In order to investigate the effects of MSHO enrollment as compared to MSC+ enrollment on a range of utilization outcome measures, this study analyzed 2010-2012 data from an extensive database containing individual-level characteristics and claims/encounter data. We conducted both descriptive analyses and multivariate analyses of the enrolment choice and service utilization, with additional stratification by urban vs. rural area. These results provide a detailed assessment of utilization differences that can be achieved with an integrated care model (MSHO) vs. a coordinated but not fully integrated care model (MSC+). This final section of the report highlights the study's principal findings, discusses policy implications, notes study limitations and outlines potential areas for additional research.
4.1. Findings and Policy Implications
First, Minnesota dual eligibles electing MSHO enrollment differed from those remaining in the MSC+ program on a range of individual characteristics. MSHO enrollees tended to be older, female, to have more medical conditions and disabilities, to have died during the year, and were slightly more likely to live in rural areas of the state. These differences were relatively minor except for age. The proportion of MSHO enrollees in the age 65-69 age group were roughly half that of MSC+ enrollees (14.4 percent vs. 27.6 percent), and MSHO enrollees were more likely found in the 90+ age group (18.9 percent vs. 13.6 percent). While most dual eligibles eventually enroll in MSHO, they do not necessarily do so upon turning age 65. Unless they explicitly choose MSHO, they will be placed in MSC+. Most dual eligibles probably have little experience with being in two plans for their care prior to becoming dually eligible, and may not understand the benefits, or even the concept, of an integrated plan. The continued educational outreach by the state and managed care plans regarding the MSHO option to dual eligibles over time reaches a greater proportion of dual eligibles, with beneficiaries age 75-79 being the first 5-year age group with a majority in the MSHO vs. MSC+ program.
Second, although MSC+ enrollees were increasingly likely to enroll in MSHO over time, MSHO enrollees rarely opted out of the MSHO program once enrolled. Very few of those who were in MSHO in January of a year ever switched to MSC+ during that year, but 12.8 percent of those who were in MSC+ in January of a year switched to MSHO by the end of the year. Although MSHO enrollees can disenroll from MSHO and elect MSC+ effective at the beginning of the next month, the finding that almost none do suggests high satisfaction with services received under MSHO. Conversely, switching to enrollment in MSHO from MSC+ increases with age, which suggests that dual eligibles may become more aware of the potential choice of MSHO or place more value on the benefits available under MSHO relative to MSC+ as they get older, particularly after attempting to coordinate care between Medicare and Medicaid themselves.
Third, compared to MSC+ enrollees, MSHO enrollees had lower hospital and ED use but greater prevalence of primary care service use. Both before and after controlling for differences in observed individual-level and area-level characteristics, MSHO enrollees received less care in hospital settings than MSC+ enrollees. This finding was in spite of MSHO enrollees being somewhat older and having somewhat greater prevalence of selected medical conditions and disabilities. At the same time, MSHO enrollees had a much higher prevalence of primary care use. Although these analyses could not assess whether more frequent primary care use led to lower hospital-based care, it is consistent with the goal of connecting those who rely more heavily on inpatient and ED use to community-based providers as a strategy to reduce reliance on hospital-based care for care more appropriately provided in the community. With the joint capitation of Medicare and Medicaid under MSHO, health plans have a strong financial incentive to expand community-based care to reduce ambulatory-care sensitive inpatient and ED use since they reap the financial benefits of any reduced hospital-based care. While we don't have individual-level data on the receipt of care coordination, such coordination is a core component of the MSHO model, including coordination across care settings (e.g., from the hospital to home) and between care providers (e.g., home and community-based care providers and PCPs). Beyond the program model that encourages an investment in care coordination, health plans may provide additional financial incentives to providers as part of their contracts to encourage greater investment in care coordination strategies to reduce potentially avoidable hospital and ED visits, including hospital readmissions.
