Housing choice, outcomes, and neighborhood characteristics in housing programs for individuals with a serious mental illness: analysis of a phoenix survey of individuals with serious mental illness.
Mushkatel, Alvin ; Guhathakurta, Subhrajit ; Thompson, Jackie 等
INTRODUCTION
While research has recently documented the importance of housing
assistance to individuals with a serious mental illness (SMI), research
is inconclusive as to the relative importance of different types of
housing assistance for individuals with serious mental illnesses in
relation to their well-being or integration into the community. More
specifically, the relative importance of individual characteristics,
including their socio-demographic factors, differently structured
housing programs, and the environmental context (neighborhood level
variables) to client success is not well researched. Yet, these factors
are of critical importance in further understanding how various
programs, including housing, can be designed and implemented to help
individuals with serious mental illnesses integrate into the community
(mainstreamed). For example, are there particular client socio-economic
characteristics which interact either positively or negatively and allow
for more successful housing program placements? Are there neighborhood
characteristics, such as a concentration of residents receiving other
types of assisted housing subsidies in a neighborhood, and are these
concentrations associated with differential reintegration outcomes?
Finally, is there an interactive effect among these three types of
factors (individual, neighborhood context, and type of housing program),
and if so, what, and under what conditions?
BACKGROUND
The 1970s brought about the deinstitutionalization era, which was a
movement to discharge individuals with serious mental illnesses from
state mental health facilities in order to reintegrate them into
mainstream society. The shift in how individuals with a serious mental
illness were treated is highlighted by the fact that in 1955 there were
559,000 individuals in state hospitals. Despite the fact that the
nation's population increased by over 100 million from 1955 to
1998, there were only just over 57,000 individuals in state hospitals by
the end of 1998 (Lamb and Bachrach 2001). The deinstitutionalization era
also corresponded to the increased use of psychotropic medications to
control the effects of severe mental illness and facilitated the
discharge of this population from state hospitals. Torrey (1997)
estimated that deinstitutionalization had resulted in 2.2 million
severely mentally ill patients without supportive psychiatric services.
The fact that individuals were more dependent on these psychotropic
medications also increased their dependence on community-based
supportive services.
Deinstitutionalization placed poorly understood demands on both
community-based alternative care for those with a severe mental illness,
as well as government-assisted housing services, therefore resulting in
a lack of community-based support services to meet the needs of this
population. Without the proper support, many of these individuals were
out on the streets, and nationally, the homeless population began to
increase. It is estimated that between one-third and one-half of the
homeless were individuals with a serious mental illness (National
Institute of Mental Health, 2000).
In order to more effectively utilize housing resources, the
Department of Housing and Urban Development (HUD) adopted a strategy of
providing housing programs that would cover the entire housing continuum
of care. These programs would not only provide housing assistance to
those of low and moderate income, but also to the nation's special
populations, including the disabled and individuals with a serious
mental illness (HUD, 1995 ) Additionally, the mental health community
realized that they, too, would need to provide a continuum of mental
health services in communities including case management, medication,
and other services designed to provide preparation and support for those
now living in the community (Fosburg, et al. 1997). In the 1980's
the "continuum approach" came under criticism for not being
able to meet the changing needs of individuals with a serious mental
illness (Dickey, et. al., 1996). The "Supported Housing"
approach, an alternative to the "continuum approach," was
viewed as more effective because it provides a range of housing options
with different levels of staff involvement based on the needs of the
individual (Carling, 1992). Studies designed to compare these two
distinct approaches produced equivocal results, suggesting that other
factors may play a more significant role in predicting effective housing
for individuals with a serious mental illness.
The research in this paper investigates many of these questions
using data collected from a series of studies completed in Phoenix,
Arizona. Before discussing these studies, the methodology employed, and
the research findings, it is necessary to provide a short review of some
of the research findings from the major studies in the field. This
review will further focus the subject of this paper, that is, the
necessity of housing for homeless individuals with a serious mental
illness and the relative importance of various factors in such a
program. This review is followed by a discussion of Arizona policies
aimed at providing greater stability and care for this population and a
brief discussion of the homeless housing programs in Phoenix. Next, an
explanation of the study's methods and major findings are offered,
as well as a discussion of their implications for homeless policy that
addresses individuals with a serious mental illness.
THE RESEARCH ON FACTORS AFFECTING INDIVIDUALS WITH SERIOUS MENTAL
ILLNESS COMMUNITY REINTEGRATION SUCCESS
One of the key findings resulting from the experience of
deinstitutionalization is the critical importance of providing far more
than just medications for those with serious mental illness: including a
host of community based services that take into account the needs and
circumstances of the patients is vital (Lamb and Bachrach, 2001). The
deinstitutionalization movement was based, in part, on the belief that a
community-based, smaller, and less isolated setting for care was
preferential to long-stay psychiatric hospitals (Bachrach, 1996).
Culhane et al. (2002) makes the argument that housing homeless
individuals with severe mental illnesses is both cost effective and more
humane solution to the problem than institutional housing of those with
sever mental.
Despite the fact that access to services and housing has been
defined as a critical component for individuals with severe mental
illnesses to be reintegrated into the community, the research, which
attempts to identify a preferable housing option, varies in the program
characteristics that are attributed to successful reintegration into the
community. Some studies claim that community-integrated living
arrangements contribute to successful reintegration (Hanrahan et al.
2001), while other studies claim that group housing is the preferable
method for housing individuals with severe mental illnesses (Goldfinger
et al. 1999). In a similar vein, Bebout et al. (1997) results suggest
that the continuum of care model, moving from supervised apartments to
independent apartments, contributes to stability of housing for the
formerly homeless with dual diagnosis of severe mental illness and
substance abuse disorder.
Other research concludes that aggressive case management is
accountable for providing stable housing (Goering et al. 1997, Lehman et
al.1997). However, Clark and Rich (2003) examined the difference between
a program with only case management services and one with a
comprehensive housing program, including case management services for
individuals with a dual diagnosis of severe mental illness and substance
abuse disorder. The results indicated that the case management service
program was only effective for those with low or moderate psychiatric
impairment, but for those with high psychiatric impairment, only the
comprehensive housing service program resulted in stable housing.
Rog (2004), in her review of the literature, concluded that the
housing with supports, regardless of the specific model, results in a
significant increase in housing stability for individuals with a serious
mental illness. However, the results of the studies are inconclusive in
identifying the preferable housing method and program characteristics
that truly contribute to stable housing and successful re-integration of
formerly homeless individuals with severe mental illnesses into the
community.
