medicaid eligibility variable is 0.089 at baseline, on average, but twice as large (0.18) when based on the ACA Medicaid expansion, if adopted in all states. Not shown in Table 1 are trends in our key variables. Over the 2002–2008 study window, VA enrollment, inpatient days, and outpatient clinic stops grew by 59%, 87%, and 141%, respectively. (In the appendix Section A.2, we provide yearly totals for these variables.) The vast majority of states experienced growth in Medicaid eligibility generosity as well. In the average state, it grew 23% between 2002 and 2008. Also not shown are year–state level statistics for our outcome variables, which are as follows: On average, across the 357 year-states in our sample, there are 49,465 enrollees (std. dev.: 49,157, min: 3857, max: 269,057) using 17,362 inpatient days (std. dev.: 18,579, min: 969, max: 106,970) and 195,992 outpatient clinic stops (std. dev.: 208,618, min: 13,172, max: 1,208,335). Table 2 presents Poisson regression coefficients for the three dependent variables: enrollment, inpatient days, and outpatient clinic stops. All three are statistically significantly and negatively associated with Medicaid eligibility. Where and when policy-driven Medicaid eligibility is higher, VA enrollment and utilization are lower. Enrollment and outpatient clinic stops are also statistically significantly and negatively associated with housing price index. Where and when home values are higher, VA enrollment and outpatient clinic stops are lower. The same is true of employmentto-population ratio. Where and when employment is higher, relative to the population, VA enrollment and outpatient clinic stops are lower. Table 2 shows a lack of statistical significance in association between inpatient days and housing price index or the employment-to-population ratio. This could occur if level and setting (VA or non-VA) of inpatient care is relatively inelastic with respect to changes in wealth or income that housing price and employment changes represent. These results are consistent with our expectations. Medicaid is a substitute for VA care, so when more veterans are eligible for it, fewer would be expected to use the VA. Similarly, when the (footnote continued) growth of the non-elderly VA population from 2006 to 2010 American Community Survey 5-year estimates (not available for earlier years) and extrapolated the proportion of the veteran population that is non-elderly back to 2002. We applied this proportion to the VA population data to obtain annual counts of non-elderly veterans only.
125
economy or housing markets are stronger, veterans are more likely to be wealthier and employed, providing a greater opportunity for take up of individual- or group-market insurance plans— which also substitute for VA care—or a greater ability to afford cost-sharing such plans require. Concerned that year-sectors with small enrollment might be exerting a disproportionate influence on our findings, we conducted a robustness check on a sample without the 5% smallest enrollment year-sectors. Results were nearly identical to those from the entire sample. Because the effects of the results of Table 2 are hard to interpret, we also estimated elasticities for the three dependent variables with respect to Medicaid eligibility and evaluated at the mean of the data. The elasticity of enrollment is 0.11 [95% CI: 0.12, 0.089, p o0.001], of inpatient days is 0.065 [95% CI: 0.12, 0.0098, p ¼0.02], and of outpatient clinic stops is 0.14 [95: CI: 0.18, 0.10, p o0.001]. A more policy-relevant way to understand the results of Table 2 is to use them to simulate what they imply for the ACA's Medicaid expansion. Combining the mean, ACA-simulated Medicaid eligibility value with the elasticity estimates above provides summary predictions. Table 1 shows that the ACA-simulated Medicaid eligibility variable is about double the baseline value at the mean (0.18 vs. 0.09). According to our estimated elasticities, a doubling of Medicaid eligibility corresponds to an 11% reduction in VA enrollment, a 6.5% reduction in VA inpatient days, and a 14% reduction in VA outpatient clinic stops. In absolute numbers, if applied to data for 2008, these