Curiously, MSHO enrollees, while having a greater prevalence of any primary care visits, if they had any, had 36 percent fewer visits than MSC+ enrollees, and, for those with a specialist visit, 36 percent fewer specialist visits. One reason for fewer visits may be that the PCP and their affiliated staff were able to provide more comprehensive, coordinated care during each visit, resulting in the need for fewer visits over time. Conversely, dual eligibles in MSC+ may have elected that program to continue to see a greater number of specialists, or to have more visits with certain specialists with whom they had long-term relationships.
Fourth, compared to MSC+ enrollees, MSHO enrollees were no more likely to have a long-term nursing home admission, were more likely to have any HCBS but less likely to have any assisted living facility use. These findings took the greater age and slightly greater prevalence of medical conditions and disabilities among MSHO enrollees into account. Potentially, the greater prevalence of HCBS and assisted living facility use helped prevent some long-term nursing home use. Even though MSHO enrollment was slightly more prevalent in rural areas, the urban/rural analyses conducted as part of the study showed there were no differences over time in the use of assisted living in rural areas resulting from MSHO enrollment, so there did not seem to be a substitution effect of assisted living for nursing home use.
The lower use of inpatient and ED services among MSHO enrollees when compared to MSC+ enrollees, and no greater prevalence of nursing home use, policymakers may question why MSHO enrollees were only 13 percent more likely to have any HCBS use and not much greater. Presumably, this likely resulted in substantial costs savings that could have been used to provide HCBS to even more enrollees, given their greater relative age and disability. However, one reason Minnesota staff gave for the MSHO HCBS use rate not being even higher was that the first assessment at age 65 is made by county-based long-term care consultation services staff as opposed to the managed care plans and, as a result, both MSHO and MSC+ enrollees with similar care needs have a common LTSS benefit package. It is possible, however, that MSHO enrollees receive higher levels of HCBS services over time if the MSHO care coordination activities are more likely to identify higher HCBS needs over time. Unfortunately, data on the level of HCBS use was not available to this study.
While the primary analysis of LTSS service use examined the average effects of MSHO over the 2010-2012 period, the estimates of the temporal trends in use that are reflected in the year dummy variables (e.g., year 2011 and year 2012 as compared to year 2010) in the logistic regression models are also informative. Independent of MSHO and MSC+ enrollment, any long-term nursing facility use was 9 percent lower in 2011 than in 2010 and 15 percent lower in 2012 than in 2010. Similarly, compared to 2010, any HCBS use was higher (5 percent and 10 percent respectively for 2011 and 2012) and any assisted higher living use was also higher (5 percent and 6 percent respectively for 2011 and 2012). These results are consistent with unpublished tabulations computed by Minnesota staff for program management5 that show declining nursing home use and increasing HCBS use over time, and likely reflect, at least in part, the state's efforts to rebalance the LTSS system toward community living.
Minnesota's overall efforts to reduce nursing home use are impressive. Minnesota staff reported as potential contributing factors to this reduction over time some market place changes for senior living in the state (e.g., expanded private investment in elderly housing, including assisted living) as well as policy changes implemented by the Minnesota Aging Division and the Minnesota Board on Aging, including initiatives to reduce nursing facility beds, education campaigns about alternatives to nursing homes and initiatives such as Return to Community which provides advice/assistance to all enrollees in nursing homes (both private pay and Medicaid) to avoid long-stays.
The MDS results showed little or minimal difference between MSHO and MSC+ enrollees in the levels of physical and cognitive impairments. A high proportion (over 90 percent) of new nursing home admissions from both the MSHO and MSC+ groups are totally dependent or require extensive assistance in at least one ADL, and about a quarter of them in both groups have severe cognitive impairment upon admission. It is likely that the use of various HCBS by MSHO and MSC+ enrollees has helped to delay nursing home entry until the declines in their health and functional status necessitate nursing home admission. We also found that a higher proportion of newly admitted nursing home residents among MSC+ enrollees are racial/ethnic minorities than among MSHO enrollees, which is most likely related to a slightly greater percentage of minorities in MSC+.
Fifth, MSHO program dual eligibles, compared with other elderly non-dual eligible Medicare beneficiaries, differed on important individual characteristics. Medicare-only enrollees were more likely to be younger, healthier, less likely to die, and to have much less service use on all outcome measures examined than MSHO enrollees. Although we do not have a specific measure of income in our datasets, MSHO enrollees, who must meet Medicaid income-eligibility standards, will also have much lower income than Medicare-only enrollees. MSHO enrollees are much more similar to MSC+ enrollees, who are also dual eligibles, than they are to Medicare-only beneficiaries. Dual eligibles overall are much different than Medicare-only beneficiaries.
Finally, it is always important to consider the potential for selection bias in analyses comparing enrollees in different programs. The risk of selection bias is a common limitation of such research, and methods for assessing its effects on impact estimates are nascent. This study lacked observed data on important individual and family characteristics that would likely affect both enrollment in MSHO and the service use measures examined here, leading to potential bias in the estimates of the direction and magnitude of MSHO effects. Our ability to quantitatively assess the potential for selection bias due to unobserved characteristics in our impact estimates using the method developed by Oster (2015) is an advancementfrom prior studies. We found that, if we had been able to incorporate the unmeasured variables, our estimates of MSHO effects would be unlikely to change direction, and, in many cases, could potentially be larger in magnitude. Overall, our findings should provide encouragement to policymakers that an MSHO-like model, applied to a similar population and system of health care, would provide strong reductions in hospital-based care.
4.1.1. Limitations of Analyses
This study faced six limitations, most related to data issues. First, we lacked data for some important individual and family characteristics that would likely be important predictors of MSHO enrollment and care utilization. For example, although our sensitivity analysis for potential selection bias indicated that the omission of unobserved characteristics would be unlikely to change the substance of the findings reported above, data on family composition and availability of non-paid helpers would have been helpful to better estimate the MSHO program impact service use. In addition, we lacked a reliable indicator of dual eligibility status prior to age 65, which is likely to be an important factor in the decision to move from MSC+ to MSHO. Additional individual characteristics for which we lacked data were race/ethnicity and a broader set of medical conditions and severity of medical conditions for case mix controls.
Second, data limitations prohibited developing as many outcome measures as desired for a complete assessment of the impacts of MSHO on health care and LTSS use. For example, we were not able to construct a measure for non-physician providers such as nurse practitioners and physician assistants, who may have been additional primary care service providers that are not captured in our data.
We also lacked data for intensity of LTSS measures. Count measures reflecting service intensity for HCBS such as hours per user for PCA service, homemaker/chore, or home health aide, or months of assisted living facility use, were not possible to calculate, given the resources of this project. Therefore, it was not possible to assess whether HCBS users in MSHO were also getting more HCBS hours for PCA services or months of assisted living services than HCBS users in MSC+. Therefore, an outstanding question regards whether the intensity of HCBS use may be associated to some degree to the lower hospital and ED service use in MSHO. In this respect, Minnesota staff suggested that higher intensity of use for HCBS in MSHO for individuals with similar needs was not likely given the uniform assessment protocol and the standard LTSS benefit package for both the MSHO and MSC+ programs; however, if the greater care coordination in MSHO results in identifying those needs more quickly MSHO enrollees could obtain HCBS services more quickly, thereby increasing HCBS services use over time.
Third, we were not able to analyze the years 2007-2009, because of a lack of consistent data over the period due to the phase-in of aspects of the program during this period. Therefore, we restricted our data analysis to 2010-2012, which corresponds to a period in which MSC+ and MSHO were both fully-phased in across the state. Our 2010-2012 year-level covariates did show interesting trends even across these three years (e.g., nursing home use across both programs was decreasing, HCBS use was increasing), which highlights the importance of analyzing multiple years of data during period of program and policy change.
Fourth, it is important to qualify these findings in terms of their generalizability to other states. The MSHO program and encounter data used in this analysis obviously are relevant for Minnesota, and they may well be generalizable to others states with similar populations, health care systems, and program and provider characteristics. Minnesota has long been a leader in health care service delivery, so may be unlike other states in their programs and policies. That said, other states that are in need of greater LTSS rebalancing and care coordination may obtain higher or lower effects from an MSHO-style program than this study found in Minnesota.
Fifth, the Office of the Assistant Secretary for Planning and Evaluation had requested that the study use LTSS assessment data maintained by the State of Minnesota. The LTSS assessment is conducted in the community annually for people who need or may need LTSS, but not for others in the community, and it is not administered in the nursing home. Furthermore, Minnesota staff said there is not a consistent criteriaabout who would receive an LTSS assessment because they can be requested by providers, plans, or a consumer. Consequently, we did not attempt to use the state's assessment data, given that program enrollees may have had different probabilities of being assessed.
Finally, the study had ambitious goals, and included some research questions for which we lacked data or high quality data to address, including the following:
Data on MSHO program and organizational features such as types and intensity of care coordination activities were not available. Such data would have helped highlight managed care plan and provider level activities related to the dramatic reductions in inpatient use.
We could not address a research question on how service use patterns changed as MSC+ and MSHO plans evolved because of the lack of data quality for the 2007-2009 during the MSC+ phase-in period.
The research question concerning differential program use and patterns of service use according to race/ethnicity (after controlling for disability) could not be analyzed because of lack of consistent race/ethnicity data.
The research question on the characteristics and service use patterns of beneficiaries who become dual eligibles before vs. at age 65 could not be addressed because of data quality concerning pre-age 65 dual eligibility status.
4.2.1. Future Analyses
The dataset used for this study, while providing a rich source of individual and service-level outcomes, had the limitations noted in the prior section. The State of Minnesota is making further improvements in their individual-level and service-level dataset on elderly programs, and future work using that data would be important to add to the contributions of this study. As noted earlier, additional analyses could be performed on LTSS service intensity for those receiving such services, and additional outcomes could be developed for nurse practitioner, physician assistant, various therapies, preventable and avoidable hospitalizations, hospital readmission, and importantly, managed care plan and provider level care coordination activities to understand how MSHO achieved the success found in this study. The state's dataset could also be used to construct additional measures of individual characteristics that would improve the analysis, including additional measures of current health and disability status as well as historical measures, such as Medicaid and Medicare eligibility prior to age 65.
Additional years of data beyond the 2010-2012 period of analysis in this study would also be important to see how the MSHO and MSC+ programs evolve in a period of great change in health care service organization with the adoption of accountable care organizations and other innovations. More recent data potentially could also be used to better understand the state's FAI with CMS that began in September 2013. Additional years of data could also be used to assess the state's continuing efforts to rebalance its LTSS system toward community living, which complements any MSHO impacts.
This study's methodology using reduced form (regression-based) analyses provided important understanding of MSHO impact, but additional, formative analyses could also be undertaken using structural equation modeling to understand the pathways through which reductions in hospital and ED use occurred. While it could not be tested as part of this study, we hypothesize that greater access to and coordination of care by PCPs, and perhaps non-physician providers, led to the finding of no difference in long-stay nursing home use between MSHO and MSC+, even though MSHO enrollees were older and had somewhat more medical conditions and disabilities. Structural equation models would allow identifying causal factors and their role and effects on downstream outcomes so that the State of Minnesota could know the relative impact of various program and provider level activities on health service outcomes. For example, this analysis potentially would show whether greater intensity of HCBS use among MSHO enrollees kept long-term nursing home use no higher and hospital (inpatient and ED) use lower than that of MSC+ enrollees.
Finally, although precluded by data constraints in the study dataset, important subgroup analyses where sample size permits could be investigated. In particular, data on those eligible for Medicaid and Medicare prior to age 65 would allow distinguishing the impacts of MSHO on those who became disabled prior to 65 from the impacts on the aged population, a potentially important distinction for understanding LTSS needs over time. Data on physician use and HCBS patterns across the two programs and over time, as well as the nature of inpatient admissions (and readmissions) and ED use would improve the understanding of the differences between a fully integrated care model represented by MSHO vs. a coordinated but not integrated care model (MSC+).