• No results found

Summary

In document Goodman_unc_0153D_16730.pdf (Page 144-148)

This study aimed to analyze what impact, if any, the post-ACA MA stars methodology had on the products and services offered by Medicare Advantage plans serving socially and economically vulnerable Medicare beneficiaries. In seeking to answer that question we examined the presence or absence of a series of supplemental and enhanced benefits in MA-PD plan offerings during 2014 and 2015, which represent, respectively, the last year of the quality bonus demonstration program and the first year that the stars bonus methodology was fully in effect. We tested the hypothesis that the larger the proportion of low-income beneficiaries who participated in an MA plan offering and the greater the level of deprivation of the county in which the plan was offered would have a positive effect on the probability that a plan offering included certain supplemental benefits designed to offset SES-related barriers to high-quality care identified by phase 1 key informants and a negative effect on the probability that a plan offering included a premium payment requirement.

The analysis examined the proportion of plans offerings that included transportation, meals, nutrition, EDM, telemonitoring, and a premium payment requirement as well as the risk differences and marginal effects of the proportion of low-income members participating in the plan and the relative deprivation in the county in which the plan was offered on the presence of these benefits. We controlled for a series of policy-relevant independent variables: contract star score, county star bonus caps, weighted average plan membership, and weighted average low-income subsidy eligible plan membership.

Our results found that among all MA plans, county-level deprivation (ADI) was significant both within and between years for the inclusion of a meals benefit. Risk differences were significant for the inclusion of a nutrition benefit between years but were not consistently significant between county ADI quintiles. Examining the EDM benefit, we found statistically significant differences by year in low, medium, and high ADI quintiles (quintiles 1, 3, and 5), but not in the intervening quintiles (2 and 4). Risk differences for the inclusion of an EDM benefit were consistently statistically significant in comparisons of high versus low ADI quintiles (quintiles 1, 2, and 3 when compared individually to quintile 5) in both

years. We found the marginal effect of the addition of 1% of low-income subsidy (LIS) eligible membership to be positive for transportation and meals but negative for all other dependent variables.

We then conducted a post-hoc analysis to determine the impact, if any, of plan SNP designation on the presence or absence of the dependent variables. In examining the risk differences between SNP and non-SNP plans we found consistent, substantial and statistically significant differences for

transportation, meals and nutrition. These differences appear to indicate that SNP plans, the vast majority of which are designed to serve individuals who are dually eligible for Medicare and Medicaid, are including supplemental benefits aligned with breaking down SES-related barriers to quality care identified by key informants in phase 1. However, while SNP plans were significantly more likely to include transportation and nutrition than non-SNP plans in both years studied, the risk differences of inclusion of those benefits for SNP plans declined significantly over time. Specifically, SNP plans were 8% less likely to offer a transportation benefit and 40.5% less likely to offer a nutrition benefit in 2015 than they were in 2014.

It is difficult to draw hard conclusions from these findings. While it is clear that SNP matters in terms of the inclusion of the studied benefits and that the differential continued but narrowed in 2015, it is unclear if the declines in the proportion of SNP plans offering all of the studied benefits other than telemonitoring (premium inclusion was stable) were the result of some SNP plans questioning the value of the benefits or the attractiveness of the benefits for marketing purposes or whether the reductions were merely the result of a reduction in available revenue.

The findings that the marginal effects of county-level ADI, stars bonus caps, and LIS eligible enrollment were nominal are surprising and warrant further research. In addition, the fact that the inclusion of a meals benefit in plan offerings is the only dependent variable that produced results consistent with our hypothesis raises a number of questions about the relationship between county-level ADI and plan benefit design. It is possible that the factors plan sponsors consider in developing plan benefit packages are less related to member-level deprivation than other plan features such as care management model, engagement of community based organizations, network design and provider-plan

collaboration and engagement. It is also possible that plans are not designing products at a county level of refinement. Given the limited number of dependent variables and the short time period of the study (2014 and 2015), further research examining the inclusion or exclusion of these benefits over future years, the inclusion of other beneficiary cost-sharing requirements (co-payments, deductibles) and the inclusion of other benefits that align with SES-related barriers identified by phase 1 key informants might shed further light on these issues.

Section 9.7 Limitations

This study has several limitations. First, all of the data used in the study was obtained from the CMS Web site. The data made publicly available is limited, requiring a series of assumptions to be made in designing the model. Those assumptions are laid out in the methods section. Specifically, because only low-income subsidy eligibility was available in these files, it was used as a proxy for the socioeconomic status of plan participants. Neither the number nor the percentage of plan members who are dually eligible for Medicare and Medicaid, a common proxy measure for SES are made publicly available and, as has been noted elsewhere, little data are available regarding the individual SES attributes of Medicare beneficiaries (Accounting for Socioeconomic Status in Medicare Payment Program, 2015).

Second, this study only looked at the years 2014 and 2015. While these years represent the transition into the full effect of the post-ACA stars methodology, they reflect a narrow window of time calling into question whether they are representative of later years. In addition, because they examine past plan practices they cannot be viewed as prognostic. Repeating the analysis to include additional years of data could assist in analyzing the policy implications of these findings.

Third, while more refined than other possible proxy measures of community SES, including the individual data elements included in the calculation of ADI, the use of ADI as proxy measure for community-level SES is not as exact. In addition, because the ADI is based on 2000 census data, it may not reflect current levels of deprivation. Finally, while ADI is published at more granular levels (nine- digit ZIP codes, ZIP code tabulation area, and U.S. Census block group code), because plan filings occur

at the county level, in order to align the ADI and the plan offerings, county-level ADI was used in this study.

Finally, this study only examined a small subset of the supplemental and enhanced benefits that a plan may offer. In addition, the use of the publicly available plan benefit package filings limited our ability to examine attributes other than benefits such as care management models, network designs and community partnerships that might, based on the phase 1 key informant feedback, impact the quality of care delivered to low SES MA participants.

The limitations of plan benefit package filings also led to the decision to include the presence or absence of a premium rather than other out-of- pocket expenditure requirements such as copayments and deductibles as a dependent variable. The presence or absence of a premium was available as a

dichotomous variable, while other out-of-pocket costs (deductibles and co-payments) appear in the plan benefit package filings individually by benefit and in some cases vary within benefit. For example, many plans vary consumer out-of-pocket costs related to hospitalization based on the length of the

hospitalization. The number of out-of-pocket cost variations made the use of those data beyond the scope of this analysis. However, premiums and cost sharing impact low-income beneficiaries in different ways with premiums generally a forming a barrier to coverage and cost sharing creating a barrier to access (Hudman & O’Malley, 2003). Whether the results found here would differ if the analysis were conducted using cost sharing rather than premiums as a dependent variable may be an important area for future research.

CHAPTER 10: ANALYSIS OF POLICY PROPOSALS IN PHASE 3

In document Goodman_unc_0153D_16730.pdf (Page 144-148)