5

Effect of Individual and Program Characteristics on Outcomes in Permanent Supportive Housing

This chapter reviews the evidence that addresses two distinct and important questions related to the effectiveness of permanent supportive housing (PSH). First, are there subgroups of individuals who are experiencing homelessness who have better outcomes when housed under a PSH model compared to usual care? Second, are there certain characteristics of PSH programs that are associated with better housing and health outcomes among clients of these programs?

INDIVIDUAL CHARACTERISTICS AND OUTCOMES

When exploring the relationship between individual characteristics of PSH and housing and health outcomes, data may be obtained through subgroup analyses in which study participants are classified after the fact into subgroups (e.g., younger versus older individuals) and outcomes are compared within or between the various subgroups. Alternatively, regression analyses can be used to identify individual characteristics associated with better outcomes. Individual characteristics that have been examined using these techniques include age and substance use.

In addition to subgroup analyses described in the previous paragraph, another possible source of data that may be obtained on the association between individual characteristics and outcomes is the assessment tool used to collect a wide range of information on the characteristics and needs of individuals experiencing homelessness when they first come into contact with a service provider (HUD, 2015a). These assessment tools can be used to facilitate engagement with persons experiencing homelessness in the community or to ascertain their eligibility for various programs. Most importantly, assessment tools can be used to identify individuals who are believed to have the greatest need for housing. This usage is based on the assumption that assessment tools can identify individuals who are at highest risk of poor health outcomes and that these individuals will derive the greatest benefit from receipt of PSH. Thus, information from assessment tools, if linked with outcome data, would be expected to provide insights

into the relationship between individual characteristics and outcomes. The committee found one study that explicitly considered nine different subpopulations and their success in PSH (Seligson et al., 2013). The New York/New York III Supportive Housing Agreement brought together state and local government agencies to provide 9,000 units of new PSH to serve these nine subpopulations. Individuals receiving PSH were compared to eligible individuals not placed in PSH. Although there are only interim evaluation data on cost and utilization measures, the results combined across all subpopulations look promising. Seligson et al. (2013) conclude that “tenants had savings in jail, shelter, state psychiatric facilities, and Medicaid utilization and costs relative to people eligible but not placed in the program” (p. 24). In terms of individual characteristics, “specific costs varied in the types of public service for which they had savings, as well as their net costs” (p. 24).

Age

An important question is whether younger or older individuals are more likely to experience benefits from PSH. Data from the At Home/Chez Soi study, conducted in five cities in Canada, randomized persons with serious mental illness experiencing homelessness to receive usual care or permanent supportive housing using the Housing First model (Stergiopoulos et al., 2015).¹ A subgroup analysis of data from this study examined housing outcomes among homeless young adults (18–24 years old)² with mental illness who were randomized to receive PSH using a Housing First model (n = 87) or treatment as usual (housing and mental health services routinely available in the community, apart from the study) (n = 69) (Kozloff et al., 2016). The percentage of time stably housed in these young adult groups over 24 months was significantly higher in the intervention group than the control group (65 percent and 31 percent, respectively, with an adjusted mean difference of +34 percent). However, there were no overall differences between the intervention and control groups in terms of quality of life, physical and mental health status, psychiatric symptoms, emergency department visits, or arrests.

Chung et al. (2018) in the At Home/Chez Soi study compared the outcomes of older (50 years and older, n = 470) and younger (18–49 years old, n = 1,678) homeless adults with mental illness. The Housing First intervention increased the percentage of time stably housed among both older and younger homeless adults

___________________

¹ In 2008 the Mental Health Commission of Canada (MHCC) undertook a 4-year research demonstration project on mental health and homelessness in Moncton, Montreal, Toronto, Vancouver, and Winnipeg. It is known as the “At Home/Chez Soi study.”

² According to the National Institutes of Health (NIH), young adults are identified as between the ages of 18 and 24. This is also the age group that the Department of Housing and Urban Development (HUD) describes as “youth.” This definition of youth may not apply to the age definition of youth used in older studies and studies conducted in other countries. Note that Kozloff and colleagues’ (2016) analysis did not compare the findings to different age groups.

compared to those receiving treatment as usual (mean differences of +43 percent and +40 percent, respectively). Older adults, however, experienced greater improvements in quality of life, psychiatric symptoms, and mental health status under the Housing First intervention compared to adults 18–49 years old.

