Abstract
This Campbell systematic review examines the effectiveness of interventions to reduce homelessness and increase residential stability for individuals who are homeless, or at risk of becoming homeless. Forty-three studies were included in the review, 37 of which are from the USA.
Included interventions perform better than the usual services at reducing homelessness or improving housing stability in all comparisons. These interventions are: High intensity case management Housing First Critical time intervention Abstinence-contingent housing Non-abstinence-contingent housing with high intensity case management Housing vouchers Residential treatment
These interventions seem to have similar beneficial effects, so it is unclear which of these is best with respect to reducing homelessness and increasing housing stability.
Plain Language Summary
Interventions to reduce homelessness and improve housing stability are effective
There are large numbers of homeless people around the world. Interventions to address homelessness seem to be effective, though better quality evidence is required.
What is this review about?
There are large numbers of homeless people around the world. Recent estimates are over 500,000 people in the USA, 100,000 in Australia and 30,000 in Sweden. Efforts to combat homelessness have been made on national levels as well as at local government levels.
This review assesses the effectiveness of interventions combining housing and case management as a means to reduce homelessness and increase residential stability for individuals who are homeless, or at risk of becoming homeless.
What is the aim of this review?
This Campbell systematic review examines the effectiveness of interventions to reduce homelessness and increase residential stability for individuals who are homeless, or at risk of becoming homeless. Forty-three studies were included in the review, 37 of which are from the USA.
What studies are included?
Included studies were randomized controlled trials of interventions for individuals who were already, or at-risk of becoming, homeless, and which measured impact on homelessness or housing stability with follow-up of at least one year.
A total of 43 studies were included. The majority of the studies (37) were conducted in the United States, with three from the United Kingdom and one each from Australia, Canada, and Denmark.
What are the main findings of this review?
Included interventions perform better than the usual services at reducing homelessness or improving housing stability in all comparisons. These interventions are: High intensity case management Housing First Critical time intervention Abstinence-contingent housing Non-abstinence-contingent housing with high intensity case management Housing vouchers Residential treatment
These interventions seem to have similar beneficial effects, so it is unclear which of these is best with respect to reducing homelessness and increasing housing stability.
What do the findings of this review mean?
A range of housing programs and case management interventions appear to reduce homelessness and improve housing stability, compared to usual services.
However, there is uncertainty in this finding as most the studies have risk of bias due to poor reporting, lack of blinding, or poor randomization or allocation concealment of participants. In addition to the general need for better conducted and reported studies, there are specific gaps in the research with respect to: 1) disadvantaged youth; 2) abstinence-contingent housing with case management or day treatment; 3) non-abstinence contingent housing comparing group vs independent living; 4) Housing First compared to interventions other than usual services, and; 5) studies outside of the USA.
How up-to-date is this review?
The review authors searched for studies published up to January 2016. This Campbell systematic review was published in February 2018.
Executive summary
Background
The United Nations Universal Declaration of Human Rights (Article 25) states that everyone has a right to housing. However, this right is far from being realized for many people worldwide. According to the United Nations High Commissioner for Refugees (UNHCR), there are approximately 100 million homeless people worldwide. The aim of this report is to contribute evidence to inform future decision making and practice for preventing and reducing homelessness.
Objectives
To identify, appraise and summarize the evidence on the effectiveness of housing programs and case management to improve housing stability and reduce homelessness among people who are homeless or at-risk of becoming homeless.
Search methods
We conducted a systematic review in accordance with the Norwegian Knowledge Centre's handbook. We systematically searched for literature in relevant databases and conducted a grey literature search which was last updated in January 2016.
Selection criteria
Randomized controlled trials that included individuals who were already, or at-risk of becoming, homeless were included if they examined the effectiveness of relevant interventions on homelessness or housing stability. There were no limitations regarding language, country or length of homelessness. Two reviewers screened 2,918 abstracts and titles for inclusion. They read potentially relevant references in full, and included relevant studies in the review.
Data collection and analysis
We pooled the results and conducted meta-analyses when possible. Our certainty in the primary outcomes was assessed using the Grading of Recommendations Assessment, Development, and Evaluation for effectiveness approach (GRADE).
Results
We included 43 relevant studies (described in 78 publications) that examined the effectiveness of housing programs and/or case management services on homelessness and/or housing stability. The results are summarized below. Briefly, we found that the included interventions performed better than the usual services in all comparisons. However, certainty in the findings varied from very low to moderate. Most of the studies were assessed as having high risk of bias due to poor reporting, lack of blinding, or poor randomization and/or allocation concealment of participants.
Case management
Case management is a process where clients are assigned case managers who assess, plan and facilitate access to health and social services necessary for the client's recovery. The intensity of these services can vary. One specific model is Critical time intervention, which is based on the same principles, but offered in three three-month periods that decrease in intensity.
High intensity case management compared to usual services has generally more positive effects: It probably reduces the number of individuals who are homeless after 12-18 months by almost half (RR=0.59, 95%CI=0.41 to 0.87)(moderate certainty evidence); It may increase the number of people living in stable housing after 12-18 months and reduce the number of days an individual spends homeless (low certainty evidence), however; it may have no effect on the number of individuals who experience some homelessness during a two year period (low certainty evidence). When compared to low intensity case management, it may have little or no effect on time spent in stable housing (low certainty evidence).
