A comprehensive review of prioritised interventions to improve the health and wellbeing of persons with lived experience of homelessness

Aliza Moledina; Olivia Magwood; Eric Agbata; Jui‐Hsia Hung; Ammar Saad; Kednapa Thavorn; Kevin Pottie

doi:10.1002/cl2.1154

. 2021 Jun 24;17(2):e1154. doi: 10.1002/cl2.1154

A comprehensive review of prioritised interventions to improve the health and wellbeing of persons with lived experience of homelessness

Aliza Moledina ¹, Olivia Magwood ², Eric Agbata ³, Jui‐Hsia Hung ⁴, Ammar Saad ⁵, Kednapa Thavorn ⁶, Kevin Pottie ^7,^✉

PMCID: PMC8356292 PMID: 37131928

Abstract

Background

Homelessness has emerged as a public health priority, with growing numbers of vulnerable populations despite advances in social welfare. In February 2020, the United Nations passed a historic resolution, identifying the need to adopt social‐protection systems and ensure access to safe and affordable housing for all. The establishment of housing stability is a critical outcome that intersects with other social inequities. Prior research has shown that in comparison to the general population, people experiencing homelessness have higher rates of infectious diseases, chronic illnesses, and mental‐health disorders, along with disproportionately poorer outcomes. Hence, there is an urgent need to identify effective interventions to improve the lives of people living with homelessness.

Objectives

The objective of this systematic review is to identify, appraise, and synthesise the best available evidence on the benefits and cost‐effectiveness of interventions to improve the health and social outcomes of people experiencing homelessness.

Search Methods

In consultation with an information scientist, we searched nine bibliographic databases, including Medline, EMBASE, and Cochrane CENTRAL, from database inception to February 10, 2020 using keywords and MeSH terms. We conducted a focused grey literature search and consulted experts for additional studies.

Selection Criteria

Teams of two reviewers independently screened studies against our inclusion criteria. We included randomised control trials (RCTs) and quasi‐experimental studies conducted among populations experiencing homelessness in high‐income countries. Eligible interventions included permanent supportive housing (PSH), income assistance, standard case management (SCM), peer support, mental health interventions such as assertive community treatment (ACT), intensive case management (ICM), critical time intervention (CTI) and injectable antipsychotics, and substance‐use interventions, including supervised consumption facilities (SCFs), managed alcohol programmes and opioid agonist therapy. Outcomes of interest were housing stability, mental health, quality of life, substance use, hospitalisations, employment and income.

Data Collection and Analysis

Teams of two reviewers extracted data in duplicate and independently. We assessed risk of bias using the Cochrane Risk of Bias tool. We performed our statistical analyses using RevMan 5.3. For dichotomous data, we used odds ratios and risk ratios with 95% confidence intervals. For continuous data, we used the mean difference (MD) with a 95% CI if the outcomes were measured in the same way between trials. We used the standardised mean difference with a 95% CI to combine trials that measured the same outcome but used different methods of measurement. Whenever possible, we pooled effect estimates using a random‐effects model.

Main Results

The search resulted in 15,889 citations. We included 86 studies (128 citations) that examined the effectiveness and/or cost‐effectiveness of interventions for people with lived experience of homelessness. Studies were conducted in the United States (73), Canada (8), United Kingdom (2), the Netherlands (2) and Australia (1). The studies were of low to moderate certainty, with several concerns regarding the risk of bias. PSH was found to have significant benefits on housing stability as compared to usual care. These benefits impacted both high‐ and moderate‐needs populations with significant cimorbid mental illness and substance‐use disorders. PSH may also reduce emergency department visits and days spent hospitalised. Most studies found no significant benefit of PSH on mental‐health or substance‐use outcomes. The effect on quality of life was also mixed and unclear. In one study, PSH resulted in lower odds of obtaining employment. The effect on income showed no significant differences. Income assistance appeared to have some benefits in improving housing stability, particularly in the form of rental subsidies. Although short‐term improvement in depression and perceived stress levels were reported, no evidence of the long‐term effect on mental health measures was found. No consistent impact on the outcomes of quality of life, substance use, hospitalisations, employment status, or earned income could be detected when compared with usual services. SCM interventions may have a small beneficial effect on housing stability, though results were mixed. Results for peer support interventions were also mixed, though no benefit was noted in housing stability specifically. Mental health interventions (ICM, ACT, CTI) appeared to reduce the number of days homeless and had varied effects on psychiatric symptoms, quality of life, and substance use over time. Cost analyses of PSH interventions reported mixed results. Seven studies showed that PSH interventions were associated with increased cost to payers and that the cost of the interventions were only partially offset by savings in medical‐ and social‐services costs. Six studies revealed that PSH interventions saved the payers money. Two studies focused on the cost‐effectiveness of income‐assistance interventions. For each additional day housed, clients who received income assistance incurred additional costs of US$45 (95% CI, −$19, −$108) from the societal perspective. In addition, the benefits gained from temporary financial assistance were found to outweigh the costs, with a net savings of US$20,548. The economic implications of case management interventions (SCM, ICM, ACT, CTI) was highly uncertain. SCM clients were found to incur higher costs than those receiving the usual care. For ICM, all included studies suggested that the intervention may be cost‐offset or cost‐effective. Regarding ACT, included studies consistently revealed that ACT saved payers money and improved health outcomes than usual care. Despite having comparable costs (US$52,574 vs. US$51,749), CTI led to greater nonhomeless nights (508 vs. 450 nights) compared to usual services.

Authors' Conclusions

PSH interventions improved housing stability for people living with homelessness. High‐intensity case management and income‐assistance interventions may also benefit housing stability. The majority of included interventions inconsistently detected benefits for mental health, quality of life, substance use, employment and income. These results have important implications for public health, social policy, and community programme implementation. The COVID‐19 pandemic has highlighted the urgent need to tackle systemic inequality and address social determinants of health. Our review provides timely evidence on PSH, income assistance, and mental health interventions as a means of improving housing stability. PSH has major cost and policy implications and this approach could play a key role in ending homelessness. Evidence‐based reviews like this one can guide practice and outcome research and contribute to advancing international networks committed to solving homelessness.

1. PLAIN LANGUAGE SUMMARY

1.1. Housing, income assistance, and case management improve housing outcomes for persons with lived experience of homelessness

Permanent supportive housing (PSH) interventions appear to improve short‐ and long‐term housing stability for persons with lived experience of homelessness. Income assistance and intensive mental health interventions show moderate benefits in housing outcomes, and evidence on standardised case management suggests potential to improve housing stability. Peer support alone does not impact housing stability. Inconsistent results on mental health, substance use and other social outcomes require additional research.

1.1.1. What is this review about?

Homelessness greatly magnifies morbidity and mortality and worsens preventable health and social inequities. We present evidence on a wide range of interventions targeting homelessness: PSH; income assistance; standard case management (SCM) and peer support; mental health interventions such as assertive community treatment (ACT), intensive case management (ICM), critical time intervention (CTI), and injectable antipsychotics; and substance use interventions such as SCFs, managed alcohol programmes (MAPs) and pharmacological interventions for opioid use disorders.

What is the aim of this review?

This systematic review and meta‐analysis examines the effects of a broad range of interventions on housing stability, mental health, quality of life, substance use, hospitalisations and health service utilisation, as well as employment and income among individuals with lived experience of homelessness.

1.2. What studies are included?

We included 86 studies across 128 publications among individuals with lived experience of homelessness. The vast majority of studies followed a randomised controlled design. Most took place in the United States (73). The rest were undertaken in Canada (8), the UK (2), the Netherlands (2) and Australia (1).

1.3. What are the main findings of this review?

Studies on housing interventions showed significant improvements in housing stability, with potential sustained benefit for up to 5.4 years. Income assistance interventions also appeared to be effective in improving housing outcomes. SCM carried the potential to improve housing, with mixed evidence suggesting its added benefit, whereas peer support programmes demonstrated no impact on housing relative to usual care.

Intensive mental health interventions demonstrated moderate improvements in housing stability and often worked in synergy with permanent housing.

Cost‐analysis studies of housing interventions reported mixed economic results. Income assistance was associated with increased costs that were offset by its added benefits. Intensive mental health interventions, such as ACT, ICM and CTI, were found to be economically beneficial. In contrast, SCM did not offer good value for money compared to other interventions.

No economic evidence was found for peer support, injectable antipsychotics or substance use interventions.

1.4. What do the findings of this review mean?

PSH may improve and maintain housing stability. Further examination of implementation barriers of housing programmes is needed to inform decisionmakers.

Income assistance, SCM and intensive mental health interventions carry the potential to improve housing outcomes, but more research is needed to examine their mechanisms. Our results on mental health and other social outcomes were mixed and inconclusive. This could be attributed to the significant proportion of study participants who were suffering from chronic mental health or substance use conditions.

1.5. What are the implications for research and policy?

More longitudinal research is needed to better examine nonhousing outcomes. Furthermore, poor reporting, lack of blinding and allocation bias reduced the certainty and precision of our results.

There are other ongoing gaps that warrant more investigation, including peer support programmes, community substance use interventions, and programmes targeted towards special populations. Further examination of implementation barriers of housing programmes is also needed.

The absence of evidence on substance use interventions for people living with homelessness represents an important research and policy gap.