would represent a decrease of 0.34M enrollees (from a baseline of 3.06M to 2.72M), of 0.07M inpatient days (1.14M to 1.07M), and 1.97M outpatient clinic stops (14.1M to 12.13M). Table 3 provides state-level results of our simulation. For each state, it lists whether it expanded Medicaid in 2014 and by what proportion VA enrollment, inpatient days, and outpatient clinic stops would fall if it did so, per our simulation. (Though we show predicted changes for all states, for those that do not actually expand, there would be no actual change.) With the exception of Vermont, in which ACA-like expansion had already occurred, VA enrollment, inpatient days, and outpatient clinic stops would fall in each state under the ACA's Medicaid expansion. Depending on state, enrollment would fall to between 0.85 and 0.98 of its baseline level; inpatient days would fall to between 0.90 and 0.99 of its baseline level; and outpatient clinic stops would fall to between 0.80 and 0.97 of its baseline level (all excluding Vermont). Appendix Section A.3 includes the results of Table 3 in figures, stratified by expansion status. The bottom three rows of Table 3 show the average, predicted effects for states that did and did not expand Medicaid in 2014, as well as for all states. On average, in expanding states, VA enrollment, inpatient days, and outpatient clinic stops are predicted to fall to 0.92, 0.95, and 0.89 of their baseline values, respectively. If non-expanding states did expand, VA enrollment, inpatient days, and outpatient clinic stops would fall more, to 0.90, 0.94, and 0.87 of their baseline values, respectively. Put another way, in states that did not expand Medicaid, demand on VA is 10 percentage points higher in terms of enrollment, 6 percentage points higher in terms of inpatient visits, and 13 percentage points higher in terms of outpatient clinic stops than it would otherwise be. If all states expanded their programs, VA demand would be 9%, 6%, and 12% lower by these three measures, respectively.
4. Discussion We used VA and publicly available data to estimate the historical relationships between policy-driven Medicaid eligibility
126
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
levels and VA enrollment and utilization. Consistent with expectations, we found that higher levels of Medicaid eligibility are associated with lower VA enrollment, inpatient and outpatient utilization. Using our estimates, we found that if the ACA's Medicaid expansion had been implemented in all states in 2008, and holding all else constant, VA enrollment, inpatient days, and outpatient clinic stops would have been 9%, 6%, and 12% lower, respectively. Our results suggest that states in which Medicaid did not expand in 2014, VA enrollment, inpatient days, and outpatient clinic stops were 10, 6, and 13 percentage points higher than they would have otherwise been, respectively. These results suggest that Medicaid expansion could considerably reduce the burden of demand placed on VA medical centers. Though not directly comparable, our estimated impact of Medicaid expansion in Massachusetts on the VA in that state is similar in magnitude to the findings of Wong et al.17 They examined the impact of Massachusetts’ 2006 health reform on the VA while we simulated the additional effect of the ACA's Medicaid expansion in that state. In one of their specifications, they found that that state's reform was associated with a 2.4 percentage point decrease in VA enrollment, relative to neighboring New England states, similar to our finding of a 3 percentage point decrease for Massachusetts after the ACA's Medicaid expansion. However, our national results are larger in magnitude, reflecting the ACA's larger pre-post change in Medicaid eligibility in much of the rest of the nation relative to Massachusetts. Our findings are accompanied by several limitations. First, our work is based on historical data available only through 2008. Compared to past Medicaid expansions, the ACA's is larger. Extrapolations warrant caution in general. In the case of the ACA's Medicaid expansion in particular, some have expressed concern
Table 1 Descriptive statistics. Variable
Mean (std. dev.)
[Min–Max]
Dependent variables Enrollmenta Inpatient daysa Outpatient clinic stopsa
4457 (3848) 1564 (1631) 17,660 (16,701)
[36–49,337] [9–19,331] [85–235,711]
Independent variables Medicaid eligibility: baselineb Medicaid eligibility: ACA-simulatedb Housing price indexb,c Employment-to-pop ratiob Exposure variable Veteran 18–64 population
5. Conclusion 0.089 0.18 1.97 62.72
(0.055) (0.030) (0.40) (3.52)
26,705 (22,631)
[0.018–0.36] [0.17–0.36] [1.07–3.48] [52.28–71.58] [132–287,567]
N ¼3962 year-sector observations. a b c
that the large increase in demand for care it motivates might outstrip available supply.1 If that is the case, then Medicaid may be a less viable substitute for VA care than the historical record suggests. Second, we are unable to simulate the effects of the ACA's individual coverage or employer mandate and other features such as the establishment of subsidized exchange coverage. Because VA enrollment satisfies the law's individual mandate, that could encourage veterans to enroll. Also, increased awareness about the mandate and exposure to outreach efforts could increase veterans’ interest in the VA.11 How firms respond to the law's employer mandate could also affect availability of employer-based coverage for veterans, thereby changing the demand for VA coverage among employed veterans.17 In principle, availability of subsidized, exchange-based coverage (for those who qualify for it) could be a substitute for VA care as well. Over a half-million nonelderly veterans (about 40% of uninsured, nonelderly veterans) are potentially eligible for exchange-based subsidies.11 However, exchanges could attempt to inform veterans of their VA eligibility, a potentially countervailing factor.13 Notably, Wong et al.17 did not find a statistically significant increase in the rate of private coverage among veterans after Massachusetts’ coverage expansion, relative to those in neighboring New England states. Finally, our study years (2002–2008) coincide with a period of rapid VA enrollment growth that has abated somewhat. It is possible that our results are driven by a reduction of that growth where and when Medicaid eligibility was relatively more expansive (extensive margin), but that there was little effect on existing VA enrollees (intensive margin). Data available for this study are not amenable to separately identifying the impact of Medicaid expansion on VA utilization by existing vs. new VA enrollees. If our results are substantially driven by a reduction in the rate of VA enrollment, not by changes in utilization by existing enrollees, then we would expect a more modest effect of the ACA's Medicaid expansion if VA enrollment growth continues to moderate. Therefore, future research should estimate these relationships during years of slower VA enrollment growth.
Year–sector level variables. Year–state level variables. Indexed to 1.00 in 1991.
The ACA's Medicaid expansion could play a large role in reducing the burden of demand on VA providers, though some of its effect could be offset by other provisions of the law. VA medical centers in states that did not expand Medicaid in 2014 are likely to have experienced a higher demand, and commensurately longer wait times, than they would have otherwise under Medicaid expansion. As policymakers continue to address VA capacity issues, they should be mindful of the potential role of Medicaid, and that it will change over time as more states adopt the expansion.
Table 2 Poisson regression coefficients. Variable
