Association of State Policies with Medicaid Disenrollment among Low-Income Medicare Beneficiaries

Abstract

For low-income Medicare beneficiaries, Medicaid provides financial protection against Medicare’s out-of-pocket costs, but many Medicare beneficiaries who qualify for Medicaid are not continuously enrolled. We examined Medicaid disenrollment among Medicare beneficiaries and the relationship between disenrollment and state policies. From 2012–2016, 18.2 percent of Medicare beneficiaries who received full or partial Medicaid disenrolled for reasons other than death. More than 50 percent of Medicare beneficiaries who remained without Medicaid one year after disenrolling continued to receive Low Income Subsidies for Medicare Part D coverage with eligibility requirements similar to those of Medicaid. Among Medicare beneficiaries with continuous Part D subsidies, the rate of Medicaid disenrollment was 24 percent lower in states that automatically enrolled recipients of the federal Supplemental Security Income program in full Medicaid; 33 percent lower in states with more generous provider payment policies; and 37 percent lower in states with less restrictive asset limits for partial Medicaid. Policies that make it easier for individuals to maintain Medicaid eligibility, and that enhance access to care in Medicaid via higher provider reimbursements, may reduce disenrollment.

INTRODUCTION

For low-income Medicare beneficiaries, Medicaid pays for services that Medicare does not (e.g., long-term nursing home and dental care) and subsidizes premiums, deductibles, and other cost sharing in the Medicare program.^1–3 The value of this coverage can amount to several thousand dollars annually – a substantial fraction of income for poor and near-poor individuals – or higher for those with complex health care needs.⁴

Despite Medicaid’s important role in financing care for vulnerable Medicare beneficiaries, many who qualify for Medicaid are not enrolled. Prior analyses estimated that only one-half of eligible Medicare beneficiaries receive full Medicaid, which pays for Medicaid-covered services in addition to Medicare premiums and cost sharing.^5,6 Participation is lower among individuals who qualify for partial Medicaid via the Medicare Savings Programs, which cover Medicare premiums and (in some cases) cost sharing.³

Low rates of Medicaid enrollment may be driven by both low take up of coverage and by disenrollment. Most states require Medicaid beneficiaries to recertify their eligibility annually, which can involve an administratively complex process of documenting income and assets and the termination of Medicaid benefits for those who fail to recertify.^7–9 One study found that 16% of Medicare beneficiaries with full Medicaid and 23% of those with partial Medicaid lost this coverage at least once within a 3-year period, and that over one-half of individuals who lost Medicaid subsequently re-enrolled.¹⁰ Avoiding coverage interruptions could increase overall Medicaid participation and ensure that individuals do not face disruptive gaps in care or financial protection from high medical expenses.^11,12

Because Medicaid is jointly administered by states and the federal government, states have discretion to establish eligibility rules, renewal procedures, and other policies that may affect individuals’ ability and incentives to retain Medicaid.² Prior work has shown that rates of Medicaid enrollment among eligible Medicare beneficiaries vary widely across states.⁵ However, no research has examined the relationship between state policies and Medicaid disenrollment among Medicare beneficiaries, for whom the loss of Medicaid may be particularly disruptive given this population’s financial vulnerability, health status, and age.³

We addressed this gap in the literature by studying the association between state policies and the loss of full and partial Medicaid among Medicare beneficiaries. We focused on Medicare beneficiaries who continuously received Low Income Subsidies for Medicare Part D prescription drug coverage (hereafter referred to as Part D subsidies), which have eligibility criteria closely aligned to Medicaid’s and thus indicate likely Medicaid eligibility.¹³ We analyzed five domains of state policy that theory and evidence suggest might affect Medicaid disenrollment in this population: the automatic enrollment of Supplemental Security Income recipients in Medicaid; the use of pre-populated forms for Medicaid renewals; income and asset limits for full and partial Medicaid; the availability of a Medically Needy program; and payment policies that affect access to care in Medicaid.

METHODS

State Policies

Prior research highlights several areas of state policy that may be salient determinants of Medicaid disenrollment among Medicare beneficiaries. Some of these policies simplify Medicaid enrollment and re-certification, while others pertain less directly to the enrollment process but may affect retention by broadening pathways to Medicaid eligibility or by increasing the value of coverage to enrollees.

Policies affecting Medicaid enrollment and recertification

For Medicare beneficiaries, one salient area of policy variation is the degree to which states coordinate Medicaid eligibility determinations with the federal Supplemental Security Income (SSI) program, which provides cash assistance to aged, blind, and disabled persons with incomes ≤75% of the Federal Poverty Level (FPL) and low assets (≤$2,000 for a single person and $3,000 for a couple).⁷ In 2016, 34 states and the District of Columbia had agreements with the Social Security Administration to automatically enroll SSI recipients into full Medicaid (so-called “Section 1634” states).¹⁴ By reducing the administrative burden of eligibility renewals, auto-enrollment of SSI recipients may reduce Medicaid disenrollment.^7,15

Some states have further simplified the Medicaid renewal process by providing recipients with pre-populated renewal forms.¹⁶ This simplification may be of particular benefit to individuals with cognitive impairments or other functional limitations, for whom complex recertification forms could present a barrier to maintaining coverage.¹⁷ In 2015, 28 states used pre-populated renewal forms for either full or partial Medicaid enrollees.²

Policies affecting Medicaid eligibility and the value of Medicaid coverage

State policies governing the restrictiveness of Medicaid’s income or asset tests may affect disenrollment in several ways. First, less restrictive eligibility criteria increase the number of individuals who qualify for Medicaid, which could reduce disenrollment provided that the income and assets of individuals near higher eligibility thresholds are not more variable than those of individuals near lower thresholds. Second, because asset tests disqualify many individuals from Medicaid and are administratively burdensome,^9,18,19 states that use less restrictive asset tests or eliminate them entirely may limit the number of individuals who lose Medicaid eligibility and deter fewer individuals from renewing their coverage.¹⁶ Third, states with higher income or asset limits for partial Medicaid allow more individuals who lose full eligibility to remain in the program by transitioning to the partial benefit, rather than disenrolling altogether. In 2016, income eligibility limits for full Medicaid were higher than SSI’s in 26 states^2,14; income limits for the Medicare Savings Programs (partial Medicaid) were higher than federal standards in four states; and asset limits for the Medicare Savings Programs were higher than federal standards or waived entirely in 11 states.^2,20

States may also increase Medicaid retention by offering additional pathways to eligibility. In 2016, 33 states offered a Medically Needy program for Medicare beneficiaries, which allows individuals with high health care costs to deduct these costs from income to qualify for Medicaid.^14,21 Prior research has linked the availability of Medically Needy programs to higher rates of Medicaid enrollment among low-income Medicare beneficiaries.¹³ Although individuals receiving Medicaid through a Medically Needy pathway may subsequently lose this coverage if they do not continue to have high medical expenses relative to income, Medically Needy programs nevertheless broaden the pathways by which Medicare beneficiaries can maintain Medicaid enrollment.

Finally, states differ in payment policies that affect access to care in Medicaid, and consequently, individuals’ incentives to maintain Medicaid coverage. Full Medicaid enrollees and individuals in the Qualified Medicare Beneficiary program (the largest of the Medicare Savings Programs) receive subsidies for Medicare cost sharing, which for Part B (physician) services includes 20% coinsurance and an annual deductible ($185 in 2019). In some states, Medicaid reimburses providers based on the difference between the state’s Medicaid fee schedule and Medicare’s payment (net of cost sharing), so that physicians may receive only partial or no payment of Medicare’s cost-sharing amount.²² In these so-called lesser-of states, a provider’s total reimbursement is based on the lesser of the state’s Medicaid fee schedule and Medicare’s payment. In lesser-of states whose Medicaid programs pay <80% of Medicare rates, providers generally will not be reimbursed for any Part B cost sharing, while providers in other states usually will receive some Medicaid reimbursement.^6,23 In 2016, 40 states had a lesser-of policy for Part B services,²² while Medicaid programs in 35 states paid physicians <80% of Medicare rates for prevalent primary care services.²⁴

