Household Accessibility and Residential Relocation in Older Adults

Abstract

Objectives

It is unclear how home environmental factors influence relocation decisions. We examined whether indoor accessibility, entrance accessibility, bathroom safety features, housing type, and housing condition were associated with relocations either within the community or to residential care facilities.

Methods

We used prospective data over 4 years from the nationally representative National Health and Aging Trends Study in the United States of Medicare beneficiaries 65 years and older living in the community (N = 7,197). We used multinomial regression analysis with survey weights.

Results

Over the 4 years, 8.2% of the population moved within the community, and 3.9% moved to residential care facilities. After adjusting for demographics and health factors, poor indoor accessibility was found to be associated with moves within the community but not to residential care facilities. No additional home environmental factors were associated with relocation.

Discussion

One-floor dwellings, access to a lift, or having a kitchen, bedroom, and bathroom on the same floor may help older adults age in place. Understanding which modifiable home environmental factors trigger late-life relocation, and to where, has practical implications for developing policies and programs to help older adults age in their place of choice.

The home environment has a vital role in supporting aging in place. Research consistently demonstrates that accessible dwellings, those without environmental barriers or home hazards, can facilitate older adults’ independence (Mahler et al., 2014; Oswald et al., 2007). The home environment is to some extent modifiable and home modifications and safety features such as grab bars, raised toilet seats, and shower stools have been shown to decrease difficulties with activities of daily living (ADLs), reduce falls, increase participation in society (Keall et al., 2015; Stark, Keglovits, Arbesman, & Lieberman, 2017; Wahl et al., 2009), and facilitate aging in place (Kendig, Gong, Cannon, & Browning, 2017). Conversely, it is unclear whether inaccessible homes and those lacking safety features trigger older adults to move. Despite increasing attention to housing needs and ways to modify homes to support aging in place, there is a dearth of knowledge concerning whether the home environment affects relocation decisions (Granbom, 2014). Although the decision making has been shown to be different for moves within the community and moves to residential care facilities (Oswald & Rowles, 2007), relocation predictors are rarely compared based on the destination of the move in nationally representative samples (van der Pers, Kibele, & Mulder, 2017). Understanding the role of the home environment in relocation decision making in later life is essential for developing policies and services that support living choices for older adults.

This study examines the role of the home environment in triggering relocation, both within the community and to residential care facilities among community-living adults ≥65 years of age. It systematically builds upon and extends previous research in this area by using data from a nationally representative U.S. sample of older adults. Specifically, the aim of this study was to examine whether indoor accessibility, entrance accessibility, bathroom safety features, housing type, and housing condition were associated with relocation within the community or to residential care facilities.

Background

Most older adults prefer to age in their current home or community, even in the context of future health decline and dependence (AARP, 2012). However, overall rates of residential relocation of older adults in the United States have been documented to vary from 5%–30% over a period of 5 years (Sergeant, Ekerdt, & Chapin, 2008).

To understand relocations and the role of the home environment, we use person–environment (P–E) fit as an overarching conceptual framework for this study. Lawton’s ecological model of aging suggests that well-being and behavior is a consequence of the interaction between older adults’ capabilities and the demands of the environment (Lawton & Nahemow, 1973). Relocation may be a consequence of the incongruence between an older adult’s functional abilities and features of the home environment. The presence of stairs or other hindering structural features may place a person at increased risk for falls, or increased difficulty to function at home. Alternately, a move can be a proactive action; a coping strategy used to relocate to a dwelling that is more likely to support daily function and accommodate anticipated future age-related health and functional changes, and a way to optimize one’s living environment (Golant, 2011; Lawton, 1989). We also draw upon the disablement model (Verbrugge & Jette, 1994), which suggests that supporting older adults to age in place may be achieved by addressing intrinsic factors (e.g., health condition) as well as by attending to extrinsic factors (e.g., the home environment). The home environment is modifiable. Making the home safe and accessible through home modifications such as installing rails, first-floor bathrooms, or safety features such as grab bars may improve P–E fit. Also, according to these conceptual frameworks, relocation may reflect an older adult’s proactive attempt to improve P–E fit.

Previous research on relocation to residential care facilities has primarily focused on intrinsic factors such as health conditions and demographics as predictors of relocation. Comprehensive reviews and meta-analyses confirm that a move is typically a consequence of functional decline and substantial need for help to manage day-to-day life. Specifically, difficulty with basic ADLs, mobility problems, frailty, cognitive decline, and dementia have consistently been shown to predict nursing home placement among community-living older adults (Gaugler, Duval, Anderson, & Kane, 2007; Kojima, 2016; Luppa et al., 2010; Miller & Weissert, 2000).

