Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Running head: SHELTER USE AT THE OLD BREWERY MISSION 1
MODELLING PATTERNS OF SHELTER USE AT THE OLD BREWERY MISSION:
DESCRIBING PROGRAM POPULATIONS, AND APPLYING A TYPOLOGY OF
HOMELESSNESS
A Research Report Submitted to
The School of Social Work
Faculty of Arts
Graduate and Post-Doctoral Studies Office
In Partial Fulfillment of the Requirements
for
The Master's Degree in Social Work
Sebastian P.W. Mott
Montreal, August, 2012
Advisor: Dr. David William Rothwell
Running head: SHELTER USE AT THE OLD BREWERY MISSION 2
TABLE OF CONTENTS
LIST OF TABLES ................................................................................................................ 4
ABSTRACT .......................................................................................................................... 5
INTRODUCTION ................................................................................................................ 6
Literature Review ...................................................................................................... 7
Definition of homelessness .......................................................................... 7
Prevalence of homelessness ......................................................................... 8
Demographics of homeless populations ..................................................... 10
Mental health .............................................................................................. 11
Substance use ............................................................................................. 13
Physical health ........................................................................................... 14
A TYPOLOGY OF HOMELESSNESS ............................................................................. 15
The Kuhn and Culhane model ................................................................................ 15
Kuhn and Culhane methodology ............................................................................ 17
RESEARCH QUESTIONS ................................................................................................ 20
METHOD ........................................................................................................................... 20
Description of the shelter programs ....................................................................... 20
Refuge ........................................................................................................ 20
The transitional programs .......................................................................... 21
Étape .......................................................................................... 21
Escale ......................................................................................... 21
Description of the data ........................................................................................... 22
The HIFIS initiative ................................................................................... 22
Format of data ............................................................................................ 23
Database creation ....................................................................................... 24
COMPARISON OF THE SHELTER'S PROGRAMS ....................................................... 25
Running head: SHELTER USE AT THE OLD BREWERY MISSION 3
Program comparisons.............................................................................................. 31
APPLYING THE TYPOLOGY TO THE OBM ................................................................ 34
Grouping clients into types ..................................................................................... 34
Comparison of clusters on background variables ................................................... 37
Prediction of chronicity........................................................................................... 39
RESEARCH FINDINGS .................................................................................................... 41
DISCUSSION ..................................................................................................................... 41
Shortcomings of data .............................................................................................. 41
Changes over time....................................................................................... 41
Varied data entry ......................................................................................... 42
Data collection by shelter program ............................................................. 42
Limitations of analysis ............................................................................................ 43
Recommendations for later data collection............................................................. 44
Implications for practice ......................................................................................... 46
Further research ...................................................................................................... 46
REFERENCES ................................................................................................................... 48
Running head: SHELTER USE AT THE OLD BREWERY MISSION 4
LIST OF TABLES
Table 1: Refuge demographics ........................................................................................... 28
Table 2: Étape demographics .............................................................................................. 29
Table 3: Escale demographics ............................................................................................ 30
Table 4: Compared programs - point in time ...................................................................... 32
Table 5: Compared programs - over time ........................................................................... 33
Table 6: Shelter use by type ................................................................................................ 36
Table 7: Shelter use by program history ............................................................................. 36
Table 8: Demographics by type .......................................................................................... 38
Table 9: Logistic regression ................................................................................................ 40
Running head: SHELTER USE AT THE OLD BREWERY MISSION 5
Abstract
Homelessness is a pervasive and complex social problem facing Canadian society. Given that the
current primary social response is the continued sheltering of the homeless through homeless
shelters, it is essential to understand both the demographic characteristics of the sheltered
populations as well as how these characteristics inform the trajectories of clients through these
spaces. This paper uses administrative data collected through the Homeless Individuals and
Families Information System as implemented through one such homeless shelter in order to
describe the population being served. It further divides this population into groups defined by the
typology of homeless chronicity as developed by Kuhn and Culhane, and attempts to statistically
analyse demographic characteristics in order to predict client belonging to one of the particular
groups. Recommendations for improved data collection and analysis follows, as well as
discussion of future research.
Keywords: homelessness, shelters, quantitative, typology of homelessness, transitionally
homeless, episodically homeless, chronically homeless
Running head: SHELTER USE AT THE OLD BREWERY MISSION 6
Introduction
While research on homelessness has been occurring for some time now, there is a
shortage of quantitative research regarding the processes and service delivery within shelters.
What does exists in the wider literature on homelessness generally speaks to describing
characteristics of the population (Hurtubise, Babin, & Grimard, 2007; Hwang et al., 2008; Kuhn
& Culhane, 1998; Rich & Clark, 2005). Other research focuses on more theoretical aspects of the
shelter system, discussing the need for low income housing, supportive housing, or better income
support (Hulchanski, Campsie, Chau, Hwang, & Paradis, 2009). Even those few studies that are
focused on program responses to homelessness tend to be oriented towards newly-developed
programs such as Housing First of the Managed Alcohol Program (Podymow, Turnbull, Coyle,
Yetisie, & Wells, 2006; Sam Tsemberis, Gulour, & Nakae, 2004). Further, these are generally
focused on specific, minority cohorts within the homeless or sheltered populations, separating
groups by gender, occupation, substance use, health status, or mental illness (Bonin, Fournier, &
Blais, 2007; D. Folsom & Jeste, 2002; McNeil & Guirguis-Younger, 2012; O’Connell, Kasprow,
& Rosencheck, 2008; Deborah K. Padgett, Stanhope, Henwood, & Stefancic, 2011; Rich &
Clark, 2005; Scott, 2007). The traditional activities of shelters, as they relate to their population
as a whole rather than the minority cohorts within it, have been explored to a lesser depth.
Studies to this effect do exist, although they are few and far between (Fitzpatrick-Lewis et al.,
2011). This study then also serves the purpose of being a small step towards filling a research
gap in this field.
This paper will first provide a brief overview of the literature on the definition of
homelessness, its prevalence, and key demographic factors. This is followed by a review of
Kuhn and Culhane's (1998) typology of homelessness. Subsequently, the source of the data, as
Running head: SHELTER USE AT THE OLD BREWERY MISSION 7
well as its manipulations and shortcomings, will be discussed. Next, the programs that operate
out of the OBM's Webster Pavilion (Refuge, Étape, and Escale) will be introduced and
described, followed by a comparison of their population on available variables. In the
penultimate section, the Kuhn and Culhane model will be applied to the OBM population under
study, with each cluster compared, and an analysis of variables predicting chronicity conducted.
Finally, a discussion will wrap up this study, with reference limitations, recommendations, and
further research.
Literature Review
Definition of Homelessness. As a term, homelessness lacks both precision and
consistency. It is also one that has a surprisingly short history, given the lengthy historical record
of urban poverty and limited housing (Ocobock, 2008; Stern, 1984).
In Homelessness: What's in a Word? Hulchanski et al. (2009) make the case that
homelessness is fundamentally a modern invention in the Western world. While homelessness
did exist elsewhere in developing countries, it is only in the 1980's that homelessness emerged as
a problem in Canadian cities. Prior to this, the poorest Canadians, those considered down and
out, were housed. A 1977 report (as cited in Hulchanski et al., 2009) on Skid Row in Toronto
never used the word "homelessness," and only used "homeless" sparingly. The authors found
that, much as in the past, the poorest were able to find housing in deteriorated buildings in the
older sections of the city, through low rent and frequent changes in address. The focus of social
intervention at the time was to find adequate housing for these, largely, men. A prior 1960 report
(as cited in Hulchanski et al., 2009) from the Social Planning Council of Metro Toronto did refer
to "homeless men," but only in that they had few or no ties to a family, and therefore lacked the
Running head: SHELTER USE AT THE OLD BREWERY MISSION 8
support traditionally found in a home. This marks the clear distinction between a home and a
house, with the home being a social and psychological space (Hulchanski et al., 2009). It was
only in the 1980s that Canadians began to be de-housed, and homelessness took on the new
meaning of not being housed. It also became more visible, as a wave of deinstitutionalisation of
long-stay psychiatric hospitals had moved many into the community without sufficient supports,
and skid rows began to fade. As a result of this, the image of the homeless shifted slightly from
the alcoholic to the "crazy" (Fischer, 1989).
In a recent review of homelessness for the Library of Parliament, Echenberg and Jensen
(2008) arrive at the conclusion that there is no official definition in Canada, and that while
homelessness seems to convey a categorical binary, homelessness actually occurs on a
continuum, with those sleeping on streets or in shelters in one end, and those at risk of losing
their housing or living in substandard living situations on the other. Due to this ambiguity,
houselessness has been proposed as a more appropriate term to describe the social phenomenon
of people living without a residence (Echenberg & Jensen, 2008; Hulchanski, 2000).
While the programs under consideration for this research study do provide beds and
meals to homeless men, and in the cases of Étape and Escale, provide stable living situations for
up to several months, these men are still considered both homeless and houseless, both within the
broader conceptualization of these terms, as well as for the purposes of this study.
Prevalence of Homelessness. Given the difficulty in arriving at a consensus for what
homelessness means, it comes as little surprise that the counting of homeless people is no more
precise. As such, there is no clear image as to how many homeless people there are in Canada.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 9
Further, this dilutes the estimates of various demographic factors, as these are necessarily drawn
from specific samples that may not be representative of the inestimable whole.
In regards to a pure census of homeless people, the estimates vary. The 2001 Canadian
Census found 14,000 staying in homeless shelters across Canada in a single night (Statistics
Canada, 2002). By the 2006 Census, this had reached over 19,000 (Echenberg & Jensen, 2008).
However, homeless counts are often controversial, due to the difficulty in capturing such a varied
population - surveying soup kitchens or shelters is often not sufficient as many use few services,
and sleep outside or with friends (Begin, Casavant, Chenier, & Dupuis, 1999) . Annualised
counts, while problematic in their own right, provide a greater understanding of the transitory
nature of homelessness. In 1998, Montreal alone was thought to account for 10-28,000 homeless,
depending on how the count was conducted (Begin et al., 1999). Statistics Canada, however, put
this number at 1785 (Statistics Canada, 2002). Counts for wider Canada are also vague, with
estimates ranging from 130,000 to 250,000 (Begin et al., 1999) . While the practical difficulties
encountered in counting the homeless certainly account for a large degree of imprecision in the
data, the lack of a single definition for homelessness also plays a part, as homelessness is
understood to vary from relative homelessness, to hidden homelessness to absolute
homelessness, with temporal factors also coming into play, and there being little congruity
between counting methods in Canada (Echenberg & Jensen, 2008; European Federation of
national Associations Working with the Homeless, 2007; Springer, 2000). Further complicating
the interpretation of the statistics is the counting methodology: Some studies may employ a point
prevalence count, while others employ a period prevalence count. The first provides a
momentary snapshot of the homeless population, while the second counts the homeless over a
specified period of time (Hulchanski, 2000).
