|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
Utilization of Community-Based Transitional Housing by Homeless Veteran Populations Diag nosed with a Mental Illness: The Association Between Predisposing, Enabling, and Need Factors with Program Outcomes by Roger Casey A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Community and Family Health College of Public Health University of South Florida Co-Major Professor: Martha Coulter, Dr. P.H. Co-Major Professor: Bruce Lubotsky Levin, Dr. P.H. Julie Baldwin, Ph.D. Roger Boothroyd, Ph.D. Robert Rosenheck, M.D. Date of Approval: October 29, 2007 Keywords: behavioral model, public health, program design, predictors, vulnerable populations Copyright 2007, Roger Casey
Dedicated to my wife and sons
ACKNOWLEDGMENTS I would like to thank my co -major professors and committee members for their support and assistance throughout this process. I am grateful to my professional colleagues and friends who have helped me and provided en couragement when neededbut most of all, I am thankful for the time and support that my wife has given me. Without her, I never would have made it.
i TABLE OF CONTENTS LIST OF TABLES iv LIST OF FIGURES vi ABSTRACT vii CHAPTER I: INTRODUCTION A ND STATEMENT OF THE PROBLEM 1 Overview 1 Need for the Study 2 Purpose of the Study 4 Research Objectives 5 Research Hypotheses 6 Participant-Level Hypotheses 6 Program-Level Hypotheses 6 Study Delimitations 7 Study Limitations 8 Definitions 10 Summary 12 CHAPTER II: REVIEW OF THE LITERATURE 13 Theoretical Framework 13 The Continuum of Care and Transitional Housing 13
ii The Behavioral Model and the Behavioral Model for Vulnerable Populations 17 Literature Review 24 The Social Epidemiology of Homelessness 24 Mental Illness Among the Homeless 26 Intervention Designs and Serv ice Provision for the Hom eless 28 Transitional Housing 30 Homelessness and Veterans 35 CHAPTER III: RESEARCH METHODS 39 Sample 39 Study Inclusion Criteria 40 Measures 40 Preliminary Data Analysis 42 Study Variables 45 Independent Variables 47 Dependent Variables 57 Statistical Analysis and Results 59 Level One Analysis 59 Level Two Analysis 72 Level Three Analysis 85 Logistic Regression Models 91 Generalized Estimating Equation (GEE) Models 96
iii Days in Residence and Outcomes 100 Discussion 101 Research Objectives 102 Implications for Public Mental Health 109 Protection of Hum an Subjects 111 REFERENCES 112 APPENDICES 121 Appendix A: NEPEC Intake Assessm ent (Form X) 122 Appendix B: NEPEC Discharge Report (Form D) 127 Appendix C: NEPEC Facility Survey Form 131 Appendix D: IRB Exemption Certification 138 ABOUT THE AUTHOR End Page
iv LIST OF TABLES Table 1 Percentage of GPD Program s Offering Vari ous Direct Services 44 Table 2 Variables of the Study 46 Table 3 Independent Variab les, Predisposing Factors 49 Table 4 Independent Variable s, Enabling Factors (Program Characte ristics) 54 Table 5 Independent Vari ables, Need Factors 56 Table 6 Outcomes Of Program Discharge 59 Table 7 Predisposing Variable Frequencies 61 Table 8 Enabling Variable Frequencies 66 Table 9 Need Variable Frequencies 68 Table 10 Outcome Variable Frequencies 70 Table 11 Study Hypotheses and B ehavioral Model Factors 74 Table 12 Dichotomous Variable Frequencies 82 Table 13 Program Completion Status (Outcome 1a ), Hypotheses One through Five 83 Table 14 Housing Status Upon Program Discharg e (Outcom e 2), Hypotheses One through Five 84 Table 15 Program Completion Status (Outcome 1a ), Hypotheses Six through Nine 90
v Table 16 Housed Upon Program Discharge (Outcome 2), Hypotheses Six through Nine 90 Table 17 Bivariate Significan ce with Outcom es 1a and 2 93 Table 18 Program Completion Status (Outcome 1a), Logistic Regression Model 94 Table 19 Housed Upon Program Discharge (Outcome 2), Logistic Regression Model 95 Table 20 Program Completion Status (Outcome 1a), Generalized Estim ating Equation Model 97 Table 21 Housed Upon Program Discharge (Outcome 2), Generalized Estim ating Equation Model 99
vi LIST OF FIGURES Figure 1. The Continuum of Care 16 Figure 2. Behavior al Model of Heal th Services Utilization 20 Figure 3. Enhanced Concept of the Behavioral Model 21 Figure 4. Behavioral Model Revision for Special Populations 23 Figure 5. Enabling Factors as Services 52
vii Utilization of Community-Based Transitional Housing by Homeless Veteran Populations Diagnosed With a Mental Illness: The Association Between Predisposing, Enabling, and Need Factors With Program Outcomes Roger Casey ABSTRACT Mental illness am ong homeless populations is a significant public health issue. Community-based programs that assist the hom eless are most often developed to meet local housing needs, not the needs of mental health populations. Transitional housing, a model frequently utilized to address homelessness in co mmunities, provides programbased housing with sup portive services. The purpose of this study was to examine the associations between participantand program-level factors on the utilizati on of community-based transitional housing by homeless veterans diagnosed with a mental illness. The study tested a revised framework of the behavioral model of utiliza tion for vulnerable populations theory. The sample was comprised of male home less veterans diagnosed with a mental illness who participated in community-based transitional housing programs in 2004 and 2005 ( n = 2,502). Data were collected on 288 programs throughout the United States, operated by local nonprofit or local government agencies and monitored by the U.S. Department of Veterans Affairs under th e Homeless Providers Grant and Per Diem Programs. Success was defined as either completion of a course of treatment as
viii determined by a masters prepared clinicia n, or if housing was obt ained upon discharge, as reported by the participant. Initial bivariate results indicated that both demographic and situational variables predicted success in transitional housing. However, upon further statistical analyses, limited predictors were revealed. Participants were more likely to be successful if they were white, reported combat experience, were interested in the program prior to admission, and were enrolled in cognitive behavioral models. Participants were more likely to be housed upon discharge if they were white, received some type of public support, were homeless less than 30 days befo re admission, and showed interest in the program at the time of the initial interview. Participants were less likely to be successful if they were diagnosed as schizophrenic. There was an indication that participants enrolled in programs designated as faith-based were less likely to be housed than those enrolled in secular programs. No statistically signifi cant associations were found between the level of services offered in the transiti onal housing programs with either successful completion or participants housing upon discharge.
1 CHAPTER I: INTRODUCTION A ND STATEMENT OF THE PROBLEM Overview The fact that so m any people in the Un ited States lack su itable housing reflects relatively recent political and socioeconomic ch anges. Individuals especially vulnerable to these changes, and thus to homelessness, are those diagnosed with a mental illness. Increases in health care costs, increases in poverty, with concurrent decreases in incomesupport programs, the need for increased job training, decreases in availability of lowcost housing units, deinstitutionalization of thos e diagnosed with mental illness, changes in vagrancy lawsall are trends associated with the rise of home lessness since the 1980s (Interagency Council on the Homeless Annual Re port, 1994). According to the National Alliance to End Homelessness, from 1970 through 1990, the number of low-income families increased by 40%, from 5.9 million to 8.5 million. At the same time, affordable housing available to these families declined by over 50%, leaving half of all low-income families without a permanent housing option (National Alliance fo r the Homeless Press Release, 2000). It is estimated that duri ng the 1970s and 1980s, the United States lost affordable housing, but the number of t hose needing low-cost housing increased, contributing to a gap of 5 million afford able housing units (Dolbeare, 1996). In addition to the evidence that may indi cate the problem of homelessness is a result of a societal change, homelessness can also be attributed to the social and
2 behavioral aspects of indi viduals in the population. Fo r instance, Plescia, Watts, Neibacher, and Strelnick (1997) identified a multitude of personal characteristics that increase an individuals risk for homelessnes s. Research conducte d through health-care outreach suggested that an individuals heal th, an underutilization of community services, and a lack informal support networks can increase the likelihood of homelessness. During the 1990s, new federal funding wa s available to create programs to provide services for homeless populations. The Stewart B. McKinney Homeless Assistance Act (1987) created a range of services including emergency shelter, transitional housing, job training, primary health care, and education. This law inspired other federal, state, and local funding initiati ves in an attempt to create programs to fill gaps in service needs through establishing local continuums of care utilizing transitional housing as a primary component. Despite this new funding and the creation of targeted services for the homeless, the population of individuals and families without permanent housing increased throughout the last three decades. Need for the Study As a social problem homelessness has enormous public health significance (Caton et al., 2005). Generall y, programs for those who ar e homeless are developed to address a local need, resulting in a program -product more likely based upon community service gaps rather than sound research. To address a communitys immediate needs, program design trends and federal funding sources have favored transitional housing models. Offering a safe place to stay for up to two years, community-based transitional
3 housing programs include various services to address the causes and effects of homelessness. Participants who utilize tr ansitional housing programs are engaged in a number of homeless-specific supportive se rvices and referred to other community agencies for specialized services such as health or mental health care. Community-based programs, although not necessarily by design, are seeing an increase in the number of individuals dia gnosed with mental illness (North, Eyrich, Pollio, & Spitznagel, 2004). Estimates of th e number of individuals diagnosed with a mental illness among the homeless population vary from 20% (Dickey, 2000) to 57% (Gelberg & Arangua, 2001). Some suggest that the prevalence of mental illness among the homeless population could be even as high as 80 to 95% (Martens, 2001). Although transitional housing has been the most widely offered service-provision model during the last 20 years, existing research indicating whether this model is the most beneficial for those homel ess individuals diagnosed with a mental illness is limited. In addition, it is difficult to determine from ex isting literature if ther e are any identifiable characteristics of this population that woul d suggest a greater likelihood of success in transitional housing type programs. A recent st udy demonstrated that interventions most likely to improve the life of a homeless pers on diagnosed with a mental illness are those found in programs that provide stable housing and basic services such as food and clothing. However, the same study demonstrated that those diagnosed with a mental illness are still compromised in terms of physical health, level of subsistence needs met, victimization, and subjective quality of lif e (Sullivan, Burnam, Koegel, & Hollenberg, 2000).
4 Currently, there is a trend away from assi sting homeless individuals in transitional housing models. This trend is based on re search that suggests specific groups of homeless populations can benefit from pl acement directly into permanent housing, avoiding the transitional hous ing step (Tsemberis, Gulcur, & Nakae, 2004). This housing first approach was conceptualized and developed primarily for those diagnosed with a mental illness; it offe rs participants direct housing pl acement, forgoing any type of transitional program. Research is limited regarding the types of homeless individuals diagnosed with a mental illness that may benefit most or may be more likely to have positive outcomes from community-based transiti onal housing. In addition, as hi ghlighted in later sections of this research, a limited number of published studies have explored the program services offered in community-based tran sitional housing that lead to successful outcomes. Community providers would bene fit from research that explores the utilization of transitional housing by those populations diagnose d with a mental illness. Purpose of the Study The purpose of this research was to exam in e the associations be tween participantand program-level factors on the utilizati on of community-based transitional housing by homeless veterans diagnosed with a mental illness. The study tested a revised framework of the behavioral model of utilization for vulnerable populations theory (Andersen, 1995; Andersen & Aday, 1978; Gelberg, Andersen, & Leake, 2000). In addition, it explored
5 the assumptions behind utilizing the tran sitional housing design for addressing homelessness. Research Objectives The research objectives were to Provide a descriptive analysis of participant chara cteristics in a national sample subset of homeless populations diagnosed with a mental illness utilizing community-based transitional housing programs; Examine and assess the intensity and types of services of communitybased transitional housing utilized by a national sample subset of the homeless population diagnosed with a mental illness; Examine the associations of program participant characteristics and mental health diagnosis with community-bas ed transitional housing outcomes; Examine the interaction of program-lev el services on the associations of program participant characteristics a nd mental health diagnosis with community-based transitional housing outcomes; and Develop and offer recommendations as to what types of community-based transitional housing programs may be best suited to meet the needs of homeless populations diagnosed with a mental illness.
6 Research Hypotheses Participant-Level Hypotheses 1. There is no significant association be tween participant d emographics and successful completion of community-based transitional housing. 2. There is a negative association between pa rticipant severity of homelessness and successful completion of community-based transitional housing. 3. There is a positive association between pa rticipant expressed interest in program utilization and successful completion of community-based transitional housing. 4. There is a positive association between pa rticipant perceived mental illness and successful completion of community-based transitional housing. 5. There is no significant associ ation between participant mental health diagnosis and successful completion of community-based transitional housing. Program-Level Hypotheses 6. There is a p ositive association betw een program certification status and participants successful completion of community-based transitional housing. 7. There is no significant association be tween the type of treatment-model philosophy of a community-based transiti onal housing program and participants successful completion. 8. There is no significant association between the religious basis of a communitybased transitional housing program and participants successful completion.
7 9. There is a positive association between th e level of (homeless-specific) program services and participants successful co mpletion of community-based transitional housing. Study Delimitations The sam ple for this study included indi viduals who were determined to be veterans by the U.S. Department of Veterans Affairs (VA) and only those veterans, identified as homeless, who sought residential services or were assessed as needing services through outreach efforts conducted by VA staff. The Literature Review and Methods secti ons explored several of the differences between the homeless-veteran and ho meless-nonveteran populations. Although differences exist in race, age, educati on level, marital stat us, employment, and vulnerability risk for homelessness, gene ralizing to the nonveteran homeless population is not unreasonable. The mental health diagnoses of the st udy subjects have been determined by masters-prepared clinicians following esta blished protocols administered by the U.S. Department of Veterans Affairs. The data set for the study included info rmation on approximately 300 transitional housing programs that represented geogra phically diverse locations throughout the United States. Programs were located in most states and the District of Columbia, in settings both urban and rural. Programs reviewed under this study represent various
8 types of transitional housing models: from low-demand, long-term housing to highdemand programs structured with limited lengths of stay. The effects of program services on particip ant outcomes included consideration of only those services offered through commun ity-based programs that received grants funded by the U.S. Department of Veterans Affa irs (VA). By law, recipients of grants are required to comply with federal regulati ons that, in part, guide the provision of services. Many similar services are offered in community-based programs not funded under these grants. However, if a type of service offered in this program had a positive effect on participan t outcomes, it would not be a reasonable assumption that the service would have the same effect in another co mmunity-based program unless the service was compared regarding type, intensity, and duration. Data on program services were collected and prepared by Veterans Affairs staff that regularly site-visit the facility under protocols administered by VA. Results of this study could have a signif icant impact on federal policy regarding transitional-housing treatment models for home less individuals diagnosed with a mental illness. Study Limitations This study used existing adm inistrative data A preliminary analysis of the data was conducted to review the feasibility of util izing this information and to determine if it was a reasonable data set for testing a particular theoretical framework and for
9 conducting the study. However, this existing data can be subject to limitations of the collection instruments an d the interviewers, as discussed in the Measures section, as well as to other limitations of the recording and co mpiling of information prior to the research. The sample for this study was made up exclusively of males who are military veterans. Although the study discusses the generalizability of this research to other VA samples, the specificity of this sa mple remains a limitation of the study. The data used for this study were collected on subjects who were contacted by VA staff conducting outreach or on those who accessed VA staff either in the community or at a VA facility. Those subjects w ho accessed VA staff were most likely seeking assistance of some type, and as such, may have been more inclined to participate in the services provided. The sample was limited to those in tr ansitional housing programs who were diagnosed with a mental illness. Participants who had co-occurring substance-abuse disorders were excluded from the sample in an effort to narrow the focus of the research to outcomes from transitional housing programs for those with mental illness. Because substance-abuse disorder frequencies in th is population are estimated at 60% to 80%, providing a more comprehensive approach to reviewing transitional housing outcomes would require further research. The services offered at each transitio nal housing program were summarized by the individual program. It was assumed that if a participant was in a particular program, the participant received that programs serv ices. Additionally, the study neither tracked
10 nor recorded services that the program par ticipant may have received in the community or at another facility. Participant outcomes were recorded immediately upon discharge from the transitional housing program. Follow-up data on participants that would capture continued success in the comm unity were not available. Definitions The following term inology was used thr oughout the study. A number of these definitions can be found in the rules and re gulations that implement Public Law 102-590 (38 CFR 61.0). Community-based: located in the community, in n ear proximity to locations the participants of the progr am frequent or where they are likely to be. Community-based also implies that the program is supported by and is coordinated with other organizations with similar missions and participants. GPD-funded program: a community-based transiti onal housing program funded under the VAs Homeless Providers Grant and Per Diem Program. GPD participant: a person who receives se rvices provided at sites funded with assistance under VAs Homeless Providers Grant and Per Diem Program. Homeless or homeless individual (From the U.S. Department of Housing and Urban Development definition as set forth in McKinney Act Legislation, 1987): an individual who lacks a fixed, re gular, and adequate nighttime residence and has a primary nighttime residence that is  a
11 supervised publicly or privat ely operated shelter designed to provide temporary living accommodations;  an institution that provides a temporary residence for persons intended to be institutionalized; or  a public or private place not designed for, or ordinarily used as, a regul ar sleeping accommodation for human beings. Note: the term homeless or homeless individual does not include any individual imprisoned or otherwise detained pursuant to an Act of the Congress or a state law. An individual on probation, parole, or under electronic custody is not considered imprisoned or otherwise detained. Nonprofit organization: an organization recognized by th e U.S. Internal Revenue Service as a 501(c)3 or 501(c)19. Mental illness: illness of the mind as determined and diagnosed by a masters-prepared clinician using standardized di agnostic procedures set forth by clinical practice, not to include substance abuse. Participant or subject: an individual who receives services provided at the sites described or in programs refere nced in this study. Supportive housing: noninstitutional housing, scattered through the community, with a limited number of particip ants, in conjunction with supportive services. Supportive services: services that address the causes and effects of homelessness, with a goal of moving participants to independent living in the community.