The demands of the physical home environment (E), with few exceptions, have not been examined as potential triggers of relocation to residential care facilities. In one of the few studies, Stineman and colleagues (2012), based on data from the 1994 Longitudinal Study on Aging, showed that self-reported built home environmental barriers (i.e., unmet need for supportive features or home modifications) increased the odds of nursing home admission (Stineman et al., 2012). A Swedish study of 387 older adults living alone found that the combination of ADL difficulties, cognitive decline, and poor household accessibility predicted relocation to residential care facilities more strongly than either having an ADL difficulty or cognitive decline alone. Household accessibility was measured in terms of the fit between a person’s functional abilities and 188 environmental factors in and around the home (Granbom, Löfqvist, Horstmann, Haak, & Iwarsson, 2014). Studies on nonphysical aspects of the environment, such as the social context and availability of care showed a trend such that risk of relocation to residential care facilities increases as care needs extend beyond what a community can offer. A consistent predictor was living alone (Gaugler et al., 2007; Miller & Weissert, 2000). Marital status, proximity to children, and availability of informal care have been associated with relocation to residential care facilities, but the principal factor was likely support coming from coresidents within the household (McCann, Donnelly, & O’Reilly, 2011).

Turning to relocation within the community, factors predicting those moves have not been as extensively studied as relocation to residential care facilities. Dutch studies have demonstrated that although relocation both within the community and to residential care facilities was associated with worsening health and disability, the associations were not as strong for relocation within the community. Although severe ADL difficulty was associated with relocation to residential care facilities, moderate ADL difficulty was associated with relocation within the community (Bloem, Tilburg, & Thomése, 2008; van der Pers et al., 2017).

The role of the home environment in triggering moves within the community has mostly been examined in qualitative studies. Findings from these studies indicate a complex decision-making process in which multiple factors are considered, including the home environment as well as health conditions, family, and cost-related factors (Roy, Dubé, Després, Freitas, & Légaré, 2018; Sergeant & Ekerdt, 2008). Moving within the community can be a way to optimize the living environment by relocating into neighborhoods with better access to services, dwellings with better amenities, downsizing, or moving closer to family (Bjelde & Sanders, 2012; van Diepen & Mulder, 2009; Oswald, Schilling, Wahl, & Gäng, 2002). A Dutch study showed that lack of home modifications and the presence of stairs predicted moves within the community for older adults who recently had become widows or widowers (Bloem et al., 2008). It is unclear whether this finding was country specific, and to date, it has not been examined in the United States.

There is a lack of prospective studies of nationally representative samples on how contextual factors, such as home environmental factors, might contribute to relocation over and above established health and sociodemographic predictors. Such knowledge is needed to deepen our understanding of the complexity of residential decision making and from which supportive programs could be developed to help older adults plan for aging in the place of their choice.

Method

In this prospective study, we used baseline data from 2011 and data on moves from yearly follow-ups between 2012 and 2015, all from the longitudinal National Health and Aging Trends Study (NHATS). NHATS uses a nationally representative sample of Medicare beneficiaries aged 65 years and older (N = 8,245) to provide a deeper understanding of national trends in late-life functioning including information on contributing contextual factors (Kasper & Freedman, 2014). Information on study design and methods have been published previously (Montaquila, Freedman, Edwards, & Kasper, 2012). Briefly, all individuals enrolled in Medicare were included in the sampling, representing 96% of the population. Cases were selected with a stratified, three-stage design: first counties or groups of counties were sampled, then ZIP codes, and finally Medicare beneficiaries enrolled as of September 30, 2010. African Americans and older participants (85 years and older) were oversampled by design. Yearly data collections were made at home visits, which started in 2011. The response rate at baseline was 71%. Proxy interviews were used for participants unable to respond, as well as for participants who died during the study, and a last-month-of-life interview was conducted. Written informed consent was provided to the participants or their proxy respondents, and the Johns Hopkins University Institutional Review Board approved the study protocol.