Critical time intervention compared to usual services may 1) have no effect on the number of people who experience homelessness, 2) lead to fewer days spent homeless, 3) lead to more days spent not homeless and, 4) reduce the amount of time it takes to move from shelter to independent housing (low certainty evidence).
Abstinence-contingent housing programs
Abstinence-contingent housing is housing provided with the expectation that residents will remain sober. The results showed that abstinence-contingent housing may lead to fewer days spent homeless, compared with usual services (low certainty evidence).
Non-abstinence-contingent housing programs
Non-abstinence-contingent housing is housing provided with no expectations regarding sobriety of residents. Housing First is the name of one specific non-abstinence-contingent housing program. When compared to usual services Housing First probably reduces the number of days spent homeless (MD=-62.5, 95%CI=-86.86 to -38.14) and increases the number of days in stable housing (MD=110.1, 95%CI=93.05 to 127.15) (moderate certainty evidence). In addition, it may increase the number of people placed in permanent housing after 20 months (low certainty evidence).
Non-abstinence-contingent housing programs (not specified as Housing First) in combination with high intensity case management may reduce homelessness, compared to usual services (low certainty evidence). Group living arrangements may be better than individual apartments at reducing homelessness (low certainty evidence).
Housing vouchers with case management
Housing vouchers is a housing allowance given to certain groups of people who qualify. The results showed that it mayreduce homelessness and improve housing stability, compared with usual services or case management (low certainty evidence).
Residential treatment with case management
Residential treatment is a type of housing offered to clients who also need treatment for mental illness or substance abuse. We found that it mayreduce homelessness and improve housing stability, compared with usual services (low certainty evidence).
Authors’ conclusions
We found that a range of housing programs and case management interventions appear to reduce homelessness and improve housing stability, compared to usual services. The findings showed no indication of housing programs or case management resulting in poorer outcomes for homeless or at-risk individuals than usual services.
Aside from a general need for better conducted and reported studies, there are specific gaps in the research. We identified research gaps concerning: 1)Disadvantaged youth; 2) Abstinence-contingent housing with case management or day treatment; 3) Non-abstinence contingent housing, specifically different living arrangements (group vs independent living); 4) Housing First compared to interventions other than usual services, and; 5) All interventions from contexts other than the USA.
Background
Description of homelessness
The United Nations Universal Declaration of Human Rights (Article 25) states that everyone has a right to housing. However, this right is far from being realized for many people worldwide. According to the United Nations High Commissioner for Refugees (UNHCR), there are approximately 100 million homeless people worldwide (1).
Defining homelessness
The term “homeless” is defined differently according to context, purpose and the geographical setting. There are three basic domains for understanding “home” and “homelessness”: 1) the physical domain (the absence of home); 2) the social domain (homelessness connected to discrimination and social exclusion), and 3) the legal domain (individuals have a right to tenancy, and people without homes still have rights and are deserving of dignity) (2, 3).
In the European Union, four categories of homelessness have been developed: roofless, houseless, insecure housing and inadequate housing (3). In the United States, the Department of Housing and Urban Development defines a person as homeless “if he or she lives in an emergency shelter, transitional housing program (including safe havens), or a place not meant for human habitation, such as a car, abandoned building, or on the streets”(4). For the purpose of this review, the following Norwegian definition of homeless should be considered:
“A person is homeless when s/he lacks a place to live, either rented or owned, and finds themselves in one of the three following situations: Has no place to stay for the night; Is referred to an emergency or temporary shelter/accommodation; Is a ward of the correctional and probation service and due to be released in two months at the latest; Is a resident of an institution and due to be discharged in two months at the latest; Lives with friends, acquaintances or family on a temporary basis” (5).
A glossary of terms related to homelessness, relevant interventions and study characteristics is included in Appendix 1.
Causes of homelessness
In discussing causes of homelessness, it is important to think of two different but related questions: ‘Why does homelessness exist?’ and ‘Who is most vulnerable to becoming homeless?’ (6). As Paul Koegel describes in Homelessness Handbook, the structural context of homelessness (why?) includes “a growing set of pressures that included a dearth of affordable housing, a disappearance of the housing on which the most unstable relied, and a diminished ability to support themselves either through entitlements or conventional or makeshift labour” while the people most affected (who?) “disproportionately include those people least able to compete for housing, especially those vulnerable individuals who had traditionally relied on a type of housing that was at extremely high risk of demolition and conversion…high numbers of people with mental illness and substance abuse…individuals with other sorts of personal vulnerabilities and problems” (6).
Homelessness around the world
Although homelessness has been defined and measured differently, some important descriptive statistics from different countries indicate the importance of the problem. Given the various ways of measuring homelessness, the following statistics are not meant to be compared among each other. A recent report stated that in the USA on a given night in January 2015, almost 565,000 people were experiencing homelessness (sleeping outside, in shelter or in transitional housing) (4). Although homelessness in the USA has decreased by 2% from 2014 to 2015, this figure is still very high (4). Homelessness is also a serious problem in Europe: 34,000 people were defined as homeless in Sweden in 2011 (7), and 14,780 households were defined as unintentionally homeless in the United Kingdom in 2016 (8). In Canada, it is estimated that approximately 1% of the population (35,000) are homeless on any given night (9) and more than 105,000 persons in Australia were counted as homeless on census night in 2011 (10). Little is known about the extent of homelessness in most developing countries due to little or no reliable data (11).
In this review we have included both individuals who are homeless (living on the streets, in shelter or temporary housing), and those who have been identified as at-risk of becoming homeless (individuals with mental illness, chronic physical illness, substance abuse, recently released criminal offenders).