Running head: SHELTER USE AT THE OLD BREWERY MISSION 10
Demographics of Homelessness. Homeless populations differ from housed populations
on a number of important variables. Perhaps most visible is the significant gender imbalance. In
the case of Montreal, 91% of the shelter population in 1994 was male (Hurtubise et al., 2007).
An earlier study, however, put this number at 70% (Fournier, 1989). Given the delay between
these studies, it is impossible to say whether the difference is due to methodological disparities,
or a relative decrease in the female homeless population over time. Nevertheless, there appears to
be consensus that men make up a large majority of the homeless population. Importantly, men
and women are thought to have different pathways into homelessness. First, men are more often
represented in the chronically homeless, as women are more often able to find shelter in
exchange for sexual or domestic services (Begin et al., 1999). Second, men and women tend to
have different pathways into homelessness. While men more frequently attribute their
homelessness to loss of employment, mental health issues, or addiction, women attribute theirs to
loss of social support, eviction, or interpersonal issues (Peressini, 2007).
While people of all ages are found to be homeless, some age brackets are more highly
represented than others: Recent estimates have 75% of the Canadian homeless population as
being between the ages of 25 and 55 (Social Planning and research Council of BC, 2005). Those
over 55 represent about 9% of the homeless population, and those over 65 represent
approximately 6% (Social Planning and research Council of BC, 2005; Stuart & Arboleda-
Florez, 2000). Of further interest is the finding by Kim et al. (2010) that homeless people over 42
years of age were more than twice as likely to have mental health problems. The
underrepresentation of seniors in the homeless population versus the housed population is
thought to be due to high mortality rates (Hwang, 2000; Stergiopoulos & Herrmann, 2003).
Running head: SHELTER USE AT THE OLD BREWERY MISSION 11
Studies in the United States frequently point to the over-representation of black people in
their samples (Fargo et al., 2012; Greenberg & Rosenheck, 2010). This does not appear to be a
problem in Canada. However, Canada does have an over-representation of First Nations people
(Canadian Institute for Health and Canadian Population Health, 2007). Hwang (2001) found that
they are over-represented by a factor of 10.
In regards to occupation, the American population again shows to have a marked
difference. Homelessness among military veterans is an ongoing issue in the United States,
where they constitute from 13% to 26% of the homeless population, which far outweighs their
proportion of the housed population (Cunningham, Henry, & Lyons, 2007; Fargo et al., 2012).
Unfortunately, similar research has not been conducted in Canada, though rates of homeless
among veterans are not thought to be as high (Ray & Forchuk, 2011)
Mental health and homelessness. Mental health, substance use, and physical health
have each been studied at great length in the context of homelessness. While each of these fields
is quite broad, and is by no means a consideration only for the homeless, their impact in
homelessness is especially acute. Not only is each on its own a potential cause of homelessness,
any problem of these types can become acute in a state of homelessness, and can become more
difficult to treat (Canadian Institute for Health and Canadian Population Health, 2007). Up to
two thirds of the homeless may suffer from a combination of mental disorders, addictive
disorder, and physical disorders (Vamvakas & Rowe, 2001).
With deinstitutionalization in the 1960's came a wave of homelessness amongst the
mentally ill (Sealy & Whitehead, 2004; Stuart & Arboleda-Florez, 2000). The homeless in
Running head: SHELTER USE AT THE OLD BREWERY MISSION 12
Canada have been shown to experience higher levels of mental illness than the non-homeless
population (Public health Agency of Canada, 2006)
There have been a large number of studies on mental health within homeless populations,
both here and abroad. Findings differ, but they all point to high levels of mental health problems.
This variation should come as no surprise, given the differing makeup of homeless populations
in different geographic regions and changes over time. Studies of mental health within this
population have also not necessarily used the same instruments, or even measured the same
things: For instance, while one may be measuring more broadly for lifetime incidence of any
mental health issue, another might look more specifically at current axis III disorders.
In a review, Frankish, Hwang and Quantz (2005) identified the lifetime prevalence of
schizophrenia among Toronto's homeless as being 6%, with affective disorders being more
common, with a 20%-40% prevalence. American studies have found higher number, with large
studies in the 1980's finding rates of 19% to 30% for affective disorders, and 11% to 17% for
schizophrenia (Bonin et al., 2007). Highlighting the variance within the homeless population,
they also note that single women are more likely to have mental illness, while female heads of
homeless families have lower rates of mental illness than other homeless people. In a systematic
literature review, Folsom and Jeste (2002) found that reported rates of schizophrenia among the
homeless vary from 2% to 45%, with a smaller 4% to 16% range in the more rigorous studies. A
sample of 150 shelter and street-based homeless people in Birmingham, Alabama scored 59% as
having "probable clinical casesness" of depression (La Gory, Ritchey, & Mullis, 1990). An
Australian study of the homeless in Melbourne found an estimated 42% lifetime prevalence of
psychotic disorders (Herrman et al., 2004). More broadly, Hurtubise et al (2007) estimated the
overall rate of mental illness among the homeless at 40-60%. Kuhn and Culhane (1998) found
Running head: SHELTER USE AT THE OLD BREWERY MISSION 13
that 6.5% of the transitionally homeless, 11.8% of the episodically homeless, and 15.1% of the
chronically homeless suffer from mental illness.
Despite these high levels of mental health problems, treatment does not appear to be
accessed as frequently as one might hope. In a Montreal and Quebec City survey in 1998-1999,
Fournier found that 60% of those accessing services for the homeless reported having mental
disorders at some point in their lifetime, and that 72% of these had experienced serious disorders
within the previous year (Bonin et al., 2007). 56% of this sample had not accessed mental health
services in this time. In a study of predictors of mental health service usage amongst homeless
service users, Bonin et al (2007) found that even within a socialized health care system, where
mental health services are available free of charge, there were significant factors associated with
service use: gender, housing situation, antisocial personality disorder, alcohol disorder, and
number of people within the service user's social network were all significant at the 0.05 level.
While numerous studies examining mental health service use among the homeless exist within
the American context (D. P. Folsom, Hawthorne, & Lindamer, 2005; Kushel, Vittinghoff, &
Haas, 2001) , this is unique in its analysis of a Canadian population.
Substance use and homelessness. Also studied to great length in homelessness is
substance use. Of particular note to this study is Kuhn and Culhane's finding that 49% of the
transitionally homeless, 66% of the episodically homeless, and 83% of the chronically homeless
had substance abuse issues (Kuhn & Culhane, 1998). While a wide variety of substances are
used by the homeless, alcohol tends to be the most common, with reported rates being between
53% and 73% (Frankish et al., 2005; Podymow et al., 2006). A Toronto study found high rates of
use of other substances as well: 60% used marijuana, 52% used cocaine, 49% used crack, 25%
used oxycontin, 18% used morphine, 14% used heroin, and 25% used other opiates (Khandor &
Running head: SHELTER USE AT THE OLD BREWERY MISSION 14
Mason, 2008). Overall, substance abuse is thought to affect anywhere between 2% and 86% of
the homeless population in the United states, with a more reliable average being between 20%
and 35% (Zerger, 2002).
Comorbidity between substance use and mental health is relatively common among
homeless populations. Toronto’s Pathways Project found that virtually every homeless person
with a lifetime diagnosis of mental illness also had a substance use disorder (Mental Health
Policy Research Group, 1998; Riordan, 2004). Other meta-analyses put the rates of dual
disorder at somewhere between 10% and 20% (Zerger, 2002).
These high rates of substance users within the homeless population has led to creative
endeavours for treatment. While the vast majority of programs continue to be focused on
abstinence prior to any further assistance (Devine, Wright, & Brody, 1995; S. Tsemberis, Gulcur,
& Nakae, 2004), a harm reduction model has emerged more recently that focuses primarily on
providing housing and caring for the health needs of clients even in the presence of active
substance use (D.K. Padgett, Gulcur, & Tsemberis, 2006; Podymow et al., 2006; S. Tsemberis et
al., 2004).
Physical health and homelessness. The physical health of homeless people has been
studied at length. Physical ailments of all kinds exist within homeless populations, and are
frequently comorbid with mental health and substance use. Again, numbers vary according to
region, time, and method. Nonetheless, one of the most robust and repeated findings is that street
involved people have much higher mortality rates than the housed population, and that they are
on par with those found in underdeveloped countries (Hwang, 2000; Roy et al., 2004; Turnbull,
Muckle, & Masters, 2007). Generally, there is a high coincidence between homelessness and
Running head: SHELTER USE AT THE OLD BREWERY MISSION 15
physical health (Eberle, Kraus, Serge, & Hulchanski, 2001). Tuberculosis, HIV, arthritis,
hypertension, diabetes, fungal infections, and parasites are all relatively common in the
population (Hwang, 2001). More recent work has also found high rates of traumatic brain injury:
Hwang et al (2008), in a survey of people using services for the homeless in Toronto, found that
that up to 53% of those surveyed had experienced some form of traumatic brain injury.
Similarly, assaults are commonplace. In studies of violence amongst the homeless, researchers
have found that 40% of men have been victims of assault in the year prior to study, and that 20%
of women have been raped (Crowe & Hardill, 1993; Kushel, Evans, Perry, Roberston, & Moss,
2003).
A Typology of Homelessness
The Kuhn and Culhane Model
In 1998, Randall Kuhn and Dennis Culhane published an article entitled "Applying
cluster analysis to test a typology of homelessness by pattern of shelter utilization: Results from
the analysis of administrative data" (Kuhn & Culhane, 1998). In it, they applied a cluster analysis
to administrative data records from the homeless shelter databases of two cities in order to test an
earlier theoretical model of homelessness for single, unaccompanied adults.
Kuhn and Culhane (1998) state that the model they set out to test was based on several
earlier theoretical articles: Lovell, Barrow, & Struening,1984; Morse, 1986; Fischer & Breakey,
1986; Koegel, 1987; Snow &Anderson, 1987; Rossi, 1986; Hopper, 1989; Sosin et al., 1990;
Jahiel, 1992. According to Kuhn and Culhane, these proposed that there are three distinct types,
or patterns, of homelessness.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 16
The transitionally homeless enter the shelter for generally one episode, and remain for
only a short time. They tend to be younger, and have fewer difficulties with mental health,
substance use, or medical issues. They generally become homeless through some crisis, and have
exhausted other resources. They are generally able to stabilize themselves in short order, and do
not return to homelessness. They make up the majority of people who become homeless, due to
their high rate of turnover.
