A latent class analysis of self-identified reasons for experiencing homelessness: Opportunities for prevention

Article (PDF Available)inJournal of Community & Applied Social Psychology 28(2):94-107 · March 2018with 218 Reads 
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
DOI: 10.1002/casp.2343
Cite this publication
Abstract
Individuals experience homelessness due to a vast number of factors. Therefore, the methods used to prevent individuals from experiencing homelessness should match their diverse needs. This study utilized survey data obtained from 577 adults experiencing homelessness to identify self-reported causes of homelessness. A latent class analysis was conducted in order to identify classes or subgroups of respondents with distinct patterns of reported causes of homelessness. A latent class analysis is a person-centred statistical approach that is used to determine groups of individuals who share similar characteristics. Findings from this analysis identified 5 distinct classes based on individuals' responses to 19 potential vulnerabilities or events that contributed to experiencing homelessness. Individuals tended to cluster around issues associated with (a) disability or physical health issues (4%), (b) substance abuse or mental health issues (30%), (c) report major life changes (3%), (d) financial crises (7%), or (e) employment difficulties (55%). Significant group differences occurred across military veteran status, history of homelessness, depression, and health-related quality of life. Results for these analyses suggest that individuals report notable differences in their reasons for becoming homeless and therefore require unique preventative solutions.
Figures - uploaded by Anna Smith Pruitt
Author content
All content in this area was uploaded by Anna Smith Pruitt
Content may be subject to copyright.
RESEARCH ARTICLE
A latent class analysis of selfidentified reasons for
experiencing homelessness: Opportunities for
prevention
John P. Barile
1
|Anna Smith Pruitt
1
|Josie L. Parker
2
1
University of Hawaii at Mānoa, Honolulu, HI,
USA
2
Pathways Community Network Institute,
Atlanta, GA, USA
Correspondence
John P. Barile, PhD, Department of
Psychology, University of Hawaii at Mānoa,
2530 Dole Street, Sakamaki Hall, C404,
Honolulu, HI 968222294, USA.
Email: barile@hawaii.edu
Funding information
National Institutes of Health, National Institute
on Minority Health and Health Disparities,
Grant/Award Number: 5U54MD00758407
Abstract
Individuals experience homelessness due to a vast number of factors.
Therefore, the methods used to prevent individuals from experienc-
ing homelessness should match their diverse needs. This study
utilized survey data obtained from 577 adults experiencing home-
lessness to identify selfreported causes of homelessness. A latent
class analysis was conducted in order to identify classes or subgroups
of respondents with distinct patterns of reported causes of home-
lessness. A latent class analysis is a personcentred statistical
approach that is used to determine groups of individuals who share
similar characteristics. Findings from this analysis identified 5 distinct
classes based on individuals' responses to 19 potential vulnerabilities
or events that contributed to experiencing homelessness. Individuals
tended to cluster around issues associated with (a) disability or phys-
ical health issues (4%), (b) substance abuse or mental health issues
(30%), (c) report major life changes (3%), (d) financial crises (7%), or
(e) employment difficulties (55%). Significant group differences
occurred across military veteran status, history of homelessness,
depression, and healthrelated quality of life. Results for these
analyses suggest that individuals report notable differences in their
reasons for becoming homeless and therefore require unique
preventative solutions.
1|INTRODUCTION
Homelessness is an international problem, widespread throughout the developed world (Shinn, 2007). In the United
States, an estimated of 578,424 people experience homelessness on any given night (Henry et al., 2014). The U.S.
Department of Housing and Urban Development defines homelessness as living in a place not meant for human
habitation, in an emergency shelter, or in transitional housing or as in a state of exiting an institution (e.g., a prison,
--------------------------------- --- -- -- --- -- --- -- -- --- -- --- -- -- --- -- --- -- --- -- -- --- -- --- -- -- --- -- --- --
Copyright © 2018 John Wiley & Sons, Ltd.
Received: 3 January 2017 Revised: 10 January 2018 Accepted: 25 January 2018
DOI: 10.1002/casp.2343
94 J Community Appl Soc Psychol. 2018;28:94107.wileyonlinelibrary.com/journal/casp
hospital, or mental health facility). There is relative consensus that homelessness in developed nations, such as the
United States and the United Kingdom, is primarily the result of a combination of systemic factors (e.g., poverty
and shortage of affordable housing stock) and individual vulnerabilities (e.g., health issues and unemployment; Main,
1998; Shinn & Gillespie, 1994). These factors are not discrete; rather, they interact, resulting in multiple pathways in
and out of homelessness (Anderson & Christian, 2003; Clapham, 2003). Subsequently, a diverse set of interventions is
needed to prevent individuals from falling into homelessness (Anderson, 2003; Anderson & Christian, 2003). First, we
must develop a complex understanding of these interacting factors. This study investigated patterns of factors
contributing to homelessness. In particular, it (a) identified common contributing factors to homelessness as reported
by individuals currently experiencing homelessness in a Southern U.S. city; (b) determined whether the reporting
individuals could be organized into unique subgroups based on patterns of selfidentified reasons for experiencing
homelessness; and (c) examined the unique challenges (e.g., health, disability, and substance abuse) associated with
being a member of each subgroup. Understanding common patterns of pathways to homelessness may help
researchers identify key opportunities for prevention.
Most research demonstrates that homelessness results from both macrolevel (structural) and microlevel
(individual) factors (Lee, Tyler, & Wright, 2010). At the macrolevel, factors such as a housing squeeze(Fitzpatrick,
Kemp, & Klinker, 2000; Kemp, Lynch, & Mackay, 2001; Wright, Donley, & Gotham, 2008), changes in social policy
(Shinn, 2007; Third & Yanetta, 2000), reduction in public housing (Third & Yanetta, 2000), income inequality (Toro,
2007), poverty (Anderson & Christian, 2003; Fitzpatrick et al., 2000), and unemployment (Kemp et al., 2001; Philippot,
Lecocq, Sempoux, Nachtergael, & Galand, 2007) produce a population of people who are at risk for experiencing
homelessness (Lee et al., 2010). Which of these atrisk people actually become homeless depends, in part, on
microlevel factors such as individual vulnerabilities. Individual vulnerabilities can include low income, changes in fam-
ily composition or relationship, military veteran status, debt, lack of social support, alcohol abuse, lack of education,
and mental and physical health issues (Anderson & Christian, 2003; Crane et al., 2005; Fitzpatrick et al., 2000; Lee
et al., 2010; Shinn, 2007). For example, when affordable housing markets are tight, individuals with financial problems,
health issues, and substance abuse issues are more vulnerable to becoming homeless (Crane et al., 2005; Shinn, 1997;
Shinn et al., 1998; Shinn & Gillespie, 1994). Without adequate buffers, such as social support or education, individual
vulnerabilities can exacerbate or be exacerbated by macrolevel impacts (Fitzpatrick et al., 2000; Lee et al., 2010).
Additionally, experiences with an institution such as prison or mental health facility may result in stigma, which can
further exacerbate macrolevel factors related to employment and housing (Roman & Travis, 2006). Ultimately,
individual vulnerabilities contribute to who gets housing, but systematic disenfranchisement places some individuals
in the front and others in the back of the line (Lee et al., 2010).
Although research suggests that homelessness results from a complex interaction of both microand macrolevel
factors, this complexity is difficult to capture in practice. Clapham (2003) argues that much research tends to focus on
one level or the other and not on the interaction between factors. Additionally, some researchers warn that this
macro/microlevel dichotomy is too simplistic, pointing out that it can be difficult to distinguish between levels
and that some factors do not neatly fitinto one level or the other (e.g., unemployment; Clapham, 2003; Fitzpatrick
et al., 2000; Neale, 1997). Instead, the causes of homelessness are complex, dynamic, and overlapping (Anderson &
Christian, 2003). These factors are neither static nor discrete; rather, they change over time and interact within and
across levels. Furthermore, causes of homelessness often cannot be distinguished from effects of homelessness
(Anderson & Christian, 2003). Therefore, understanding factors contributing to homelessness necessitates a focus
on interactions among overlapping factors at multiple levels.