1.6. How up‐to‐date is this review?

The review authors searched for studies that had been published up until February 10, 2020.

2. BACKGROUND

2.1. The problem, condition or issue

Worldwide, over 1.8 billion people lack adequate housing and almost 25% of the world's urban population reside in informal accommodation (UN HRC, 2019). “People with a lived experience of homelessness” is a term coined to describe individuals who are, have been, or at risk of becoming homeless. This population lacks stable, permanent, appropriate housing, or may be without immediate prospect, means and ability to acquire it (Canadian Observatory on Homelessness, 2017). This population continues to grow, giving rise to a major international clinical and public health priority. Homelessness is strongly associated with high levels of morbidity (Hwang, 2009) and mortality (Nordentoft & Wandall‐Holm, 2003). People with lived experience of homelessness are at an increased risk for acute illnesses such as traumatic injury (including brain injury), frostbite, peripheral vascular disease, soft tissue infections, and dental decay (Hwang & Bugeja, 2000). Many homeless people also suffer from chronic medical conditions such as diabetes (Hwang & Bugeja, 2000), cardiovascular disease (Lee et al., 2005), cancer (Krakowsky et al., 2013) and respiratory illnesses (Raoult et al., 2001). Rates of serious mental illness (Fazel et al., 2014), cognitive impairment (Stergiopoulos et al., 2015b) and drug and alcohol use (Aubry et al., 2012; Grinman et al., 2010; Kennedy et al., 2017; Kerr et al., 2009; Torchalla et al., 2011) are disproportionately high, as are rates of homicide and suicide (Cheung & Hwang, 2004). Moreover, people who are homeless experience a disproportionately high prevalence of infectious diseases such as hepatitis C, HIV and tuberculosis (Beijer et al., 2012; Corneil et al., 2006; Roy et al., 2001). Despite the significant burden of disease, people with lived experience of homelessness are less likely to access and maintain the care required for their cure and treatment (Milloy et al., 2012; Palepu et al., 2011). People with lived experience of homelessness encounter many barriers to health and social care. The competing need to find food and shelter results in delays in accessing health care services (Gelberg et al., 1997) and those who do seek health care often experience discrimination that precludes adequate uptake of preventative health services (Wen et al., 2007). The structural stigma they experience when accessing health or social services is a major cause of their health inequities (Hatzenbuehler et al., 2013). For example, people with lived experience of homelessness with coexisting mental health conditions report specific barriers to accessing care, such as being unaware of the location of care, affordability, wait times and having experienced previous rejection from health or social services (Rosenheck et al., 1997). In addition, many health care recommendations, such as dietary advice, can prove impossible without access to resources, such as proper nutrition and cooking facilities (Hwang, 2001). This lack of appropriate access to community based care and reliable social contexts to implement preventive health behaviours results in disproportionately high acute care use by people with lived experience of homelessness (Saab et al., 2016). This population frequently experiences longer hospital stays and a higher risk of unplanned readmission than the general population (Saab et al., 2016), as discharge planning is compromised by inadequate housing to return to and suboptimal structures to support proper follow up care (Kushel, 2016).

Substantial research demonstrates that people with lived experience of homelessness benefit from receiving tailored, patient‐centred care within interprofessional teams with an integrated approach to community and social services (Coltman et al., 2015; Hwang & Burns 2014; James et al., 2005). A systematic review on health interventions for marginalised and socially excluded populations identified a range of potentially effective interventions that have relevance for marginalised and excluded populations, but it was not specific to people with lived experience of homelessness (Luchenski et al., 2017). Additionally, numerous studies have looked at the effectiveness of patient‐centred care for people with lived experience of homelessness within community services and social services (Coltman et al., 2015; Hwang & Burns 2014; James et al., 2005). Our review aims to evaluate current evidence on the effectiveness and cost effectiveness of interventions that directly or indirectly improve the health of those with lived experience of homelessness.

2.2. Description of the condition

2.2.1. The intervention

We evaluated the effectiveness and cost‐effectiveness of interventions for people with lived experience of homelessness that aim to improve these people's health, service usage, and social outcomes. Prior to conducting this systematic review, to rank the priority topics and the needs that were the most important for this vulnerable population, we used a Delphi Consensus process (Keeney et al., 2010) to engage 76 people with lived experience of homelessness and 84 healthcare workers and researchers with professional experience in the field of homelessness (Shoemaker et al., 2020). A literature review and consensus process informed the final selection of the five categories of interventions to be included in this review (see Figure 1).

2.3. Description of the intervention

2.3.1. Housing interventions

PSH is long‐term housing in the community combined with the provision of individualised supportive services that are tailored to participants' needs and choices. PSH follows the principles of “Housing First,” whereby the ability to access housing is not contingent on sobriety/abstinence, or the ability to follow through with treatment plans (Benston, 2015). This approach runs in contrast to what has been the orthodoxy of “treatment‐first” approaches, whereby people experiencing homelessness are placed in emergency services and must address certain personal issues (e.g., addictions, mental health) prior to being deemed “ready” for housing. Rather, in “Housing First,” the priority is to provide an individual with permanent housing and the choice to access treatment and supports, such as ACT or ICM offered by a multidisciplinary team (Aubry, Nelson, et al., 2015; Somers et al., 2017).

2.3.2. Income assistance interventions

Income assistance is a fundamental intervention for preventing and addressing homelessness. It consists of interventions that directly increase an individual's available income or improve their access to basic living necessities. Some examples include government assistance (i.e., income‐supplement programme (Brownell et al., 2016), charity donation or panhandling (Poremski et al., 2015), provision of cheques, tax benefits or cash transfers. Cash transfers are a form of financial aid offered on either a conditional or an unconditional basis (Lagarde et al., 2009). Further examples include support finding and maintaining employment or offering information on income benefits or financial literacy/debt‐management counselling (Abbott & Hobby, 2000), and the provision of food, daycare and fuel or rent supplements (Gruber et al., 2000; Power et al., 2015; Whittle et al., 2015).

2.3.3. SCM and peer support

Case management entails a myriad of services in which people with multiple morbidities or issues are supported by case managers who assess, plan and facilitate the access to health and social services that are necessary for the person's plan of care and recovery (De Vet et al., 2013). There are many different types of care‐coordination models, and these vary according to approach and caseload. The more intensive models of case management will be examined under “Mental‐health interventions”. Here, we will examine two specific areas of less intensive‐care coordination.

Standard case management

SCM allows for the coordination of an array of social, health‐care and other services to help individuals maintain good health and strong social relationships. This is achieved by “including engagement, assessment, planning, linkage with resources, consultation with families, collaboration with psychiatrists, patient psycho‐education, and crisis intervention” (Kanter, 1989). The Case Manager or navigator's role is performed by either a clinician, nurse, community outreach worker or social worker with an average caseload of 35 clients (De Vet, 2013; Guarino, 2011). The target population for standard case‐management models are people who are experiencing homelessness or those who are vulnerably‐housed and have complex health concerns and are presenting to primary‐care practitioners and are often provided with this type of care as a time‐limited service (De Vet, 2013).

Peer support

Peer support includes the sharing of knowledge, experience, emotional, social or practical help by or with an individual who has experienced a similar background to the service user (Mead et al., 2001). Peer support workers may be termed differently in different settings, either as mentors, recovery coaches, or life coaches, and all offer emotional and social support to individuals who are newly homeless or on a treatment plan or path to recovery from substance use or homelessness (Barker & Maguire, 2017). Thus, due to their shared experiences and ability to establish a relationship built on trust, peers are uniquely positioned to assist persons experiencing homelessness due to their shared experiences and ability to establish a relationship built on trust (Barker & Maguire, 2017; Faulkner & Basset, 2012; Finlayson et al., 2016).

2.3.4. Mental health interventions

We assessed four evidence‐supported interventions that are relevant to serious mental illness, which is defined as conditions that substantially limit major life activities due to functional impairment (SAMHSA, 2016).

Assertive community treatment

ACT consists of a multidisciplinary group of healthcare workers in the community, that offers team‐based care to persons with high levels of needs. This team has 24‐h/day, 7‐days/week availability and provides services tailored to the needs and goals of each service user (Coldwell & Bender, 2007; De Vet et al., 2013). There is no time limit on the services provided, but transfer to lower intensity services is common after a period of stability (Homeless Hub, n.d.) Ten service users per case manager is the typical caseload, and services are offered in a natural setting, such as the workplace, home or social setting (De Vet et al., 2013).

Intensive case management

ICM is offered to persons with serious mental illness but who typically have moderate needs, such as fewer hospitalisations or less functional impairment, as well as for people experiencing addictions (Dieterich et al., 2017). ICM helps service users through the support of a case manager that brokers access to an array of services. The case manager accompanies the service user to meetings and can be available for up to 12 h/day, 7 days a week. Case managers for ICM often have caseloads of 15–20 service users each (De Vet, 2013).

Critical time intervention

CTIs are a form of time‐limited ICM, defined as a service that supports continuity of care for service users during times of transition; for example, from a shelter to independent housing or following discharge from the hospital. This service strengthens the person's network of support in the community (Silberman School of Social Work, 2017). It is administered by a CTI worker and is usually limited to a period of 6–9 months after institutional discharge or placement in housing. It comprises of three phases: Phase 1: Transition—Provide support and begin to connect the client to people and agencies that will assume the primary role of support; Phase 2: Tryout—Monitor and strengthen support network and client's skills; Phase 3: Transfer of care—Terminate CTI services with support network safely in place (Gaetz et al., 2013; Herman & Mandiberg, 2010).

