Patterns of Shelter Use Among Men New to Homelessness in Later Life

Abstract

People who become homeless for the first time in late life are a growing but understudied population. This study draws on administrative data from one shelter (N = 1,214 first-time homeless) to assess the extent to which age is related to shelter stay and, to examine psychosocial factors that may be associated with shelter departure. Our bivariate and survival analysis results suggest that older homeless men stay in the shelter 2 weeks longer than younger clients. Older men with pending legal issues and mobility concerns were more likely to leave the shelter than those without such concerns. Findings highlight the impact of age and other psychosocial variables on shelter stay, and provide direction from which to address homelessness among men who are new to homelessness in later life.

Keywords

homeless men social services

Despite the anticipated growth of homelessness in late life, especially for men in Organization for Economic Co-Operation and Development (OECD) countries (Crane et al., 2005), limited research exists. Within this small body of emerging literature (Grenier, Barken, Sussman, Rothwell, & Lavoie, in press), only a handful of researchers have focused on the trends and typologies of older people who find themselves homeless for the first time in late life (Petersen & Parsell, 2015). Yet, adults facing homelessness for a first time in older age are reported to have different experiences and service needs than those who are aging in situations of chronic homelessness (Aubry, Farrell, Hwang, & Calhoun, 2013), and such differences can be important where policy and community-based planning are concerned. This article analyzes how age and psychosocial vulnerability work together to affect patterns of shelter use among men who use the shelter system for the first time (i.e., first-time homelessness) to inform policies and practices for older people who are homeless.

Prevalence and Vulnerabilities Experienced by Older Homeless Adults

Homeless adults are typically considered “older” in their early to mid-50s as they tend to show physical and cognitive signs of aging approximately 10 years earlier than non-homeless adults (Gonyea, Mills-Dick, & Bachman, 2010; Ploeg, Hayward, Woodward, & Johnston, 2008). Recent estimates suggest that older homeless adults represent between 10% to 15% of the total homeless population in the United States (Gonyea et al., 2010) and that 10% of shelter users in Canada are older homeless adults (McDonald, Dergal, & Cleghorn, 2004).

Individual vulnerabilities associated with either becoming or remaining homeless in later life (age 50 and over) include substance abuse, mental health issues, legal problems, and life events such as widowhood, marital breakdown, eviction, and physical health problems (Garibaldi, Conde-Martel, & O’Toole, 2005). It is noteworthy that two of the life events identified as triggering homelessness—health problems and widowhood—increase in prevalence in advanced age (National Institute on Aging, 2011). It is also noteworthy that individual risk factors that are associated with becoming or remaining homeless in old age can be exacerbated by structural forces such as shortages of available low-income housing for older adults with health needs, substance abuse programs that target younger cohorts, and ageism (Watson, George, & Walker, 2008).

Prevalence and Vulnerabilities of Older Homeless Men

The majority of older homeless adults are reported to be men. Estimates from England, the United States, and Australia suggest the proportion of total homeless who are male ranges between 63% and 92% (Crane et al., 2005; Crane & Warnes, 2010). What is problematic from a service perspective is that the average duration of homelessness appears to be higher among older homeless men than homeless adults in general, and older homeless women in particular (Hecht & Coyle, 2001; North & Smith, 1993).

Although many of the individual vulnerabilities associated with entering or remaining homeless appear similar for men and women (McDonald, Dergal, & Cleghorn, 2007; Rich & Clark, 2005), there are some important gender differences. Older men’s homelessness is most often attributed to loss of employment, severity of mental illness, substance abuse, and criminality or legal problems (Kim et al., 2010; Peressini, 2007). Furthermore, men (and especially those new to homelessness in later life) appear to be particularly affected by a rapidly changing and globalized economy. For example, labor market changes in the steel, automobile production, and pulp and paper industries have resulted in a substantial decline in weekly earnings over the past half century and/or permanent job loss (Fortin, Green, Lemieux, Milligan, & Riddell, 2012). Within this structural context of growing economic precarity, a rise in men who are new to homelessness in later life is expected (Crane & Warnes, 2010).

Newly Homeless Older Adults

Only a handful of researchers have examined the needs and experiences of older people who are newly homeless. Yet, all of these point to the significance of age and the complexity of homelessness in later life. Findings based in the United States, Canada, the United Kingdom, and Australia flag a number of possible issues. Caton and Colleagues (2005) conducted a longitudinal study with 377 men and women new to the shelter system in New York City. Their findings revealed that the median number of days in homelessness was greatest among the “older” group (age 44+) suggesting they were at heightened risk for becoming chronically homeless. In Toronto, Canada, McDonald and colleagues (2007) conducted structured interviews with 68 older homeless persons to compare differences in profiles and service utilization of those who became homeless prior to and after age 50. Their findings suggested that older adults new to homelessness feel less safe in shelters, less informed about available services, and prefer informal assistance to formal assistance than adults who have grown old homeless. In the United Kingdom and Australia, Crane et al. (2005) and Petersen and Parsell (2015) conducted studies on pathways into first-time homelessness in later life, with samples of 377 and 561, respectively. They found that eviction, relational breakdown, monetary challenges with rent, inability to continue living with family and inaccessible rental housing could lead to late-life homelessness.