As the understanding of the importance of adequate housing and
community social services for recovery grew, the context in which these
services were delivered was often overlooked. Yet, there was ample
reason to hypothesize that environmental or neighborhood factors would
play an important part in the success or failure of community
integration for individuals with a serious mental illness. It is
possible that seriously mentally ill patients who are in the community
setting will be influenced by the neighborhood and the conditions
surrounding their housing.
The impact environmental context can have on individuals with a
serious mental illness has been recognized as an important factor but
not in the context of housing programs designed for individuals with a
serious mental illness. For example, Silver (2000a,b) used data from the
MacArthur violence study to better understand the relationship between
neighborhood context and violence among patients discharged from
in-patient psychiatric clinics. He hypothesized that the lack of control
variables measuring neighborhood context would explain the inconsistency
of past studies that had attempted to use race to explain the likelihood
of violence. His study found that when neighborhood disadvantage was not
controlled for, the effect of race on predicting violence was grossly
overestimated. Knowing this, it could explain why Goldfinger et al.
(1999) found that "consumers who were African American or Hispanic
experienced disproportionately more days homeless if--but only if they
were assigned to independent housing." His study failed to consider
the environmental context of the neighborhood. Swanson et al. (2002)
also found that violence in the surrounding area was one predictor of
violent behavior. In short, neighborhood context is shown to be an
important factor when predicting behavioral outcomes of individuals who
are severely mentally ill.
THE IMPORTANCE OF CONTEXT
The provision of housing for persons with a serious mental illness
has become an integral element in the continuum of care. In addition, it
is clear from the studies that some form of housing program (supervised
living, supportive housing etc.) is associated with positive outcomes,
including greater stability, higher quality of life, less likelihood of
homelessness, and less use of clinical services. Yet, the importance of
the environment (neighborhood context) that surrounds the housing has
been largely ignored in such studies until recently. It is important to
emphasize that without the knowledge of the surrounding environment, we
can not attribute various behavioral outcomes to only the housing
program since either the neighborhood environment, or the individual
risk factors, or both may be causing the observed outcomes. Indeed, a
recent paper suggests that the inconsistency of research results that
observed the impact of different housing programs may be the result of
the failure to account for the neighboring environment (List, 2004).
Increasingly there is a recognition of the importance of examining the
impact of the environmental context. For example, The National Institute
of Mental Health (NIMH) identified contextual influences on mental
illness and its care as a priority area for research (NIMH, 2000). It
argued that both individual and contextual effects "... need to be
identified and assessed to aid in designing and/or developing
interventions appropriate to the needs and circumstances of specific
individuals or groups suffering from mental disorder" (NIMH, 2000).
To date, such research has been limited (Wong et al., 2002; Newman,
2001; Newman et al., 1994; Guhathakurta and Mushkatel, 2000, 2002).
The importance of neighborhood context was noted as early as Faris
and Dunham's (1939) study examining Chicago neighborhoods to
determine the prevalence of mental illness. By examining the geographic
spatial distribution of psychiatric patients prior to hospitalization,
they documented an association between the nature of the mental illness
and the neighborhood. Those patients suffering from schizophrenia and
substance abuse disorders were far more likely to live near the city
center, regardless of their race. Faris and Dunham (1939) suggested that
these city center neighborhoods were more likely than other
neighborhoods to be characterized by disorganization and deterioration,
the very conditions they felt supported social isolation confusion and
other conditions associated with mental disability.
In the 1980s, a number of researchers attempted to use neighborhood
contextual factors to predict psychological impairment; however, these
studies resulted in mixed findings. For example, Gray et al. (1983)
found no relationship between environmental context and mental
well-being, but Nagy et al. (1988) discovered a relationship between
urban location and productive activities. As List (2004) notes, these
studies were limited by their rudimentary measures of environmental
context (type of neighborhood--residential or commercial, and locale
urban, suburban or rural). The nature of the neighborhood context is not
well operationalized by such efforts at measurement.
More germane are the studies by Ross (2000) and Ross et al. (2000),
which found neighborhood context to be an important variable in
determining behavioral outcomes in the general population and depression
in adults. These studies found that those who live in neighborhoods with
a high percentage of poor and mother-only households display higher
levels of depression than those living in areas without these
characteristics. In addition, one study found that while neighborhood
factors were not as important as individual factors, they still could
not be disregarded because in some cases they were found to be more
important than many individual factors in predicting the prevalence of
schizophrenia and substance abuse disorder (Goldsmith, et al. 1998).
Finally, Silver et al.'s(2002) analysis of 11,686 participants
residing in 261 census tracts revealed that neighborhood disadvantage
was associated with higher rates of major depression and substance
abuse, while neighborhood residential mobility was associated with
higher rates of schizophrenia, major depression, and substance abuse.
Neighborhood context can readily be seen as a potentially powerful
factor in influencing the success or failure of various community-based
programs for individuals with serious mental illnesses. For example, a
national report in 2002 clearly notes that social segregation of the
mentally ill hinders their recovery (Onken, et al. 2002). The goal of
various independent housing programs for individuals with a serious
mental illness is to place these individuals in a normalized
environment. In a similar vein, HUD requires public housing units to be
properly distributed within an area and among the population (HUD 1999).
Yet, despite goals for a balanced spatial distribution of such housing,
Guhathakurta and Mushkatel (2000) found that HUD's housing programs
(including Section 8 vouchers) and housing for individuals with a
serious mental illness were very concentrated in the city of Phoenix.
For example, 31 percent of the Section 8 housing was distributed among
only 10 census tracts, and 15 census tracts contain a little more than
50 percent of all city-administered subsidized housing (ibid. :525).
These tracts were found to have a higher concentration of minorities,
fewer college-educated persons, lower median household incomes, and a
higher percentage of renters than is the average for the city. Indeed,
these 15 tracts had more than twice as many non-whites and fewer than
one-half the average number of college-educated than the city average.
Finally, these tracts were concentrated in the inner-city areas of
Phoenix.
In comparing the Shelter Plus Care housing (S+C SMI housing-
independent living arrangements) to these city run programs,
Guhathakurta and Mushkatel (2000) found a similar spatial concentration
of the SMI housing, albeit slightly less concentrated than the City/HUD
programs. Just 10 tracts in the city accounted for 29 percent of all S+C
housing in the metropolitan area (ibid. :529). Based on their regression
and binomial logit analysis, they find that the traditional
city-administered housing programs and the S+C for individuals with a
serious mental illness tend to be located in similar neighborhoods that
are highly associated with the number of non-whites in the tract. They
conclude that assisted housing is concentrated in similar locations and
that the presence of one type of this housing is a strong predictor of
the presence of the other. Hence, current policy seems to result in
increasing concentrations of stigmatized populations in the city
centers.