The episodically homeless often cycle between homelessness and non-homelessness,
either through occasionally finding short term accommodation, living on the street, or being
institutionalised. They also are likely to be young, though they experience higher rates of mental
health problems, substance abuse, and medical problems. They may not fit the profile of the
chronically homeless simply due to their frequent interactions with institutions such as prison,
hospitals and detoxification centres. They are likely to have numerous episodes of varying
lengths.
The chronically homeless are those who use shelters more as a long term housing
solution than as emergency relief. They tend to be older, chronically unemployed, and are more
frequently disabled or experiencing substance abuse problems. They have few separate episodes
at shelters, but remain there for much longer periods.
Before moving on, it is important to clarify the definition of an episode, as it is not
synonymous here with a stay. Furthermore, episodes, along with total days of service use, make
up the units of analysis through which the authors conducted their cluster analysis.
There can be numerous irregularities in terms of shelter use: Clients may frequently book
in and book out at a shelter. They might have a large number of stays that are separated by only a
night or two each, when they might sleep elsewhere. Further, homeless shelters generally require
Running head: SHELTER USE AT THE OLD BREWERY MISSION 17
clients to return before a certain hour in order to retain their bed from the previous night,
otherwise, they get booked out. In order to smooth these irregularities in service use, stays can be
recalculated into episodes. Whereas a stay is counted as any number of consecutive days booked-
in, an episode is distinguished by stays that a gap between stays of thirty days or more. For
example, a client who has five different one night stays, each separated by eight days, is seen as
having one singular episode. Another client who has two single night stays, separated one from
the other by 30 days, is seen as having two distinct episodes. This definition is used by Kuhn and
Kulhane, and is based on previous work by P. Koegel and M.A. Burnam (Kuhn & Culhane,
1998).
Kuhn and Culhane methodology. Kuhn and Culhane collected a large amount of data
from city-wide homeless shelter utilization records from both Philadelphia and New York. This
data included information not only on stays in the shelter, but also on other variables, such as
mental health, physical health, substance abuse, age, education, and gender. In the case of New
York, these health indicators are the result of self-report or interviewer determination, and as a
result there may be under-identification. In the case of Philadelphia, the client stay records were
merged with local health records in order to supplement the information, potentially making it
more reliable.
In the case of New York, Kuhn and Culhane included data collected over the three years
subsequent to a client's first stay at a shelter. While the authors had access to nine years worth of
data, due to the implementation of a city-wide computer booking system for shelters, the
decision to only collect data over three years per client allowed for a larger sample of clients for
whom a lengthy stay history could be recorded. In order to achieve this three year window, the
authors excluded from the sample any clients whose first stay was less than three years prior to
Running head: SHELTER USE AT THE OLD BREWERY MISSION 18
the last day of data collection. Likewise, any client for whom there was evidence of service use
prior to data collection was excluded from the sample. This would have included, for example,
any clients for whom there was some record of shelter stays prior to the implementation of the
computer system from which the data for this study was extracted.
The Philadelphia data had a more limited time frame, and so the window for client data
collection was reduced to two years. Both data sets had a final date of October 1, 1995, though
New York's began in 1986, while Philadelphia's began in 1991. As with the New York data,
clients who were known to be homeless prior to the creation of the system, as well as clients who
were first booked in later than two years prior to the end of data collection were excluded from
the study. The Philadelphia dataset was augmented with data from local databases on publicly
reimbursed mental health and substance abuse treatment.
Kuhn and Culhane used a nearest centroid sorting cluster analysis, using the variables
"number of episodes" and "number of days" as the seed variables. In order to ensure that the
number of days, which could number over a thousand for a client, would not outweigh the
number of stays, which numbered up to a maximum of fourteen, Kuhn and Culhane rescaled the
days and episodes variables to have a mean of zero and a variance of one. These records are then
submitted to the cluster analysis based on a three cluster model. While the authors lament the
restriction on clusters, they argue that while the analysis requires a distinct number of clusters, in
this case they are drawn from a significant amount of background theoretical research, and are
therefore not as arbitrary as they might otherwise seem.
Testing indicated that the clusters accurately represented the theoretical case models.
Reliability of the clustering was tested by splitting the sample into two subsamples and running
the cluster analysis again on the first half, then this model was used to predict cluster
Running head: SHELTER USE AT THE OLD BREWERY MISSION 19
membership in the second subsample. 99.1% of observations were accurately predicted. Further,
not only were the findings replicated within the subsamples of the dataset, but there was also
agreement between the both the New York data and the Philadelphia data.
The findings from Kuhn and Culhane's cluster analysis coincide highly with the
predictions of background characteristics from the theoretical model: Clients belonging to the
chronic cluster are older, while those belonging to the transitional and episodic clusters are
younger (significant at the 0.05 level). There does appear to be some variance in health
indicators between cities however. In New York, the rates of mental illness and medical
difficulties were lowest among the transitional cluster (6.5%/14.2%) and highest among the
chronic cluster (15.1%/24.0%), with the episodic cluster (11.8%/19.8%) falling in between.
Substance abuse, however, was lowest for the transitional cluster (28.2%) and highest for the
episodic cluster (40%), with the chronic cluster falling in between (37.9).
In Philadelphia, both the episodic (6.4%) and chronic (5.3%) had higher rates of self-
reported mental health issues than the transitional (3.4%). On self report of medical problems
and substance abuse, however, the authors saw the familiar pattern of lowest rates among the
transitional (14.0%/31.2%), highest rates among the chronic (28.7%/69.5%), and the transitional
(18.7%/50.5%) falling in between. Interestingly, while there were more substance abusers in the
chronic cluster, significantly more of the episodic clients had a history of substance abuse
treatment, which may lend credence to the concept that they are episodic by virtue of their
frequent stays in institutions. Or, to the contrary, it may simply be a result of "clean and sober"
shelter rules in Philadelphia (Kuhn & Culhane, 1998). All of the above findings were significant
at the 0.05 level.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 20
There are a number of different cluster or typology models that have been published in
the literature on homelessness (for a detailed analysis, see Keith Humphreys, 1995;
William McAllister, Li Kuang, & Mary Clare Lennon, 2010). However, the Kuhn and Culhane
three cluster model appears to be that most widely cited. Furthermore, it provides a
categorization that is easily replicated given the administrative data made available to this
project, yet one which is nonetheless theoretically supported. As such, it is a useful means by
which to assess the shelter use patterns within the population being studied. Many of the
variables included in this analysis, such as <30, >50, and all the combinations of mental health,
substance use and physical health, were incorporated due to their existence in Kuhn and
Culhane's analysis.
Research Questions
The purpose of this study is to develop an understanding of the population being served
by the Old Brewery Mission (OBM). To this end, it seeks to answer three primary research
questions: First, what are the demographic characteristics of those who populate the three
primary residential programs at the OBM? Are there any significant differences between them?
And what are the significant predictors of chronicity, as defined by Kuhn and Culhane (1998),
among this population?
Methodology
Description of the Shelter's Programs
Refuge. The Refuge is the OBM's emergency shelter. Clients accessing this program are
given a bunk for the night, access to showers, and are given three meals per day. Beyond this,
service is limited. There is no case management available to them. They must remain indoors
Running head: SHELTER USE AT THE OLD BREWERY MISSION 21
overnight, and must vacate the premises after breakfast. While education, citizenship, and
financial data is generally not collected on these clients, many of them have previously taken part
in the transitional programs, where this data is collected. The data is then kept in their permanent
file.
The Transitional Programs. There are two transitional programs available at the OBM's
Webster Pavilion: Étape and Escale. These are partially conceived of as being sequential in
nature.
Étape. Like the Refuge, Étape provides a bunk, meals, showers, and laundry to clients.
However, the Refuge and Étape are on different floors, meaning that Étape clients have exclusive
use of their space, which includes a common area. Furthermore, they have 24 hour access to their
bunks, and have case management services provided to them. This program is free of charge, and
typically lasts thirty days, though extensions are sometimes granted. Particular program goals are
unspecified, but can be understood to be helping clients to overcome their crisis situations in
order to return to stable housing.
Escale. Escale is considered a transitional program with more intensive services than
Étape. Clients are entitled to stay for up to three months, and are required to pay rent. Unlike
Refuge and Étape clients, Escale clients have the option between a semi-private dormitory, a
room shared with one other person, or an individual room. Greater privacy, of course, comes
with an increase in rent. As with Étape, clients of Escale have 24 hour access to the shelter,
receive case management, and have access to meals, showers and laundry. Clients can remain for
up to three months, though extensions can be granted. As with Étape, particular program goals
Running head: SHELTER USE AT THE OLD BREWERY MISSION 22
are unspecified, but can be understood to be helping clients to overcome their crisis situations in
order to return to stable housing.
Description of the Data
The HIFIS Initiative. As part of the Homelessness Partnering Strategy, Human
Resources and Skills Development Canada and the Canadian Mortgage and Housing Corporation
(CMHC) launched the Homeless Individuals and Families Information System (HIFIS) in 2001.
As part of this program, HIFIS software is provided to Canadian homeless shelters free of
charge, along with regular updates and toll-free technical support. It was designed to be both an
administrative tool, used to help shelters and policy makers to "collect the data needed to better
address the needs of the local, regional and national homeless population", as well as a means by
which homeless people could "participate in the development of flexible and responsive
programs and services that meet their needs" (Peressini & Engeland, 2004). While having been
developed by the CMHC, it has now been completely transferred into the hands of the Canadian
Homelessness Secretariat (Peressini & Engeland, 2004). As of 2007, HIFIS was active in 400
shelters country-wide (Government of Canada, 2007).
In regards to the day to day operation of the shelter, HIFIS is a software program that is
used to track clients and beds. Clients are booked in and out to specific beds through the system,
and their intake and discharge details are maintained within the system. The system does allow
for creation and modification of new data fields, and so the type of information collected can be
quite varied. For instance, the OBM's client files contain the system defaults of first nations
status, gender, and age, as well as more detailed information pertaining to citizenship status,
Running head: SHELTER USE AT THE OLD BREWERY MISSION 23
education, health indicators, financial details, correctional history, identification numbers, case
management sessions, etcetera.