Sample

In order to focus on how home environments in the community influence relocation, for the present analyses, we excluded individuals living in residential care facilities at baseline—nursing homes, assisted living facilities, and independent living settings (n = 1,048, 12.7%). This yielded a study sample of community-living participants (N = 7,197; 87.3%) with a proxy response rate of 5.8%. Participants were coded as movers if they responded as having a new address in 2012, 2013, 2014, or 2015 and as non-movers if they stated no change of address. Address information could be obtained from 6,082 participants in 2012; 5,032 in 2013; 4,184 in 2014; and 3,743 in 2015. For all 4 years, 1,157 (16.1%) last-month-of-life interviews were conducted. In all, n = 961 (13.4%) participants moved, including n = 117 (12.2%) participants who moved but died between rounds (information retrieved from last-month-of-life interviews).

We distinguished between two categories: movers within the community, that is, those who moved to traditional community housing, to a retirement community, or to senior housing (n = 588; 8.2%); and movers to residential care (n = 373; 5.2%), those who moved to nursing homes, assisted living facilities, and independent living settings (n = 373; 5.2%). Type of facility was determined on follow-up questions and observations made by the NHATS interviewers and described in detail elsewhere (Freedman & Spillman, 2014). Participants who moved more than once during the study were categorized based on the first move (n = 170, 17.7%).

Measures

Home environmental factors

Several aspects were collected using interviewer observations during or directly after the home visit. For indoor accessibility, the interviewers were asked whether the homes were a single-floor unit or multi-floor unit. The interviewer recorded whether the multi-floor units had a lift (yes/no), stair-glide (yes/no), or whether the bathroom bedroom and kitchen were on the same floor (yes/no). We categorized indoor accessibility as having no stairs (2); having stairs but lift, stair-glide, or bathroom, bedroom, and kitchen on the same floor (1); and having stairs (0). Mobile homes were coded as single-floor units, thus, having no stairs. For entrance accessibility, interviewers rated whether the home had steps (yes/no) and ramp (yes/no) at the entrance of the home. We categorized entrance accessibility into having no steps at the entrance (2), steps but ramp at the entrance (1), and steps at the entrance (0). Information on bathroom safety was collected by asking the participants whether they had a shower or only bathtub, grab bars in bath/shower area, grab bars in the toilet area, raised toilet or raised toilet seat, bath seat in bath/shower area, and medical emergency system. Items present were coded as 1, and a summary score was derived (range 0–6). The interviewer recorded whether type of housing (a building’s physical structure) was either house (detached single house, attached single house, or mobile home) or apartment (multiunit dwelling or other). For housing condition, the interviewer observed whether the home had broken or boarded-up windows, a crumbling foundation, missing bricks or sidings, or uneven surfaces or broken steps leading to the home (yes/no). A summary score was derived by adding number of conditions present (range 0–4).

Confounding variables

Demographic variables included age (65–69, 70–74, 75–79, 80–84, 85–89, and 90+ years), sex, race and ethnicity (white/Caucasian, black/African American, and Other, with the majority of Other being Hispanic ethnicity or Asian), education (less than high school, high school or General Educational Development or trade, some college, and bachelor’s degree or greater), living alone, Medicaid eligibility, length of residence in current dwelling, and income. For income, we used a derived variable provided by the NHATS team in the open file on total income from all sources during the prior year with imputed values for 43.6% of the sample (see Montaquila et al., 2012). To reduce the influence of outliers, we recorded categories based on quintiles (<12,000, 12,000–19,999, 20,000–33,999, 34,000–59,999, and ≥60,000). We treated income and age as ordinal variables in the analysis.

We controlled for several health-related factors. On self-rated health, participants were asked to rate their overall health on a scale from 1 (excellent) to 5 (poor). For our analyses, we reversed the scale so the higher the rating, the better health. To measure ADL difficulty, we created a count (0–5) of selected activities for which participants reported either receiving help or having difficulty performing without help in the last month. The included activities are eating, bathing/showering, using the toilet, dressing, and moving around indoors. On cognitive functioning, we used the NHATS dementia classification scheme as a categorical variable: no dementia (2), possible dementia (1), and probable dementia (0; Kasper, Freedman, & Spillman, 2013). Participants were asked whether they had had any falls in the last year (yes/no) and whether they had been admitted to hospital in the last year (yes/no).

