Permanent Supportive Housing: Assessing the Evidence

Recovery can be difficult for individuals with mental or substance use disorders who are homeless. Research shows that individuals who are homeless have worse health status than those who are at risk of homelessness (1). Compared with those who are housed, individuals who are homeless have worse outcomes in areas such as living situation, family and social relations, employment, daily activities, and legal and safety problems (2). An understanding of the impact of housing on health provides the foundation for including housing programs as a part of the treatment and recovery process for individuals with mental or substance use disorders.

This article reports the results of a literature review that was undertaken as part of the Assessing the Evidence Base (AEB) Series (see box on next page). The objectives of this review were to describe the components of permanent supportive housing programs, rate the level of research evidence of existing studies (that is, methodological quality), and describe its effectiveness compared with other housing models and treatment as usual. The review also examined differences in the effectiveness of permanent supportive housing programs for various subgroups of the population.

Description of permanent supportive housing

For purposes of the AEB Series, the Substance Abuse and Mental Health Services Administration (SAMHSA) describes permanent supportive housing as a direct service that helps adults with mental and substance use disorders who are homeless or disabled identify and secure long-term, affordable, independent housing. Service providers offer ongoing support and collaborate with property managers to preserve tenancy and help individuals resolve crisis situations and other issues. Permanent supportive housing programs differ from other living arrangements by providing a combination of flexible, voluntary supports for maintaining housing and access to individualized support services. Table 1 presents a description of permanent supportive housing and its components. Beginning in the late 1990s, SAMHSA funded a multisite demonstration with explicit attention to the principles of permanent supportive housing. Fidelity to these principles was measured in this demonstration (4). SAMHSA also recently published a tool kit that adapted these key elements of the model (4), which are listed in a box on the next page.

This review examined the evidence for permanent supportive housing, as defined by SAMHSA. Some established housing models that meet this service definition are Housing First and the U.S. Department of Housing and Urban Development–Veterans Affairs Supportive Housing program (HUD-VASH). The Housing First or Pathways to Housing programs consider housing a basic right and provide individuals with housing first. Housing is then combined with supportive treatment services in the areas of mental health and general medical health, substance use, education, and employment. Abstinence from substance use and participation in services are not required for program participation. The HUD-VASH program for homeless veterans combines housing choice via voucher rental assistance with case management and clinical services. Services are provided for participating veterans at VA medical centers and community-based outreach clinics. In addition to studies of Housing First and HUD-VASH, this review included studies of other models reported to have high to moderate adherence to the components of permanent supportive housing.

Methods

Search strategy

We conducted a survey of major databases: PubMed (U.S. National Library of Medicine and National Institutes of Health), PsycINFO (American Psychological Association), Applied Social Sciences Index and Abstracts, Sociological Abstracts, Social Services Abstracts, Published International Literature on Traumatic Stress, the Educational Resources Information Center, and the Cumulative Index to Nursing and Allied Health Literature. We also examined bibliographies of major reviews and meta-analyses. We searched for meta-analyses, research reviews, and individual studies published from 1995 through 2012. Search terms included permanent supportive housing, supported housing, supportive housing, Housing First, and services-enriched housing.

Inclusion and exclusion criteria

This review was limited to U.S. and international studies in English and included the following types of articles: randomized controlled trials (RCTs), quasi-experimental studies, single-group time-series design studies, and review articles such as meta-analyses and systematic reviews; and studies that focused on individuals with mental disorders or co-occurring mental and substance use disorders. We excluded studies focused on populations with other health conditions and studies of families, children, and adolescents. We also excluded studies of transitional, congregate, recovery, sober living, and abstinence-contingent housing models because these studies do not meet the SAMHSA definition for permanent supportive housing. Similarly, we excluded studies that focused on service interventions that support housing but that did not meet the SAMHSA definition, such as intensive case management and critical time intervention.

Strength of the evidence

The methodology used to rate the strength of the evidence is described in detail in the introduction to this series (5). The research designs of the studies that met the inclusion criteria were examined. Three levels of evidence (high, moderate, and low) were used to indicate the overall research quality of the collection of studies. Ratings were based on predefined benchmarks that considered the number of studies and their methodological quality. In rare instances when the ratings were dissimilar, a consensus opinion was reached.

In general, high ratings indicate confidence in the reported outcomes and are based on three or more RCTs with adequate designs or two RCTs plus two quasi-experimental studies with adequate designs. Moderate ratings indicate that there is some adequate research to judge the service, although it is possible that future research could influence reported results. Moderate ratings are based on the following three options: two or more quasi-experimental studies with adequate design; one quasi-experimental study plus one RCT with adequate design; or at least two RCTs with some methodological weaknesses or at least three quasi-experimental studies with some methodological weaknesses. Low ratings indicate that research for this service is not adequate to draw evidence-based conclusions. Low ratings indicate that studies have nonexperimental designs, there are no RCTs, or there is no more than one adequately designed quasi-experimental study.

We accounted for other design factors that could increase or decrease the evidence rating, such as how housing, populations, and interventions were defined; use of statistical methods to account for baseline differences between experimental and comparison groups; identification of moderating or confounding variables with appropriate statistical controls; examination of attrition and follow-up; use of psychometrically sound measures; and indications of potential research bias.

Effectiveness of permanent supportive housing

We described the effectiveness of permanent supportive housing—that is, how well the outcomes of the studies met its goals. We compiled the findings for separate outcome measures and study populations, summarized the results, and noted differences across investigations. We considered the quality of the research design in our conclusions about the strength of the evidence and the effectiveness of permanent supportive housing. We also considered whether permanent supportive housing should be covered by insurance plans on the basis of the level of evidence.

Results and discussion

Level of evidence

We identified eight literature reviews of various housing models including permanent supportive housing (6–13), seven RCTs (14–20), and other quasi-experimental studies (21–25). Tables 2 and 3 summarize the literature reviews and individuals studies included in this review, including the types of housing, the outcomes measured, and the results.

The level of evidence for permanent supportive housing was rated as moderate. Although seven RCTs have been conducted, methodological flaws, such as a lack of a definition for permanent supportive housing or inconsistently defined program elements and small samples, are reflected in the moderate rating. Stronger designs compared no housing or “nonmodel housing” (8) with various housing approaches, including community residences (22), group homes (15), and various case management interventions (17). Some studies, such as the one by Culhane and colleagues (25), used control groups that were constructed from existing data. A number of the studies also had small samples with low statistical power to detect differences, and not all used standardized measures or scientific conventions when measuring outcomes. Results were stronger for studies that compared permanent supportive housing with treatment as usual or no housing.

Most studies reviewed did not provide evidence of fidelity or did not use consistent measurements of the fidelity principles. For example, a recent comprehensive review of 25 studies on supportive and supported housing and similarly labeled programs concluded that 16 of the 25 studies used the supported housing label to describe the programs, even though the programs adhered to less than 50% of the elements of the model (7). There was a lack of clear distinction between supported and supportive housing. The terms have been used interchangeably, and different housing approaches often were clustered under the same label—even within a single study (7,26).

Effectiveness of permanent supportive housing

The outcome measures most consistently used in studies of permanent supportive housing were housing stability, hospital inpatient and emergency room use, consumer satisfaction, and behavioral health measures.

Housing and hospitalization.

Despite the shortcomings in the body of research, a consistent finding was that the provision of housing—regardless of model—had a strong, positive effect in promoting housing stability and reducing homelessness (9). Compared with no housing or the status quo, permanent supportive housing programs yielded positive outcomes on these domains. Some studies that compared permanent supportive housing programs with other types of housing have been unable to detect significant differences (6,9,19). However, a meta-analysis by Leff and colleagues (8) suggested that although a range of housing models improved tenure in housing and decreased homelessness, permanent supportive housing may have had the largest effect of all models.

A recent evaluation compared permanent supportive housing provided by the Collaborative Initiative on Chronic Homelessness with usual care and showed that when all outcome data obtained over two years were averaged, participants were housed an average of 52% more days during the previous three-month period (27). Furthermore, participants in the intervention experienced 60% fewer days of homelessness during the previous three months than those in the comparison group and had 33% fewer days of institutional care over two years. Comparisons with other specific housing models, however, did not find benefit for permanent supportive housing. When improvements among models have been found to differ (28), one explanation offered is that the intervention with stronger outcomes is more responsive to consumer needs and provides more consistent treatment to each consumer.

Three RCTs examined Housing First; for two of the RCTs, which involved people with severe mental illness, results were presented across four articles (14,18,28,29), and for one RCT, which involved people with cocaine dependence, results were presented across two articles (19,20). All studies found that participants in Housing First had significantly less homelessness compared with participants receiving standard care, day treatment with no housing, or housing that was contingent on treatment and sobriety. Housing First participants obtained housing earlier and remained stably housed. The authors interpreted their results as indicating that people with mental disorders are able to live independently. Studies also showed less hospitalization (8,9,18,21,25) and fewer emergency room visits (8,21) for participants in Housing First.

Several additional studies used strong quasi-experimental designs, either involving a comparison group of individuals in another type of housing (22,23) or a comparison group receiving another type of service (24). Findings included better housing outcomes (24) and increased housing tenure (23).

Consumer satisfaction.

Another relatively consistent finding was that low-demand permanent supportive housing, compared with other housing models in which service participation is required or services are not offered, received the highest ratings of consumer preference (28) and satisfaction (8,22). For example, Leff and colleagues (8) examined 44 unique housing alternatives described in 30 studies; those that met the principles of permanent supported housing achieved the highest effect size (.73) for satisfaction, which differed significantly from the effect sizes for residential care and treatment and for treatment as usual. RCTs also found that compared with the control group, Housing First participants reported greater perceived choice (14,23).

Behavioral health.

A majority of studies, including the meta-analysis by Leff and colleagues (8), found no effect of permanent supportive housing on psychological symptoms or alcohol or drug use. Tsemberis and associates (28) and Rosenheck and colleagues (17) found no significant difference in alcohol and drug use between groups, and Padgett and colleagues (29) found no significant difference over time that could be attributed to permanent supportive housing. Conversely, a study by Larimer and colleagues (21) found significant reduction in alcohol use for individuals in a Housing First model for homeless adults with severe alcohol problems compared with a wait-list group; however, the specific needs of this group may account for the difference in the findings from other studies on permanent supportive housing programs that focus on individuals with severe mental illness as the primary condition.

Outcomes for special populations.

Compared with studies of other treatment models for individuals with mental disorders, supportive housing studies provided little detail about residents’ diagnoses or other intake criteria. Studies that were sponsored by the National Institute of Mental Health and SAMHSA focused on individuals with severe and persistent mental illness but used different strategies for screening and measuring the disorder. Some housing programs excluded individuals with active substance use; others included individuals with co-occurring mental and substance use disorders.

The effects of permanent supportive housing on individuals with primary or co-occurring substance use disorders have not been studied thoroughly. Two exceptions include Larimer and colleagues (21), who examined the effectiveness of a Housing First model for individuals who were chronically homeless and had a primary diagnosis of a substance use disorder, and O’Connell and colleagues (30), who examined the effectiveness of HUD-VASH for male veterans who were homeless and had substance use disorders or co-occurring mental disorders. Larimer and colleagues (21) found that Housing First participants had improvements in shelter use, sobering unit use, and total utilization costs compared with control groups. They also found that housing tenure was related to better improvement in the outcome measures. O’Connell and colleagues (30) found that compared with intensive case management alone, HUD-VASH was associated with more positive housing outcomes for veterans with co-occurring mental disorders and veterans who were active substance users.

Several of the studies, such as that by Larimer and colleagues (21), focused on individuals who were homeless at the time of the investigation and had histories of homelessness. Other studies did not require homelessness as an entry criterion (22).

Studies typically reported the demographic characteristics of the individuals, including age, race and ethnicity, veteran status, education, and marital status. Most studies had a diverse study population with respect to race and ethnicity (8,21), and some examined the moderating effects of race-ethnicity on outcomes. Leff and colleagues (8) reported inconsistencies regarding outcomes for racial and ethnic subgroups for supportive housing and all other housing intervention models. Examining pre-post effect sizes, the authors of this meta-analysis found that studies with a majority of nonwhite participants reported greater reductions in drug and alcohol use and hospitalization, but these studies reported less housing stability and less satisfaction than studies in which a majority of participants were white. Another study, which compared HUD-VASH with intensive case management alone, found more positive housing outcomes for white participants than for African Americans and more positive clinical outcomes for African-American participants than for white participants (30). Similarly, studies with a majority of non-Hispanic participants found greater reductions in drug and alcohol use but less housing stability compared with studies in which a majority of participants were Hispanic (8). Edens and colleagues (31) found that men and women participating in the 11-site Collaborative Initiative on Chronic Homelessness demonstrated equal and significant improvements in the number of days housed and modest improvements in mental health outcomes over two years.

Conclusions

Studies have found that permanent supportive housing for individuals with mental and substance use disorders, compared with treatment as usual, reduced homelessness, increased housing tenure over time, and resulted in fewer emergency room visits and hospitalizations. Moreover, consumers consistently rated permanent supportive housing more positively than other housing models and preferred it over other more restrictive forms of care. On the basis of this evidence, the authors recommend that permanent supportive housing be included as a covered service as part of a full spectrum of options that support recovery for individuals with mental and substance use disorders.

The importance of the findings on consumer preference should not be underestimated. Consumer choice and consumer preference are at the center of many evidence-based practices, based on an understanding that consumers are more likely to embrace services that are tailored to their preferences and less likely to terminate services early or abruptly. Choice is recognized as an important factor in recovery, as it engages a consumer’s willingness and motivation to make life changes (32).

A key issue for permanent supportive housing is clearly defining and implementing the elements of the model. Currently, several housing programs meet the SAMHSA definition for permanent supportive housing. However, the quality of the evidence on the outcomes of permanent supportive housing is varied, and studies differ in the strength of their research design and methodological rigor. Even though multiple RCTs and some strong quasi-experimental designs have been implemented, methodological flaws prevent current research on permanent supportive housing from achieving more than a moderate level of evidence (see box on this page). Inconsistencies in the definition and implementation of models, the comparison condition used, and the outcomes measured, as well as small samples and variability in rigor, limit the ability to draw firm conclusions from the literature (6,7,8,11,26,27).

As consistently noted, there is a need to determine fidelity to permanent supportive housing principles, operationalize the interventions, and examine the program components that are related to outcomes (6,7,8,10,11). In particular, interventions should match service intensity to individual needs for various populations and capture the impact of these tailored interventions on outcomes. Absent to date is a careful review of the inclusion and exclusion criteria of various housing models that may shape the service population; this is needed to determine the generalizability of the findings.

Controlled comparative effectiveness research that systematically examines differences in outcomes among different models of permanent supportive housing is also needed. As Leff and colleagues (8) noted, because a wide variety of housing models have been found to be effective in helping participants achieve residential stability and other positive outcomes, random assignment of individuals with mental and substance use disorders to different housing models should pose no ethical concerns. Finally, it would be helpful to have sensitivity within these studies to the moderating effects of individual characteristics, especially race, ethnicity, and age. Further research would provide a more complete understanding of which models yield the greatest improvement on a range of outcomes for various subpopulations.

Policy makers including payers (for example, state mental health and substance use directors, managed care companies, and county behavioral health administrators) should consider mechanisms that would support permanent supportive housing. A defined evaluation component is needed to continue to build the evidence base for permanent supportive housing as policy makers and others determine how best to incorporate it into a full continuum of care. Individuals with mental and substance use disorders can benefit from increased access to permanent supportive housing as a long-term support for a life in recovery in the community.