Description of the intervention
A serious problem, affecting any effort to synthesize research on housing programs and case management for homelessness, is a lack of consistency in the use of program labels (12). Below is a short description of the groups of interventions included in this review.
Case management
Case management (CM) is a “collaborative process of assessment, planning, facilitation and advocacy for options and services to meet an individual's health and social needs through communication and available resources” (13). In an early review of case management, Morse (1998) summarized the research on why case management has been widely implemented with homeless individuals (14): people who are homeless have multiple serious problems and their service needs are often unmet (15, 16), and these services, and the necessary resources, are difficult to access (17). Furthermore, patients with a mental illness may refuse help and/or miss appointments and/or show aggressive or antisocial behaviour which leads to exclusion from care in many instances (16). Case managers are intended to help guide the individual through the system and facilitate their access to resources and services.
Morse (14)suggested that case management can be described in terms of seven process variables that impact on the intensity of care provided: Duration of services (varying from brief or time limited to ongoing and open-ended) Intensity of services (involving frequency of client contact, and client-staff ratios) Focus of services (from narrow and targeted to comprehensive) Resource responsibility (from system gatekeeper responsible for limiting service utilization to client advocate responsible for increasing access or utilization of services) Availability (from scheduled office hours to 24-hour availability) Location of services (from all services delivered in office to all delivered in vivo) Staffing ratios and composition (from individual caseloads to interdisciplinary teams with shared caseloads)
Case management interventions can be categorized into the following five models: broker case management (BCM), standard case management (SCM), intensive case management (ICM), assertive community treatment (ACT), and critical time intervention (CTI). See Table 3.1 in Appendix 3 for an adapted overview of case management models (14, 18).
In this review, we have organized case management according to intensity: high versus low. The following is a description of the interventions included under high intensity case management:
Assertive Community Treatment (ACT) is an example of intensive case management in which a high level of care is provided. The distinguishing features of ACT are described as follows: “case management provided by a multidisciplinary team of professionals, including psychiatrists, social workers, nurses, occupational therapists, vocational specialists, etc.; 24-hour, 7 days a week coverage; assertive outreach; and providing support to clients in the community where they live rather than office-based practice” (19).
Intensive case management (ICM) is similar to ACT. However, the primary difference (McHugo et al., 2004; Meyer and Morrissey, 2007) is that while ACT involves a shared caseload approach, ICM case managers are responsible for their individual caseloads. Furthermore, each staff member of an ACT team provides direct services, while this is not the case when ICM is applied. Finally, ICM usually lacks a validated model including a manual for treatment fidelity. We will use the term intensive case management when referring to both categories (ICM and ACT). When it is necessary to separate the two alternatives, this is explicitly emphasized in the text.
Intensive case management (ICM and ACT) is intended to make sure that the client receives sufficient service, support and treatment when and where it is needed. In this way intensive case management (one case manager per 15 or fewer clients, available 24-7, and the combined competence of a multidisciplinary team), may help homeless people to obtain accommodation, and once housed avoid eviction.
Low intensity case management refers to all other types of case management where 1) the case manager has responsibility for more than approximately 15 clients, is less available, and where meetings are scheduled less frequently than, for example, once per week, 2) the intervention is described as standard or broker case management, or 3) where intensity was not described.
Housing programs
Housing programs for homeless people typically provide accommodation and include goals such as long term residential stability, improved life-skills and greater self-determination (20, 21). These programs are complex and may include various forms of support and services, such as case management, work therapy, treatment of mental illness and substance abuse (22).
The objective, to find accommodation and avoid eviction, is assumed to be facilitated by combining case management with housing programs. The housing programs are more or less based on housing philosophies. The philosophy may determine the sequence of how specific program elements are introduced and removed. The intended endpoint is usually the same, i.e., independent living with as high degree of normality as possible, e.g., apartments owned or rented by the client, integrated among apartments for ordinary tenants, where housing is neither contingent on sobriety nor on treatment compliance, and with no on-site staff (23).
Non-abstinence-contingent housing programs
According to one philosophy, stable and independent housing is needed for the client to become treatment ready(24). Housing should neither be contingent on sobriety nor on treatment compliance, but only on rules that apply for ordinary tenants(24). These housing programs aim to provide a safe and predictable living arrangementin order to make the clients treatment ready. The client's freedom to choose is crucial for treatment to be successful(25). Therefore, housing programs are neither contingent on treatment compliance nor on sobriety. In other words, housing is parallel to and not integrated with treatment, or with other services. This type of treatment is also sometimes referred to as Parallel housing, or Housing First.
“Housing First” is a specific model of non-abstinence-contingent housing developed by Pathways to Housing. The program is founded on the idea that housing is a basic right. The two core foundations of the program include psychiatric rehabilitation and consumer choice. Individuals are encouraged to define their own needs and goals. Housing is provided immediately by the program if the individual wishes, and there are no contingencies related to treatment or sobriety. The individual is also offered treatment, in the form of an adapted version of Assertive Community treatment (addition of a nurse practitioner to address physical health problems, and a housing specialist)(24).
Abstinence-contingent housing programs
An alternative philosophy assumes that clients need a transitional period of sobriety and treatment compliance, before they can live independently in their own apartments. Without the transitional phase they will soon become evicted, and return to homelessness. In other words, this phase may be necessary for many clients to become housing ready. According to this philosophy housing is integrated with treatment. This approach has been referred to as treatment first, continuum of care, and or linear approach(22, 26).
Housing vouchers
Housing vouchers are financial support (usually) from the government where the individual can choose any free market rental property they wish, with no conditions based on tenancy other than financial contribution of 30% of their income(27).
Housing programs and case management
Housing programs and case management tend to appear in various combinations. Evaluations are typically based on comparison of one type of combination with another, or with “usual care” (often drop in centres, after care services, outpatient clinics, brokered case management, etc.). This means that housing programs are often not implemented and evaluated in similar forms. Any effort to analyse and synthesize evaluations of housings programs, case management and other included services, must therefore consider this complexity and lack of clarity. In addition to this complexity, the population of homeless people consists of subgroups that may respond differently to alternative interventions: mentally ill, substance abusers, veterans, women, etc., and each of these subgroups can be divided further.
In order to make the intervention complexity more comprehensible, two dimensions are outlined: (1) case management care intensity, and (2) contingency of tenancy in housing programs. On the one end of the case management scale there are teams with caseloads of maximum 15 clients per case manager, and full on-site availability (24 hours, 7 days a week) for services and support. In the middle there is CM with caseloads with between 15 to 40 clients per case manager, and service and support only available duringoffice hours at the office. At the other end of the scale there are no case managers, and clients have to rely on drop-in centres, outpatient clinics, after care services, charities, etc. With respect to contingency in housing programs, there appears to be a dichotomy where programs either require that individuals adhere to agreed-upon treatment or sobriety obligations in order to remain in housing (abstinence-contingent) or no conditionality is placed on tenancy, other than in some cases of financial contributions (non-abstinence-contingent).
How the interventions may work
There are two objectives of the interventions: first to get accommodation, and then to avoid eviction. Housing programs provide accommodation to individuals. Case management (low or high intensity) is intended to compensate for the clients’ lack of resources and to help them either obtain accommodation, and/or after they have become housed, avoid eviction. It is a collaborative process, including assessment, planning, facilitation and advocacy for options and services.
Why it is important to do this review
Efforts to combat homelessness have been made on national levels as well as at local government level, including specific treatments for particular types of clients. In addition, there have been many evaluations of housing and treatment programs for homeless individuals and/or persons at risk of homelessness. Several reviews and meta-analyses have also been published (12, 18, 20, 28-31). Yet, a large share of the reviews are out of date, or do not focus on homelessness and residential stability as primary outcomes, or are not systematic reviews of effectiveness.
Tabol and colleagues (2010) (12) aimed to determine how clearly the supported/supportive housing model is described and the extent to which it is implemented correctly (treatment fidelity). Another recent systematic review by de Vet and colleagues focussed on case management for homeless persons. They identified 21 randomized controlled trials or quasi-experimental studies, but did not conduct a meta-analysis, or GRADE the certainty of the evidence. A review by Chilvers and colleagues published in 2006 looked specifically at supported housing for adults with serious mental illness, but did not identify any relevant studies(32).
This review differs from previous attempts at reviewing the evidence in that we have only included randomized controlled trials that examine a broad range of interventions with follow-up of at least one year. Furthermore, we have pooled the results where possible which has allowed us to look at the evidence across studies and not conclude based on small sample sizes from individual studies. Finally, we have applied GRADE to the outcomes, thus providing a more concrete indication of our certainty in the evidence.
Objectives
The primary objective was to assess the effectiveness of various interventions combining housing and case management as a means to reduce homelessness and increase residential stability for individuals who are homeless, or at risk of becoming homeless. Interventions include: Abstinence-contingent housing, non-abstinence contingent housing, housing vouchers and residential treatment High intensity case management (intensive case management and assertive community treatment), and low (ordinary or brokered) case management Housing programs combined with case management programs.
Methods
This systematic review of the effectiveness of interventions to reduce homelessness and increase residential stability for people who are homeless was conducted in accordance with the guidelines in the NOKC Handbook for Summarizing Evidence (33) and the Cochrane Handbook for Systematic Reviews of Interventions (22).
This review was carried out in two phases. The first phase began with a literature search in 2010. The project was taken over in 2014 by the current review team and two updates to the original search were conducted in addition to a search for grey literature. We reassessed studies included by the original review team for inclusion, and excluded those with a quasi-experimental design (see further details below). Due to problems with archiving, there is no documentation of reasons for exclusion for some of the studies excluded in the first phase of the project.
A protocol was approved and published by the review team in the Campbell Library in 2010. The protocol was used as the basis for the development of a protocol by the current review team which was approved and published on the NOKC website in 2014(34). The updated searches (2014 and 2016) were based on the search specified in the Campbell approved protocol, and the inclusion criteria aresimilar, aside from study design. There are four main differences between the protocol published in Campbell Library and the protocol for the current review: Firstly, in this review protocol we only included RCTs. This decision was based on the number of RCTs identified, which seemed sufficient even after the original search. Secondly, we did not include data or analyses related to cost effectiveness as these outcomes were not prioritized by our commissioners. Thirdly, we did not exclude studies if they did not sufficiently report the results. The results from these studies were reported narratively. Finally, we applied the GRADE approach to all primary outcomes.
Literature search
We systematically searched for literature in the following databases. Unless otherwise noted, the databases were searched in 2016, 2014, and 2010. Any databases that were not searched in 2016 and 2014 is due to lack of access. There were no limitations on the search with respect to date of publication (i.e. the databases were searched for their entirety since indexing began). PsycINFO ASSIA (2014, 2010) Campbell Library (2016) Cochrane Library (including CENTRAL) PsychInfo (2016, 2014) PubMed Social Services Abstracts Sociological Abstracts ERIC (2016, 2014) CINAHL ISI Web of Science (2016, 2014)
In addition, we conducted a search for grey literature through Google and Google Scholar and reference lists of identified and included studies using terms related to homelessness and housing. This search for grey literature was conducted in English, Norwegian, Swedish and Danish.
A research librarian planned and executed all the searches. The complete search strategy is published as an appendix to this report (Appendix 2). The search was last updated in January 2016.
Inclusion criteria
Exclusion criteria
We originally included quasi-experimental designs for consideration when they met the other study criteria and used propensity score matching at baseline. However, given the number of randomized controlled trials identified in the updated literature search, we decided to limit inclusion to randomized controlled trials only. We thus excluded eleven studies from the final review. Given the inherent methodological limitations of quasi-experimental designs in answering effectiveness questions, we do not believe that this decision influenced the final results of this review.
Article selection
Two reviewers independently read and assessed references (titles and abstracts) for inclusion according to pre-defined inclusion criteria (see above). When at least one review author considered the reference potentially relevant, the reference was ordered to be read in full-text. Two reviewers independently read and assessed each article in full-text for inclusion according to a pre-defined inclusion form. Where differences in opinion emerged, the reviewers discussed until consensus was achieved. A third reviewer was brought in in instances where agreement was not possible, to assist in the decision.
Critical appraisal
The included studies were assessed for methodological limitations using the Cochrane Risk of Bias (RoB) tool (37). Studies were assessed as having low, unclear or high risk of bias related to: (1) randomization sequencing, (2) allocation concealment, (3) blinding of personnel and participants, (4) blinding of assessors for subjective outcomes and (5) objective outcomes, (6) incomplete outcome data, (7) selective reporting and (8) any other potential risks of bias. One reviewer assessed each study and a second reviewer checked each assessment and made comments where there were disagreements. Results of the Risk of Bias assessments were discussed until consensus was reached.
Data extraction
One reviewer systematically extracted data from the included studies using a pre-designed data recording form. A second reviewer then checked the data extraction for all included studies. Any differences or comments were discussed until consensus was achieved.
The following core data were extracted from all included studies: Title, authors, and other publication details Study design and aim Setting (place and time of recruitment/data collection) Sample population characteristics (age, gender, ethnicity, mental health/substance use status, homelessness status, criminal activity) Intervention characteristics (degree and type of housing support and degree/type of service support and/or therapy offered) Methods of outcome measurement (clinical, self-report, physical specimens for substance use outcomes) Primary outcomes related to number of days spent in stable housing or homeless Secondary outcomes related to housing (satisfaction with housing, type of housing, etc.), addiction status, mental or physical health, criminal activity, and/or quality of life.
Many of the studies were reported in more than one publication. One publication was identified as the main publication (usually the one with results related to the primary outcomes), and we only extracted data from publications in addition to the identified main publication when they added more information regarding the methods or results on relevant outcomes. We excluded studies if they reanalysed already included data using different techniques.
Given the complexity of the interventions being investigated, we attempted to categorize the included interventions along four dimensions: (1) was housing provided to the participants as part of the intervention; (2) to what degree was the tenants’ residence in the provided housing dependent on, for example, sobriety, treatment attendance, etc.; (3) if housing was provided, was it segregated from the larger community, or scattered around the city; and (4) if case management services were provided as part of the intervention, to what degree of intensity. We created categories of interventions based on the above dimensions: Case management only Abstinence-contingent housing Non-abstinence-contingent housing Housing vouchers Residential treatment with case management
Some of the interventions had multiple components (e.g. abstinence-contingent housing with case management). These interventions were categorized according to the main component (the component that the primary authors emphasized). They were also placedin separate analyses. We then organized the studies according to which comparison intervention was used (any of the above interventions, or usual services).
For each comparison, we evaluated the characteristics of the population. In those cases where they were considered sufficiently similar (specifically with respect to individuals versus families, mental illness, substance abuse problems, literally homeless versus at risk of homelessness), and had comparable outcomes, the results from the studies were pooled in a meta-analysis when possible. In those cases where the populations of studies with the same comparisons were considered too different to analyse together we have not pooled the results.
We extracted dichotomous and continuous data for all outcomes where available. We also extracted raw data and, when such data were available, adjusted outcome data (adjusted comparison (effect) estimates and their standard errors or confidence intervals). When information related to outcome measurement (e.g. sample sizes, exact numbers where graphs were only published in the article) were missing in the publication, we contacted the corresponding author(s) via e-mail and requested the data.
Data synthesis
Results for the primary outcomes (number of days spent in stable housing or homeless) are presented for each comparison along with a GRADE assessment. Results for secondary outcomes (for longest follow-up time) for each comparison were not synthesized, but are presented in Appendix 4. For comparisons where more than two studies are included, we present the primary outcomes with the longest follow-up time. Results for secondary outcomes are described in Appendix 4.
We summarized and presented data narratively in the text and table for each comparison. We also conducted a meta-analysis with random effects model and presented the effect estimate, relative risk and the corresponding 95% confidence interval (CI) using risk ratio for dichotomous outcomes. For continuous outcomes we analysed the data using (standardized) mean difference ((S)MD) with the corresponding 95% CI. We used SMD when length of time was measured different between pooled studies (e.g. in days versus months, etc.). We conducted meta-analyses using RevMan 5,using a random-effects model and inverse-variance approach(38). This method allowed us to weight each study according to the degree of variation in the confidence in the effect estimate.