The combined findings suggest that persons who became homeless for the first time in later life (generally 50+, although some studies have used other thresholds) have different pathways into homelessness, barriers to exiting homelessness, and service needs than both newly homeless adults whom are younger, and older homeless adults with experiences of homelessness across the life course. In addition to age, psychosocial complexity among older adults who are homeless seems to emerge as a key consideration for policy and planning. Although these studies make important and significant contributions to understandings of homelessness, they are based on relatively small sample sizes (McDonald et al., 2007), self-selected participants (Crane et al., 2005; McDonald et al., 2007), non-standardized measures (Crane et al., 2005), inconsistent operationalization of “older” homelessness (e.g., Caton et al., 2005, used 44+, others used 50+ Gonyea et al., 2010: Grenier et al., in press), and for the most part are descriptive rather than explanatory (Crane et al., 2005; McDonald et al., 2007).

Our study aims to advance knowledge in four ways. First, our longitudinal sample is considerably larger than previous work on first-time homelessness in later life. Our sample includes 1,214 homeless men who were followed through the shelter system for 6 months. Second, our study includes all male clients served by the agency, not only those who self-selected to participate in the study, which would suggest greater generalizability of results. Third, we use a standardized and validated measure to document psychosocial vulnerability. Fourth, for the first time, we examine shelter departure in addition to length of stay. Our rationale is that factors associated with length of stay (a continuous measure of days in a shelter) may differ from those associated with risk of departure (a dichotomous outcome of either saying or going on any given day). Exploring both outcomes allows for a more comprehensive portrait of the psychosocial issues associated with shelter use and departure. Our study was guided by the following research questions and hypotheses:

Research Question 1: Do men who use a shelter for a first time in later life (50+) stay longer than men who use a shelter for a first time in young to mid-adulthood (younger than age 50)?

Hypothesis 1: We hypothesize that older clients stay longer in the shelter and are less likely to depart from shelter than younger clients.

Research Question 2: What are the patterns of departure among the older men and what factors are associated with departure?

Hypothesis 2: We hypothesize that psychosocial factors associated with shelter stay and departure will differ for older and younger clients.

Method

Setting and Context

This study used administrative data gathered in partnership with a major homeless service provider in Montreal, Canada. Montreal is a large urban city with a diverse metropolitan population of 3.8 million people, and like many large urban centers, has a fairly large and diverse homeless population. Although accurate measures of individuals occupying positions on the spectrum of homelessness are unavailable for Montreal, a 1996/1997 survey found that approximately 28,214 people in Montreal used shelters, soup kitchens, and stay centers for homeless people. Of these respondents, 12,666 indicated that they lacked a permanent fixed address in the previous year (Chevalier & Fournier, 1998).

Canada’s response to homelessness is premised on the Housing First (HF) approach, as outlined in the 2014 renewal of the Homeless Partnering Strategy of the federal government (Employment and Social Development Canada, 2013). HF aims to provide immediate housing to persons identified as chronically or recurrently homeless. Once housed, other forms of support such as linkages to mental health, substance abuse, or community services are provided (Tsemberis, Gulcur, & Nakae, 2004). Canada’s adoption of HF is based on At Home/Chez Soi, a nationwide demonstration project operating between 2008 and 2013 that compared HF with traditional service approaches in the five cities of Montreal, Toronto, Winnipeg, Moncton, and Vancouver. As in HF, the At Home/Chez Soi model specifically targets people who are chronically and episodically homeless (Goering et al., 2014). The shift toward HF for chronically or episodically homeless persons, suggests that shelters can be expected to continue to be the first-stop for the first-time homeless who are not the target population of the HF strategy.

The service provider (hereafter referred to as the agency) where our research was carried out has a number of pavilions and apartments that correspond with programming. It is the largest homeless shelter for men in the province of Quebec and the largest homeless shelter for women in Canada. In 2013, the agency delivered over 269,000 meals and provided some 118,000 overnight stays in its emergency shelters and transition units. Programs were provided as part of transitional programming which offered daytime access to the shelter, an assigned bed for the duration of their stay, as well as an assigned case manager. Transitional programs have been designed to provide a maximum of 6 months of services, but in some cases serve a clientele that are more chronically homeless, some of whom are older men.

Participants

The study population included adult men who experienced first-time homelessness, which was defined as the first time the person had presented themselves for service at this shelter. Upon admission, workers entered data into the Homeless Individuals and Families Information System (HIFIS) software system. HIFIS is part of Canada’s National Homelessness Information System (NHIS)—a federal data development initiative designed to collect and analyze baseline data on the use of shelters in Canada. For this study, the research team worked with the service provider to extract the data from the complex relational database. Files were merged and variables recoded to develop the database for analysis. Our sample was restricted to those who stayed at the men’s shelter for the first time between July 2009 and September 2011. Over this time period, 1,214 men enrolled in the transition programs at the agency. The purpose of the agency’s two transitional housing programs is to provide the support needed for clients to re-establish independent living. The first program named Etape is geared toward stabilization and accessing eligible benefits; the second program named Escale provides autonomous living arrangements inside the shelter. Importantly, participants in the Escale program are required to pay rent.¹ The age range of these 1,214 men was 18 to 76. The study received ethics approval from McMaster University and the McGill University Research Ethics Board (File No. 235-1210).

Measurement

The data analyzed for this study were derived during the intake and initial enrollment at the shelter. At the agency, the general process of enrollment involves meeting with a case manager. The first meeting includes a basic psychosocial assessment where the client and worker discuss the cause(s) of the homeless condition, current and future needs, and plans to end their homelessness.

Psychosocial Vulnerability

During the initial meeting, the client and caseworker complete a modified version of the Arizona Self-Sufficiency Matrix (SSM; Culhane, Parker, Poppe, Gross, & Sykes, 2007). The SSM is an assessment tool that is used to understand the complex psychosocial-economic issues experienced by homeless people. The instrument captures many of the dimensions found to be associated with homelessness for adult men and/or older adults (Crane et al., 2005; Crane & Warnes, 2010; Peressini, 2007); thus, allowing us to get a detailed portrait of the psychosocial factors that may affect their trajectories. Our study used 10 items from the original matrix: income, education, legal issues, life skills, mental health, substance use, family, mobility, community involvement, and safety, with the agency opting to omit items such as food and housing that were not considered relevant in a shelter environment. With regard to aging, the life skills element is akin to the ability to perform activities of daily living—a level of functioning that is typically used to establish older adults’ eligibility for gerontological services (Huxley, Evans, Munroe, & Cestari, 2008). Community involvement can be considered an operationalization of social engagement, a commonly cited protective factor for “aging well” (Gray, 2009). Mobility represents access to viable and appropriate transportation, one of the most commonly identified concerns for older people in Canada and the United States (Turcotte & Schellenberg, 2007; U.S. Department of Transportation, 2013).

Each item of the SSM contains five ordinal responses. The items are scored from 1 to 5, with 1 indicating more severe problems and 5 indicating higher functioning. For more information on scoring, see Supplementary Appendix A. The SSM has demonstrated reliability in another large sample (Cronbach’s α = .81; Culhane et al., 2007). A correlation matrix of the SSM variables is provided in Supplementary Appendix B. To our knowledge, there is limited research on the scale consistency across dimensions in the SSM. In other words, it is not clear whether an increase from a score of 2 to 3 on one dimension (e.g., education) is consistent to the same increase on another dimension (e.g., legal). As such, we took a conservative approach to treat the SSM as vulnerability by creating a dichotomous cutoff. To score the SSM as a vulnerability index, we followed the guidelines for the SSM (Abt Associates, 2005) and established a dichotomous vulnerability cutoff score of 2. A score of 2 or less on the original scale indicates severe vulnerability on the measured dimension. A score greater than 2 on the original scale indicates less vulnerability which we coded as 0 for absence of vulnerability. Furthermore, we validated this vulnerability threshold with shelter staff who confirmed the appropriateness of the cutoffs. Using mobility as an example, if a client scored 1 (“No access to transportation, public or private”) or 2 (“Transportation is available, but unreliable, unpredictable, unaffordable; may have car but no insurance, license, etc.”) on the original scale, they were considered vulnerable on this item and given a score of 1. If they scored 3 (“transportation is reliable and available”), 4 (“basic travel needs are met”), or 5 (“in addition to 4, transportation is affordable and adequately insured”) on the original scale, they were considered to be absent of vulnerability on this item and given a score of 0.

Operationalizing Age

Based on a growing consensus in research on aging and on homelessness, we used the age of 50 as a marker of homelessness among older people (Gonyea et al., 2010: Grenier et al., in press). Age at first stay in the shelter was transformed into a binary variable, generating two groups: those 50 years of age or greater, and those 49 and under.

Covariates

In addition to the SSM variables and age, we extracted both language (French, English, and Other) and immigrant status from the HIFIS system. Language was coded as French = 1, Other = 0. Immigrant was coded as 1 = immigrant and non-citizen; 0 = Canadian citizen. Data was not collected on race/ethnicity.

Dependent Variables

We isolated two dependent variables that would help us to better understand patterns of service use. The first dependent variable was total nights spent in shelter during the first 6 months of a client’s first stay. Length of stay is important to identify the profiles of men that tend to stay longer and might therefore benefit from longer term more complex intervention at the shelter. Other types of men tend to leave more quickly and may benefit from different services such as information about the local homeless and housing systems. We ran analyses with shorter (3 months) and longer time stay lengths (9 and 12 months), but for programmatic and empirical reasons the 6 months window was the most appropriate option. Overall, 95% of cases with complete data exited the shelter within 180 days.

The second dependent variable, shelter departure, was used in the survival analysis. Homeless clients depart from the shelter for many reasons. Some men find independent housing, some transition back to a previous housing arrangement, others may be barred from the shelter, and some leave for unknown reasons. There are also a number of people who return to the shelter at a later point in time, or continue to use services through other programs such as the transitional programs or community supports. The episodic and cyclical nature of homelessness is well documented, but understanding length of stay and departure could play an important role in refining the types of services people receive, and when, so that they are better able to secure long-term housing and avoid falling into chronic patterns of homelessness and shelter use.

Analytical Plan

To address our research questions, we compared bivariate means between older and younger homeless clients across the SSM items. Wilcoxon signed-rank test for differences in continuous means (number of days) and Pearson’s chi-square tests for differences in proportions (SSM variables) were run to understand differences between the two age groups. We then conducted a survival analysis using a Cox proportional hazard model. Survival analysis is a form of multivariate regression that is useful for time-to-event analysis (Kalbfleisch & Prentice, 2011). We were primarily interested in the relationship between the SSM variables and the likelihood of exiting the shelter on a given day. In addition to age and SSM variables, language and immigration status were included as covariates. The model can be written as,

[H (t) = H_{0} (t) \exp (β_{1} X_{1} + β_{2} X_{2} + β_{3} X_{3} \dots \dots + β_{k} X_{k})]

Where $H_{0} (t)$ is the baseline hazard of leaving the shelter at time t and X₁ . . . X_k represents a set of independent variables that are hypothesized to be related to shelter departure. After dividing each side of the equation by $H_{0} (t)$ and taking the natural log, the dependent variable becomes $H (t) / H_{0} (t)$ , that is, the risk of leaving the shelter. Coefficients $β_{1} \dots β_{k}$ were estimated with Cox proportional hazard regression in Stata 13. Estimates are presented as hazard ratios (HRs) which represent the risk of leaving the shelter for each independent variable X₁, at any time, holding the other covariates constant. Standard errors were estimated with 100 bootstrap replications. Survival analysis is useful here because the procedure allows us to account for right-censored information on stay length. In our study, those who did not leave the shelter within 180 days were censored in the analysis. In the Cox proportional hazard model, clients were considered at risk of shelter departure until they left the shelter or were censored. Clients who were “censored” remained at the shelter and continued to receive services after the study time period (180 days in this study) expired.