Format of Data. While HIFIS can produce reports detailing stay lengths and client
details, the utility of these are limited in a number of ways: First, there are only a limited variety
of reports available as options: There is a list of reports that are available to generate, which is
largely composed of details that would be useful for the day to day operation of the shelter. For
example, there is a report that lists all clients currently booked in at the shelter. Second, these
reports are only customizable in regards to date ranges and shelter programs. Working with the
previous example, this it would be possible to generate a report of the clients booked in to one of
the shelters unique programs on any given date. Third, HIFIS reports do not come with the
details necessary to merge data between independent reports.
Given these shortcomings, accessing the data for analysis required exporting a copy of
the entire database. Unfortunately, the database does not exist as a simple Excel file. The
database is divided into 75 individual data categories, with up to 55 different variables each.
Each of these individual data categories is encoded as a combination of .FPT, .DBF, and .CDX
files. Accessing them through Excel requires the installation of the Microsoft OLE DB Provider
for Visual FoxPro. These data categories can then be imported and saved as data tables within
Excel.
The coding of variables differs: dates, names, memos, integers, codes, numeric identifier
codes, phone numbers, and bed numbers, among others, are included. Fortunately, one of the
many data categories is a table of the 2,858 codes and their corresponding meaning.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 24
Database Creation. The database for analysis was created using STATA 12. This
software allows for the direct importing of Excel tables. Unique client identifiers were found in a
number of data tables that allowed for data to be merged. Included in this analysis were the
client, health, financial information, contributing factors, and stay details files. A number of
manipulations were made to the data. Most importantly, the data pertaining to stays is calculated
from the day the client booked in to the day the client was booked out. The very next time the
client booked in, even if it was on the same day as the book in, was considered a separate stay. In
order to align this with the Kuhn and Culhane research, stays were converted into episodes,
whereby stays that were separated by 30 days or less were merged into one episode. Any stay
that began more than 30 days after a book out were considered the beginning of the subsequent
episode.
Unfortunately, there is no single variable within HIFIS for mental health, physical health,
or substance use. Instead, there are a number of locations where information pertaining to these
issues may be entered. Therefore, in order to construct useful binary variables, STATA was
programmed to generate the affirmative of the binary variable in question if a relevant code
appeared in any one of a number of different locations.
In order to create an eligibility window for client data collection not unlike Kuhn and
Culhane's, the time frame was kept to two years. This corresponds to the Philadelphia data's time
frame, and was necessary due to the short time frame of viable data - data collection only began
in May of 2009, and data entry on a number of key variables only began as late as May 2010.
Since the period of data collection for this paper ended in June of 2012, this left only two years
Running head: SHELTER USE AT THE OLD BREWERY MISSION 25
and two months for data collection, and therefore only a two month window for clients to have a
first stay, if this analysis was to stay true to the Kuhn and Culhane analysis. This was impossible
due to the extremely limited sample this produced. Consequently, eligibility limitations were
relaxed so that any client who had a stay during those two months, regardless of previous history,
was included in the analysis, and their episode history was tracked for the two years following
their first stay during that two month window.
Comparison of the Shelter's Programs
The following sections detail the three main programs at the Webster Pavilion at the
OBM: Refuge, Étape, and Escale. These are the three programs at for which data exists in the
Webster Pavilion's HIFIS server. While there is also data collection through HIFIS at the
women's shelter at the Patricia McKenzie Pavilion, it exists for an even shorter time period.
The following tables of descriptive statistics use the same variables as those in the Kuhn
and Culhane study where possible. This explains the variables < 30 and > 50, as well as mental
health, physical health, and substance use. Unfortunately, data on race was not available. As
well, analysis on gender data would be fruitless given that these programs are open only to men.
Mean age, education, citizenship, and financial data were included in order to test the
significance of a greater number of variables.
The first three tables are of descriptive statistics for each of the three main shelter
programs: Refuge (Table 1), Étape (Table 2), and Escale (Table 3). Each of these has two
different counts. The first of these counts is a point-in-time count. This is a count of each of the
variables for those client booked in to each program on the night of May 20 2011. The second of
Running head: SHELTER USE AT THE OLD BREWERY MISSION 26
these counts is an over-time count. This is a count of each of the variables for each client booked
in over the course of two years, from January 1 2009 to December 31 2011.
While Étape and Escale have relatively complete statistics, with little in the way of
missing data, Refuge data unfortunately comes up short on many fronts. This is due to the fact
that clients complete more lengthy intake questionnaires upon their entry into Étape or Escale.
Age data is complete for each of the shelter programs, while education, citizenship, health, and
financial data is lacking for Refuge clients. The reason this data exists at all for this group is
largely due to historical contact with either Étape or Escale: The majority of Refuge clients for
whom this more indepth data exists have had stays in one of these programs. There is a minority,
however, that do not have any stays in Étape or Escale. It is hypothesized that these clients
completed intake questionnaires for Étape or Escale, but did not follow through with any stays.
Point-in-time statistics for both Étape and Escale have little in the way of missing data,
while the over time counts are much less complete. This is hypothesized to be due to historical
changes in what information is collected through intake questionnaires. While it is known that
the intake questionnaires have changed since the inception of the programs, the extent of these
changes is not fully known. Furthermore, as stated previously, data collection is becoming more
consistent over time.
Age, education, citizenship, and financial data exist within HIFIS as discrete categories:
They are responses to specific questions in the intake questionnaires, and are entered as such in
HIFIS. In contrast to this, penal history and health data can exists in a large number of
categories. Health data indicating mental illness, for example, can exist within the categories of
Health Issues, Contributing Factors, or Reason for Service. As such, it is much harder to
Running head: SHELTER USE AT THE OLD BREWERY MISSION 27
determine whether there is missing data for these questions. While the point-in-time data appears
complete, given the completeness of the rest of the intake sections, the over time count of these
variables remains questionable.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 28
Table 1
Refuge Descriptive Statistics
Variables Point-in-time count Over time count
Mean SD Mean SD
Age 49.06 12.43 44.73 11.85
Monthly income 777.52 285.5 768.42 311.7
Monthly expenses 433.92 256.8 408.50 267.7
N (%) N (%)
Age at book in
<30 13 (8.33) 1,775 (12.88)
>50 81 (51.92) 5,024 (36.46)
Education
No Schooling 0 (0.00) 37 (1.05)
Elementary 6 (22.22) 703 (19.95)
High school 12 (44.44) 1,890 (53.65)
Post secondary 9 (33.33) 893 (25.35)
Citizenship
Canadian citizen 21 (77.77) 3,000 (85.15)
Permanent res 5 (18.52) 255 (7.24)
Refugee 1 (3.70) 145 (4.12)
Visa holder 0 13 (0.37)
Known penal history 0 (0.00) 280 (7.95)
Health Status
Mental Health (MH) 8 (12.50) 1,149 (16.81)
Substance Use (SU) 17 (26.56) 2,009 (29.38)
Physical Health (PH) 14 (21.86) 1,744 (25.51)
SU or MH 22 (34.38) 2,615 (38.25)
SU and MH 3 (4.69) 543 (7.94)
MH or PH 20 (31.25) 2,425 (35.47)
MH and PH 2 (3.13) 468 (6.85)
PH or SU 26 (40.63) 3,071 (44.92)
PH and SU 5 (7.81) 682 (9.98)
SU or MH or PH 30 (46.88) 3,447 (50.42)
SU and MH and PH 1 (1.56) 238 (3.48)
Note. For age variables, N = 156 for point-in-time, and N = 13,778 for
over time. For education and citizenship variables, N = 27 for point-in-
time, and N = 3,523 for over time. For health status variables, N = 64
for point-in-time, and N = 6,837 for over time. Income N = 27 for
point-in-time and N = 3,018 for over time. Expense N = 13 for point-
in-time, and N = 1,460 for over time.
< 30, > 50, and combinations of health factors included so as to
provide comparable variables to Kuhn and Culhane (1998).
Running head: SHELTER USE AT THE OLD BREWERY MISSION 29
Table 2
Étape Descriptive Statistics
Variables Point-in-time count Over time count
Mean SD Mean SD
Age 42.6 10.55 42.47 11.07
Monthly income $ 772.15 265.89 761.96 317.49
Monthly expenses $ 529.23 624.00 418.83 299.34
N (%) N (%)
Age at book in
<30 6 (11.32) 206 (14.26)
>50 14 (26.42) 393 (27.20)
Education
No Schooling 0 (0.00) 5 (0.48)
Elementary 10 (19.23) 188 (17.87)
High school 23 (44.23) 541 (51.43)
Post secondary 19 (36.54) 318 (30.23)
Citizenship
Canadian Citizen 49 (92.45) 849 (86.90)
Permanent resident 2 (3.77) 82 (8.39)
Refugee 2 (3.77) 41 (4.20)
Visa holder 0 (0.00) 5 (0.51)
Known penal history 10 (18.87) 179 (12.39)
Health Status
Mental Health (MH) 15 (28.30) 345 (23.88)
Substance Use (SU) 19 (35.85) 490 (33.91)
Physical Health (PH) 17 (32.08) 483 (33.43)
SU or MH 30 (56.60) 671 ( 46.44)
SU and MH 4 (7.55) 164 (11.35)
MH or PH 28 (52.83) 674 (46.64)
MH and PH 4 (7.55) 154 (10.66)
PH or SU 29 (54.72) 779 (53.91)
PH and SU 7 (13.21) 194 (13.43)
SU or MH or PH 38 (71.70) 890 (61.59)
SU and MH and PH 2 (3.77) 84 (5.81)
Note. For age variables, N = 53 for point-in-time, and N = 1,445 for
over time. For education, citizenship, and health issue variables, N =
53 for point-in-time. For over time, education N = 1052, while health
issues and citizenship N = 977. Income N = 39 for point-in-time and N
= 833 for over time. Expense N = 26 for point-in-time, and N = 366
for over time.
< 30, > 50, and combinations of health factors included so as to
provide comparable variables to Kuhn and Culhane (1998).
Running head: SHELTER USE AT THE OLD BREWERY MISSION 30
Table 3
Escale Descriptive Statistics
Variables Point in time count Over time
Mean SD Mean SD
Age 47.18 9.28 45.90 11.33
Monthly income $ 802.53 323.55 789.04 294.06
Monthly expenses $ 473.92 124.41 484.14 274.56
N (%) N (%)
Age at book in
<30 1 (2.00) 41 (8.40)
>50 22 (44.00) 189 (38.73)
Education
No Schooling 0 (0.00) 2 (0.51)
Elementary 3 (6.12) 54 (13.78)
High school 24 (48.98) 202 (51.53)
Post secondary 22 (44.90) 134 (34.18)
Citizenship
Canadian citizen 45 (91.84) 345 (90.55)
Permanent resident 4 (8.16) 28 (7.35)
Refugee 0 (0.00 7 (1.84)
Visa holder 0 (0.00) 1 (0.26)
Known penal history 2 (4.00) 20 (4.10)
Health Status
Mental Health (MH) 24 (48.00) 163 (33.40)
Substance Use (SU) 18 (36.00) 197 (40.37)
Physical Health (PH) 24 (48.00) 203 (41.60)
SU or MH 36 (72.00) 284 (58.20)
SU and MH 6 (12.00) 76 (15.57)
MH or PH 35 (70.00) 292 (59.84)
MH and PH 13 (26.00) 74 (15.16)
PH or SU 32 (64.00) 300 (61.48
PH and SU 10 (20.00) 100 (20.49)
SU or MH or PH 42 (84.00) 357 (73.16)
SU and MH and PH 5 (10.00) 44 (9.02)
Note. For age variables, N = 49 for point-in-time, and N = 1,445 for
over time. For point in time education, citizenship, and health issue
variables, N = 49. For over time education and N = 392 and citizenship
= 381. Income N = 49 for point-in-time and N = 401 for over time.
Expense N = 49 for point-in-time, and N = 296 for over time.
< 30, > 50, and combinations of health factors included so as to
provide comparable variables to Kuhn and Culhane (1998).
Running head: SHELTER USE AT THE OLD BREWERY MISSION 31
Program Comparisons
Table 4 and Table 5 each compare variables across the transitional programs. Table 4
examines the differences between the point-in-time counts of each program, while Table 5
examines the differences between the over time counts of each program. Table 7 provides a
summary of shelter use by shelter program history.
As discussed previously, the over time counts have significant and sometimes unknown
numbers of missing data. However, given that they are counted over the same time frame, and
that intake questionnaires for both programs have always been equivalent to each other, they
remain comparable.
In order to ascertain the p-value of the difference between programs on any given
categorical variable, chi-square ( χ2)
were calculated. However, in those cases where expected
cell counts were less than 5, Fisher's Exact was employed instead. Finally, in those cases where
data were continuous, as in "Mean Age at Book-in", "Mean Monthly Income", and "Mean
Monthly Expenses", ttests were conducted.
In both tables, there is a significant difference between the "No data" variables under both
education and citizenship. This is due to the separate intake questionnaires completed by clients
of Étape and Escale. Significant in both tables is the younger age of Étape clients, as well as the
greater frequency of penal history. Also, in Table 5, the Étape clients are less likely to have post
secondary education.
In regards to health status, on almost all variables, Escale clients are significantly more
frequently identified than Étape clients, and Étape clients more frequently than Refuge clients.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 32
Table 4
Crosstabulation of Étape and Escale Point in Time Variables
Point in Time
Variables
Refuge Étape Escale
Three-
way χ2
Étape vs
Escale χ2
Mean SD Mean SD Mean SD
Age 49.06 12.43 42.6 10.55 47.18 9.28 6.19**a -2.33*t
Monthly income $ 777.52 285.5 772.15 265.89 802.53 323.55 1.08a -1.47t
Monthly expenses $ 433.92 256.8 529.23 624.00 473.92 124.41 8.94**a -2.83**t
N (%) N (%) N (%)
Age at book in
<30 13 (8.33) 6 (11.32) 1 (2.00) 0.178F 0.113F
>50 81 (51.92) 14 (26.42) 22 (44.00) 10.43* 3.50
Education
No Schooling 0 (0.00) 0 (0.00) 0 (0.00)
Elementary 6 (22.22) 10 (19.23) 3 (6.12) 0.08 0.07F
High school 12 (44.44) 23 (44.23) 24 (48.98) 0.27 0.23
Post secondary 9 (33.33) 19 (36.54) 22 (44.90) 1.21 0.73
Citizenship
Canadian citizen 21 (77.78) 49 (92.45) 45 (91.84) 4.57 0.01
Permanent res 5 (18.52) 2 (3.77) 4 (8.16) 0.09F 0.42F
Refugee 1 (3.70) 2 (3.77) 0 (0.00) 0.42F 0.50F
Visa holder - - - - -
Known penal history 0 (0.00) 10 (18.87) 2 (4.00) 0.00**F 0.03*
Health Status
Ment. Health (MH) 8 (12.50) 15 (28.30) 24 (48.00) 51.48** 4.24*
Subs. Use (SU) 17 (26.56) 19 (35.85) 18 (36.00) 23.54** 0.00
Phys Health (PH) 14 (21.86) 17 (32.08) 24 (48.00) 39.16** 2.72
SU or MH 22 (34.38) 30 (56.60) 36 (72.00) 16.42** 2.65
SU & MH 3 (4.69) 4 (7.55) 6 (12.00) 0.313F 0.82
MH or PH 20 (31.25) 28 (52.83) 35 (70.00) 17.16** 3.19
MH & PH 2 (3.13) 4 (7.55) 13 (26.00) 0.001**F 0.02*F
PH or SU 26 (40.63) 29 (54.72) 32 (64.00) 6.36* 0.92
PH & SU 5 (7.81) 7 (13.21) 10 (20.00) 3.65 0.86
SU or MH or PH 30 (46.88) 38 (71.70) 42 (84.00) 18.38** 2.25
SU & MH & PH 1 (1.56) 2 (3.77) 5 (10.00) 0.14F 0.26
* = p < .05, ** = p < .005.
F = Fisher's Exact used rather than χ2
due to low expected cell counts.
t = ttest
a = ANOVA
Note. For age variables, Refuge N = 156, Étape N = 53, Escale N = 50. For education and citizenship
variables, Refuge N = 27, Étape N = 52, Escale N = 49. For health status, Refuge N = 64, Étape N = 53,
and Escale N = 50. For income, Refuge N = 27, Étape N = 39, and Escale N = 49. For Expenses, Refuge
N = 13, Étape N = 26, Escale N = 49.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 33
Table 5
Crosstabulation of Étape and Escale Over Time Variables
Over Time
Variables
Refuge Étape Escale
Three-
way χ2
Étape vs
Escale χ2
Mean SD Mean SD Mean SD
Age 44.73 11.85 42.47 11.07 45.90 11.33 27.59**a -5.89**t
Monthly income $ 768.42 311.7 761.96 317.49 789.04 294.06 1.08 a -1.47t
Monthly expenses $ 408.50 267.7 418.83 299.34 484.14 274.56 8.94** a -2.83**t
N (%) N (%) N (%)
Age at book in
<30 1,775 (12.88) 206 (14.26) 41 (8.40) 11.17** 11.22**
>50 5,024 (36.46) 393 (27.20) 189 (38.73) 50.97** 23.06**
Education
No Schooling 37 (1.05) 5 (0.48) 2 (0.51) 0.19F 1.00F
Elementary 703 (19.95) 188 (17.87) 54 (13.78) 9.90* 3.43
High school 1,890 (53.65) 541 (51.43) 202 (51.53) 1.98 0.00
Post secondary 893 (35.35) 318 (30.23) 134 (34.18) 20.65** 2.08
Citizenship
Canadian citizen 3,000 (87.90) 849 (86.90) 345 (90.55) 3.44 3.44
Permanent res 255 (7.47) 82 (8.39) 28 (7.35) 0.97 0.40
Refugee 145 (4.25) 41 (4.20) 7 (1.84) 5.2 4.47*
Visa holder 13 (0.38) 5 (0.51) 1 (0.26 0.86F 1.00F
Known penal history 280 (2.03) 179 (12.39) 20 (4.10) 476.33** 27.14**
Health Status
Ment Health(MH) 1,149 (16.81) 345 (23.88) 163 (33.40) 613.44** 17.09**
Subs. use (SU) 2,009 (29.38) 490 (33.91) 197 (40.37) 534.57** 6.64*
Phys. Health (PH) 1,744 (25.51) 483 (33.43) 203 (41.60) 694.46** 10.64**
SU or MH 2,615 (38.25) 671 ( 46.44) 284 (58.20) 98.63** 20.19**
SU & MH 543 (7.94) 164 (11.35) 76 (15.57) 45.10** 5.99*
MH or PH 2,425 (35.47) 674 (46.64) 292 (59.84) 160.47** 25.40**
MH & PH 468 (6.85) 154 (10.66) 74 (15.16) 60.68** 7.12*
PH or SU 3,071 (44.92) 779 (53.91) 300 (61.48 80.23** 8.47**
PH & SU 682 (9.98) 194 (13.43) 100 (20.49) 60.16** 14.12**
SU or MH or PH 3,447 (50.42) 890 (61.59) 357 (73.16) 139.95** 21.30**
SU & MH & PH 238 (3.48) 84 (5.81) 44 (9.02) 46.53** 6.05*
* = p < .05, ** = p < .005.
F = Fisher's Exact used rather than χ2
due to low expected cell counts.
t = ttest
Note. For age variables, Refuge N = 13,778, Étape N = 1,445, and Escale N = 488. For education
variables, Refuge N = 3,023, Étape N = 1,052, and Escale N = 392. For citizenship variables, Refuge N =
3,413, Étape N = 977, Escale N = 381. For health status variables, Refuge N = 6,837, Étape N = 1, 445,
and Escale N = 488. For Income, Refuge N = 3, 018, Étape N = 833, and Escale N = 401. For expense,
Refuge N = 1, 460, Étape N = 366, and Escale N = 391.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 34
Applying the Typology to the OBM Dataset
Table 6 and Table 8 attempt to answer whether the population of the OBM approximates
that found by Kuhn and Culhane in their survey of homeless shelter users in New York and
Philadelphia (Kuhn and Culhane, 1998).
Table 6 provides a breakdown of shelter use by cluster. It examines primarily the number
of days and the number of episodes by cluster. Note that as in Kuhn and Culhane (1998), the
majority of clients are transitional, and that the chronic clients, while accounting for a small
minority of clients, consume an enormous majority of client days.
Grouping Client Into Clusters
It should be reiterated that this is not an attempt to conduct the same analysis as Kuhn
and Culhane. It is, to the contrary, a rather primitive attempt at clustering clients into the same
categories using the category boundaries for the two year Philadelphia sample as published by
Kuhn and Culhane. Unfortunately, the boundaries they published were not complete, and so for
the sake of this analysis, some had to be arbitrarily erected. Their cluster analysis resulted in
transitional clients who had an average of 1.19 stays with a total of 20.4 days on average, with
episodes lasting an average of 17.1 days. Episodic clients had between 3 and 320 days (72.7 on
average) over 3 to 10 episodes (3.84 on average), and a much lower days per episode than the
chronic, at 18.9. Chronic clients had between 132 and 730 days (252 on average) over 5 or fewer
episodes (1.53 on average), with a higher number of days per episode at 165.
Given the limited data with which to work, as well as the evident overlap between
variables, it was impossible to perfectly recreate the Kuhn and Culhane clusters. However, this
attempt used the following rules:
Running head: SHELTER USE AT THE OLD BREWERY MISSION 35
Transitional clients
o two or fewer episodes, and a total of 50 days or fewer.
Episodic clients
o more than 5 episodes, or;
o fewer than 2 episodes, but between 51 and 131 total days, or;
o between 3 and 5 episodes, but 320 total days or less, so long as mean days per
episode is 50 or less
Chronic clients
o 5 or fewer episodes and 132 total days or more, or;
o between 2 and 5 episodes, but 320 or less, so long as mean days per episode was
over 50, or;
o total days over 320
Despite these rules, there were still some clients who fell into both chronic and episodic
categories, so the following rule was added:
Chronic designation removed if
o Client is also episodic and mean days per episode is under 50, or;
o Total episodes number more than 5
While not a particularly elegant solution, this does provide and exhaustive and exclusive
clustering of clients.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 36
Table 6
Description of Shelter Use by Type - Two Year Window
Cluster type
Variables Transitional Episodic Chronic Total
Sample size 1,044 548 289 1,881
Percentage of clients 55.5 29.13 15.36 100
Average No. of episodes 1.26 3.06 1.92 1.89
Average No. of days 11.5 67.22 406.85 88.53
Average days per episode 8.02 13.90 157.14 59.69
Client days 12100 36835 117581 166516
Percentage of client days 7.27 22.12 70.61 100
Ratio % days/% clients 0.13 0.76 4.60 1.00
Table 7
Description of Shelter Use by Shelter Program History - Two Year Window
Shelter use history
Variables Refuge only
Étape and
not Escale
Escale and
not Étape
Escale and
Étape Total
Sample size 1,091 503 60 227 1,881
Percentage of clients 58.00 26.74 3.19 12.07 100
Average No. of episodes 1.82 2.1 1.6 1.78 1.89
Average No. of days 77.72 81.18 128.02 146.28 88.53
Average days per episode 42.7 38.66 80.01 82.18 46.84
Client days 84,796 40,833 7,681 33,206 166,516
Percentage of client days 50.92 24.52 4.61 19.94 100
Ratio % days/% clients 0.88 0.92 1.45 1.65 1.00
Table 6 and Table 7 examine the histories of the same clients, but classify them
differently. While Table 6 is composed of variables and clusters included in the Kuhn and
Culhane study, Table 7 is included so as to provide some context as to the potentially
distortionary effects of having clients moving through programs which have prescribed stay
Running head: SHELTER USE AT THE OLD BREWERY MISSION 37
lengths, rather than through simple emergency shelter services as per the Kuhn and Culhane
samples. Comparing the average number of days and the average days per episode between the
tables, it becomes evident that those clients who progress through the transitional programs of
Étape and Escale may be naturally labeled as chronic. The variable Ratio %days/% clients also
seems to support this: This variable derives a ratio by dividing the percentage of client days by
the percentage of clients belonging to that particular group. The chronic clients in Table 6 are
shown to consume the more days in this respect than any other group, as are the transitional
program clients in Table 7, though to a lesser extent.
Comparison of clusters of background variables
Surprisingly, very few of the variables in Table 8 differ significantly between clusters.
While age is always significant, education and citizenship do not differ significantly. The only
other variables with any significant difference between clusters involve physical health, which
appears with increasing frequency between clusters.
All cases for which there was missing data on any of the variables was dropped. This
resulted in a loss of 744 or 71% of transitional cases, 271 or 49% of episodic cases, and 143 or
49% of chronic cases. Overall, this is a loss of 62% of cases.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 38
Table 8
Background Characteristics by Type
Variables Transitional Episodic Chronic
Three-
way χ2
Epis. vs.
Chron. χ2
Mean SD Mean SD Mean SD
Age 42.66 11.20 44.40 10.87 48.43 10.54 3.64t***
Monthly income $ 726.52 255.34 782.64 347.28 819.70 322.11 -0.92t
Monthly expenses $ 422.74 427.69 414.54 340.78 435.06 182.89 -0.43t
% % %
Age at book in
<30 14.67 10.47 4.79 9.89* 3.95*
>50 27.33 31.41 45.21 14.54** 7.89*
Education
No Schooling 0.74 1.18 0.00 0.79F 0.54F
Elementary 18.52 19.59 14.58 1.05 1.03
High school 59.26 50.89 53.13 2.18 0.12
Post secondary 21.48 28.40 32.29 3.62 0.44
Citizenship
Canadian Citizen 95.20 91.88 87.76 4.09 1.18
Permanent resident 3.20 6.25 8.16 0.26F 0.34
Refugee 0.80 1.88 3.06 0.49F 0.68F
Visa holder 0.80 0.00 1.02 0.34F 0.38F
Known penal history 11.33 6.5 6.16 5.58 0.02
Health status
Mental Health (MH) 24.00 24.19 26.71 0.43 0.32
Substance Use (SU) 27.33 31.41 26.03 1.78 1.33
Physical Health (PH) 28.67 36.46 45.21 12.23** 3.06
SU or MH 42.67 45.85 45.21 0.64 0.02
SU & MH 8.67 9.75 7.53 0.60 0.57
MH or PH 43.33 49.82 60.27 11.34** 4.20*
MH & PH 9.33 10.83 11.64 0.66 0.06
PH or SU 47 54.15 54.11 3.58 0.00
PH & SU 9.00 13.72 17.12 6.61* 0.87
SU or MH or PH 55.33 62.82 66.44 6.12* 0.55
SU & MH & PH 2.33 5.05 4.79 3.28 0.01
* = p < .05, ** = p < .005.
F = Fisher's Exact used rather than χ2
due to low expected cell counts.
t = ttest
Note. For transitional, N = 300. For episodic, N = 277. For chronic, N = 146.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 39
Prediction of Chronicity
In conducting their cluster analysis to test the three cluster typology of homelessness,
Kuhn and Culhane (1998) found that a number of variables differed significantly between
clusters. Namely, age, race, sex, mental health, substance abuse, and physical health. The
analysis in this paper used stepwise logistic regression to determine if any of the variables
available in the OBM dataset predicted membership in the chronic cluster.
The descriptive statistics alone show that there are large swaths of missing data. Due to
the influence this would have on the logistic regression, it was decided that the cases with data
missing in either education level or in citizenship status would be removed. Only 367 cases
remained.
While the first logistic regression included the complete model, it was rejected due to
collinearity of both elementary education and non-citizen. This reduced the operable variables to
mental health, substance use, physical health, age at book-in, > 50, < 30, Canadian citizen, high
school, and post secondary.
As per Hosmer and Lemeshow (1989) as referenced in Tabachnick and Fidell (2001, p.
535), the criterion for inclusion was set to .20, in order to "ensure entry of variables with
coefficients different from zero". Even at this generous inclusion criterion, the only variables that
had a sufficiently low confidence interval were "age at book-in", "high school", and "post
secondary", with "age at book-in" being the only predictor that achieved a significance level
better than .05. Results suggest that these predictors explain relatively little of the variance
(Pseudo R2 = 0.05, Log Likelihood = -195.59). However, the "age at book-in" variable indicates
Running head: SHELTER USE AT THE OLD BREWERY MISSION 40
that an increase in the age at book-in does significantly increase the odds of being in the chronic
cluster. Every other model tested achieve lower Pseudo R2 and Log Likelihood scores.
These findings align very little with those of Kuhn and Culhane (1998). While there was
agreement between this model and theirs in regards to older clients more frequently being
chronic, there was no further agreement on the remainder of the background variables. Where
Kuhn and Culhane found significant differences between the groups in mental health, substance
use, and physical health, as well as < 30 and > 50, this study found nearly none. < 30, > 50 and
physical health were significantly different between groups in simple chi square comparisons,
but were not significant in logistic multivariate regression analysis.
Table 9
Stepwise LogisticRregression Predicting Client Placement in Chronic Cluster
Number of obs = 367
LR chi2(3) = 22.13
Prob > chi2 = 0.00
Log likelihood = -195.59 Pseudo R2 = 0.05
Chronic Odds
Ratio
Std. Err. z P>z [95% Conf. Interval]
Age at book-in 1.06 0.01 4.26 0.00 1.03 1.08
High school 1.70 0.64 1.41 0.16 0.81 3.57
Post secondary 1.58 0.56 1.31 0.19 0.80 3.15
Constant 0.01 0.01 -5.78 0.00 0.00 0.07
Running head: SHELTER USE AT THE OLD BREWERY MISSION 41
Research Findings
The research questions this paper sought to answer were have been addressed through the
tables presented herein,. The demographic details, though limited in scope, have been elucidated.
Interestingly, some differences emerged. Notable differences included Étape clients being
generally younger, not as highly educated, and more likely to have a penal history. Escale
clients, for their part, were more likely to have mental health, substance use, or physical health
issues. In regards to the prediction of chronicity, the findings are somewhat less clear: While age
and education predict chronicity, they do so to a very small extent, as indicated by the low R2
value. Importantly, the p values for the education variables are such that they are non-significant.
Consequently, the current capacity for predicting chronicity is quite low.
Discussion
Shortcomings of the Data
The data was limited in a number of ways that hampered the analysis. First, the fields of
data evolved over time. Second, data entry varied between employees. Third, the clients partook
not only in the regular shelter service, but also in programs that offered more long term beds with
greater access to the shelter. The clients of these transitional programs have much more data
collected due to the more intensive intake questionnaires, as well as the ongoing follow-up
provided by case managers.
Changes over time. Fields for data entry, as well as labels available in limited data entry
fields have changed over time. As stated in the section on database creation, some variables
appeared later than others. While some variables, such as citizenship status, education level and
correctional history have their first entries in April 2009, they only see semi-regular use as of a
Running head: SHELTER USE AT THE OLD BREWERY MISSION 42
much later date: October 7 2010. Other variables, though in existence as of the implementation
of HIFIS, see the creation of new and important variables pertaining to substance use and mental
health as of October 2010.
Varied data entry. A number of labels within frequently-used fields in HIFIS have never
been defined at the OBM. For example, there have been up to 37 different discharge options
available for use, but there is little agreement on what each means. Similarly, there is no policy
on entering new health data subsequent to the client's intake: some case managers may add to the
client's health file if they discover the existence of a health problem, whereas others may not,
assuming that the data there is a reflection of the status at the point of intake intake alone.
Finally, data entry in general has not been consistent, though it is growing more so. Whereas
client citizenship and education data was once a rarity, it is now being entered with greater
regularity.
Data collection by shelter program. As stated earlier, data analysed from HIFIS comes
from three different shelter programs: the Refuge, Étape, and Escale. The Refuge has a relatively
barebones intake package, requiring among other things name, date of birth, reason for service,
First Nations status, gender, and relatively recently, any drug, alcohol, or mental health issues of
concern, though this last is entered at the discretion of the employee performing the intake and
without client input - further, its implementation was for the purpose of security rather than
responding to client needs. Étape and Escale, however, have much more intensive intake
packages that frequently take over an hour to complete. These include detailed questions
regarding family history, health history, employment history, financial details, etcetera. Further,
due to the continued follow-up provided by these programs, it is possible that clients have their
files updated and case managers learn more about them. Consequently, clients in Étape and
Running head: SHELTER USE AT THE OLD BREWERY MISSION 43
Escale may be identified much more frequently when they have mental health, physical health,
or substance use issues. Nonetheless, in all cases, age, gender, and stay data is entered.
Limitations of Analysis
While this was largely a theoretical and academic exercise, there are nonetheless some
lamentable limitations to the analysis. First and foremost among these is the shortage of useful
data. While Étape and Escale clients have relatively complete intakes that create comprehensive
images of clients, the intake questionnaire for the Refuge is quite limited. This means that not
only are a majority (62%) of clients missing important citizenship and education data, but they
are also lacking in data related to health indicators, as the refuge intake questionnaire does not
specifically ask these questions, simply leaving a space for intake workers to enter suspected
cases. This large gap in the data has reduced the sample so greatly that analyses are subsequently
hampered.
Second, the stay data from which the clusters are constructed are directly affected by the
shelter programs that clients participate in. Those clients who participate in Étape and Escale
generally have longer stays as compared to Refuge clients, as seen in Table 6. This means that
those clients who are by definition more likely to be lumped into the episodic or chronic
categories, by virtue of their longer stays, are also more likely to have better background data.
Third, this analysis was unable to follow the Kuhn and Culhane model of counting clients
as of their first date booked-in, due to the reduced sample size this would produce. Consequently,
the analysis may over-represent episodic and chronic client vis-a-vis the Kuhn and Culhane
findings. Due to their longer periods of homelessness, a shelter would be likely to accrue more
Running head: SHELTER USE AT THE OLD BREWERY MISSION 44
and more people from these clusters over time, as their stays would overlap. This could explain
the difference in the relative sizes of the clusters.
Finally, the data collected was limited to one homeless shelter, while it is entirely
possible that clients move from shelter to shelter. Consequently, the model created here does not
only deal with the limitation that it does not account for street homelessness, as does the Kuhn
and Culhane model, but it also cannot account for shelter users who stay at multiple shelters.
Recommendations for Later Data Collection
In order to better know the population at the OBM, and consequently in order to better
serve them, it is important to collect information more broadly, more consistently, more reliably,
and more widely. Additionally, data validity is a concern that may be helpful to address.
The first recommendation is that the basic client intake questionnaires for Refuge are
modified so as to include information about health indicators. This could simply take the form of
adding a few direct questions that must be asked as a part of every intake. Conversely, the intake
questionnaire could be reworked more fundamentally. Intake questionnaires from other shelters
or other programs for at risk populations similar to those dealt with at the OBM could be
assessed in comparison to that currently used at the OBM.
The second recommendation is that there be better quality assurance for the completeness
of intake questionnaires. While completeness of intake questionnaires has been getting more
consistent of late, it would be a relatively simple task to identify patterns by which they are not
completed properly. For example, it may be a matter of time of day, particular staff, or specific
questions.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 45
The third recommendation is that all data collected become more reliable between
workers. As things stand, there is little in the way of training, policies, and procedures for staff
who complete intakes, discharges, and follow up with clients. As a result, there is disagreement
on the definition of terms, as well as the updating of client files. Ensuring that all employees
share a common understanding of terms and that each completes intakes, discharges and follow-
up similarly would likely require a combination of training and revision of forms. However, it
would go a long way towards making data more reliable, and therefore opening doors for more
analyses.
Fourth, it would be useful for larger policy planning as well as program development if
the homeless shelters of Montreal were to share their client usage database. Currently, the OBM
does not share client data with other shelters. Not only is this a utility made possible already
through HIFIS, but it has already been adopted by other jurisdictions in Canada, and cities in the
United States have had shared data systems since the mid-1980s (Kuhn & Culhane, 1998).
Finally, health indicators are currently assessed by intake workers and case managers.
Their assessment of client health issues are a result either of client self-disclosure or the worker's
own suspicion. While this is interesting in its own right, in terms of what this means for long
term outcomes for clients, it is also an issue for both reliability and validity. Adopting validated
instruments for the assessment of clients would address both of these issues. The concern is, of
course, that these instruments would render the intake process even more cumbersome than it
already is. This is by no means a necessary step, but it is certainly one to take under
consideration.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 46
Implications for Practice
The implications for practice here are important to consider for the context under study.
Perhaps first and foremost is the implication that the OBM quantitatively knows very little about
its clientele. Therefore, it appears to be relatively ill-equipped for serving its clientele. In order to
redress this, workers should be made aware of the importance of collecting good data, and
should also be fed back important information derived from this in order to have their efforts
reinforced.
A second, but perhaps equally important implication for practice is that workers ought to
be aware of the relationship to chronicity of age, mental health, physical health, and substance
use. A greater awareness of the distribution of the population in regards to these issues may
compel workers to be more astute in detecting them, as well as more ambitious in providing
referrals.
Finally, the extent to which those clients in the chronic cluster dominate the consumption
of services may imply that these clients ought to be identified and targeted for more intense
interventions, thus freeing up resources for other clients.
Further Research
There are several interesting avenues of research in this field. First, nothing has yet been
done with the data from case management meetings between case managers and clients. Much of
this data is in the form of longhand notes, and so would be of little use in a quantitative analysis.
However, there is simple quantitative data that includes the date of the meeting, as well as a
categorical variable for subject discussed. These could be compared to outcomes in order to test
effectiveness of case management meetings.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 47
Second, individual employee data has not yet been included in any analyses. Each data
entry in HIFIS records the name of the employee who entered it, as well as that of the last
employee to alter it. Consequently, it would be possible to analyse patterns of use by employees,
as well as their influence on outcomes of clients. These differences between employees may be
indicative of performance differences as much as they could be about differential caseloads,
allocation of clients, or simply reliability of data entry.
Third, it would be interesting to assess changes in age in the population. As Canada
undergoes a larger demographic shift towards an older population (Statistics Canada, 2010),
determining concomitant changes in shelter demographics may allow for the timely creation of
needed programs.
Finally, and more in line with this particular study, once data exists for a long enough
time frame, acluster analysis not unlike Kuhn and Culhane's (1998) could be performed to test
for what variables predict the different clusters in the Montreal context. This information could
then be used to inform the creation of programs as well as the targeting of services to particular
clients in order to head off potential future chronicity.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 48
References
Begin, P., Casavant, L., Chenier, N. M., & Dupuis, J. (1999). Homelessness ( No. PRB 99-1E) (p. 47). Library
of Parliament. Retrieved from
http://www.parl.gc.ca/Content/LOP/ResearchPublications/prb991-e.pdf
Bonin, J.-P., Fournier, L., & Blais, R. (2007). Predictors of Mental Health Service Utilization by People
Using Resources for Homeless People in Canada. Psychiatric Services, 58(7), 936–941.
Canadian Institute for Health and Canadian Population Health. (2007). Mental health and homelessness.
Improving the health of Canadian, 2007-2008. Ottawa: Canaidan Institute for Health
Information.
Crowe, C., & Hardill, K. (1993). Nursing research and political change: The street health report. Canadian
Nurse, 89(1), 21–24.
Cunningham, M., Henry, M., & Lyons, W. (2007). Vital mission: Ending homelessness among veterans (p.
32). National Alliance to End Homelessness. Retrieved from
http://www.google.ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CGAQFjAB&url=ht
tp%3A%2F%2Fwww.endhomelessness.org%2Ffiles%2F1839_file_Vital_Mission_Final.pdf&ei=YZ
krUP1Nir_JAez8gbAE&usg=AFQjCNGLNhsDy9aIGg5yqOkkFvjqsxU8dQ
Devine, J. A., Wright, J. D., & Brody, C. J. (1995). An evaluation of an alcohol and drug treatment program
for homeless substance abusers. Evaluation Review, (19), 620–645.
Eberle, M., Kraus, D., Serge, L., & Hulchanski, J. D. (2001). Volume I - The relationship between
homelessness and the health, social services and criminal justice systems: A review of the
literature. Homelessness: Causes and effects. B.C.: Ministry of Social Development and
Economic Security.
Echenberg, H., & Jensen, H. (2008). Defining and enumerating homelessness in Canada (p. 7). Library of
Parliament.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 49
European Federation of national Associations Working with the Homeless. (2007). ETHOS - European
typology of homelessness and housing exclusion. European Federation of National Associations
Working with the Homeless. Retrieved from
http://www.feantsa.org/files/freshstart/Toolkits/Ethos/Leaflet/EN.pdf
Fargo, J., Metraux, S., Byrne, T., Munley, E., Montgomery, A. E., Jones, H., Sheldon, G., et al. (2012).
Prevalence and risk of homelessness among US Veterans. Preventing Chronic Disease: Public
Health Research, Practice and Policy, 9(110112), 1–9.
Fischer, P. J. (1989). Estimating the prevalence of alcohol, drug and mental health problems in the
contemporary homeless population: a review of the literature. Contemporary Drug Problems,
16, 333–389.
Fitzpatrick-Lewis, D., Ganann, R., Krishnaratne, S., Ciliska, D., Kouyoumadjian, F., & Hwang, S. W. (2011).
Effectiveness of interventions to improve the health and housing status of homeless people: a
rapid systematic review. BMC Public Health, 11(1), 638–651. doi:Article
Folsom, D., & Jeste, D. V. (2002). Schizophrenia in Homeless Persons: A Systematic Review of the
Literature. Acta Psychiatrica Scandinavica, 105, 404–413.
Folsom, D. P., Hawthorne, W., & Lindamer, L. (2005). Prevalence and risk factors for homelessness and
utilization of mental health services among 10,340 patients with serious metnal illness in a large
public mental health system, (162), 370–376.
Fournier, L. (1989). Énumération de la clientèle des centres d’hébergement pour itniérants à Montréal.
Les sans-abri au Québec: étude exploratoire (p. 46). Montréal: Ministère de la Main-d’oeuvre et
de la Sécurité du Revenu. Retrieved from
http://www.cmis.mtl.rtss.qc.ca/pdf/publications/1989_heberg_itinerants.pdf
Frankish, C. J., Hwang, S. W., & Quantz, D. (2005). Homelessness and health in Canada: Research lessons
and priorities. Canadian Journal of Public Health, (96), s23–s29.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 50
Government of Canada. (2007, March 19). HIFIS Initiative fact sheet. Government of Canada. Retrieved
from
http://www.homelesshub.ca/%28S%28o2xgg1zddivjrba1chgazxvb%29%29/Resource/Frame.asp
x?url=http%3a%2f%2fwww.hifis.ca%2finitiative%2ffiche-information_fact-sheet%2ffact_sheet-
eng.pdf&id=34649&title=HIFIS+-+HIFIS+Initiative+Resources&owner=121
Greenberg, G. A., & Rosenheck, R. A. (2010). Mental Health Correlates of Past Homelessness in the
National Comorbidity Study Replication. Journal of Health Care for the Poor and Underserved,
21(4), 1234–1249.
Herrman, H., Evert, H., Harvey, C., Oye, G., Pinzone, T., & Gordon, I. (2004). Disability and service use
among homeless people living with psychotic disorders. Australian and New Zealand Journal of
Psychiatry, (38), 965–974.
Hulchanski, J. D. (2000). A new Canadian pastime? Counting homeless people. University of Toronto.
Retrieved from
http://www.urbancenter.utoronto.ca/pdfs/researchassociates/Hulch_CountingHomelessPeople
Hulchanski, J. D., Campsie, P., Chau, S. B. Y., Hwang, S. W., & Paradis, E. (Eds.). (2009). Homelessness:
What’s in a Word. Finding Home: Policy Options for Addressing Homelessness in Canada.
Toronto: Cities Centre, University of Toronto. Retrieved from
http://www.homelesshub.ca/FindingHome
Hurtubise, R., Babin, P. O., & Grimard, C. (2007). Understanding Shelters: An Overview of the Scientific
Literature. Colloque CRI 2007 - Shelters at a Crossroads. Departement de service social,
Universite de Sherbrooke.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 51
Hwang, S. W. (2000). Mortality among men using homeless shelters in toronto, ontario. JAMA: The
Journal of the American Medical Association, 283(16), 2152–2157.
doi:10.1001/jama.283.16.2152
Hwang, S. W. (2001). Homelessness and health. Canadian Medical Association Journal, 164(2), 229 –233.
Hwang, S. W., Colantonio, A., Chiu, S., Tolomiczenko, G., Kiss, A., Cowan, L., & Levinson, W. (2008). The
effect of traumatic brain injury on the health of homeless people. Canadian, 179(8), 779–784.
Keith Humphreys. (1995). Sequential validation of cluster analytic subtypes of homeless veterans.
American Journal of Community Psychology, 23(1), 75–98. doi:10.1007/BF02506923
Khandor, E., & Mason, K. (2008). Reserach Bulletin #3: Homelessness and crack use, 8.
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.
Kuhn, R., & Culhane, D. P. (1998). Applying Cluster Analysis to Test a Typology of Homelessness by
Pattern of Shelter Utilization: Results from the Analysis of Administrative Data. American Journal
of Community Psychology, 26(2), 207–232. doi:10.1023/A:1022176402357
Kushel, M. B., Evans, J. L., Perry, S., Roberston, M. J., & Moss, A. R. (2003). No door to lock: victimization
among homeless and marginally housed persons. Archives of Internal Medicine, 163(20), 1292–
2499.
Kushel, M. B., Vittinghoff, E., & Haas, J. S. (2001). Factors associated with the healthcare utilization of
homeless persons. Journal of the American Medical Association, (285), 739–752.
La Gory, M., Ritchey, F. J., & Mullis, J. (1990). Depression among the homeless. Journal of Health and
Social Behavior, 31(March), 87–101.
McNeil, R., & Guirguis-Younger, M. (2012). Illicit drug use as a challenge to the delivery of end-of-life
care services to homeless persons: Perceptions of health and social services professionals.
Palliative Medicine, 26(4), 350 –359. doi:10.1177/0269216311402713
Running head: SHELTER USE AT THE OLD BREWERY MISSION 52
Mental Health Policy Research Group. (1998). Mental illness and pathways into homelessness:
Proceedings and recommendations. Toronto: Canadian Mental Health Association.
O’Connell, M. J., Kasprow, W., & Rosencheck, R. A. (2008). Rates and Risk Factors for Homelessness After
SUccessful Housing in a Sample of Formmerly Homeless Veterans. Psychiatric Services, 59(3),
268–275.
Ocobock, P. (2008). Introduction: agrancy and Homelessness in Global and Historical Perspective. In A.
Beier & P. Ocobock (Eds.), Cast Out: Vagrancy and Homelessness in Global and Historical
Perspective. Athens, OH: Ohio University Press.
Padgett, D.K., Gulcur, L., & Tsemberis, S. (2006). Housing first services for people who are homeless with
co-occurring serious mental illness and substance use. Research on Social Work Practice, (16),
74–83.
Padgett, Deborah K., Stanhope, V., Henwood, B. F., & Stefancic, A. (2011). Substance Use Outcomes
Among Homeless Clients with Serious Mental Illness: Comparing Housing First with Treatment
First Programs. Community, 47, 227–232.
Peressini, T. (2007). Perceived reasons for homelessness in Canada: Testing the heterogeneity
hypothesis. Canadian Journal of Urban Research, (16), 112–126.
Peressini, T., & Engeland, J. (2004). The Homelessness Individuals and Families Information System: A
Case Study in Canadian Capacity Building. Canadian Journal of Urban Research, 13(2), 347–361.
doi:Article
Podymow, T., Turnbull, J., Coyle, D., Yetisie, E., & Wells, G. (2006). Shelter-Based Managed Alcohol
Administration to chronically Homeless People Addicted to Alcohol. Canadian Medical
Association Journal, 174(1), 45–49.
Public health Agency of Canada. (2006). The human face of mental health and mental illness in Canada.
Ottawa: Public Health Agency of Canada.
Running head: SHELTER USE AT THE OLD BREWERY MISSION 53
Ray, S. L., & Forchuk, C. (2011). The experience of homelessness among Canadian Forces and allied forces
veterans (p. 25). London, ON: Homelessness Partnering Secretariat. Retrieved from
http://homelesshub.ca/ResourceFiles/Homelesss%20Vets%20Article.pdf
Rich, A., & Clark, C. (2005). Gender Differences in Response to Homelessness Services. Evaluation and
Program Planning, 28(1), 69–81.
Riordan, T. (2004). Exploring the circle: Mental illness, homelessness and the criminal justice system in
Canada ( No. PRB 04-02E). Ottawa: Library of Parliament.
Roy, E., Haley, N., Leclerc, P., Sochanski, B., Boudreau, J. F., & Boivin, J. F. (2004). Mortality in a cohort of
street youth in Montreal. Journal of the American Medical Association, 292(5), 569–574.
Scott, S. (2007). All Our Sisters: Stories of Homeless Women in Canada. Peterborough: Broadview Press.
Sealy, P., & Whitehead, P. C. (2004). Forty years of desinstitutionalization of psychiatric services in
Canada: an empirical assessment. Canadian Journal of Psychiatry, 49(4), 249–257.
Social Planning and research Council of BC. (2005). On our streets and in our shetlers... Results of the
2005 Greater Vancouver Homeless Count. Vancouver, BC: Social Research and Planning Council
of BC.
Springer, S. (2000). Homelessness: A proposal for a global definition and classification. Habitat
International, 24, 475–484.
Statistics Canada. (2002). 2001 census: Collective dwellings ( No. 96F0030XIE2001004). 2001 Census
Analysis Series (p. 24). Retrieved from
http://www12.statcan.gc.ca/english/census01/Products/Analytic/companion/coll/pdf/96F0030
XIE2001004.pdf
Statistics Canada. (2010). Demographic change. Retrieved September 10, 2012, from
http://www.statcan.gc.ca/pub/82-229-x/2009001/demo/int1-eng.htm
Running head: SHELTER USE AT THE OLD BREWERY MISSION 54
Stergiopoulos, V., & Herrmann, N. (2003). Old and homeless: A review and survey of older adults who
use shelters in an urban setting. Canadian, 48(6), 374–380.
Stern, M. J. (1984). The Emergence of the Homeless as a Public Problem. Social Service Review, 58(2),
291–301.
Stuart, H. L., & Arboleda-Florez, J. (2000). Homeless shelter users in the post deinstitutionalization era.
Canadian Journal of Psychiatry, (45), 55–62.
Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston: Allyn & bacon.
Tsemberis, S., Gulcur, L., & Nakae, M. (2004). Housing first, consumer choice, and harm reduction for
homeless individuals with a dual diagnosis. American Journal of Public Health, 94(4), 651–656.
Tsemberis, Sam, Gulour, L., & Nakae, M. (2004). Housing First, Consumer Choice, and Harm Reduction
for Homeless individuals With a Dual Diagnosis. American Journal of Public health, 94(4), 651–
656.
Turnbull, J., Muckle, W., & Masters, C. (2007). Homelessness and health. Canadian Medical Association
Journal, 177(9), 1065–1068.
Vamvakas, A., & Rowe, M. (2001). Mental health training in emergency homeless shelters. Community
Mental Health Journal, 37(3), 287–295.
William McAllister, Li Kuang, & Mary Clare Lennon. (2010). Typologizing Temporality: Time‐Aggregated
and Time‐Patterned Approaches to Conceptualizing Homelessness. Social Service Review, 84(2),
225–255.
Zerger, S. (2002). Substance abuse treatment: What words for homeless people? A review of the
literature. Nashville: National Health Care for the Homeless Council.