In cases where the means, number of participants and test statistics for t-test were reported, but not the standard deviations, and there was the opportunity to include results in a meta-analysis, we calculated standard deviations, assuming same standard deviation for each of the two groups (intervention and control).
Heterogeneity
We assessed statistical heterogeneity using I2. Where I2was less than 25% we considered the results to have low heterogeneity. Where I2 was greater than 50% we considered the results to have high heterogeneity. Where this heterogeneity could be explained, we proceeded to pool results. However, if heterogeneity could not be explained, we did not pool the results and presented the results separately for each study.
Subgroup analysis
We did not plan or conduct moderator or subgroup analyses.
Dependent effect sizes
We did not include a comparison group more than once in an analysis. Where we were interested in an intervention and it was compared to two or more comparison interventions that were both considered to be within the realm of “usual services”, we combined the two comparison arms into one comparison group and compared the means of the combined control groups to the intervention for a given outcome (39).
In one study we have combined two intervention arms that both employed slightly differing versions of an intervention (assertive community treatment) into one intervention group and compared that to the usual services comparison condition (40).
Primary outcomes
Outcomes related to housing and homelessness were reported using multiple measurements/scales/methods in some studies. These included number of days spent in stable housing or homeless, length of time to move from shelter to permanent housing (measured in days), number or percentage of participants who reported being homeless during a given period, or at a certain measurement point, and the change in number/proportion of days spent in various living conditions between baseline and follow-up points.
Secondary outcomes
We did not synthesize or report results for secondary outcomes. They are described in Appendix 4 as they are reported in the original primary publications.
GRADING of the evidence
We assessed the certainty of the synthesized evidence for each primary outcome using GRADE (Grading of Recommendations Assessment, Development, and Evaluation). GRADE is a method for assessing the certainty of the evidence in systematic reviews, or the strength of recommendations in guidelines. Evidence from randomized controlled trials start as high certainty evidence but may be downgraded depending on five criteria in GRADE that are used to determine the certainty of the evidence: i) methodological study quality as assessed by review authors, ii) degree of inconsistency, iii) indirectness, iv) imprecision, and v) publication bias. Upgrading of results from observational studies is possible according to GRADE if there is a large effect estimate, or a dose-response gradient, or if all possible confounders would only diminish the observed effect and that therefore the actual effect most likely is larger than what is suggested by the data. GRADE has four levels of certainty:
Assessments are done for each outcome and are based on evidence coming from the individual primary studies contributing to the outcome. For more information on GRADE visit www.gradeworkinggroup.org, or see Balshem and colleagues 2011 (41).
For a detailed description of the Norwegian Knowledge Centre's procedures, see the Norwegian Knowledge Centre's Handbook(33).
Results
The search was conducted in three stages. The original systematic search of databases in 2010 resulted in 1,764 unique references (Figure 1). We identified a further831 unique references from the update search in2014, and 323 more in the January 2016 update search. Altogether we identified 2,918potentially relevant references through database searches. In addition, a grey literature search identifiedan additional 2 relevant studies (and 11 references). We excluded 2,526references based on title and abstract. We read 394 references in full and excluded 316 based on the predefined inclusion and exclusion criteria. In total, we critically appraised 43 studies that were described in 78 publications. A list of excluded studies with reasons for exclusion is included in Appendix 5. Problems related to archiving from the first search in 2010 resulted in missing the references and the reasons for exclusion for 50 excluded studies.
Flowchart of the literature selection process
Description of the included studies
We identified 43 randomized controlled studies (RCTs) reported in 78 publications (24, 26, 27, 39, 40, 42-81)that met our inclusion criteria, and two studies in progress (31, 82). See Appendix 9 for a description of the studies in progress.
Thirteen of the included studies were published in or after 2010, thirteen were published between 2000 and 2009, and seventeen studies were published before 2000.
The majority of the studies were conducted in the United States (n=37), and other included studies came from other high-income countries, including United Kingdom (n=3), Australia (n=1), Canada (n=1), and Denmark (n=1). Eleven of the studies were conducted at multiple sites (cities/institutions).
The duration of the intervention was not reported in all of the included studies. It appears that in most of these cases the intervention was available/offered until the longest follow-up. There were also some discrepancies between the number of participants randomized and the number of participants included in analyses in some cases. We have highlighted where we think this is a concern.
From these 43 RCTs we have summarized findings from 28 comparisons in five categories of interventions (see Table 1).
Overview of comparisons of case management interventions
Risk of bias in the included studies
The majority of the RCTs were assessed as having high risk of bias. In many instances this was due to inadequate reporting of methods in general (unclear risk of bias). In particular, most studies were at unclear or high risk of selection bias because they either did not report randomization or allocation concealment procedures or reported inadequate methods of randomization or allocation concealment. The vast majority of studies were assessed as having unclear or high risk of performance bias: Blinding of participants and personnel was either not described in many studies (unclear risk), or not possible and reported as such (high risk). In the majority of studies outcome assessors were not blinded (high risk), or blinding was not mentioned (unclear risk). The risk of bias was separated into blinding of outcome assessment for subjective and objective outcomes due to the poor reporting, or lack, of blinding. The intention behind this was to indier4’;cate that the blinding might have an impact on subjective outcomes, but not objective outcomes such as death or number of days housed when the data came from administrative records. Some studies also were assessed as being at high risk for attrition bias because they used inappropriate methods for dealing with missing data, or reporting bias because the results were not reported for all outcomes. It is not clear how much attrition has occurred in many of the primary studies, and in some cases the level of attrition differs between results within the same study but is not discussed by the primary authors. See Appendix 6 for a more detailed explanation of the risk of bias assessment for each study.
Interventions and comparisons
We included and extracted data from 43RCTs(this information was presented in 78 publications). Some studies included multiple comparisons (multiple interventions), and some publications reported results from multiple studies (for example information related to two studies in one publication). Details on all of the included comparisons are described below. Details regarding data related to secondary outcomes is not reported in the main text of this report but can be found in Appendix 4.
The case management component in the included studies varied in terms of approach, intensity and case-load for case managers. We have therefore categorized case management components as either low intensity (case management with no further details, brokered case management), high intensity (Assertive Community Treatment or Intensive Case Management), or Critical Time Intervention (intensive case management for a shorter defined period of time). In addition, some interventions included a housing component and a treatment component that could not be described as case management (e.g. day treatment or Community Reinforcement Approach). Interventions including these treatment components have been analysed separately from interventions that include low or high intensity case management components. Most of the interventions evaluated in the included comparisons were complex in that they were made up of multiple components, and there was a large degree of flexibility in terms of how the interventions were implemented (including varying levels of treatment fidelity). Furthermore, many of the studies reported that the interventions and control conditions changed and evolved during the course of the studies in terms of organization, and availability of resources and services. More details on the interventions evaluated in each study is reported under the relevant comparison.
The comparison groups varied considerably, and in many cases it is difficult to ascertain what kind of interventions participants in these groups received/were offered due to poor reporting. The comparison groups were described as usual services (care as usual), other types of housing programs or case management interventions, or other types of interventions. All of the comparison groups, however, received some type of active intervention. That is, even participants in the usual services groups had access to drop in centres, and to some degree case management and/or shelter.
Population in the included studies
A total of approximately 10,570 participants were included in the identified studies. This is an approximate number due to poor reporting in many of the studies. The majority of the studies included adults who had a mental illness or substance dependence and were homeless or at-risk of becoming homeless due to the previous mentioned illnesses. More detail on the populations in the included studies is available under each comparison.
Description of outcomes reported in the included studies
All of the included studies reported at least one outcome related to homelessness or housing stability. This was reported in various ways including the number of days participants reported being housed/homeless, proportion of participants homeless or housed at follow-up, time to exit from/return to shelter, and frequency of address change. Many of the included studies also included outcomes related to employment, mental or physical health, quality of life, social support and criminal activity. Details regarding outcomes are described under each comparison.
Secondary outcomes for each comparison are presented in Appendix 8.
Category 1: Case management
Description of included studies
We identified 26 studies with four comparisons that evaluated the effect of case management on housing stability and/or homelessness (26, 39, 40, 44-48, 50, 52-54, 56, 59, 60, 64, 69-72, 74, 76, 77, 79, 80, 83). The majority of the studies were conducted in the USA (N=22), with the remaining studies from either Australia (N=1), Denmark (N=1) or the United Kingdom (N=3). Data for the included studies were collected between the 1980s (earliest published study from 1990, but it is unclear when data was collected) and 2009, and thus represent varying populations and settings in terms of political and social climate in the various countries and states where the studies are conducted. The exact number of participants is not always clearly reported. We have reported the total number randomized and included in analyses where possible.
Within the category of case management, we identified four subcategories of interventions which were compared to usual services or other interventions. See Table 2 for an overview.
Overview of case management comparisons
Table 3 presents an overview of the populations, interventions, comparisons and outcomes in the included studies. The total number of participants indicates the number of participants randomized. The number of participants for each group does not always add up to the total number of participants because most studies reported the number included in analyses, but not always the number randomized. Participants in the included studies were adults (>18 years old) unless otherwise specified. We report the longest outcome assessment for each study (shorter follow-up assessments were also done in some studies).
Description of studies that evaluated effects of case management interventions (N=26)
Description of the intervention
The case management intervention in the included studies varied considerably in terms of intensity, organization and length. The interventions are described in more detail under the relevant comparison and in Appendix 7.
Category 1.A: High intensity case management
We identified 18 studies that evaluated the effect of high intensity case management on housing stability and/or homelessness (39, 40, 44-46, 48, 50, 52-54, 59, 60, 69-71, 76, 80, 83). High intensity case management included interventions which were described as using either Assertive Community Treatment (ACT; N=12) or intensive case management (ICM; N=6). The included interventions varied in terms of ratio of clients per case manager, frequency of contact, length of treatment and follow-up, location of appointments, degree of service provision versus referral, and team versus individual approach to case management.
The interventions in the majority of the included studies (N=13) are compared to usual services (44-46, 48, 50, 54, 59, 60, 69-71, 80, 83). One study compared the intervention to another type of high intensity case management (76) and two studies compared it to low intensity case management (53, 69). In two of the included studies, multiple intervention arms or comparison arms were relevant for this category of interventions (39, 40). In one study we have combined two intervention arms that both employed slightly differing versions of assertive community treatment into one intervention group compared to usual services (40). In the other study (39), we combined two comparison arms that both offered usual services to participants into one comparison group compared to the intervention.