Statistical Analysis

Analyses incorporated analytic weights to adjust for the complex survey design and nonresponse (Montaquila et al., 2012) in the NHATS survey. We used bivariate multinomial logistic regression to examine home environmental factors, demographics, and potential confounders individually, and their relation to relocation within the community and to residential care within 4 years from baseline. Non-movers were the reference category. Then we estimated a multivariable model with all home environmental factors, adjusted for demographics and potential confounders. In the adjusted models, we tested two additional functional measures: instrumental ADLs (IADL) and unmet need of assistance with ADLs, first by adding them individually to the adjusted models and second by replacing ADL difficulty with each measure. These tests did not significantly change the results. We, therefore, presented the models with ADL difficulty only. In addition, we did a subgroup analysis of the residential care movers (n = 373) to examine differences between movers to nursing homes and those to other types of residential care facilities (assisted living facilities and independent living settings). Information on type of residential care facility was missing for n = 75 cases. Thus the analytic sample included n = 118 participants who moved to nursing homes and n = 180 and cases who moved to assisted living facilities and independent living settings. We used binomial logistic regression analysis with weights and adjusted for demographics and potential confounders. We used STATA svy 14 (Stata Corp., College Station, TX) for all analyzes. A significance level of p value < .05 was considered statistically significant.

Results

In our analytic sample of 7,197 participants, 4,147 (57.6%) were women and 3,050 (46.4%) were men. The vast majority lived in houses (n = 6,367, 88.5%), and 2,169 (30.1%) lived alone. In all, 961 (13.4%) participants moved in 4 years.

Applying weights to reflect the total population of community-dwelling Medicare beneficiaries in the United States in the 4 years of the study, an estimated 2,738,455 was found to have moved within the community (8.2%), 1,287,507 moved to residential care facilities (3.9%), and 29,329,152 did not move (87.9%). Table 1 shows baseline characteristics of the weighted sample.

Table 1.

Participant Characteristics (N = 7,197, weighted sample = 33,355,114)

Variables	Non-movers (n = 6,193)		Movers within the community (n = 588)		Movers to residential care (n = 373)
	%	n	%	n	%	n
Weighted sample	87.9	29,329,152	8.2	2,738,455	3.9	1,287,507
Demographics
Age, Md (range)	70–74	(65–90+)	70–74	(65–90+)	80–84***	(65–90+)
Sex					**
Women	55.2		58.2		62.9
Men	44.8		41.8		37.1
Race					**
White/Caucasian	80.0		76.4		88.6
Black/African American	8.3		9.0		5.8
Other	11.7		14.7		5.6
Education
Less than high school	22.5		25.7		23.9
High school/GED/Trade	34.8		30.2		34.0
Some collage	18.2		21.0		20.5
Bachelor’s or higher	2.4		23.2		21.6
Living alone	25.8		33.5**		41.6***
Medicaid eligible	11.2		15.8**		13.1
Length of residence, m (SE)	23.3	(0.40)	14.1***	(0.60)	23.3	(1.08)
Total annual income, Md (range)	20,000–33,999	<12,000–≥ 60,000	20,000–33,999**	<12,000–≥ 60,000-	20,000–33,999***	<12,000–≥ 60,000-
Health
Self-rated healtha					**
Poor	6.6		6.0		11.1
Fair	18.2		16.1		20.2
Good	30.3		32.0		30.4
Very good	29.7		30.6		27.3
Excellent	15.2		15.4		11.1
ADL difficulty, m (SE)b	0.4	(0.01)	0.42	(0.04)	0.9***	(0.08)
Dementia					***
Probable dementia	7.9		7.4		26.7
Possible dementia	11.0		13.1		16.3
No dementia	81.1		79.6		57.0
Fallen	28.9		31.8		44.7***
Hospital stay	19.3		23.6*		37.0***
Housing
Indoor accessibility			**
Single floor	44.9		46.7		49.8
Multiple floors but lift, glide, BBKc on same floor	42.4		32.1		38.8
Multiple floors	12.8		21.2		11.4
Entrance accessibility
No step entrance	21.7		24.9		23.5
Steps at entrance but ramp	5.3		4.5		6.6
Steps at entrance	73.1		70.5		69.8
Bathroom safety, m (SE)d	1.8	(0.02)	1.8	(0.07)	2.6***	(0.09)
Type of housing			***		**
Apartment	9.5		18.1		15.4
House	90.5		81.9		84.6
Housing condition, m (SE)e	0.3	(0.02)	0.3	(0.04)	0.3	(0.05)

Note: GED = General Educational Development; m = mean; SE = standard error.

^a1–5, higher means better health (reversed).

^b0–5, higher means more difficulty.

^cBBK = bathroom, bedroom, and kitchen on the same floor.

^d0–6, number of present home modifications and safety features.

^e0–4, higher means worse housing condition.

*Statistically significant difference from non-movers with p value < 0.05.

**Statistically significant difference from non-movers with p value < 0.01.

***Statistically significant difference from non-movers with p value < 0.001.

When we tested each factor in bivariate analyses, living alone, Medicaid eligibility, shorter length of residence in current dwelling, lower annual income, and experiencing one or more hospital admissions in the last year increased the likelihood of moving within the community.

As to the home environment, indoor accessibility and type of housing were associated with moves within the community. Community-living older adults were less likely to move if they lived in a single-floor unit (relative risk ratio [RR] = 0.625, 95% confidence interval [CI] = 0.451–0.867) or had a lift, stair glide or bath, bedroom, and kitchen on the same floor (RR = 0.457, 95% CI = 0.322–0.647). The likelihood of moving within the community was greater for those living in an apartment than for those living in a house (RR = 2.091, 95% CI = 1.604–2.725; see Table 2).

Table 2.

Demographic Factors, Health Factors, and Housing Factors Associated With Relocation Within 4 years Within the Community or to Residential Care Facilities: Unadjusted

Variables	Movers within the community vs non-movers				Movers to residential care vs non-movers
			95% CI for RR				95% CI for RR
	RR	p Value	Lower bound	Upper bound	RR	p Value	Lower bound	Upper bound
Demographics
Age	0.975	.449	0.912	1.042	1.900	<.001	1.748	2.056
Sex: women vs men	1.130	.170	0.947	1.348	1.379	.007	1.100	1.737
Race
Black/African American vs white/ Caucasian	1.130	.319	0.885	1.443	0.635	.016	0.441	0.914
Other vs white/Caucasian	1.318	.063	0.984	1.765	0.433	.001	0.267	0.702
Education	0.976	.641	0.879	1.083	0.960	.382	0.875	1.053
Living alone vs not living alone	1.450	.001	1.164	1.806	2.051	<.001	1.579	2.666
Medicaid eligible: yes vs no	1.484	.003	1.151	1.912	1.196	.321	0.835	1.713
Length of residence	0.963	<.001	0.956	0.970	0.999	.996	0.993	1.007
Annual total income	0.889	.001	0.832	0.951	0.794	<.001	0.728	0.867
Health
Self-rated healtha	1.037	.449	0.943	1.140	0.845	.001	0.770	0.927
ADL difficultyb	1.046	.269	0.964	1.134	1.395	<.001	1.299	1.498
Dementia
Probable dementia vs no dementia	0.950	.730	0.705	1.280	4.808	<.001	3.526	6.556
Possible dementia vs no dementia	1.208	.156	0.928	1.573	2.109	<.001	1.427	3.118
Fallen last year: yes vs no	1.147	.101	0.973	1.352	1.990	<.001	1.543	2.565
Hospital stay: yes vs no	1.292	.039	1.014	1.646	2.449	<.001	1.947	3.080
Housing
Indoor accessibility
Single floor vs multiple floors	0.625	.006	0.451	0.867	1.239	.333	0.798	1.925
Multiple floors but lift, stair glide, or BBKc on same floor vs multiple floors	0.457	<.001	0.322	0.647	1.024	.915	0.6421	1.638
Entrance accessibility
No step entrance vs steps at entrance	1.192	.084	0.976	1.455	1.136	.436	0.820	1.574
Steps at entrance but ramp vs steps at entrance	0.891	.574	0.591	1.342	1.316	.280	0.787	2.202
Bathroom safetyd	0.986	.695	0.916	1.061	1.381	<.001	1.289	1.481
Housing: apartment vs house	2.091	<.001	1.604	2.725	1.727	.002	1.238	2.412
Housing conditione	0.972	.557	0.882	1.071	0.901	.225	0.760	1.068

Note: Multinomial logistic regression model, with weights. The reference category is non-movers. ADL = activities of daily living; CI = confidence interval; RR = relative risk ratio.

^a1–5, higher means better health (reversed).

^b0–5, higher means more difficulty.

^cBBK = bathroom, bedroom, and kitchen on the same floor.

^d0–6, number of present home modifications and safety features.

^e0–4, higher means worse housing condition.