12 Transitional housing: housing in a collective or semi -institutional setting with supportive services, intended to facilitate the move ment of homeless in dividuals and their dependents to permanent housing within 24 months. Veteran: a person who served in the active militar y, naval, or air service, and who was discharged or released under c onditions other than dishonorable. Summary Assisting hom eless individuals requires s ound research regarding the provision of housing for various populations. A study that can suggest which types of transitional housing services are likely to be most beneficial for those who are homeless and diagnosed with a mental illness will have potential significance for public health policy. This study offered both an assessment of the types of homeless serv ices available in a national sample and an examination of the pr ogram-level services that may increase the effectiveness of the transitional housing model. This study aims to assist providers who struggle day to day with help ing the homeless, researchers who study this complex and persistent social phenomenon, and decision makers, e.g., offi cials and state and federal legislators, who influence the allocation and utilization of limited public health resources, especially as those resources relate to the homeless dia gnosed with a mental illness.
13 CHAPTER II: REVIEW OF THE LITERATURE This chapter provides both a discussion of the theoretical basis of this study and a review of the relevant resear ch literature on hom eless popula tions and programs to assist the homeless. The Theoretical Framework sect ion of this chapter includes a discussion on the transitional housing model as an inte gral component to th e continuum of care promoted by the federal government thr ough housing policies of the 1980s and 1990s. Following this discussion, the theoretical f oundations of the transitional housing model design are contrasted with recent program re search. Finally, the behavioral model and the revised behavioral model for vulnerable populations are discussed and proposed as the theoretical framework for this study. Theoretical Framework The Continuum of Care and Transitional Housing This study focuses on the utilization of transitional housing by vulnerable populations. Community-based, transitional-type housing addresses the need for, or lack of, low-cost housing: displaced individuals are provided temporary housing through transitional programs while they are able to make a living. As vulnerable populations drop out of low-cost housing, community-bas ed transitional housing-type programs fill the gaps. The transitional housing programs assist the participants in addressing the
14 causes and effects of homelessness (Barrow & Rita, 1998). Upon completion of these programs, it is assumed the vulnerable (relative to this study, veterans diagnosed with a mental illness) will be better system-fit indi viduals, that is, ready to secure housing, presumably with enhanced social networks, increased entitlement revenues, job skills, and improved mental and physical health. The term transitional housing is used to describe a wide variety of setting types designed for those populations who have an unassured residence. There is no national listing of transitional housing programs; ma ny exist through informal arrangements with nonprofit organizations or self-help groups. Transitional housing programs for those participants focusing on sobriety maintenan ce remain intentionally anonymous. During the past 10 years, the U.S. Departments of Health and Human Serv ices (HHS), Housing and Urban Development (HUD), Justice (DOJ), and Veterans Affairs (VA) have all offered funding to construct and/or op erate community-based programs that are transitional in nature. The Urban Institute (1999) estimates that there are approximately 4,400 transitional housing program s in operation nationally. A systems approach to addressing the homeless problem has been evident in national policy. In 1993, President Clinton issued Executive Order 12848 to provide for streamlining and strengthening U.S. efforts to break the cycle of homelessness. Under this Executive Order, federal agencies through the U.S. Interagency Council on Homelessness, were charged with developi ng a coordinated fede ral plan with the necessary administrative and legislative in itiatives to address the nations homeless problem. This plan, detailed in the document entitled Priority: Home! The Federal Plan
15 to Break the Cycle of Homelessness (Interagency Council on the Homeless, 1994), established federal policy that encouraged specific program design models to address homelessness. This federal plan reformulated the wa y communities could request federal funding under the McKinney Act. Instead of community agencies making applications directly to federal funding sources, local groups were required to initiate and formulate planning councils to submit coordinated and colla borative applications establishing a comprehensive continuum of care thus encouraging a systemic approach to solving local homelessness. Although many communities had previously attempted to develop coordinated homeless assistance, the federal plan es tablished national policy by prioritizing funding based upon a communitys description of and commitment to a local continuum of care for the homeless. For ex ample, to be competitive in seeking HUD funding under the McKinney Act 1994 Homeless Super Notice of Fund Availability (NOFA) process, a community must have shown evidence of local planning to develop a comprehensive continuum of care. Other fe deral funding agencies (HHS, DOJ, VA), in subsequent funding announcements, also required continuums of care. A core component of the continuum of care is transitional housing. The diagram below (Figure 1) was included in the 1994 federal plan:
16 Figure 1. The Continuum of Care. From Priority: Home! The Federal Plan to Break the Cycle of Homelessness According to the federal plan, implementation of the continuum of care would help enhance a localized systemic effort and move existing homeless assistance programs with diverse rules and requirements toward a si ngle coordinated approach to dealing with homelessness (Interagency C ouncil on the Homeless, 1994, p. 74). Evident in this plan, and thus in the c ontinuum, is the assigned importance of the transitional housing component. Transitional housing was, and continues to be, viewed as a way to assist individua ls in addressing the causes and effects of homelessness, enabling individuals to become better system -fit. According to the Department of Housing and Urban Development (HUD), transitional housing should provide temporary residence with supportive services to he lp people develop the skills necessary for permanent housing. The theory of transitional housing has been embraced and supported by federal policy, and adopted by localities. However, a review of the literatu re, presented in the Permanent Housing Transitional Housing Supportive Housing Emergency Shelter Outreach and Intake The Continuum of Care
17 second section of this chapter, does not provi de conclusive evidence of the effectiveness of this model, especially for those homeless individuals diagnosed with a mental illness. The Behavioral Model and the Behavioral Model fo r Vulnerable Populations Researchers have utilized various theoretical fram eworks to examine and study homeless populations. Based upon intervie ws with mothers living in temporary emergency shelters, Banyard (1995) suggests that coping theory is useful. Studying a group of homeless individuals diagnosed with a mental illness, Benda (2004) chose predictors of readmission to psychiatric care based upon life-course th eory. Berne, Dato, Mason, and Rafferty (1990) utiliz ed a poverty model to study conditions that contribute to significant physical and mental health proble ms of families. Other theories or models have been utilized to study homeless indivi duals or like groups of homeless populations, for example, systems integration modeli ng (Dennis, Steadman, & Cocozza, 2000), learned helplessness theory (Flynn, 1997), lear ned helplessness with social disaffiliation theory (Goodman, Saxe, & Harvey, 1991), a nd attachment theory (Gwadz, Clatts, Leonard, & Goldsamt, 2004). A number of theoretical models have b een used to study the homeless population as well as the societal costs of homelessne ss and utilization of services by homeless populations. Through a discussion of syst ems theory, Caplan & Caplan (2000) demonstrated how the public health approach to homelessness is not based upon primary, secondary, or tertiary prevention but on crisis theory. As a result, many systems of care may actually harm the people seeking assistance. The resources and adaptive
18 characteristics of the homeless population are discussed as solutions to the problem by Haber & Toro (2004) through a social orga nization model in a broad ecological development perspective. Kreider & Nic holson (1997) discuss the impediments that homeless populations face in accessing health care, based upon other, nonfinancial barriers. Perhaps one of the more provocativ e studies is Lyon-Callos (2000) research demonstrating that the medicalizing of homelessness may reinforce service organizations blame of the homeless populati on for their situation. Such medicalizing avoids the larger political economic processe s that are the root cau ses of homelessness. This study focused on the uti lization of a particular se rvice, transitional housing, by a homeless population. The individual caus es of, or the systemic reasons for, homelessness were avoided as specific topics of this research. Therefore, a revised version of Andersens behavioral model of he alth services theory served as this studys analytic framework and basis for data analysis. The behavioral model (Andersen, 1995; Andersen & Aday, 1978) is a theory frequently used to study patterns of health-car e service utilization. The initial framework developed by Andersen suggest s that a populations use of he alth care services is a function of the populations predisposition to use it, of the factors that enable or impede use, and of an individuals need for care (Pruchno & McMullen, 2004). The theory and its contributing models assist in defining a sequence of conditions that may be factors in whether or not populations use services a nd the volume of services they consume (Andersen & Aday).
19 In the behavioral model, variable s are organized under three domains ( predisposing factors, enabling factors, or need factors), each characterized as possible determinants of utilization (see Figure 2). Predisposing factors, such as demographic characteristics (age, sex, marital status, ethnicity, and education), exist before the illness. Enabling factors include those environmental or indivi dual characteristic s that increase the likelihood of service utilization, such as fi nancial resources, ability to locate services, or health care insurance. Need factors are considered to be the number of current or past health conditions; need factors may also incl ude subjects perceptions of their health conditions or professional ev aluations of their health. The behavioral model has been widely used as a framework for research exploring access to and utiliz ation of health care serv ices. Studies based upon the behavioral model have expl ored use of services by popul ations with developmental disabilities (Pruchno & McMullen, 2004), ut ilization of support groups by family caregivers of adults w ith mental illne ss (Biegel, Shafran, & Johnsen, 2004), use of health services among the elderly (Saag et al., 1998), ethnic differences in the utilization of inpatient mental health service (Padgett, Pa trick, Burns, & Schlesinger, 1994), and the influence of health beliefs of elderly a dults on access to and utilization of care (Evashwick, Rowe, Diehr, & Branch, 1991).
20 Figure 2. Behavioral Model of Health Se rvices Utilization (Andersen, 1995) Initially, the primary focus of the Andersen model was to explain access and service availability. More recently, however, research has expanded the model, and the conceptualization of service access has include d not only access, but also utilization and the receipt of services (Phillips, Mo rrison, Andersen, & Aday, 1998; Pruchno & McMullen, 2004). Although not fr equently cited in earlier studies, Aday and Andersen (1995) originally highlighted the importance of the utilization of services and the completion of the course of treatment when considering access to care. Implicit in the characterizations of access as properties of the individual or the system, then, is the assumption that the quantity and quality of an individuals passage through the medical care system are affected by th ese factors (Aday, 1974, p. 210). Enhancing the receipt of treatment as an important component of service utilization, Phillips et al. (1998) discuss the importance of pr ovider-related variables in Predisposing: Factors that represent an individuals predisposition to utilization (age, gender, marital status). Enabling : Factors that encourage or impede utilization (environment, social Health Care supports, finances). Utilization Need : Factors that represent an individuals need for care (current health or professional evaluation).
21 the behavioral model of utilization. Provider -related variables include those factors that may be influenced by providers, as well as pr ovider characteristics that interact with patient characteristics to influence utilization (see Figure 3). For example, factors that may be influenced by providers include the me thod of service delivery or the types of services (within the existing service model). Provider characteristics that interact with recipients of service could be the gender of the healthcare provider or the context where care occurs. These variables measure the context where the utilization occurs. According to Phillips, the infl uence of these variables has be en relatively unexplored. Of the research reviewed that cited the behavioral m odel between 1975 through 1995, only 51% included provider-related variables Figure 3. Enhanced Concept of the Behavior al Model (Phillips et al., 1998)
22 Further revisions of the behavioral mode l were suggested by Gelberg, Andersen, and Leake (2000). Considering special needs of various subject groups, Gelberg et al. (2000) presented the behavioral model for vulnerable populations enhancing the initial model to include domains especially relevant to understanding th e health and healthseeking behavior of vulnerable populations. According to the authors, this adaptation includes factors to consider wh en studying the use of health services and health outcomes of vulnerable populations with special need s (see Figure 4). The categories can be tailored to the types of speci fic populations when the model is applied to those groups. In a study of a homeless population diagnosed with a mental illness, Desai, Rosenheck, and Kasprow (2003) found that vulnerable domain factors were important supplements in assessing determinants of receipt of medical care. In this revised model, Gelberg et al. ( 2000) also emphasized Andersens original premise that course of treatment is a nece ssary consideration when studying access. The authors state that while most models of health service utiliz ation stop at utilization, with this study we were able to examine the effects of realized access (i.e., utilization) on health outcomes. Health status is both an outcome as well as a determinant of use (p. 27).
23 Figure 4. Behavioral Model Revision for Sp ecial Populations (Gelberg et al., 1997) Although the behavioral model and its revi sions have been a framework widely used to study health care util ization, there has been limited use of this model in research with homeless populations. Studies have included research on competing priorities including barriers to medical care among homeless adults (Gel berg, Gallagher, Andersen, et al., 1997), medical service use by sheltered homeless (Wei nreb, Goldberg, & Perloff, 1998), predictors of the course of health serv ices utilization of homeless people (Gelberg et al., 2000), the accessibly of medical care for homeless wo men (Lim, Andersen, Leake, Cunningham, & Gelberg, 2002), and determinants of medical care (Des ai et al., 2003). This study tested the behavioral mode l for vulnerable populations for those homeless individuals diagnosed with a mental illness. Variables we re identified within the context of the Andersen behavior model as predisposing enabling or need factors The study design considered outcome measures as an important factor in evaluating
24 utilization of care. The design also incorpor ated provider-related va riables to determine the context where utilization occurs. R ecognizing the major revision to the model proposed by Gelberg, Gallagher, Andersen, and Koegel (1997), additional domains tailored for the subject population were added. Literature Review The following section reviews the relevant literature pertaining to hom elessness and those in the homeless population diagnosed with a mental illness. An overview of the social epidemiology of homelessness is followed by a review of the literature on homelessness and mental illness. Next, inte rvention methodologies are discussed through a review, and this discussion le ads to an examination of the literature relevant to the specific program designs of transitional housi ng. Finally, a review is provided on the programs and research pertaining to homeless veterans, the s ubject group of this study. The Social Epidemiology of Homelessness A num ber of studies have shown psychos ocial attributes that correlate with homelessness. Increased risk for homelessness has been associated with mental illness (Breakey et al., 1989; Isaac, 1990; Koegel, Burnam, & Farr, 1988; Phelan & Link, 1999; Pollio, North, Thompson, Paquin, & Spitznagel, 1997; Sullivan et al., 2000). Increased risk for homelessness has also been associat ed with substance abuse (Calsyn & Morse, 1991; Johnson, Freels, Parsons, & Vangeest, 1997; Vangeest & Johnson, 2002; Wenzel, 1993;) and with individuals who have co-occu rring disorders (Bla nkertz, Cnaan, White,
25 Fox, & Messinger, 1990; Fischer & Breakey, 1991; Wenzel, Ebener, Koegel, & Gelberg, 1996). In addition, an increased risk for homelessness has been associated with other, environmental characteristics rather than indi vidual traits. The limited availability of low-cost housing, increased poverty, and incr easingly weaker social ties evident in contemporary family units have all been lin ked to the rise in the homeless population (Dolbeare, 1996; Schutt & Gerret, 1992). Additionally, environmentally exclusive determinants have been associated with homelessness, for instance, increased community violence, enactment of rigid local vagran cy laws, and reduction of community-based social services (Foscarinis, 1991; Haugland, Siegel, Hopper, & Alexander, 1997). Highlighting the multiplicity of possible causes of homelessness, yet other research has identified determinants exclusive of those mentioned above. The causes for homelessness have also been linked to poor health (Rosenheck, Gallup, & Frisman, 1993) and childhood abuse (Koegel, Melamid, & Burnam, 1995; Susser, Struening, & Conover, 1987). The most recent demographic informati on available on the homeless population can be found in a report published by the Urban Institute (1999), entitled Homelessness: Programs and the People They Serve: National Survey of Homeless Assistance Providers and Clients. Of those homeless individuals s eeking services nationally, 80% were between 25 and 54 years old. Sixty-eight pe rcent were male, and 32% were female. Forty-one percent were white, non-Hispanic ; 40% were black, nonHispanic; 11% were Hispanic; and 8% were Native American. Forty-eight percent of the homeless individuals were never married; of the 52% that were married at one time, 24% were
26 divorced, and 15% were separa ted. The proportion of those who graduated from high school was 34%; 28% reported high er educational attainment. Of the homeless population, approximately 60% stayed in shelters; however, 20% lived literally on the streets. Slightly over half did not have any paid employment within 30 days prior to being interv iewed. Only 40% received any type of government benefits; however, that figure was higher for those reporting as part of a family system (52%). Mental Illness Among the Homeless The prevalence of m ental i llness among the homeless is not static. A study that examined data collected in 1980, 1990, and 2000, revealed a dramatic increase in mood and substance-use disorders among the homel ess population (North et al., 2004). The authors stated that service systems need to be aware of potential prevalence changes and the impact of these changes on service needs. The most prominent mental disorders among the homeless were found to be depression affective disorders, substance abuse, psychotic disorders, schizophrenia, and personality disorders (Martens, 2001). Additionally, Martens reported that the prevalence of mental disorders among the homeless may be as high as 80 to 95%. Acco rding to this research, in the United States homelessness is a major, complex, public health problem. Most of the studies explor ing the relationship between homelessness and mental illness began appearing in the literature in th e 1980s. Early studies identified a homeless sample as unaffiliated persons living in extr eme poverty with high levels of physical and mental disability (Rossi, Wright, Fisher, & Willis, 1987). Early studies also found that rates of schizophrenia were elevated am ong individuals who had been homeless many
27 times or for long periods of time (Koegel, Burnam, & Farr, 1988). Using standardized diagnostic criteria, Fischer, Shapiro, Br eakey, Anthony, and Kramer (1986) found that about one-third of homeless individuals had a current psychiatric disorder. In the same sample, homeless individuals exhibited higher prevalence rates in every diagnostic category, and homeless men were found to have higher rates of hospitalization for mental disorders. During the late 1980s, other studies demons trated the high rates of co-morbidity, that is, mental illness with substance abuse, among the homeless population. These studies also identified the need for mental health and substance abuse services (Breakey et al., 1989). Koegel & Burnam (1988) found that the homeless were characterized by a substantially higher prev alence of other mental disorders in addition to substance abuse disorders, particularly the major mental illnesses. In the 1990s and into 2000, research continued to show the high prevalence rates of mental illness among homeless individuals (Caton, Shrout, Eagle, & Felix, 1994; Fische r & Breakey, 1991; Johnson & Barrett, 1995; North, Thompson, Pollio, Ricci, & Smith, 1997). As studies continued to document the pr oblem of mental illness among homeless individuals, other research ers were exploring why thes e rates were unusually high. Koegel demonstrated that childhood experiences increase adults vulnerability to homelessness and that adults vulnerability to homelessness could be affected by factors that include age, gender, and r ace/ethnicity (Koegel et al., 199 5). Variables that were the strongest predictors included the number of stressful even ts before becoming homeless, age, current life satisfaction, psychopathology, and prior mental hos pitalization (Calsyn
28 & Morse, 1991). Variables, including the av ailability of social and economic resources, were also associated with ho melessness (Johnson et al., 1997). Other studies suggested that resource problems may determine homelessness. Individuals that were homeless and diagnosed with a mental illness had reduced protection afforded by social networks and increased impact of disaffiliation (Sosin & Bruni, 1997). Sullivan et al. (2000) found that me ntal illness may play a role in initiating homelessness for some but that it is unlikely in and of itself to be a sufficient risk factor for homelessness. Intervention Designs and Servic e Pr ovision for the Homeless As homeless populations were being studied, so were methods to address their needs. Researchers discovered that conve ntional methods of treatment for those homeless persons diagnosed with a mental illness were not always effective. Indeed, effective approaches to address the needs of this pop ulation were thought to require significant modifications of tr aditional techniques and changes in the implementation of specific interventions (Cal syn & Morse, 1991). However, according to Wenzel et al. (1996), homeless persons appear to have no less commitment to achieving treatment goals than their nonhomeless counterparts. Furthermore, the life-inhomelessness cycle might actually inhibit the success of tr aditional treatment methods; for some homeless persons with a mental illness, the homeless shelters, programs, jails, and prisons were found to function as a ma keshift alternative to inpatient care or supportive housing and thus possibly to rein force the marginalization of the population
29 (Haugland et al., 1997). In a study that de termined access to treatment for the homeless adult population, Koegel, Sullivan, Burnam, Mo rton, and Wenzel (1999) found that only one-fifth of those who had either a chronic substance abuse disorder or chronic mental illness received treatment for those disorders within a 60-day period. In the Center for Mental Health Se rvices Access to Community Care and Effective Services and Supports (ACCESS) study, baseline and follow-up data on 1,828 homeless individuals were coll ected to evaluate the relatio nship between individuals socio-demographic and clinical characteristic s, social support, a nd levels of formal service use. Social support was determined to be positively related to acquiring or accessing services (Lam & Rosenheck, 1999). A ccording to the authors, social support was most strongly associated with improved ac cess to an array of di fferent services, and this improved access was determined to be an important need among the homeless population. Other studies focused on prevention effort s. Olfson, Mechanic, Hansell, Boyer, and Walkup (1999), indicated that at the time of hospital discharge, psychiatric symptoms and impaired functioning posed a risk of homelessness among patients with schizophrenia. Researchers found that an enhanced community-based mental health system was not sufficient to prevent homelessness among high-risk persons with a serious mental illness, and 11% of their study sample experienced homelessness after referral to an extended acute care facility (Kuno, Rothbard, Averyt, & Culhane, 2000). The authors suggested that strategies to prevent homelessness should be considered,
30 perhaps at the time of discharge from the re ferring community hospital or extended acute care facility. Considered a significant intervention methodology and utilized by communitybased organizations for the provision of homeless services, transitional housing is meant to offer a temporary residence while program participants can work toward residential stability. Distinct from emergency shelters and permanent housing, transitional housing is viewed as an integral component of a communitys continuum of care for the homeless population. Transitional Housing W ide diversity exists among transitional housing program models. Transitional housing models are based upon differing philoso phical and disciplinary traditions; they also target different subgroups, vary in physical structures an d intensities of services, and place varying degrees of demands upon reside nts (Barrow & Rita, 1998). Research on the effectiveness of transitional housing is limited, in part, because of the various definitions of transitional housing. In addition, the effectiveness of transitional housing can be measured through a number of diffe rent domains: housing, employment, or service outcomes; provision or linkage to serv ices; clinical status; or assessment of the immediate or long-term benefits of pr ogram participation (Barrow & Rita). Early U.S. Government Accounting Offi ce (GAO) reviews indicated that HUDs transitional housing programs may be succe ssful in reaching the intended target population, discharging residents to comm unity independent housing, and increasing
31 participants income upon discharge ( Homelessness: Transitional Housing Shows Initial Success but Long-Term Effects Unknown 1991). However, it was noted in the GAO report that these results were from data collected from participants immediately after discharge from transitional housing. As the title of the report i ndicates, the long-term benefits were not studied. Several earlier studies have shown transitional housing programs to be effective in linking homeless populations to indepe ndent living. Murray and Baier (1995) evaluated a transitional housi ng program for the homeless diagnosed with a mental illness and found that of the 228 part icipants in the sample, upon discharge 48% obtained and maintained permanent housing and secured in come supports either through entitlements or employment. No association was found between psychiatric diagnosis and individual goal attainment (Murray & Baier). In th e same sample, over 78% maintained housing one year after discharge; this group was more likely to have utilized psychiatric day programs while in transitional housing reside nce (Murray, Baier, No rth, Lato, & Eskew, 1997). In general, those most likely to comp lete the transitional housing program were more significantly involved in activities of the program (Murray, Baier, Lato, & Eskew, 1995). Although these studies may suggest that residents of transiti onal housing are more successful if the program is structured (with more activities and options for treatment), Carr, Murray, Harrington, and Oge (1998) f ound that satisfaction with a transitional housing program was inversely re lated to program structure. Regardless, there seems to be a relationship between su ccess in transitional housing and the number or intensity of the services available in the program.
