Utilization of VA and Medicare Services by Medicare-Eligible Veterans: The Impact of Additional...

The Veterans Health Administration (VA) is the largest integrated healthcare system in the United States, providing comprehensive healthcare services to approximately 4 million U.S. veterans-a population that is older, sicker, and more likely to be male than the general population (Wilson and Kizer 1997). Because of its relatively small service population, VA access points have historically covered large service areas. Travel distances for those veterans who are remote from VA medical centers (VAMCs) may restrict access to these services, and restricted access may result in underutilization of services (Leape et al. 1999; Mirvis and Graney 1999). Indeed, long travel distance to care has been associated with veterans' using less outpatient (Burgess and DeFiore 1994) and general VA care (Mooney et al. 2000).

To address this problem, the VA has tried to improve access to care by establishing community-based outpatient clinics (CBOCs) (Chapko et al. 2002). These clinics were designed both to improve access to care for historically underserved veteran populations, such as rural veterans who live a significant distance from VA hospital-based care, and to improve care coordination by encouraging veterans' use of the VA as the primary source of healthcare services (U.S. Congress 2001). Although initial studies of CBOCs concluded that the clinics were attracting new patients to the VA, these patients appeared to be using the VA mostly for primary care services; referral to VA specialty care services at VA hospitals was not seen to the degree anticipated (Fortney et al. 2002).

Andersen's (1995) evolving behavioral model of health services use identifies the healthcare system environment (including improved care access) in concert with enabling resources (such as insurance or enrollment coverage for healthcare) as influential in patients' use of health services. We were interested in determining whether VA patients who were taking advantage of the improved access to care afforded by CBOCs and enrolled in multiple systems of healthcare might be using health services differently than those who did not use CBOCs. Therefore, using older, Medicare-enrolled VA patients as the unit of analysis, we sought to determine the relationship between use of CBOCs for VA primary care and overall reliance on and use of the VA and the private sector for primary care, specialty care, and inpatient care. To examine a population with high healthcare needs (Weeks et al. 2004) and restricted healthcare access, we limited our analysis to the predominantly rural setting of northern New England.

METHODS

Our method of identifying patients, obtaining data, and performing analyses to determine the relationship described here is outlined in Figure 1 and described in detail below.

From the VA Patient Treatment and Outpatient Clinic files we obtained patient-level information on age; date of birth; gender; ZIP code of residence; social security number; and inpatient and outpatient VA utilization, including the location of that care (i.e., a CBOC or a VA medical center). Because geographic variation in health services utilization has been demonstrated for the VA (Ashton et al. 1999) and for the general Medicare population (Wennberg 1999), we focused our study on veterans living within a single VA regional service delivery network: VA's New England Health Care System. This system was established in 1996, serves veterans in the six New England states, and has the highest Medicare enrollment in the VA (Wright, Lamkin, and Petersen 2000). Because we were interested in examining veterans who had increasing access to CBOCs and who lived in a predominantly rural setting, we limited our analysis to those who lived in the three northern New England states of Vermont, New Hampshire, and Maine. During the time period examined, access to CBOCs increased in this region. On January 1, 1997, ten CBOCs were operational in northern New England, two of which had opened during 1996. Over the three-year study period, four new CBOCs were opened: one in 1997 and three in 1998.

We then used ZIP codes to identify VA patients who lived in northern New England and VA treatment files to limit the analysis to those who obtained all of their VA services through the New England Health Care System. For analytical purposes, we developed three cross-sectional cohorts of veterans, each of which was enrolled in the VA, with enrollment defined as having used the VA within a three-year period. For instance, the 1997 cohort had used at least one VA service between the beginning of 1995 and the end of 1997. We collected information on VA inpatient and outpatient utilization from these files.

Veterans age 65 and older are likely to be enrolled in Medicare and use the private sector for some healthcare services and the VA for others (Passman et al. 1997; Wright et al. 1997). Because we were interested in understanding whether CBOC utilization might be associated with service use in the private sector, we used social security numbers, birth dates, and gender to probabilistically link these identified veterans to private-sector utilization data obtained from the Medicare denominator, 100 percent Medicare Provider Analysis and Review, and Outpatient and Physician Supply files for 1997 to 1999 as previously described (Fleming et al. 1992). From Medicare files, we obtained private-sector inpatient and outpatient utilization data and determined whether each patient's enrollment was in fee-for-service or HMO Medicare and whether each patient's Medicare eligibility was a result of age or disability. We excluded all HMO enrollees because Medicare databases do not collect service utilization data on HMO patients; without excluding this group, Medicare utilization in this subpopulation would have been underestimated. Further, we limited our analysis to Medicare-enrolled veterans who were at least 65 years old and Medicare eligible by virtue of age. (Although some veterans younger than 65 may be eligible for Medicare because of disabilities, they may have inherently different utilization patterns than veterans who are eligible for Medicare by virtue of age.) Finally, we eliminated females from the analysis because relatively few females are enrolled in the VA system.