For residential care, several demographic and health-related factors were associated with relocation. Increasing age, being a woman, living alone, ADL difficulties, falls and hospital admissions in the last year, and possible or probable dementia increased the relative risk of moving to a residential care facility compared to not moving. Not being white/Caucasian, having a lower annual income, and higher self-rated health, on the other hand, decreased the likelihood of moving to a residential care facility.

For the home environment, bathroom safety and type of housing were associated with moves to residential care. That is, compared to non-movers, older adults were more likely to move the more bathroom safety features they had (RR = 1.381, 95% CI = 1.289–1.481). They were also more likely to move and if they lived in an apartment rather than a house (RR = 1.727, 95% CI = 1.238–2.412) and see Table 2.

After adjusting for demographic and health-related factors, we found the associations with length of residence in the current dwelling and hospital admissions in the last year remained statistically significant for relocation within the community. Also, higher self-rated health increased the relative risk of moving within the community compared to not moving.

Regarding the home environment, after adjusting, indoor accessibility remained associated with moves within the community. Specifically, the likelihood of moving decreased for older adults living in a single-floor unit (RR = 0.580, 95% CI = 0.412–0.815) or having a lift, stair glide or bath, bedroom, and kitchen on the same floor (RR = 0.499, 95% CI = 0.350–0.710; see Table 3).

Table 3.

Housing Factors Associated With Relocation Within 4 years Within the Community or to Residential Care Facilities, Adjusted for Demographic Factors and Health-Related Factors (n = 6,840, Weighted Sample = 31,826,831)

Variables	Movers within the community vs non-movers				Movers to residential care vs non-movers
			95% CI RR				95% CI RR
	RR	p Value	Lower bound	Upper bound	RR	p Value	Lower bound	Upper bound
Demographics
Age	0.040	.326	0.961	1.125	1.719	<.001	1.538	1.920
Sex: women vs men	1.040	.668	0.867	1.245	1.089	.558	0.815	1.455
Race
Black/African American vs white/Caucasian	1.001	.996	0.763	1.312	0.551	.003	0.374	0.811
Other vs white/Caucasian	1.203	.209	0.899	1.611	0.278	<.001	0.142	0.541
Education	0.991	.880	0.885	1.111	1.274	.001	1.108	1.465
Living alone vs not living alone	1.258	.073	0.978	1.618	1.571	.004	1.160	2.129
Medicaid eligible: yes vs no	1.067	.699	0.763	1.492	1.105	.648	0.715	1.708
Length of residence	0.963	<.001	0.956	0.969	0.992	.019	0.986	0.999
Annual total income	0.937	.146	0.858	1.024	0.915	.254	0.783	1.068
Health
Self-rated healthb	1.155	.008	1.040	1.282	1.036	.528	0.927	1.158
ADL difficultya	1.058	.257	0.958	1.168	1.078	.139	0.975	1.191
Dementia
Probable dementia vs no dementia	0.756	.151	0.514	1.111	2.510	<.001	1.662	3.788
Possible dementia vs no dementia	1.090	.564	0.809	1.469	1.596	.033	1.039	2.450
Fallen last year: yes vs no	1.086	.399	0.894	1.320	1.237	.115	0.948	1.615
Hospital admission last year: yes vs no	1.484	.005	1.135	1.939	1.779	<.001	1.309	2.417
Housing
Indoor accessibility
Single floor vs multiple floors	0.580	.002	0.412	0.815	1.154	.536	0.727	1.833
Multiple floors but lift, stair glide or BBKc on same floor vs multiple floors	0.499	<.001	0.350	0.710	0.843	.508	0.506	1.407
Entrance accessibility
No step entrance vs steps at entrance	0.921	.439	0.747	1.137	0.846	.316	0.608	1.920
Steps at entrance but ramp vs steps at entrance	0.892	.543	0.615	1.295	0.736	.264	0.428	1.268
Bathroom safetyd	0.985	.687	0.914	1.061	1.087	.105	0.982	1.202
Housing: apartment vs house	1.105	.559	0.787	1.550	1.446	.061	0.982	2.128
Housing conditione	0.991	.841	0.903	1.087	0.905	.251	0.762	1.075

Note: Multinomial logistic regression model, with weights. The reference category is non-movers. ADL = activities of daily living; CI = confidence interval; RR = relative risk ratio.

^a0–5, higher means more dependence.

^b1–5, higher means better health (reversed).

^cBBK = bathroom, bedroom, and kitchen on the same floor.

^d0–6, number of present home modifications and safety features.

^e0–4, higher means worse housing condition.

For moves to residential care, increasing age, higher education, living alone, hospital admission in the past year, and probable and possible dementia all increased the likelihood of a move to residential care. Not being white/Caucasian and having a longer length of residence in the current dwelling decreased the likelihood of moving.

No home environmental factors were statistically significantly associated with moves to residential care (see Table 3).

Our subgroup analysis comparing movers to nursing homes and movers to other forms of residential care (assisted living facilities and independent living settings) showed that home environmental factors had no statistical significance in the model. Instead, lower income and lower self-rated health increased the odds of moving to a nursing home (odds ratio [OR] = 0.705, 95% CI = 0.526–0.944, p value = .020 and OR = 0.754, 95% CI = 0.611–0.932, p value = .010, respectively; see Supplementary Material Online).

Discussion

This study examines the relationship between aspects of the physical home environment and relocations either within the community or to residential care for community-living older adults. The findings show that the home environment, specifically, indoor accessibility, is associated with relocation to another living arrangement within the community. Specifically, older adults living in dwellings with more than one floor, without access to a lift or stair glide, or without the kitchen, bedroom, and bathroom on the same floor are more likely to move within the community. The association between indoor accessibility and a move within the community remained after adjusting for well-established relocation predictors such as ADL difficulties, increased age, cognitive limitations, and living alone (Gaugler et al., 2007; Luppa et al., 2010; Miller & Weissert, 2000). Home environment was not, however, found to be related to relocation to residential care.

Our study is the first using a U.S. nationally representative sample to show that poor indoor accessibility is associated with relocation within the community. Our results align with theoretical assumptions on P–E dynamics. They suggest that relocation within the community may be a proactive strategy to enhance congruence between the older adult’s abilities and their living environment (Golant, 2011; Lawton, 1989). As such, relocation to another dwelling in the community can be viewed as an approach to reduce disablement.

Our results suggest that dwellings without stairs or dwellings with lifts, or with stair glides, or with bathroom, bedroom, and kitchen on the same floor can support older adults to age-in-place. Home environments designed to accommodate or modify age-related functional limitations could be highly significant, given that a recent study estimates that 50% of all community-living older adults in the United States have difficulties performing basic or IADLs (Freedman & Spillman, 2014). Home environments designed to accommodate or modify age-related functional limitations thus appear to be significant.

As stated, we found no home environmental factors to be associated with relocation to residential care. This is in contrast to the study by Stineman and colleagues (2012) where unmet need of home modifications was associated with moves to a nursing home. In our study, home modifications were designated as the presence of bathroom safety features, rather than unmet need. Alternatively, more detailed measures of the P–E fit might be needed to capture the role of the home environment in relocation to residential care. European studies using the Housing Enabler to measure household accessibility in terms of fit between a person’s functional abilities and 188 environmental factors in and around the home suggest that this could be the case. In those studies, a higher score, reflecting more household accessibility problems, increased the likelihood of moving to residential care (Granbom et al., 2014, Iwarsson et al., 2016).

Also, it is likely that measures capturing the emotional attachment to home are needed to understand the complexity of relocation decisions better. Our results on length of residence showed that the longer a person had lived in their current home, the less likely they were to move at all. They are in line with recent conceptual models on aging in place that emphasize that older adults’ agency and belonging in the home influence the desire to stay (or move) (Wahl, Iwarsson, & Oswald, 2012). Length of residence has been used as a proxy for home attachment and housing satisfaction (Lewicka, 2011). Further research is needed to investigate the differences between objective and perceived measures of the home environmental features and how they trigger relocation. Constructs such as usability, home attachment, and housing-related control beliefs could be valuable for exploring these issues in nationally representative relocation studies (see e.g., Oswald et al., 2006).

Our study highlights that relocation within the community and to residential care constitutes different types of moves triggered by different factors. It is unusual to compare moves to different destinations in the same analytic model (van der Pers et al., 2017). By doing so, our results strengthen findings from qualitative research showing that demographics, the home environment, and underlying health and care needs can be very different for moves within the community and to residential care (Roy et al., 2018).

Our study also showed interesting differences between those who move within the community and those who did not move, despite having similar sociodemographics. The movers within the community were more likely than the non-movers to have worse indoor accessibility and rate their health better but were more likely to have been admitted to a hospital. These seemingly contradictory results suggest that even when older adults perceive their health as good, living in an inaccessible dwelling could pose challenges in the future. Hospital admission may make older adults and their families review their living situation and begin to recognize that their homes are becoming increasingly challenging (Pope & Kang, 2010). Further, it has been proposed that to move, older adults need to view such a move as manageable, that is, that they have the resources and strength to enact the move (Golant, 2011). If not, they are more likely to stay in place despite increasing challenges in day-to-day life in the home. Unfortunately, in this study as in others, these assumptions have not been explicitly measured.