It is important to note, housing location in relation to support
services is also important to account for when considering environmental
context. Wong and Solomon (2002) note that access to services is
critical for individuals with a serious mental illness if they are to
maintain housing within a normalized community. This is an additional
factor that must be considered when examining environmental context.
These findings are disheartening in several respects. First, a
major portion of the program logic underlying independent living
arrangements for individuals with a serious mental illness is based on
the understanding of placing them in a normalized environment. However,
even the limited analysis by Guhathakurta and Mushkatel (2000) indicates
that the neighborhoods in Phoenix where such housing is located are far
from average, and that they are in areas where a large number of
households receive some form of housing assistance. Second, because
these neighborhoods are disadvantaged, it is possible that they are also
characterized by higher crime rates, which would suggest that
individuals with a serious mental illness are more likely to become
victims of such crimes (Brekke et al. 2001; Silver, 2000a, b). Finally,
the research in this area raises more questions than it answers. We
address some important gaps in the literature by analyzing the relative
importance of socio-demographic, neighborhood context and housing
program factors leading to successful reintegration of individuals with
a serious mental illness to the community. Before addressing these
important issues, a discussion of the Phoenix and Arizona policy history
for individuals with a serious mental illness is necessary, along with a
discussion of the current study.
THE PHOENIX INDIVIDUALS WITH SERIOUS MENTAL ILLNESS CONTEXT
In 1988 the Maricopa Association of Governments (MAG) created a
Homeless Task Force, which produced a report on conditions that would
alleviate the homeless crisis in the county and return these individuals
to self-sufficiency (MAG, 1989). This report was followed by another
report in 1998 and reports from the State Homeless Coordination Office
of the Arizona Department of Economic Security (DES), which estimated
the number of homeless individuals with a serious mental illness and
assessed their needs.
As a result of court settlements (Arnold versus Arizona Department
of Health Services, 160 Ariz. 1989,) Maricopa County and the Arizona
Department of Health Services through a private entity are now the
principal providers of housing for individuals with a serious mental
illness. In fact, the court established a monitor to oversee the
provision of services to individuals with a serious mental illness and
to provide frequent reports to assure adequate levels of services were
being provided to these individuals. Partially due to the Arnold case, a
comprehensive study was initiated in 1998, it found that to establish a
comprehensive mental health system within the county would cost around
$293 million. While the Arizona legislature subsequently allocated
considerable funding to meet this goal, the State has still not resolved
its problems in providing adequate services to the individuals with a
serious mental illness.
In 2002, 1000 formerly homeless individuals were housed using funds
from the Stewart McKinney Act, and another 1000 formerly homeless
persons diagnosed with a serious mental illness had moved from McKinney
funded housing to Section 8 permanent housing (DES, 2002). In 2003, the
Arizona Department of Health Services' Division of Behavioral
Health produced its Strategic Plan for Housing for Maricopa County for
Individuals with a Serious Mental Illness. The report notes that in the
last eight years the resources for such housing have doubled, and the
agency restates its commitment to a residential continuum in which
community-based services are provided throughout a range of residential
settings that vary by the amount of independent living arrangements and
services provided (ADHS, 2003). The report also notes that Supported
Housing (independent housing) was developed to afford individuals with a
serious mental illness a choice in their housing situation, while
providing necessary community-based services. The plan also states that,
to date, there has been no research rigorously comparing the outcomes of
varying housing approaches. These independent or supported housing
approaches were the focus of the SAMHSA/CMHS Housing Initiative Study, a
nation-wide study that started in 1997-98 and in which Arizona
participated and provided a portion of the data utilized in this
research.
THE SAMHSA/CMHS HOUSING INITIATIVE STUDY
Client data from individuals with a serious mental illness was
collected as part of a larger multi-site research project funded by HUD
and the Substance Abuse and Mental Health Services Administration
(SAMHSA). The Arizona Department of Health Services' Division of
Behavioral Health directed the Community Mental Health Study (CMHS) and
assembled three multidisciplinary research teams to implement the
evaluation using several different methods. One team conducted an
evaluation of two housing programs designed for individuals with a
serious mental illness. The programs varied by the degree of
independence they had. The Supported Housing (SH) program allows far
more choice in housing situation and services than does the other
Supervised Independent Living (SIL) program.
A total of 185 individuals participated in the study and were
interviewed within two weeks of moving into their housing (baseline),
and again at three, six and tweleve months (Community Rehabilitation
Division, 2003). The assignment or referral to either "... the SH
or SIL program was made by the consumer and the clinical team lead by
the Regional Behavioral Health Agency case manager (ibid. : 7).
Unfortunately, random assignment of participants proved to be impossible
and the number of participants in each program is strongly over-weighted
to the SH program. In addition, it is important to note the shift to a
for-profit RBHA, along with other important changes that took place
shortly after the study started (ibid. :8). Additionally an 18-month
interview was added to the protocol after the study began. The data from
several of these points in time is used in this paper to examine the
interaction between individual characteristics, different housing
programs, and neighborhood factors in explaining the success of
individuals with a serious mental illness in integrating into the
community.
DATA AND METHOD
This study of housing outcomes of SMI population in Phoenix is
based on a survey of 185 severely mentally ill and homeless individuals
who were counseled and housed under two different SMI housing programs
between (month) 2000 and (month) 2001 in Maricopa County. The
respondents who participated in the survey were contacted at the time
they were initially housed and after each 3-month interval up to 18
months since the initial contact if they continued to be in one of the
SMI housing programs. This study is based on a subset of the entire data
as it extracts information for only those respondents who were initially
housed in the city of Phoenix. In addition, the study only uses the
surveys conducted at baseline, 6-month, and at 12-month intervals. The
data for Phoenix consisted of 136 initial participants. This number
declined to 113 after 6 months and further to 108 after 12 months due to
participants dropping out of the programs. The gender, racial, and age
composition of the participants at baseline, after 6 months and after
12-months are provided in Table 1.
As may be expected, the respondents have a range of physical and
psychological disorders that are associated with both legal and illegal
drug use. A significant majority of the interviews at baseline (42.2%)
were conducted in substance abuse treatment program facilities. Another
35.6 percent of the interviews at that time were conducted at mental
health program facilities. Also, almost 90 percent of the respondents
had been hospitalized at least once during their lifetime. About
three-quarters of the interviewees have received health benefits from
Medicare and/or Medicaid programs. A typical Phoenix participant in the
study had initially availed of mental health services in the early 20s
and subsequently been hospitalized for emotional and psychological
disorders about four times. Few admitted to current drug use (about 5%
used cocaine or crack in the current month and only 6% had some drug
problem).
The overall purpose of the multi-site study was to examine the
effectiveness of various types of housing for the SMI population.
Specifically, it was designed to evaluate whether the supported housing
program provided better housing outcomes for the SMI population than
other comparable programs. An extensive questionnaire was used to
collect information from the respondents. The respondents were asked
questions related to their residential history, housing satisfaction,
livelihood, health and functioning, quality of social support, choice
and empowerment, among others. These interviews were conducted
"face to face" at the interviewees housing site or close to it
(such as in the office of the apartment complex). The random allocation
of housing failed due to several extenuating circumstances resulting in
a significant majority of the respondents being housed in supported
housing. Less than 20 percent of them were housed through the
comparative housing program.
The focus of this paper is different from the multi-site study. In
this paper we seek to determine the relative importance of individual
and neighborhood factors in determining housing outcomes for the SMI
population. We hypothesize that the quality of the neighborhood in terms
of housing quality, demographics, incidence of crime, and accessibility,
among others, are important factors determining housing satisfaction
among the study participants after controlling for the individual
characteristics of the participants. The study also examines the
relative influence of each of these neighborhood characteristics in
determining housing outcomes of SMI population in Phoenix. Finally, the
paper discusses the policy implications for better designing housing
programs for the SMI population.
A variety of sources were tapped to access information on
neighborhood context that were subsequently integrated with the data
from SMI housing survey discussed earlier. Information about household
and housing characteristics in the census tracts were SMI respondents
were housed was obtained from 2000 census of population and housing. The
transit accessibility of the neighborhoods was calculated from the
distance of tract centroids to the nearest bus route. Data on crime
incidence were obtained from the City of Phoenix Police Department at a
finer resolution than the census tract. This information was abstracted
to the level of the census tract with the help of spatial matching
performed in Arcview GIS software package. In addition, information on
spatial concentrations of various assisted and supported housing within
the tracts was also obtained from several sources. Table 2 provides a
listing of the most important variables used in this study together with
their sources.
The SMI housing provided for the survey participants were located
in 50 census tracts out of a total of 327 tracts (15%) within the city
of Phoenix. Among the tracts that included SMI housing, only 12 census
tracts housed 52 percent of the respondents. Figure 1 shows the census
tracts that include one or more housing as provided to the participants
of this study. Table 3 provides some demographic and housing
characteristics of the 12 census tracts where majority of the
participants have been housed.
[FIGURE 1 OMITTED]
The analysis of the SMI population's perceived quality of life
in relation to their individual characteristics and the characteristics
of the neighborhood in which they were housed follows a two-pronged
approach. First, the strength of association between perception of
quality of life of the respondents and other outcome measures such as
housing satisfaction, empowerment, and service need are determined
through non-parametric statistics. These associations are also examined
after controlling for a number of individual and neighborhood attributes
to see whether such attributes mediate the association between quality
of life and the other three outcome variables. Second, the unique
contribution of the various attribute variables (neighborhood and
individual) after controlling for other explanatory variables is
established through ordinal regression analysis. The generic form of
this model is provided below:
Link ([y.sub.ij]) = [[theta].sub.j] - [[[beta].sub.1] [x.sub.i1] +
... [[beta].sub.n][x.sub.in]]
Where: Link([y.sub.ij]) = log(-log(1 - y)) this is the
complementary log-log (extreme value) function
[[theta].sub.j] = the threshold for the jth category
[[beta].sub.1] .. [[beta].sub.n] = regression coefficients
[x.sub.i1] .. [x.sub.ip] = values of the predictors for the ith
case
The complementary log-log link function was determined to be the
most appropriate given the responses to "feeling of life in
general" were skewed towards the higher values. The interpretation
of the model is based upon several assumptions. First, the model should
suggest a latent continuous variable, which is discretized by the j
ordinal categories. Second, the constants in the model are only
determined by the category's probability of being predicted
(without the contribution of the independent variables). Third, the
prediction part of the model depends only on the predictors. The second
and third assumptions guarantee that the results will be a set of
parallel planes (or lines), one for each category of the outcome
variable. The frequencies and marginal percentages of all the outcome
variables are presented in Table 4.
ANALYSIS
Among the various outcomes-related questions that were asked of SMI
respondents in the survey, the perception of their overall quality of
life is perhaps the most robust. This question was asked twice during
the survey; once at the beginning of all other perceptual questions and
again at the end of the survey. The tests of association including chi
square and sommers'd are highly significant suggesting strong
relationship between the two answers on a similar question as may be
expected. However, the strength of the relationship is marginally higher
than 60% (tau-b and somers' d) indicating that the two responses
are less than perfectly aligned. Further tests showed that the responses
received at the end of the survey had higher variation and provided a
better outcome measure of "quality of life" than the initial
response.
The analysis of the determinants of quality of life of the surveyed
SMI population shows that overall housing satisfaction, feeling of
empowerment and overall need for service are all significant variables
explaining quality of life (Table 5). Among the three indicators noted,
the feeling of empowerment had the strongest and the most significant
impact on respondents' quality of life. The somers' d
statistic (.567) suggests a 57 percent reduction in error when
predicting quality of life after taking overall feeling of empowerment
into consideration. The corresponding reduction in error for predicting
quality of life accounting for the other two explanatory variables,
housing satisfaction and service need, are 19 percent and 17 percent
respectively. Also of note is the expected negative association between
quality of life and service need. That is, respondents who indicated
higher need for service also scored low on their assessment of
"quality of life".
While it is clear that quality of life is affected by the
respondents' attitude towards their housing, feeling of
empowerment, and their need for service, this relationship is mediated
by a number of individual and neighborhood attributes. We control for
several of such attributes to determine which among them strengthen or
weaken the initial associations between quality of life and the three
explanatory variables described earlier. The type of housing program,
supervised independent living (SIL) or supported housing (SH), is an
important policy variable that was factored in to the relationship
between housing satisfaction and quality of life. As shown in Table 5,
this relationship is strong and significant for individuals in SILs,
while for the respondents in the SH program this relationship was not
significant. In contrast, the SMI population in supported housing showed
a stronger association between their feeling of empowerment as well as
their service need with their perception of quality of life. The
corresponding statistics for those in supervised independent living
situations did not have a high level of significance. Hence it can be
concluded that survey respondents in supported housing did not associate
quality of life with their housing satisfaction but with their feeling
of empowerment. However, if they had high service needs, their
perception of quality of life was low.
Four neighborhood attribute variables were also tested to determine
their effect on the relationship between quality of life and the
explanatory variables in Table 5. These are; 1) concentration of
subsidized housing of any type in census tract; 2) median household
income in census tract; 3) the incidence of drug related crimes in the
tract; and 4) housing quality in the census tract as determined through
a survey conducted between June and August, 2004. Curiously, SMI
population housed in neighborhoods with high or somewhat high
concentrations of other forms of subsidized housing showed a greater
propensity to associate housing satisfaction, overall empowerment, and
service need with quality of life. Those living in census tracts with
low and somewhat low levels of subsidized housing concentrations did not
associate quality of life with any of these three explanatory variables.
This suggests that the perception of quality of life is sensitive to the
individuals' sentiments about their own condition when they are
living in census tracts with moderate to high concentrations of
subsidized housing. This is also reflected in the census tracts with low
or somewhat low median household incomes. Those living in neighborhoods
with higher median incomes did not associate their quality of life with
the attitudinal variables reported here.
The contribution of incidence of drug crimes in the census tract is
significant in predicting the association between quality of life and
the explanatory variables discussed. Except for the tracts with low
levels of drug crimes, all other levels had some impact on one or more
of the explanatory variables and their ability to predict quality of
life. Curiously, there is a negative relationship between quality of
life and housing satisfaction in tracts with somewhat high levels of
drug crime. This is perhaps an aberration given that this relationship
is significant and positive for both high and somewhat low levels of
drug crimes in the census tract. The relationship between quality of
life and overall empowerment, as well as, between quality of life and
service need is as expected for somewhat high levels of drug crimes in
the neighborhood. That is, in these neighborhoods, SMI populations
indicating high overall empowerment also registered highly on quality of
life scale. On the other hand high service need reduced their quality of
life.
Housing quality had an expected impact on the relationship between
quality of life and housing satisfaction. At high levels of housing
satisfaction, this relationship is significant and positive but at lower
levels of housing quality the association between quality of life and
housing satisfaction is not significant. Although, overall feeling of
empowerment does contribute highly to the respondents' quality of
life, this relationship is significant only at medium levels of housing
quality in tract. At low and high levels of housing quality, the
magnitude of association between the two outcome variables is high and
positive but not significant. Service need is negatively associated with
quality of life in census tracts with high levels of housing quality.
However, this relationship is less apparent in census tracts with medium
and low levels of housing quality.
In addition individual level variables were examined and found to
be statistically insignificant in mediating the relationship between
quality of life and other explanatory variables than did neighborhood
level variables. Gender did not feature as a significant variable as
both men and women had similar low levels significance in predicting any
of these relationships tested. Similarly, Hispanics and non-Hispanics
had no impact on the association between quality of life and the three
explanatory variables. However, racial differences can be detected when
Whites and non-Whites are separated out. SMI respondents who are White
showed a strong propensity to relate quality of life with housing
satisfaction and overall empowerment, but not service need. For
non-Whites, none of these relationships was significant.
The other two important individual attributes accounted for were
personal health, and whether the respondent reported being victimized
during the past six months. SMI individuals who rated their health as
excellent, fair, or poor, did not significantly attribute quality of
life with housing satisfaction, overall empowerment , or service need
(except for some weak relationships for those with poor health). In
contrast, those who were victimized attributed their quality of life
significantly to their overall empowerment. The non-victims did not make
any significant association between empowerment and quality of life.
From this analysis of nonparametric statistics of association
between quality of life, housing satisfaction, overall empowerment and
service need a number of significant conclusions can be drawn. First,
the program types have a significant role to play in predicting quality
of life and its relationship to the variables included in this analysis.
Second, neighborhood variables such as median household incomes,
concentrations of other subsidized housing, and housing quality also
contribute to enhancing the relationship between quality of life,
housing quality, empowerment, and service need. Finally, individual
variables are less relevant in mediating the relationship between
housing quality and the three other explanatory variables. Given the
impact of the neighborhood and the program type in this analysis, we can
safely conclude that they are the key attributes of a well designed
policy for successfully housing the severely mentally ill population.
Ordinal regression results presented in Table 6 provide additional
evidence of the strength of contextual variables in explaining
respondents "feeling about life in general". The fit of the
model is fairly good as indicated by the change in chi-square of the
final model over intercept only model and the pseudo R-square values
presented in Table 6. More importantly, the prediction of this model for
the two categories "good" and "mixed" of the
dependent variable "feeling about life in general" is 80
percent and 64 percent respectively. The prediction of the lowest
category of "bad" is, however, lower at 46 percent. The
coefficients of the explanatory variables do not present an intuitive
explanation similar to OLS but the direction and significance of these
variables are important indicators to note.
The reduced form model presented in this study does not include
several of the individual variables dealing with age, gender, race, and
ethnicity because these variables were not significant enough in
explaining "feeling about life in general". The only
self-perception variables that were significant were the
respondents' feeling of overall empowerment and their attitude
about their personal health. Their perception about their own health has
an expected impact on their overall feeling of "good",
"mixed" or "bad", that is, better health is
associated with higher levels of quality of life holding all else
constant. However, the attitude towards empowerment is more complex.
Respondents who felt somewhat empowered seem to also have a higher
propensity for a lowered sense of feeling "good" about their
quality of life, ceteris paribus. This relationship disappears for those
who have stronger feeling of overall empowerment.
The variables representing the neighborhood contextual attributes
of the respondents are more numerous in this model. One significant
result is that in the census tracts with lowest concentrations of
existing subsidized housing leads to higher likelihood of low
satisfaction with quality of life, holding all else constant. This
relationship is, however, not significant at "somewhat low",
"somewhat high" concentrations of subsidized housing. Also,
those respondents housed in the supervised independent living program
housing seemed to have a lower propensity to feel good about their
"life in general". This suggest that the supported housing
programs are probably better able to provide for the housing needs of
the SMI/Homeless population in Phoenix when compared to the supervised
independent living arrangement. The other neighborhood quality variables
such as percent of city's median rent, percent unemployed in census
tract, percent of African Americans and percent of detached
single-family homes are also significant. Higher proportions of
unemployed, single family detached homes and African Americans increase
the likelihood of lower feelings about quality of life. The strong
negative relationship of respondents' quality of life at low and
somewhat low levels of housing satisfaction with quality of life is also
a significant finding. Higher likelihood of being low or somewhat low on
levels of housing satisfaction is associated with lower likelihood of
feeling "good" about "life in general". This
reinforces the significance and direction of the relationship between
housing satisfaction and quality of life found earlier in non parametric
tests.
The overarching conclusion of the analyses of nonparametric
statistics and ordinal regression is that context matters. Several
contextual variables measuring attributes of the neighborhood (defined
by a census tract) are significant and strong predictors of SMI
respondents' "feeling of life in general". Notable
findings with special significance for policy are the contributions of
two variables, program type and concentration of subsidized housing in
census tract in explaining respondents' quality of life. Clearly,
supervised independent living arrangements fared poorly when compared to
the supported housing program. There is evidence to suggest that
lowering concentrations of subsidized housing in a neighborhood may have
a threshold beyond which satisfaction levels are lowered rather than
elevated. This is perhaps an artifact of neighborhoods with low levels
of subsidized housing concentrations being also associated with low
levels of social contact and accessibility. Similarly, neighborhoods
(census tracts) with high levels of single-family detached housing also
contributed to low levels of satisfaction with quality of life, all else
being constant. Therefore the picture that appears from this analyses is
a complex one, in which, the preferred neighborhoods have higher rents,
lower unemployment rates, and lower concentrations of racial minorities
(especially African Americans) but also has fewer single family detached
homes and moderate concentrations of subsidized housing. Further
research is necessary to determine the observed threshold effects in
these contextual variables.
CONCLUSION
The analysis has confirmed our initial hypothesis concerning the
importance of neighborhood contextual variables in understanding SMI
clients' perceptions of their quality of life. Not only has the
importance of the different housing programs impacts on SMI perceived
quality of life been demonstrated, but in several instances these
programmatic variables seem to interact with neighborhood factors and
feelings of empowerment, service need, and housing satisfaction to
either increase or decrease clients' perceived quality of life.
Individual level factors have been shown to have much lower levels of
impact in mediating the relationship between perceived quality of life
and the other explanatory variables, including housing satisfaction,
empowerment and service need, than did neighborhood variables. Two
individual level factors that did emerge as having some importance were
whether a respondent had been the victim of a crime in the last six
months, and their own perceived health.
The ordinal regression performed clearly indicates that housing
program impacts clients' perceived quality of life with supported
housing clients having higher satisfaction with their quality of life
than do participants in the supervised independent living program. In
addition, the concentration of assisted housing in a tract is a
significant neighborhood factor associated with perceived quality of
life. However, the direction of some of these neighborhood contextual
effects is not in the hypothesized direction. For example, lower
concentrations of subsidized housing in an area are associated with
lower perceptions of quality of life among respondents.
The key findings from this study for policy-makers, however is that
housing satisfaction, and the neighborhood contextual variables
associated with it, are also associated with higher perceived quality of
life. Furthermore, program type and concentration of subsidized housing
are notable factors in explaining respondents' perceived quality of
life. There is evidence in the data that there may be a threshold effect
with regard to lower levels of assisted housing concentrations below
which satisfaction with quality of life is actually lowered. We have
suggested that this may either be a result of lowered access to needed
services in such areas or increased social isolation in these
neighborhoods. These findings suggest that policies designed to house
the SMI population in the community can be better designed than they are
currently. When implemented, such continuum of care policies must take
neighborhood factors into consideration as clients and program
implementers select housing. Indeed, these factors appear to be far more
important than most individual factors in explaining the SMI's
perceived quality of life. Additionally, some neighborhood factors were
found to be important only in the more independent living supported
housing program, and not in the SIL program. Hence, the selection of
housing location in neighborhoods is particularly critical for
recipients of supported housing. Our analysis strongly suggests that
current practices can be improved with additional knowledge of how these
neighborhood factors interact with programmatic and individual factors.
It also suggests that policies not accounting for such neighborhood
factors are far more likely to fail than those that do.
ACKNOWLEDGEMENTS
This study was funded in part by a cooperative agreement from the
U.S. Department of Health and Human Service, Substance Abuse and Mental
Health Service Administration, and the Center for Mental Health Services
(grant no. 2 UIE52060-02). Additional funding for this research was
provided by the Neighborhood Service Administration of the City of
Phoenix.
REFERENCES
Arizona Department of Economic Security. (2002). Annual Report on
Homelessness.
Arizona Department of Health Services, Division of Behavioral
Health. (2003). Strategic plan for housing for Maricopa County for
individuals with a serious mental illness. Unpublished and available
through ADHS.
Bachrach, L. (1996). Deinstitutionalization: Promises, problems,
and prospects, in mental health service evaluation. Cambridge, England:
Cambridge University Press.
Bebout, R., Drake, R., Xie, H., McHugo, G., and Harris, M.. (1997).
Housing status among formerly homeless dually diagnosed adults.
Psychiatric Services 48:936-941.
Brekke J., Prindle, C., Bae, S., and Long, J.. (2001). Risks for
individuals with schizophrenia who are living in the community.
Psychiatric Services 52:1358-1366.
Carling, P.J. Housing, community support, and homelessness:
Emerging policy in mental health systems. New England Journal of Public
Policy, 8, 281-295.
Clark C., and Rich, A. (2003). Outcomes of homeless adults with
mental illness in a housing program and in case management only.
Psychiatric Services 54:78-83.
Community Rehabilitation Division. (2003). Final report for grant
entitled: Housing approaches for persons with serious mental illness.
Arizona Department of Health Services, Behavioral Health Services.
Culhane D., Metraux, S., and Hadley, T. (2002). Public service
reductions associated with placement of homeless persons with severe
mental illness in supportive housing. Housing Policy Debate 13:107163.
Dennis, D., Buckner, J., Lipton, F.R., and Levine, L.S. (1991). A
decade of research and services for homeless mentally ill persons: Where
do we stand? American Psychologist 46:1129-1138.
Dickey, B., Gonzalez, O., Latimer, E., Powers, K., Schutt, R., and
Goldfinger, S.M. (1996). Use of mental health services by formerly
homeless adults residing in group and independent housing. Psychiatric
Services, 47, 152-158.
Faris R., and Dunham, H. (1939). Mental Disorders in urban areas:
An ecological study of schizophrenia and other psychoses. New York NY:
Hafner Publishing Co.
Fosberg, L., G. Locke, and M. Finkel. (1997). National evaluation
of the shelter plus care program. Washington, DC: U.S. Department of
Housing and Urban Development.
Goering P., D. Wasylenki, S. Lindsay, D. Lemire, and A. Rhodes.
(1997). Process and outcome in a hostel outreach program for homeless
clients with severe mental illness. American Journal of Orthopsychiatric
67:607-617.
Goldfinger S., Schutt, R., Tolomoiczenko, G., Seidman, L., Penk,
W., Turner, W. and Caplan, B. (1999). Housing placement and subsequent
days homeless among formerly homeless adults with mental illness.
Psychiatric Services 50:674-679.
Goldsmith H., Holzer, C., and Manderscheid, R. (1998). Neighborhood
characteristics and mental illness. Evaluation and Programming Planning
21:211-225.
Gray L., Goldsmith, H., Livieratos, B. and Dupuy, H. (1983).
Individual and contextual social status contributions to psychological
well-being. Sociology and Social Research 68:79-95.
Guhathakurtha, S. and Mushkatel, A. 2002. Race, ethnicity and
household characteristics of section 8 housing clients and their impact
on adjacent housing quality. Urban Affairs Review 37:521-542.
Guhathakurtha, S. and Mushkatel, A. (2000). Does locational choice
matter? Housing programs in Phoenix, Arizona. Urban Affairs Review
35:520-540.
Hanrahan P., Luchins, D., Savage, C. and Goldman, H. (2001).
Housing satisfaction and service use by mentally ill persons in
community integrated living arrangements. Psychiatric Services
52:1206-1209.
Lamb R., and Bachrach, L. (2001). Some perspectives on
deinstitutionalization. Psychiatric Services 52:1039-1045.
Lehman A., Dixon, L., Kernan, E., DeForge, B., and Postrado, L.
(1997). A randomized trial of assertive community treatment for homeless
persons with severe mental illness. Arch Gen Psychiatry 54:1038-1043.
List, J. (2004). Understanding environmental context: Development
of a conceptual model to measure the influence of neighborhood context
on housing for the homeless severely mentally ill. Unpublished
professional project report. School of Planning and Landscape
Architecture, Arizona State University.
Maricopa Association of Governments. (1989). A regional partnership
to end homelessness in the valley of the sun. Homeless Task Force
Report.
Maricopa Association of Governments. (2003). Regional plan to end
homelessness. Regional Continuum of Care Committee on Homelessness.
Nagy, M., Fischer, G., and Tessler, R. (1988). Effects of facility
characteristics on the social adjustment of mentally ill residents of
board and care homes. Hospital and Community Psychiatry 38:1281-1286
National Institute of Mental Health. (2000). Priority areas for
behavioral translation research. (Online)
Available:www.nimh.nih.gov/tbsia/priority.cfm [10-14-02].
Newman S. 2001. Housing attributes and serious mental illness:
Implications for research and practice. Psychiatric Services
52:1309-1317.
Newman S., Rechocsky, J., Kaneda, K. and Hendrick, A. (1994). The
effects of independent living on persons with chronic mental illness: An
assessment of the section 8 certificate program. The Milbank Quarterly
72:171-199.
Onken S., Ridgeway, P., Dornan, D. and Ralph, R. (2002). Mental
health recovery: What helps and what hinders? A National Research
Project for the Development of Recovery Facilitating System Performance
Indicators: Phase One Research Report.
Ridgeway, P. and Rapp, C. (1997). The active ingredients of
effective supported housing: A research synthesis. Unpublished
manuscript. School of Social Welfare, The University of Kansas.
Rog, D.J. (2004) The Evidence on Supported Housing. Psychiatric
Rehabilitation Journal, 37, 334-343.
Ross, C. (2000). Neighborhood disadvantage and adult depression.
Journal of Health and Social Behavior 41:177-187.
Ross, C., Reynolds, J. and Geis, K. (2000). The contingent meaning
of neighborhood stability for residents' psychological well-being.
American Sociological Review 65:581-597.
Silver, E. (2000)a. Extending social disorganization theory: A
multilevel approach to the study of violence among persons with mental
illness. Criminology 38:1043-1074.
Silver, E. (2000)b. Race, neighborhood disadvantage, and violence
among persons with mental disorders: The importance of contextual
measurement. Law and Human Behavior 24:449-456.
Silver, E., Mulvey, E. and Swanson, J. (2002). Neighborhood
structural characteristics and mental disorder: Faris and Dunham
revisited. Social Science and Medicine 55:1457-1470.
Spaniel, L. (2001) Recovery from psychiatric disability:
Implications for Rehabilitation Counseling Education. Rehabilitation
Education, 15, 167-175.
Swanson, J., Swartz, M., Essock, S., Osher, F., Wagner, R.,
Goodman, L., Rosenberg, S., and Meador, K. (2002). The
Social-Environmental Context of Violent Behavior in Persons Treated for
Severe Mental Illness. American Journal of Public Health, 92(9).
Torrey, E. (1997). Out of the shadows. New York, NY: John Wiley and
Sons Inc.
U.S. Department of Housing and Urban Development. (1995).
Supportive housing application: Moving toward a continuum of care. HUD
Document No. HUD-40076 (2/95). Washington, DC: Author.
U.S. Department of Housing and Urban Development. (1999).
McKinney-Vento Homeless Assistance Act Part 576, SubpartH, Sec.882.803:
Section 8 Moderate Rehabilitation Single Room Occupancy Program for
Homeless Individuals. (Online) Available:
http://www.hud.gov/offices/cps/homeless/rulesandregs/regulations/
882sro/882803.
Wong, Y. and Solomon, P. (2002). Community integration of persons
with psychiatric disabilities in supportive independent housing: A
conceptual model and methodological considerations. Mental Health
Services Research 4:13-28.
ALVIN MUSHKATEL
SUBHRAJIT GUHATHAKURTA
JACKIE THOMPSON
KATHY THOMAS
Arizona State University
MICHAEL FRANCZAK
Executive Vice President Marc Center
Table 1
Age, Gender, Race, and Ethnicity of the SMI survey
Participants
Baseline 6-Month 12-Month
Percent Female 38.5% 36.3% 38.9%
Percent White 67.4% 69.9% 72.2%
Percent Hispanic 15.6% 15.0% 13.9%
Median Age 44 44.5 45
Table 2
Sources of Variables Used in this Study
Variables Source
Feeling about Life in General SAMSHA/CMHS SMI survey
Housing Satisfaction SAMSHA/CMHS SMI survey
Overall Empowerment Aggregated from responses in
SAMSHA/CMHS SMI survey
Service Need Aggregated from responses in
SAMSHA/CMHS SMI survey
Program type SAMSHA/CMHS SMI survey
Health SAMSHA/CMHS SMI survey
Victimization SAMSHA/CMHS SMI survey
Gender SAMSHA/CMHS SMI survey
Race SAMSHA/CMHS SMI survey
Ethnicity SAMSHA/CMHS SMI survey
Concentration of subsidized Guhathakurta and Mushkatel 2000,
housing in tract 2002
Median household income of Census of Population and Housing
tract 2000
Drug Crimes in tract City of Phoenix Police Department
Percent single-family detached Census of Population and Housing
housing 2000
Median rent of tract as percent Calculated from Census of
of city's median rent Population and Housing 2000
Median household income in Calculated from Census of
tract as percent of city's Population and Housing 2000
median household income
2004 Housing Quality Calculated from City of Phoenix
Housing Quality Survey 2004
Table 3
Characteristics of 12 Census Tracts with the Highest
Concentrations of SMI Respondents
Census SMI Cum % Med age % collge med
tract resp * educ houshld
inc
1043 10 7.4 32.7 19.8 35165
1091 8 13.3 26.3 9.7 31055
1108 8 19.3 29.7 17.3 30947
1072.01 7 24.4 23.1 6.7 21068
1055 6 28.9 26 23.8 28682
1086.02 5 32.6 29.3 11.5 26476
1090 5 36.3 24.2 6.3 24394
1111 5 40.0 38 37.5 45893
1045.01 4 43.0 26.2 7.4 30938
1086.01 4 45.9 25.9 8.3 24412
1089.02 4 48.9 31.7 19.7 30841
1109 4 51.9 100 35.4 16.1 28438
Phx 135 30.8 22.7 41207
City
Census % White % unpd % rent %
tract singlefam
il det
1043 83.2 6.1 49.6 38.6
1091 49.1 8.9 31.9 75.9
1108 61.2 5.1 56.0 33.3
1072.01 60.7 19.8 81.9 6.0
1055 66.5 6.2 88.4 7.0
1086.02 57.2 6.5 74.2 27.2
1090 45.7 8.5 81.9 12.1
1111 87.9 4.6 34.4 67.0
1045.01 63.6 4.8 70.9 24.3
1086.01 47.9 9.5 78.1 16.7
1089.02 66.8 4.4 64.8 27.7
1109 69.9 7.5 67.9 31.6
Phx 71.1 3.7 39.3 57.6
City
* Data from baseline survey
Table 4
Marginal Percentages of the Outcome and Proximal
Variables at 12-Month Survey_
Value Marginal
Variables labels N Percentage
Feeling about life in general in bad 13 12.6%
three categories
mixed 44 42.7%
good 46 44.7%
Program Type SIL 15 14.6%
SH 88 85.4%
Tract concentrations of subsidized Low 27 26.2%
housing in groups
Somewhat low 20 19.4%
Somewhat high 25 24.3%
High 31 30.1%
recoded housing satisfaction into bad 13 12.6%
three variables
mixed 18 17.5%
good 72 69.9%
Overall psychological somewhat 5 4.9%
empowerment low
somewhat high 95 92.2%
high 3 2.9%
Health Excellent 32 31.1%
good 31 30.1%
poor 40 38.8%
Valid (for all above) 103 100.0%
Missing (for all above) 5
Housing Quality High 22 20.4%
Medium 26 24.1%
Low 26 24.1%
Missing 34 31.5%
Total 108
Table 5
Strength of Association Between "Feeling about Life in
General" and other Outcome Variables
Housing Overall
Satisfaction Empowerment
Control Variables None 0.195 * 0.567 **
Program type SIL 0.455 *
SH 0.166 0.701 **
Conc. Of subsidized Low 0.215 0.647
housing in tract Somewha 0.054 0.895
t low
Somewha 0.410 * --
t high
High 0.113 0.669 *
Median household Low 0.503 ** 0.185
income of tract Somewha 0.180 0.659 *
t Low
Somewha 0.212 0.944
t High
High -0.132 0.783
Health Excellent 0.188 0.446
fair 0.173 0.935
poor 0.238 (#) 0.490 (#)
Victimization Yes 0.168 0.754 *
No 0.236 0.348
Drug Crimes in low 0.237 0.273
tract Somewha 0.366 * 0.615 (#)
t Low
Somewha -0.407 * 0.559 (#)
t High
High 0.495 * 0.964
Housing Quality High 0.354 * 0.575
(2004 survey) Medium 0.138 0.804 *
Low 0.207 0.920
Service
Need
Control Variables -0.171 *
Program type -0.086
-0.225 **
Conc. Of subsidized -0.164
housing in tract -0.120
0.013
-0.333 *
Median household -0.076
income of tract -0.320 **
-0.109
-0.126 (#)
Health -0.166
-0.210
-0.051
Victimization -0.107
-0.256 (#)
Drug Crimes in -0.099
tract -0.038
-0.587 **
-0.093
Housing Quality -0.251 *
(2004 survey) -0.120
-0.099
Directional Somers'd: Dependent "Feeling about Life in General
"Significance levels indicated by: ** < .01; * <.05; (#) < .1
Table 6
Parameter Estimates from Ordinal Regression Analysis
Estimate Std. Wald
Error
Threshold Quality of life--bad -5.651 2.077 7.401
Quality of life--mixed -2.759 1.963 1.977
Median rent as % of .029 .015 3.469
city
% -.389 .120 10.450
unemployed in
tract
% single -.023 .009 6.208
family
detached
% African -.155 .060 6.596
American
Single -1.314 .468 7.866
independent
living
Supported Housing 0(a) -- --
Location Supported
housing -1.523 .470 10.522
concentration
--low
Supported
housing
concentration -.278 .543 .263
--somewhat
low
Supported
housing
concentration -.284 .481 .349
--somewhat
high
Supported
housing concentration 0(a) -- --
--high
Overall
empowerment--somewhat -4.807 1.323 13.205
low
Overall
empowerment--somewhat -1.105 1.046 1.115
high
Overall
empowerment 0(a) -- --
--high
Health--excellent 1.192 .395 9.111
Health--good .819 .381 4.618
Health--poor 0(a)
Housing
satisfaction-- -1.169 .495 5.563
bad
Housing
satisfaction-- -1.323 .439 9.080
mixed
Housing
satisfaction-- 0(a) -- --
good
Model fit Intercept Only 188.723
-2 log Final 118.511
likelihood Chi-Square 70.21**
Pseudo R Cox and Snell .494
squares Nagelkerke .574
McFadden .346
95% Confidence
Interval
df Sig. Lower Upper
Bound Bound
Threshold 1 .007 -9.722 -1.580
1 .160 -6.606 1.087
1 .063 -.002 .059
1 .001 -.625 -.153
1 .013 -.041 -.005
1 .010 -.273 -.037
1 .005 -2.232 -.396
0 -- -- --
Location
1 .001 -2.443 -.603
1 .608 -1.344 .787
1 .555 -1.226 .658
0 -- -- --
1 .000 -7.399 -2.214
1 .291 -3.156 .946
0 -- -- --
1 .003 .418 1.966
1 .032 .072 1.565
0
1 .018 -2.140 -.198
1 .003 -2.183 -.462
0 -- -- --
Model fit
-2 log
likelihood
Pseudo R
squares
Link function: Complementary Log-log. a This parameter is set to zero
because it is redundant. Significance < .01