32 Vocational and housing assistance (Grella, 1993), a therapeutic milieu (Murray & Baier, 1993), comprehensive rehabilitative treatment (Prabucki, Wootton, McCormick, & Washam, 1995), and case management (Conrad et al., 1998) were all found to enhance transitional housing and increase particip ants likelihood of successful outcomes. Through observational studies, othe r researchers found that basi c interventions (similar to transitional housing, that is, stressing stable housing, incl uding provision of food and clothing, addressing physical heal th problems, and training individuals to minimize their risk of victimization) would most likely improve the quality of life of homeless persons with a mental illness (Sullivan, et al., 2000). Several studies demonstrated the positive association between in tegrated services and residential care. Bebout, Drake, Xie, McHugo, and Harris (1997) studied residential outcomes of homeless adults with severe mental illness and found that if formerly homeless persons are provided in tegrated dual diagnosis trea tment, they can gradually achieve stable housing. Drake, Yovetich, Bebout, Harris, and McHugo, (1997) examined the effects of integrating mental health in terventions for homeless persons with severe mental illness and found positive quality of li fe outcomes. Integrating psychiatric treatment (Kasprow, Rosenheck, Fr isman, & DeLella, 1999) and offering multidimensional treatment (Leda & Rosenhec k, 1992) also increased the likelihood of positive outcomes from temporary housing pr ograms. Rosenheck (2000) found that innovative programs for the homeless with me ntal illness are more effective than standard care.
33 Although most of the studies cited above ar e limited in scope a nd do not utilize an experimental design, they do provide some evidence of the e ffectiveness of the transitional housing model or at least of the model type us ed in the particular research. However, these studies may also demonstrate the usefulness of the services associated with this model, in addition to, or rather th an, demonstrating evidence of the effectiveness of transitional housing itself. Many homeless advocates disagree with the concept of tr ansitional housing, arguing that it stigmatizes populations utiliz ing the programs while institutionalizing a problem that can be solved by increasing th e availability of affordable housing. More recently, studies have discussed the disadvantages of transitional housing and the advantages of housing-first models. In the late 1990s, several studies reported on the weaknesses of transitional housing programs. Hopper, Jost, Hay, Welber, and Haugland (1997) reported that institutional settings coupl ed with shelters provi de a circuit for the mentally ill homeless that may prevent or s ubstitute for more stable and appropriate housing. Tsemberis, Moran, Shinn, Asmussen, and Shern (2003) showed that when homeless individuals diagnosed with a mental illness were placed directly into a housingfirst model, that is, into permanent hous ing with supports, housing retention was remarkably high after six months. Although utilizing a small sample ( n = 225), the authors found that 79% of t hose placed retained housing. A follow-up study (Tsemberis, Gulcur, & Nakae, 2004) on the same sample and additional research conducted in a similar program design model (Tsemberis & Eisenberg, 2000) yielded equally positive
34 housing retention results for those who were homeless and diagnosed with a mental illness. Similar to the housing first model is supportive housing. Designed to be permanent independent housing, supportive housing programs utilize single occupancy apartments integrated within the community. Participants are pr ovided with significant ongoing case management services to address a variety of needs. They can enter supportive housing from other pr ograms (a distinction from the housing-first model), but when they are admitted directly to supportiv e housing, the program model takes the shape of the housing-first design. Mares, Kaspro w, and Rosenheck (2004) found that there were no significant differences in outcom es between those who had received prior residential treatment and those placed dire ctly in supportive housing. Supportive housing has been found to produce better outcomes than case management alone (Rosenheck, Kasprow, Frisman, & Liu-Mares, 2003), and, when associated with case management, the effectiveness of the model can be de monstrated (Kasprow, Rosenheck, Frisman, & DiLella, 2000). However, in the Mares et al. (2004) study, there did not seem to be an association between the part icipants preference for housing and housing outcome. The discussion above provides evidence linki ng mental illness with homelessness. In addition, housing with treatment seems to o ffer this population an alternative to living on the streets and, in many cases, leads to mo re stable conditions with improved mental health. The review of the literature concerning effectiv eness of transitional housing seems to reveal that assessment of the pr ogram model is limited. Past studies were moderate in scope, sample sizes were small, and perhaps most important, there was no
35 standardized comparison considering the dive rsity of program models and the various definitions of success. In addition, most st udies did not discriminate to the extent necessary to demonstrate wh ether the program model or level of services led to successful outcomes. Homelessness and Veterans The VA reports that on any given night nearly 200,000 veterans are hom eless ( U.S. Department of Veterans Affair s Fact Sheet Homeless Veterans, 2005). In the mid-1990s, it was found that the overall proporti on of veterans among homeless men was 41%, somewhat higher than the 34% of veterans in the general population (Rosenheck et al., 1994). Homeless veterans, as opposed to homeless nonveterans, are more likely to be white and older, to have higher education, a nd to be married or to have been married (Rosenheck & Koegel, 1993). No differences were found between homeless veterans and nonhomeless veterans in terms of resident ial instability, current social functioning, physical health, mental illness, or substance abuse (R osenheck & Koegel, 1993). However, a subsequent study indicated that there may indeed be a higher rate of substance abuse and unemployment among homeless veterans than among nonhomeless veterans (Rosenheck et al., 1994). Perhap s the most significant difference between homeless veterans and homeless nonveterans was that veterans between the ages of 20 to 34 were 4.76 times more likely to be homeless than those who were nonveterans in the same age group (Rosenheck et al., 1994).
36 Findings from the National Survey of Ho meless Assistance Providers and Clients ( Homelessness: Programs and the People They Serve, 1999) showed that 33% of homeless adult men were veterans, and approximately 31% of nonhomeless adult men were veterans. Gamache, Rosenheck, and Tessl er (2001) used the National Survey data to determine that the cohort at highest risk in earlier studies (ages 20 to 34), although now older, is still at highest risk. Of all veterans, homeless veterans have an increased mortality risk, particularly those that are older (Kas prow & Rosenheck, 2000). Among all homeless veterans, African-American veterans were likely to be younger; they had more problems with drugs, but white homeless veterans were more likely to have diagnos es of alcohol abuse or serious psychiatric disord ers (Leda & Rosenheck, 1995). The VA Northeast Program Evaluation Ce nter (NEPEC) compiles demographic data on veterans using specialized VA homele ss programs. For federal fiscal year 2003, demographic characteristics for this sample ( n = 41,696) can be summarized as follows: The average age was 49, and most served duri ng the Vietnam era (46%). The next largest percentage was the post-Vietnam service era (40%). Of the total sample, 46% were African-American, and 45% were white; 45% were divorced, and 30% were never married; 29% were unemployed, and 29% were working part-time or held irregular employment. Only 44% received any public assistance or support ( Health Care for Homeless Veterans Programs: The 17th Annual Report 2004). The impact of homeless veterans on the VAs medical care can be highlighted by a 1998 survey, which found that 12% of all inpatient s had been homeless at admission or
37 had lost their housing while in the hosp ital (Rosenheck & Se ibyl, 1998). The VAs response to serving the homeless veteran has been to establish specialized treatment programs to assist with housing and psychosocial treatment. The Health Care for Homeless Veterans (HCHV) Program was established by Public Law 100-6 on February 12, 1987. The HCHV Program was developed to provide health and mental health care, and othe r needed services, to homeless veterans. Rosenheck and Fontana (1994) revealed that homeless veterans indi vidual vulnerability to homelessness is most likely due to a multiplicity of psychiatric and nonpsychiatric factors. Close to 50% of the veterans enrolled nationally in HC HV programs manifested one or more severe psychiatric symptoms at screening (Rosenheck et al., 1989). Particularly relevant to treat ment was that participation in the programs was found to be associated with improvement in all areas of mental health and community adjustment. Additionally, improvement in psychiatric sy mptoms was associated with superior housing outcomes and improvement in community adjustment (Leda & Rosenheck, 1992). Homeless veterans represent a sample not too dissimilar to the general homeless population. The differences highlighted above are evident (Rosenheck & Koegel, 1993; (Rosenheck et al., 1994; Gamache et al., 2001), and adjusting for these differences could be accomplished with other samples. The homeless veteran population treated in VA services provides a national sample; they are a group of individuals who have been interviewed under consistent protocols and who are housed in programs with standardized admission and discharge data collection instruments. A study of this
38 population and the community-based transiti onal housing programs providing services to these individuals will provide useful inform ation for the study of other homeless groups and/or services designed to address the causes and effects of homelessness.
39 CHAPTER III: RESEARCH METHODS Sample In 1994, the U.S. Departm ent of Veterans Affairs, authorized by Public Law 102590, initiated the Homeless Providers Grant and Per Diem (GPD) Program. The GPD Program provides grants and operational funds for nonprofit or ganizations to create and maintain transitional housing programs for homeless veterans. Since 1994, the GPD Program has funded over 200 organizations, cr eating more than 300 residential programs and establishing more than 7,000 community-based transitional housing beds nationally. The GPD-funded programs currently represent th e nations largest integrated network of community-based transitional housing program s. Ranging from a three-bed, low-demand program for homeless who are disabled and di agnosed with a mental illness to a hundredbed sober living facility for vocationally-ori ented individuals, the programs are operated by the host organizations and monitored thr ough standardized protocols developed by VA. GPD-funded programs offer a viable and extensive transiti onal housing setting throughout the country. This sample reflects a diversity of transitional housing programs on a national level, with data available and collected utilizing consistent standards. Programs are required by federal law to opera te transitional housing model designs. Data consistency is ensured by the practice of each participant being interviewed by a clinician
40 adhering to standardized monitoring protocol s; services offered at each program are categorized by level and type using the same instrument. To date, no study has been conducted examining the relationships between participant characteristics, level of program services, and outcomes using Andersen s behavioral model of service utilization as the theoretical framework in this setting. This study was thus a secondary analysis of existing administrative data. In 2004 through 2005, approximately 21,908 homeless vete rans were served in community-based transitional housing programs operated by nonprofit organizations funded by VA under the GPD Program. Programs exist nationally in 45 states and the District of Columbia. Study Inclusion Criteria Participant inclusion in the study was pr edicated on adm ission into one of the designated community-based transitional housing programs and a diagnosis of mental illness with no co-occurring drug or alcohol a buse diagnosis. Diagnosis was determined by a clinician at the time of the initial interview. Female veterans were less than 3% of the total population and were exclud ed. The sample for this study was n = 2,502. Measures At adm ission and discharge from co mmunity-based tran sitional housing, participant interviews were conducted by VA clinical staff designated as liaisons to community-based homeless provider organi zations operating transitional housing projects for homeless veterans. Liaisons we re advanced-degree sta ff, most often holding masters degrees in social work or nursi ng curriculums. Structur ed interviews were
41 conducted using standardized procedures as de lineated in monitoring protocols published by VAs Northeast Program Evaluation Center (NEPEC). Program services information was documented by liaison staff on structur ed interview forms. These program assessments were completed by the VA liaisons in consultation with program managers of the community-based tran sitional housing organization. For this study, data utilized were from the following three data sets: 1. Participant-level admission data: Ad mission data contained descriptive information on each participant, including standard demographic information as well as combat experience, employment st atus, level of public support, amount of income, length of homelessness, number of homeless episodes, current living situation, perceived mental illness in a ddition to clinically evaluated mental illness diagnosis (see Intake Form X, Appendix A). 2. Participant-level discharge data: Di scharge data included the reason the participant left the program and the participants plans for living in the community (see Discharge Form D, Appendix B). 3. Program-level characteristics and services information: Program characteristics and services data included program certi fication status, whether the program was faith-based or secular, and the level of homeless-specific services offered (see Facility Survey Form, Appendix C).
42 Preliminary Data Analysis A prelim inary data analysis was performed on GPD participant and program data from 2004 to determine if the data set would provide a reasonable test of the studys theoretical framework and to ascertain the feasibility of conducting the study as proposed. An examination of program participant data from 2004 showed that 13% of the participants (1,641) had no substance abuse or mental health diagnosis; 39% (4,974) had no mental health diagnosis but were determin ed to have a substance abuse disorder; 39% were determined to have both a mental he alth and substance abuse disorder; and 9% (1,121), the portion of the sample under study, had a mental health disorder with no indications of substance abuse. To narrow the focus of this research on mental illness and homelessness, participants in the transitional housing programs diagnosed with a substance disorder or co-occurring disorders were excluded from the sample. As expected, the inclusion of 2005 data sa mple size approximately doubled program participant data (21,908). Of those 21,908 total program participants, 10% (2,189) had no substance abuse or mental health diagnos is; 77% (16,886) were determined to have a substance abuse disorder; a nd 13% (2,831), the sample under study, had a mental health disorder with no indications of substance a buse. It should be not ed that this figure (2,831) represented episodes in tran sitional housing not individual pa rticipants; the number of participants was 2,502 when duplicates were removed.
43 The total number of operating programs surveyed in 2004 was 274; for 2004 and 2005 combined, the total number of programs was 288. The program-related variables of interest in this study include d the program certification le vel, the progr am treatment philosophy, and the type of services provide d. Of the programs surveyed in 2004, 13% had a state mental health license, 32% reporte d having a state public health or state board of health certifica tion, and 15% of the programs had a national accreditation. Most programs reported their treatment philosophy as either a therapeutic community (23%) or adhering to a psychosocial rehabilitation model (20%). Other programs reported treatment philosophies including cognitive/behavioral models (15%) and 12-step models (16%). Sixty-seven percent of the progr ams were reported as secular or having no religious base; 33% were faith-based or historically a faith-based organization. The 2004 data set indicated that a wide-rang e of services was offered in the GPDfunded programs. Services and the percentage of programs that offered services directly (as opposed to referral to othe r staff or agencies) are illust rated in Table 1, as follows: discharge planning, 88%; case management, 85%; group or indivi dual therapy, 83%; housing assistance, 73%; money manageme nt, 70%; transportation assistance, 69%; social security assistan ce, 65%; outcome follow-up, 55%; vocational/educational counseling, 38%; aftercare counseling, 36%; family counseling, 27%; nutritional counseling, 19%; spiritual counseling, 18%; AI DS screening and counseling, 8%; payee services, 8%; and legal counseling 5%.
44 Table 1 Percentage of GPD Programs O ffering Various Direct Services Type of Service % Direct Service Discharge Planning 88 Case Management 85 Group or Individual Therapy 83 Housing Assistance 73 Money Management 70 Transportation Assistance 69 Social Security Assistance 65 Outcome Follow-Up 55 Vocational/Educational Counseling 38 Aftercare Counseling 36 Family Counseling 27 Nutritional Counseling 19 Spiritual Counseling 18 AIDS Screening & Counseling 8 Payee Services 8 Legal Counseling 5 The theoretical framework required va riables to be regarded under the predisposing, enabling and need factors and within those factors under the traditional and vulnerable population domains Hypotheses were based upon the level of association between these various factors and domains. A review of the 2004 and 2005
45 data sets indicated that variability of the sample and the sample size were adequate and that establishing these categories of factors and domains was feasible. Study Variables Data sets from 2004 and 2005 were com bined. Individual identifiers were removed and, because several fields contained limited responses, some variables were collapsed and reported in summary form. As de scribed later in this section, creation and modification of several variables from the initial data set was required to establish bivariate relationships. Based upon Andersens behavioral mode l (Andersen, 1995; Andersen & Aday, 1978), variables or individual determinants of service utilization are defined as predisposing, enabling or need factors. Predisposing factors are those preexisting subject characteristics. Enabling factors are those personal, family, or community resources that affect care. Need f actors are the perceived/subjective or professional/objective assessments of urgenc y for services or illness level. The enhancement of Andersens model by Gelberg et al. (2000), that is, the behavioral model for vulnerable populations, provides a distinction between the traditional and vulnerable domains to accommodate special populations. This enhanced mode l also allows the study of service utilization impact on health status outcomes. Using the Desai et al. (2003) and Gelberg et al. (2000) studies as a guide, variables were categorized according to factor and domain as illustrated in Table 2.
46 Table 2 Variables of the Study Independent Variables Domains Specific Factors Predisposing: Traditional: Age Ethnicity Marital Status Military Combat Status Vulnerable: Employment Public Support Entitlements Amount of Income Length of Homelessness Episodes of Homelessness Current Living Situation Enabling: Traditional: Program Certification Status Treatment Model/Philosophy Program Religious Basis Vulnerable: Level of Homeless-Specific Services Offered by Program Need: Traditional: Expressed Interest in Program Vulnerable:
47 Independent Variables Domains Specific Factors Perceived Mental Illness Mental Health Diagnosis (as Determined by Clinical Interview) Dependent Variables Program Outcomes Outcome 1: Particip ant Program Completion Status (Clinical Assessment) Outcome 2: Participant Housi ng Status at Discharge (Subject Response) Independent Variables Predisposing Factor Variables The predisposing-traditional domain variables included the dem ographic characteristic variables related to the prope nsity of the individual to use transitional housing. These variables represented the partic ipants characteristics existing before the illness, or in this case before the homeless ep isode, such as age, ethnicity, marital status, and military combat exposure. The predisposing-vulnerable domain variables included those variables existing before seeking transitional housing and rele vant for the study of homeless populations. Included in this domain were the variables representing the participants employment status, level of public support/ entitlements, amount of income, length of homelessness, number of homeless episodes in the last three years, and current living situation.
48 Data Collection. Data for predisposing-traditional and -vulnerable domains were obtained from the following items on th e program admission forms (see Form X, Appendix A): Questions 3 through 6, 8 th rough 10a, 27, and 29 through 34. This information was collected by the clinician inte rviewer at the time of contact with the participants and was used as screening fo r transitional housing programs. Time of contact with participan ts was usually within seven days of admission. Answers to all interview questions listed above were recorded as the s ubjects responses (see Table 3).
49 Table 3 Independent Variables, Predisposin g Factors (Admission Data Set) PredisposingTraditional Demographics Measure Type of Measure Test Data Source Age: Years Continuous T-Test Form X, #3 Ethnicity: Categorical Chi-Square Form X, #5 Hispanic White Black Asian American Indian Other Marital: Categorical Chi-Square Form X, #6 Married Separated Divorced Never Married Widowed Combat: Categorical Chi-Square Form X, #8 Yes No Employment: Days Worked Continuous T-Test Form X, #27
50 PredisposingVulnerable Characteristics of Vulnerability Measure Type of Measure Test Data Source Public Support: Categorical Chi-Square Form X, #29-33 None One Type Two Types Three Types Amount of Income: Income in Dollars Categorical Chi-Square Form X, #34 $ 0 1 49 50 99 100 499 500 999 1000 + Length: Days/Mos./Yrs. Categorical Chi-Square Form X, #10 0 1 29 days 30 days -<6 mos. PredisposingVulnerable Characteristics of Vulnerability Measure Type of Measure Test Data Source Length: Days/Mos./Yrs. Categorical Chi-Square Form X, #10 6 mos. <1 year 1-< 2 years 2 yrs + / unknown
51PredisposingVulnerable Characteristics of Vulnerability Measure Type of Measure Test Data Source Episodes: (last three years) Categorical Chi-Square Form X, #10 a 0 1 2 3 4 5+ Current Living Situation: Categorical Chi-Square Form X, # 9 Own Apt./Room Inst./Shelter None Enabling Factor Variables Andersen (1978) describes the enabling fact or as the variable s that depict the means individuals have avai lable for the use of services. In earlier studies that explored access to services using the behavio ral model framework, this component was restrictively defined as individual, family, or other supportive resources to utilize services in addition to the supportive aspects of the community. As discussed above, Gelberg et al. (2000) expanded the model to consider the actual utilization of services and its impact on outcome measures This study defined the enabling factors not as the
52 supportive resources of the individual, family, or community but as the actual level of services provided by the service, which is transitional housing. This concept of the enabling factor is evident in a number of earlier studies that incorporated the use of provider-related variables in the behavioral model framewor k as illustrated in Figure 5 (Phillips et al., 1998). Figure 5. Enabling Factors as Services The enabling-traditional domain variables for this study included the variables that represent the certification status of the community-based transitional housing program, the treatment model, and any religi ous basis of the orga nization that operates the program. The enabling-vulnerable domain variables included the variables that represent the level of services offered specifically for the homeless in each of the transitional
53 housing programs. These variables were desc ribed in the Preliminary Data Analysis section. Data collection. The following program categorie s were retained for analysis: program certification status, treatment model, religious basis, and level of program services (see Facility Survey form items: V II. 1.a. through 1.i.; VI. 1.; VIII 2.; and V. 1 through 23, Appendix C). Levels of certification were subset to create four levels of certification: no certification; st ate mental health licensed; national accreditation; and a sum category of multiple state licensing and/or national accreditations. Treatment models included eight categories: Medica l Model; Therapeutic Community; CognitiveBehavioral Model; 12-Step Model; Psyc hosocial Rehabilitation Model; FaithBased/Moral Training; Supportive Housing; a nd Other. Religious basis was further categorized as no religious base; a historical but not current religi ous base; and a clear religious orientation. A numeri cal score, which was calculate d for the level of services offered, reflected the number of types of different services provided in the program (see Table 4). As described in the preceding Measures section, program services information was documented by VA staff on structured interview forms. Information gathered on the Facility Survey form included the types of services offered in the GPD-funded program and whether the services were offered dire ctly by the program or through referral to another agency. The numerical score calculated to obtain the level of services offered was the total number of the various types of services that were provided directly by staff on-site at the GP D-funded program.
54 Table 4 Independent Variables, Enabling Fact ors (Program Characteristics) (Facility Survey Data Set) EnablingTraditional Program Services Nonspecific Measure Type of Measure Test Data Source Certification Status: Categorical Chi-Square Fac. Survey VII. #1 a-i None State License Natl. Accreditation Multiple sum Treatment Model: Categorical Chi-Square Fac. Survey VI. #1 Medical Therapeutic community Cognitive behavioral 12-step Psychosocial rehab Faith-based/moral Supportive housing Other Religious Basis: Categorical Chi-Square Fac. Survey VII. #2 None Historical not current Clear religious base EnablingVulnerable Program Services Specific Measure Type of Measure Test Data Source Services Offered: Total Score = Number of direct services Continuous T-Test Fac. Survey V. #1-23
55 Need Factor Variables Andersen refers to the need factor as th e illnes s level interpreted to be the most immediate cause of health care use. A di stinction in both the early model and later revisions of Andersens theory was the need as perceived by the individual and the need as evaluated by the delivery system. The need-traditional domain variables represent a measure of the participants perc eived need, that is, the subjects stated interest in the transi tional housing service. The need-vulnerable domain variables included the variable s relevant to the homeless population diagnosed with a mental illness: the subjects perceived mental illness as well as the clinically evaluated mental illness diagnosis. Data Collection. The participants stated intere st in the transitional housing service was indicated by code d item 59 on admission data (see Form X, Appendix A). Perceived mental illness data were also gathered from Fo rm X admission data according to the participants responses to item VI 23. The participants diagnosis is determined by items VIII 37 through 45 (see Table 5).
56 Table 5 Independent Variables, Need Factors (Admission Data Set) NeedTraditional Expressed Interest in Program Measure Type of Measure Test Data Source Subjects Interest: Categorical Chi-Square Form X, cd #39 -Did not talk, was not interested -Interested in only basic services -Interested in full range of services NeedVulnerable Mental Illness Measure Type of Measure Test Data Source Subject Perception: Categorical Chi-Square Form X, #VI 23 Yes No NeedVulnerable Mental Illness Measure Type of Measure Test Data Source Clinician Diagnosis: Type of Disorder Categorical Chi-Square Form X, #10 Schizophrenia Other psychotic Mood Personality Clinician Diagnosis: Type of Disorder Categorical Chi-Square Form X, #10 PTSD Adjustment Other Psychiatric
57 Dependent Variables Outcomes As defined by Gelberg et al. (2000), outcom es transcend the predisposing, enabling, and need factors. Outcome measures included two program measures related to success: how or under what circumstances the subject left the program and the subjects housing status upon discharge. Each outco me measure and asso ciation with the independent variables was reported separately. The first measure indicated the clinicians evaluation of the course of the individuals participation in the pr ogram, and the second represented the housing status upon discharg e from transitional housing as reported by the subject. These two measures were used to further distinguish between and offer a discussion of any differences between the cl inicians assessment of success and the subjects stated anticipated plans for housing upon discharge from a program. Data Collection. A program outcome was determin ed as either successful or nonsuccessful. Success was determined in tw o ways, subset into the following two outcome categories: Outcome 1: Participant Completion Status at Program Discharge (Completion Status): Success was defined by the VA clinician and indicated on question 11 (see Form D, Appendix B) if Number 1 was selected. The question reads as follows: The veteran ended the program because: 1. Successful completion of the program; 2. Veteran violated the program rules; 3. Veteran left the program on own decision without staff approval;
58 4. Veteran became too ill to continue the program; 5. Contract was terminated; and 6. Other. VA monitoring protocol states that if a program participant made substantial progress toward a documented treatment plan then, at the clinicians discre tion, that subject may be determined successful in the program. Upon discovery of limited responses in Numbers 5 and 6, these fields were collapsed into one field defined as Other. Outcome 2: Participant Housing Status at Program Discharge (Housing Status): The second determination of success in th e program was the program participants response to Question 16, anticipated living situation after discharge. Single room, halfway house, apartment, or other institution was defined as a successful outcome while no residence or leaving the program without indication of a residence was considered unsuccessful (see Table 6).
59 Table 6 Outcomes of Program Discharge (Report of Discharge Data Set) Program Outcomes How and under what circumstances the subject left the program. Measure Data Source Outcome 1: Program Completion Status Clinician Assessment Form D III. #11 Successful completion Violation of rules Own decision without staff advice Became too ill to continue Other Outcome 2: Housed Upon Program Completion Subject response Form D #13 Apartment, room, institution No residence, no response Statistical Analysis and Results Level One Analysis Frequencies of distribution were con ducted for both the independent and dependent variables to provide a descriptive analysis of the study group, that is, m ale homeless veterans who were diagnosed with a mental illness with no co-occurring substance abuse disorders and who utilized GPD-funded transitional housing throughout
60 the country for 2004 and 2005. For veterans who enrolled in a program more than once (329), their first episode of enrollment was uti lized for the sample. Also included in this analysis were the characteristics of the pr ograms that provided these services and the program outcomes. Table 7 represents a demographic analysis utilizing the independent predisposing variables and provides subject pr ofiles that reflect population types and characteristics of the study group (n = 2,502). For comparison, the table also includes subject profiles of those without a mental illness diagnosis in the population housed in these transitional programs. Most study participants were betw een the ages of 40 to 49 or 50 to 59 (35.7% and 40.7%, respectively). A majority were white (58.8%); most were divorced (48.7%) or never married (29.5%); and 20.6% reported being in combat while in the military.
61 Table 7 Predisposing Variable Frequencies (Admission Data Set) PredisposingTraditional Demographics Field Study Participants No Mental Illness Group n = 2,502 % No Response n = 2,189 % No Response Age: 0 5 (.2%) 20-29 69 2.8 66 3.0 30-39 252 10.1 240 11.0 40-49 894 35.7 847 38.7 50-59 1019 40.7 766 35.5 60-69 212 8.5 201 9.2 70-79 41 1.6 57 2.6 80+ 15 0.6 12 0.5 Ethnicity: 4 (.2%) 5 (.2%) Hispanic 135 5.4 138 6.3 White 1470 58.8 975 44.5 Black 810 32.4 984 45.0 Asian 10 .4 13 0.6 Am. Indian 37 1.5 37 1.7 Other 36 1.4 37 1.7
62 PredisposingTraditional Demographics Field Study Participants No Mental Illness Group Marital: 6 (.2%) 5 (.2%) Married 124 5.0 135 6.2 Separated 322 12.9 237 10.8 Divorced 1219 48.7 1006 46.0 Never Married 739 29.5 739 33.8 Widowed 92 3.7 67 3.1 Combat: 0 16 (7%) Yes 515 20.6 354 16.2 No 1987 79.4 1819 83.1 Employment: 9 (.3%) 10 (.5%) None 1936 77.3 1504 68.7 1-10 261 10.4 274 12.5 Employment: 11-20 201 8.0 245 11.2 21-31 95 3.7 156 7.1 Public Support: 3 (.1%) 5 (.2%) None 1269 50.7 1402 64 1Type 999 39.9 709 32.4 2 Types 208 8.3 67 3.1 3 Types 23 .9 6 0.3
63 PredisposingVulnerable Characteristics of Vulnerability Study Participants No Mental Illness Group Amount of Income: 19 (.8%) 18 (.8%) $0 803 32.1 793 36.2 1-49 85 3.4 89 4.1 50-99 90 3.6 103 4.7 100-499 583 23.3 554 25.3 500-999 667 26.7 456 20.8 1000 + 255 10.3 176 0.8 Length: 5 (.2%) 14 (.6%) 0 days 184 7.4 150 6.9 1 29 days 686 27.4 673 30.7 30 days <6 months 752 30.1 729 33.3 6 months <1 year 300 12.0 231 10.6 1 <2 years 208 8.3 142 6.5 2 years + 367 14.7 234 10.7 Unknown 0 .0 15 0.7 Episodes: (last three years)1 1502 (60%) 1392 (63.6) 0 119 4.8 105 4.8 1 577 23.1 485 22.1 2 167 6.7 131 6.0 3 72 2.9 37 1.7 4 25 1.0 12 0.5 5+ 40 1.6 28 1.3
64PredisposingVulnerable Characteristics of Vulnerability Study Participants No Mental Illness Group Current Living Situation: 15 (.6) 28 (1.3) Own apt/rm 525 21.0 478 21.8 Inst/shelter 1492 59.6 1209 55.2 None 470 18.8 474 21.7 *Note: question added to the su rvey late in the study period. Under the predisposing factors in the vulne rable domainthose characteristics that represent vulnerability to homelessnessthe percentage of those reporting no employment within the last thirty days was 77.3%; 50.7% reported receiving no public support, and 32.1% reported no income prior to admission to a program. Over 29% in the study group were homeless between 30 days and 6 mont hs prior to admission (29.7%); 27.4% were homelessness 1 to 29 days before enteri ng a GPD-funded program. The study group was not likely to have multiple episodes of homelessness: less than 13% reported two or more episodes in the last three years.1 More than half of the study group reported living in a shelter or institution at the time of the interview (59.6%). The enabling-traditional and enablingvulnerable domains of the study included those variables that repres ented characteristics of the transitional housing programs utilized by the study particip ants. In 2004 and 2005, 288 programs were surveyed (Table 8). Over 46% of the programs surveyed had no state or national cer tification or license. 1 Episodes of homelessness was added as a survey category in early 2004. Data collection instruments were not distributed nationally for use until the end of the study period.
65 Of the remaining programs, 50.0% reported ha ving state licensure, and 13.5% reported having a national accreditation. Most frequently reported program treatment philosophies were therapeutic communities (23.6%); cognitive-behavioral (15.3%), 12step (15.6%), or psychosocial models ( 20.8%); or supportive housing (14.6%). A majority of the programs had no religious ba sis (66%); 18% had a historical religious basis but were not currently a religious model, and only 14.9% had a clear and current religious base. Under the enabling-vulnerable domain, program services offered directly to and specifically for the study group were calculated as a services quotient the sum of services provided within the program offered specifically for the sample and only directly by program staff. Most programs service quot ient was in the range of 31-40 (41.0%). The next highest frequency was in the range of 41-50 (31.4%).
66 Table 8 Enabling Variable Frequencies (Facility Survey Data Set) EnablingTraditional Program Services nonspecific Measure No of programs utilized n = 288 % Certification Status: 1 missing (.3%) None 134 46.7 State license 144 50 Natl. accreditation 39 13.5 Multiple sum 153 53.1 Treatment Model: 2 missing (.7%) Medical 6 2.1 Therapeutic 68 23.6 Cognitive behavioral 44 15.3 12-step 45 15.6 Psychosocial rehab 60 20.8 Faith-based/moral 9 3.1 Supportive housing 42 14.6 Other 12 4.2 Religious Basis: 3 missing (1.0%) None 190 66 Historical but not current 52 18 Clear religious base 43 14.9
67 EnablingVulnerable Program Services Specific Measure No of programs utilized n = 288 % Services Offered: Total Score number of services offered directly (subset by participant) 0-10 52 1.8 11-20 61 2.2 21-30 301 10.6 31-40 1160 41.0 41-50 889 31.4 51-60 368 13.0 The traditional and vulnerable domains of the need factor frequencies are represented in Table 9. Need was determined by the subjects interest in the program (traditional domain) and the subjects perceived mental health as well as the clinicians diagnosis (vulnerable domain). Most of those interviewed in the study group (78.8%) expressed an interest in a fu ll range of homeless services; and 64.3% of those interviewed reported that they believed they had a curr ent psychiatric or emotional problem other than alcohol or drug use. A majority of the study group was diagnosed with a mood disorder (56.8%). The next most recurri ng diagnosis was an adjustment disorder (39.3%). The remaining diagnostic frequenc ies were other psychiatric disorders (15.5%), post traumatic stress disorder (PTSD, 10.9%), sc hizophrenia (8.7%), other psychotic disorder (7.6%), and personality disorder (7.6%).
68 Table 9 Need Variable Frequencies (Admission Data Set) NeedTraditional Expressed Interest in Program Measure Study Frequency n = % No Response Subjects Interest: 227 (9.1%) -Did not talk, was not interested 12 .5 -Interested in only basic services 292 11.7 -Interested in full range of services 1971 78.8 NeedVulnerable Mental Illness Measure Study Frequency n = % No Response Subjects Assessment: 0 Yes 1610 64.3 No 892 35.7 Clinician Diagnosis: Type of Disorder Schizophrenia 217 8.7 Other psychotic 189 7.6 Mood 1421 56.8 Personality 189 7.6 PTSD 272 10.9 Adjustment 983 39.3 Other psychiatric 388 15.5
69 The outcomes measures from the transitional housing programs were the dependent variables (Table 10). Outcomes were recorded as bot h program participant completion status as determined by VA staff (O utcome 1) and housing status at discharge as reported by the participan t (Outcome 2). A majority of those in the study group discharged from GPD-funded programs left successfully as measured by Outcome 1 (52.2%). A relatively equal number of participan ts left either because of a violation of rules (17.5%) or by their own decision (18.5%). Only 6.9% became too ill to continue the program. As measured by housing status upon discharge (Outco me 2), 84.4% of the participants stated that th ey would be living in an apartment, room, or other institution/program upon discharge; 15.4% reported they had no residence upon discharge or did not respond.
70 Table 10 Outcome Variable Frequencies (Report of Discharge Data Set) Program Outcomes How and under what circumstances the subject left the program. Measure Study Frequency n = % No Response Outcome 1: Program Participant Completion Status Successful completion 1307 52.2 Violation of rules 438 17.5 Own decision without staff advice 463 18.5 Became too ill to continue 172 6.9 Other 122 4.9 Outcome 2: Participant Housing Status at Discharge 5 (.2%) Apartment, room, institution 2112 84.4 No residence, no response 385 15.4 Variable frequencies were also examined for those participants in the programs who were not diagnosed with a mental illnes s or co-occurring substa nce abuse disorder. As indicated in Table 7, popul ations of the study group and those program participants not diagnosed with a mental illness were similar. The study group had a larger percentage between the ages of 50 and 59 (40.7% as opposed to 35.5% in the group without a diagnosis of mental illness). A larger percentage of those in the study group were white (58.8%) as opposed to those in the group with no mental illness (44.5%).
71 Those in the study group were less likely to have worked prior to admission to the program; as expected, however, more of those in the study group reported receiving public support. Frequencies of the number of episodes of homelessness in the last 3 years between the two groups were similar as we re the frequencies reported on respondents current living situation. Summary, Level One Analysis The sam ple size for this study was 2,502. Mo st subjects were between the ages of 40-49 or 50-59; the majority were white, and most had been divorced or never married. Similar to nonhomeless veterans, 20% repor ted experiencing combat while in the military. Most were not employed prior to being admitted to one of the GPD-funded programs and about one-half received some t ype of public support. A majority were homeless between 1 to 29 days or 30 days to 6 months, and more than half reported living in a shelter prior to admission. Almost half of the programs reported no state or national certific ation, license, or accreditation. Of those that had some type of independent review or certification, most were licensed by the state. Slightly more than 10% had an accreditation by a national accrediting body. With the exception of medical models or the faith-based/moral training program models, the various program treatment philosophies were equally represented in the sample. Most programs reported no clear or specific religious basis. A majority of the programs had a mid-range services-quotient score (31-40). Appr oximately one-third of the sample was represented by programs w ith the next higher-ra nge services-quotient (41-50).
72 Most of those interviewed in the study group expressed an in terest in the full range of services available. As expected, fr equencies were high in both the subjects selfreport of perceived mental i llness as well as in all dia gnostic categories reported by the clinicians. The cumulative analysis of the dependen t variables offers a summary of the outcomes of GPD-funded transitional housing programs. Later sections of this dissertation provide discussi on on the usefulness of these programs for the sample population. Outcome summaries also will add to discussions on program-designeffectiveness assumptions of the transitional housing model for addressing homelessness. Level Two Analysis The second level analysis includ ed tests of bivariate association for each hypothesis as described in Table 3. Aneshensel (2002) emphas izes the importance of the bivariate analysis. The au thor states that, although often overlooked on the way to multivariate methodology, the multivariate desi gn rests upon the foundation laid through analysis of the two-variable model. A delib erate bivariate analysis linked with each of the hypotheses of this study provided specifi c levels of significance of the prime theoretical variables of inte rest rather than estimates. In addition, this analysis contributed to the development of focal relationships as recognized by Aneshensel: The first analytic step is to establish that the focal relationship is feasible, that two variables may be related to one another. This goal is realized by demonstrati ng that the two variables are
73 empirically associated with one another [;] further analysis serves to evaluate whether the focal relationship is indeed a relationship or merely an association (p. 11). The hypothesis testing that follows uses bi variate analysis (Table 11). Analysis was conducted on data sets with and without duplicates removed. The data set without duplicates removed provided anal ysis of associations with dependent variables and an enrollment episode in a program without re gard to subject. When duplicates were removed, the data set represented associations with dependent variables and individual subjects; tests consisted of pa rticipant outcomes, not episode outcomes. For participants who had more than one transi tional housing experience, only the first enrollment episode was included in the sample. To develop odds ratios for significant re lationships, program completion status (Outcome 1) was subset in to two categories to reflect successful or nonsuccessful completion of the program. Success was determined as the first response on Question 11 and nonsuccess by Questions 2 through 5, Form D (see Appendix B). This redefined Outcome 1 was used in further analyses of hypotheses la ter in this section and is referred to as Outcome 1a Several independent variable s were subset to calculate odds ratios and, in some cases, to compensate for limited responses in several fields.
74 Table 11 Study Hypotheses and Behavioral Model Factors ParticipantLevel Hypotheses Factors and Domains Speci fic Factor Association with Outcomes One Predisposing Traditional Subject Demographics No Association Two Predisposing Vulnerable Vulnerability of homelessness Negative Three Need Traditional Expressed interest in program Positive Four Need Vulnerable Perceived mental illness Positive Five Need Vulnerable Mental health diagnosis No Association Program-Level Hypotheses: Factors and Domains Speci fic Factor Association with Outcomes Six Enabling Traditional Program certification status Positive Seven Enabling Traditional Treatment model No Association Eight Enabling Traditional Religious Basis No Association Nine Enabling Vulnerable Homeless-specific services offered Positive Participant-Level Hypotheses Hypothesis One. There is no significant a sso ciation between participant demographics and successful completion of community-based transitional housing.
75 The category of Participant Age was gathered on the collection instrument by recording the Date of Birth. Other demogra phic variables of Ethni city, Marital Status, and Military Combat were collected as cate gorical level data. Levels of association between program outcomes and date of birth we re determined through t-tests; chi-square was utilized to determine levels of a ssociation between program outcomes and the categorical independent variables. No statistically significan t relationships were found be tween participant age and either of the program outcomes for the samp le without duplicates removed. Significant associations existed between ethnicity a nd Outcome 1 (program completion status) [ X 2 (20, n = 2,831) = 63.27, p < .05)] as well as ethnicity and Outcome 2 (housing status) [ X 2 (5, n = 2,831) = 12.178, p < .05)]. Ethnicity was categor ized as Hispanic, White, Black, Asian, American Indian, and Other. Consider ing that several cells had limited responses (Asian, American Indian, and Ot her), this variable was furthe r subset into White versus Nonwhite. When reexamining the variable of race dichotomously with the duplicates removed, significant associations were f ound between both Outcome 1a (dichotomous variable representing prog ram completion status) [ X 2 (1, n = 2,502) = 7.05, p < .05] and Outcome 2 [ X 2 (1, n = 2,502) = 6.27, p < .05)]. A significant association was also found between participants combat experience and Outcome 1a [ X 2 (1, n = 2,502) = 5.17, p < .05]. As determined by Outcome 1a, whites ha d a 24% better chance of being successful from a GPD-funded program than nonwhites. As determined by Outcome 2, whites were 32% more likely to be housed at discha rge from the program. Participants who experienced combat while in the military were 25% more likely to be successful in the
76 programs as opposed to those who did not experience combat, as determined by Outcome 1a. At the p < .05 level of significance, there wa s no association between combat and participants likelihood of being housed at discharge from the program (Outcome 2). Hypothesis Two There is a negative association between participant severity of homelessness and successful completion of community-based transitional housing. The characteristics of vulnera bility to homelessness were determined as follows: The category of Employment was represented by the number of days worked in the last 30 days. The category of Public Support En titlements was subset into the following: None; VA, Social Security, or other public supports; In receipt of two of the types of public support; and In receipt of all three types. Amount of Income was divided into categor ies represented by the dollar amount of income received in the last 30 days. Lengt h of Homelessness was recorded according to the following categories: Not homeless; At least one night but le ss that one month; At least one month but less than six months; At least six months but less than one year; At least one year but less than two years; and Two years or more.
77 Episodes of Homelessness were recorded as the number of separate episodes the participant experienced in the la st 3 years, one to five epis odes (or more). Current Living Situation was subset into the following categories: Own apartment, room, or house or sharing with friend or family; Institution or shelter/temporary housing program; and No residence, living outdoors or in an abandoned building. Levels of association between empl oyment (days worked) and both outcomes were determined by t-tests. Levels of asso ciation with categorical independent variables and program outcomes were determined by conducting tests of chi-square. The number of sources of public support pa rticipants reported before admission to the program was statistically significant with Outcome 1 (p rogram completion status) [ X 2 (12, n = 2,831) = 21.70, p < .05)]. However, when this variable was further separated into public support versus no public support a nd duplicate participants were removed, this variable was not significantly associated with Outcome 1a (the dichotomous variable representing program completion status). The amount of income in the last thirty days was significantly associated with Outcome 1 [ X 2 (20, n = 2,831) = 36.05, p < .05)] but no significance was found between this indepe ndent variable and Outcome 1a with the non-duplicated sample. Chi-square values fo r length of homelessness and current living situation before admission to the program were not statistically significant with either Outcome 1 or 1a. Statistically significant re lationships were found betw een several predisposing factors in the vulnerable doma in with a participants hous ing status upon discharge from
78 the program (Outcome 2). Although employment was not statistically significant for this outcome, a significant relationship was evid ent between public support and Outcome 2 when the public support variable wa s expressed in bivariate form ( none versus any ) and duplicates were removed [ X 2 (6, n = 2,502) = 4.34, p < .05)]. Those participants who reported any public support before admission to a GPD-funded program were 26% more likely to be housed upon discharg e. A significant relationship was also evident between length of homelessness and Outcome 2 [ X 2 (6, n = 2,831) = 22.99, p < .05)]. This variable was further subset into dichotomous categories of length of homelessness (0-30 days versus 31 days or more) and duplicat es were removed. For this dichotomous variable, a significant associati on with Outcome 2 was evident [ X 2 (1, n = 2,502) = 7.75, p < .05)]. As determined by Outcome 2, partic ipants in the program were 29% less likely to be housed when leaving the transitional hous ing if they were homeless more than 30 days prior to program admission. The current living situation of th e participant at the time of the interview was significantly associated with Outcome 2 [ X 2 (2, n = 2,831) = 17.05, p < .05)]. However, when this relationship was subset and expressed in bivariate form and duplicates were removed, no statistically significant association was evident. Hypothesis Three. There is a positive associatio n between participant expressed interest in program utilization and successful completion of community-based transitional housing. The participants Expressed Interest in Program utiliz ation was separated into categories that represent a measure of interest as follows: Did not talk to interviewer or not interested in services;
79 Interested in only basic services; and Interested in a full range of services for the homeless. Association with categorical outcome measures and expresse d interest in the program was determined through chi-square. A significant association existed between expressed interest in the program and both dependent variables, program completion status (Outcome 1) [ X 2 (8, n = 2,831) = 26.25, p < .05)] and housed upon discharge (Outcome 2) [ X 2 (2, n = 2,831) = 26.10, p < .05)]. This variable was further categorized as either those who e xpressed interest in basic services or no interest in the program, or those who e xpressed interested in a fullrange of services. When duplicates were removed, significant associations were found between this dichotomous variable with both Outcome 1a (the dichotomous variable representing program completion status) [ X 2 (1, n = 2,502) = 14.86, p < .05)] and Outcome 2 (housing status upon program discharge) [ X 2 (1, n = 2,502) = 9.12, p < .05)]. As determined by both Outcome 1a and Outcom e 2, those who expressed interest in a full range of services were 46% more likely to be successful as well as housed after participation in GPD-funded programs. Hypothesis Four. There is a positive associati on between partic ipant perceived mental illness and successful completion of community-based transitional housing. The category of Perceived Mental Illness was recorded and measured by the participants response to interviewers questions as a yes or a no. Chi-square was the method of analysis to determine asso ciation with the outcome measures.
80 No significant association was found betw een perceived mental illness and either dependent variable. Hypothesis Five. There is no significant associ ation between participant mental health diagnosis and successful completion of community-based transitional housing. Mental Health Diagnosis, determined by a clinician at the time of interview, was categorized as one of the following seven diagnostic typologies: 1. Schizophrenia; 2. Other Psychotic Disorder; 3. Mood Disorder; 4. Personality Disorder; 5. Post-Traumatic Stress Disorder (PTSD) from Combat; 6. Adjustment Disorder; and 7. Other Psychiatric Disorder. The association of Mental Health Di agnosis with outcome measures was determined utilizing chi-square. Several diagnostic variables were associat ed with both dependent variables. For Outcome 1 (program completion status), schizophrenia [ X 2 (4, n = 2,831) = 20.99, p < .05)]; mood disorder [ X 2 (4, n = 2,831) = 11.66, p < .05)]; PTSD [X 2 (4, n = 2,831) = 12.21, p < .05)]; and adjustment disorder [ X 2 (4, n = 2,831) = 13.13, p < .05)] were found to be statistically signi ficant. To express the rela tionship between diagnosis and program success in bivariate form, Outcome 1a (the dichotomous va riable representing program completion status) was used as th e dependent variable, and duplicates were
81 removed. Significant associations were found for schizophrenia [ X 2 (1, n = 2,502) = 11.03, p < .05] and for adjustment disorder [ X 2 (1, n = 2,502) = 4.72, p < .05]. As determined by Outcome 1a, those diagnosed wi th schizophrenia were 38% less likely to be successful in a GPD-funded program. Ho wever, those who were diagnosed with adjustment disorder had a 19% greate r chance of success in the program. For Outcome 2 (housing status upon progr am discharge), both mood disorder [ X2 (1, n = 2,502) = 4.42, p < .05)] and other psyc hiatric disorder [ X 2 (1, n = 2,502) = 5.39, p < .05)] were statistically significant. For those who were diagnosed with a mood disorder, there was a 26% gr eater chance of be ing housed upon discharge from a GPDfunded program. However, those participants diagnosed as other psychiatric disorders were 28% less likely to be housed after a GP D-funded program than those participants not diagnosed as such.
82 Table 12 Dichotomous Variable Frequencies n = 2,502 n % No Response Completion Status (Outcome 1a): 0 Success 1307 52.2 Nonsuccess 1195 47.8 Housing Status (Outcome 2): 5 (.2) Housed 2112 84.4 Not Housed 385 15.4 Ethnicity: 4 (.2) White 1470 58.8 Nonwhite 1028 41.1 Marital Status: 6 (.2) Married 124 5.0 Not Married 2372 94.8 Combat: 0 No 1987 79.4 Yes 515 20.6 Public Support: 3 (.1) No 1269 50.7 Yes 1230 49.2 Income: 19 (.8) No 803 32.1 Yes (any) 1680 67.1
83n = 2,502 n % No Response Length of Homelessness: 5 (.2) 0-30 days 870 34.8 31 days or more 1627 65.0 Living Situation: 15 (.6) Room, institution, shelter 2017 80.6 None 470 18.8 Program Certification: 1 None 134 Any 153 Religious Base: 3 None 190 History or current 95 Table 13 Program Completion Status (Outcome 1a): Hypotheses One Through Five n = 2,502 Value df Asymp. Sig. (2-sided) Odds Ratio 95% Confidence Intervals Lower Upper White/Nonwhite 7.056 (b) 1 .008 1.241 1.058 1.456 Combat 5.172 (b) 1 .023 1.254 1.032 1.524 Interest in Program 14.863 (b) 1 .000 1.459 1.203 1.769 Schizophrenia 11.033 (b) 1 .001 .622 .469 .825 Adjustment Disorder 4.716 (b) 1 .030 1.195 1.017 1.404 N of Valid Cases 2502
84 Table 14 Housing Status upon Program Discharge (Out come 2): Hypotheses One through Five n = 2,502 Value df Asymp. Sig. (2-sided) Odds Ratio 95% Confidence Intervals Lower Upper White/Nonwhite 6.272 (b) 1 .012 1.321 1.062 1.643 Public Support 4.339 (b) 1 .037 1.261 1.013 1.570 Length of Homelessness 7.747 (b) 1 .005 .714 .562 .906 Interest in Program 9.119 (b) 1 .003 1.465 1.142 1.879 Mood Disorder 4.419 (b) 1 .036 1.263 1.016 1.570 Other Psychiatric Disorder 5.398 (b) 1 .020 .719 .544 .951 N of Valid Cases 2502 Summary, Level Two Analysis Statistical associations were found in several of the above bivariate relationships between the dependent outcome variables and those independent variables represented as the predisposing and need factors in both the traditional and vulnerable domains (Tables 13 and 14). To express odds ratios for Outcome 1 (program completion status), a dichotomous variable was constructed (Outco me 1a). Also, when indicated, duplicate participants were removed from the sample to determine outcomes based on individuals rather than on episodes, and selected indepe ndent variables were subset dichotomously in order to express odds ratios or because limited responses existed in multiple fields. Significant relationships were found between both ethnicity and combat experience for program completion status (Outcome 1a) a nd ethnicity for housing status upon discharge
85 (Outcome 2). Those participants who were white and reported experiencing combat in the military had a greater lik elihood of success in GPD-funded programs. Those who received public support and those who we re homeless less than 30 days prior to admission were more likely to be housed upon discharge from a GPD-funded program (Outcome 2). Participants who expressed an interest in a full range of services were more likely to be successful in GPD-funded programs and more likely to be housed upon discharge. No association was found between perceived mental illness and either dependent variable. However, several diagnostic indepe ndent variables were associated with both dependent variables. For Outcome 1a, those who had a diagnosis of schizophrenia were less likely to be successful; those diagnosed with adjustment disorder had a greater chance of success. As determined by Outc ome 2, those diagnosed with mood disorder had a greater chance of being housed upon discharge from a GPD-funded program; participants diagnosed with oth er psychiatric disorders were less likely to be housed. Level Three Analysis The third le vel of statistical analysis pr ovided the data necessa ry to assess the theoretical framework of this study. This study proposed to test a revised framework of the behavioral model of util ization for vulnerable populations theory and to explore the assumptions of the utilization of the transitional housing model for addressing homelessness. To pursue this theoretical di scussion, the Level Three Analysis focused on and examined the influence of program cer tification, program philosophy, and program
86 services on transitional housing outcomes as well as participant characteristics and mental health diagnostic predictors of transitional housing program completion. As described previously, the level of serv ices offered (program certification status and program services) and type of overall service philosophy (treatment model and religious basis) were considered the enabling factors of the be havioral utilization model. Gelberg et al. (2000) examined impact of servic es on outcomes. In this analysis, services were enabling factors. As such, this study expl ored the implications for defining services as enabling factors as well as the influence of these factors on the relationships between predisposing and need factors with pa rticipant success in the program. Hypotheses Six through Nine begin the discussion regarding the effects of both the traditional and vulnerable enabling fact or domains on the associations between domains representing program participant char acteristics and mental health diagnoses with outcomes. First, bivariate associations between the enabling factors and outcomes were examined. Program-Level Hypotheses Hypothesis Six. There is a positive associa tion between program certification status and participants successful comple tion of community-based transitional housing. The category of Program Certification Stat us was subset to represent a graduated level of certification as follows: 1. None; 2. State licensed; 3. National accreditation; and
87 4. State license and national accreditation. Chi-square was utilized to determine asso ciation with the outcome variables. No statistical significance was found between the above program cer tification status variable with either Outcome 1 or 2 at the p < .05 level. This variable was further analyzed by recategorizing the programs as having any certification/accreditation or having none. No significant relationships we re revealed under this di chotomous categorization. Hypothesis Seven. There is no significant asso ciation between the type of treatment model of a community-based tran sitional housing program and participants successful completion. The Treatment Model was subset to repr esent a categorical level variable as follows: Medical model; Therapeutic community; Cognitive-behavioral therapy or social learning model; 12-Step model; Psychosocial rehabilitation model; Faith-based/moral training model; Supportive housing with no specifi c treatment philosophy; and Other. Chi-square was utilized to determine a ssociations with outcome variables. Several significant associations were found between program treatment models and Outcome 1 (program completion status): therapeutic community model [ X 2 (4, n =
88 2,831) = 13.68, p < .05)]; cognitive behavioral model [ X 2 (4, n = 2,831) = 9.97, p < .05)]; 12-step model [ X 2 (4, n = 2,831) = 14.13, p < .05)]; and psychosocial rehabilitation model [ X 2 (4, n = 2,831) = 20.07, p < .05)]. Treatment categories of the medical model and faith-based/moral trai ning model yielded limited frequencies ( n = 6; n = 9, respectively) and were excluded from this analysis. To express these relationships in bivariate form, Outcome 1a (the dichotomous variable representi ng program completion status) was utilized, and duplic ate participants were removed; the following statistically significant relationships were determined: therapeutic community model [ X 2 1, n = 2,502) = 9.77, p < .05]; cognitive behavioral model [ X 2 1, n = 2,502) = 8.43, p < .05 ]; 12-step model [ X 2 1, n = 2,502) = 9.23, p < .05 ]; and psychosocial rehabilitation model [ X 2 1, n = 2,502) = 11.08, p < .05]. As opposed to ot her programs models, participants were 27% less likely to be successful if they were enrolled in a therapeutic community model and 38% less likely to be successful if enrolled in a 12-step model program. Those that were admitted to a cognitive behavioral model program had a 40% greater chance of success and those admitted to a psychosocial rehabilitation model program had a 35% greater chance of achieving success as determined by Outcome 1a. No associations were found between progr am models and progr am participants likelihood of being housed upon discharge (Outcome 2). Hypothesis Eight. There is no significant associ ation between the religious basis of a community-based transitional housi ng program and participants successful completion. The program Religious Basis represented a categorical level variable as follows:
89 A private or public secular agency wi th no religious base or history; A private agency that at one time had a re ligious orientation but has evolved into an agency that is largely secularly based; and A private agency that conti nues to have a clear religi ous base and orientation. Chi-square was utilized to determine a ssociations between a programs religious basis with outcome variables. No statistic ally significant relationship was determined between the programs religious basis and Outcome 1 (program completion status). However, a significant association existed between this independent variable and Outcome 2 (housing status upon discharge) [ X 2 (2, n = 2,831) = 9.43, p < .05)]. The variable was further divided by categorizing programs as either faith-based (as determined by current faith-based status or a historical orientation of being faith-based) versus those programs that were identified as secular. Statistical significance was determined between Outcome 2 and this dichotomous variable [ X 2 (1, n = 2,502) = 4.05, p < .05)]. Participants in faith-based pr ograms were 19% less likely to be housed upon discharge than participan ts in secular programs. Hypothesis Nine. There is a positive associati on between the level of homelessspecific program services and participants successful completion of community-based transitional housing. The levels of Homeless-Specific Services were determined by the total score of the number of different services directly pr ovided within each program. T-tests were utilized to examine significance of associations with outcomes.
90 No statistical significance was observed be tween services and either of the two outcome measures. This variable was further defined categorically in dichotomous fields (below/above total score of 41) and utilizing chi-square, no statistically significant association with Outcome 2 was determined. Table 15 Program Completion Status (Outcome 1a): Hypotheses Six Through Nine Value df Asymp. Sig. (2-sided) Odds Ratio 95% Confidence Intervals Lower Upper Therapeutic Community 9.771 (b) 1 .002 .727 .595 .888 Cognitive Behavioral 8.432 (b) 1 .004 1.403 1.116 1.766 12Step 9.230 (b) 1 .002 .621 .456 .846 Psychosocial Rehabilitation 11.080 (b) 1 .001 1.348 1.130 1.607 N of Valid Cases 2502 Table 16 Housed Upon Program Discharge (Outco me 2): Hypotheses Six Through Nine Value df Asymp. Sig. (2-sided) Odds Ratio 95% Confidence Intervals Lower Upper Religious Basis 4.048 (b) 1 .045 .807 .635 .994 N of Valid Cases 2502
91 Summary, Hypotheses Six through Nine Hypotheses Six through Nine were tested by exploring bivariate associations between both outcom e variables and independent variables categorized as the enabling factors in both the traditional and vulnerable domains (Table s 15 and 16). No statistical significance was found between pr ogram certification status a nd either outcome. Several significant associations were found between program treatment models and Outcome 1a (dichotomous program completion). Those participants in cognitive behavioral or psychosocial rehabilitation models were mo re likely to achieve success; those in therapeutic community or 12-st ep models were more likely to be unsuccessful. It is interesting to note that these same treatment model associations were not evident in associations with housing at discharge (Outcome 2). No statistically significant relationship was determined between a pr ograms religious base and Outcome 1a; however, an association existed between the religious basis of a program with Outcome 2. Participants were less likely to be housed at discharge if they were enrolled in faithbased programs. Within the vulnerable domain, no statis tical significance was observed between level of services offered in the pr ograms and the two outcome measures. Logistic Regression Models The bivariate analyses of the study hypot heses perform ed above were a deliberate effort to identify the prime theoretical v ariables of interest. As discussed by Aneshensel (2002), this analytic step is n ecessary to establish the focal relationships
92 through which further analysis serves to eval uate whether these relationships are indeed relationships or merely associations (p. 11). Additionally, as suggested by Gelberg et al. (2000), enabling factors influence services util ization and therefore are potential effectors of outcomes. For the purposes of the Le vel Three Analysis, these transitional housing services are enabling factors that may contribute to the subjects improved outcome. However, as evidenced above, utilizing bivari ate tests of signifi cance, the effects of program philosophy and services (the enabli ng factors) on successf ul outcomes may not be as evident as theorized. Additionally, pa rticipant characteristics and mental health diagnosis may not be obvious predictors of success in transitional housing outcomes. To investigate these relationships further, logi stic regression models were established. Logistic regression was used to test associations while controlling for variables identified in the bivariate analyses. Models were constructed for both Outcome 1a (the dichotomous variable representing program completion status) and Outcome 2 (housing status upon discharge from a GPD-funded progr am). Table 17 illustrates variables of interest that were the focus of this analysis.
93 Table 17 Bivariate Significance with Outcomes 1a and 2 Factor Domain Outcome 1a Outcome 2 Predisposing Traditional White/Nonwhite White/Nonwhite Combat Vulnerable (none) Public support Length homelessness Need Traditional Interest in program Interest in program Vulnerable Schizophrenic Mood disorder Adjustment disorder Other psychiatric disorder Enabling Traditional Therapeutic community Faith-based (history) Cognitive behavioral 12Step Psychosocial rehab Vulnerable (none) (none) The first model of logistic regression (Tab le 18) represents va riables of interest identified in the bivariate analysis with Outcome 1a.
94 Table 18 Program Completion Status (Outcome 1a): Logistic Regression Model n = 2,502: Dependent Variable: Success: 1307 (52.2%); Nonsuccess: 1195 (47.8%) 95.0% CI for EXP (B) Independent Variable B df Sig. Exp (B) Lower Upper White .183 1 .030 1.201 1.018 1.416 Combat .205 1 .043 1.228 1.007 1.498 Interest in Program .351 1 .000 1.420 1.168 1.727 Schizophrenia -.389 1 .008 .678 .507 .905 Adjustment Disorder .126 1 .136 1.135 .961 1.340 Therapeutic Community -.135 1 .253 .874 .693 1.102 Cognitive Behavioral .355 1 .007 1.427 1.103 1.845 12-Step Program -.309 1 .071 .734 .525 1.027 Psychosocial Rehab .320 1 .003 1.376 1.118 1.695 As indicated in Table 18, several variab les were significantly associated with Outcome 1a (program completion status) wh en controlling for various predisposing, enabling, or need factors. This model indica ted that whites were 20% more likely to be successful in GPD-funded programs than nonw hites. Those who experienced combat while in the military were 23% more likel y to be successful, and those who were interested in a full range of services when interviewed for transitional housing were 42% more likely to be successful. In this m odel, only schizophrenia was a significant predictor when controll ing for other variables. Those di agnosed with schizophrenia were 32% less likely to be succe ssful. And, unlike the bivari ate analysis where several
95 program models were determined to either im prove or lessen the likelihood of success in the program, only the cognitive behavioral and psychosocial rehabilitation models were significantly associated with success when cont rolling for other variables in these logistic regression tests. Par ticipants enrolled in cognitive behavioral programs were 43% more likely to be successful, and those enrolled in psychosocial rehabilitation programs were 38% more likely to achieve success as determined by Outcome 1a. The second model of logistic regression (T able 19) represents those variables of interest identified in the biva riate analysis and th eir significance with housing status upon program discharge (Outcome 2). Table 19 Housed Upon Program Discharge (Outco me 2): Logistic Regression Model n = 2,502 Dependent Variable: Housed: 2112 (84.6%); Not Housed: 385 (15.4%) 95.0% CI for EXP (B) Independent Variable B df Sig. Exp (B) Lower Upper White/Nonwhite .281 1 .015 1.324 1.057 1.659 Public Support .263 1 .023 1.301 1.037 1.632 Length of Homelessness -.337 1 .006 .714 .560 .910 Interest in Program .364 1 .005 1.439 1.117 1.854 Mood Disorder .142 1 .220 1.153 .918 1.447 Other Psychiatric Disorder -.277 1 .060 .758 .568 1.014 Faith-Based -.215 1 .066 .807 .642 1.014
96 Similar to associations with Outcome 1a and bivariate tests, the variables of ethnicity, public support, le ngth of homelessness before admission to the program, and interest in the program were statistically sign ificant. Whites were 32% more likely to be housed upon discharge than nonwhites. Those th at received any public support prior to admission to a program were 30% more likely to be housed when discharged. Those that were homeless more than 30 days prior to admission were 29% less likely to be housed, and those that expressed interest in a full range of services were 44% more likely to be housed upon discharge from a GPD-funded program than those interested in basic or no services. As was determined in the bivari ate analysis, no diagnostic categories were statistically significant with housing status. Additionally, no statistical significance was revealed between program models and this housing outcome. Alt hough in the bivariate analysis it was determined th at those participants in faith-based programs were less likely to be housed, this logistic regression model revealed only marginal significance at p < .06 (19% less likely to be housed upon disc harge from a GPD-funded program). Generalized Estimating Equation (GEE) Models Generalized Estimating Equation (GEE) models were constructed to further analyze predictors of successful outcomes fr om GPD-funded programs. GEE procedures recognize the possibility of repeated or clustered data and provide a method for analyzing data collected in groups where obs ervations within a group may be correlated but observations in separa te groups are independent (Lumley, 2007, p. 475). GEE models were constructed for this studys variables of interest identified in the bivariate
97 analysis for both Outcomes 1a and 2. The sample tested was the enrolled GPD-funded program participants; the first admission was used for those participants that enrolled in a GPD-funded program more than once during the study period ( n = 2502). The subject variable for GEE analysis purposes was th e individual program identification code. Table 20 Program Completion Status (Outcome 1a) : Generalized Estimating Equation Model n = 2,502 Dependent Variable: Success: 1307 (52.2%); Non-Success: 1195 (47.8%) 95% Wald Confidence Interval Hypothesis Test Parameter B Std. Error Lower Upper Wald Chi-Square df Sig. White/Nonwhite .183 .0963 -.006 .372 3.613 1 .057 Combat .205 .0972 .015 .396 4.467 1 .035 Interest in Program .351 .1106 .134 .568 10.055 1 .002 Schizophrenia -.389 .1477 -.678 -.099 6.934 1 .008 Adjustment Disorder .126 .1027 -.075 .328 1.514 1 .218 Therapeutic Community -.135 .1497 -.428 .158 .814 1 .367 Cognitive Behavioral .355 .1364 .088 .623 6.792 1 .009 12-Step Program -.309 .1848 -.671 .054 2.789 1 .095 Psychosocial Rehab .320 .1906 -.054 .693 2.810 1 .094 Results (Table 20, Outcome 1a) indicated that most variables identified as statistically significant in the logistic re gression model remained significant when compensating for clustered participant data in the GEE analysis. Perhaps somewhat revealing was that the earlier identified stat istically significant relationship between white
98 participants and the likelihood of success in the program decreased (p = .057). In addition, GEE analysis indicated that if pa rticipants were enro lled in a psychosocial rehabilitation model program, they were no more likely to achieve success at discharge than those participants enrolled in other program models. Through logistic regression analysis, it was indicated that enrollment in this type of program was a significant predictor of success. Consis tent with bivariate and logi stic regression models, if a participant showed interest in the program at the time of the initial interview or reported experiencing combat while in the milita ry, the participant would more likely be successful. The diagnosis of schizophrenia remained significantly associated with nonsuccessful program completion in both the logistic regression and GEE model, and adjustment disorder was no longer statistically significant with Outcome 1a in the GEE analysis.
99 Table 21 Housed Upon Program Discharge (Outcome 2): Generalized Estimating Equation Model n = 2,502 Dependent Variable: Housed: 2112 (84.4%); Not Housed: 385 (15.4%) 95% Wald Confidence Interval Hypothesis Test Parameter B Std. Error Lower Upper Wald Chi-Square df Sig. White/Nonwhite .277 .1322 .018 .536 4.403 1 .036 Public Support .258 .1045 .053 .463 6.093 1 .014 Length of Homelessness -.336 .1290 -.589 -.084 6.805 1 .009 Interest in Program .361 .1348 .097 .626 7.180 1 .007 Mood Disorder .138 .1046 -.067 .343 1.736 1 .188 Other Psychiatric Dis. -.290 .1608 -.605 .025 3.248 1 .072 Faith-Based -.212 .1733 -.552 .127 1.502 1 .220 Days Worked -.005 .0080 -.021 .011 .426 1 .514 Results illustrated in Table 21 for Ou tcome 2 (housing status upon discharge) indicated that, when bivariate models are ut ilized, participant clustering may inaccurately determine predictor determination. However, the GEE models constructed with Outcome 2 showed similar statistically significant rela tionships with those identified in logistic regression models. Not sta tistically significant in the above GEE analysis was the relationship between housing status at discharge and enrollment in faith-based programs. Logistic regression indicated th at participants in faith-based programs may be less likely to be housed upon discharge; this rela tionship was not evident in Table 21.
100 Days in Residence and Outcomes This study was limited to those particip ant and program factors that may be predictors of success in tran sitional housing. Success was defined as completion status, as determined by VA staff (Outcome 1a), and housing status upon discharge from the program (Outcome 2) as reported by the particip ant. The influence or effect of length of stay in the transitional housing on program outcomes was not fully examined within the context of this study. However, a preliminary cross tab and chi-square analysis of length of stay versus outcomes was performed: the results may indicate that this dose-response approach to determining factors that infl uence success in transitional housing programs could be revealing and a consider ation for future research. The data set utilized for this study included the number of days in residen ce of each participant. Length of stay was subset into 0-90; 91-180; 181-270; and 271+ days. Without duplicates removed, the percentage of those successfully comple ting and housed upon discharge increased with each subset days-in-residence variable. The percentage of participants determined to have successful completion status (Outcome 1a) increased in each subset (0-90 days: 39%; 91-180 days: 62%; 181-270 days : 68 %; and 271 days or greater: 72 %). For those housed upon discharge (Outcome 2), the same trend was revealed (0-90 days: 78%; 91180 days: 90%; 181-270 days: 91%; and 271 days or greater: 93%). When length of stay was further subset into a dichotomous variab le (less than 180 days versus 181 days or greater), a positive correlation existed betw een both program success (Outcome 1a) and housed upon discharge (Outcome 2). For both groups, veterans were at least 62% more
101 likely to be successful and housed upon discharg e if their length of stay in a GPD-funded program was 181 days or more. Discussion Hom elessness is identified as a signifi cant public health issue attributed to societal changes as well as to the soci al and behavioral aspects of the homeless individual. Frequently, ho meless programs are developed to address a communitys immediate gaps rather than homeless individu als needs. One of the most common types of homeless program is transitional housing offering a stable place to live for up to two years while providing an overlay of social support and servic es. As discussed herein, there is documented evidence of the increased prevalence of mental illness in the homeless population. The purpose of this study was to examine the associations between participantand program-level factors with community-based transitional housing. This study used a revised framework of the behavior al model of utiliza tion and explored the assumptions of the utilization of the tr ansitional housing design for addressing homelessness. The following is a summary of the research objectives and hypotheses of this study. A general discussion will follow addressing the implications for theory and public mental health.
102 Research Objectives Objective One Provide a d escriptive analysis of participant characteristics in a national sample subset of the homeless population diagnosed wi th a mental illness, utilizing communitybased transitional housing programs. The Department of Veterans Affairs ope rates the countrys largest network of homeless transitional housing programs under the Homeless Providers Grant and Per Diem Program. This paper examined GP D Program data sets from 2004 and 2005. Frequencies were categorized under the predisposing-traditional and predisposingvulnerable domains of utilization theory. Study inclusion was based on admission into the program and a mental health diagnosis, ex cluding substance abuse. An examination of demographic frequencies provided th e characteristics of the study group. Objective Two Examine and assess the intensity and types of services of community-based transitional housing utilized by a national sa mple subset of the homeless population diagnosed with a mental illness. Transitional housing service intensity and type was ex amined under the enablingtraditional and the enabling-vulnerable domai ns of the utilization theory. Although almost half of the programs surveyed had no certification or accred itation, it was evident that there was considerable intensity and variet y of services offered to participants. More than two-thirds of the programs service quotient scores were between 30 and 50, representing a high level of services on-site, offered directly to the participants while
103 enrolled in housing. The Preliminary Data Analysis section of this study delineated the types of services offered, while the Level One Analysis provided the frequencies of services and program type, as categorized und er the domains of the utilization theory. Treatment philosophies in the programs diffe red; however, those programs that were therapeutic communities or s upporting housing or programs that followed principals of the cognitive behavioral, 12-step, and psychosoc ial rehabilitation models were relatively equally represented. Objective Three Examine the associations between program participant characteristics and mental health diagnosis, with community-b ased transitional housing outcomes. Associations between participant char acteristics and commun ity-based housing outcomes were examined through bivariate tests and structured within the studys participant-level hypotheses (Hypotheses On e through Five) discussed below. Hypothesis One. There is no significant asso ciation between participant demographics and successful completion of community-based transitional housing. Statistically significan t associations were determined through the use of bivariate tests between participant ethnicity and combat experience with completion of the programs. Hypothesis Two. There is a negative association between participant severity of homelessness and successful completion of community-based transitional housing. Bivariate tests demonstrated statistica lly significant associations between outcomes and participants receipt of public support and their le ngth of homelessness
104 before admission. Several characteristics of vul nerability to homelessness, as determined by these variables, were negatively asso ciated with success in the program. Hypothesis Three. There is a positive association between participant expressed interest in program utilization and successful completion of community-based transitional housing. As measured by bivariate tests, ther e was a positive association between participant expressed interest in th e program and successful completion. Hypothesis Four. There is a positive association between participant perceived mental illness and successful completion of community-based transitional housing. No association was found between the participants perceived mental illness status and program outcomes through bivariate analysis. Hypothesis Five. There is no significant associat ion between participant mental health diagnosis and successful completion of community-based transitional housing. Significant associations were determined through the use of bivariate tests between participants mental h ealth diagnosis and outcome. However, these associations differed depending on the outcome measure. Schizophrenia and adjustment disorder were associated with program completion status (Outcome 1a) while mood disorder and other psychiatric disorder were associated with housing status (Outcome 2). Objective Four Examine the effects of program-level services on the associations between program participant characteris tics and mental health diagnosis with community-based transitional housing outcomes.
105 Effects of program-level services on the association between program participant characteristics and mental health diagnoses were examined through t-tests and bivariate methods and structured within this st udys program-level hypotheses (Hypotheses Six through Nine) discussed below. Hypothesis Six. There is a positive associati on between program certification status and participants successful comple tion of community-based transitional housing. No statistically significan t association was found be tween program certification status and program outcomes. Hypothesis Seven. There is no significant associ ation between the type of treatment model philosophy of a community-based transi tional housing program and participants successful completion. Through the use of bivariate tests, statis tically significant associations with Outcome 1a (program completion) were de termined with therapeutic communities, cognitive behavioral, 12-step, and psychos ocial rehabilitation program models. Hypothesis Eight. There is no significant associa tion between the religious basis of a community-based transitional housi ng program and participants successful completion. A statistically significant a ssociation existed between Outcome 2 (housing status) and those programs that were determined to be faith-based or thos e programs operated by an organization with religious origins.
106 Hypothesis Nine. There is a positive association between the level of program services and participants successful co mpletion of community-based transitional housing. No statistically significan t association was found between the level of program services and program outcomes through the use of bivariate analysis. Associations reported above in Hypotheses One through Nine and the significance of the relationships between the predisposing, enabling, and need fact ors with transitional housing program outcomes were initially discus sed and tested in this study within the context of bivariate analysis or performing t-te sts. These tests, discussed as part of the methodology section of this paper in the Leve l Two Analysis and the first portion of the Level Three Analysis, promoted further examination through additional analytic methods. Logistic regression models were de veloped to further explore any associations that were determined and to more accurately delineate any differe nces between each of the dependent variables. As discussed in the Level Three Analysis, significant relationships existed between various particip ant and program characteristics as revealed by the regression models. According to logistic regression mode l interpretation, those in the study group that were more likely to be successful in the transitional housing programs and more likely to be housed upon discharge were white. Participants that re ported experiencing combat while in the military were more likely to be successful, but there was no evidence that they would more likely be housed at discharge. Those participants who were homeless less than 30 days before admission or received public support were more likely
107 to be housed upon program discharge. Those th at were not diagnosed with schizophrenia were more likely to be successful, and those th at showed an interest in the program were more likely to be successful and housed upon program discharge. Participants enrolled in cognitive behavioral or ps ychosocial rehabilitation m odel programs had a better likelihood of success. And, participants were more likely to be housed upon discharge if they were in programs that we re not religious or had no religious history. There was no indication through t-test or biva riate analysis that the serv ices offered directly by the program, on-site, were associated with success. GEE models were utilized in this stu dy to compensate for subject clustering. These models, based on variables of interest from the bivari ate and logistic regression tests, demonstrated typical risks of inaccura te study assumptions. Limited statistically significant associations with both outcomes were revealed through the use of this analysis. White participants in the program were more likel y to be successful and housed as were those participants who showed an interest in program services upon admission. Participants who reported comb at experience while in the military were more likely successful, but there was no relationship to combat and housing upon discharge. Schizophrenia was the only diagnostic categor y that was significan tly associated with program success. Additionally, GEE analysis demonstrated that onl y those participants in cognitive behavioral programs had a greate r chance of success th an those in other program types. If participants were homele ss less than thirty days before admission or received public support, they were more likely to be housed upon discharge. Through further cross tab analyses, it appeared that wh ites were less likely to be diagnosed with
108 schizophrenia and more likely to be enroll ed in cognitive behavioral model programs; nonwhites were more likely to be enrolled in programs with lower level services. Objective Five Develop and offer recommendations as to what types of community-based transitional housing programs may be best suited to meet the needs of the homeless population diagnosed with a mental illness. The tests performed through this study di d not demonstrate c onclusive evidence that would indicate what type s of community-based transi tional housing program may be best suited to meet the needs of the homeless population diagnosed with a mental illness. Evidence provided suggests there may be a likelihood that programs offering cognitive behavioral or psychosocial re habilitation models are bett er suited for the homeless population. However, this could not be confirmed when comparing outcomes as measured by both dependent variables and through GEE analysis. The two participant characteristics consistently associated w ith positive program outcomes in bivariate, logistic regression, and GEE test s were participant ethnicity a nd interest in the program. Those that were white and those that expre ssed an interest in the program were more likely to be successful as measured by both outcomes. And, although a focus of this study was to determine if program services were related to positive outcomes, considering that program serv ices were not associated with either outcome, this assumption could not be supported.
109 Implications for Public Mental Health This study exam ined the relationship be tween participant characteristics and completion of transitional housing programs. Additionally, this research explored the possibility of program-level f actors affecting participants su ccess. Assumptions of the behavioral utilization theory proposed by A ndersen (1995) suggest that factors, both individualand program-level, traditional and vulnerable, co uld be predictors of careoutcomes. Through this studys various sta tistical tests reveal ed limited factors influencing outcomes from transitional housi ng programs. Those significant factors, initially demonstrated through first level an alysis, were not confirmed by further tests instituting statistical controls. Participants who showed an interest in the program prior to admission were those that would most lik ely be successful and housed upon discharge. When implementing a number of statistical te sts to control for vari ables and compensate for clustered data, interest in the program was consistently associated with positive outcomes as measured by both outcome measures. This study also proposed to explore the models of transitional housing. As a preferred method to address a communitys immediate need, tr ansitional housing has been a foundation of most local areas homeless assistance plans. Success in the programs studied was measured by two outcomes. Frequencies of success for 2 years of data were 52% (clinician determination of completion status ) and 84% (subj ect report of housing status upon discharge). Although limite d associations were revealed between participantor programlevel characteristics, simple fre quency reports supported by this
110 studys large sample size would sustain the argument that transitional housing is a supportable and reasonable method to provide care for homeless populations. A series of articles regarding the need for national mental he alth services was featured in the American Journal of Public Health (2006). A number of studies cited earlier in this study reference the increased incidence and prevalen ce of mental health diagnoses in the homeless population, the chronicity of mental illness among the homeless, and the difficulty in performing outreach to or providing mental health care for those on the streets. The hea lth of a particular population can be seen as dependent and interdependent on various levels of conn ections with and among components of its environment. The implications for public mental health become an uncertain and infrequently studied mix of mental illness, homelessness, and access to or provision of treatment. Accessing services for this popul ation equates to seeking primary care, which in many cases actually becomes a pur suit of housing. Homeless service providers are becoming the mental health care institutions for an increasingly larger segment of this nations populationthe homeless diagnosed with a mental illness. It was expected that the results of th is study would assist the providers who struggle with helping the homeless, the resear chers who study this social phenomena, as well as the decision makers who influence th e allocation of resources. Program designs for transitional housing should be develope d specifically for the population and the individual, as those that show an interest in the program will more likely be successful. However, it appears services are not always equally distri buted, as programs studied did not seem to offer the same services to all populations. Providers n eed to ensure that
111 services offered to participants who ar e nonwhite equal those provided to white populations. Although not conclusive, there is an indication that pr ograms more aligned with cognitive behavioral or psychosocial rehabilitation approaches have better outcomes with the homeless population. Finally, at leas t as demonstrated by one of the outcome measures, participants in secular as opposed to faith-based programs were more likely to find housing when reintegrat ing into the community. Protection of Hum an Subjects The individual subject data used in th is study was obtained through a survey of existing records. These records and the comp iled data are maintained by the Department of Veterans Affairs Northeast Program Eval uation Center under authority of the U.S. Department of Veterans Affair s. Individual identifiers were not queried for this study. This study, submitted under the title Utilization of Community-Based Transitional Housing by Homeless Veteran Populations Diagnosed with a Mental Illness was determined as exempted by the Institutional Review Board (IRB), University of South Florida, Tampa Florida (See Appendix D).
112 REFERENCES Aday, L. A., & Andersen, R.M. (1974). A fr a mework for the study of access to medical care. Health Services Research, Fall 208-220. Andersen, R. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior, 36 1-10. Andersen, R., & Aday, L. A. (1978). Access to medical care in the U.S.: Realized and potential. Medical Care Research and Review, 16 (7), 533-546. Aneshensel, C. S. (2002). Theory-Based Data Analysis for the Social Sciences London: Pine Forge Press. Banyard, V. L. (1995). Taking another route: Daily survival narr atives from mothers who are homeless. American Journal of Community Psychology, 23 (6), 871-891. Barrow, S., & Rita, Z. (1998). Transitional Housing Services: A Synthesis. Paper presented at the 1998 National Sympos ium on Homeless Research, Washington, D.C. Bebout, R. R., Drake, R. E., Xie, H., McHugo, G. J., & Harris, M. (1997). Housing status among formerly homeless dually diagnosed adults. Psychiatric Services, 48 (7), 936-941. Benda, B. B. (2004). Life-course theory of readmission of s ubstance abusers among homeless veterans. Psychiatric Services, 55 (11), 1308-1310. Berne, A. S., Dato, C., Mason, D. J., & Rafferty, M. (1990). A nursing model for addressing the health needs of homeless families. Image: The Journal of Nursing Scholarship, 22 (1), 8-13. Biegel, D. E., Shafran, R. D., & Johnsen, J. A. (2004). Facilitators a nd barriers to support group participation for family caregivers of adults with mental illness. Journal of Community Mental Health, 40 (2), 151-166. Blankertz, L., Cnaan, R., White, K., Fox, J ., & Messinger, K. (1990). Outreach efforts with dually diagnosed homeless persons. Families In Society, 71 387-395.
113 Breakey, W. R., Fischer, P. J., Kramer, M., Ne stadt, G., Romanoski, A. J., Ross, A., et al. (1989). Health and mental health prob lems of homeless men and women in Baltimore. Journal of the American Medical Association, 262 (10), 1352-1357. Calsyn, R. J., & Morse, G. A. (1991). Corre lates of problem drinking among homeless men. Hospital and Community Psychiatry, 42 (7), 721-725. Caplan, G., & Caplan, R. (2000). Principles of community psychiatry. Journal of Community Mental Health, 36 (1), 7-24. Carr, S., Murray, R., Harrington, Z., & Oge, J. (1998). Discharged residents' satisfaction with transitional housing for the homeless. Journal of Psychosocial Nursing and Mental Health Services, 36 (7), 27-33. Caton, C. L., Dominguez, B., Schanzer, B., Hasi n, D., Shrout, P., Felix, A., et al. (2005). Risk factors for long-term homelessness: Findings from a longitudinal study of first-time homeless single adults. American Journal of Public Health, 95 (10), 1753-1759. Caton, C. L., Shrout, P. E., Eagle, P. F., Opler, L. A., & Felix, A. (1994). Correlates of codisorders in homeless and never ho meless indigent schizophrenic men. Psychological Medicine, 24 (3), 681-688. Conrad, K. J., Hultman, C. I., Pope, A. R., Lyon s, J. S., Baxter, W. C., Daghestani, A. N., et al. (1998). Case managed residentia l care for homeless addicted veterans: Results of a true experiment. Medical Care Research and Review, 36 (1), 40-53. Dennis, D. L., Steadman, H. J., & Cocozza, J. J. (2000). The impact of federal systems integration initiatives on services for mentally ill homeless persons. Mental Health Services Research, 2 (3), 165-174. Desai, M. M., Rosenheck, R. A., & Kasprow, W. J. (2003). Determinants of receipt of ambulatory medical care in a national samp le of mentally ill homeless veterans. Medical Care Research and Review, 41 (2), 275-287. Dickey, B. (2000). Review of programs for pe rsons who are homeless and mentally ill. Harvard Review of Psychiatry, 8 (5), 242-250. Dolbeare, C. (1996). Housing policy: A ge neral consideration. In Baumohl (Ed.), Homeless in America (p. 291). Westport, CT: Oryx. Drake, R. E., Yovetich, N. A., Bebout, R. R., Harris, M., & McHugo, G. J. (1997). Integrated treatment for dually diagnosed homeless adults. Journal of Nervous and Mental Disease, 185 (5), 298-305.
114 Evashwick, C., Rowe, G., Diehr, P., & Branc h, L. (1984). Factors explaining the use of health care services by the elderly. Health Services Research, 19 (3), 357-382. Fischer, P. J., & Breakey, W. R. (1991). The epidemiology of alcohol, drug, and mental disorders among homeless persons. American Psychologist, 46 (11), 1115-1128. Fischer, P. J., Shapiro, S., Breakey, W. R ., Anthony, J. C., & Kramer, M. (1986). Mental health and social characteristics of th e homeless: A survey of mission users. American Journal of Public Health, 76 (5), 519-524. Flynn, L. (1997). The health practices of homeless women: A causal model. Applied Nursing Research, 46 (2), 72-77. Foscarinis, M. (1991). The politic s of homeless: A call to action. American Psychologist, 6, 1231-1239. Gamache, G., Rosenheck, R., & Tessler, R. (2001). The proportion of veterans among homeless men: A decade later. Social Psychiatry and Psych iatric Epidemiology, 36(10), 481-485. Gelberg, L., Andersen, R., & Leake, B. ( 2000). The behavioral model for vulnerable populations: Application to medical care use and outcomes. Health Services Research, 34 1273-1302. Gelberg, L., & Arangua, L. (2001). Homeless persons. In R. Andersen, T. Rice & G. Kominski (Eds.), Changing the U.S. Health Care System, Gelberg, L., Gallagher, T. C., Andersen, R. M., & Koegel, P. (1997). Competing priorities as a barrier to medical car e among homeless adults in Los Angeles. American Journal of Public Health, 87 (2), 217-220. Goodman, L., Saxe, L., & Harvey, M. (1991) Homelessness as psychological trauma: Broadening perspectives. American Psychologist, 46 (11), 1219-1225. Grella, C. E. (1993). A residential recovery program for homeless al coholics: Differences in program recruitment and retention. Journal of Mental He alth Administration, 20(2), 90-99. Gwadz, M. V., Clatts, M. C., Leonard, N. R ., & Goldsamt, L. (2004). Attachment style, childhood adversity, and behavioral ri sk among young men who have sex with men. Journal of Adolescent Health, 34 (5), 402-413.
115 Haber, M. G., & Toro, P. A. (2004). Homelessness among families, children, and adolescents: An ecologicaldevelopmental perspective. Clinical Child and Family Psychology Review, 7 (3), 123-164. Haugland, G., Siegel, C., Hopper, K., & Alexander, M. J. (1997). Mental illness among homeless individuals in a suburban county. Psychiatric Services, 48 (4), 504-509. Health Care for Homeless Veterans Programs: The 17th Annual Report. (2004). West Haven, Connecticut: VA Northeas t Program Evaluation Center. Homelessness: Programs and the people they serve. National Survey of Homeless Assistance Providers and Clients. (19 99). Washington, D.C.: The Urban Institute. Homelessness: Transitional Housing Shows Initial Success but Long-Term Effects Unknown. (No. GAO/RCED-91-200)(1991). Washington, D.C.: General Accounting Office. Hopper, K., Jost, J., Hay, T., Welber, S., & Haugland, G. (1997). Homelessness, severe mental illness, and the institutional circuit. Psychiatric Services 48 (5), 659-665. Interagency Council on the Homeless Annual Report (1994). Washington DC: Interagency Council on the Homeless. Isaac, R. (1990). Madness in the Streets New York: The Free Press. Johnson, T. P., & Barrett, M. E. (1995). Substance use and treatment needs among homeless persons in Cook County, Illinois. International Journal of Addictions, 30(5), 557-585. Johnson, T. P., Freels, S. A., Parsons, J. A ., & Vangeest, J. B. (1997). Substance abuse and homelessness: Social sel ection or social adaptation? Addiction, 92 (4), 437445. Kasprow, W. J., & Rosenheck, R. (2000) Mortality among homeless and nonhomeless mentally ill veterans. Journal of Nervous and Mental Disease, 188 (3), 141-147. Kasprow, W. J., Rosenheck, R., Frisman, L., & DiLella, D. (1999). Residential treatment for dually diagnosed homeless veterans: A comparison of program types. American Journal on Addictions, 8(1), 34-43. Kasprow, W. J., Rosenheck, R. A., Fris man, L., & DiLella, D. (2000). Referral and housing processes in a long-term s upported housing program for homeless veterans. Psychiatric Services, 51 (8), 1017-1023.
116 Koegel, P., Burnam, M. A., & Farr, R. K. ( 1988). The prevalence of specific psychiatric disorders among homeless individuals in the inner city of Los Angeles. Archives of General Psychiatry, 45 (12), 1085-1092. Koegel, P., Melamid, E., & Burnam, A. (1995). Childhood risk factors for homelessness among homeless adults. American Journal of Public Health, 85 (12), 1642-1649. Koegel, P., Sullivan, G., Burnam, A., Morton, S. C., & Wenzel, S. (1999). Utilization of mental health and substance abuse se rvices among homeless adults in Los Angeles. Medical Care Research and Review, 37 (3), 306-317. Kreider, B., & Nicholson, S. (1997). Health insurance and the homeless. Health Economics, 6(1), 31-41. Kuno, E., Rothbard, A. B., Averyt, J., & Culhane, D. (2000). Homelessness among persons with serious mental illness in an enhanced community-based mental health system. Psychiatric Services, 51 (8), 1012-1016. Lam, J. A., & Rosenheck, R. (1999). Soci al support and service use among homeless persons with serious mental illness. International Journal of Social Psychiatry, 45(1), 13-28. Leda, C., & Rosenheck, R. (1992). Mental hea lth status and community adjustment after treatment in a residential treatment program for homeless veterans. American Journal of Psychiatry, 149(9), 1219-1224. Leda, C., & Rosenheck, R. (1995). Race in the treatment of homeless mentally ill veterans. Journal of Nervous and Mental Disease, 183 (8), 529-537. Lim, Y. W., Andersen, R., Leake, B., Cunningham, W., & Gelberg, L. (2002). How accessible is medical care for homeless women? Medical Care Research and Review, 40 (6), 510-520. Lumley, T. Department of Biosta tistics, University of Washington, XLISP-Stat tools for building Generalized Es timating Equation models www.jstatsoft.org, 2007. Lyon-Callo, V. (2000). Medicalizing homelessn ess: The production of self-blame and self-governing within homeless shelters. Medical Anthropology Quarterly, 14(3), 328-345. Mares, A. S., Kasprow, W. J., & Rosenheck, R. A. (2004). Outcomes of supported housing for homeless veterans with psyc hiatric and substance abuse problems. Mental Health Services Research, 6 (4), 199-211.
117 Martens, W. H. (2001). A review of physical and mental health in homeless persons. Public Health Review, 29 (1), 13-33. Murray, R., Baier, M., North, C., Lato, M., & Eskew, C. (1995). Components of an effective transitional residential progra m for homeless mentally ill clients. Archives of Psychiatric Nursing, 9 (3), 152-157. Murray, R., Baier, M., North, C., Lato, M., & Eskew, C. (1997). One-year status of homeless mentally ill clients who comple ted a transitional residential program. Journal of Community Mental Health, 33 (1), 43-50. Murray, R. B., & Baier, M. (1993). Use of th erapeutic milieu in a community setting. Journal of Psychosocial Nursing and Mental Health Services, 31 (10), 11-16. Murray, R. L., & Baier, M. (1995). Evaluation of a transitional residential programme for homeless chronically mentally ill people. Journal of Psychiatric Mental Health Nursing, 2(1), 3-8. National Alliance for the Homeless Press Release. (2000). New York Times North, C. S., Eyrich, K. M., Pollio, D. E., & Spitznagel, E. L. (2004). Are rates of psychiatric disorders in the homeless population changing? American Journal of Public Health, 94 (1), 103-108. North, C. S., Thompson, S. J., Pollio, D. E ., Ricci, D. A., & Smith, E. M. (1997). A diagnostic comparison of homeless and nonho meless patients in an urban mental health clinic. Social Psychiatry and Psychiatric Epidemiology, 32 (4), 236-240. Olfson, M., Mechanic, D., Hansell, S., Boyer, C. A., & Walkup, J. (1999). Prediction of homelessness within three months of discharge among inpatients with schizophrenia. Psychiatric Services, 50 (5), 667-673. Padgett, D. K., Patrick, C., Burns, B. J., & Schl esinger, H. J. (1994). Ethnic differences in use of inpatient mental health services by blacks, whites, and Hispanics in a national insured population. Health Services Research, 29 (2), 135-153. Phelan, J. C., & Link, B. G. (1999). Who are "the homeless"? Recons idering the stability and composition of the homeless population. American Journal of Public Health, 89(9), 1334-1338. Phillips, K. A., Morrison, K. R., Andersen, R., & Aday, L. A. (1998). Understanding the context of healthcare utilization: Asse ssing environmental and provider-related variables in the behavior al model of utilization. Health Services Research, 33 (3 Pt 1), 571-596.
118 Plescia, M., Watts, G.R., Neibacher, S., & Strelnick, H. (1997). A multidisciplinary health care outreach team to the ho meless: A 10-year experience of the Montefiore Care for the Homeless Team. Family and Community Health, 20(2), 58-60. Pollio, D. E., North, C. S., Thompson, S., Paquin, J. W., & Spitznagel, E. L. (1997). Predictors of achieving stable housing in a mentally ill homeless population. Psychiatric Services, 48 (4), 528-530. Prabucki, K., Wootton, E., McCormick, R ., & Washam, T. (1995). Evaluating the effectiveness of a residential rehabi litation program for homeless veterans. Psychiatric Services, 46 (4), 372-375. Pruchno, R. A., & McMullen, W. F. (2004). Pattern s of service utilizat ion by adults with a developmental disability: Type of service makes a difference. American Journal of Mental Retardation, 109 (5), 362-378 Rosenheck, R. (2000). Cost-effectiveness of services for mentally ill homeless people: The application of research to policy and practice. American Journal of Psychiatry, 157(10), 1563-1570. Rosenheck, R., & Fontana, A. (1994). A mode l of homelessness among male veterans of the Vietnam War generation. American Journal of Psychiatry, 151 (3), 421-427. Rosenheck, R., Gallup, P., & Frisman, L. K. (1993). Health care utilization and costs after entry into an outreach program for homeless mentally ill veterans. Hospital and Community Psychiatry, 44 (12), 1166-1171. Rosenheck, R., Kasprow, W., Frisman, L., & Liu-Mares, W. (2003). Cost-effectiveness of supported housing for homeless persons with mental illness. Archives of General Psychiatry, 60 (9), 940-951. Rosenheck, R., & Koegel, P. (1993). Characteris tics of veterans and nonveterans in three samples of homeless men. Hospital and Community Psychiatry, 44 (9), 858-863. Rosenheck, R., Leda, C., Gallup, P., Astrachan, B., Milstein, R., Leaf, P., et al. (1989). Initial assessment data from a 43-site program for homeless chronic mentally ill veterans. Hospital and Community Psychiatry, 40 (9), 937-942. Rosenheck, R., & Seibyl, C. (1998). Costs of homelessness: Health service use and cost. Medical Care 36(8), 1256-1264 Rossi, P. H., Wright, J. D., Fisher, G. A ., & Willis, G. (1987). The urban homeless: Estimating composition and size. Science, 235 (4794), 1336-1341.
119 Saag, K. G., Doebbeling, B. N., Rohrer, J. E., Kolluri, S., Peterson, R., Hermann, M. E., et al. (1998). Variation in tertiary prevention and hea lth service utilization among the elderly: The role of urban-rural residence and supplemental insurance. Medical Care Research and Review, 36 (7), 965-976. Schutt, R., & Gerret, G. (1992). Responding to the Homeless, Policy and Practice New York: Plenum Press. Sosin, M. R., & Bruni, M. ( 1997). Homelessness and vulnerabi lity among adults with and without alcohol problems. Substance Use and Misuse, 32 (7-8), 939-968. Stewart B. McKinney Home less Assistance Act (1987). Strain, L. A. (1991). Use of health services in later life: The influence of health beliefs. Journal of Gerontology, 46 (3), S143-150. Sullivan, G., Burnam, A., & Koegel, P. (2000). Pathways to homelessness among the mentally ill. Social Psychiatry and Psychiatric Epidemiology Epidemiol, 35 (10), 444-450. Sullivan, G., Burnam, A., Koegel, P., & Hollenbe rg, J. (2000). Quality of life of homeless persons with mental illness: Results from the course-of-homelessness study. Psychiatric Services, 51 (9), 1135-1141. Susser, E., Struening, E., & Conover, S. (1987). Childhood experiences of homeless men. American Journal of Psychiatry, 144 (22), 1599-1601. Tsemberis, S., & Eisenberg, R. F. (2000). Pathways to housing: Supported housing for street-dwelling homeless individuals with psychiatric disabilities. Psychiatric Services, 51 (4), 487-493. Tsemberis, S., Gulcur, L., & Nakae, M. ( 2004). Housing First, consumer choice, and harm reduction for homeless indivi duals with a du al diagnosis. American Journal of Public Health, 94 (4), 651-656. Tsemberis, S., Moran, L., Shinn, M., Asmussen, S. M., & Shern, D. L. (2003). Consumer preference programs for individuals who are homeless and have psychiatric disabilities: A drop-in center and a supported housing program. American Journal of Community Psychology, 32 (3-4), 305-317. U.S. Department of Veterans Affairs. Fact Sheet Homeless Veterans, September, 2005. Vangeest, J. B., & Johnson, T. P. (2002). Subs tance abuse and homelessness: Direct or indirect effects? Annals of Epidemiology, 12 (7), 455-461.
120 Weinreb, L., Goldberg, R., & Perloff, J. (1998). Health characteristics and medical service use patterns of sh eltered homeless and low-income housed mothers. Journal of General Internal Medicine, 13 389-397. Wenzel, S. L. (1993). Indicators of chronic homelessness among veterans. Hospital and Community Psychiatry, 44(12), 1172-1176. Wenzel, S. L., Ebener, P. A., Koegel, P., & Gelberg, L. (1996). Drug-abusing homeless clients in California's substance abuse treatment system. Journal of Psychoactive Drugs, 28 (2), 147-159.
122 Appendix A: NEPEC Intake Assessm ent (Form X)
127 Appendix B: NEPEC Discharge Report (Form D)
131 Appendix C: NEPEC Facility Survey Form
138 Appendix D: IRB Exemption Certification
ABOUT THE AUTHOR Roger Casey earned a B achelors degree from Colorado State University and a Masters in Social Work from the Univers ity of South Florida. Upon graduation, Dr. Casey was employed in the Homeless Chronically Mentally Ill (HCMI) Veterans Program at the James A. Haley Veterans Administration Medi cal Center, Tampa, Florida. Designed to provide outreach, case management, and residential treatment, the HCMI Program received permanent authorization in 1992. In 1994, Dr. Casey was invited to VA Central Office, Mental Health Strategic Health Care Group in Washington, D.C., to implement and manage the VA Home less Providers Grant and Per Diem (GPD) Program. Working with nonprof it organizations, federal and state government agencies, and veterans service groups, Dr. Casey de veloped the GPD program into the nations largest integrated homeless residential services program: GPD serves over 15,000 homeless veterans annually at more than 300 community-based residential sites in every state, Washington, D.C., Puerto Rico, and Guam.
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001935521
007 cr mnu|||uuuuu
008 080424s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002237
Utilization of community-based transitional housing by homeless veteran populations diagnosed with a mental illness :
b the association between predisposing, enabling, and need factors with program outcomes
h [electronic resource] /
by Roger Casey
[Tampa, Fla.] :
University of South Florida,
ABSTRACT: Mental illness among homeless populations is a significant public health issue. Community-based programs that assist the homeless are most often developed to meet local housing needs, not the needs of mental health populations. Transitional housing, a model frequently utilized to address homelessness in communities, provides program-based housing with supportive services. The purpose of this study was to examine the associations between participant- and program-level factors on the utilization of community-based transitional housing by homeless veterans diagnosed with a mental illness. The study tested a revised framework of the behavioral model of utilization for vulnerable populations theory. The sample was comprised of male homeless veterans diagnosed with a mental illness who participated in community-based transitional housing programs in 2004 and 2005 (n = 2,502).Data were collected on 288 programs throughout the United States, operated by local nonprofit or local government agencies and monitored by the U.S. Department of Veterans Affairs under the Homeless Providers Grant and Per Diem Programs. Success was defined as either completion of a course of treatment as determined by a master's prepared clinician, or if housing was obtained upon discharge, as reported by the participant. Initial bivariate results indicated that both demographic and situational variables predicted success in transitional housing. However, upon further statistical analyses, limited predictors were revealed. Participants were more likely to be successful if they were white, reported combat experience, were interested in the program prior to admission, and were enrolled in cognitive behavioral models.Participants were more likely to be housed upon discharge if they were white, received some type of public support, were homeless less than 30 days before admission, and showed interest in the program at the time of the initial interview. Participants were less likely to be successful if they were diagnosed as schizophrenic. There was an indication that participants enrolled in programs designated as faith-based were less likely to be housed than those enrolled in secular programs. No statistically significant associations were found between the level of services offered in the transitional housing programs with either successful completion or participants' housing upon discharge.
Dissertation (Ph.D.)--University of South Florida, 2007.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 139 pages.
Advisors: Martha Coulter, Dr.P.H. and Bruce Lubotsky Levin, Dr.P.H.
x Public Health
t USF Electronic Theses and Dissertations.