Injectable antipsychotics

Injectable antipsychotics have a major role to play in the treatment for psychosis of patients living in precarious situations as these individuals often have a limited ability to follow through with oral medication treatment plans (Llorca et al., 2013). The clinical effectiveness of newer (aripiprazole, olanzapine, paliperidone and risperidone) and older antipsychotics (haloperidol, fluphenazine, flupenthixol) is similar (Castillo & Stroup, 2015), but their effectiveness among homeless and vulnerably housed persons is unknown.

2.3.5. Interventions for substance use

We assessed three interventions relating to substance‐use disorders (SUDs) that apply to people experiencing homelessness and those who are vulnerably housed.

Supervised consumption facilities (SCFs)

SCFs are legally sanctioned facilities where people who use substances can consume pre‐obtained substances under supervision (Drug Policy Alliance, n.d.). There exist various terminologies for these facilities, including supervised injection facilities (SIF), supervised consumption sites (SCS), medically supervised injection centres (MCIS), among others. Such facilities are frequently used as safe spaces for people experiencing homelessness and those who are vulnerably housed as well as substance users.

Managed alcohol programmes

A MAP provides shelter, medical assistance, social services and the provision of regulated alcohol to help residents cope with severe alcohol use disorder (Shepherds of Good Hope Foundation, n.d.). This programme is provided by professional staff and nurses.

Pharmacological interventions for opioid use disorder

The effectiveness and cost‐effectiveness of opioid therapy medications, including buprenorphine/naloxone, naloxone, naltrexone (British Columbia Centre on Substance Use, 2017) methadone, and injectable diacetylmorphine (heroin; Haasen et al., 2007), have been documented in general‐population studies. The literature demonstrates the effectiveness of naltrexone (Krupitsky et al., 2011), buprenorphine (with or without naloxone), and methadone (McKeganey et al., 2013), as well as injectable diacetylmorphine (heroin; Haasen et al., 2007) for treating opiate dependence, but the evidence specific to homeless populations is yet to be synthesised.

2.4. How the intervention might work

2.4.1. Housing interventions

Access to adequate housing is an end‐all objective of most homeless individuals. The unconditional provision of stable housing that is permanent in tenure, supportive in nature, and scattered across the rental market has been associated with a positive impact on the long‐term residential stability of homeless individuals (Aubry et al., 2016; Tsemberis et al., 2004). As well, synchronising the provision of permanent housing with supportive services is found to improve social functioning and quality of life (Aubry et al., 2016; Stergiopoulos et al., 2015).
Another model of providing PSH is to congregate accommodation units with supportive services in a single location that may be situated in a residential or commercial area and equipped with commonly used facilities. Such models were found to positively improve the housing and health‐related outcomes of homeless individuals, as per the scattered model (Somers et al., 2017).

2.4.2. Income assistance interventions

Financial hardships halt all efforts to end the cycle of homelessness (Institute of Medicine (US) Committee on Health Care for Homeless People, 1988). Evidence suggests that increasing income is an effective strategy to improve both access to health services and health status (Lagarde et al., 2009). This correlation may, very well, be the result of reduced financial stressors and increased ability to afford fundamental life necessities, such as housing, food, and medications (Richards et al., 2008). Moreover, it was found that providing information on income‐assistance resources has the potential to improve the physical and psychosocial health of disadvantaged populations (Adams et al., 2006).

2.4.3. SCM and peer support

Standard case management

Regardless of the heterogeneity and complexity of case‐management models, literature suggests that case management is associated with improved residential stability and substance‐use outcomes (De Vet, 2013). Patients who are provided with the services of a case manager are more likely to feel supported and guided in their quest to access and maintain fundamental health and social services (Conrad et al., 1998).

Peer support

The peer‐support model employs workers with shared life experiences to provide social support, advocacy, education, and role modelling to homeless individuals (Barker & Maguire, 2017). Among homeless populations, these elements have been found to improve quality of life and reduce problematic substance use (Barker & Maguire, 2017).

2.4.4. Mental health interventions

Assertive community treatment

Evidence suggests that individuals experiencing homelessness may benefit from receiving support in the form of ACT. This model of care is found to reduce days on the street or hospitalised, increase community functioning, and improve life satisfaction and mental‐health‐related outcomes for this vulnerable population (Coldwell & Bender, 2007; De Vet, 2013). The positive impact of this model of care could be attributed to the strength of the provider‐service‐user relationship and the sincere effort of the multidisciplinary team providing the care (Stuart, 2009).

Intensive case management

Literature is abundant with evidence that suggests the added benefits of providing ICM to disadvantaged populations. This model of care carries the potential to decrease hospitalisations, increase access to care and reduce economic hardships (De Vet, 2013; Dieterich et al., 2017).

Critical time intervention

The benefit of CTI lies in its ability to prevent the discontinuity of care during periods of transition. Providing CTI to homeless individuals was associated with increased residential stability, decreased mental‐health symptomatology and strengthened ties with services, family, and friends (De Vet, 2013; Jones et al., 2003).

Injectable antipsychotics

Injectable antipsychotics have been found to be effective for managing serious mental illnesses and reducing episodes of mental‐health emergencies (Llorca et al., 2013).

2.4.5. Interventions for substance use

Supervised consumption facilities

There is increasing evidence suggesting the benefit of SCFs in reducing precarious substance‐use‐related public behaviours and increasing access to and maintenance of treatment services (Kennedy et al., 2017; Wood et al., 2007). These facilities are believed to provide a safe environment for accessible services that provide nonjudgemental staff who help connect clients to health and social services as needed.

Managed alcohol programmes

MAPs follow a harm‐reduction approach by providing clients with shelter and regulated alcohol dispensing, accompanied by health and social support as needed (Podymow et al., 2006). The scarce literature on these interventions among homeless individuals suggest its effectiveness in reducing alcohol intake, hospitalisations and incarcerations (Podymow et al., 2006).

Pharmacological interventions for opioid use disorder

Opioid agonist therapy (OAT) such as methadone, buprenorphine/naloxone, and oral morphines have been associated with decreased mortality and morbidity (British Columbia Centre on Substance Use, 2017). As well, evidence suggests that opioid antagonist therapy such as naltrexone is successful in mitigating overdose‐related mortalities and incarceration rates (Roozen et al., 2006).

2.5. Why it is important to do this review

There has been a long‐standing social discourse on effectively tackling the negative consequences of the urban homelessness that has plagued communities internationally. Persons with lived experience of homelessness face higher rates of infectious and chronic disease and disproportionately poorer outcomes, with reduced rates of access to effective quality care. In order to better understand the needs and resources available for people with lived experience of homelessness, policymakers, practitioners, and allied health professionals need high‐quality systematic reviews and knowledge‐translation strategies on interventions that are specific to this population. Our review aims to provide a comprehensive overview of the benefits and cost‐effectiveness of interventions designed to indirectly or directly improve the health and well‐being of persons with lived experience with homelessness. We present updated perspectives on the effect of PSH, income assistance, SCM/peer support and mental‐health and substance‐use interventions, on the housing stability, mental health, quality of life, hospitalisations, earned income and employment statuses of persons with lived homelessness. This review is part of a series of other publications that inform a national practice guide on homelessness (Pottie et al., 2020), and serves to provide best‐practice updates to health‐care professionals and policymakers and guide the future care of persons with lived experiences of homelessness.

3. OBJECTIVES

The objective of this systematic review is to identify, appraise, and synthesise the best available evidence on the effectiveness and cost‐effectiveness of interventions to improve the health and social outcomes of people experiencing homelessness and those who are vulnerably housed. Our outcomes of interest include housing stability, mental health, quality of life, substance use, hospitalisations, employment and income. The following research questions were developed to guide the formation of the systematic review:

1.
What is the effectiveness of PSH on the health and social outcomes of people experiencing homelessness and those who are vulnerably housed?
2.
What is the effectiveness of income assistance on the health and social outcomes of people experiencing homelessness and those who are vulnerably housed?
3.
What is the effectiveness of SCM and/or peer support on the health and social outcomes of people experiencing homelessness and those who are vulnerably housed?
4.
What is the effectiveness of mental‐health interventions (ACT, ICM, CTI and injectable antipsychotics) on the health and social outcomes of people experiencing homelessness and those who are vulnerably housed?
5.
What is the effectiveness of interventions for substance use (SCFs, MAPs and pharmacological interventions for opioid use disorder) on the health and social outcomes of people experiencing homelessness and those who are vulnerably housed?
6.
What are the costs and cost‐effectiveness of the aforementioned interventions for people experiencing homelessness and those who are vulnerably housed?

4. METHODS

4.1. Criteria for considering studies for this review

4.1.1. Types of studies

The protocol was registered with the Campbell Collaboration (Pottie et al., 2019) and reported according to the preferred reporting items for systematic reviews and meta‐analyses for protocols (PRISMA‐P; Moher et al., 2015). The results of the review are reported using the PRISMA reporting guidelines (Moher et al., 2009).