According to the Corporation for Supportive Housing (CSH, 2016b), homelessness is also on the rise among older adults (individuals over age 50). Over the past three decades, the median age of single homeless adults in San Francisco increased from 37 in 1990 to 46 in 2003 (Hahn et al., 2006). Across the United States, the modal age of single adults experiencing homelessness increased from 34–36 years in 1990 to 49–51 years in 2010, with people in the last half of the Baby Boom cohort at highest risk. By contrast, the modal age of adults in homeless families remained 21–23 years throughout this period (Culhane et al., 2013). The percentage of sheltered homeless single adults over 50 years increased from 16.5 percent in 2007 to 21.4 percent in 2015³ (HUD, 2016a, Part 2).

Compared to their younger peers, older homeless adults have higher rates of chronic illnesses, may suffer from multiple physical and psychological comorbidities (Garibaldi et al., 2005), and are more likely to die from chronic conditions including cardiovascular disease and cancer (Baggett et al., 2013). Homeless adults over age 50 also have been shown to have rates of chronic illness comparable to or higher than community-dwelling adults 15–20 years their senior, including conditions more commonly linked to much older individuals, such as memory loss and difficulty performing activities of daily living (Brown et al., 2012, 2017). The combination of issues typically associated with homelessness such as mental health and substance abuse with those related to aging such as reduced mobility and a need for assistance with daily activities is challenging providers who serve this population to develop creative solutions.

Different communities have developed PSH programs to address the housing and health needs of older adults experiencing homelessness (Brown et al., 2013; Henwood et al., 2015c). CSH suggests that these may be more cost-effective than nursing homes for addressing the needs of this population (CSH, 2011). Although some argue that older adults facing chronic health problems would benefit from PSH, most studies show no evidence of differential health outcomes in older adults across residential settings (Bamberger and Dobbins, 2015; Kogan et al., 2016), in contrast to the results presented on the previous page from the At Home/Chez Soi study (Chung et al., 2018).

___________________

³ This study is also discussed in Chapter 2.

Substance Use

The provision of PSH is usually not contingent on abstinence from alcohol and drug use. In fact, evidence from the Housing First model suggests that requiring abstinence from substance use prior to PSH provision is likely to prolong the duration of chronic homelessness (see below for further details). However, it is reasonable to question whether high levels of substance use are associated with less positive outcomes after the provision of housing. In an evaluation of the Collaborative Initiative on Chronic Homelessness, which provided PSH and mental health services in 11 communities, participants were classified as high-frequency substance users (>15 days alcohol use or illicit drug use [including marijuana] in the past 30 days, n = 120) or nonusers (no days of use in the past 30 days, n = 290) (Edens et al., 2011). Participants with intermediate levels of substance use were excluded from this analysis. Over a 24-month follow-up period, high-frequency substance users and nonusers experienced comparable levels of improvement in days housed, days of inpatient hospitalization, and days in prison or jail. High-frequency substance users continued to have higher levels of substance use than nonusers, but their frequency of substance use declined over time. There was no evidence of deterioration in mental health outcomes among high-frequency users compared to nonusers.

Collins et al. (2013) examined individual characteristics associated with housing outcomes among individuals experiencing chronic homelessness who entered the single-site Eastlake Housing First program (in Seattle, Washington). Among 111 individuals with outcome data at 2 years, 23 percent returned to homelessness during the follow-up period. Age, sex, race/ethnicity, homelessness history, substance use, medical comorbidities, or psychiatric symptoms were not significant predictors of a return to homelessness. When examining the outcome of cumulative time housed over the follow-up period, any alcohol use in the past 30 days at baseline was associated with increased time housed, whereas any drug use in the past 30 days at baseline and psychotic symptoms at baseline were associated with decreased time housed. This finding suggests that among substance-using individuals experiencing homelessness in PSH, housing stability is more easily attained for persons whose substance of choice is alcohol rather than drugs, but it is not clear whether the same finding would hold for individuals who did not receive PSH. Overall, it is not clear whether PSH is differentially effective for people who use different substances.

Other Individual Characteristics

The At Home/Chez Soi study also provides information regarding the outcomes of a number of other subgroups participating in the study (Stergiopoulos et al., 2015). One subgroup study by Volk et al. (2016) found that during the first 12 months after randomization, 86.5 percent of participants assigned to the Housing First intervention achieved housing stability (defined as being housed more than 50 percent of the time during months 3–12 and/or being housed 100 percent of

the time during months 9–12), whereas 13.5 percent of participants did not achieve housing stability. All participants had a diagnosis of mental illness. Individual characteristics associated with lack of housing stability in the first 12 months were longer cumulative lifetime duration of homelessness, proportion of time spent in jail during the 3 months prior to randomization, and lower community psychological integration. A diagnosis of post-traumatic stress disorder (PTSD) or panic disorder—in contrast to the most common psychiatric diagnoses among participants, major depression and psychotic disorders—was a predictor of housing stability. However, predictive models correctly identified only 3.8 percent of individuals who failed to achieve housing stability.

Assessment Tools

A number of assessment tools are available to guide the allocation of housing assistance, and Continuum of Care programs funded by HUD are required to use a standardized assessment tool with clients (HUD, 2015a). An expert panel recently convened by HUD concluded, however, that it is unknown whether any assessment tool identifies individuals whose housing or health outcomes are more likely to be improved by the provision of PSH (HUD, 2015a). No peer-reviewed studies have examined this question. Despite their widespread use, assessment tools cannot be assumed to be accurate predictors of an individual’s responsiveness to PSH, particularly as there is no evidence that individuals who screen in or out with these tools are qualitatively different from each other. It may be possible, however, to have a measure based on housing-sensitive conditions (Recommendation 3-2) that would provide a stronger basis for allocation of resources.

PROGRAM CHARACTERISTICS AND OUTCOMES

There is substantial diversity among existing models of PSH (see Chapter 2). It is therefore important to consider whether certain characteristics of PSH programs are associated with better outcomes for the clients of these programs. Equally important is the question of whether similarly positive outcomes can be achieved with models of PSH that are easier or require fewer resources to implement. The three main PSH dimensions to be considered are (1) characteristics of the housing, (2) characteristics of the supportive services, and (3) the level of resources available for housing and supportive services.

Ideally, the assessment of the effect of program characteristics on outcomes would be based on data from controlled trials in which individuals experiencing homelessness were randomized to different PSH programs. Given limitations in the published literature, findings were also considered from quasi-experimental studies in which individuals experiencing homelessness were assigned to different models of PSH on a nonrandom basis, as well as data from observational studies that examined the association between various program characteristics and individual-level outcomes. No attempt was made to compare outcome data from stud-

ies examining different models of PSH if these studies were conducted and reported independently of one another. Such comparisons would have a very high risk of bias due to differences in study populations and study designs.

Characteristics of Housing

Single-Site Versus Scattered-Site Models

A critical differentiating characteristic between PSH models is the use of the single-site versus scattered-site model, as described in Chapter 2. In the single-site model, housing is provided at a dedicated building with support services attached to the site. This model is sometimes referred to as congregate housing or project-based housing. In the scattered-site model, housing is provided in existing private market rental units dispersed throughout the community. These models have significant implications for how additional PSH is created, as the single-site model requires new construction or the identification of entire buildings that are available to lease, whereas the scattered-site model requires availability of rental units in the private housing market. Other differentiating features related to costs, availability, housing environment, and support services are detailed in Table 5-1. An important outcome consideration is the impact of the single-site and scattered-site models on the social integration of clients; however, research findings on this question have been mixed (Quilgars and Pleace, 2016).

TABLE 5-1 Features of Single-Site Versus Scattered-Site PSH

With respect to costs, a report by the General Accounting Office (GAO) estimated that the average total 30-year costs for one-bedroom units in the same general location are 8–19 percent higher for programs that produce housing (such as the construction of a single-site supportive housing building) compared to housing vouchers (which are used in scattered-site supportive housing programs) (GAO, 2002).

As noted earlier, with scattered-site housing, the individual/family receives a rental subsidy in the form of a Section 8 voucher. Tenants are required to pay 30 percent of their monthly income toward rent and utilities; the remaining funds

are provided via the voucher to the landlord. The major advantage of scattered-site housing is that it allows for renting units on the private housing market, rather than constructing new housing units or identifying empty apartment buildings that are available to lease. As noted earlier, under the Pathways Housing First model, individuals choose their own residence.

Scattered-site housing typically involves working with individual landlords to secure agreements to rent to individuals experiencing chronic homelessness, which may not be easy. A disadvantage of scattered-site housing is that supportive services must be accessed via mobile support or travel to the services site. For example, Barnes (2012) described barriers to service delivery for scattered-site housing in Toronto that included the time and costs of providers traveling from one location to the next. However, Hogan (1996) cited several surveys indicating that residents prefer scattered-site housing to single-site housing.

Single-site housing means that the units are clustered in the same building, block, or neighborhood. One advantage of single-site programs is that it is far more likely that supportive services are part of the building or in its immediate proximity, which likely plays a role in encouraging residents to access these services (e.g., Collins et al., 2013). Residents also report increased feelings of security (Parsell et al., 2015). Improved social integration at single-site housing is considered an advantage, but neighborhood opposition and resident concerns about declining property values, often described as “not in my backyard” (NIMBY), can be a disadvantage (Hogan, 1996). More recent research, however, appears to indicate that NIMBY concerns are minimal (Palmer, 2016) (as discussed in Chapter 7).

“Mixed housing” developments provide PSH alongside housing to low-income tenants not participating in PSH in the same community or development (Wilkins et al., 2011, 2014). Supportive services may be delivered either on- or off-site. Examples of mixed-housing models include the New Haven Project in New Haven, Connecticut, where a percentage of units are set aside for individuals experiencing homelessness (Wilkins and Burt, 2012). The New Haven project is administered by the state Department of Mental Health and Addiction Services.

A number of studies have examined the effect of single-site versus scattered-site models on housing and health outcomes. For example, in a randomized controlled trial (RCT) conducted in Boston, 118 homeless adults with major mental illness were randomly assigned to group housing with staff support and gradually increasing self-governance versus placement in independent apartments (Dickey et al., 1996; Goldfinger et al., 1999). Both groups received case management support. There was no significant difference between the two groups in terms of housing status at the end of 18 months, although individuals assigned to group housing had fewer days of homelessness over the course of the follow-up period. The use of inpatient and outpatient mental health services was similar in both groups. Neuropsychological functioning was not significantly different between the two groups on 10 of 11 measures (Seidman et al., 2003).

McHugo and colleagues (2004) conducted a study in which adults with severe mental illness who were homeless or at high risk of homelessness were randomized to receive scattered-site housing with Assertive Community Treatment (ACT) services versus single-site congregate housing with integrated case management. However, there was a large degree of crossover in housing types between the two groups, with the percentage of participants living in their own apartment at 18 months being 53 percent and 47 percent, respectively. As a result, it is not possible to draw conclusions from this study regarding outcomes in scattered-site versus congregate-site housing. In an assessment of 125 individuals experiencing chronic homelessness with mental illness in a PSH program that provided a modified ACT services and primary care to meet health needs, Henwood et al. (2011) found that through collaborative primary care partnerships, ACT can serve as a medical home for individuals with psychiatric disabilities and co-occurring serious health problems. With ongoing effort to measure outcomes, this program can help inform the development of a comprehensive model of integrated care.

A quasi-experimental study conducted in New York City compared outcomes among 157 individuals with severe mental illness and a history of homelessness who were entering either “supported housing” (residential hotels or scattered-site apartments) or single-site “community residences” for persons with mental illness with on-site dining and intensive support services (Siegel et al., 2006).⁴ Using propensity scoring methods to adjust for differences between the two groups, the proportion of individuals who remained housed over the 18-month observation period did not differ significantly between the two programs. Scattered-site residents reported significantly greater housing satisfaction in terms of autonomy and use of discretionary funds, but they also tended to report greater feelings of isolation. The authors noted that the supported housing sites were substantially less costly to operate than the single-site community residences.

At the Vancouver site of the At Home/Chez Soi study, participants who received the PSH intervention were assigned on a nonrandom basis to either scattered-site units in market housing (n = 90) or units at a single-site congregate housing building reserved for study participants (n = 107) (Somers et al., 2017). During the 24-month follow-up period, the proportion of time spent in stable housing was similar among participants in scattered-site versus single-site congregate housing (74.5 percent versus 74.3 percent). However, differences were observed on certain secondary outcome measures. Individuals in congregate housing had greater improvements in community functioning, psychological community integration, and mental health recovery than individuals in scattered-site units. There were no significant differences in terms of changes in physical community integration, psychiatric symptoms, quality of life, or substance use problems. In addition, there were no differences in daily substance use, as assessed by the Maudsley Addiction Profile (Somers et al., 2015).

___________________

⁴ This study is based on a non-PSH model.

Housing Quality

The At Home/Chez Soi study examined whether housing quality is associated with housing stability (Adair et al., 2016). Housing quality was assessed using a newly developed multidimensional standardized Observer-Rated Housing Quality Scale (OHQS) that involved in-person assessment of housing units and buildings by trained research assistants, based on unit (safety/security, utilities, etc.), building (staff, inside and outside conditions, etc.), and neighborhood (location, transportation access, etc.) scales. OHQS scores were obtained for 204 individuals randomized to scattered-site housing and 228 individuals randomized to usual care but who nonetheless obtained housing. OHQS scores ranged from 13.5 (lowest possible quality) to 135 (highest possible quality). Housing quality scores were positively associated with housing stability: 73.4 (95 percent confidence interval [CI] = 68.3–78.5) for those housed none of the time; 91.1 (95 percent CI = 89.2–93.0) for those housed some of the time; and 93.1 (95 percent CI = 91.4–94.9) for those housed all of the time. In regression models, housing quality was significantly associated with housing stability at 24 months of follow-up, even after adjustment for city, housing characteristics, participant ethnicity, community functioning, and social support. The study demonstrates that average individuals considered the unit condition to be most important, then the neighborhood, then the building; however, additional research is needed to assess the impact of the environment on housing retention, including individual choice as well as family and community circumstances (Adair et al., 2016).

Characteristics of Supportive Services

Housing First Versus Treatment First

In a landmark RCT, Tsemberis and colleagues (2004) compared two approaches to providing PSH for chronically homeless adults with serious mental illness. (See also Chapter 2 for more about the two approaches.) In the treatment-first approach, individuals experiencing homelessness are transitioned from living on the street to shelters, from shelters to transitional housing, and from transitional housing to permanent housing when the individual is deemed “housing ready.” Compliance with psychiatric treatment and abstinence from substance use are expected to achieve “housing readiness.”

In contrast, the Housing First approach offers immediate housing in an independent apartment, without any requirement that the individual comply with psychiatric treatment or abstain from substance use. Individuals experiencing homelessness are offered support and treatment by an ACT team and encouraged to define their own recovery-oriented goals. Participants randomized to the scattered-site Housing First program (n = 87) achieved housing much more rapidly than those assigned to treatment first (n = 119), with the proportion of time stably housed over the first 6 months approximately 65 percent versus 15 percent, respectively. This significant difference in housing stability was sustained over 24

months. Housing First participants spent less time in the hospital, but there were no significant differences between the two groups in psychiatric symptoms, alcohol use, or drug use.

Several studies have examined how the programs’ adherence to the Pathways to Housing model of PSH relates to outcomes. Gilmer et al. (2014) examined housing outcomes across 96 “full-service partnership programs” in California. The programs provided subsidized permanent housing and multidisciplinary team-based services focused on rehabilitation and recovery. Programs with higher fidelity, especially on dimensions of separation of housing and services and participant rights to choose and reject services had better housing outcomes. The Canadian At Home/Chez Soi study examined fidelity to the Pathways Housing First model across 12 PSH programs. Those with greater fidelity had greater improvements in housing stability and also in community functioning as rated by observers and quality of life as rated by participants (Goering et al., 2016). Davidson et al. (2014) examined fidelity to the Pathways Housing First program across nine PSH programs in New York City. Clients in programs with higher fidelity on consumer participation (service plans driven by clients with no requirements for substance abuse treatment) had better housing retention and were less likely to report using opiates and stimulants.

Intensity of Services and Concordance with Need

Common sense suggests that the intensity and type of supportive services provided to persons experiencing homelessness should be tailored to the severity of the individual’s illnesses, level of need, and personal preferences. Supportive housing programs that serve individuals with higher levels of physical and mental comorbidities or more severe behavioral issues will generally need to provide more intensive and therefore more costly services to appropriately support their clients. However, no studies have directly compared the provision of supportive services of different intensities to a group of individuals experiencing homelessness with a defined level of need to assess for differences in effects on housing and health outcomes.