Our linking of VA and Medicare datasets produced a stream of VA and private-sector utilization data for each enrolled veteran for each cohort year. We wanted to compare utilization of common outpatient settings and types of visits across both systems. Therefore, from VA "stop codes" and Medicare "coding categories" we identified the following utilization categories: primary outpatient care, specialty outpatient care, and inpatient admissions. For each inpatient admission, we also determined lengths of stay. We assigned each VA and Medicare-funded outpatient visit to either primary or specialty care and enumerated the total number of visits in each category by each veteran each year.

Because we were interested to determine whether use of a CBOC was associated with VA patients' reliance on the VA as a healthcare system, we developed a VA reliance ratio: the ratio of VA to private-sector utilization. For example, if an enrolled veteran used the VA for three outpatient visits and used Medicare-funded private-sector services for six outpatient visits, the VA reliance ratio would be 3:9 or 0.3. We felt that this relative measure, together with an overall measure of utilization, would help us understand whether VA patients who had access to multiple systems of healthcare might be using VA CBOCs for primary care and the private sector for specialty care.

To ensure that the population we were examining did indeed live in a predominantly rural setting where veterans' healthcare needs have been demonstrated to be the greatest during the time period studied (Weeks et al. 2004), we categorized patients using the U.S. Department of Agriculture's rural-urban commuting area (RUGA) code designation (Economic Research Service 2001), a ten-point designation of rural and urban status based on travel and shopping patterns and designated at the county level. We then used the University of Washington's probabilistic ZIP code to county crosswalk file to match veterans' ZIP codes to RUCA designations, and we defined veterans as living in a rural setting if they had rural RUCA codes (i.e., codes 7 through 10). We found that 79 percent of the 1997 cohort, 81 percent of the 1998 cohort, and 81.8 percent of the 1999 cohort lived in rural ZIP codes.

To develop comparison groups that would answer our question, we used the VA patient treatment files to identify the locations of VA primary care utilization. We classified these older, male, northern New England VA patients into three groups: those who used CBOCs exclusively for the primary care services that they obtained through VA (CBOC only), those who used VA medical center (VAMC) hospitals exclusively for the primary care services they obtained through the VA (VAMC only), and those who accessed both CBOCs and VAMCs for the primary care services they obtained through the VA (mixed).

Analysis

Because utilization data were highly skewed, we used Wilcoxon nonparametric tests to compare differences between systems of care-VA and private sector-and Kruskal-Wallace nonparametric tests to compare trends across time within a system of care. We present mean values in Table 1, discussed in the "Results" section below. We analyzed all data using SPSS version 11.0 (2001).

RESULTS

Older male veterans in northern New England who received their primary care exclusively from CBOCs obtained the vast majority of their care through the private sector and funded by Medicare (see Table 1). The CBOC-only group obtained very few specialty care services or inpatient services through the VA. For this group, utilization of all VA services, except bed days of care, trended significantly downward over time. On the other hand, this group used statistically increasing amounts of private-sector primary and specialty care outpatient services over the time period examined.

The mixed group used more VA and fewer private-sector healthcare services when compared to the CBOC-only group. The mixed group received the majority of primary care services through the private sector. While the mixed group obtained statistically more private-sector inpatient bed days of care and specialty care outpatient services in 1998 and 1999 and more private-sector inpatient stays in 1999, differences were not nearly as dramatic as for the CBOC-only group and were not likely to have much clinical meaning. In this group, VA utilization of each service type trended significantly downward, while private sector use of each service type trended significantly upward.

By definition, the VAMC-only group obtained all of their VA primary care through a VAMC and did not use CBOCs. Even within this group, the proportion of care obtained from the VA was lower than that obtained through the private sector, except for specialty care, where statistically more VA care was provided each year, and for inpatient care in 1997. Although VA specialty care remained the same over time, all other VA healthcare utilization trended significantly downward while all private-sector care trended significantly upward.

Over the three-year period, the mixed group grew the fastest (99 percent increase), followed by the CBOC-only group (67 percent) and the VAMC-only group (13 percent). For every service type and within every group, the reliance on the VA for healthcare services trended downward (see Figure 2). The VAMC-only group remained the most reliant on VA for specialty and inpatient care, although that group's reliance decreased over time. The mixed group was the most reliant on the VA for primary care services; a decline in reliance for specialty and inpatient care was more dramatic for the mixed group. The CBOC-only group had virtually no reliance on the VA for inpatient or specialty care services; a decline in reliance on the VA for primary care services was apparent.

CONCLUSIONS

Older Medicare-eligible male veterans from northern New England who obtain the entirety of their VA primary care services from CBOCs appear to rely on the private sector for the large majority of their primary, specialty, and inpatient care needs. This finding suggests that this group uses the VA primarily for supplemental care, possibly as an access point for a pharmacy benefit that is likely to be less costly than that available through Medicare. Although the CBOCs enhance access to care for rural veterans, they provide a minority of the veterans' primary care and do not appear to act as a feeder system to local VA medical centers. Of some concern is the finding that trends moving away from reliance on the VA for primary, specialty, and inpatient care were apparent for all three groups examined. These trends, effected through an absolute decrease in utilization of VA services and an absolute increase in utilization of private-sector services, suggest decreasing relevance of the VA to this service population.

This finding is counterintuitive. Over the recent past, in an effort to attract new patients, the VA has expanded the number of CBOC access points dramatically. Despite the increased access to outpatient care, relatively few in our study obtained their VA primary care exclusively from such clinics. For those who obtained primary care services from both CBOCs and the VAMC (the mixed group), the overall higher utilization through both the VA and the private sector suggests that some of these services were duplicative-and possibly wasteful-and that the care lacks coordination.

These findings have implications for managers of healthcare delivery systems. First, similar to the risks of a loss leader in the business world, it appears that providing a new service may not result in the provision of additional services by the host organization. In a rural setting in particular, local access for one type of service may not translate into use of a more distant service with the same provider. In addition, our findings suggest that efforts to improve access may be associated with lack of coordination of care. The CBOC-only group used the most overall primary care services but considerably fewer specialty and inpatient care services. Again, this suggests that the primary care provided at CBOCs was additive to existing care being provided through Medicare.

Our study has several limitations. First, we did not control for case mix. Although a previous study found that case mix was similar between veterans using a mobile clinic in a rural area and those using a VA medical center (Wray et al. 1999), it is possible that veterans "mix and match" services-using a variety of VA and Medicare services-to meet their individual Healthcare needs (Petersen and Wright 1999).

Second, we did not correct for socioeconomic factors. It is possible that factors such as income and insurance coverage influenced the system in which the different groups obtained care, as has been found in other work (Weeks et al. 2002, 2003).

Third, we were unable to track individual-level data over time. It is possible that utilization patterns of individual patients change over time, as patients are exposed to particular types of service. However, from a management standpoint, the sequential, cross-sectional analysis that we used can give a good indication of static needs.

Fourth, we examined only VA care and private-sector care that was paid for by Medicare. Given that we limited our analysis to VA patients who were also enrolled in Medicare, we feel confident that we captured the vast majority of private-sector care obtained by the patients we studied. However, it is possible that we missed care provided to military retirees at active military bases. Veterans who live in northern New England are reasonably close to the Brunswick Naval Air Station Health Clinic in Brunswick, Maine; the Portsmouth Naval Shipyard Health Clinic in Portsmouth, New Hampshire; and the Hanscomb Air Force Base Health Clinic in Bedford, Massachusetts. Thus, we could have missed some outpatient care provided to military retirees in these locations. However, given their location in the far eastern part of the geographic area studied, the limited nature of the services provided at these locations, and their limited hours of operation, we do not believe that our population of study obtained enough healthcare services through these venues to change our findings.

Fifth, the makeup of the cohorts that we examined may have changed over the study period. While a large majority of patients in one year's cohort are likely to be in the following-year and previous-year cohorts, the growth in the number of patients over the study period suggests that these new patients-patients who had not previously been enrolled in the VA system and could have different utilization patterns than more established patients-are likely to influence the average rates and reliance figures reported. Thus, it is possible that our results reflect a system transitioning from a hospital-based system to an outpatient-clinic-based system, one that has not reached a steady state. From a management perspective, these changes reflect the realities of an ever-changing service population that VA managers must face and suggest that the trends that we identified may persist in the future should enrollment patterns continue.

Finally, our study was limited to a specific population in a specific area of the country that established CBOCs early. During the study period, the number of CBOCs grew by only 44 percent in the study area compared with a growth of 124 percent in the rest of the nation; by the end of the study period, 78 percent of all currently operating CBOCs in this area were operational, whereas in the rest of the country only 64 percent were operational (Planning System Support Group 2003), suggesting a higher rate of expansion of CBOCs in the rest of the country compared with that in the study area both during and since the study. In other words, since the study, the number of operating CBOCs has increased by 28 percent in the study area and 56 percent outside of the study area. While our findings may be useful for managers still establishing CBOCs and may anticipate future findings, particularly in rural areas, generalizability may be limited.

Despite these limitations, our study sheds light on utilization patterns of elderly, male veterans in a rural setting and suggests actions for policymakers. Our findings suggest that, in this setting, improved access to VA care for those veterans who live some distance from specialty and inpatient care access appears to provide complementary, not substitutive, services for those who use both VAMCs and CBOCs. This raises the question of the efficiency of these access points from the system perspective: although CBOCs appear to be efficient when analyzed as discrete units (Chapko et al. 2000), their overall contribution to care may be limited if the services provided are duplicative. The very low reliance on VA for some veteran users suggests a lack of coordination of care between the VA and Medicare. Improving coordination of care across systems may not only enhance patient outcomes (Wasson et al. 1984) but also result in cost savings to taxpayers. Analyses of the effectiveness and efficiency of adding access points to healthcare systems should be conducted from the organizational perspective, with particular emphasis on examining the possibilities of encroachment, worsened coordination of care, and potential overuse of healthcare services.

PRACTITIONER APPLICATION

Normand E. Deschene, FACHE, president and CEO, Lowell General Hospital, Lowell, Massachusetts

The relevancy of the Veteran's Health Administration's (VA) strategy to improve access to care is targeted in this study, and the authors specifically review the effectiveness of the community-based outpatient clinics (CBOCs). They point out that travel time has historically been tied to veterans' utilization of the VA, and the large service areas of VA medical centers (VAMCs) have contributed to underutilization of these facilities. As a means of improving access to healthcare services, the VA has focused its efforts on expanding the number of these community-based clinics. Beyond the primary goal of improving access, the strategy is also aimed at improving the coordination of care as well as increasing the utilization of VA hospitals by eligible veterans. Complicating these issues is the fact that most veterans are Medicare eligible and thus have the opportunity to pursue healthcare outside of the VA delivery system in the private sector.

The methodology employed by the researchers to test the effectiveness of the VA's strategy was to study the relationship between Medicare-enrolled VA patients' use of CBOCs for primary care versus use of VA and private-sector providers for primary, specialty, and inpatent care in a rural setting. They used files of VA patients in the New England Health Care System. Because of their interest in rural patients, the authors limited their study group to those veterans living in northern New England. This is especially relevant because, over the three-year study period, the VA has opened a number of new CBOCs in this area. The use of Medicare files necessitated the elimination of certain cohorts such as those enrolled in Medicare HMO programs; however, the goal of comparing utilization data across the VA and Medicare system did provide a sample needed to extrapolate the study's findings.

The findings were interesting in that they clearly demonstrate that veterans in the study group who received some care from a CBOC also received the majority of their healthcare from the private sector. Further, even veterans who got most of their primary care from CBOCs obtained little specialty care services or inpatient services from VAMCs. Perhaps the most surprising discovery was that despite the opening of more CBOCs during the study period, the overall reliance on the VA for healthcare services declined in all of the measured groups. The authors suggest that instead of replacing utilization of private-sector healthcare, the CBOCs have been utilized for additional or complementary services that perhaps are not covered by Medicare.

This study cautions healthcare managers who plan to establish new points of access for a capitated service population: patients may use the new access points somewhat differently than anticipated. In particular, patients may target particular services to enhance their own benefits package, use the new access points to complement rather than reduce overall service utilization, or cannibalize market share from other clinics or hospitals in the same system, thereby driving up overall costs of care. Healthcare managers should anticipate and monitor changes in patients' patterns of healthcare utilization when new access points are established.