Recognizing the dynamic and complicated nature of homelessness causes, the pathways framework provides a
more holistic approach to understanding microlevel issues within a structural context. Focusing on the interaction
can help determine pathways into and out of homelessness (Anderson & Christian, 2003; Clapham, 2003). Similarly,
Shinn (2007) suggests a multilevel framework that examines the interaction of policy, sociocultural, and individual
factors. She argues that researchers must examine interactions across multiple levels if they are to fully understand
causes of homelessness and develop appropriate interventions. By studying the complex interaction of individual
BARILE ET AL.95
and structural factors (both within and across levels), researchers can identify distinct patterns of causes that will be
useful in homelessness prevention.
Identifying naturally occurring subgroups or classes of individuals based on the complex interaction of specified
multilevel factors is one way researchers can target intervention and prevention efforts. Researchers have approached
these questions using both qualitative (e.g., Chamberlain & Johnson, 2013) and quantitative approaches (e.g., Kuhn &
Culhane, 1998; Morse, Calsyn, & Burger, 1992). Taking a qualitative approach to examining over 4,000 case histories
of individuals who had experienced homelessness, Chamberlain and Johnson (2013) identified five common pathways
to homelessness in Australia. These pathways included individuals that transitioned into homelessness from youth
(35% of their sample), individuals who became homeless due to a housing crisis (19%), and individuals with a history
of substance abuse (17%), mental health issues (16%), or family breakdown (11%; Chamberlain & Johnson, 2013).
Taken together, these approaches have provided a strong for unique pathways to experiencing homelessness based
on service records.
Previous literature also has demonstrated the value of quantitative approaches, such as cluster analysis tech-
niques, in classifying individuals experiencing or at risk for homelessness into distinct subgroups based on their
pattern of social services use (Gleason, Barile, & Baker, 2017; Kuhn & Culhane, 1998) and their reported service needs
(Morse et al., 1992). For example, using archival shelter data in New York City, Kuhn and Culhane (1998) found that
users of homeless shelters could be classified as Traditionally Homeless (80% of their sample), Episodically Homeless
(10%), or Chronically Homeless (10%). Traditionally Homeless individuals tended to be younger (compared with
Chronically Homeless individuals), to not abuse substances, and to be in good health. Episodically Homeless individ-
uals also tended to be younger, but they presented with mental health, substance abuse, and medical problems. The
Chronically Homeless individuals also had substance abuse and medical problems but were older than individuals in
the other two classes. This study has been useful in identifying service use patterns and in targeting interventions.
Morse et al. (1992) also used cluster analysis to identify four subgroups of individuals experiencing homelessness.
Based on interviews with emergency shelter users (N= 248) in St. Louis, the study classified individuals based on ser-
vice needs as opposed to service use. The Economically Disadvantaged group (53% of the sample) was similar to Kuhn
and Culhane's Traditionally Homeless group in that it also made up the largest proportion of the sample and did not
report notable health or substance abuse issues. They also identified a Drinking Problems group (20%; similar to Kuhn
and Culhane's Chronically Homeless group), with a mental health subgroup (17%; similar to Kuhn and Culhane's
Episodically Homeless group). A smaller proportion of the sample was classified as Socially Advantaged group (5%),
which was unique in that individuals reported greater income and wider social support networks. Finally, they were
unable to classify 6% of their sample. Although both studies were useful in identifying important service use and
needs patterns that helped identify directions for intervention, neither analysis included a direct query of actual
causes of homelessness. Therefore, these classes represent a clustering of individuals with similar demographic and
health conditions but do not shed light into what leads individuals to experience homeless.
A more recent study used a similar method to examine individual risk factors for homelessness (Tsai, Kasprow, &
Rosenheck, 2013). Using latent class analysis (LCA), Tsai et al. (2013) examined data on 120,852 homeless military
veterans collected from 142 Department of Veterans Affairs sites across the United States. Similar to cluster analysis,
LCA can be used to identify naturally occurring groups based on a set of variables. Using nine variables, they identified
four latent classes: Poverty/Substance Abuse/Incarceration (40%), Dual Diagnosis (28%), Relatively Few Problems (22%),
and Disabling Medical Conditions (10%), which suggested particular pathways in and out of homelessness (Tsai et al.,
2013). For example, the veterans classified into the Relatively Few Problems class were less likely to be chronically
homeless, to ever been incarcerated, or to have a substance use disorder. These individuals were also more likely
to be referred to permanent housing programmes. Individuals in the Poverty/Substance Use/Incarceration class were
more likely to have been incarcerated, to report an income less than $600 dollars a month, and to have a substance
use disorder (although substance use rates were slightly lower than the Dual Diagnosis class). Individuals in the Dual
Diagnosis class had the highest probability of having a substance use disorder, a psychotic disorder, or a psychiatric
hospitalization. Finally, those in the Disabling Medical Problems class had a very high probability of having a chronic
96 BARILE ET AL.
medical problem and to have been unemployed the previous 3 years. These findings suggest that even within a
subgroup of veterans, multiple pathways to homelessness exist, and these pathways have implications for
intervention and prevention.
Taken together, these studies (along with other studies on the topic; see (Aubry, Klodawsky, & Coulombe, 2012;
McAllister, Lennon, & Kuang, 2011; Muñnoz, Panadero, Santos, & Quiroga, 2005) suggest that individuals experienc-
ing homelessness have different rates of substance abuse, physical and mental health issues, and economic challenges
and that these issues are associated with different patterns of homelessnessfor example, transitional, episodic, or
chronic. Moreover, these patterns suggest that a number of factors are associated with an increase in an individual's
vulnerability to experiencing homelessness when combined with systemic factors, such as lack of affordable housing.
These three studies demonstrate the value of cluster and LCA in understanding experiences of homelessness and in
identifying key opportunities for targeting intervention efforts. Given the value of cluster and LCAs in determining
service use and needs patterns and in developing interventions, this study applied this technique to identifying
patterns of selfreported causes of homelessness in order to inform prevention efforts.
1.1 |Current investigation
The current investigation built on previous work devoted to identifying subgroups of individuals with a history of
homelessness, but unlike these studies, the current investigation focused on what factors led individuals to experience
homelessness. Whereas previous work in this area has largely characterized subgroups based on reports of service
use, medical conditions, employment history, and/or history of institutionalization, this study directly asked individ-
uals experiencing homelessness what factors led to their homelessness and used these data to identify subgroups.
Specifically, the current investigation sought to determine the following:
1. Based on selfidentified reasons for becoming homeless, can a large sample of individuals experiencing homeless-
ness be organized into unique groups using LCA?
2. Is class membership associated with differences in the amount of time or frequency in which individuals have
experienced homelessness?
3. Do the class compositions differ based on demographic factors (e.g., gender, age, military veteran status, family
status, and disability status)?
4. Do classes differ in reported substance use, healthrelated quality of life, depression, or chronic health
conditions?
Answers to these questions will be useful in tailoring prevention programmes to individuals based on patterns of
selfidentified factors.
2|METHODS
2.1 |Participants
A communitylevel survey with individuals experiencing homelessness was conducted from February to April 2013 in
a large Southern U.S. metropolitan city (Atlanta, Georgia). A community homeless count that was conducted in
January 2013 revealed a homeless population that was majority male (68%), unaccompanied (i.e., individuals; 88%),
and living in a shelter (69%; Parker, 2013). Purposive sampling was conducted based on these three main character-
istics (gender, sleeping location, and family status) in order to obtain a sample of individuals that were similar to
those identified during the community homeless count. In total, 603 surveys were conducted, and the surveyed
BARILE ET AL.97
adults closely reflected the homeless count on all three characteristics, with 63% adult male, 91% unaccompanied,
and 65% sheltered.
Of the 603 surveys, 577 were complete (26 questionnaires were unusable due to several factors, such as
duplication). The resulting respondents (N= 577) were overwhelmingly unaccompanied (91%) and male (83%). The
respondents were also majority Black/African American (88%) and middleaged (4554 years of age; 42%), which is
representative of the homeless population in this area. The majority of the survey respondents (61%) had experienced
homelessness multiple times over the previous 3 years, which indicates a population that moves in and out of
homelessness. Additionally, almost half of the sample (46%) had experienced homelessness for 1 year or longer.
Although the survey results are generally representative, because the sample is not random, survey respondents
may not precisely mirror the homeless population in general on certain factors. Specifically, the study sample does
not include individuals who were doubled up or temporarily staying with a friend or relative but otherwise had no
permanent residence. It also does not include individuals that may have been temporarily residing in jail or the hospital
with no permanent residence. Instead, the participants in this study are more representative of individuals staying in
emergency shelter or classified as unsheltered.
2.2 |Procedures
The surveys were primarily administered at eight locations ranging from service provider agencies and meal sites to
emergency shelters and transitional housing programmes. Survey administrators were military veterans who them-
selves had experienced homelessness. They received extensive training on the questionnaire, survey techniques,
and protocol. Having veterans who have experienced homelessness conduct the survey created a peertopeer
interview scenario whereby the respondents felt more comfortable answering personal questions about their home-
less situation. The questionnaires were also designed to be selfadministered for respondents who had difficulty with
the facetoface interview process (e.g., persons who are deaf). Potential participants were simply asked if they would
be willing to complete the survey at each location. The survey took approximately 15 min, and respondents received a
$5 fast food restaurant gift certificate as compensation for their time. Institutional review board approval was granted
for this study.
2.3 |Measures
The core sections of the survey identified demographic characteristics, homeless history, and disabling conditions.
Special sections focused on reasons for becoming homeless and known individual vulnerabilities, such as depression,
healthrelated quality of life, and chronic health conditions.
2.3.1 |History of homelessness
History of homelessness was measured by asking respondents how many times they had experienced homelessness
in the last 3 years (from 1 to 6 or more times) and how many months or years they had experienced homeless
continuously since their last permanent housing (from less than 1 month to more than 5 years). These variables were
recoded to represent whether it was respondents' first time homeless (60%) and whether they had experienced
homeless for 3 years or more (24%).
2.3.2 |Disability status and activity of daily living
To determine the impact of disability, the survey asked whether any of the following made it difficult for respondents
to carry out their daily activities: alcohol use (18%), drug use (16%), physical disability (18%), mental illness (24%),
posttraumatic stress disorder (8%), HIV/AIDS (2%), chronic health problems (12%), or other conditions (2%).
98 BARILE ET AL.
2.3.3 |Reasons for becoming homeless
The survey asked respondents, what are the primary reasons that caused you to become homeless? Respondents could
respond affirmatively to any of 20 potential causes, including substance abuse factors (e.g., alcohol or drug use), eco-
nomic factors (e.g., lost job), family factors (e.g., argument with family or friends), health factors (e.g., illness or medical
problem), housing factors (e.g., housing loss due to low public housing inventory), and other factors (e.g., relocation).
These specific items were chosen based on common responses provided during previous administrations of this
survey. No individuals reported losing housing due to eviction from a foreclosed rental property; therefore, this factor
was not used in further analyses, leaving 19 potential causes of homelessness (Table 1).
2.3.4 |Healthrelated quality of life
A single item asking in general, how is your current health? measured healthrelated quality of life. Similar items have
been found to be strong predictors of mortality and overall health and are used on multiple national U.S. surveys
(Barile et al., 2013; Brown, Thompson, Zack, Arnold, & Barile, 2015).
2.3.5 |Patient Health Questionnaire
Depression was measured using the Patient Health Questionnaire (PHQ8). The PHQ8 has been validated as a
measure of current depression in U.S. samples (Kroenke et al., 2009; Kroenke & Spitzer, 2002; Kroenke, Spitzer, &
Williams, 2001). A PHQ8 score of 10 has an 88% sensitivity and 88% specificity for major depression and repre-
sents clinically significant depression (M= 7.82, SD = 6.99; 35% had scores 10; Corson, Gerrity, & Dobscha,
TABLE 1 Frequencies for each potential cause of homelessness
n%
Alcohol/drug use 187 32%
Rehab release 39 7%
Job loss 318 55%
Unable to pay rent 161 28%
Loss of money 169 29%
SSI or SSD cutoff
a
35 6%
Argument with family/friend 101 18%
Family violence 41 7%
Divorce 60 10%
Death in family 65 11%
Physical illness 98 17%
Mental illness 118 21%
Released from hospital 13 2%
Disabled 82 14%
Loss of Sec. 8 34 6%
Foreclosure 34 6%
Eviction 66 11%
Relocation 110 19%
Release from jail 106 18%
Note. Participants could mark more than one reason.
a
Supplemental Security Income (SSI) and Social Security Disability Insurance (SSDI) are U.S. governmentfunded programmes.
SSI provides lowincome individuals a monetary stipend in the event that an individual is at least 65 years old or has a doc-
umented disability. SSDI provides income supplements to people who cannot work due to a disability and have a record of
past employment earnings.
BARILE ET AL.99
2004). Items included questions such as How often during the past 2 weeks were you bothered by feeling down,
depressed, or hopeless? For each question, respondents could answer on a 4point scale, ranging from Not at all to
Nearly every day.
2.4 |Analytic procedures
LCA was conducted in an exploratory fashion to identify classes or subgroups of respondents with distinct patterns of
reasons for becoming homeless. LCA is a personcentred (rather than a variablecentred) statistical approach for
identifying groups of individuals who share similar characteristics (Hagenaars & McCutcheon, 2002). In this study,
LCA used the recognition of 19 reasons for becoming homeless to divide the population under study into subgroups
that shared a distinct interpretable pattern of relationships among the indicators. LCA assumes that two or more
classes or subgroups can be ascertained from the sample and that a categorical latent variable indicating membership
in a class or subgroup can explain the relationship among the observed indicators. Because membership in the
subgroups is unknown, the latent categorical variable must be inferred from the data. LCA, therefore, permits classes
of participants with differing profiles of reasons for becoming homeless to be determined probabilistically based on
response patterns to questions regarding their selfidentified causes of homelessness.
This study assessed a series of LCA models, beginning with a singleclass model and adding classes until the model
fit no longer improved significantly. As recommended (Nylund, Asparouhov, & Muthén, 2007), we chose a best fitting
model based on two statistical goodnessoffitindices: the Akaike information criterion (AIC) and the Bayesian infor-
mation criterion (BIC; Hagenaars & McCutcheon, 2002). Lower scores on both indices indicate a better model fit.
Graphs of the latent classes were also reviewed to consider the substantive interpretability (meaningfulness and dis-
tinctiveness) of the resultant latent class solutions. Finally, we examined the average posterior probabilities by class.
Posterior probabilities represent the probability that an individual would be classified in each of the classes, with
higher scores indicating higher probability.
3|FINDINGS
LCA models estimating three to six classes were tested. The threeclass solution resulted in AIC = 8,590.77;
BIC = 8,847.88; entropy = .74. The fourclass solution improved slightly with AIC = 8,548.48; BIC = 8,892.75;
entropy = .74. The fiveclass solution resulted in AIC = 8,528.56; BIC = 8,959.99; entropy = .78, and the sixclass
solution resulted in AIC = 8,517.39; BIC = 9,035.98; entropy = .80. We chose the fiveclass solution for further
interpretation over the fourclass and the sixclass solutions because the fiveclass solution resulted in an AIC that
was lower than the fourclass model and a BIC that was lower than the sixclass model, suggesting the fiveclass
model best fit the data. Moreover, the fiveclass model also resulted in the most interpretable classes. An examination
of the posterior probabilities revealed that individuals had a strong probability of belonging to a single class, with
coefficients ranging from .83 to .92 for the fiveclass solution. The posterior probabilities for each cause of
homelessness by class for the fiveclass solution appear in Figure 1. We labelled each of the five classes based on
the distribution of these posterior probabilities.
We labelled Class 1 (4% of participants) the Disability/Physical Health Class (DPHC) because individuals in this class
had a high probability of reporting that their government benefits (e.g., Supplemental Security Income or disability
benefits) had been cut, that they had a physical illness, and that a disability contributed to their homelessness. Class 2
(30% of participants), the Substance Abuse/Mental Health Class (SAMHC), included individuals with a high probability of
listing alcohol/drug use and mental illness as causes of their homelessness. Class 3 (3% of participants), the Life Tran-
sitions Class (LTC), reported rehab release, death in the family, eviction, and release from jail as the primary reasons for
becoming homeless. Class 4 (7% of the sample), the Financial Class (FC), indicated that loss of a job, inability to pay
rent, and loss of money were the primary causes of homelessness. Finally, Class 5 (55% of the sample), the JobRelated
100 BARILE ET AL.
Class (JRC), reported that loss of a job was the main reason for becoming homeless. Although members of the JRC did
not report job loss any more often than members of other classes, this class is unique because these individuals were
very unlikely to report any other cause for their homelessness.
Although each latent class differed significantly across demographic variables (Table 2), the majority of all classes
were male (range 8186%) and Black/African American (range 8194%). Additionally, the largest proportion of
individuals within each class fell between the ages of 4554 years of age, with the FC and the JRC having the highest
TABLE 2 Demographic characteristics by class membership
Class 1: Disability/
Physical Health
Class 2: Substance
Abuse/Mental Health
Class 3: Life
Transitions
Class 4:
Financial
Class 5: Job
Related
n%n%n%n%n%
Age
1824 ‐‐ ‐‐ 21212‐‐ ‐‐ 12 4
2534 1 5 9 6 1 6 5 12 45 14
3544 3 14 24 16 1 6 11 26 46 15
4554 11 52 74 50 8 47 17 41 117 37
5564 6 29 33 22 5 29 9 21 85 27
6574 ‐‐ ‐‐ 32‐‐ ‐‐ ‐‐ ‐‐ 11 4
75 and above ‐‐ ‐‐ 21‐‐ ‐‐ ‐‐ ‐‐ 21
Female 4 18 21 14 3 18 8 19 59 18
Black/African American 17 81 129 85 16 94 293 81 293 89
Has at least one child 3 13 12 8 2 12 4 9 32 9
Military veteran 9 39* 33 21 3 18 10 23 62 19
Note. Class 5 served as the comparison class for statistical tests.
*p< .05.
FIGURE 1 The estimated probabilities by class based on 19 potential reasons for experiencing homeless.
Supplemental Security Income (SSI) and Social Security Disability Insurance (SSDI) are U.S. governmentfunded
programmes. SSI provides lowincome individuals a monetary stipend in the event that an individual is at least
65 years old or has a documented disability. SSDI provides income supplements to people who cannot work due to a
disability and have a record of past employment earnings
BARILE ET AL.101
percentages of adults under 35 (38% and 33%, respectively). Veteran status was the most notable demographic
difference between classes, with the DPHC having the largest percentage of veterans (39%).
In addition to differing demographically, the classes also differed significantly across their history of
homelessness, health, and disability indicators (Table 3). For example, individuals in the DPHC reported the longest
duration of homelessness and had the highest reported rates of physical disability, mental disability, chronic health
problems, poor health, and depression. The SAMHC contained individuals with high rates of alcohol use, drug use,
physical disability, and mental disability (although their rates of physical disability and mental disability were still
considerably lower than individuals in the DPHC). The LTC also included individuals with high rates of drug use
but, who otherwise, had reported relatively fewer health problems. Additionally, the FC and JRC, compared with
the DPHC and SAMHC, reported fewer disabilities, better health, and lower depression scores.
Finally, classes also differed on the amount of causes listed. The JRC, the largest class (55% of respondents),
reported minimal causes for experiencing homelessness outside of job loss (50% of the sample reported job loss).
Other classes were much more likely to indicate multiple reasons for experiencing homeless. For example, on average,
the DPHC (4% of respondents) reported an average of 9.13 different reasons for becoming homeless; the LTC (3% of
respondents) reported an average of 5.65; the SAMHC (30% or respondents) reported an average of 4.9; and the FC
(7% of respondents) reported an average of 4.02.
4|DISCUSSION
These findings support and build upon previous literature and suggest possible avenues for prevention. Selective
prevention aims to prevent homelessness by targeting individuals who are at the highest risk (Gordon, 1983).
Unfortunately, because the potential causes of homelessness are so vast, determining who is at the highest risk can
be difficult, and a onesizefitsall approach to prevention is likely to be ineffective. Outside of increasing the pool
TABLE 3 Differences in history of homelessness, health, and disability by class
Class 1: Disability/
Physical Health
Class 2: Substance
Abuse/Mental Health
Class 3: Life
Transitions
Class 4:
Financial
Class 5: Job
Related
%%%%%
History of homelessness
First time homeless 45 27* 19* 50 45
Homeless less than 6 months 36 37 29 44 46
Homeless 3 years or more 38** 35*** 35 21 17
Health and disability
Poor health
a
61** 24** 12 5 14
Alcohol use
b
22 34*** 6 12 11
Drug use
b
9 29*** 53*** 9 9
Physical disability
b
57*** 25** 0 9 15
Mental disability
b
57*** 38*** 24 12 17
Chronic health problems
b
44*** 14 12 14 9
No disability reported
b
9** 14*** 18 61* 41
Mean Mean Mean Mean Mean
PHQ depression score 13.70*** 10.32*** 6.53 5.44 6.61
a
In response to the question how is your current health? with possible responses of excellent,very good,good,orpoor.
b
In response to the question please indicate if any of the things on this list make it difficult for you to carry out your daily activities.
Compared to Class 5:
*p< .05. **p< .01. ***p< .001.
102 BARILE ET AL.
of available affordable homes, it may be necessary to target prevention efforts at specific subgroups. Therefore, we
explore the unique patterns of reasons reported by individuals within each of these classes and consider possible
opportunities for prevention. For example, to prevent homelessness for individuals that matched the profile identified
by the DPHC, interventions aimed at streamlining the securement of disability benefits and preventing future loss of
these benefits may assist individuals with chronic physical and mental health issues that manage financial constraints
imposed by their disability. Furthermore, veterans represented a significantly larger percentage of this class (39%)
compared with any other class, which suggests that stronger ties between veteran health services and traditional
social services are likely needed. Additionally, because this class had the largest number of listed causes, and likely
corresponding risk factors, a wraparound service model may be appropriate for prevention with individuals
demonstrating a DPHC pattern (Apicello, 2010; Smelson et al., 2013).
Individuals within the SAMHC also have a clear pattern of precursors to becoming homeless, which offer oppor-
tunities for prevention. Persons in this class reported a high probability of individual vulnerabilities, such as abusing
alcohol or drugs, being released from rehabilitation programmes, and experiencing a mental illness. Although not as
high as those in the DPHC, the SAMHC reported significantly higher depression scores and poorer health compared
with individuals in the LTC to JRC. For this group, a dual diagnosis was likely, requiring treatment that must cut across
several systems of care. In fact, post hoc examination of the data revealed that 67% of individuals in the SAMHC
reported both substance abuse (drug or alcohol) and a mental health disability, which is in stark contrast to only 5%
of the DPHC, 5% of the LTC, 3% of the FC, and 21% of the JRC. This finding suggests that a combination of
alcohol/drug rehab, mental health assistance, and supportive housing could be critical to preventing people who fit
this classification from falling into homelessness.
People within the LTC, like those in the SAMHC, reported that drug use made it difficult for them to carry out
their daily activities. On the other hand, they reported significantly lower depression scores than both the DPHC
and SAMHC. What makes this class unique is that individuals reported recent release from rehab or jail, eviction,
and death in the family as causes of homelessness at rates much higher than individuals in other classes. Situational
crises coupled with vulnerabilities, such as drug use, created a unique pattern of causes. Therefore, although it repre-
sented a minority of the total sample (3%), the LTC individuals may be easily targetable for homelessness prevention.
Prevention efforts for this group could work to ensure that individuals, particularly those with a history of drug abuse,
are provided smoother transitions to affordable housing upon release from an institution or after experiencing an
eviction. This finding mirrors previous findings that situational crises can lead to homelessness when combined with
other microand macrolevel factors (Crane et al., 2005).
Our analyses also indicated that financial constraints, and job loss, in particular, led to episodes of homelessness
for many of the people living on the streets and in temporary shelters, suggesting that prevention efforts should con-
sider macrolevel economic conditions. For example, unlike members of the LTC, DPHC, and SAMHC, persons in the
FC reported that losing a job, inability to pay rent, and a general lack of funds were the major contributors to becoming
homeless. Similar to individuals in the JRC, individuals in this class were unlikely to report any disabling conditions. It is
likely that individuals in this class had experienced some financial difficulties prior to becoming homeless. Situations
such as losing a job or enduring major financial expenses often result in an inability to keep up rent payments.
Moreover, insecure job markets that adversely affect individuals in lowpaying jobs are particularly vulnerable to
becoming homeless.
While also reporting job loss, the largest class, the JRC (55% of the sample), reported a very limited number of
reasons for becoming homeless. Unlike all other classes, the only notable reason for becoming homeless was job loss,
which is not entirely surprising given the economic climate at the time when this survey was conducted. In January
2013, the unemployment rate for the region was 8.9% and had sat above 10% for much of the previous 3 years. This
was a dramatic increase from previous years; from 2004 to 2008, the average unemployment rate was 5.1% (United
States Department of Labor, 2017). The significant increase in unemployment may have created a new class of
individuals experiencing homelessness. Compared with the other classes, individuals in the JRC tended to be younger,
in decent health (only 14% stated they had poor health), and report few disabilities. Based on these characteristics of
BARILE ET AL.103
the FC and the JRC, greater attention to the macrolevel economic conditions that affect individuals working in
lowwage jobs warrants more consideration. Prevention efforts for individuals in these groups may include increasing
the availability of lowcost housing, securing a living wage for the lowest earning workers, and providing free or low
cost educational opportunities for individuals impacted by market shifts. We recommend the expansion of robust
reemployment programmes, such as those proposed by the Brookings Institution, that provide support while
reintegrating individuals into the workforce (Kugler, 2015). Providing support while helping individuals reengage in
the workforce would likely limit the number of individuals who slip into chronic homelessness.
Unfortunately, many homeless services focus solely on meeting the needs of individuals in the DPHC and
SAMHC. Although these services are necessary, in the spirit of prevention, more attention on individuals on the cusp
of experiencing homelessness due to financial and jobrelated conditions (such as individuals in the JRC and FC) is
warranted. Such a focus may reduce the number of individuals that experience chronic homelessness, chronic disease,
and substance abuse disorders. Data suggest that the longer an individual goes unsheltered, the greater the impact on
their health and wellbeing, demonstrating that factors are not static or discrete. Our analyses revealed that
individuals in both the JRC (33% were under 45 years of age) and the FC (38% were under 45 years of age) tended
to be younger than individuals in the DPHC (19% were under 45 years of age), the SAMHC (24% were under 45 years
of age), and the LTC (24% were under 45 years of age) and had less experience with homelessness. It is quite possible
that if these individuals were followed over time without intervention, many would eventually be reclassified as
members of a different latent class. In other words, they may develop health and substance abuserelated issues
after becoming homeless for the first time, further pointing to the need to intervene sooner (Johnson &
Chamberlain, 2008).
In addition to having implications for prevention, our study adds to and supports previous findings regarding the
clustering of individuals experiencing homelessness into unique subgroups. For example, classes such as the JRC,
DPHC, and SAMHC mirror previously identified subgroups. The JRC is similar to Kuhn and Culhane's (1998)
Traditionally Homeless class based on age, substance abuse rates, and health. In both studies, these classifications
represented the largest proportion of the sample. In this study, the JRC represented 55% of the sample, and in
Kuhn and Culhane's study, the Traditionally Homeless class represented 80% of the sample. It is quite possible that
these numbers would have been even closer if the JRC was combined with the FC (7%) and/or the Kuhn and
Culhane study had included unsheltered adults. This finding suggests that the Traditionally Homeless class likely
experienced homelessness almost exclusively due to jobrelated factors. Additionally, the DPHC mirrored Kuhn
and Culhane's Chronically Homeless class. In both studies, these classes represented individuals that were more
likely to be older and have a history of substance abuse and medical problems. Finally, Kuhn and Culhane's
Episodically Homeless classification is similar to the SAMHC found in the current study and both the Substance
Abuse and Mental Health classifications identified by Chamberlain and Johnson (2013). Both groups were defined
by mental health and substance use but also reported some physical health issues and tended to be younger, at least
compared with the DPHC. Because many of the previous studies used service history data to determine groups,
together, these findings suggest that patterns of perceived causes of homelessness can predict later patterns of
service use and needs.
In addition to supporting previous research, this study builds upon previous work by identifying a new classthe
LTC. The smallest classification in the current investigation, the LTC (3% of the sample), does not have an equivalent
classification in other studies. The lack of equivalent class in other studies is likely due to their reliance on the
consequences of experiencing homelessness (i.e., services use) in identifying classes instead of on potential causes
of homeless. For example, no service use equivalent exists for release from jail or rehab, eviction, or a death in the
familyfactors reported at high rates by individuals in the LTC. It does share some similarities to Chamberlain and
Johnson's (2013) Family Breakdown classification in that both can be associated with a disruption in family life.
However, the Family Breakdown class is associated with a partner leaving a relationship (either due to domestic
violence or as a result of new economic pressure), and the LTC is associated with a death in the family (partner or
otherwise). Overall, the LTC captures a much broader pattern of a variety life transitions that increase vulnerability
104 BARILE ET AL.
for homelessness. Identifying this subgroup has implications for prevention efforts because bridging services could be
made available to target individuals undergoing these life transitions.
Although this study builds upon previous research and helps identify opportunities for selective prevention, it does
have notable limitations. First, the list of causes of homelessness was not exhaustive. Although the survey gave the
option for respondents to describe circumstances that were not listed, few respondents chose to offer additional
causes. Therefore, a number of circumstances could have contributed to homelessness that were not recorded. This
study also relied on selfreported causes of homelessness, and it is possible that some issues, such as loss of a job, were
overreported, whereas other issues, such as substance abuse or mental health, were underreported due to perceived
social stigma. However, selfreport is also one of the strengths of this investigation because tapping into experiential
knowledge of individuals who have experienced homelessness is imperative for understanding the comprehensive
experience of homelessness. Future research would benefit from the pairing of selfreport data to service use data
and/or a longitudinal followup, which would enable researchers to determine whether individuals' diverse entry into
homelessness resulted in differences in service use and in varying exits out of homelessness. Additionally, this study
relied on a relatively small sample to conduct LCA. This resulted in some classes with small sample sizes (e.g., the LTC
only included 17 individuals). Additional research is needed to ensure that these classes can be replicated. It is also
critical to understand moderators. If we understand how factors, such as length of time homeless, impact transitions
across classes, we may be better equipped to devise interventions that allow individuals to exit homelessness. Future
research should examine how individual attributes (e.g., gender and age) and history (e.g., experience of prejudice or
length of time homeless) interact with causes of homelessness reported by individuals.
Despite these limitations, the current investigation identified five distinct classes based on 19 selfreported
causes of homelessness, and these classes overlapped with previous research that investigated the classification of
individuals based on service use and needs. This study sheds light on the events that occurred prior to individuals
becoming homeless and suggests that although factors contributing to homelessness are complex and multilevel, they
do tend to interact in detectable patterns. Identifying these patterns can be useful for guiding homelessness preven-
tion efforts. Future work might consider the ways in which these patterns and opportunities for prevention shift or
remain stable across geographic, political, and cultural contexts.
ACKNOWLEDGEMENTS
Barile received funding support from National Institutes of Health, National Institute on Minority Health and Health
Disparities (5U54MD00758407).
ORCID
John P. Barile http://orcid.org/0000-0003-4098-0640
REFERENCES
Anderson, I. (2003). Synthesizing homelessness research: Trends, lessons and prospects. Journal of Community and Applied
Social Psychology,13(2), 197205. https://doi.org/10.1002/casp.721
Anderson, I., & Christian, J. (2003). Causes of homelessness in the UK: A dynamic analysis. Journal of Community and Applied
Social Psychology,13(2), 105118. https://doi.org/10.1002/casp.714
Apicello, J. (2010). A paradigm shift in housing and homeless services: Applying the population and highrisk framework to
preventing homelessness. The Open Health Services and Policy Journal,3,4152. https://doi.org/10.2174/
1874924001003010041
Aubry, T., Klodawsky, F., & Coulombe, D. (2012). Comparing the housing trajectories of different classes within a diverse
homeless population. American Journal of Community Psychology,49(12), 142155. https://doi.org/10.1007/s10464
0119444z
Barile, J. P., Reeve, B. B., Smith, A. W., Zack, M. M., Mitchell, S. A., Kobau, R., Thompson, W. W. (2013). Monitoring
population health for Healthy People 2020: Evaluation of the NIH PROMIS
®
Global Health, CDC Healthy Days, and
BARILE ET AL.105
satisfaction with life instruments. Quality of Life Research,22(6), 12011211. https://doi.org/10.1007/s11136012
0246z
Brown, D., Thompson, W., Zack, M., Arnold, S., & Barile, J. (2015). Associations between healthrelated quality of life and
mortality in older adults. Prevention Science,16(1), 2130. https://doi.org/10.1007/s111210130437z
Chamberlain, C., & Johnson, G. (2013). Pathways into adult homelessness. Journal of Sociology,49(1), 6077. https://doi.org/
10.1177/1440783311422458
Clapham, D. (2003). Pathways approaches to homelessness research. Journal of Community and Applied Social Psychology,
13, 119127. https://doi.org/10.1002/casp.717
Corson, K., Gerrity, M. S., & Dobscha, S. K. (2004). Screening for depression and suicidality in a VA primary care setting:
2 items are better than 1 item. American Journal of Managed Care,10(11 Pt 2), 839845.
Crane, M., Byrne, K., Fu, R., Lipmann, B., Mirabelli, F., RotaBartelink, A., Warnes, A. M. (2005). The causes of homelessness
in later life: Findings from a 3nation study. The Journals of Gerontology: Social Sciences,60B(3), S152S159.
Fitzpatrick, S., Kemp, P., & Klinker, S. (2000). Single homelessness: An overview of research in Britain. Bristol: Policy Press.
Retrieved 9 Jan. 2018 from https://www.jrf.org.uk/report/singlehomelessnessoverviewresearchbritain
Gleason, K., Barile, J. P., & Baker, C. K. (2017). Describing trajectories of homeless service use in Hawai'i using latent class
growth analysis. American Journal of Community Psychology,59, 158171.
Gordon, R. S. Jr. (1983). An operational classification of disease prevention. Public Health Reports,98(2), 107109.
Hagenaars, J. A., & McCutcheon, A. L. (2002). Applied latent class analysis. Cambridge, United Kingdom: Cambridge University
Press.
Henry, M., Cortes, A., Shivji, A., Buck, K., Khadduri, K., & Culhane, D. P. (2014). The 2014 Annual Homelessness Assessment
Report (AHAR) to Congress: Part 1 point in time counts. Retrieved 9 Jan. 2018 from: https://www.hudexchange.info/
resources/documents/2014AHARPart1.pdf
Johnson, G., & Chamberlain, C. (2008). Homelessness and substance abuse: Which comes first? Australian Social Work,61(4),
342356. https://doi.org/10.1080/03124070802428191
Kemp, P., Lynch, E., & Mackay, D. (2001). Structural trends and homelessness: A quantitative analysis. Edinburgh: Scottish
Executive Central Research Unit.
Kroenke, K., & Spitzer, R. L. (2002). The PHQ9: A new depression diagnostic and severity measure. Psychiatric Annals,32(9),
509515. https://doi.org/10.3928/004857132002090106
Kroenke, K., Spitzer, R. L., & Williams, J. B. (2001). The PHQ9: Validity of a brief depression severity measure. Journal of
General Internal Medicine,16(9), 606613. https://doi.org/10.1046/j.15251497.2001.016009606.x
Kroenke, K., Strine, T. W., Spitzer, R. L., Williams, J. B., Berry, J. T., & Mokdad, A. H. (2009). The PHQ8 as a measure of
current depression in the general population. Journal of Affective Disorders,114(13), 163173. https://doi.org/
10.1016/j.jad.2008.06.026
Kugler, A. (2015). Strengthening reemployment in the unemployment insurance system. Prepared for The Hamiliton Project,
Washington, D.C. Retrieved 9 Jan. 2018 from https://www.brookings.edu/wpcontent/uploads/2016/06/strengthen-
ing_reemployment_in_unemployment_insurance_syste_kugler.pdf
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), 207232. https://doi.
org/10.1023/A:1022176402357
Lee, B. A., Tyler, K. A., & Wright, J. D. (2010). The new homelessness revisited. Annual Review of Sociology,36, 501521.
https://doi.org/10.1146/annurevsoc070308115940
Main, T. (1998). How to think about homelessness: Balancing structural and individual causes. Journal of social distress and the
homeless,7(1), 4154. https://doi.org/10.1023/A:1022966631533
McAllister, W., Lennon, M. C., & Kuang, L. (2011). Rethinking research on forming typologies of homelessness. American
Journal of Public Health,101(4), 596601. https://doi.org/10.2105/AJPH.2010.300074
Morse, G. A., Calsyn, R. J., & Burger, G. K. (1992). Development and crossvalidation of a system for classifying homeless
persons. Journal of Community Psychology,20(3), 228242. https://doi.org/10.1002/1520-6629(199207)20:3%3C228::
AID-JCOP2290200306%3E3.0.CO;2-0
Muñnoz, M., Panadero, S., Santos, E. P., & Quiroga, M. Á. (2005). Role of stressful life events in homelessness: An intragroup
analysis. American Journal of Community Psychology,35(12), 3547. https://doi.org/10.1007/s1046400518886
Neale, J. (1997). Theorising homelessness: Contemporary sociological and feminist perspectives. In R. Burrows, N. Pleace, &
D. Quilgars (Eds.), Homelessness and social policy (pp. 3549). London: Routledge.
106 BARILE ET AL.
Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of classes in latent class analysis and growth
mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling,14(4), 535569. Retrieved 9 Jan. 2018
from http%3A%2F%2Fciteseerx.ist.psu.edu%2Fviewdoc%2Fdownload%3Fdoi%3D10.1.1.461.9619%26amp%3Brep%
3Drep1%26amp%3Btype%3Dpdf
Parker, J. (2013). The 2013 Metro Atlanta trijurisdictional collaborative homeless census: City of Atlanta, Fulton and Dekalb
County. Prepared for the Metro Atlanta TriJurisdictional Collaborative on Homelessness, Atlanta, GA.
Philippot, P., Lecocq, C., Sempoux, F., Nachtergael, H., & Galand, B. (2007). Psychological research on homelessness in West-
ern Europe: A review from 1970 to 2001. Journal of Social Issues,63, 483503. Retrieved 9 Jan. 2018 from http%3A%
2F%2Fciteseerx.ist.psu.edu%2Fviewdoc%2Fdownload%3Fdoi%3D10.1.1.471.4761%26amp%3Brep%3Drep1%26amp%
3Btype%3Dpdf
Roman, C. G., & Travis, J. (2006). Where will I sleep tomorrow? Housing, homelessness, and the returning prisoner. Housing
Policy Debate,17, 389418. https://doi.org/10.1080/10511482.2006.9521574
Shinn, M. (1997). Family homelessness: State or trait? American Journal of Community Psychology,25(6), 755769. https://doi.
org/10.1023/A:1022209028188
Shinn, M. (2007). International homelessness: Policy, sociocultural, and individual perspectives. Journal of Social Issues,63(3),
657677. https://doi.org/10.1111/j.15404560.2007.00529.x
Shinn, M., & Gillespie, C. (1994). The roles of housing and poverty in the origins of homelessness. American Behavioral
Scientist,37, 505521. https://doi.org/10.1177/0002764294037004006
Shinn, M., Weitzman, B. C., Stojanovic, D., Knickman, J. R., Jimenez, L., Duchon, L., Krantz, D. H. (1998). Predictors of home-
lessness among families in New York City: From shelter request to housing stability. American Journal of Public Health,
88(11), 16511657.
Smelson, D. A., Kline, A., Kuhn, J., Rodrigues, S., O'Connor, K., Fisher, W., Kane, V. (2013). A wraparound treatment
engagement intervention for homeless veterans with cooccurring disorders. Psychological Services,10(2), 161167.
https://doi.org/10.1037/a0030948
Third, H., & Yanetta, A. (2000). Homelessness in Scotland: A summary of research evidence. Edinburgh: Scottish Homes.
Toro, P. A. (2007). Toward an international understanding of homelessness. Journal of Social Issues,63, 461481. https://doi.
org/10.1111/j.15404560.2007.00519.x
Tsai, J., Kasprow, W. J., & Rosenheck, R. A. (2013). Latent homeless risk profiles of a national sample of homeless veterans and
their relation to program referral and admission patterns. American Journal of Public Health,103(S2), S239S247. https://
doi.org/10.2105/AJPH.2013.301322
United States Department of Labor. (2017, July). Local area unemployment statistics, Bureau of Labor Statistics Retrieved
from: https://www.bls.gov/lau/#tables
Wright, J. D., Donley, A. M., & Gotham, K. F. (2008). Housing policy, the low income housing crisis, and the problem of home-
lessness. In R. McNamara (Ed.), Homelessness in America (Vol. 2)Causes of homelessness (pp. 3148). Praeger: New York.
How to cite this article: Barile JP, Pruitt AS, Parker JL. A latent class analysis of selfidentified reasons for
experiencing homelessness: Opportunities for prevention. J Community Appl Soc Psychol. 2018;28:94107.
https://doi.org/10.1002/casp.2343
BARILE ET AL.107
  • ... Similarly, the Caton et al. [54] study constructed employment and earning an income as having a preventive effect against homelessness. Participants in the current study and in other studies reported that losing a job, inability to pay rent, and a general lack of funds were the major contributors to becoming homeless, and indicated that insecure job markets adversely affect individuals in lowpaying jobs who are particularly vulnerable to becoming homeless [55]. It is worth noting that in welfare societies, poor people have a secure income; however, participants' stories indicated that income without a job does not provide a sufficient protective mechanism against homelessness. ...
    Article
    Full-text available
    Background: Homelessness is rising in the United Kingdom, despite investment in measures to eradicate it made by the government and charity organisations. Aim: The aim is to examine the stories of homeless people in order to document their perceptions of their social status, the reasons that led to their homelessness, and propose a conceptual explanation. Method: We conducted 26 semi-structured interviews in three centres for homeless people in Cheshire, North West of England. Results: Three categories—education, employment, and health—emerged from the data and provided a theoretical explanation for the reasons that led to their homelessness. These are vital not only for the successful negotiation of one’s way out of homelessness, but also for achieving other social goods, including social connections, social mobility, and engaging in positive social relationships. Conclusion: Participants catalogued the adverse childhood experiences, which they believe limited their capacity to meaningfully engage with the social institution for social goods, such as education, social services, and institutions of employment. Since not all people who have misfortunes of poor education, poor health, and loss of job end up being homeless, we contend that a combination of these with multiple adverse childhood experiences may have weakened their resilience to traumatic life changes, such as loss of job and poor health.
  • ... The purpose of this paper is to examine these differences in order to understand and address service gaps. This paper builds on previous research that identified five unique classes of individuals with distinct patterns in pathways into homelessness (Barile, Pruitt, & Parker, 2018). Using these classifications, the current investigation examined the unique service needs associated with each class and explored how individuals in these classes differed in how they learned about social services, what services they used, how they determined which services to use, and what barriers they faced to obtain services. ...
    Article
    Gaps between service needs and availability often prevent individuals with a history of homelessness from accessing services and reestablishing permanent housing. This paper examines self‐identified service utilization and service needs based on data collected from an urban sample (N = 577) of adults experiencing homelessness. This study found that individuals differ in their use and continued need of services depending on the reasons they identified as contributing to their homelessness. The majority reported that they learned of services through word of mouth, had difficulty accessing services due to limited transportation options, and were most likely to use services that were in convenient locations that fulfilled their needs and where they were treated with respect. These findings have implications for developing coordinated intakes and the development and dissemination of services aimed at assisting individuals experiencing homelessness.
  • ... These groupings were made based on theoretical and statistical reasoning. Barile, Pruitt, and Parker (2018). Participants could indicate their perception of how likely each of 26 factors were to cause homelessness in Honolulu on a fivepoint scale from "Definitely Unlikely" to "Definitely Likely." ...
    Thesis
    Full-text available
    Research suggests that dominant cultural narratives, in part, perpetuated by the media, may influence attitudes and beliefs about social problems. In turn, these attitudes and "lay theories" that the public has about a social issue can affect the type of policies that they support to solve the problem. Unfortunately, few studies have directly assessed this relationship. It is important to understand these associations because dominant cultural narratives potentially can impact the level of stigma persons experiencing homelessness experience as well as the types of policies implemented to alleviate their struggles. This study combined exploratory narrative analysis and observational research design 1) to identify the dominant narrative(s) surrounding homelessness in Hawai‘i as perpetuated by the local media; 2) to analyze community members’ exposure to and degree of endorsement of these dominant narratives among stakeholder groups; and 3) to assess the impact of these narratives on community members’ attitudes, beliefs and policies endorsed. Results showed that negative media narratives were the most common narratives in local media coverage, and community members indicated the most exposure to and endorsement of these narratives. Exposure to negative media narratives also had the most impact on endorsed solutions to homelessness. In particular, negative media exposure predicted increased endorsement of basic services and individual-level solutions when mediated by beliefs that homelessness is caused by individual deficits. Exposure to media narratives also interacted with gender and previous contact with persons experiencing homelessness to impact community member attitudes. This work discusses these findings and their implications for research on media, homelessness, and public opinion as well as for local homeless policy and interventions aimed at reducing stigma.
  • ... Similarly, the Caton et al. [54] study constructed employment and earning an income as having a preventive effect against homelessness. Participants in the current study and in other studies reported that losing a job, inability to pay rent, and a general lack of funds were the major contributors to becoming homeless, and indicated that insecure job markets adversely affect individuals in low-paying jobs who are particularly vulnerable to becoming homeless [55]. It is worth noting that in welfare societies, poor people have a secure income; however, participants' stories indicated that income without a job does not provide a sufficient protective mechanism against homelessness. ...
  • Article
    The State of Hawai‘i, like many other areas across the United States, has large numbers of individuals and families experiencing homelessness, many of whom seek support through statewide shelters and services. This study explored the diversity of ways in which individuals and families moved through Hawai‘i's homeless service system. Using administrative data, a cohort of new service users was tracked across time to trace the developmental trajectories of their homeless service use. The sample consisted of adults who had entered the service system for the first time in the fiscal year (FY) of 2010 (N = 4655). These individuals were then tracked through the end of FY 2014, as they used emergency shelter, transitional shelter, and outreach services. A latent class growth analysis was conducted and identified four distinct patterns of service use: low service use (n = 3966, 85.2%); typical transitional shelter use (n = 452, 9.7%); atypical transitional use (n = 127, 2.7%), and potential chronic service use (n = 110, 2.4%). Multinomial logistic regression models were then used to determine if select demographic, family, background experience (e.g., education, employment), or health variables were associated with class membership. The distinct profiles for class membership are discussed.
  • Article
    Objective: While considerable attention has focused on improving the detection of depression, assessment of severity is also important in guiding treatment decisions. Therefore, we examined the validity of a brief, new measure of depression severity. Measurements: The Patient Health Questionnaire (PHQ) is a self-administered version of the PRIME-MD diagnostic instrument for common mental disorders. The PHQ-9 is the depression module, which scores each of the 9 DSM-IV criteria as "0" (not at all) to "3" (nearly every day). The PHQ-9 was completed by 6,000 patients in 8 primary care clinics and 7 obstetrics-gynecology clinics. Construct validity was assessed using the 20-item Short-Form General Health Survey, self-reported sick days and clinic visits, and symptom-related difficulty. Criterion validity was assessed against an independent structured mental health professional (MHP) interview in a sample of 580 patients. Results: As PHQ-9 depression severity increased, there was a substantial decrease in functional status on all 6 SF-20 subscales. Also, symptom-related difficulty, sick days, and health care utilization increased. Using the MHP reinterview as the criterion standard, a PHQ-9 score > or =10 had a sensitivity of 88% and a specificity of 88% for major depression. PHQ-9 scores of 5, 10, 15, and 20 represented mild, moderate, moderately severe, and severe depression, respectively. Results were similar in the primary care and obstetrics-gynecology samples. Conclusion: In addition to making criteria-based diagnoses of depressive disorders, the PHQ-9 is also a reliable and valid measure of depression severity. These characteristics plus its brevity make the PHQ-9 a useful clinical and research tool.
  • Article
    Full-text available
    This study measures the use and relative importance of different measures of health-related quality of life (HRQOL) as predictors of mortality in a large sample of older US adults. We used Cox proportional hazards models to analyze the association between general self-reported health and three "healthy days" (HDs) measures of HRQOL and mortality at short-term (90-day) and long-term (2.5 years) follow-up. The data were from Cohorts 6 through 8 of the Medicare Health Outcomes Survey, a national sample of older adults who completed baseline surveys in 2003-2005. At the long term, reduced HRQOL in general health and all categories of the HDs were separately and significantly associated with greater mortality (P < 0.001). In multivariate analysis of long-term mortality, at least one HD category remained significant for each measure, but the associations between mental health and mortality were inconsistent. For short-term mortality, the physical health measures had larger hazard ratios, but fewer categories were significant. Hazard ratios decreased over time for all measures of HRQOL except mental health. In conclusion, HRQOL measures were shown to be significant predictors of short- and long-term mortality, further supporting their value in health surveillance and as markers of risk for targeted prevention efforts. Although all four measures of HRQOL significantly predicted mortality, general self-rated health and age were more important predictors than the HDs.
  • Article
    Objectives: We identified risk and need profiles of homeless veterans and examined the relation between profiles and referrals and admissions to Department of Veterans Affairs (VA) homeless service programs. Methods: We examined data from the VA's new Homeless Operations Management and Evaluation System on 120,852 veterans from 142 sites nationally in 2011 and 2012 using latent class analyses based on 9 homeless risk factors. The final 4-class solution compared both referral and admission to VA homeless services. Results: We identified 4 latent classes: relatively few problems, dual diagnosis, poverty-substance abuse-incarceration, and disabling medical problems. Homeless veterans in the first group were more likely to be admitted to the VA's permanent supportive housing program, whereas those in the second group were more likely to be admitted to more restrictive VA residential treatment. Homeless veterans in the third group were more likely to be admitted to the VA's prisoner re-entry program, and those in the fourth group were more likely to be directed to VA medical services. Conclusions: The heterogeneous risk and need profiles of homeless veterans supported the diversity of VA homeless services and encouraged the development of specialized services to meet their diverse needs.
  • Article
    The nine-item Patient Health Questionnaire depression scale is a dual-purpose instrument that can establish provisional depressive disorder diagnoses as well as grade depression severity.
  • Article
    Prevention is critical in any effort to end homelessness. Unfortunately, the practice of homelessness prevention is still in its infancy and there is little science base for its implementation. Risk factors for homelessness have been identified at multiple levels: the individual, institutional, and societal levels. Addressing all three in prevention practice is necessary. The population/high-risk framework is the most appropriate framework for conceptualizing how to design programs and policies to prevent homelessness because it draws attention to the need for direct intervention among those at most risk, and also for modifying the overall context. This review of the literature and technical reports points to a number of strategies that demonstrate preliminary effectiveness or are in need of rigorous evaluations. Reductions in homelessness as a result of targeted, high-risk approaches alone are achievable, but will be short-lived unless low-cost and affordable housing and income are addressed at the population level. Simultaneous implementation and evaluation of both population and high-risk prevention strategies will bring us closer to reaching our goal of ending homelessness.
  • Article
    Full-text available
    This year, over 630,000 prisoners will be released from state and federal prisons across the country—more than four times as many as were released in 1980. In this article, we examine the scope of the prisoner reentry issue—what is known about the intersection of housing, homelessness, and reentry and about the barriers returning prisoners face in securing safe and affordable housing.Although the housing challenges are formidable, progress is being made on numerous fronts. We seek to frame the dynamics of the reentry housing discussion by highlighting the promising strategies that are emerging. These strategies, taken to scale, could help create a very different national policy on prisoner reentry. Ultimately, effective reentry strategies have the potential not only to reduce re‐arrest and increase public safety, but also to reduce homelessness.
  • Article
    The present study used cluster analysis to empirically derive and cross-validate a system for classifying homeless persons on the basis of their service needs. Variables used in the subgroup identification phase of the study included psychopathology, alcoholism, social support, socioeconomic status, and health status. Four subgroups were identified: (1) an economically disadvantaged group, (2) an alcoholic group, (3) a mentally ill group, and (4) a relatively advantaged group. Subsequent analysis revealed few differences between the four groups in background characteristics. Moreover, subgroups were similar in service willingness and utilization, with the exception of utilization and willingness to receive treatment for alcohol and mental health problems. The results suggest the need for policies and services that address the core socioeconomic needs that cut across all subgroups, as well as for additional, specialized services for those with psychiatric or alcohol problems.