We included studies as recommended by the Cochrane Effective Practice and Organisation of Care (EPOC) Cochrane group for reviews of effectiveness (Effective Practice and Organisation of Care (EPOC)‐Cochrane, 2015). We included randomised control trials (RCTs), non‐RCTs, controlled before‐after studies, interrupted time‐series studies, and repeated‐measures studies. We considered studies published in both peer‐reviewed journals and grey literature.

4.1.2. Types of participants

We included populations experiencing homelessness, defined as those who lack stable, permanent, appropriate housing, or who may be without immediate prospects, means, and ability to acquire it. Such physical living situations can include emergency shelters or provisional accommodations (Canadian Definition of Homelessness, 2017). Studies must have reported whether participants were experiencing homelessness in order to be included in this review. We included studies that included a subset of the sample experiencing homelessness as long as 50% of the participants were homeless. We included studies that were among individuals or families and we did not restrict our inclusion criteria by age, sex or gender. We excluded studies that were specific to indigenous populations experiencing homelessness, as this line of inquiry is being pursued by an indigenous‐specific research team (Thistle & Smylie, 2020). We excluded all other populations.

4.1.3. Types of interventions

We included the following interventions, as outlined in Section 1.1.2: PSH: income assistance, SCM, ICM, ACT, CTI, peer support, SCFs, MAPs, injectable antipsychotics, and OAT. We included studies that had multicomponent interventions as long as one of the interventions applied to those mentioned above.

All of the included interventions were either compared to an inactive control (i.e., a placebo, no treatment, standard care) or to an active control intervention (alternative or variant of the intervention; Effective Practice and Organisation of Care (EPOC)‐Cochrane, 2015). In the case of an interrupted time series, the control must have been a historical control with three data points. If a study with more than two intervention arms was included, then we only included the intervention and control arms that met the eligibility criteria.

4.1.4. Types of outcome measures

Studies were included in this review if they reported the use of validated measures and reported at least one of the following outcomes.

Primary outcomes

1.
Housing stability: Any measures assessing participants housing status, such as the number of days in stable housing, number of days homeless (on the street or in shelters), number of participants in stable housing, and number of participants homeless (on the street or in shelters). Typical tools to measure housing stability include the Residential Timeline Followback Inventory (RTLFB) (Tsemberis et al., 2007).

Secondary outcomes

2.
Mental health: any measures assessing psychological status and wellbeing, including but not limited to, psychological distress, self‐reported mental health status, or mental illness symptoms. Typical tools to measure mental health include the Colorado Symptom Index (CSI) (Boothroyd & Chen, 2008), and the self reported mental status SF‐12 (Nelson et al., 2012).
3.
Quality of life: which include any assessment of well‐being, life and personal satisfaction, quality of social relationships, and any specific physical or mental quality of life measures. Typical tools to measure quality of life include the Lehman Quality of Life Interview (Lehman et al., 1996), and the EuroQoL 5‐D scale (Lamers et al., 2006).
4.
Hospitalisation: any measures of participants' use of hospital and emergency services, such as the number of days hospitalised or number of visits to the emergency department. Such measures can be assessed through a standard questionnaire asking participants for their service use, or through accessing the hospital data sets.
5.
Substance use: As measured by the number of days using alcohol or substance, the rate and frequency of using alcohol or substances, number of days of abstinence from alcohol or substances or physical and mental consequences of using alcohol or substances. Typical tools to measure substance use outcomes include the Global Appraisal of Individual Needs‐Short Screener of Substance Use Problems GAIN‐SS (Dennis et al., 2006).
6.
Income: Any measures of money that participants acquire from different resources including social assistance, disability benefits, donations, and part or full time employment. Such measures can be assessed through a standard questionnaire asking participants for their weekly, monthly or annual income from different sources.
7.
Employment: Any measures of employment that participants partake during the study period, including but not limited to, number of days of paid employment, number of employed or unemployed participants, hourly wage, and employment tenure. Such measures can be assessed through a standard questionnaire asking participants about their employment rates during the study period.
8.
Economic outcome: any measures of cost, cost benefit, cost utility or incremental cost‐effectiveness ratio. Such measures can be obtained from administrative databases and cost reports. An example of how an individualised programme cost could be calculated is dividing the sum of all on‐site operation and services costs (maintenance, utilities, insurance, etc.) by the capacity of the project (Larimer, 2009).

Duration of follow‐up

All durations of follow‐up were included. We collected data at each available time‐point.

Types of settings

We included studies where the intervention took place in any setting where the primary care of people experiencing homelessness takes place. Primary care is known as the “entry point to the larger health care system” (Tarlier, 2007) and can be provided by professionals from many disciplines, such as family physicians, psychiatrists, social workers, emergency physicians, and so forth. We also included community‐based interventions provided in social‐service or shelter/supervised consumption locations, private or nonprivate clinics, hospital emergency rooms, outreach care, street patrols, mobile care units, and so forth.

We included studies that occurred in high‐income countries and excluded studies that occurred in low‐ and middle‐income countries (World Bank, 2019).

4.2. Search methods for identification of studies

We searched for relevant literature in consultation with an information specialist (librarian).

4.2.1. Electronic searches

We searched the following bibliographic databases from database inception to February 10, 2020:

Ovid MEDLINE(R) and Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Daily and Versions(R) (1946 to February 7, 2020)
EBM Reviews—Cochrane Central Register of Controlled Trials (January 2020)
EBM Reviews—Cochrane Database of Systematic Reviews (2005 to February 4, 2020)
EBM Reviews—Database of Abstracts of Reviews of Effects (1st Quarter 2016)
EBM Reviews—Health Technology Assessment (4th Quarter 2016),
EBM Reviews—NHS Economic Evaluation Database (1st Quarter 2016)
Embase (1974 to February 7, 2020)
EBSCO CINAHL (–2020)
EBSCO PsycINFO (–2020)
Epistemonikos (–2020)

We used a combination of subject headings and keywords including “homeless”, “marginalized” and “shelter”. Full strategies for each database can be found in Appendix A.

4.2.2. Searching other resources

The reference lists of all articles selected for full‐text review were manually searched for relevant citations. These were cross‐referenced against our original search results and any additional potentially relevant citations were screened. Further, we consulted content experts for any publications or resources that might enrich our findings and we screened their suggestions against our inclusion criteria.

4.3. Data collection and analysis

We collected and analysed data according to our protocol (Pottie et al., 2019).

4.3.1. Selection of studies

Teams of two reviewers independently screened titles and abstracts in duplicate. We pilot tested the screening criteria at both the title‐and‐abstract‐screening stage and the full‐text stage. We used the PRISMA flow diagram to report the eligibility of studies. We retrieved the full text of all of the studies that passed this first‐level screening. The full‐text reviews were also done in duplicate by two reviewers, and agreement was reached by consensus. Disagreements were resolved by consultation with a third reviewer.

4.3.2. Data extraction and management

We developed a standardised extraction sheet for each topic variable and created a table of characteristics to describe a summary of findings from our included studies. The data extraction sheet was piloted by two independent reviewers. We collected and utilised all relevant numerical data (SDs, effects estimates, confidence intervals [CIs], test statistics, p values, etc.). Teams of two reviewers extracted data in duplicate and independently. The reviewers compared their results and resolved disagreements by discussion or with help from a third reviewer.

4.3.3. Assessment of risk of bias in included studies

We assessed risk of bias according to guidance for Cochrane Effective Practice and Organisation of Care (EPOC) reviews (Effective Practice and Organisation of Care (EPOC)‐Cochrane, 2015). Nine standard criteria are suggested for all randomised trials, nonrandomised trials and controlled before‐after studies, including random‐sequence generation, allocation concealment, baseline‐outcome measurements, baseline characteristics, incomplete‐outcome data, knowledge of the allocated interventions adequately prevented during the study, protection against contamination, selective outcome reporting, and other biases.

Risk of bias was assessed by teams of two review authors, in duplicate. Discrepancies were resolved through discussion or by a third reviewer.

It is important to highlight that we did not subject any of the identified economic evaluations to critical appraisal, in accordance with guidance from the Cochrane Handbook (Shemilt et al., 2019).

4.3.4. Measures of treatment effect

Whenever possible, we performed statistical analyses using RevMan 5. For dichotomous data, we used odds ratios (OR) and risk ratios (RR) with 95% CIs. For continuous data, we used the mean difference (MD) with 95% CI, if the outcomes were measured in the same way between trials. We used the standardised mean difference (SMD) with 95% CI to combine the trials that measured the same outcome but used different methods of measurement.

4.3.5. Unit of analysis issues

The results from some studies were reported in multiple publications. Therefore, to prevent the double counting of data, individual records were screened to identify unique studies and were evaluated for potential overlap by comparing the study design, enrolment and data‐collection dates, authors and their associated affiliations, and the reported selection and eligibility criteria. When reviewing multiple publications, we included only unique data from each study. Several studies included outcome data for multiple time points. Comparisons were therefore carried out separately for periods of 6 months and less (short term), 6–18 months (midterm), and 18 months or more (long‐term). If multiple measures of the same outcome were reported, we prioritised outcomes measured using validated scales for meta‐analyses and reported all available outcome data narratively.

4.3.6. Dealing with missing data

We will contact authors once for any missing data. We will use any supplementary data provided by authors in our analysis, or report findings as extracted otherwise.

4.3.7. Assessment of heterogeneity

We assessed heterogeneity among studies in two ways. First, we assessed clinical heterogeneity: heterogeneity in population, interventions or outcomes. We used I ² statistics as a guide to assess heterogeneity along with a visual inspection of forest plots.

4.3.8. Assessment of reporting biases

There were very few outcomes that provided enough data for a meta‐analysis; therefore we could not assess for reporting bias. For future updates, funnel plots would be used if there are 10 or more studies in a meta analysis for one outcome and an investigation would be conducted for reporting biases, for example, publication bias.

4.3.9. Data synthesis

We aimed to conduct a separate meta‐analysis for each outcome and intervention. We assessed clinical heterogeneity by considering the study population, intervention, comparison, outcome measure, and timing of outcome assessment. The assessments were made in consultation with members of the research team with clinical and statistical expertise. We pooled data from studies we judged to be clinically homogeneous. If more than one study provided usable data in any single comparison, we performed a meta‐analysis. We standardised all the reported effect sizes as RRs for the dichotomous outcomes and MDs or SMDs for the continuous outcomes. For the majority of findings, characteristics of studies (such as study designs, intervention types, or outcomes) were too diverse to yield a meaningful summary estimate of effect. Similarly, we deemed forest plots of single studies to be of limited value to the review. When heterogeneity precluded a meta analysis, we synthesised our findings narratively as recommended by the synthesis without meta‐analysis (SWiM) reporting guideline (Campbell et al., 2020).

4.3.10. Subgroup analysis and investigation of heterogeneity

Based on the availability of the data, we had planned to conduct subgroup analyses for the following subgroups: women, youth, and people with disabilities. However, since very few studies were included in each comparison within the review, we could not conduct any of the aforementioned subgroup analyses.

4.3.11. Sensitivity analysis

We had planned to conduct sensitivity analyses; however, since very few studies for each intervention were included in the review, we could not conduct any sensitivity analyses.

4.3.12. Summary of findings and assessment of the certainty of the evidence

We assessed the certainty of evidence for housing‐stability outcomes by using the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) approach (Balshem et al., 2011). Housing‐stability outcomes were selected as critical patient‐important outcomes by our review team, in consultation with content experts and people with lived experience of homelessness. GRADE rates certainty of evidence are as follows:

Certainty of evidence	Definition
High	There is a lot of confidence that the true effect lies close to that of the estimated effect
Moderate	There is moderate confidence in the estimated effect: The true effect is likely to be close to the estimated effect, but there is a possibility that it is substantially different
Low	There is limited effect on the estimated effect: The true effect might be substantially different from the estimated effect
Very low	There is very little confidence in the estimated effect: The true effect is likely to be substantially different from the estimated effect

Certainty assessment							No of patients		Effect
No of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	HF + ACT/ICM	UC	Relative (95% CI)	Absolute (95% CI)	Certainty	Importance
Number of participants in stable housing long term (Aubry et al., 2016 ⁽¹³⁾; HF + ACT, Stefancic, 2007⁽⁶⁴⁾; HF + ACT) (follow up: +20 months)
2	Randomised trials	Serious^*	Not serious	Not serious	Not serious	None	376/578	152/388	OR 3.58 (2.36–5.43)	306 more per 1000 (from 211 more to 386 more)	⊕⊕⊕◯ MODERATE	CRITICAL
Percentage of participants in stable housing long term (Aubry et al., 2016; ⁽¹³⁾ HF+ACT) (follow up: 24 months)
1	Randomised trial	Serious^a	Not serious	Not serious	Not serious	None	273/369 (74%, CI 69%‐78%)	138/337 (41%, CI 35%‐46%)	OR 4.10 (2.98–5.63)	330 more per 1000 (from 264 more to 387 more)	⊕⊕⊕◯ MODERATE	CRITICAL
1	Randomised trial	Serious^a	Not serious	Not serious	Not serious	None	273/369 (74%, CI 69%‐78%)	138/337 (41%, CI 35%‐46%)	RR 1.81 (1.58–2.08)	330 more per 1000 (from 264 more to 387 more)	⊕⊕⊕◯ MODERATE	CRITICAL
Total number of days housed long term (Aubry et al., 2016; ⁽¹³⁾ HF+ACT, Stergiopoulos et al., 2015; ⁽⁵⁶⁾ HF+ICM) (follow up: 24 months)
1	Randomised trial	Serious^a	Not serious	Not serious	Not serious	None	411 (280.74 days, SD278.92)	369 (115.33 days, SD191.43)	‐	AAD 161.8 more days (82.5 more to 241.1 more)	⊕⊕⊕◯ MODERATE	CRITICAL
1	Randomised trial	Serious^a	Not serious	Not serious	Not serious	None	689	509	‐	SMD 1.38 more days (1.03 more to 1.73 more days)	⊕⊕⊕◯ MODERATE	CRITICAL
Percentage of days stably housed over 24 months (Aubry et al., 2016; ⁽¹³⁾ HF+ACT, Stergiopoulos et al., 2015; ⁽⁵⁶⁾ HF+ICM)
1	Randomised trial	Serious^a	Not serious	Not serious	Not serious	None	469	481	At Home/Chez Soi: High needs participants (HF+ACT): Adjusted absolute difference [AAD] = 42%; 95% CI [38%‐45%; p<0.01] (Aubry et al., 2016).		⊕⊕⊕◯ MODERATE	CRITICAL
1	Randomised trial	Serious^a	Not serious	Not serious	Not serious	None	689	509	Moderate needs participants (HF+ICM): Adjusted mean difference:		⊕⊕⊕◯ MODERATE	CRITICAL
									Site A, 33.0% [95% CI, 26.2–39.8%];
									Site B, 49.5% [95% CI, 41.1–58.0%];
									Site C, 35.6% [95% CI, 29.4–41.8%];
									Site D, 45.3% [95% CI, 38.2–52.5%]; (Stergiopoulos et al., 2015). (C).

Certainty assessment							No.of patients		Effect		Certainty	Importance
No. of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Intervention	Comparison	Relative (95% CI)	Absolute (95% CI)	Certainty	Importance
Number of homeless episodes (Compensated Work Therapy (CWT) versus control services; Kashner, 2002 ⁽²³⁾)
1	Randomised trials	Serious^a	Not serious	Not serious	Serious^b	None^c	Not reported	Not reported	OR 0.10 (0.1–0.3)	‐	⊕⊕◯◯ LOW	CRITICAL
Number of days spent homeless (HUD Section 8 housing vouchers paired with ICM versus treatment as usual; Rosenheck 2003 ⁽¹²⁾)
1	Randomised trials	Serious^d	Not serious	Not serious	Serious^e	None	182 (13.5 days)	188 (20.45 days)	‐	36.2% fewer days (p < .001)	⊕⊕◯◯ LOW	CRITICAL
Number of days stably housed (HUD Section 8 housing vouchers paired with ICM versus treatment as usual; Rosenheck, 2003 ⁽¹²⁾)
1	Randomised trials	Serious^d	Not serious	Not serious	Serious^e	None	182 (59.39 days)	188 (47.60 days)	‐	25% more days (p < .001)	⊕⊕◯◯ LOW	CRITICAL

Certainty assessment							No. of patients		Effects
No. of patients	No. of studies	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Case Management	Comparison	Relative (95% CI)	Absolute (95% CI)	Certainty	Importance
Number of days homeless (Cox et al., 1998, Grace & Gill, 2014, Toro et al., 1997) (follow up: 13+ months)
3	Randomised trials	Very serious^a	Not serious	Not serious	Not serious	None	358	308		SMD 0.22 fewer (0.4 fewer to 0.03 fewer)	⊕⊕◯◯ LOW	CRITICAL
Proportion of participants housed (Korr & Joseph, 1996) (follow‐up: 6 months)
1	Randomised trials	Serious^b	Not serious	Not serious	Serious^c	None	36/48 (75%)	15/44 (34.1%)	OR 5.80 (2.35 to 14.31)	409 more per 1000 (from 208 more to 540 more)	⊕⊕◯◯ LOW	CRITICAL
Time spent on the street (Shern et al., 2000) (follow‐up: 24 months)
1	Randomised trials	Serious^d	Not serious	Not serious	Serious^e	None	91	77	‐	MD 26.71 fewer (39.21 fewer to 14.2 fewer)	⊕⊕◯◯ LOW	CRITICAL
Time spent in shelters (Shern et al., 2000) (Follow‐up: 24 months)
1	Randomised trials	Serious^d	Not serious	Not serious	Serious^e	None	91	77	‐	MD 20.29 more (13.38 more to 27.19 more)	⊕⊕◯◯ LOW	CRITICAL
Time spent in community housing (Shern et al., 2000) (follow‐up: 24 months)
1	Randomised trials	Serious^d	Not serious	Not serious	Serious^e	None	91	77	‐	MD 11.07 more (1.52 more to 20.61 more)	⊕⊕◯◯ LOW	CRITICAL