Results

In response to the first research question, we found that older men (n = 301) had statistically significant longer stays (p < .01) when compared with younger men, F(1, 212) = 38.21 (see Table 1). The median length of stay of the older group was 39 days compared with a median length of stay of 25 days in the younger sample (mean stay length was 63 [SD = 60.21] for older, compared with 41 [SD = 47.52] for younger). Table 1 shows that immigrants comprised a greater proportion of the younger group (15%) compared with the older group (7%). We see several differences across age groups for vulnerability. First, younger men with an income vulnerability rate of 79% were statistically significantly higher than the older men group at 66% (χ² = 18.03, p < .01). Second, there was a statistically significant difference across age groups in the proportion of men who experienced community vulnerability on the community involvement dimension of the scale (37% for the older group compared with 28% for the younger group).

Table 1.

Description of the Sample by Older and Younger Clients.

	n	Younger (n = 913)	Older (n = 301)	Test statistic
Days in shelter	1,214	25	39	38.21***
Age	1,214	37	55	1,541.29***
French language	1,214	55%	54%	0.12
Immigrant	1,207	15%	7%	14.48***
Vulnerability type
Income	1,207	79%	66%	18.03***
Education	1,210	17%	21%	1.9
Legal	1,209	13%	13%	0.03
Life skills	1,211	35%	35%	0.01
Mental health	1,212	12%	13%	0.16
Substance use	1,209	18%	20%	0.47
Family	1,209	61%	59%	0.33
Mobility	1,209	38%	35%	0.91
Community	1,201	28%	37%	8.46**
Safety	1,204	19%	22%	1.63

Note. Median reported for continuous variables. n is the number of non-missing values. Tests used: Wilcoxon test for continuous variables; χ² Pearson test for categorical variables.

**p < .01. ***p < .001.

Patterns of Shelter Use

The finding that older men stay between 2 and 3 weeks longer than younger men demonstrates that the homeless experience for older adults is different than younger adults. However, our first analysis was only bivariate and did not control for multiple influences. To better understand and target interventions, we were interested in exploring how age might be related to shelter stay after holding psychosocial vulnerabilities constant. Thus, the next step was to examine shelter patterns in a multivariate framework. One option was to estimate an OLS linear regression model predicting total days spent in the shelter. We found that older age was related to a longer stay length of about 23 days (b = 22.89, SE = 3.75, t = 6.10, p < .001). However, this approach was problematic for at least two reasons. First, the distribution of the total days in shelter was strongly positively skewed and thus violated the assumption of normal distribution for the dependent variable. Second, in this OLS approach, clients who never left the shelter were observed as staying for 180 days, despite actually staying longer than 180 days and perhaps indefinetly. Results from models examining stay length are not presented here but available upon request.

As a stronger approach, we exploited the time dimension of the data and focused on the likelihood of shelter departure on a given day. The key advantages of survival analysis include maximizing the time-to-event nature of the data and inclusion of censored data in the analysis. A list-wise deletion of cases with items missing resulted in a loss of n = 181 observations. Because our primary interest was in understanding differences in shelter departure patterns between the older and younger men, we first tested the equality of survivor functions across age groups. The log-rank test was statistically significant (χ² = 26.95, p < .01) meaning that the survival distributions of the two groups were different. Figure 1 shows the Kaplan–Meier survival estimates over time across older and younger men. At no point in the study time frame did the 95% confidence intervals for age groups overlap. Our results from both the statistical test and the graphical presentation reveal that, across time, older men had a statistically significant lower likelihood of leaving the shelter—a finding that is extremely important in terms of recognizing the challenges that may exist for older men, and adapting services to better meet their needs.

Figure 1.

Graph of the survivor function for shelter stay across older and younger homeless male clients.

A cox proportional hazard was then used to model shelter departure within 180 days using client age and the 10 SSM indicators of vulnerability as covariates. The dependent variable time to shelter departure had an average of 47 days (SD = 51.61); median = 29. The model (see Table 2) was statistically significantly different than a null model (χ² = 59.26, p < .01). Visual plots using the Schoenfeld and scaled-Schoenfeld residuals as well as statistical tests confirmed the proportionality assumption was upheld (χ² = 22.45, p < .05).

Table 2.

Cox Proportional Hazard Model of Shelter Departure.

Variable	Hazard ratio
Older	0.654***
	[0.554, 0.771]
French language	1.125
	[0.997, 1.269]
Immigrant	0.955
	[0.806, 1.132]
Vulnerability type
Income	0.828*
	[0.686, 0.999]
Education	1.031
	[0.861, 1.235]
Legal	0.914
	[0.712, 1.174]
Life skills	1.036
	[0.898, 1.195]
Mental health	1.026
	[0.824, 1.277]
Substance use	1.427***
	[1.229, 1.658]
Family	1.155*
	[1.021, 1.306]
Mobility	1.030
	[0.886, 1.198]
Community	0.865
	[0.743, 1.006]
Safety	1.103
	[0.926, 1.315]

N	1,021

Note. Total number of events = 1,021 and censored = 56, resulting in 5% censored. Standard errors estimated from 100 bootstrap replications. Exponentiated coefficients; 95% confidence intervals in brackets.

p < .05. **p < .01. ***p < .001.