Medicaid eligibility Housing price index Employment-to-pop ratio
Coefficient (std. dev.) Enrollment
Inpatient days
Outpatient clinic stops
1.20 (0.10)nnn 0.30 (0.0056)nnn 0.019 (0.0018)nnn
0.73 (0.32)n 0.28 (0.016) 0.0065 (0.0056)
1.60 (0.23)nnn 0.068 (0.0082)nnn 0.0077 (0.0029)nn
N ¼3962 year-sector observations. Exposure variable: Veteran 18–64 population. Year and sector fixed effects not shown. n
po 0.05. p o0.01. nnn p o0.001. nn
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
127
Table 3 Simulation results. State
Expanded Medicaid in 2014a
AK No AL No AR Yes AZ Yes CA Yes CO Yes CT Yes DC Yes DE Yes FL No GA No HI Yes IA Yes ID No IL Yes IN No KS No KY Yes LA No MA Yes MD Yes ME No MI Yes MN Yes MO No MS No MT No NC No ND Yes NE No NH No NJ Yes NM Yes NV Yes NY Yes OH Yes OK No OR Yes PA No RI Yes SC No SD No TN No TX No UT No VA No VT Yes WA Yes WI No WV Yes WY No Ave for all Medicaid-expanding states (baseline total) Ave for all non-expanding states (baseline total)b Ave for all states (baseline total)
Assuming expansion, relative proportional change in (and baseline values of) … Enrollmentb
Inpatient daysb
Outpatient clinic stopsb
0.92 (15,129) 0.90 (63,947) 0.91 (43,830) 0.93 (77,661) 0.92 (262,479) 0.87 (54,047) 0.86 (22,598) 0.92 (5813) 0.96 (9017) 0.93 (227,040) 0.86 (109,911) 0.96 (12,074) 0.92 (32,828) 0.92 (19,327) 0.88 (97,547) 0.85 (63,733) 0.86 (29,939) 0.88 (54,081) 0.90 (49,578) 0.97 (36,468) 0.88 (53,956) 0.94 (19,600) 0.90 (74,005) 0.96 (52,304) 0.92 (68,721) 0.87 (38,804) 0.89 (17,593) 0.89 (114,708) 0.86 (10,291) 0.86 (21,366) 0.92 (14,361) 0.93 (43,121) 0.86 (27,938) 0.91 (39,151) 0.98 (150,647) 0.92 (111,550) 0.91 (47,994) 0.92 (46,094) 0.90 (106,399) 0.89 (8168) 0.89 (66,559) 0.87 (14,989) 0.89 (74,409) 0.91 (258,593) 0.89 (18,864) 0.86 (90,543) 1.00 (7623) 0.92 (76,070) 0.9 (56,784) 0.92 (32,913) 0.90 (9485) 0.92 (1.44M) 0.90 (1.62M) 0.91 (3.06M)
0.95 0.94 0.94 0.95 0.95 0.92 0.91 0.95 0.98 0.95 0.91 0.98 0.95 0.95 0.93 0.90 0.91 0.92 0.94 0.98 0.93 0.96 0.93 0.98 0.95 0.92 0.93 0.93 0.92 0.91 0.95 0.95 0.91 0.95 0.99 0.95 0.94 0.95 0.94 0.93 0.93 0.92 0.93 0.95 0.93 0.91 1.00 0.95 0.94 0.95 0.94 0.95 0.94 0.94
0.90 (45,491) 0.87 (288,633) 0.88 (227,522) 0.90 (373,539) 0.89 (1,178,442) 0.83 (231,889) 0.82 (108,549) 0.89 (43,461) 0.95 (33,290) 0.90 (1,216,681) 0.82 (442,031) 0.95 (35,594) 0.90 (146,826) 0.89 (81,741) 0.85 (457,956) 0.80 (287,069) 0.82 (139,512) 0.84 (256,588) 0.87 (230,560) 0.96 (167,982) 0.84 (248,157) 0.92 (92,990) 0.86 (333,467) 0.95 (253,784) 0.90 (317,633) 0.84 (186708) 0.85 (76,303) 0.85 (499,915) 0.82 (34,888) 0.82 (108,171) 0.89 (61,314) 0.90 (168,542) 0.82 (141,894) 0.89 (179,323) 0.97 (620,958) 0.89 (615,658) 0.88 (233,522) 0.90 (218,748) 0.87 (481,676) 0.86 (45,462) 0.85 (280,025) 0.83 (70,355) 0.86 (345,951) 0.88 (1,224,139) 0.86 (78,130) 0.81 (350,784) 1.00 (29,114) 0.90 (308,929) 0.87 (279,329) 0.90 (179,908) 0.87 (44,224) 0.89 (6.64M) 0.87 (7.46M) 0.88 (14.10M)
(3437) (18,869) (25,035) (27,930) (95,675) (17,074) (9162) (3770) (2643) (95,330) (37,727) (2978) (10,331) (7095) (37,705) (21,112) (13,606) (20,976) (22,935) (16,601) (16,323) (5806) (27,028) (14,570) (28,536) (15,705) (6809) (42,338) (2585) (6927) (4045) (11,773) (10,704) (13,680) (43,207) (42,895) (22,833) (18,096) (36,482) (4005) (25,776) (4455) (32,211) (112,962) (8247) (31,446) (1735) (24,951) (19,400) (13,641) (3757) (0.52M) (0.63M) (1.14M)
a
Source: The Advisory Board.2 For non-expanding states, values shown are those simulated as if those states did expand. Proportional changes would be, by construction, 1.00 (no change) for a simulation of non-expansion. b
Conflict of interest This statement accompanies the article "The Effect of Medicaid Expansions on Demand for Care from the Veterans Health Administration," authored (co-authored) by Austin B Frakt, Amresh Hanchate, Steven D Pizer and submitted to Healthcare as an original article. Below all authors have disclosed relevant commercial associations that might pose a conflict of interest:
Consultant arrangements: None Stock/other equity ownership: None Patent licensing arrangements: None Grants/research support: None Employment: None Speakers' bureau: Austin Frakt was under contract with Speakers on Healthcare during preparation of this manuscript. Expert witness: None Other: None
128
A.B. Frakt et al. / Healthcare 3 (2015) 123–128
Acknowledgments This work was funded by the VA HSR&D (Project MRC 13-427) and supported by the VHA Office of the Assistant Deputy Under Secretary for Health for Policy and Planning (10P1A). The views expressed are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs, Boston University, or Northeastern University. The authors have no conflicts of interest to disclose. This work has been approved by the VA Boston Healthcare System's Institutional Review Board.
Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at: http://dx.doi.org/10.1016/j.hjdsi.2015.02.004. References 1. Abraham J. How might the Affordable Care Act's coverage expansion provisions influence demand for medical care? Milbank Q. 2014;92(1):63–86. 2. The Advisory Board. Where the states stand on Medicaid expansion. February. 〈http://www.advisory.com/daily-briefing/resources/primers/medicaidmap〉; 2014 19.05.14.
3. Cameron AC, Trivedi PK. Microeconometrics: Methods and Applications. New York: Cambridge University Press; 2005. 4. Cameron AC, Trivedi PK. Microeconometrics Using Stata, vol. 5. College Station, TX: Stata Press; 2009. 6. Currie J, Gruber J. Health insurance eligibility, utilization of medical care, and child health. Q J Econ. 1996;111(2):431–466. 7. Cutler DM, Gruber J. Does public insurance crowd out private insurance? Q J Econ. 1996;111(2):391–430. 8. Frakt AB, Carroll AE. Sound policy trumps politics: states should expand medicaid. J Health Polit Policy Law. 2013;38(1):165–178. 11. Haley J, Kenney GM. Uninsured Veterans and Family Members: State and National Estimates of Expanded Medicaid Eligibility Under the ACA. Urban Institute. March. 〈http://www.urban.org/publications/412775.html〉; 2013. 14.05.14. 12. Haley J, Kenney GM. Uninsured veterans and family members: national and state estimates and new coverage options under the ACA. US Department of Veterans Affairs Cyber-Seminar. 21 November. 〈http://www.hsrd.research.va. gov/for_researchers/cyber_seminars/archives/773-notes.pdf〉; 2013 14.05.14. 13. Heiss C, McMahon SM. Veterans and the ACA: How Health Reform Boosts Eligibility for VA Health Care. Hamilton, NJ: Center for Health Care Strategies; 2012. 14. Hendricks A, Gardner J, Frakt A, et al. What can Medicaid data add to research on VA patients? J Rehabil Res Dev. 2010;47(8):773–780. 15. Kizer KW, Jha AK. Restoring trust in VA health care. New Engl J Med. 2014. 16. Office of the Assistant Deputy Under Secretary for Health for Policy and Planning. 2011 Survey of veteran enrollees’ health and reliance upon VA. Department of Veterans Affairs. 17. Wong ES, Maciejewski ML, Herbert PL, Bryson CL, Liu CF. Massachusetts health reform and veterans affairs health system enrollment. Am J Manag Care. 2014;20(8):629–636.