In the At Home/Chez Soi study, individuals experiencing homelessness with a current mental health disorder were classified as having either high needs or moderate needs for treatment using a complex algorithm based on psychiatric and substance use diagnoses, community functioning score, and previous pattern of hospitalizations or incarceration (Stergiopoulos, 2015). Over a 2-year period, participants with high or moderate needs who were either experiencing homelessness or were “precariously housed” were randomized to a Pathways Housing First model of primarily scattered-site permanent supportive housing with ACT versus usual care (Aubry et al., 2016). Housing First participants spent more time being stably housed, reported having a higher quality of life, and had better community functioning. However, at the end of the second year, the two groups both “improved substantially” (Aubry et al., 2016, p. 278).

Level of Resources for Housing and Supportive Services

No studies have specifically examined the relationship between higher versus lower levels of funding for housing and supportive services and client outcomes. Given the association between housing quality and housing stability noted above, the committee expects that program funding levels that enable individuals experiencing homelessness to be housed in relatively higher-quality units may increase the likelihood of successful housing outcomes.

CONCLUSIONS

There is some evidence that individual characteristics of the people using PSH have a modest impact on the outcomes achieved with PSH. According to Chung et al. (2018), in the At Home/Chez Soi study, persons 50 years and older may derive somewhat greater mental health benefits from PSH than younger individuals, although reductions in homelessness are similar across age groups. The evidence is inconclusive as to whether persons who abuse alcohol and/or drugs derive generally comparable housing and health benefits from PSH, compared to persons who do not abuse substances. There is no evidence to support the use of current predictive models to identify individuals who are unlikely to achieve housing stability through PSH programs. As Toros and Flaming (2017) suggest, “additional predictive tools are needed to effectively target segments of the population experiencing homelessness that are appropriate for earlier interventions” (p. 27). As described above, assessment tools are widely used to collect information on the characteristics and needs of people experiencing homelessness. Most importantly, they are used to identify individuals who are believed to have the greatest need for housing. Despite their widespread use, there is a lack of evidence regarding the use of these tools to identify individuals who are more likely to have improved outcomes if provided with PSH.

With respect to program characteristics, there is good evidence from multiple studies that single-site and scattered-site supportive housing programs result in comparable levels of housing stability over follow-up periods of up to 2 years. There is less conclusive evidence with respect to health and other outcomes. One study found that scattered-site models resulted in greater housing satisfaction but also possibly higher rates of feelings of isolation. Another study found that a single-site model was associated with improved community functioning, psychological community integration, and recovery, but that scattered-site housing resulted in significantly fewer criminal sentences. Based on the available data, definitive claims cannot be made regarding the relative effectiveness of single-site versus scattered-site PSH with respect to outcomes other than housing stability.

There is evidence that the Housing First model of immediate housing in scattered-site units with ACT support results in better housing outcomes and possibly fewer days in hospital compared to a treatment-first approach that uses transitional housing and treatment for psychiatric illness and substance use to help

individuals achieve “housing readiness.” Further, fidelity to the Pathways Housing First model, with emphasis on client choice over services, leads to better housing outcomes and, in some studies, better quality of life, better community functioning, and reduced substance use. Apart from this, there is a notable lack of studies comparing the effect of different characteristics of supportive services on housing, health, or health care utilization outcomes.

The positive association between housing quality and housing stability suggests the need to ensure good housing quality in the selection of scattered-site buildings and the construction and management of single-site buildings. Careful attention is needed to ensure that PSH programs receive funding that is sufficient to achieve this goal.

In its assessment of existing studies, the committee was hampered by a less than robust literature to assess the effect of individual and program characteristics on outcomes in PSH. The PSH models, for example, are required to include the appropriate supportive services for the individuals being served. However, the existing literature lacks information on the type, intensity, frequency, or length of these services, nor are there clear details of what constitutes “usual services” when comparing the efficacy of different models. As such, it remains difficult to generalize who among individuals experiencing homelessness are most likely to benefit from them. Furthermore, there is no agreement on what the best supportive service models are for the different groups of individuals that are housed. To address these gaps, the committee makes the following recommendation: