Cognition and Take-up of the Medicare Savings Programs

Abstract

This cross-sectional study examines the association between cognition and Medicare Savings Program enrollment among qualified elderly Medicare beneficiaries.

The Medicare Savings Programs (MSPs) are limited Medicaid benefits that reduce out-of-pocket costs in Medicare Parts A and B for individuals with low income and assets. Despite the substantial financial assistance the MSPs offer, fewer than one-half of eligible individuals are enrolled.¹

The complexity of the application and reattestation process for the MSPs, which requires individuals to provide detailed documentation of their income and assets at least annually, has been identified as one explanation for the low take-up of these benefits.² Individuals with cognitive impairments, such as dementia, which 6.5 million elderly Americans experience, may face particular challenges navigating the MSP benefit. Prior research identified poor cognition as a contributor to the low take-up of insurance benefits other than the MSPs.^3,4

In this study, we examined the association between cognition and MSP enrollment among elderly Medicare beneficiaries who qualify for these programs. We also examined enrollment in the Low-Income Subsidy (LIS), a separate program that provides premium and cost-sharing assistance in Medicare Part D that Medicare beneficiaries automatically receive if they are enrolled in an MSP.⁵

Methods

We analyzed Medicare beneficiaries 65 years and older from biennial 2008 to 2016 waves of the Health and Retirement Study (HRS; https://hrs.isr.umich.edu/sites/default/files/rda-forms/MedRIC-Documentation-for-HRS-Data-Requestors.pdf) who were linked to contemporaneous indicators of MSP and LIS enrollment from Medicare enrollment files. The University of Pittsburgh institutional review board approved this study, which was an analysis of secondary survey data. Survey respondents provided informed consent when the survey was conducted. We limited our sample to financial respondents (ie, individuals most knowledgeable about a household’s finances) who qualified for the MSPs based on their income and assets, excluding individuals receiving Veterans Health Administration benefits and those eligible for or receiving full Medicaid (eAppendix in the Supplement).

We used responses to a validated 27-point scale that assessed attention, memory, and basic mathematical reasoning to assign financial respondents to 1 of 3 cognitive status categories: likely dementia (scores from 0-6), mild cognitive impairment without dementia (scores 7-11), and normal cognition (scores 12-27).⁶ Missing observations were imputed using responses from the prior HRS survey wave if available.

We estimated logistic regression models to assess differences in MSP and LIS enrollment among financial respondents associated with mild cognitive impairment or likely dementia (vs normal cognition), adjusting for respondents’ socioeconomic, demographic, and family characteristics (marital status and presence of living children); education; tobacco use; chronic conditions, depression, and disability status; Medicare Advantage enrollment; and indicators for state and survey wave. To facilitate interpretation, we converted estimates into adjusted enrollment probabilities for each cognitive status category, holding the other covariates constant. We used HRS survey weights to make our estimates representative of the community-dwelling Medicare population. Analyses were conducted using Stata, version 16 (StataCorp), and statistical significance was set at P < .05.

Results

The study sample consisted of 3263 MSP-eligible respondent-years (Table). In adjusted analyses, we estimated that 29.7% of respondents who likely had dementia were enrolled in 1 of the MSPs vs 39.5% of those with normal cognition (P = .006; Figure). Low-Income Subsidy enrollment was also lower among respondents with dementia (30.0%) vs normal cognition (38.7%; P = .02). We did not detect significant differences in MSP or LIS enrollment between respondents with mild cognitive impairments vs normal cognition.

Table. Characteristics of MSP–Eligible Respondents in the HRS by MSP Enrollment Status.

Respondent characteristica	MSP, %		P valued
	Enrolled (n = 1210 respondent-years)b	Not enrolled (n = 2053 respondent-years)c
Cognitive function (% in category)e
Cognitively intact	54.0	57.4	.01
Cognitively impaired without dementia	35.6	29.5
Dementia	10.4	13.1
Age, y	75.4	76.8	<.001
Female	70.0	69.8	.93
Race/ethnicity (% in category)
White	61.9	69.8	.002
Black	19.5	17.8
Hispanic	16.2	11.0
Asian	1.0	0.5
Other	1.4	0.9
Education, y	10.3	11.1	<.001
Married	15.4	20.6	.004
Any living childrenf	86.2	90.1	.03
Ever smoked tobacco	64.4	54.1	<.001
Disabledg	27.6	13.9	<.001
Difficulty with ≥2 activities of daily living	17.6	14.6	.06
Count of CCW chronic conditions, meanh	5.1	4.7	.01
Depressedi	27.2	17.6	<.001
Household, mean, $
Incomej	12 763	13 637	<.001
Net assetsk	−1102	8915	<.001
Enrolled in a Medicare Advantage Plan	45.9	41.3	.04

Abbreviations: CCW, Chronic Conditions Data Warehouse; HRS, Health and Retirement Study; MSP, Medicare Savings Plan.

^aAnalyses conducted among financial respondents 65 years or older in HRS (https://hrs.isr.umich.edu/sites/default/files/rda-forms/MedRIC-Documentation-for-HRS-Data-Requestors.pdf) households that met eligibility criteria for the MSPs based on income (net of disregards) and countable assets. We excluded individuals receiving Veterans Administration benefits and those eligible for or receiving full Medicaid. See the eAppendix in the Supplement for how we identified MSP enrollees.

^bWhen weighted, this sample represented 4 538 777 person-years in the community-dwelling Medicare population.

^cWhen weighted, this sample represented 7 133 750 person-years in the community-dwelling Medicare population.

^dP value for the difference in means or proportions, comparing MSP enrollees with eligible nonenrollees.

^eWe used responses to a validated 27-point scale to categorize individuals into 1 of 3 cognitive status categories: likely dementia (scores 0-6), mild cognitive impairment without dementia (scores 7-11), and normal cognition (scores 12-27). We imputed missing observations using responses from the prior HRS survey wave, if available.

^fProportion of respondents or spouses (if present) with at least 1 living child.

^gDisability was original reason for Medicare entitlement.

^hAssessed from the Medicare CCW, which draws from claims since 1999 to capture beneficiaries’ cumulative burden of chronic disease. In regression analyses, we adjusted for the count of CCW conditions reported on claims before the HRS survey year and for indicators denoting the presence of 6 or more or 9 or more CCW conditions.

ⁱBased on responses to the Center for Epidemiologic Studies depression questionnaire administered in the HRS.

^jWe reported the income net of the following annual disregards used to determine MSP eligibility: $240 in income from any source, $780 in earned income, and 50% of earned income exceeding $780. We applied additional state-specific disregards for the MSPs in Illinois ($60 per year) and Mississippi ($360 per year).

^kHousehold assets (eg, balances in checking, retirement, and savings accounts, stocks, bonds, trusts, and the value of properties not used for residence) are the reported net of unsecured debts (eg, credit card debt) and a per-person burial allowance of $1500. We did not include the value of assets that are not counted for purposes of determining MSP eligibility (principally, the value of a home used as one’s residence and vehicles).

Figure. Adjusted Associations Between Cognitive Status and Enrollment in the Medicare Savings Programs and the Part D Low-Income Subsidy.

Authors’ analyses of Medicare beneficiaries in the biennial 2008 to 2016 waves of the Health and Retirement Study (HRS; https://hrs.isr.umich.edu/sites/default/files/rda-forms/MedRIC-Documentation-for-HRS-Data-Requestors.pdf) linked to contemporaneous indicators of enrollment in the Medicare Savings Programs (MSPs) and Part D Low-Income Subsidy (LIS) from Medicare enrollment files (eAppendix in the Supplement). Adjusted rates of MSP and LIS enrollment were calculated from survey-weighted logistic regression models that predicted program-specific enrollment as a function of cognitive status (likely dementia and mild cognitive impairment without dementia, omitting normal cognition as the reference group). Estimates were adjusted for respondents’ age, sex, race and ethnicity, education, marital status, presence of any living children, tobacco use, disability, difficulty performing 2 or more activities of daily living (eg, toileting or dressing), chronic conditions reported on Medicare claims before the survey year (assessed from the Chronic Condition Data Warehouse), depression status assessed from the Center for Epidemiological Studies depression questionnaire, household income and assets, Medicare Advantage enrollment, and indicators for state and the survey wave. We used HRS survey weights to make the estimates representative of the community-dwelling Medicare population. We assessed the statistical significance of differences in enrollment between individuals with mild cognitive impairment or dementia (vs normal cognition) using standard errors that accounted for within-person clustering.

^aStatistically different from enrollment among individuals with normal cognition at P = .006.

^bStatistically different from enrollment among individuals with normal cognition at P = .02.

Discussion

Medicare beneficiaries who qualify for MSPs and the LIS are less likely to receive these benefits if they have dementia. Our findings are consistent with evidence that individuals with severe cognitive impairments are prone to making suboptimal insurance decisions or foregoing potentially valuable insurance coverage altogether.³ One limitation of this observational study is that unmeasured beneficiary characteristics that are associated with cognitive status and program take-up could have biased our estimates. Our results also complement those of a prior study that linked cognitive status to beneficiaries’ awareness of and likelihood of applying for the LIS, which analyzed individuals who did not automatically receive the LIS because they did not have full Medicaid or MSPs.⁴

Simplifying or automating the process by which Medicare beneficiaries apply to and recertify eligibility for the MSPs could reduce barriers to the take-up and retention of these benefits, particularly among individuals with cognitive impairment. For example, states could simplify asset tests or autoenroll recipients of other means-tested assistance (eg, food stamps or home heating assistance) in MSPs. Given how enrollment in MSPs and the LIS is linked, these reforms would also likely increase LIS participation among low-income Medicare beneficiaries with dementia.

Supplement.

eAppendix.