The 21st Century Cures Act Impact on Medicare Advantage Risk Adjustment: Part III – Adjustments for Dual Enrollees

Aug 24, 2021 | Policy, Risk Adjustment

Introduction

This blog is the third in a series that overviews key changes to the Medicare Advantage (MA) risk adjustment models brought on by the 21st Century Cures Act (CCA). In this blog, we describe the changes made to adjust for the different types of MA enrollees eligible for Medicaid benefits. The first and second installments of this series described changes made to account for the number of conditions each member has and the changes made to account for additional conditions specifically related to mental health disorders, substance abuse disorders, and chronic kidney disease (CKD).

Pre-CCA Adjustments for Dual Enrollees

In Part 1 of this series, we provided an overview of how the MA risk adjustment model works and why it is needed. Below, we review additional details of Medicare eligibility and risk adjustment that provide important context for the CCA changes related to dual enrollees.

There are two types of dual-eligible beneficiaries – those receiving full Medicaid benefits and those receiving partial Medicaid benefits. Full-benefit dual-eligible (full-duals) beneficiaries comprise roughly 70% of all duals. They receive cost sharing assistance and medical benefits from Medicaid not paid for by Medicare (e.g., dental, long-term care). In contrast, partial-benefit duals generally receive assistance only with Medicare premiums and, in some cases, assistance with cost sharing.

The MA risk adjustment model—known as the CMS-HCC model—is used to calculate risk scores to adjust the per member per month payments that MA plans receive for each of their enrollees. Prior to the CCA, the adjustment factors in the CMS-HCC model were historically calibrated using two different risk segments (or subgroups) of the Medicare population; (1) beneficiaries residing in the community (representing well over 90% of all Medicare beneficiaries) and (2) beneficiaries residing in long term care institutional facilities. This allows for separate adjustment factors to be estimated for both populations to reflect the unique cost patterns of each segment.

The types of adjustment factors used to determine the relative risk of each enrollee include a series of demographic- and diagnosis-related variables. Included in the series of demographic variables are indicators for Medicaid enrollment. There are different Medicaid factors by gender, aged/disabled entitlement status, and whether a beneficiary lives in the community or

in an institution. The relative factors for the Medicaid indicators reflect that, on average, dual-eligible beneficiaries cost more than non-dual eligible beneficiaries with otherwise similar

disease and demographic profiles. The Centers for Medicare & Medicaid Services (CMS) reports that approximately 70% of duals have three or more chronic conditions compared to half of Medicare-only enrollees; approximately 30% of duals have six or more, compared to 16% of non-duals. Moreover, having to manage benefits from two different insurance coverage programs can result in duals being vulnerable to fragmented and poorly coordinated health care. For these and other reasons, the costs and resource use involved in managing the health of dual-eligible enrollees are different than non-duals.

However, under the pre-CCA CMS-HCC model specifications, there was no adjustment for the different types of dual-eligible enrollees. Moreover, the relative adjustment for the diagnosis factors in the model were the same for the different types of duals, relative to the non-duals, in both the community and institutional models. Thus, there was the underlying assumption in these pre-CCA models that the additional costs due to having a particular illness or condition (i.e., HCC) would be similar for partial- and full-duals.

Prior research has shown that there may be problems with that underlying assumption. The Medicare Payment and Advisory Commission (MedPAC) found that this approach resulted in systemic underpayments—by about 5 percent—to MA plans for full-duals and overpayments— also by about 5 percent—for partial-duals. A similar analysis by CMS suggests over 9 percent of under- and over-payments, on average. It should be noted that these under- and over-payments were concentrated in the community segment of the CMS-HCC model. This is because the institutional segment is made up of predominantly full-duals (over 80 percent of the institutionalized population). Given the better representation of costs for duals in this population, the under and over payments are well below one percent. Hence, when CMS explored options to improve the MA risk adjustment model for duals, the focus was on the community segment.

CCA Adjustments for Dual Enrollees

CMS considered several different approaches to account for dual-eligibility status in the risk adjustment model for the community segment. These alternatives included applying a multiplicative adjustment factor in the single community segment model to account for the differences in dual-eligibility status; developing models with independently derived risk factors for each of three segments—non-duals, full-duals and partial-duals; and developing models with independently derived risk factors for six segments based on dual-eligibility status and Medicare-eligibility status:

  1. Full-duals eligible for Medicare because of disability
  2. Full-duals eligible for Medicare because of age
  3. Partial-duals and eligible for Medicare because of disability
  4. Partial-duals and eligible for Medicare because of age
  5. No Medicaid benefits and eligible for Medicare because of disability
  6. No Medicaid benefits and eligible for Medicare because of age

CMS considered the six-segment approach because it found that the cost and disease patterns of the non-dual, full-dual, and partial-dual, and the aged versus disabled segments, were distinct, both within the dual types (e.g., full-dual aged versus full-dual disabled) and between the dual types (e.g., full-dual disabled versus partial-dual disabled). Hence, using the six segments could better control for the inherent difference in how these populations need and use health care.

Moreover, even though CMS found that the approach of using three segments and six segments yielded similar improvements in the predictive accuracy of the risk adjustment models for duals on average, the six-segment version out-performed the three-segment version when looking at high- and low-cost enrollees within each segment. Based on these analyses, CMS ended up proposing the six-segment approach, which was eventually finalized after the public-comment period.

Impact to the MA Risk Adjustment Model

As with the other adjustments from the CCA that we have discussed so far, a principle reason for this change is to improve the accuracy of payments being made to MA plans. As noted above, prior to this change, systemic under and over payments for full-dual and partial-dual enrollees, respectively, have been documented. This issue is becoming increasingly significant given the rapid growth of MA products that cater to the dual-eligible population, such as dual-eligible special needs plans (D-SNPs) (see prior blog for more information on D-SNPs). Over the last five years, enrollment in D-SNPs has grown by over 60%, reaching approximately 3 million enrollees.

An evaluation by MedPAC found that making distinctly different adjustments for full-duals and partial-duals eliminates the systematic underpayments for the full-duals and overpayments for the partial-duals that had occurred in the prior versions of the CMS-HCC models that do not distinguish between these populations.

Impact to MA Plans

These adjustments for full-duals and partial-duals will impact MA plans differently depending on how many of each type of dual is enrolled in the plan. MA plans that have a relatively large proportion of full-benefit dual-eligible enrollees will see higher payments and vice versa for plans with partial duals. These changes could be particularly helpful for certain D-SNPs that generally take on a large proportion of full-duals.

Looking Forward

The next blog in this series will address the potential expansion of the amount of diagnosis data available for risk adjustment calculations.