The sociodemographic and health-related differences between non-movers and movers to residential care in our study were substantial and consistent with previously reported predictors for relocation to residential care (Gaugler et al., 2007; Luppa et al., 2010; Miller & Weissert, 2000). The movers to residential care were older, lived alone to a greater extent, were more educated, more often white, and had been admitted to hospital more often than the non-movers. ADL difficulty was not statistically significantly associated with moves to residential care or to the community, which was likely the result of including several additional health-related factors in the adjusted model. Our unadjusted analysis agreed with previous research on predictors for relocation to residential care.

In our study, bathroom safety features were not significantly associated with any type of relocation. This was somewhat surprising considering that safety features such as grab bars, raised toilet seats, and bath seats have been shown to be associated with the ability to age in place in previous studies (Kendig et al., 2017). Indoor accessibility problems such as living on more than one level or not having access to a lift may be more challenging to address and also financially out of reach than more common modifications such as installing grab bars and second banisters. Indoor accessibility problems may, therefore, be more likely to trigger a move to another dwelling. Because safety features are likely correlated with several health factors and sociodemographic factors (Meucci, Gozalo, Dosa, & Allen, 2016), we also ran an adjusted model without safety features. It did not significantly alter any associations. As with relocation, installation of safety features could be a proactive or reactive strategy to manage ADLs and day-to-day activities. Considering that a substantial portion of the U.S. housing stock has problematic home environmental features (U.S. Department of Housing and Urban Development, 2015), it is important to improve building legislation and policies (Pynoos, 2018). In addition, to promoting aging in place and avoiding unwanted relocations in the existing housing stock, we need to improve the availability of innovative home modification services.

Looking closer at moves to different types of residential care, we found no housing factors that predicted relocation to nursing homes compared to other forms of residential care (assisted living facilities and independent living settings). However, our results confirm that older adults who move to assisted living facilities and independent living settings are better off health wise and financially than older adults moving to nursing homes. Surprisingly, we found no associations on race/ethnicity and relocation as we did when comparing residential care moves to non-movers. Our findings align with the suggestion that previously shown differences by race/ethnicity in access to different residential care settings might be changing (Feng, Fennell, Tyler, Clark, & Mor, 2011).

It should be noted, 3.9% of the sample moved to residential care within 4 years, which would be in the lower range of what Sergeant and colleagues reported in 2008 (5%–30% in 5 years). However, NHATS includes a data set that is nationally representative and the lower rate found in this data set may be more accurate.

A limitation of our study is that we did not account for reasons for moving, which could be valuable in further studies investigating decision making and effects of relocation in later life. By using a prospective design, we took a first step toward exploring associations between the home environment and relocation. NHATS provide longitudinal data from yearly data collections beginning in 2011. Thus, in future studies, information from the NHATS study on personal reasons for relocation (e.g., move to follow family members) or financial reasons (unexpected change in financial situation) and whether or not movers relocated to home environments that better accommodated their health care needs, can add important knowledge to the field.

Conclusions

Indoor accessibility is associated with residential mobility of community-dwelling older adults, however, only for moves from one home to another within the community and not to residential care facilities. One-floor dwellings, access to a lift, or having kitchen, bedroom, and bathroom on the same floor may help older adults age in place. The results of our study call for more exploration of the role of modifiable home environment factors and late-life relocation but also have practical implications for developing policies and programs to help older adults aging in place, as well as relocation counseling and societal planning for older adults.

Funding

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. The current study was completed at the Center for Innovative Care in Aging at Johns Hopkins University School of Nursing. M. Granbom is supported by the Swedish Research Council FORMAS (942-2015-403), The Crafoord Foundation, Sweden (20160604), and The Helge Ax:son Johnsons Foundation, Sweden.

Conflict of Interest

The authors declare that they have no conflict of interest.

Author Contributions

M. Granbom planned the study, contributed to the statistical analysis, and wrote the article. N. Perrin performed the statistical analysis and contributed to revising the article. S. Szanton, T. Cudjoe, and L. N. Gitlin helped to plan the study and to revise the article.