Low-income Medicare beneficiaries may be less likely to apply for or renew Medicaid coverage in states where providers have fewer incentives to serve dual enrollees due to Medicaid’s nonpayment of Medicare cost-sharing. Prior research has linked lesser-of payment to lower use of physician services,^25,26 consistent with evidence that provider participation in Medicaid is responsive to reimbursement policy.²⁷

Study Population

Our study population consisted of fee-for-service Medicare beneficiaries who received at least one month of full or partial Medicaid during 2012–2016, whom we identified in a random 5% sample of Medicare beneficiaries. We limited our analyses to individuals enrolled in fee-for-service Medicare for at least one year of the study period (while alive for decedents) and in the prior year to assess baseline health conditions. We excluded a small number of individuals who moved states, among whom we saw a notably different pattern of disenrollment (Appendix Exhibit 1²⁸).

Medicaid Enrollment

We assessed the receipt of full vs. partial Medicaid using monthly Medicare-Medicaid dual eligibility codes in the Medicare Master Beneficiary Summary File. We constructed a monthly longitudinal record of beneficiaries’ Medicaid enrollment, starting from the first month of enrollment in the study period and continuing until death, the end of the study period, or the beneficiary’s exit from fee-for-service Medicare (e.g., enrollment in Medicare Advantage). We identified “episodes” of Medicaid enrollment, defined as one or more consecutive months of full or partial Medicaid followed by a coverage gap of one or more months. For beneficiaries with one or more Medicaid enrollment episodes, we analyzed the duration of the index episode (Appendix Exhibit 2²⁷).

Medicaid enrollment data do not indicate whether individuals left Medicaid because they no longer qualified for coverage or disenrolled despite remaining eligible. To assess likely Medicaid eligibility, we used data on beneficiaries’ receipt of Low Income Subsidies that cover premiums and co-payments in the Medicare Part D program. Almost all individuals who qualify for a “full” Part D subsidy (in which individuals face $0 Part D premiums and deductibles as well as reduced co-payments) qualify for full or partial Medicaid, and federal policy stipulates that both full and partial Medicaid recipients should automatically receive full Part D subsidies. Individuals with slightly higher incomes who are ineligible for Medicaid may qualify for a “partial” Part D subsidy, which provides a sliding-scale Part D premium, a partial deductible, and reduced co-payments.^13,29

Individuals receiving a partial Part D subsidy following Medicaid disenrollment are likely to have lost Medicaid eligibility due to changes in income or assets, while those who continued to receive the full Part D subsidy might have lost Medicaid despite remaining eligible. To examine these outcomes, we assessed the proportion of Medicare beneficiaries who received full versus partial Part D subsidies after disenrolling from Medicaid. Of beneficiaries who received a full Part D subsidy without interruption, we examined the relationship between state policies and Medicaid disenrollment.

State-Level Variables

We included the following state policy variables. Unless otherwise noted, we measured these policies annually (Appendix Exhibit 3–4²⁷). First, we included an indicator for whether a state had a Section 1634 agreement with the Social Security Administration to automatically enroll SSI recipients in full Medicaid.^14,30–32 Second, we included an indicator for whether states provided aged, blind, or disabled individuals with pre-populated forms for Medicaid coverage renewals in 2015 (the only year this variable was reported for all states).² Third, we categorized states by their income eligibility limits for full Medicaid (≤75% of the FPL, >75% but <100% of the FPL, or 100% of the FPL) and included a separate indicator for whether states had asset limits higher than the SSI limit used by most states.^2,14,30–32 Fourth, we included two indicators for the restrictiveness of states’ eligibility criteria for the Medicare Savings Programs: whether income limits were higher than federal standards, and whether asset limits were higher than federal standards or waived.^2,20 Fifth, we included an indicator for whether states offered a Medically Needy program for Medicare beneficiaries.^2,14,30–32 Sixth, we categorized states according to a combination of their lesser-of payment policies for Part B cost sharing^22,33,34 and Medicaid payments for primary care services relative to Medicare (≥80% vs. <80% of Medicare rates).^7,24,35,36 The resulting four-level categorical variable captured differences in physicians’ expected reimbursement for Medicare cost sharing for primary care services, ranging from the least generous states (lesser-of payment and Medicaid rates <80% of Medicare) to the most generous states (payment of some or all Part B cost sharing and Medicaid rates ≥80% of Medicare).

Covariates

We included several demographic and clinical characteristics of Medicare beneficiaries. First, we used the Medicare Master Beneficiary Summary File to assess beneficiaries’ age, sex, race and ethnicity, disability, and end-stage renal disease status in the base year of Medicaid enrollment. Second, we used the Medicare Chronic Conditions Data Warehouse to assess the presence of 26 chronic conditions prior to beneficiaries’ index Medicaid enrollment episodes. Third, we constructed a prospective annual Hierarchical Condition Category (HCC) risk score based on the demographic characteristics of beneficiaries and diagnoses from prior-year Medicare claims. Fourth, we used the American Community Survey to assess the following characteristics of ZIP Code Tabulation Areas (among residents age 65 and older) in which beneficiaries resided: educational attainment, the poverty rate, household income, and the proportion of residents living alone. Fifth, we used the Area Health Resources File to assess residence in a metropolitan area and the number of primary care physicians per 1,000 residents in the beneficiary’s county.

Statistical Analyses

We performed two sets of analyses. First, we assessed patterns of Medicaid disenrollment and re-enrollment among Medicare beneficiaries. We estimated the proportion of Medicare beneficiaries who lost Medicaid, the duration of coverage losses, and the proportion of individuals who re-enrolled. We also examined the proportion of Medicare beneficiaries who received full versus partial Part D subsidies following Medicaid disenrollment.

Second, we estimated multivariate Cox proportional hazard regression models to assess the relationship between state policies and Medicare beneficiaries’ risk of disenrollment, adjusting for person- and area-level characteristics. We conducted these analyses among Medicare beneficiaries who continuously received the full Part D subsidy, which we used as a proxy for likely Medicaid eligibility. Beneficiaries who died, exited fee-for-service Medicare, or remained in Medicaid until the end of the study period were analyzed as censored. (See Section 6 of the Appendix for additional details.²⁷)

Our primary analyses examined time to the loss of Medicaid altogether (i.e., ≥1 month without full or partial coverage). In secondary analyses, we separately analyzed the loss of full Medicaid among beneficiaries who initially received full coverage vs. the loss of full or partial Medicaid among beneficiaries who initially received partial coverage. Analyzing the loss of either full or partial Medicaid among partial recipients allowed us to treat cases in which beneficiaries switched from partial to full Medicaid as continuous coverage.

Sensitivity Analysis

In 2013 and 2014, the Affordable Care Act raised Medicaid payments to equal Medicare rates for a subset of primary care services. Because not all services were affected by this “fee bump,” our main analyses categorized states using their base Medicaid rates, net of this fee increase. In a sensitivity analysis, we assumed states had higher Medicaid payment rates in 2013–14 and in subsequent years for a subset of states that extended this policy.²⁴

Limitations

Our analyses had several limitations. First, our observational study design precludes causal inference and is susceptible to bias from unmeasured differences in beneficiary- or state-level factors that may be correlated with the specific policies we analyzed. Second, since we only studied individuals enrolled in the traditional fee-for-service Medicare program, our analyses may not generalize to the Medicare Advantage population, including Special Needs Plans targeting duals. However, because 78% of duals are enrolled in fee-for-service Medicare,³ our analyses are relevant to a majority of Medicare beneficiaries with Medicaid. Third, to the extent states adopted policies in response to Medicaid attrition, estimates of the association between these policies and disenrollment may be biased due to policy endogeneity (i.e., reverse causality).

Finally, we could not definitively determine whether individuals exited Medicaid because they lost eligibility or failed to renew coverage. We used receipt of the full Part D subsidy as a marker of likely Medicaid eligibility and the partial Part D subsidy to identify individuals who lost Medicaid eligibility. However, because a sizeable proportion of individuals who qualify for partial Part D subsidies do not receive them,²⁹ partial subsidy enrollment is likely to undercount the total number of Medicare beneficiaries who became ineligible for Medicaid. Administrative lags across programs may also limit the reliability of full Part D subsidy enrollment as a proxy for continued Medicaid eligibility.

RESULTS

Our study population consisted of 505,336 Medicare beneficiaries, of whom 378,850 initially received full Medicaid and 126,386 initially received partial Medicaid (Exhibit 1). Beneficiaries who initially received full Medicaid were more likely to be non-white, had more chronic conditions, and had higher average HCC risk scores than those initially enrolled in partial Medicaid.

Exhibit 1:

Characteristics of Medicare beneficiaries receiving Medicaid

	Medicare beneficiaries initially receiving:a, b
	Full or partial Medicaid (n=505,336)	Full Medicaid (n=378,950)	Partial Medicaid (n=126,386)
Age in years, mean	65.7	66.1	64.7
Male, %	38.4	37.7	40.3
Race/ethnicity, %
White	60.9	59.1	66.5
Black	18.6	18.0	20.7
Hispanic	12.8	13.9	9.6
Asian	5.3	6.6	1.6
Other	2.3	2.6	1.7
Disabledc, %	50.4	48.9	54.9
Beneficiary has end-stage renal diseased, %	1.5	1.5	1.4
HCC Scoree, mean	1.56	1.66	1.23
Count of Chronic Condition Data Warehouse (CCW) conditionsf
≥ 6 conditions, %	53.5	55.9	46.3
≥ 9 conditions, %	32.6	35.9	22.9

Source: Authors’ analysis of Medicaid enrollment data from Medicare Master Beneficiary Summary Files for the period 2012–16 linked to patient characteristics.

Notes:

^aBeneficiaries were categorized as receiving full or partial Medicaid based on their first month of Medicaid enrollment in the study period. See Appendix Exhibit 2 for details of our study sample.

^bUnless otherwise noted, patient characteristics were assessed for the year in which a beneficiary’s first episode of Medicaid coverage began.

^cDisability was the original reason for Medicare entitlement.

^dEnd-stage renal disease is current reason for Medicare entitlement.

^eHierarchical Condition Categories (HCC) risk scores are derived from demographic and diagnostic data in Medicare enrollment and claims files, with higher risk scores indicating higher predicted spending in the subsequent year. Here, we report HCC scores constructed using diagnoses in Medicare claims from the year preceding beneficiaries’ index Medicaid enrollment episodes.

^fChronic conditions reported for each patient prior to the year in which beneficiaries’ initial episode of Medicaid coverage began, as assessed from the Medicare Chronic Conditions Data Warehouse.

Over the 5-year study period, 18.2% of beneficiaries in our pooled sample lost Medicaid for reasons other than death or exit from fee-for-service Medicare, 18.5% of full Medicaid recipients lost full coverage, and 29.9% of partial Medicaid recipients lost Medicaid altogether (Exhibit 2). Older Medicare beneficiaries and those in poorer health were less likely to lose Medicaid (Appendix Exhibit 6²⁷). Of individuals in the pooled sample who disenrolled, 54.2% lacked Medicaid for a period of 1–6 months and 30.8% had no Medicaid coverage for 7–24 months. Cumulatively over the study period, 60.8% of individuals who lost Medicaid re-enrolled.

Exhibit 2:

Medicaid disenrollment and re-enrollment among Medicare beneficiaries

	Medicare beneficiaries initially receiving:
	Full or partial Medicaid (n=505,336)	Full Medicaid (n=378,950)	Partial Medicaid (n=126,386)
Loss of Medicaid:
Lost full or partial Medicaid, %a	18.2	14.3	29.9
Lost full Medicaid, % of beneficiaries initially receiving full Medicaidb	-	18.5	-
Among Medicare beneficiaries who lost Medicaid: c
Months without any Medicaidd, %
1–6 months	54.2	56.3	51.1
7–12 months	16.1	16.8	15.1
13–24 months	14.7	14.0	15.7
25–59 months	15.0	12.9	18.0
Regained Medicaid, %e	60.8	62.1	59.0

Source: Authors’ analysis of Medicaid enrollment data from Medicare Master Beneficiary Summary Files for the period 2012–16.

Notes:

^aMedicare beneficiaries disenrolling from full or partial Medicaid (i.e., a loss of Medicaid altogether). Cases in which a beneficiary switched from full to partial Medicaid (or vice versa) in adjacent months are considered continuous coverage.

^bPercentage losing full Medicaid. Some individuals who initially had and then lost full Medicaid switched to partial Medicaid the following month (Appendix Exhibit 7). Thus, the percentage of individuals losing full Medicaid is greater than the percentage losing Medicaid altogether.

^cPercentages that follow are of Medicare beneficiaries who lost full or partial Medicaid (i.e., a loss of Medicaid altogether). There were 91,943 Medicare beneficiaries who lost Medicaid altogether, of whom 54,152 initially received full Medicaid and 37,791 initially received partial Medicaid.

^dCensored months due to death or exit from fee-for-service Medicare are not counted towards the number of months of lost Medicaid coverage.

^eCumulative proportion of individuals who resumed full or partial Medicaid following ≥1 month without any Medicaid coverage. Appendix Exhibit 7 reports the proportion of Medicare beneficiaries who regained Medicaid relative to time since disenrollment.

Exhibit 3 plots the proportion of Medicare beneficiaries who received full versus partial Part D subsidies by month after Medicaid disenrollment (person-months in which individuals resumed Medicaid coverage are excluded). Approximately 97% of Medicare beneficiaries received the full Part D subsidy after losing Medicaid and up to 5 months thereafter. Of beneficiaries who remained without Medicaid one year after disenrolling, 52% received the full Part D subsidy while 13% received a partial subsidy. After two years, receipt of the full Part D subsidy among individuals who remained without Medicaid declined to 2%, while receipt of the partial Part D subsidy was approximately 30%.

Exhibit 3: Proportion of Medicare beneficiaries who continued to receive the Part D Low-Income Subsidy after disenrolling from Medicaid.

Source: Source: Authors’ analysis of Part D Low-Income Subsidy data from Medicare Master Beneficiary Summary Files for the period 2012–16.

Notes: Proportion of Medicare beneficiaries who received the full vs. partial Part D Low-Income Subsidy following disenrollment from full or partial Medicaid (i.e., a loss of Medicaid altogether), excluding beneficiaries who moved states, person-months in which a beneficiary subsequently regained Medicaid, and months after a beneficiary died or exited fee-for-service Medicare. On the horizontal axis, month 0 corresponds to the month in which Medicaid coverage was initially lost. The sample size in month 0 is 91,943 Medicare beneficiaries, and is smaller in subsequent months due to re-enrollment in Medicaid and censoring due to death, end of the study period, or beneficiaries’ exit from fee-for-service Medicare.

Exhibit 4 reports estimates of the adjusted association between state policies and Medicaid disenrollment among Medicare beneficiaries who received the full Part D subsidy beginning in their first month of Medicaid coverage and continuing without interruption thereafter. In the pooled sample, the rate of losing any Medicaid was 24% lower in Section 1634 states that automatically enrolled SSI recipients in full Medicaid (Hazard Ratio [HR]: 0.76, 95% CI: 0.59 to 0.97), 33% lower in states whose Medicaid programs reimbursed providers for some or all Medicare cost-sharing and paid ≥80% of Medicare rates for primary care services (HR: 0.67, 95% CI: 0.46 to 0.99), and 37% lower in states whose asset limits for partial Medicaid were higher than federal standards or waived altogether (HR: 0.63, 95% CI: 0.49 to 0.82).

Exhibit 4:

Adjusted association between state policies and Medicaid disenrollment^a

	Population:	Medicare beneficiaries initially receiving full or partial Medicaidb		Medicare beneficiaries initially receiving full Medicaidc		Medicare beneficiaries initially receiving partial Medicaidd
State policy:f	Outcome:	Loss of full or partial Medicaide		Loss of full Medicaid		Loss of full or partial Medicaide
		Adjusted Hazard Ratiog	(95% CIh)	Adjusted Hazard Ratiog	(95% CIh)	Adjusted Hazard Ratiog	(95% CIh)
State automatically enrolls SSI recipients in full Medicaid (§1634 state)		0.76	(0.59, 0.97)	0.66	(0.53, 0.83)	0.92	(0.68, 1.24)
State uses pre-populated forms for Medicaid coverage renewals		1.30	(0.99, 1.69)	1.26	(0.99, 1.62)	1.26	(0.90, 1.76)
Income Limit for full Medicaid:
≤ 75% of FPL (ref.)		-		-		-
76–99% of FPL		0.99	(0.70, 1.42)	1.02	(0.71, 1.47)	1.09	(0.71, 1.68)
100% of FPL		0.78	(0.59, 1.02)	0.88	(0.69, 1.13)	1.18	(0.84, 1.66)
Asset limit for full Medicaid: higher than SSI limit or waived		0.97	(0.73, 1.29)	0.95	(0.69, 1.30)	0.66	(0.43, 1.02)
Income limit for partial Medicaid: higher than federal standard		1.09	(0.74, 1.62)	-i		0.78	(0.46, 1.32)
Asset limit for partial Medicaid: higher than federal standard or waived		0.63	(0.49, 0.82)	-i		0.60	(0.42, 0.87)
Medically Needy program for Medicare beneficiaries		0.81	(0.52, 1.25)	0.83	(0.57, 1.21)	0.97	(0.60, 1.56)
Provider payment policy:
Lesser of payment of Part B cost-sharingj and a primary care fee ratio <80%k (ref.)				-		-
Partial or full payment of Part B cost-sharingj and a primary care fee ratio <80%k		1.41	(1.14, 1.75)	1.25	(0.95, 1.64)	1.31	(0.98, 1.74)
Lesser of payment of Part B cost-sharingj and a primary care fee ratio ≥80%k		0.94	(0.66, 1.34)	0.99	(0.69, 1.43)	0.82	(0.58, 1.16)
Partial or full payment of Part B cost-sharingj and a primary care fee ratio ≥80%k		0.67	(0.46, 0.99)	0.57	(0.40, 0.81)	0.63	(0.38, 1.05)

Source: Authors’ analysis of Medicaid coverage variables in the Medicare Master Beneficiary Summary files for the period 2012–16 linked to state policy, patient, and area-level variables.

Notes:

^aEstimated from Cox proportional hazard regression models adjusting for age, sex, disability, presence of end-stage renal disease, indicators of 26 chronic conditions from the Medicare Chronic Condition Data Warehouse, an annual patient-level Hierarchical Condition Category risk score, and ZCTA- and county-level characteristics. Each column contains estimates from a separate regression model. See Appendix Exhibit 9 for full model estimates. Analyses exclude Medicare beneficiaries who relocated states or lacked continuous receipt of the full Part D Low Income Subsidy during the study period. See Appendix Exhibit 2 for details of our analysis sample.

^bEstimates for time to the loss of full or partial Medicaid in our full sample (466,103 Medicare beneficiaries initially receiving full or partial Medicaid during the study period).

^cEstimates for time to the loss of full Medicaid. The sample size is 356,991 Medicare beneficiaries initially receiving full Medicaid during the study period.

^dEstimates for time to the loss of full or partial Medicaid. The sample size is 109,112 Medicare beneficiaries initially receiving partial Medicaid during the study period.

^eThat is, ≥1 month without any Medicaid coverage.

^fSee Appendix Exhibit 3 and 4 for detailed state policies and our data sources.

^gHazard ratios less than 1 indicate a lower rate of Medicaid disenrollment.

^h95% Confidence Intervals were constructed using standard errors clustered at the state level.

ⁱOnly state policy variables that pertained primarily to full Medicaid were included in this regression model.

^jWe assessed state Medicaid programs’ policies for reimbursing providers for Medicare Part B deductibles and coinsurance. In lesser-of states, Medicaid reimburses providers for Part B deductibles and co-insurance only up to the lesser of the state’s Medicaid fee schedule or the Medicare paid amount.

^kThat is, a state’s Medicaid fee schedule for primary care services was <80% vs ≥80% of Medicare’s for primary care services, net of temporary fee increases provided under the Affordable Care Act’s primary care fee “bump.” Model estimates inclusive of this fee increase are shown in Appendix Exhibit 10.

Compared to estimates in the pooled sample, the rate of disenrollment from full Medicaid was lower in states that automatically enrolled SSI recipients in full Medicaid and that had more generous provider payment policies. Among partial Medicaid recipients, the rate of Medicaid disenrollment was 40% lower (HR: 0.60, 95% CI: 0.42 to 0.87) in states with less restrictive asset limits for partial Medicaid compared to federal standards.

When we analyzed states’ Medicaid payment rates inclusive of Affordable Care Act fee increases, we found a significantly lower rate of disenrollment in states whose Medicaid programs paid ≥80% of Medicare rates for primary care services (Appendix Exhibit 10²⁷), supporting findings from our main analysis that the generosity of Medicaid provider reimbursements is a salient determinant of disenrollment.

DISCUSSION

In a national sample of Medicare beneficiaries who received full or partial Medicaid during the period 2012–2016, we found that 18.2% of beneficiaries disenrolled from Medicaid for reasons other than death, and that 60% of those who disenrolled regained Medicaid at some point later in our study period. These patterns of disenrollment and re-enrollment – termed “churning” – were less frequent than has been reported among non-elderly and non-disabled adults (who do not receive Medicare)³⁷, but were notable for two reasons. First, one year after disenrolling, more than one-half of Medicare beneficiaries who remained without Medicaid continued to receive “full” Low Income Subsidies for Medicare Part D coverage. Virtually all Medicare beneficiaries receiving these subsidies are eligible for Medicaid due to these programs’ closely aligned eligibility criteria. Thus, our study provides evidence of misalignment between Medicaid and Part D subsidies, despite federal policies to harmonize eligibility standards and coordinate application processing for these programs.³⁸ Second, the Medicaid disenrollment rates we found were very similar to those reported fifteen years ago,⁷ highlighting a continued challenge of reducing gaps in Medicaid coverage in the Medicare population.

We identified several state policies that were associated with lower rates of Medicaid disenrollment among Medicare beneficiaries. First, states that automatically enrolled SSI recipients in full Medicaid (Section 1634 states) retained higher proportions of Medicaid recipients for a longer time. This finding suggests the potential benefits of improving coordination between separately administered safety-net programs that serve similar populations (SSI and the Part D Low Income Subsidies are administered by the Social Security Administration, while states administer Medicaid). In recent guidance to state Medicaid directors, CMS identified improved coordination between Medicaid, SSI, and the Part D Low Income Subsidy as an opportunity for states to reduce administrative barriers to program participation and to better manage services for vulnerable Medicare beneficiaries.³⁹

Second, states with high or no asset limits for the Medicare Savings Programs (partial Medicaid) have lower rates of disenrollment. This finding is consistent with evidence that some Medicare beneficiaries do not receive partial Medicaid because they do not meet asset tests¹⁸ or find these tests overly complex.^9,18,19 Policies that reduce the stringency of asset tests and the burden of demonstrating eligibility could reduce disenrollment.

Third, states whose Medicaid programs pay Medicare cost-sharing amounts for Part B services and reimburse providers at least 80% of Medicare rates for primary care services exhibit lower rates of disenrollment, particularly among individuals with full Medicaid. This finding is consistent with evidence that enrollment in Medicaid is responsive to the scope and accessibility of care in the program, which is driven in part by providers’ economic incentives to serve Medicaid recipients.^6,23,27

While some policy changes we identified (increasing Medicaid’s payment of Medicare cost-sharing) would require additional state expenditures, others are administrative and would cause states to incur additional costs only when individuals retain Medicaid for longer periods. For example, 16 states currently do not auto-enroll SSI recipients into full Medicaid, including 8 that use more restrictive income or asset criteria than the SSI standard (Section 209(b) states). Since 2014, two states – Ohio and Indiana – switched from 209(b) to 1634 status, while Indiana concurrently raised income limits for full and partial Medicaid.^14,30 Research investigating the impact of these changes on Medicaid disenrollment will be instrumental for guiding future state reforms.

CONCLUSION

Over a 5-year period, 18.2% of Medicare beneficiaries receiving full or partial Medicaid were disenrolled from the program despite frequently continuing to receive full Part D subsidies whose income and asset eligibility criteria align closely with Medicaid’s. State policies that reduce administrative barriers to re-enrollment, make it easier for individuals to maintain Medicaid eligibility, and offer higher provider reimbursements that enhance access to care in Medicaid, may enhance low-income Medicare beneficiaries’ ability and incentives to maintain Medicaid benefits for which they are eligible.