Methods	Designs: Cost analysis and randomised controlled trial
	Setting: Toronto, Montreal, Moncton, Winnipeg, Vancouver, Canada
	Follow up: up to 24 months
	Recruitment: Referred by health and social service agencies
	Randomization: Participants were randomized by a central data collection system that used an adaptive randomization algorithm
	Allocation: 1:1 allocation ratio
	Blinding: Not possible
	Timing of outcome assessment: At baseline and every six months for 21 or 24 months
	Outcome assessor: The investigators conducted in person interviews
Participants	Population: Homeless individuals or individuals living in a single room occupancy, rooming house, or hotel who have had two or more episodes of being absolutely homeless with serious mental illness and high support needs.
Participants	Sample size: Total n = 950, Intervention n = 469, TAU n = 481
Interventions	Intervention:
	Housing First provides immediate access to independent housing and mental health supports; rent supplements were provided that ensured housing costs did not exceed 30% of participant's income; housing coordinators provided assistance to find and move into housing; support services were provided via assertive community treatment; study participants agreed to observe the terms of their lease and be available for a weekly visit by programme staff.
	Comparator:
	Treatment as usual (TAU): Participants had access to the already existing interventions and programs available in their community including any housing and community support other than the Housing First program
Outcomes	Housing stability, mental health, quality of life, substance use, hospital admission, employment, income
Outcomes	Cost or cost‐effectiveness: The Housing First programme led to a reduction in the mean cost to $21,367 per person per year; this cost offset was associated with office visits, hospital admissions, emergency shelter visits, home visits, and incarceration; the savings gained by Housing First did not fully offset its cost.
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomization was performed via adaptive randomization procedures. Adaptive randomization can ensure better balance between groups in small and moderate sized studies than strict randomization, by continually adjusting the probability of assignment to either group, depending on existing group assignment.
Allocation concealment (selection bias)	Low risk	The allocation algorithm was concealed from both the participants and the research staff.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	High risk	Several aspects of the study prohibited blinding, including the nature of the administered questionnaires (detailed housing history and service use), location of participant interviews (some participants elected to be interviewed at their place of residence)
Incomplete outcome data (attrition bias)	Low risk	No evidence of incomplete outcome data provided
Selective reporting (reporting bias)	Low risk	No evidence of reporting outcomes selectively is provided
Other bias	Low risk	No other potential risk of biases

Methods	Design: 3‐arm randomized control trial
	Setting: Philadelphia, PA, USA
	Follow up: 15 months
	Recruitment: Participants were recruited by research staff at three county assistance offices
	Randomization: Single‐blind randomization was performed by assigning participants to 150 random numbers previously assigned.
	Allocation: Not reported
	Blinding: Single blinding was used so as program facilitators and instructors were instructed not to discuss allocation
	Timing of outcome assessment: At baseline, and 9,12, and 15 months
	Outcome assessor: Self‐administered survey
Participants	Population: Low‐income families with a child; 65% of whom had housing insecurity
Participants	Sample size: Total n = 103, Partial intervention n = 35, Full intervention n = 37, Control n = 31
Interventions	Intervention:
	Partial intervention included assistance opening a credit union savings account, with participants' savings matched throughout trial and 28 weeks of 3 h financial empowerment education classes, which focused on developing internal and external resources to support self‐sufficiency; education included saving for education, retirement, housing, entrepreneurial activities, improving credit and reducing debt; the full intervention was the same as the partial intervention plus a simultaneous 28 weeks of trauma‐informed peer support.
	Comparator:
	Control group: standard TANF programming consisting of 20 h per week of scheduled supervised job training and job search activities.
Outcomes	Housing stability, mental health, employment, income
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Through single‐blind randomization participants were assigned into each group.
Allocation concealment (selection bias)	Unclear risk	No description of allocation concealment was provided.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	High risk	“All data are self‐reported, and therefore may be prone to bias, where people may minimize symptoms or other concerns and exaggerate income and employment experiences.
Incomplete outcome data (attrition bias)	Low risk	No evidence of incomplete outcome reporting or attrition bias was detected.
Selective reporting (reporting bias)	Low risk	No evidence of reporting outcomes selectively was reported.
Other bias	Unclear risk	Class participation rates were limited and potentially hindered our ability to assess full treatment effects.

Methods	Design: Randomized control trial
	Setting: Denver, Colorado, United States
	Follow up: 10 months
Participants	Population: Homeless individuals 18 years or older with alcohol or other substance abuse problems
Participants	Sample size: Total sample n = 323, Intervention n = 163, control n = 160
Interventions	Intervention:
	Intensive Case Management (ICM) with access to services available at Arapahoe House (detoxification facility). ICM included proactive outreach, client identification and assessment, development of an individually‐tailored and comprehensive service plan for each client, establishment of linkages between service systems and clients such that services were matched to client needs, continuity of service utilization monitoring, and assertive advocacy with community agencies on behalf of clients. Case managers worked in pairs (dyads) with a caseload averaging 15 clients per dyad (and never exceeding 17 clients per dyad) and with a consequent high intensity of client contact.
	Comparator:
	Access to the full range of services offered by Arapahoe House. Services include a comprehensive array of substance abuse treatment and rehabilitation services, including detoxification, residential and outpatient services, substance abuse counseling, literacy and vocational assessment, and job training and placement.
Outcomes	Housing stability, Mental health, Quality of life, Substance use, Employment
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The random assignment of clients to the experimental (ICM) and control groups in the present study was implemented with a computer program specifically designed to protect the integrity of the assignment process
Allocation concealment (selection bias)	Unclear risk	No description of allocation concealment provided
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	Unclear risk	No description of blinding of outcome assessment provided.
Incomplete outcome data (attrition bias)	Low risk	There was no evidence that differential attrition introduced meaningful nonequivalence between the groups.
Selective reporting (reporting bias)	Low risk	No evidence that outcomes were selectively reported.
Other bias	Low risk	No other biases identified.

Methods	Design: 3‐arm Randomized control trial
	Setting: The Westside area, Los Angeles county, United States
	Follow up: 9 months
Participants	Population: Homeless adults with both a serious mental illness and substance dependence
Participants	Sample size: Total sample size n=276, Experimental arm1 n = 144, Experimental arm2 n = 67, Control n = 65
Interventions	Experimental arm 1: A social model residential program providing integrated mental health and substance abuse treatment
	Experimental arm 2: A community‐based nonresidential program using (1) curriculum‐based groups focused on substance abuse and mental health education and rehabilitation; (2) 12‐step programs including participation in community‐based AA or NA meetings (3) discussion of issues of importance to the clients (4) individual counseling and case‐management (5) psychiatric consultation and ongoing medications management (6) general community activities
	Comparator: Participants randomized to the control group did not receive interventions but were allowed free access to community services.
Outcomes	Housing stability, mental health, substance use
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Probability of assignment to the nonresidential group was set at twice that of the other groups. Random assignment was made within two blocking variables, gender and primary type of mental disorder.
Allocation concealment (selection bias)	Unclear risk	Allocation concealment not reported
Blinding of participants and personnel (performance bias)	High risk	Participants and personnel could not be blinded to the intervention.
Blinding of outcome assessment (detection bias)	Unclear risk	Blinding of outcome assessment not reported.
Incomplete outcome data (attrition bias)	High risk	Rates of completed follow‐up were 79% for the 3‐month follow‐up, 76% for the 6‐month follow up, 70% for the 9‐month follow‐up, and 58% completed all three follow ups. At the 9‐month follow‐up the completion rate among those assigned to the control group (57%) was significantly lower than among those assigned to nonresidential treatment (76%).
Selective reporting (reporting bias)	Low risk	No evidence of selective outcome reporting.
Other bias	Low risk	None Identified.

Methods	Design: Randomized control trial
	Setting: Seattle, Washington, United States
	Follow up: 3 months
Participants	Population: Homeless adolescents 13‐21 years old
Participants	Sample size: Total sample size n = 115, Intervention n = 55, control n = 60
Interventions	Intervention:
	Intensive case management: Development of treatment plans, linkage to appropriate services, monitor and track clients, advocating for basic entitlements, 24 h crisis service. Maximum caseload of 12
	Comparator:
	Those randomized to the control group received regular case management with caseload 18‐30
Outcomes	Mental health, quality of life, substance use
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“Random assignment was accomplished by preparing a stack of sequentially numbered envelopes, and placing in each a card with a matching number and group assignment”.
Allocation concealment (selection bias)	Unclear risk	No description of allocation concealment was provided.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	Unclear risk	No description of blinding of outcome assessment was provided.
Incomplete outcome data (attrition bias)	Unclear risk	Results are based on the data collected from the first 115 adolescents who completed the first quarterly follow‐up assessment of the 229 total recruited ‐ no significant difference from the overall pool of participants on any of the measures of psychological and social adjustment previously described (p<.10). Also, the exact number in each analysis varies from 104 to 115 due to missing data on some measures, there's no further explanation.
Selective reporting (reporting bias)	Low risk	No evidence of reporting outcomes selectively was detected.
Other bias	Low risk	No other potential biases were detected.

Methods	Design: A cost analysis based on administrative databases (before‐ and‐after study)
Participants	Population: Homeless people with mental illness who were formerly homeless in the state of Maine (USA)
Interventions	Intervention:
	Before and after providing permanent supported housing
	Comparator:
	N/A
Outcomes	Cost or cost‐effectiveness: Supportive housing was associated with a 57% reduction in mental health care costs, a 97% reduction in emergency shelter costs, a 14% reduction in jail costs, and 32% savings in ambulance costs; total savings to the system was US$584,907 after 12 months in housing, representing a mean saving of $2,182 per participant in the study
Notes

Methods	Design: Quasi‐experimental trial
	Setting: Ottawa, Canada
	Follow up: 24 months
	Recruitment: Invitation from participants' case manager
	Randomization: No randomization was performed
	Allocation: Not mentioned
	Blinding: Not mentioned
	Timing of outcome assessment: At baseline and every 6 months for 24 months
	Outcome assessor: Not mentioned
Participants	Population: Homeless individuals or people at risk of homelessness with problematic substance use
Participants	Sample size: Total n=178, Intervention n=89, TAU n=89
Interventions	Intervention:
	A partnership between a community mental health agency and a programme located in a community health centre; each participant received a rent supplement and paid a maximum of 30% of their income towards rent; housing comprised private market rental units of participant's choice; participants also had access to opioid agonist treatment and substance use treatment.
	Comparator:
	Treatment as usual (TAU) including all social and health services available in the community other than the Housing First program. The services were scattered across a service rich city and included supportive housing, mental health, and substance use services available to people who are homeless as well as services that can be accessed while people are in a shelter.
Outcomes	Housing stability, mental health, quality of life, substance use
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Description on random sequence generation was not provided.
Allocation concealment (selection bias)	High risk	Selection bias: Recruitment for the comparison group occurred after all Housing First clients participating had been admitted to the program. Since the program selected adults with the greatest health and substance needs, this sampling strategy may have resulted in Housing First clients having more needs and more severe health and substance use problems than the comparison participants.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	Unclear risk	No description of blinding of outcome assessment was provided
Incomplete outcome data (attrition bias)	Low risk	No evidence of incomplete outcome data.
Selective reporting (reporting bias)	Low risk	No evidence that outcomes were reported selectively.
Other bias	Low risk	No other potential biases detected.

Methods	Design: Cost‐effectiveness analysis conducted alongside an RCT, societal perspective. Costs were reported in 1995 US$
Participants	Population: Persons with schizophrenia, schizoaffective disorder, or bipolar disorder and a concurrent substance use disorder
Interventions	Intervention:
	Specialized treatment for dual disorders in an ACT team
	Comparator:
	SCM program with targeted substance abuse services
Outcomes	Cost/Cost‐effectiveness:
	ACT more effective and less costly than SCM in both substance abuse and quality of life comparisons.
	The relationship between the treatment received and benefit to cost ratio was curvilinear. During earlier periods SCM produced better outcomes per $10,000 invested than did ACT. During the final year of the study, ACT produced substantially better outcomes per $10,000 than SCM.
Notes

Methods	Design: 3‐arm randomized control trial
	Setting: Portland, Oregon, United States
	Follow up: 24 months
Participants	Population: Clients with serious and persistent mental illness who were at risk of being homeless
Participants	Sample size: Total sample n = 163, Consumer ACT n = 57, Non‐consumer ACT n = 57, Usual care n = 49
Interventions	Intervention:
	The ACT programs: Each team consisted of four full‐time and one part‐time case managers, one of whom was the team leader. Both teams shared a psychiatrist, a nurse practitioner, and a clinical director. The psychiatrist conducted initial psychiatric assessments, prescribed and monitored medication, and participated in treatment planning. The clinical director provided consultation to the two teams and handled administrative tasks. The Consumer ACT staff held a bachelor's degree, had more average experience in the mental health field, had more mental disorders themselves than the non‐consumer team. The average caseload for the ACT team during the study period was 4.6 clients per case manager. Non‐consumer ACT: Staff reported no diagnosable mental illness. The majority of the non‐consumer‐ACT staff held a master's degree. The average caseload for the ACT team during the study period was 5.4 clients per case manager.
	Comparator:
	The "Usual Care" control condition:
	Most subjects assigned to usual care were served by one of four community mental health centers (CMHCs) and a number of smaller, more specialized agencies in the Portland metro area usual care agencies set up special, intensive, community‐based programs to better serve these high‐need clients in the community. The intensity of services in these programs rivaled the intensity of the ACT teams. Average caseload size for usual care overall was 26.9
Outcomes	Hospitalization
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No description of random sequence generation was provided.
Allocation concealment (selection bias)	Unclear risk	No description of allocation concealment was provided.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	Unclear risk	No description of blinding of outcome assessment was provided.
Incomplete outcome data (attrition bias)	Unclear risk	No description of reporting outcome data incompletely or attrition bias was provided.
Selective reporting (reporting bias)	Low risk	No evidence of reporting outcomes selectively was detected.
Other bias	Low risk	No other potential biases detected.

Methods	Design: Quasi experimental trial
	Setting: Pinellas county, Florida, United States
	Follow up: 12 months
Participants	Population: Homeless adults with mental illness
Participants	Sample size: The total sample size n = 152, Comprehensive housing program n = 83, Case management n = 69
Interventions	Intervention:
	Comprehensive housing programs: Developed a program specifically to prevent and reduce homelessness in this population. The program features guaranteed access to housing and housing support services, case management, and priority access to everything from medication management to vocational services. Project Return in Tampa, Florida, also provides comprehensive housing services to homeless persons with severe mental illness, including guaranteed access to housing, housing support services, and case management.
	Comparator:
	Case management only: developed a homeless outreach and support team (HOST) to provide short‐term case management services for homeless individuals with severe mental illness. The activities of this blended case management program (11) include active outreach and engagement, some on site counseling, medication and medication management, assistance with obtaining housing, and linkages to other psychosocial services.
Outcomes	Housing stability, mental health
Notes

PERMALINK

A comprehensive review of prioritised interventions to improve the health and wellbeing of persons with lived experience of homelessness

Aliza Moledina

Olivia Magwood

Eric Agbata

Jui‐Hsia Hung

Ammar Saad

Kednapa Thavorn

Kevin Pottie

Abstract

Background

Objectives

Search Methods

Selection Criteria

Data Collection and Analysis

Main Results

Authors' Conclusions

1. PLAIN LANGUAGE SUMMARY

1.1. Housing, income assistance, and case management improve housing outcomes for persons with lived experience of homelessness

1.1.1. What is this review about?

1.2. What studies are included?

1.3. What are the main findings of this review?

1.4. What do the findings of this review mean?

1.5. What are the implications for research and policy?

1.6. How up‐to‐date is this review?

2. BACKGROUND

2.1. The problem, condition or issue

2.2. Description of the condition

2.2.1. The intervention

Figure 1.

2.3. Description of the intervention

2.3.1. Housing interventions

2.3.2. Income assistance interventions

2.3.3. SCM and peer support

Standard case management

Peer support

2.3.4. Mental health interventions

Assertive community treatment

Intensive case management

Critical time intervention

Injectable antipsychotics

2.3.5. Interventions for substance use

Supervised consumption facilities (SCFs)

Managed alcohol programmes

Pharmacological interventions for opioid use disorder

2.4. How the intervention might work

2.4.1. Housing interventions

2.4.2. Income assistance interventions

2.4.3. SCM and peer support

Standard case management

Peer support

2.4.4. Mental health interventions

Assertive community treatment

Intensive case management

Critical time intervention

Injectable antipsychotics

2.4.5. Interventions for substance use

Supervised consumption facilities

Managed alcohol programmes

Pharmacological interventions for opioid use disorder

2.5. Why it is important to do this review

3. OBJECTIVES

4. METHODS

4.1. Criteria for considering studies for this review

4.1.1. Types of studies

4.1.2. Types of participants

4.1.3. Types of interventions

4.1.4. Types of outcome measures

Primary outcomes

Secondary outcomes

Duration of follow‐up

Types of settings

4.2. Search methods for identification of studies

4.2.1. Electronic searches

4.2.2. Searching other resources

4.3. Data collection and analysis

4.3.1. Selection of studies

4.3.2. Data extraction and management

4.3.3. Assessment of risk of bias in included studies

4.3.4. Measures of treatment effect

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	This study was a quasi experimental trial with non‐random assignment design.
Allocation concealment (selection bias)	High risk	This study was a quasi experimental trial with non‐random assignment design.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	Unclear risk	No description of blinding of outcome assessment was provided.
Incomplete outcome data (attrition bias)	Unclear risk	The attrition rate in the group receiving only case management was significant and was probably due to the time‐limited nature of these services.
Selective reporting (reporting bias)	Low risk	No evidence of reporting outcomes selectively.
Other bias	Low risk	No other potential biases detected.

Methods	Design: Randomized control trial
	Setting: Illinois, United States
	Follow up: 24 months
Participants	Population: Homeless addicted male veterans
Participants	Sample size: Total sample n=358, Intervention n = 178, Control n = 180
Interventions	Intervention:
	3‐6 months of stay at hospital facility (residential care) plus case management. The ratio of residents to case managers was 10:1 in the residency phase and approximately 25:1 in the community follow‐up phase. Relapse prevention skills training, consisting of assertive drink and drug refusal, coping with relapse, social networking, and anger management, was an essential component of CMRC treatment. Self‐help groups such as Narcotics Anonymous (NA) and Alcoholics Anonymous (AA), both on‐site and community‐based, were emphasized for emotional support and to enhance coping behavior while achieving abstinence
	Control:
	21 days in customary care: During the hospital stay in customary care, patients received substance abuse and abstinence education and individual and group therapy. Patients were seen by a social worker for assessment, psychotherapy, and discharge planning.
Outcomes	Housing stability, Substance use, Employment
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Participants were randomized to treatment condition. However, no description of random sequence generation was provided.
Allocation concealment (selection bias)	Low risk	“The 358 subjects were measured at the pretest on the variables on interest before random assignment and before receiving the experimental or control intervention. This procedure ensured that neither the investigators nor the subjects would know who was to be assigned to the experimental or control group after the pre‐test”.
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design.
Blinding of outcome assessment (detection bias)	Unclear risk	No description of blinding of outcome assessment was provided.
Incomplete outcome data (attrition bias)	Low risk	“… we concluded that differential attrition was not a significant problem in this study”.
Selective reporting (reporting bias)	Low risk	No evidence of reporting outcome selectively was detected.
Other bias	Low risk	No other potential biases were detected.

Methods	Design: Randomized controlled trial
	Recruitment: by the CBPR team members' wide dissemination of flyers
	Setting: Clinics and homeless shelters in Metropolitan Chicago
	Follow‐up: 4, 8, and 12 months
Participants	Population: Homeless African Americans with serious mental illness
	Eligibility criteria: People had to self‐identify as African American and report being currently homeless. People also had to self‐report whether they currently were challenged by mental illness and their current diagnosis. Excluded were those who reported no mental illness, or who were receiving case management somewhere else.
	Sample size: Total n = 67, Intervention n = 33, control n = 33
	Baseline characteristics: “Overall, 39% (N = 26) of research participants were female, and the sample had a mean (SD) age of 52.9(68.0). The group was 87% (N=58) heterosexual and somewhat varied in education, with 64% (N = 43) having a high school diploma or less. Thirteen percent (N = 10) reported having some kind of employment. The two groups did not differ significantly on any demographic characteristics nor on primary diagnosis” (Corrigan et al., 2017)
Interventions	Peer Navigator Program (PNP): The peer navigators conducted face‐to‐face meetings with service recipients in places and at times that were convenient to the recipient. Goals of the meetings were to review all health concerns and actions to address these concerns. Goals and actions could include activity related to alleviating homelessness, improving diet, and reducing criminal justice involvement because each of these factors will influence health. PNs were expected to contact participants at least once a week. However, frequency was as high as five times a week, depending on participants' needs. All three PNs were African Americans who were homeless during their adult life and in recovery from serious mental illness.
Interventions	Treatment as usual: usual care which may have included services provided by the Together for Health system (T4H), a coordinated care entity funded by the Illinois Medicaid Authority to engage and manage care for individuals with multiple chronic illnesses. T4H was a network of more than 30 mental and other health care programs in Chicago to provide integrated care to people with serious mental illness. One of the goals of T4H (and for the PNP, for that matter) was to engage and enroll people with disabilities into its network.
Outcomes	Housing stability: Frequency of homelessness
	Mental health: Frequency of experiencing mental health problems in the past 30 days using the Texas Christian University Health Form (TCU‐HF), and emotional wellbeing using the Short Form‐36 (SF‐36)
	Quality of life: Generic quality of life was assessed with Lehman's Quality of Life Scale (QLS).
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The investigators mentioned randomization but did not report on the random sequence generation
Allocation concealment (selection bias)	Unclear risk	No reporting was made on allocation concealment
Blinding of participants and personnel (performance bias)	High risk	Blinding of participants and personnel was not possible due to the nature of the intervention and study design
Blinding of outcome assessment (detection bias)	Unclear risk	No reporting was made on blinding of outcome assessors
Incomplete outcome data (attrition bias)	Low risk	Seven participants were lost to follow‐up, with two of these participants dying during the course of the project and three being incarcerated
Incomplete outcome data (attrition bias)	Low risk	Reason for dropping out were not related to the nature of the outcome
Selective reporting (reporting bias)	Low risk	No evidence of selective outcome reporting
Other bias	Low risk	No evidence of other concerns for bias

Methods	Design: Cost analysis, based on administrative databases maintained by eight agencies
Participants	Population: Homeless people who received services or support from eight agencies in New York
Interventions	Intervention:
	New York‐New York housing placement
	Comparator:
	No supportive housing placement
Outcomes	Cost or cost‐effectiveness: The total mean cost of service utilization for the New York‐New York placement period was US$40,451 per placement per year; the annualised cost per placement was $13,570; NY/NY housing was associated with a $12,146 net reduction in health, corrections, and shelter service use annually per person over the first 2 years of the intervention; a New York‐New York placement had a net additional cost of $1,425 per placement per year.
Notes

Methods	Design: Randomized control trial
	Setting: Netherlands
	Follow up: up to 9 months
Participants	Population: Homeless adults who had resided in a shelter for less than 14 months and were moving to housing
Participants	Sample size: Total sample n = 183, Intervention n = 94, control n = 89
Interventions	Intervention:
	Critical Time Intervention (CTI). Time‐limited, strength‐based intervention for vulnerable people, which bridges the gap between services during times of transition, generally consisted of discharge planning and referral services and access to a range of community‐based services. The CTI worker provides practical and emotional support and helps to develop and strengthen links with community resources, creating a network that will continue to provide support long after CTI has ended. Delivered in three phases of 3 months: transition to the community (phase 1), try‐out (phase 2), and transfer of care (phase 3)
	Comparator:
	Care‐as‐usual
Outcomes	Housing stability, Mental health, Quality of life, Substance use
Notes

Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomization was stratified by shelter and balanced within blocks of four. A list of random numbers was computer generated by a member of the research team and concealed in a secure digital file until assignment.
Allocation concealment (selection bias)	Low risk	Clients, shelter staff, and this research assistant did not have any foreknowledge of condition assignment.
Blinding of participants and personnel (performance bias)	High risk	Clients remained blind until they met their CTI worker or case manager for the first time. However, both participants and personnel were aware of allocation once CTI intervention was initiated.
Blinding of outcome assessment (detection bias)	High risk	Condition assignment was withheld from data collectors. However, some of them did become aware of the allocated condition during follow up, because participants would sometimes spontaneously disclose this information during an interview.
Incomplete outcome data (attrition bias)	Low risk	Low attrition in study (5%)
Selective reporting (reporting bias)	Low risk	Outcomes pre‐specified in published protocol.
Other bias	Low risk	Bias limited by use of standardized instruments.

Methods	Design: Cost analysis based on a prospective experimental design from state or local government
Participants	Population: Individuals with a current diagnosis of severe mental illness
Interventions	Intervention:
	Evolving consumer household
	Comparator:
	Independent apartment living
Outcomes	Cost or cost‐effectiveness: Individuals assigned to evolving consumer households had mean annual housing expenditures of $42,829 compared with $13,042 for individuals assigned to independent living apartments; difference in annual mean treatment costs of individuals in the ECH group ($11,293) and the independent apartment living group ($14,541) were not statistically different; the mean annual cost per person the evolving consumer households group was statistically higher than those in independent living apartments ($56,434 vs $29,838).
Notes

Methods	Design: Randomized controlled trial
	Recruitment: Potential participants were identified by the housing program case managers or by medical records.
	Setting: 2 Veterans Affairs medical centers, blinded city, US
	Follow‐up: At baseline, midpoint (about 4–6 mo), and endpoint (about 9–12 mo)
Participants	Population: Formerly homeless Veterans housed using housing vouchers (HUD‐VASH)
	Eligibility criteria: having a history of abusing alcohol and/or drugs, having a serious mental illness, and having a history of homelessness
	Sample size: Total n = 166, Intervention n = 85, control n = 81
	Baseline characteristics: “Participants were older mostly men with nearly equal proportions of white and African American Veterans. Modal education status was high school graduate and most were unemployed and unmarried. 40.1% of Veteran participants had been homeless 2–4 times, for 1–3 years (32.7%). There were no statistically significant differences for any baseline characteristics between treated and control group Veterans” ‐Ellison et al., 2020
Interventions	Peer specialist services: Alongside the HUD‐VASH housing vouchers, the peer specialist‐Veteran meetings were designed to focus on mental health and substance use recovery and community integration skills delivered in both 20 “structured” sessions and 20 “unstructured” meetings designed for community engagement and relationship building. The peer specialists were intended to meet with Veterans at their home or in the community for 1 h each week for 40 sessions to occur over the course of 9–12 months. The peer specialists had to be Veterans with significant recovery from mental health issues, which could include substance abuse. Seven male peer specialists were hired over the course of the study, 6 were white and 1 was African American. It was not possible to provide peer specialists matched to sex and racial profiles of the Veterans.
Interventions	Treatment as usual: Included the HUD‐VASH housing vouchers, as well as usual care provided by the HUD‐VASH program.
Outcomes	Housing stability: Housing stability was measured with a count of days spent in HUD‐VASH housing within the past 30 days and with a count of days housed that included HUD‐VASH housing and other types of community housing (such as with a friend or parents).
	Mental health: Symptoms of mental illness were identified using the 24‐item Behavior And Symptom Identification Scale (BASIS‐24).
	Substance use: alcohol and drug use were identified using questions from the Addiction Severity Index (ASI).
Notes