In its investigations, the committee found no substantial published evidence that permanent supportive housing (PSH) improves the health of people experiencing chronic homelessness. As described previously, most studies did not explicitly include people with serious health problems who are the most likely to benefit from housing. Of the studies that were more rigorous, the committee found that housing increases the well-being of persons experiencing homelessness. In its review of studies examining cost-effectiveness of PSH, the committee found that the literature is limited with few randomized controlled studies available, a majority using a quasi-experimental design. Further, the available studies have not been conducted in a manner that is methodologically aligned with generally accepted health care cost-effectiveness research design. In principle, the most robust scientific evidence to answer the question would come from studies using a randomized design and that cover a comprehensive array of costs and effectiveness measures. Overall, the committee found few studies evaluating the cost-effectiveness of PSH programs and the studies that have been done provide incomplete data that do not fully capture the health benefits of PSH.
This chapter addresses the major research gaps associated with developing programs to address the housing and health needs of individuals experiencing homelessness. Among the gaps identified by the committee:
- Inconsistencies in definitions and characteristics of PSH in the existing research literature and limited description of key services or minimum standards of PSH;
- Limited evidence base for screening tools used in allocating housing services assistance;
- Barriers to collection of data on health outcomes of PSH;
- Need for additional randomized controlled trials (RCTs) or strong quasi-experimental data, which may bolster and refine evidence of the impact of PSH and other forms of permanent housing on health outcomes and health care costs;
- Limited university-agency partnerships that represent lost opportunities to evaluate and monitor health outcomes and costs;
- Insufficient application of “big data” science to integrated health data systems, homeless management information systems, and other data resources;
- Need for testing innovations in payment models to support housing and services; and
- Research focused on societal barriers to promote acceptance of persons who have experienced homelessness as neighbors in communities.
HARMONIZATION OF PSH DEFINITIONS AND PROGRAM CHARACTERISTICS
Research has assessed fidelity to the Pathways Housing First approach within PSH (Gilmer et al., 2014) and shown better fidelity to be associated with housing stability, substance abuse reduction, community integration, and quality of life (Davidson et al., 2014; Goering et al., 2016). Assessments of fidelity have not routinely accompanied evaluations of impact on patient health, however, raising a possibility that PSH models that have been compared to treatment-as-usual control conditions were “diluted” versions of those ideal models. To the extent that PSH has not been implemented in accordance with some minimum standard, service differences between PSH and treatment-as-usual clients might be diminished, thus providing at least partial explanation for no-difference findings in health outcomes. What PSH and control conditions consist of in each study must be understood to evaluate study findings.
Case management is fundamental in linking PSH clients to supportive services and in providing a point of contact and support for clients (Hannigan and Wagner, 2003). The array of other services that a client may access should be flexible, voluntary, and tailored in accordance with individual needs (SAMHSA, 2010; Tsemberis, 2015), and thus no minimum or required set of services has been specified, except case management. Since PSH typically targets individuals with physical and behavioral health problems experiencing chronic homelessness, however, identification of a minimum set of services to be made available on a voluntary basis and specification of key ingredients would seem reasonable, including types of services provided and effective versus ineffective client-to-staffing ratios needed to foster housing retention and housing outcomes. Such standardization would facilitate efforts to understand effectiveness in influencing health outcomes and could aid in further development of quality and performance guidelines for PSH. Frequency and intensity of services may also vary within service type, depending on provider training and other resource considerations, along with resident preferences, thus further complicating interpretation of impact on health outcomes.
Furthermore, studies thus far have not “isolated” the effect of housing on health outcomes as a distinct ingredient separate from the services provided through enrollment in PSH. As noted in Chapter 5, any direct health benefits of
housing are difficult to isolate because clients in treatment-as-usual study conditions may have received services comparable to those in the treatment group. Practically, this is a not a research gap that can be easily addressed, but it deserves attention in a discussion of essential ingredients of PSH. Because findings on direct physical and behavioral health outcomes of PSH have been less than robust, future studies must carefully document characteristics of all services accessed by persons in PSH and other study conditions. See Chapter 5 for recommendations on the need to conduct research on PSH to specify and delineate certain characteristics of supportive services in PSH.
UNDERSTANDING WHO BENEFITS MOST FROM PSH
The need for housing that is available to individuals experiencing homelessness continues to outpace supply (Steffen et al., 2015). Additionally, there are differences among those experiencing homelessness in terms of length of time spent homeless and number and severity of their morbidities. Because of a limited housing supply and the costs associated with hospitalization and other public system contacts, federal, state, and local entities have typically prioritized the allocation of PSH resources to “high-need, high-cost” individuals (Levitt, 2015). This focus has necessitated a systematic means of sorting individuals along dimensions of need in order to make decisions about who benefits from housing and accompanying services. The U.S. Department of Housing and Urban Development (HUD) has required that governmental and nongovernmental agencies receiving federal funding to serve individuals experiencing homelessness use standardized assessments of need to aid in allocation decisions (HUD, 2012a; Levitt, 2015).
Because assessment tools used in determining housing eligibility emerged from urgency in response to HUD policy, rather than from a series of carefully conducted studies over time, and because the tools are relatively new and not yet subjected to careful research to examine reliability and validity, the base of scientific evidence for existing assessment tools is scant (Levitt, 2015). The Service Prioritization Decision Assistance Tool (SPDAT) and the Vulnerability Index Service Prioritization Decision Assistance Tool (VI-SPDAT)1 are two such tools employed nationally in determining the allocation of housing and services to homeless single adults (OrgCode Consulting, 2015). The 10th Decile Tool was developed specifically to prioritize people experiencing homelessness with high levels of health risk in Los Angeles County (CSH, 2015). See Chapter 6 for a discussion of assessment tools.
There is consensus among experts regarding the gap in research to understand reliability and validity of assessment tools but no consensus on the outcomes that the tools should predict (Levitt, 2015). Screening to determine eligibility for PSH can pose challenges for providers, as reported during the committee’s site visits. Providers may struggle in turning away people who are screened but who fall short of a standardized (yet what seems arbitrary to providers) cut point for
1 Available through OrgCode Consulting Inc. and Community Solutions.
receiving service despite displaying some level of need. As articulated by a provider during a committee site visit, how should one extend some level of hope for housing to a person in need who receives a score that falls just short of service eligibility? Providers also raised concerns about the availability and adequacy of standardized training on the SPDAT and VI-SPDAT assessment tools to ensure fidelity of implementation within and across evaluators.
Another point regarding validity of the assessment tools used in determining access to housing and services is that they were developed for use with adults experiencing chronic homelessness, the population that has received the most focus in efforts thus far to address homelessness through PSH. Such tools cannot be assumed to be valid for other populations for whom PSH might also be a viable solution to homelessness and a vehicle for improved health outcomes. The tools also cannot be assumed to have validity for determining access to other forms of permanent housing.
IMPROVING THE QUALITY OF EVIDENCE EXAMINING PSH AND HEALTH
RCTs are the gold standard in demonstrating intervention effects. When an RCT is well executed, it effectively controls for selection bias and other threats to internal validity. Some quasi-experimental comparison group designs (e.g., utilizing propensity score matching or regression discontinuity designs) are good choices when randomization is not feasible. To the extent that evidence rests on weaker quasi-experimental designs or other observational data, confidence in the causal effect of PSH on health outcomes is reduced.
The clinical trial process is lengthy and resource intensive. However, because trials to understand the effect of PSH have thus far shown inconsistencies and limited direct impact across most clinical health and behavioral health outcomes, and few studies have rigorously assessed cost-effectiveness, additional research using carefully executed designs could serve to bolster and refine understanding of health and cost outcomes. (See Chapter 4 for a discussion of research needed on cost-effectiveness of PSH.)
Several considerations should be made in planning and conducting future RCTs to ensure their thoroughness and rigor, and thus interpretability and value to decision making. Some researchers have argued the importance of process evaluations in understanding the implementation of interventions in the context of large social experiments (Epstein and Klerman, 2016) and have noted the importance of using more complex experimental designs to better understand how and whether specific components of complex interventions work (Bell and Peck, 2016). Generally, additional, high-quality research and RCTs would aid in achieving a comprehensive understanding of how, why, when, and specifically for whom PSH works and for what health outcomes.
As already noted, future trials should strive to implement a “standardized” or “minimum required” model of PSH as much as that is feasible, assess fidelity
to that model, and carefully assess the characteristics of services and housing received in all study conditions to facilitate interpretation of findings. Qualitative data should also be routinely collected and analyzed as part of future trials to enhance understanding of housing and services and the experiences of providers and clients in delivering and accessing services. Such data will assist in defining services and in elucidating “key ingredients” of PSH.
Additionally, as the review of experimental literature on health outcomes made clear in Chapter 3, reporting on health outcomes has been limited and focused primarily on utilization of services and selected chronic conditions. Future trials should perhaps focus on a broader range of mutable health outcomes and assess whether there are health conditions whose course and medical management are more significantly influenced than others by having safe and stable housing (i.e., “housing-sensitive” conditions). For example, as noted in Chapter 3, residents with HIV/AIDS in PSH have shown greater rates of survival with intact immunity (Buchanan et al., 2009), and resident well-being has also shown improvement over time (e.g., Aubry et al., 2015). Self-reported assessments of health, including perceived pain, are accepted measures in assessing quality in patient-centered care and may play a role in strengthening the evidence for the impact of PSH on health outcomes (Gordon et al., 2005; IOM, 2001). Self-reported assessments could include those that can be directly translated into the quality-adjusted life-years (QALYs), if there is a desire for better understanding of the cost-effectiveness analysis (CEA) principles. Chapter 3 includes recommendations related to the need for additional research on conditions that could be considered housing sensitive.
Additional consideration of populations and individual differences is warranted in efforts to understand the effect of PSH on health and costs of health care. Very little is known about the health impacts of PSH for populations such as youth and families, or about other permanent housing models, including for the majority of persons who experience homelessness but who are not chronically homeless. Health care should aim to reduce disparities in quality and access based on race, ethnicity, age, gender, and other characteristics, and should promote health equity (IOM, 2001; HHS, 2017). While the importance of considering individual differences in health and social needs is acknowledged in PSH models, differences based on characteristics such as race, ethnicity, culture, gender, and sexual orientation have not received as much attention (Waegemakers Schiff and Rook, 2012). In a review of U.S. Housing First studies, Waegemakers Schiff and Rook (2012) noted that none of the investigations had addressed issues of diversity and ethnicity, and that all of the investigations were conducted in major metropolitan areas and were focused on single adults, most of whom had a mental illness. The Canadian At Home/Chez Soi study included a small city, and modifications tailored to indigenous communities in one site. There is a need for integration of social determinants of health into disease-focused research.
Attention to culturally sensitive delivery of services might affect a resident’s experience with housing and services and ultimately her or his health out-
comes (Netto, 2006; Hoeft et al., 2016; Diaz et al., 2017). Attention to characteristics that have been linked to health disparities is therefore important. Analysis of these characteristics as “effect modifiers” or moderating variables may facilitate refinement and tailoring of services, and aid in the interpretation of health outcomes reported within PSH studies. By systematically investigating the question, “What works best for whom?,” quality of care and improved health outcomes might be better facilitated for all residents. Studies to address this question will require large sample sizes.
An additional consideration for future research and experimental studies is that variations in housing types and models, as well as service mix within studies, complicate interpretation of evidence. An inherent complexity in interpreting the impact of PSH on health outcomes lies in the variation across studies in housing characteristics and services associated with housing. As an example, some studies have evaluated Pathways Housing First programs while others have focused on other PSH approaches. Although there are similarities between the two, Pathways Housing First,2 as strictly defined, refers to a specific model that embraces the elements of consumer choice and harm reduction, and utilizes the Assertive Community Treatment (ACT) model of psychiatric case management and service delivery (Tsemberis, 2015). A small amount of research has suggested that fidelity is not always maintained in implementation of the HF model (Gilmer et al., 2014). Other PSH approaches may further vary in level of adherence to elements of the specific Housing First model (Rog et al., 2014). The ACT model may not be needed by all PSH clients and more-flexible, less-intensive models should be developed and evaluated. For example, the Canadian At Home/Chez Soi study offered intensive case management rather than ACT teams to participants judged to have moderate needs.
An additional difference that may arise among PSH (and Housing First programs) is whether the housing provided to homeless clients is scattered site or single site. Future studies should account for housing types, models, and service mix to improve understanding of housing impact on health outcomes including for patients with chronic disease, and to ensure that “key ingredients” of PSH can be identified and included in scale-up efforts. When possible, studies should compare the effectiveness of different PSH models.
Built environment and neighborhood characteristics may also explain health outcomes. Individuals in PSH are nested within buildings and neighborhoods, where each of these levels may influence factors pertinent to health. Guidelines embraced by the Corporation for Supportive Housing state that physical aspects of housing should ensure a safe and attractive environment that promotes health of tenants (CSH, 2014b). Whether scattered site or single site, the quality and attributes of PSH and the characteristics of neighborhoods in which that housing is located may affect health outcomes and should be examined to understand the influence of PSH on health outcomes. Previous studies have found associations
2 The language can be confusing. Although Pathways Housing First is a specific model, as defined above, it is sometimes referred to in the literature as Housing First.
between the characteristics of housing and risk of disease, injury, and mental health and well-being (Krieger and Higgins, 2002; Hwang et al., 2003; Osman et al., 2008; Ram et al., 2016). Characteristics of the neighborhood could either promote or hinder behaviors such as walking and healthy eating, which in turn affect health, yet studies have given scarce consideration to such factors in efforts to understand the association between PSH and health outcomes (Henwood et al., 2013). Research has not yet determined whether or to what extent characteristics of the built environment or neighborhood may moderate an effect of PSH on health outcomes. See Chapter 3 for recommendations related to the need for additional research on PSH and health.
APPLYING NEW TOOLS TO INTEGRATE AND LEARN FROM DATA SYSTEMS
As described in Chapter 7, siloed databases and data collection systems stand as barriers to decision making and policies that affect persons experiencing homelessness and the continuity of services they receive.
For many years, HUD’s Homeless Management Information System (HMIS) has been the primary tool for collecting data at both the aggregate and the individual and family level. Grantees receiving HUD funds are required to input data into an HMIS. HUD created standards for local HMIS databases with its mission and priorities in the forefront. Thus, HMIS captures data into an HMIS database on housing-related outcomes, but only very minimal data on health outcomes or cost savings to other agencies. In turn, the data systems used in the health care field have not been designed to capture data on patients’ housing status. However, HMIS data systems are local choices from many different vendors; these systems are not integrated from county to county, much less integrated at the national level. The metrics and databases used by these different systems are generally not compatible.
The U.S. Interagency Council on Homelessness (USICH) is leading an effort to improve and integrate data collection across federal agencies (USICH, 2015c). USICH reports that this includes efforts to increase the role of mainstream federal programs to assess and track housing status and homelessness, and to provide information to Medicaid agencies, health care providers, and hospitals on assessing homelessness and housing status, such as use of the Z59.0 homelessness diagnostic code in ICD-10 (USICH, 2015c).
Despite this progress, much work remains to be done to develop a common federal vocabulary and data standards regarding housing status across agencies and programs and then to link those data with outcomes in health and other domains.
In recent years, however, sophisticated efforts to utilize diverse administrative and archival data resources to diagnose, monitor, and enhance service systems and policies have increased. Abundant opportunities remain to apply “big data”
science in research to understand and address homelessness, particularly in enhancing understanding of the associations among PSH, health, health care, and costs.
According to the 2018 Bipartisan Policy Center report, “between the various programs operated by HUD and HHS, an immense amount of data is being collected on the housing and health conditions of the U.S. population. Matching and utilizing [those] data across programs is crucial to better aligning health and housing services and ensuring that federal investments are efficiently targeted to achieve the best results.” The report goes on to recommend that “a formal data collaboration initiative between HUD and HHS could expand previous efforts, and better evaluate both local and federal opportunities to match datasets that overlap the health and housing nexus” (BPC, 2018).
In addition, the 21st Century Cures Act, signed into law in December 2016, may serve to further data integration and analysis efforts through its stipulations on the interoperability of health care data technology systems, including U.S. Department of Veterans Affairs (VA) and non-VA health and housing data, private and public claims data, criminal justice data, mortality data, etc. This, in turn, should facilitate better research.
BUILDING UNIVERSITY-AGENCY PARTNERSHIPS FOR BETTER DATA AND ANALYSIS
Individual agencies providing housing and supportive services to people who are currently experiencing homelessness or have formerly experienced homelessness typically lack sufficient resources for ongoing in-house performance monitoring and evaluation of client outcomes. Systematic, longitudinal studies following clients as they reside in permanent housing and access health and other services, for example, are rare outside of externally funded efforts initiated by researchers. Agency-initiated evaluations utilizing comparison groups are rarer still. Even the externally funded, researcher-initiated longitudinal control group designs that could prove valuable, for example, in understanding the effectiveness of PSH in addressing housing-sensitive symptoms and conditions (suggested in Chapter 3) are time limited by the grant award period. Additionally, an emphasis on scientific innovation in federally sponsored research, while critical for advancing science, may lead to an untoward consequence of hindering replication attempts and time extensions in longitudinal studies. These caveats argue for innovation and cultivation of alternative approaches in funding and conducting PSH research to enable longer-term, regular monitoring of health outcomes, programming, and costs.
University-agency partnerships may prove useful as an alternative or supplemental approach in monitoring the impact of PSH. Universities, as centers of diverse intellectual resources and other capital, are uniquely equipped to partner with and enhance the communities in which they are located. Indeed, some observers have argued that universities have a moral responsibility to serve their communities (Watson et al., 2011), and place-based community interventions in
partnership with universities have received increased support (Gewirtz, 2007; Bellamy et al., 2008). Models for building and maintaining university-agency partnerships exist in the field of child welfare for purposes of research, training, and promotion of effective policy (Zlotnick, 2010; Drabble et al., 2013). Such examples may prove helpful in developing partnerships for evaluation and policy surrounding PSH; providers in the committee’s site visits voiced strong interest in such partnerships. Community-based participatory approaches are also relevant in that, when carefully followed, they promote equitable and beneficial experiences for community-based collaborators (Wallerstein and Duran, 2010).
TESTING INNOVATIONS IN PAYMENT MODELS TO SUPPORT HOUSING AND SERVICES
The Patient Protection and Affordable Care Act (ACA) has enhanced innovation and experimentation in using Medicaid dollars to improve health and contain health care costs among Medicaid recipients. Notably relevant to serving individuals experiencing homelessness through state-level managed care organizations is the flexibility in using Medicaid dollars for housing-related costs and “health homes,” which support provision of integrated and coordinated primary and behavioral health care for disabled persons experiencing homelessness and other Medicaid recipients. A panel of experts convened through the Center for Health Care Strategies (Moses et al., 2016) issued a series of recommendations for enhancing access to housing and services for persons who are chronically homeless and Medicaid eligible. Their recommendations for further innovation in health care financing included testing the impact of different incentive arrangements in state managed care organizations on availability and access to coverage, and the effectiveness of extending coverage for housing-related services delivered in conjunction with health care.
PSH providers interviewed in site visits additionally noted the need for research to determine the most efficient means of ensuring continuity of integrated care and funding for care when clients transition from homelessness to one or more different PSH or other care settings over the years. Ensuring ongoing care across sites, systems, and the adult lifespan is increasingly important as homeless and formerly homeless populations grow older and require more medical services (Henwood et al., 2013; Brown et al., 2017).
Although future opportunities to pursue and test innovation in delivery of health care and Medicaid-reimbursable housing-related services for individuals experiencing homelessness through ACA and Medicaid expansion are unclear, research to demonstrate efficacy and effectiveness of efforts to increase quality and reduce costs of care for vulnerable populations in health plans will continue to be important. Determining effective and efficient “sharing” or braiding of Medicaid dollars and other funds is critical to the ongoing efforts to scale up PSH throughout the United States.
Research to enhance other means of cooperation and coordination across providers and systems of care to promote cost savings and quality is also critical
to the ongoing efforts to scale up PSH throughout the United States. As a case in point, the VA’s experience in expanding supportive housing to veterans experiencing homelessness across different U.S. regions illustrates not only successes, but also specific and major challenges including increased staffing demands, need for greater coordination across systems, and rental market limitations (Austin et al., 2014). Research in public health and health care improvement initiatives has documented both the opportunities and challenges inherent in widespread implementation of programs (Spoth et al., 2013; Barker et al., 2016). Challenges argue for careful organizational and implementation research to understand and facilitate ongoing scale-up efforts for PSH. See Chapter 7 for recommendations related to how federal agencies can examine their policies and programs with the goal of maximizing flexibility and the coordinated use of funding streams for supportive services and health- and housing-related care and services.
PSH holds potential for reducing the number of persons experiencing chronic homelessness, although much additional research is needed to determine the effectiveness of PSH in improving health and to clarify for whom and in which circumstances it is most beneficial. There have been relatively few well-designed, high-quality research studies about the effectiveness of PSH for health, and the evidentiary base upon which conclusions in this regard can be drawn is incomplete and suffers from a paucity of health outcomes data, inconsistencies in the definition of terms, and limited descriptions of the characteristics of different PSH models. The evidence is limited in its ability to delineate key services and minimum standards of PSH and in predicting who is most likely to benefit from it. Additional randomized controlled trials, when ethically appropriate to undertake, could bolster and refine the evidence of the impact of PSH on health outcomes and health care costs. More partnerships between universities and PSH providers to perform evaluation and monitoring of health outcomes and costs, test innovative financing models for housing and services, and mine health data and homeless management information systems could fill in many of the research and data gaps.