Several observations are noteworthy in Table 2. On any given day, older men, compared with younger men, were less likely to leave the shelter (expressed as a HR this quantity was .65 [p < .01]). The HRs are interpreted as ratios of incidence rates. Therefore, on any given day, .35 as many older adults departed proportionally to the younger group. Three SSM variables were related to shelter departure, in both positive and negative directions. Men with income vulnerability were much less likely to leave the shelter on any given day—an important finding where policy and programming are concerned. To be specific, men with income vulnerability were at a 17% reduced risk for shelter departure (HR = 0.83). On the contrary, two types of vulnerability—substance use and family—were associated with an increased risk of shelter departure. Substance abuse vulnerability had the strongest association (HR = 1.43). This is not surprising, as actively using substances while seeking services is not permitted. Men with family vulnerability were also more likely to depart (HR = 1.16), suggesting that clients with stronger social support are perhaps more likely to engage with services at the shelter. A sum vulnerability variable was tested but was not statistically significant and did not change the substantive meaning of the Table 2 results (results not shown, available upon request).

To more explicitly test our second hypothesis, we built on Table 2 results by examining interactions between age and vulnerabilities. These analyses helped us understand the types and magnitudes of risk of shelter departure specifically for the older homeless group. An additional 10 models were analyzed that included all combinations of age and vulnerability indicators (results not shown but available upon request). Table 3 reports only the interactions that met criteria for statistical significance (p < .05). Models displayed similar fit characteristics as the base model reported in Table 2. Table 3 Model 1 highlights the interaction between age and legal vulnerability. We see the relationship between older men and shelter departure was affected by legal vulnerability (HR = 1.59). In other words, older men with outstanding legal problems had sizably increased risk of shelter departure. Model 2 showed how the relationship between age and shelter exit was affected by mobility vulnerability (HR = 1.42 p < .05), with persons with mobility issues less likely to leave the shelter.

Table 3.

Cox Proportional Hazard Model of Shelter Departure With Selected Interactions.

Variable	Model 1	Model 2
Older	0.616***	0.583***
	[0.526, 0.721]	[0.493, 0.689]
French language	1.123	1.131
	[0.987, 1.277]	[0.997, 1.283]
Immigrant	0.957	0.956
	[0.796, 1.149]	[0.808, 1.131]
Vulnerability type
Income	0.828*	0.832*
	[0.701, 0.977]	[0.715, 0.969]
Education	1.038	1.033
	[0.840, 1.284]	[0.859, 1.244]
Legal	0.824	0.914
	[0.632, 1.073]	[0.738, 1.132]
Life skills	1.035	1.035
	[0.913, 1.174]	[0.898, 1.193]
Mental health	1.038	1.039
	[0.847, 1.273]	[0.848, 1.274]
Substance use	1.436***	1.430***
	[1.220, 1.691]	[1.183, 1.728]
Family	1.159*	1.153*
	[1.013, 1.327]	[1.000, 1.328]
Mobility	1.032	0.954
	[0.891, 1.195]	[0.808, 1.126]
Community	0.863*	0.863
	[0.749, 0.994]	[0.731, 1.018]
Safety	1.099	1.105
	[0.887, 1.360]	[0.921, 1.326]
Older × Legal	1.589*
	[1.070, 2.359]
Older × Mobility		1.422*
		[1.066, 1.897]

n	1,021	1,021

p < .05. **p < .01. ***p < .001.

Discussion

Summary of Main Findings

Our study results show that older men, who are new to homelessness, are at heightened risk of remaining in a shelter relative to their younger counterparts. More specifically, older men spend on average 2 weeks longer in a shelter and have a higher probability of remaining in that shelter each day than younger adults. Given the research design, the results provide some of the strongest evidence documenting a pattern of prolonged length of stay for older newly homeless men in North America.

Our study further suggests that particular psychosocial profiles significantly affect the pattern of shelter use among older men who use a shelter for the first time in old age. Older men with pending legal and mobility (transportation) issues face a sizably higher risk of departing the shelter on any given day than older men without these issues. Older homeless men with legal vulnerability—defined as outstanding or current legal involvement or noncompliance—may leave the shelter for a variety of reasons. It is plausible that clients with this vulnerability leave to address their legal concerns, including possible incarceration. Importantly, SSM measure does not distinguish between types of crime, for example, violent versus non-violent. The positive relationship between mobility vulnerability and age is somewhat counterintuitive: We would expect the opposite, that is, that older clients without transportation would be more likely to stay in the shelter until securing transportation. Montreal has a well-developed public transportation system of buses and metros. However, we speculate the system may be unaffordable for older homeless men. Single passes are now US$3.25 and the Transport Society of Montreal offers a relative 40% discount on monthly transit passes for persons aged 65 and older.

Substance abuse was a strong predictor of shelter departure in the overall model (Table 2), which is not surprising. Many shelters have no tolerance policies for substance use. Importantly, the interaction between substance abuse and age was not statistically significant—suggesting the statistically significant finding reported in Table 2 is driven mostly by younger clients. This is particularly interesting as both younger and older men have about the same rates of substance use vulnerability (18% for younger and 20% for older, see Table 1).

Overall, these issues or risk factors and their associated patterns of shelter use are unique to older men compared with younger men, suggesting that older men are a vulnerable yet diverse group with psychosocial issues that warrant specific policy and practice attention. Considering these findings in relation to the context of service gaps and policy issues leads us to identify both potential reasons for such patterns, but also important challenges in addressing the needs of older men who become homeless for the first time after age 50.

Policy and Practice Implications

Although our findings do not speak to service use beyond the shelter system, they hold important implications for how shelter workers and administrators may better predict older men’s likelihood of staying in a given shelter and respond accordingly. They also raise questions about the policy initiatives that do not include specific strategies for older adults, and approaches that uniformly shift funding from shelters or other emergency homeless services to stable housing. In particular, the most significant findings of legal and transportation vulnerability increasing risk of shelter departure for older men point to a need for particular types of support that are not usually integrated into HF such as access to low-cost transportation, subsidized legal services, and a case manager to link older adults new to homelessness to these services.

The increased length of stay noted in our study also suggests that shelter workers and administrators are pivotal resources for many older men new to homelessness as they are in a position to form relational connections that could support successful linkages to other required services. For older men with limited family and community connections, for example, shelter workers may actively facilitate connections with gerontological services such as case management and social and recreational activities to create a stronger support network (Gonyea et al., 2010), a protective factor for chronic homelessness (Gelberg, Linn, & Mayer-Oakes, 1990). This could be done by inviting representatives of publicly funded home care services, or not-for-profit community-based services into the shelter to meet older men before they depart, offering joint meetings and discussions about services, and providing some follow-up beyond the shelter to ensure connections have been made and sustained (Gilbody, Bower, Fletcher, Richards, & Sutton, 2006; Pauzé, Gagné, & Pautler, 2005; Shibusawa & Padgett, 2009). More than a system navigator, the shelter worker would also be positioned to help older adults new to homelessness grapple with the emotional reactions that may cause further distress and interfere with accepting assistance (e.g., feeling one is no longer successful; fearing that accepting ongoing help means one needs to accept a position of neediness; Snow and Anderson, 1993). Identifying and developing this type of relational connection requires that services providers in both sectors (the shelter system and gerontological services) receive training that challenges their views around aging, homelessness, and the intersection of the two (Crane & Warnes, 2007).

Our findings also suggest that not all older men who enter the shelter system can be expected to stay for long periods. Men with limited transportation options or pending legal issues may be expected to have shortened length of stays. We expect that these patterns do not suggest these men are exiting homelessness earlier, but rather have complexities that shelters may be less able to address. In these instances, we suggest immediate targeted information and referral upon shelter entry to the combination of legal services for low-income adults, specialized legal services for older adults, viable housing options, transportation services, and community-based case management (also see McDonald et al., 2007). Although necessary, we expect these shorter term shelter-based interventions may be insufficient to avert chronic homelessness, especially for men with legal issues who face many challenges to re-housing (Caton et al., 2005; Davis-Berman, 2011). More work is needed to explore the unique experiences of this group of newly homeless older men and develop practice approaches specific to their needs.

Shifts in policies that govern homeless and gerontological services also appear warranted. HF policies that are now prevalent throughout North America, for example, aim to channel funding to long-term housing rather than the crisis response services provided through shelters and emergency health care (Tsemberis, Gulcur, & Nakae, 2004). Although funding for viable housing options for older adults with different psychosocial challenges is imperative (especially for older men with mobility limitations, mental health challenges, and criminal records) three issues emerge. First, long-term safe and affordable housing must be available for older people. Second, HF must be accompanied by the specific types of supports that are needed by older people. These could be related to health care, psychosocial needs, or adapted spaces. Third, policies and associated funding formulas should also ensure that shelters and other emergency community-based services are financially equipped to develop practices that encourage successful linkages to appropriate ongoing supportive services that can facilitate housing resettlement.

Improved coordination between shelter systems and gerontological services also requires a shift in eligibility criteria. Currently, eligibility is available for adults aged 60 to 65+ in most jurisdictions across North America. Our research findings affirm that around age 50 is a time in which age begins to play a role in homeless trajectories and may represent a more accurate threshold for accessing community services for older adults, including subsidized supportive housing, ongoing case management and social/recreational services (Grenier et al., in press; also see Gonyea et al., 2010; Ploeg et al., 2008).² Without such a change in eligibility for service provision, shared care between gerontological services and homeless shelters will not be possible for many at risk homeless men.

Perhaps most important, our research suggests that older men new to homelessness are a vulnerable yet diverse group who require unique strategies and services. National, State/Provincial/Territorial and Municipal policies on homelessness must make this group visible in their proposed frameworks and action plans. Quebec’s 2014 National Policy on Homelessness may represent an example of a policy that recognizes the complexities of homelessness in old age. Unique both within Canada and the United States, the policy identifies older adults as a vulnerable group and emphasizes the need for a continuum of housing and support services which could include more comprehensive supportive both within and beyond the shelter walls (see Barken et al., 2015). It will be important to monitor the extent to which the continuum of housing and support are realized in the province and to study the impact of the Quebec model on older adults new to homelessness over time.

Limitations

We acknowledge the methodological limitations of this study. First, the study does not measure changes over time in the SSM variables. We assume that SSM variables are time variant. Further, the findings are associational and do not indicate causality. With a large sample size, we assume the findings here have a reasonable level of external validity to other urban settings. However, the limited or unavailable data on the population of homeless men means that we cannot investigate this empirically. Our data do not include seasonal variations. This may or may not be influential as some research has pointed to seasonal effects in shelter stays (Culhane & Kuhn, 1998), while others have found the magnitude of seasons to be minimal (O’Flaherty & Wu, 2008). Last, we acknowledge that clients may have been homeless for some time before seeking services at the agency. Clients may have stayed at other shelters in Montreal or elsewhere. As such, it is possible that not all the clients who presented at this agency were homeless for the first time.

Implications for Research

Understanding the potential for a positive/“successful” departure, and the type of departure, is a critical next step in future research. It will be important to have accurate data on the available housing options for older people in particular regions, and to track the nature of the transition out of the shelter. Specifically, we want to know: did the client depart for independent living, another shelter, incarceration, hospitalization, continued homelessness, or other? Were they able to secure housing, and for how long? If not, what happened, and how can professionals intervene to ensure long-term housing? Future work can improve our understanding by exploring and testing underlying latent factors associated with shelter stay and other dependent variables of interest. It should also focus on differences that may exist in the trajectories into and out of homelessness, as experienced by groups at varying social locations (e.g., gender, education, ethno-cultural minority status, health/impairment, etc.; Grenier et al., in press). Income, precarious work and patterns of low pension contributions for example, may, in combination with psychosocial factors, provide insight into men’s homelessness in late life. The HIFIS system can play a major role in the next steps of research aimed to understand these trajectories. Discussions are underway between shelter provides in Montreal and other parts of Canada to explore how to coordinate and share data across shelters using the HIFIS platform.

Conclusion

We demonstrate that older men who become homeless in later life have unique psychosocial needs in relation to younger homeless men, and that these vulnerabilities impact shelter use. Understanding how different challenges—such as legal and mobility vulnerabilities—may impact tendencies to stay or leave the shelter system can provide service providers, administrators and decision-makers with important directions on when, where, and how to address the particular issues that render older homeless men vulnerable to remaining homeless for extended periods of time.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was conducted with support from the Social Sciences and Humanities Research Council of Canada (grant number 435-2012-1197).

Notes

Author Biographies

David W. Rothwell is an assistant professor at the College of Public Health and Human Sciences at Oregon State University. He conducts research on poverty, homelessness, and social development.

Tamara Sussman is an associate professor in the School of Social Work at McGill University and associated researcher at the Centre for Research and Expertise in Social Gerontology (CREGÉS). Drawing on more than 15 years of practice experience with older adults and their families, her program of research focuses on how health and social service systems affect older adults facing a variety of challenges including physical frailty, poverty, and mental health issues.

Amanda Grenier is an associate professor in the Department of Health, Aging and Society, Gilbrea Chair in Aging and Mental Health, and Director of the Gilbrea Centre for the Study of Aging (McMaster University, Canada). Her work explores the intersections of theory, policy frameworks, organizational practices, and lived experience in areas such as frailty, late-life transitions, exclusion, and older homelessness.

Sebastian Mott was previously the research coordinator for the Evaluation, Research and Improvement Project for Transition Services—a joint project between McGill University and the Old Brewery Mission. His research interests focus on homelessness, with particular attention paid to chronic homelessness, data, policy, and harm reduction approaches.

Valérie Bourgeois-Guérin is an assistant professor in the Department of Psychology at Université du Québec à Montréal (UQAM), and a psychologist and a researcher at the Centre for Research and Expertise in Social Gerontology (CREGÉS). Her researches explore aging, end of life, bereavement, and qualitative research.

References

Abt Associates. (2005). The National HMIS Data Standards and the Arizona Self-Sufficiency Matrix: Collecting Uniform Information on Homeless Person. Retrieved from http://monarchhousing.org/wp-content/uploads/2008/08/ProjectEvaluationTraining.pdf

Aubry

Farrell

Hwang

S. W.

Calhoun

(2013). Identifying the patterns of emergency shelter stays of single individuals in Canadian cities of different sizes. Housing Studies, 28(6), 1-18.

Barken

Grenier

Budd

Sussman

Rothwell

D. W.

Bourgeious-Guerin

(2015). Aging and homelessness in Canada: A review of frameworks and strategies. Hamilton, Ontario, Canada: Gilbrea Center for Studies in Aging, McMaster University. Retrieved from aginghomelessness.com/wp-content/uploads/2012/10/PolicyReview-FINAL-REPORT-April-2015.pdf

Caton

C. L.

Dominguez

Schanzer

Hasin

D. S.

Shrout

P. E.

Felix

. . . Hsu

(2005). Risk factors for long-term homelessness: Findings from a longitudinal study of first-time homeless single adults. American Journal of Public Health, 95, 1753-1759.

Chevalier

Fournier

(1998). Dénombrement de la clientèle itinérante dans les centres d’hébergement, les soupes populaires et les centres de jour des villes de Montréal et de Québec [A count of the homeless clients in shelters, soup kitchens and day centers in the cities of Montreal and Quebec]1996-97. Québec City, Québec, Canada: Santé Québec.

Crane

Byrne

Lipmann

Mirabelli

Rota-Bartelink

. . . Warnes

A. M.

(2005). The causes of homelessness in later life: Findings from a 3-nation study. The Journals of Gerontology, Series B: Psychological Sciences & Social Sciences, 60, S152-S159.

Crane

Warnes

A. M.

(2007). The outcomes of rehousing older homeless people: A longitudinal study. Ageing and Society, 27, 891-918.

Crane

Warnes

A. M.

(2010). Homelessness among older people and service responses. Reviews in Clinical Gerontology, 20, 354-363.

Culhane

Parker

Poppe

Gross

Sykes

(2007). Accountability, cost-effectiveness, and program performance: Progress since 1998. In Toward Understanding Homelessness. Retrieved from http://aspe.hhs.gov/hsp/homelessness/symposium07/culhane/

10.

Culhane

D. P.

Kuhn

R. H.

(1998). Patters and determinants of public shelter utilization among homeless adults in New York City. Journal of Policy Analysis and Management, 17, 23-43.

11.

Davis-Berman

(2011). Older women in the homeless shelter: Personal perspectives and practice ideas. Journal of Women & Aging, 23, 360-374.

12.

Employment and Social Development Canada. (2013). Homelessness partnering strategy directives 2014-2019. Retrieved from http://www.esdc.gc.ca/eng/communities/homelessness/funding/directives.shtml

13.

Fortin

Green

D. A.

Lemieux

Milligan

Riddell

W. C.

(2012). Canadian inequality: Recent developments and policy options. Canadian Public Policy, 38, 121-145.

14.

Garibaldi

Conde-Martel

O’Toole

T. P.

(2005). Self-reported comorbidities, perceived needs, and sources for usual care for older and younger homeless adults. Journal of General Internal Medicine, 20, 726-730.

15.

Gelberg

Linn

L. S.

Mayer-Oakes

S. A.

(1990). Differences in health status between older and younger homeless adults. Journal of the American Geriatrics Society, 38, 1220-1229.

16.

Gilbody

Bower

Fletcher

Richards

Sutton

A. J.

(2006). Collaborative care for depression: A cumulative meta-analysis and review of longer-term outcomes. Archives of Internal Medicine, 166, 2314-2321.

17.

Goering

Veldhuizen

Watson

Adair

Kopp

Latimer

Aubry

(2014). National At Home/Chez Soi final report. Calgary, Alberta: Mental Health Commission of Canada.

18.

Gonyea

J. G.

Mills-Dick

Bachman

S. S.

(2010). The complexities of elder homelessness, a shifting political landscape and emerging community responses. Journal of Gerontological Social Work, 53(7), 575-590.

19.

Gray

(2009). The social capital of older people. Ageing and Society, 29(01), 5-31.

20.

Grenier

Barken

Sussman

Rothwell

D. W.

Lavoie

J.-P.

(in press). Homelessness and aging in Canada: Setting a research agenda to inform policy and practice. Canadian Journal on Aging.

21.

Hecht

Coyle

(2001). Elderly homeless: A comparison of older and younger adult emergency shelter seekers in Bakersfield, California. American Behavioral Scientist, 45, 66-79.

22.

Huxley

Evans

Munroe

Cestari

(2008). Integrating health and social care in community mental health teams in the UK: A study of assessments and eligibility criteria in England. Health & Social Care in the Community, 16, 476-482.

23.

Kalbfleisch

J. D.

Prentice

R. L.

(2011). The statistical analysis of failure time data (Vol. 360). Hoboken, NJ: John Wiley.

24.

Kim

M. M.

Ford

J. D.

Howard

D. L.

Bradford

D. W.

(2010). Assessing trauma, substance abuse, and mental health in a sample of homeless men. Health & Social Work, 35(1), 39-48.

25.

McDonald

Dergal

Cleghorn

(2004). Homeless older adults research project: Executive summary. Toronto, Ontario, Canada: Institute for Human Development, Life Course and Aging. Retrieved from http://www.toronto.ca/legdocs/2005/agendas/committees/rse/rse050301/it005.pdf

26.

McDonald

Dergal

Cleghorn

(2007). Living on the margins. Journal of Gerontological Social Work, 49(1-2), 19-46. doi:10.1300/J083v49n01_02

27.

National Institute on Aging. (2011). Global health and ageing. Bethesda, MA. Retrieved from http://www.who.int/ageing/publications/global_health/en/

28.

North

C. S.

Smith

E. M.

(1993). A comparison of homeless men and women: Different populations, different needs. Community Mental Health Journal, 29, 423-431.

29.

O’Flaherty

(2008). Homeless shelters for single adults: Why does their population change? Social Service Review, 82, 511-550.

30.

Pauzé

Gagné

M. A.

Pautler

(2005, December). Collaborative mental health care in primary health care: A review of Canadian initiatives. Vol. I: Analysis of initiatives. Mississauga, Ontario: Canadian Collaborative Mental Health Initiative.

31.

Peressini

(2007). Perceived reasons for homelessness in Canada: Testing the heterogeneity hypothesis. Canadian Journal of Urban Research, 16, 112-126.

32.

Petersen

Parsell

(2015). Homeless for the first time in later life: An Australian study. Housing Studies, 30, 368-391. doi:10.1080/02673037.2014.963522

33.

Ploeg

Hayward

Woodward

Johnston

(2008). A case study of a Canadian homelessness intervention programme for elderly people. Health & Social Care in the Community, 16, 593-605.

34.

Rich

A. R.

Clark

(2005). Gender differences in response to homelessness services. Evaluation and Program Planning, 28, 69-81.

35.

Shibusawa

Padgett

(2009). The experiences of “aging” among formerly homeless adults with chronic mental illness: A qualitative study. Journal of Aging Studies, 23, 188-196.

36.

Snow

D. A.

Anderson

(1993). Down on their luck: A study of homeless street people. Berkeley: University of California Press.

37.

Tsemberis

Gulcur

Nakae

(2004). Housing first, consumer choice, and harm reduction for homeless individuals with a dual diagnosis. American Journal of Public Health, 94, 651-656.

38.

Turcotte

Schellenberg

(2007). A portrait of seniors in Canada, 2006. Ottawa, Ontario: Statistics Canada, Social and Aboriginal Statistics Division.

39.

U.S. Department of Transportation. (2013). Transportation statistics annual report 2012. Washington, DC: Research and Innovative Technology Administration Bureau of Transportation Statistics. Retrieved from http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/transportation_statistics_annual_report/2012/index.html

40.

Watson

D. P.

George

Walker

(2008). “Falling through the cracks”: Health care needs of the older homeless population and their implications. Research in the Sociology of Health Care, 26, 187-204.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB