The Effects of Housing Cost Burden and Housing Tenure on Moves to a Nursing Home Among Low- and Moderate-Income Older Adults

Abstract

Background and Objectives

In the United States, a growing number of older adults struggle to find affordable housing that can adapt to their changing needs. Research suggests that access to affordable housing is a significant barrier to reducing unnecessary nursing home admissions. This is the first empirical study we know of to examine whether housing cost burden (HCB) is associated with moves to nursing homes among older adults.

Research Design and Methods

Data include low- and moderate-income community-dwelling older adults (N = 3,403) from the nationally representative 2015 National Health and Aging Trends Study. HCB (≥30% of income spent on mortgage/rent) and housing tenure (owner/renter) are combined to create a 4-category housing typology. Multinomial logistic regression models test (a) if renters with HCB are most likely (compared with other housing types) to move to a nursing home over 3 years (2015–2018) and (b) if housing type interacts with health and functioning to predict moves to a nursing home.

Results

Across all models, renters with HCB had the greatest likelihood of moving to a nursing home. Moreover, self-rated health, physical capacity, and mental health were weaker predictors of nursing home moves for renters with HCB.

Discussion and Implications

Results suggest that older renters with HCB are most likely to experience unnecessary nursing home placement. The growing population of older renters experiencing HCB may not only signal a housing crisis, but may also challenge national efforts to shift long-term care away from nursing homes and toward community-based alternatives.

In the United States, population aging is coinciding with a growing affordable housing crisis. Older Americans are spending more of their income on housing (Joint Center for Housing Studies, 2019) and many struggle to find and maintain affordable, stable housing that can adapt to their changing needs (Pearson et al., 2019; Stone, 2013). Housing assistance is not an entitlement and the availability of publicly subsidized housing falls substantially short of demand. Only 33% of eligible low-income older households receive federal housing assistance (HUD, 2017) and many older adults remain on waitlists for months or years (Vandawalker, Locke, & Lam, 2012). Older adults waiting for housing assistance experience housing instability, food insecurity, and poor health outcomes (Carder, Kohon, Limburg, & Becker, 2018). Evidence also suggests that homelessness among older adults is on the rise (HUD, 2018). In the last 10 years, the number of older adults living in homeless shelters or transitional housing increased by 69%, from roughly 45,000 to nearly 76,000 people, and this trend is expected to continue (HUD, 2018).

Although homeownership rates among older adults are generally high compared with younger populations, with an estimated 79% of community-dwelling older adults owning their home (United States Census Bureau, 2018), evidence suggests that the proportion of older homeowners, compared with older renters, will continue to decline as younger cohorts enter older adulthood (Joint Center for Housing Studies, 2019). This trend threatens financial security in later life because home equity is the largest source of savings for most older Americans (Sass, 2017). Rising rents also threaten the financial well-being of older renters because renters (54%) compared to homeowners (26%) are more likely to experience housing cost burden (HCB), defined as spending 30% or more of monthly income on housing-related expenses (Joint Center for Housing Studies, 2019). Older adults experiencing HCB often face trade-offs between covering housing costs or paying for food, medications, and other health-related expenses (Alley et al., 2011; Carder, Luhr, & Kohon, 2016; Carder et al., 2018). Among older adults, HCB increases with age, with low-income renters (72%) being most at risk of experiencing HCB in later life (Joint Center for Housing Studies, 2019).

Intersection of Affordable Housing and Long-Term Care

Approximately 70% of older adults will need long-term care (LTC) services in their lifetime (Bipartisan Policy Center, 2016) and most prefer to receive care at home, or in assisted living, rather than a nursing home (Kasper, Wolff, & Skehan, 2018). Medicaid is the nation’s primary payer for LTC (Harris-Kojetin et al., 2019) and federal Medicaid funding can generally only cover room and board expenses in nursing homes, not in community-based alternatives (Rosenbaum, 2016). Given the current LTC context in the United States, homeowners are better positioned than renters to afford preferred LTC options, although often with significant out-of-pocket costs. Indeed, evidence suggests that homeowners, compared with renters, are more likely to age in place and avoid moving to a nursing home (McCann, Grundy, & O’Reilly, 2012; Rouwendal & Thomese, 2013). However, it is not known whether differences in housing affordability between renters and homeowners partially drive this association.

Coinciding with the growing affordable housing crisis, the U.S. formal LTC system continues to undergo a major transformation, characterized by a shift away from nursing home care and toward community-based alternatives, a process known as rebalancing LTC (Degenholtz, Park, Kang, & Nadash, 2016). Some scholars have argued that the use of nursing homes among Medicaid recipients with disabilities is partially driven by their need for affordable housing (Crossley, 2018; Rosenbaum, 2016). In support of this claim, formal evaluations of federal programs designed to reduce unnecessary nursing home use among Medicaid beneficiaries have often cited lack of affordable and accessible housing as a significant barrier to success (Bardo, Applebaum, Kunkel, & Carpio, 2013; Hoffman, Kehn, & Lipson, 2017; Lipson, Stone Valenzano, & Williams, 2011).

Although gaps in affordable housing may challenge national efforts to rebalance LTC, surprisingly few empirical studies have examined the association between HCB and nursing home use among older adults (Anguelov & Frank, 2020; Burr & Mutchler, 2007; Mutchler & Burr, 2003). An earlier cross-sectional study using census data from 1990 suggests that older adults living in counties with low housing availability and unaffordable rents are more likely to live in a nursing home than similar older adults living in counties with greater access to affordable housing (Mutchler & Burr, 2003); however, other work suggests that community-level HCB is not associated with higher nursing home use (Anguelov & Frank, 2020; Burr & Mutchler, 2007). We know of no studies examining the effect of individual-level HCB on subsequent moves to a nursing home. This study intends to address this gap by examining whether HCB and housing tenure (owner/renter) are associated with moving to a nursing home among low- and moderate-income older adults.

Conceptual Framework

We draw from the person–environment (P–E) fit perspective (Lawton & Nahemow, 1973) and the Wiseman Behavioral Model of Elderly Migration (Wiseman, 1980) to help inform our understanding of what it means to “need” LTC services in a nursing home. According to the P–E fit perspective (Lawton & Nahemow, 1973), an individual’s capacity (e.g., health and functioning) interacts with environmental demands to influence adaptive behavior. Therefore, individuals can reach their full potential when their individual capacities are aligned with environmental demands and opportunities. Building from the P–E fit perspective, the Wiseman model (1980) suggests that environmental incongruence, characterized by a poor fit between individual and environmetal resources, acts as a triggering mechanism that can influence residential relocation. As such, it is likely that older renters with HCB, compared with homeowners and renters without HCB, are more at risk of experiencing environmental incongruence because they often have less control over their home environment and have fewer financial resources to reduce environmental press and compensate for declines in health and functioning (Spillman, Biess, & MacDonald, 2012). The Wiseman (1980) model also delineates push factors (e.g., environmental stress, loss of independence) and pull factors (e.g., environmental amenities) that influence relocation. As such, it is possible that older adults are “pushed” out of community housing by a combination of declines in health and functioning and experiencing HCB, while also being “pulled” toward nursing homes as both an LTC and affordable housing option. Therefore, we would expect that the association between health and functioning and nursing home use would be weaker for older adults with HCB, compared with those without HCB, because HCB operates as a push factor and nursing homes work as a pull factor.

The Current Study

We use a nationally representative sample of community-dwelling Medicare beneficiaries from the 2015 NHATS to examine how housing tenure (owner/renter) and HCB are associated with moves out of community residence and to a nursing home over 3 years (2018) among low- and moderate-income older adults. This is the first study we know of to examine individual-level HCB as a potential push factor that contributes to moves out of community housing and to a nursing home. Drawing from our conceptual framework, we test the following hypotheses:

Hypothesis 1: Among low- and moderate-income older adults living in the community, renters compared with homeowners will be more likely to move to a nursing home, with renters with HCB most likely to move.

Hypothesis 2: Differences in health and functioning [dementia, self-rated health, mental health, physical capacity, activities of daily living (ADL) limitation, instrumental activities of daily living (IADL) limitation] and other economic factors (Medicaid, annual income) will not fully account for the greater likelihood of moving to a nursing home among renters with HCB.

Hypothesis 3: The association between health and functioning and moves to a nursing home will be moderated by community housing type such that the association is strongest for homeowners compared with renters and weakest for renters with HCB.

Research Design and Methods

Data

We limit our sample to low- and moderate-income older community-dwelling Medicare beneficiaries from the nationally representative 2015 NHATS (N = 3,403), and examine predictors of moves out of community residence and to a nursing home over 3 years (2015–2018). NHATS data were first collected in 2011, with a baseline response rate of 71% and 7% of interviews completed by a participant serving as a proxy (Kasper & Freedman, 2014). NHATS data are collected annually and in 2015 the NHATS sample was replenished to adjust for those no longer in the study due to death or attrition. Detailed information on the NHATS study design and methods can be referenced elsewhere (DeMatteis, Freedman, & Kasper, 2016b).

We first limit our analytic sample to homeowners or renters living in the community in 2015, which excludes older adults living in Continuing Care Retirement Communities, assisted living, nursing homes, or other arrangements (e.g., living with family) (n = 6,535). We limit our sample to low- and moderate-income older adults, defined as those in the bottom 75% of the income distribution, with annual income of less than $65,000 per year (n = 4,993). Only 2% of older adults in the top 25% of the income distribution experience HCB (n = 28) and they are less likely to face the same financial trade-offs due to high housing costs compared with low- and moderate-income older adults. Seventy-one percent (n = 3,537) remained in our sample after attrition between baseline and follow-up 3 years later. For participants with missing baseline data on covariates of interest, we use data from the subsequent round (n = 314) when possible. Four percent of remaining cases were listwise deleted due to missing data on covariates (n = 134). Bivariate analyses comparing participants lost to attrition (n = 1,258) with those remaining in our final analytic sample (n = 3,403) show that our sample overrepresents younger, white, higher-income older adults in better health and underrepresents renters with HCB.

Measures

Nursing Home Admission

The outcome of interest is nursing home admission. Over 3 years (2015–2018), participants either remained in the community (referent), moved to a nursing home, or died. Though not the focus of our analyses, we include death as a possible outcome in the analyses and Supplementary Materials. Participants who moved to a nursing home and then died between baseline and follow-up were categorized as having moved to a nursing home (n = 16) and those who moved to assisted living or another LTC arrangement were categorized as remaining in the community. An advantage of the NHATS study over other national studies (e.g., Health and Retirement Survey) is that facility staff and participants/proxies are interviewed to distinguish between assisted living and nursing home settings (Freedman & Spillman, 2014) and between long-term and short-term rehabilitation nursing home stays (Montaquila, Freedman, Spillman, & Kasper, 2012). As such, we are able to well-categorize long-term moves to a nursing home.

Community Housing Type

We consider both housing tenure (homeowner vs renter) and experiencing HCB in 2015 to create the following typology as the baseline community housing measure: owners without HCB (referent), owners with HCB, renters without HCB, and renters with HCB. We use the publicly available imputed total income measure (DeMatteis, Freedman, & Kasper, 2016a) and self-reported monthly mortgage or rent to calculate HCB, defined as paying more than 30% of income on housing (Joint Center for Housing Studies, 2019).

Background Characteristics

Self-reported gender (female = 1; male = 0), living arrangement (lives alone = 1; lives with others = 0), race/ethnicity (person of color = 1; non-Hispanic white = 0), and age were included as controls. Age was categorized ordinally from 1 to 6 to protect the identity of participants in the publicly available NHATS data and was included in the analysis as a continuous measure (1 = 65–69 years old; 6 = 90+ years old).

Health and Functioning

We include a variety of health and functioning measures: self-rated health as a continuous measure (1 = poor; 5 = excellent), physical capacity [0–12 scale (α = 0.88), with higher scores reflecting greater self-reported physical capacity on activities such as walking, lifting, opening jars] (Freedman et al., 2011), and mental health by examining depression and anxiety symptoms using the Patient Health Questionnaire-4 (PHQ-4). The PHQ-4 assesses how often (0 = not at all, 1 = several days, 2 = more than half the days, 3 = nearly every day) a participant felt on edge, experienced uncontrolled worrying, felt depressed, or had little interest in doing things, with scores greater than two coded as having a mental health concern (Kroenke, Spitzer, Williams, & Löwe, 2009). We also included a dichotomous measure of dementia status (1 = probable dementia). A participant was coded as having probable dementia if they met one of the following critieria: (1) participant or proxy reported a dementia diagnosis, (2) a score indicating probable dementia on the AD8 Dementia Screening Interview, or (3) a score at least 1.5 SDs below the mean on at least two domains of cognitive tests (executive functioning, memory, and orientation) (Kasper, Freedman, & Spillman, 2013).

We also adjust for self-reported difficulty independently completing ADL (any ADL limitation = 1) or IADL tasks (any IADL limitation = 1) in the last month. ADL tasks included eating, bathing, toileting, dressing, getting around inside the home, and getting out of bed. IADL tasks included getting outside the home, doing laundry, shopping, preparing meals, handling bills/banking, and managing medications. Participants who received help (for health or functioning reasons) and never completed the task alone in the last month were also coded as having a limitation.

Other Economic Factors

We adjusted for self-reported Medicaid receipt (Medicaid = 1). We used the imputed total annual income variable available in the NHATS (DeMatteis et al., 2016a) based on the participant’s estimate of individual income or income for co-residing couples from multiple sources, including earned income, social security, supplemental security income, pensions, retirement account withdrawals, and income earned from interest/dividends (Kasper & Freedman, 2018).

Analysis

We conducted bivariate analyses to examine differences across community housing types (Table 1) and to test whether community housing type was associated with nursing home admission over 3 years (Table 2). As seen in Table 2, only three homeowners with HCB moved from the community to a nursing home over 3 years and therefore homeowners with HCB (n = 298) were excluded from subsequent multivariate analyses. To test our second and third hypotheses, we used multinomial logistic regression to estimate relative risk (RR) ratios for the likelihood of moving to a nursing home, compared with remaining in the community (referent) over 3 years (Table 3). All independent variables were measured at baseline (2015) predicting moves to a nursing home (or death) over 3 years (2018). Results of multinomial regression models with death included as an outcome can be found in Supplementary Table 1.

Table 1.

Weighted Characteristics of Low- and Moderate-Income Community-Dwelling Older Adults in 2015 by Housing Type (N = 3,403)

	Homeowners		Renters		p
	No HCB (n = 2,302)	HCB (n = 298)	No HCB (n = 344)	HCB (n = 459)
Female (%)	59.3	65.9	53.5	66.6	.01
Age (1–6 scale)	2.7	2.0	2.5	2.7	<.001
Person of color (%)	14.4	32.3	32.7	43.6	<.001
Lives alone (%)	33.0	27.6	62.5	50.6	<.001
Probable dementia (%)	5.6	4.3	8.5	13.1	<.001
Self-rated health (1–5 scale)	3.4	3.2	3.0	2.9	<.001
Mental health concern PHQ-4 (%)	25.4	27.5	35.4	41.1	<.001
Physical capacity (0–12 scale)	9.5	9.5	8.3	8.3	<.001
ADL limitation (%)	29.5	28.1	45.7	39.7	<.001
IADL limitation (%)	30.2	34.3	49.3	44.8	<.001
Medicaid (%)	6.6	9.3	40.7	40.2	<.001
Average annual income (dollars)	35,198	28,303	26,222	17,383	<.001
Total sample (%)	68.8	10.1	9.4	11.7

Note: The mean is presented for continuous variables. The p values reflect the design-based F-test that corrects for the NHATS complex survey design and tests for differences in the weighted distributions between community housing types (owners without HCB, owners with HCB, renters without HCB, and renters with HCB). ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living; PHQ-4 = Patient Health Questionnaire-4.

Table 2.

Weighted Characteristics of Low- and Moderate-Income Community-Dwelling Older Adults in 2015 by Place of Residence 3 Years Later (N = 3,403)

	Community	Nursing home	Deceased	p
Total	94.2	1.7	4.2
Housing typea				<.001
No HCB/own	94.6	1.2	4.2
HCB/own	97.2	0.5	4.0
No HCB/rent	90.6	3.4	6.1
HCB/rent	91.9	4.0	4.1
Gender				.013
Female	95.0	1.6	3.4
Male	93.0	1.7	5.4
Age (years)				<.001
65–79	96.3	1.0	2.7
80+	88.0	3.4	8.6
Race				.079
Person of color	95.7	0.9	3.4
White	93.8	1.8	4.4
Living situation				.018
Lives with others	94.7	1.2	4.1
Lives alone	93.3	2.5	4.2
Probable dementia	77.4	7.8	15.0	<.001
Fair/poor self-ratedhealth	89.7	3.0	7.3	<.001
Mental healthconcern (PHQ-4)	90.5	2.9	6.6	<.001
Physical capacitylimitation	92.4	2.2	5.4	<.001
ADL limitation	89.8	3.1	7.0	<.001
IADL limitation	90.2	3.4	6.6	<.001
Medicaid status				.004
Medicaid	91.2	3.5	5.4
No Medicaid	94.7	1.4	4.0
Annual income				.002
Less than $29,999	92.6	2.3	5.1
$30,000+	95.7	1.0	3.3

Note: p Values are based on design-based F-test that corrects for the NHATS complex survey design. ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living; PHQ-4 = Patient Health Questionnaire-4.

^aSample sizes for nursing home movers by housing type: no HCB/own (n = 36); HCB/own (n = 3); no HCB/rent (n = 12); HCB/rent (n = 23).

Table 3.

Multinomial Logistic Regression Models Predicting Moves From the Community in 2015 to a Nursing Home by 2018 (N = 3,105)

	1	2	3	4	5	6	7	8	9
Housing type (ref = no HCB/own)a
No HCB/rent	3.51*	2.21	1.91	2.68*	0.55	1.82	0.78	1.20	5.01*
HCB/rent	4.14***	2.97**	2.79*	3.22*	0.63	6.63**	0.92	2.28†	3.66*
Female	0.70	0.55*	0.56†	0.56*	0.55*	0.52*	0.56*	0.57†	0.54*
Age (1–6 scale)	1.48***	1.22†	1.24*	1.26*	1.23*	1.24*	1.25*	1.23*	1.23*
Person of color	0.37*	0.19**	0.16**	0.17**	0.17**	0.16**	0.17**	0.16**	0.16**
Lives alone	1.36	1.79*	1.76†	1.75†	1.79*	1.80*	1.71†	1.76†	1.79*
Probable dementia		4.05***	3.91***	5.81***	3.89***	3.92***	3.83***	3.90***	4.00***
Self-rated health (1–5 scale)		0.86	0.88	0.87	0.72	0.91	0.89	0.88	0.89
Mental health (PHQ-4)		1.22	1.20	1.19	1.19	2.06	1.20	1.18	1.21
Physical capacity (0–12 scale)		0.87*	0.87*	0.88*	.88*	0.88*	0.81**	0.88*	0.87*
ADL limitation		1.15	1.15	1.14	1.16	1.19	1.13	0.95	1.16
IADL limitation		1.47	1.50	1.51	1.47	1.62	1.47	1.50	2.21†
Medicaid			1.59	1.60	1.69	1.62	1.64	1.57	1.64
Income (logged)			1.02	1.02	1.01	1.03	1.03	1.02	1.02
Interactions (ref = no HCB/own)
Dementia × no HCB/rent				0.21†
Dementia × HCB/rent				0.57
Self-rated health × no HCB/rent					1.55
Self-rated health × HCB/rent					1.67†
Mental health × no HCB/rent						0.89
Mental health × HCB/rent						0.14*
Physical capacity × no HCB/rent							1.15
Physical capacity × HCB/rent							1.18*
ADL limitation × no HCB/rent								1.90
ADL limitation × HCB/rent								1.35
IADL limitation × no HCB/rent									0.21
IADL limitation × HCB/rent									0.60
Constant	0.02***	0.03***	0.02*	0.05†	0.02*	0.08	0.02*	0.05†	0.06†

Note: Referent nonmovers. Coefficients are reported as relative risk ratios; regressions employ weights and adjust for standard errors to account for complex survey design. ADL = activities of daily living; HCB = housing cost burden; IADL = instrumental activities of daily living; PHQ-4 = Patient Health Questionnaire-4.

^aOnly three homeowners with housing cost burden (HCB/own) moved from the community to a nursing home between baseline and follow-up. Due to the inadequate sample size, HCB/own was dropped from the regression models (n = 298).

^† p < .10. *p < .05. **p < .01. ***p < .001.

Our multivariate analysis builds from previous work examining black–white disparities in moving out of community residence and to assisted living or a nursing home (Jenkins Morales & Robert, 2019). A stepwise approach was used to examine whether health and functioning (Model 2) and other economic factors (Model 3) explained the association between community housing type and nursing home use. To test our third hypothesis, we added interaction terms between community housing type and measures of health and functioning (dementia, self-rated health, mental health, physical capacity, ADL limitation, and IADL limitation) to our final model specification (Model 3). The corresponding NHATS 2015 analytic weights were used in all analyses and standard errors were adjusted to account for NHATS’ complex sample design (Kasper & Freedman, 2018). Analyses were conducted in Stata Version 16.0 (Stata Corp., College Station, TX).

Results

Table 1 presents weighted baseline characteristics by community housing type with p values that reflect the design-based F-test that corrects for the NHATS complex survey design and tests for differences in the weighted distributions between community housing types (owners without HCB, owners with HCB, renters without HCB, and renters with HCB). Roughly 22% of low- and moderate-income community-dwelling older adults experienced HCB in 2015, with 57% of renters (n = 459) having HCB compared with 11% of homeowners (n = 298). Renters were more likely to live alone, have Medicaid, be in worse health, and have lower levels of functioning than homeowners. Among renters, those with HCB were more likely to have dementia (13.1% vs 8.5%) and potential mental health concerns (41.1% vs 35.4%) and those without HCB were more likely to live alone (62.5% vs 50.6%), and have an ADL (45.7% vs 39.7%) or IADL (49.3% vs 44.8%) limitation. Renters with HCB were signficantly more likely to be a person of color (43.6%) compared with all other housing types (p ≤ .001).

Table 2 presents weighted baseline characteristics for our analytic sample by residence type (or death) 3 years later. By 2018, 94.2% of the sample remained in the community, 1.7% moved to a nursing home, and 4.2% died. Participants with probable dementia in 2015 were most likely to move to a nursing home (7.8%) or die (15.0%) by 2018. Overall, those who remained in the community over 3 years were more likely to be homeowners, female, younger, living with others, higher income, and in better health (all p ≤ .05). Based on community housing type in 2015, renters with HCB were most likely to move to a nursing home (4.0%) and renters without HCB (6.1%) were most likely to die 3 years later (p < .001). Only three homeowners with HCB moved to a nursing home by 2018. Due to sample size constraints, we were not able to include homeowners with HCB in subsequent multivariate analyses.

Table 3 presents results from our multinomial logistic regression predicting moves from the community in 2015 to a nursing home by 2018. Model 1 shows that after adjusting for gender, age, race, and living arrangement, renters (with and without HCB) were significantly more likely than homeowners without HCB to move from the community to a nursing home over 3 years, with renters with HCB most likely (RR: 4.14, p < .001). The likelihood of moving to a nursing home for renters without HCB was 3.51 times that of homeowners without HCB (p = .02) and for renters with HCB the likelihood was 4.14 times that of homeowners without HCB (p < .001). There was no statistically signficant difference when comparing renters with and without HCB (RR: 1.18, p = .73). Taken together, the bivariate results shown in Table 2 and the multivariate results shown in Table 3 (Model 1) support our first hypothesis that renters compared with homeowners are more likely to move to a nursing home over 3 years and that renters with HCB are most likely to move to a nursing home.

As seen in Table 3 (Model 2), after adding health and functioning measures, there was no longer a statistically significant difference between renters without HCB and homeowners without HCB (RR: 2.21, p = .15); however, renters with HCB remained significantly more likely to move to a nursing home compared with homeowners without HCB (RR: 2.97, p = .02). Compared with homeowners without HCB, poorer health and lower levels of functioning among renters with HCB in 2015 only partially acounted for the higher likelihood of moving to a nursing home 3 years later. In Model 3, adding income and Medicaid receipt to our analysis also partially mediated the association between renters with HCB and likelihood of moving to a nursing home (compared with homeowners without HCB), but a significant direct effect remained (RR: 2.79, p = .01).

Even after accounting for a variety of health and functioning measures and other economic factors, being a renter with HCB (compared to a homeowner without HCB) was associated with a 179% increase in the likelihood of moving to a nursing home over 3 years (p = .01). Although only at the trend level, being a renter with HCB was associated with a 46% higher likelihood of moving to a nursing home compared to renters without HCB (RR: 1.46, p = .44). Results presented in Table 3 support our second hypothesis that differences in health and functioning (Model 2) and other economic factors (Model 3) do not fully account for the greater likelihood of moving to a nursing home among renters with HCB.

To test our third hypothesis that the association between health and functioning and likelihood of moving to a nursing home will be weakest for renters with HCB, we included a series of interactions in Table 3 (Models 4–9). In support of our hypothesis, Models 5–7 demonstrate that a similar proportion of renters with HCB and homeowners without HCB reporting poor self-rated health (2%), a mental health concern (2%), or low physical capacity (5%) will move to a nursing home over 3 years; however, the likelihood of moving to a nursing home among renters with HCB, compared with homeowners without HCB, was higher for renters with HCB in excellent health (4% vs 1%), without an identified mental health concern (5% vs 1%), or at full physical capacity (3% vs 1%) (all p < .10). But not in support of our third hypothesis, for the dementia interaction, the association between having dementia and moving to a nursing home was weakest for renters without HCB, compared with homeowners without HCB (RR: 0.21, p = .08). The likelihood of moving to a nursing home among participants with dementia, compared with those without dementia, was considerably higher for renters with HCB (8% vs 3%) and homeowners without HCB (5% vs 1%), but estimates were similar for renters without HCB regardless of probable dementia status (3% vs 2%). Although not statistically significant, the interaction results for community housing type and ADL limitation and IADL limitation also did not support our third hypothesis. Taken together, the interaction results in Models 4–9 partially support our third hypothesis that poor health and functioning is a weaker predictor of moves to a nursing home among renters with HCB, compared to other housing types.

Discussion and Implications

This is the first empirical study we know of to examine whether HCB and housing tenure are associated with future moves to a nursing home among older adults. Using contemporary national data, the results support our hypothesis that renters with HCB are most likely to move to a nursing home (compared with other housing types), even after adjusting for differences in demographic characteristics, health and functioning, and other economic factors. The results are consistent with prior evidence that renters are more vulnerable to nursing home admission (McCann et al., 2012; Rouwendal & Thomese, 2013). We expand on this prior work by demonstrating that higher levels of HCB among renters, compared to homeowners, partially drives this association. Results also suggest that renters with HCB are more likely to move to a nursing home in better health compared to homeowners without HCB.

Bivariate results suggest significant racial/ethnic disparities in homeownership and experiencing HCB, with older adults of color representing 44% of renters with HCB and only 14% of homeowners without HCB. As renters with HCB are most likely to move to a nursing home, future research and policy should consider how racial/ethnic disparities in access to affordable housing might in turn influence disparities across LTC settings. Recent evidence suggests that black overrpresentation in nursing homes is explained by black older adults having fewer financial resources (including homeownership) and worse health compared with whites (Jenkins Morales & Robert, 2019); however, more research is needed to better understand how racial disparities in housing and community-based LTC options influence future nursing home use across different racial and ethnic groups. Living in multigenerational households also significantly varies by race and ethnicity (Joint Center for Housing Studies, 2019), which has unique implications for the experience of HCB and requires further research.

Limitations of this study include attrition in the NHATS unrelated to mortality, and the short 3-year follow-up period, which limited the number of moves we were able to study. Renters with HCB were significantly more likely to be lost to attrition, compared with all other housing types, which potentially downwardly biases our estimates because renters with HCB remaining in our sample are likely less vulnerable than the general population of renters with HCB. Also, as our measure of HCB only includes monthly rent or mortgage payments, we underestimate the proportion of older adults experiencing HCB, in particular homeowners with HCB, compared with national estimates (Joint Center for Housing Studies, 2019) that include other housing-related expenses (e.g., taxes, utilities, and association fees). Attrition in the NHATS, the short follow-up period, and lower estimates of HCB all contributed to sample size constraints and our inability to include owners with HCB in the mulivariate analysis. To better understand the specific needs of homeowners with HCB, future research should examine the use of LTC services among this population. Strengths of this study include the use of recent data from a nationally representative sample of older adults, prospective follow-up on moves to a nursing home, strong measures of nursing home use, a combined examination of housing tenure and HCB, and multiple measures of physical and mental health.

The results demonstrate that health and functioning do not fully explain why older adults use LTC services in a nursing home and suggest that the housing context of older adults also drives nursing home use regardless of health and functioning. Drawing from the Wiseman model (1980), the results suggest that being a renter and experiencing HCB operate as push factors, putting renters with HCB at greatest risk of nursing home admission. Given the current LTC and housing contexts in the United States, renters with HCB also might be pulled toward nursing homes as the only LTC and affordable housing option. The results and conceptual framework used in this study can inform current efforts to develop a measure of P–E fit among older adults (Weil, 2019) to adequately assess and summarize “aging in the right place” (Golant, 2015). More primary research is also needed to better understand the experiences of older adults with HCB and how living in affordable housing influences the ability of older adults to age in an environment that appropriately matches their needs.

Of the roughly 4.9 million very-low-income older households that were eligible for subsidized housing in 2015, just 1.6 million (33%) received assistance (HUD, 2017; Joint Center for Housing Studies, 2019), and by 2030 the number of older households eligible for subsidized housing is projected to increase by over 2 million (Stone, 2018). Although evidence suggests that affordable supportive housing options can prevent nursing home admissions (Castle & Resnick, 2016; Park, Kim, Kwon, & Kown, 2019), high out-of-pocket costs and few public financing options limit the ability of low- and moderate-income older adults to access community-based alternatives to nursing home care (Stone, 2018). Consistent with formal evaluations of nursing home prevention programs (Bardo et al., 2013; Hoffman et al., 2017; Lipson et al., 2011), these empirical results strengthen the evidence that access to affordable housing across the continuum of LTC is critical to prevent costly nursing home use and allow people to live in an environment that appropriately matches their needs.

Given the growing population of older renters with HCB (Joint Center for Housing Studies, 2019), these results encourage further collaboration across aging, housing, and health care sectors to develop affordable housing solutions that meet the needs of the aging population. As the LTC landscape continues to evolve, a growing number of state Medicaid programs are providing managed long-term services and supports (MLTSS) through capitated contracts with managed care organizations (MCOs) (Lewis, Eiken, Amos, & Saucier, 2018). Some states have incorporated housing needs into their risk-based payment formulas to MCOs (Ash et al., 2017), suggesting growing recognition of housing as a social determinant of health (Taylor, 2018); however, little is known about whether MLTSS models are addressing the housing needs of their members to promote healthy aging in the community (Archibald, Kruse, & Somers, 2018).

Affordable housing has been described as “the glue that holds everything together; without access to such housing and the stability it provides, it becomes increasingly difficult, if not impossible, to introduce a system of community-based services that can enable successful aging” (Bipartisan Policy Center, 2016, p. 23). The current study provides empirical evidence of disparities in the ability of older renters and those experiencing HCB to remain in the community and suggests that access to affordable housing is an important protective factor against future nursing home use. In order to uphold the 1999 U.S. Supreme Court Olmstead decision affirming the rights of persons with disabilities to live in the community, advocates, policy makers, and program administrators must work to eliminate disparities in nursing home use based on access to affordable housing. To inform these efforts, more research is needed to better understand the intersection of affordable housing and use of formal LTC services among economically vulnerable older adults.

Funding

This work was supported by National Institutes of Health funding to the University of Wisconsin-Madison Center for Demography of Health and Aging P30 AG017266, and the University of Wisconsin-Madison Office of the Vice Chancellor for Research and Graduate Education with funding from the Wisconsin Alumni Research Foundation. The National Health and Aging Trends Study (NHATS) is sponsored by the National Institute on Aging NIA U01AG032947 through a cooperative agreement with the Johns Hopkins Bloomberg School of Public Health.

Conflict of Interest

None reported.