Services provided as part of “usual services” varied greatly between and within the studies. We have chosen to include all studies that compared high intensity case management to “usual services” in one comparison. The term “usual services” covers a wide variety of services, but generally refers to the variety of services available to any person meeting the eligibility criteria of the study and not an alternative intervention which participants who are not randomized to the intervention group receive. Usual services in the included studies included drop-in centres, provision of a list of services and information (69), case management style services (59)and limited peer coaching(83). Control conditions were too poorly described in most studies to accurately document what participants had access to.
1.A.1. High intensity case management compared to usual services
We identified 18 studies (39, 40, 44-46, 48, 50, 52-54, 59, 60, 69-71, 76, 80, 83) which evaluated the effect of high intensity case management compared to usual services on housing stability and homelessness in the USA (N=15), United Kingdom (N=2) and Denmark (N=1). The included studies were conducted over a long span of time; however, the majority of studies were conducted or began before the end of 2000 (N=12).
Fifteen of the included studies focused on adults with mental illness and/or substance abuse issues (39, 40, 44, 45, 48, 50, 52-54, 59, 60, 69-71, 76). One study focused on disadvantaged youth (46), one study included adults with families (80), and one study targeted recently released criminal offenders (83). While the studies differed slightly in the populations targeted, all of the studies included participants with mental illness and/or substance abuse even when that was not the main identifying characteristic of the target population. Information regarding mental illness and substance abuse was not reported for the study on disadvantaged youth; however, there was little reason to assume that this group would react differently to the intervention. More importantly, given the outcomes analysed here, housing stability and homelessness, one can assume that this is a universally sought after outcome, and the characteristics of the population might not be considered to be important. Below is a description of the results.
Primary outcome: Housing stability
Six of the included studies examined housing stability for adults with mental illness and/or substance dependence issues (45, 46, 50, 54, 59, 60, 69).
We carried out a meta-analysis for number of days in stable housing, pooling available data from four included studies (46, 50, 60, 69, 71) to examine the effect of high intensity case management compared to usual services on number of days in stable housing. As evident from the forest plot (Figure 2), the pooled analysis indicates that the high intensity case management leads to an increase in the number of days spent in stable housing compared to usual services (SMD=0.90, 95%CI=0.00 to 1.79). Although considerable heterogeneity is indicated by I2 and Chi2(I2=98%, chi2=186.17), this is expected due to the complexity of the included interventions, the geographical range of included studies (multiple cities across USA, and Australia) and the wide range of when the interventions were implemented.
Number of days in stable housing, 12-24 months follow-up, high intensity case management vs usual services
We carried out a meta-analysis to estimate the number of participants in stable housing at 12-18 months after the start of the intervention, pooling available data from two included studies (45, 54). As evident from the forest plot (Figure 3), the pooled analysis indicates that high intensity case management leads to a greater number of individuals living in stable housing compared to usual services (RR=1.26, 95%CI= 1.07 to1.49). While the heterogeneity was assessed as being high (I2=73%, chi2=3.64), this can be accounted for by differences in when the interventions were implemented (approximately 15 years between publications) and assessed and geographical differences (UK and USA). Together these differences may have implications for political or social contexts which may, in turn, have impacted, for example, the type of usual services being provided.
Number of participants in stable housing, 12-18 months follow-up, high intensity case management vs usual services
It is uncertain whether high intensity case management improves either the length of time individuals spend in their longest recorded residence, the number of clients who do not move (45), or the number of moves during the last half of a one or two year period (45).
One study reported that there was no difference between the intervention and control groups in the number of moves reported during the previous 12 months as measured at 24 months MD=0.30 (-0.04, 0.64)(46).
Primary outcome: Homelessness
Thirteen of the included studies examined homelessness (39, 44-46, 48, 50, 54, 59, 60, 70, 71, 80, 83). Seven studies reported outcomes related to length of time homeless, either in terms of number of months (44) or number of days (39, 46, 50, 60, 71, 80).
We carried out a meta-analysis for the number of days spent homeless, pooling available (adjusted) data from six included studies (39, 46, 50, 60, 71, 80). One of the studies adjusted the results for demographic characteristics, specifically ethnicity (60). This study (60)also reported both number of days homeless in shelter and number of days homeless on streets. It was not possible to combine the data from these two outcomes (means and the standard error of the mean (SEM) were reported, but not the number of participants who reported experiencing these living arrangements), so we have chosen to include the number of days homeless in shelter in this meta-analysis. The pooled estimate indicates that high intensity case management leads to fewer days spent homeless compared to usual services. Although there is considerable heterogeneity (I2=58%, chi2=11.77), this may be explained by a wide range of geographical settings (USA and Australia), and large differences in when the interventions were implemented and assessed (from 1990s to 2006). Together these differences may have implications for political or social contexts which may, in turn, have impacted, for example, the type of usual services being provided.
Number of days homeless, 12-24 months, high intensity case management vs usual services
In one study (44), high intensity case management seemed to lead to fewer months homeless (mean number of months per 100 months homeless). However, the 95% confidence interval indicates that high intensity case management might make little or no difference the amount of time spent homeless (results as reported in original publication: n=-1.5 [95%CI -4.3 to 1.3], p=0.29).
One study reported that participants in the high intensity case management group reported spending almost half as many days living on the street than participants in the usual services group (MD=0-14.10 (-15.77, -12.43))(60)
Three studies reported whether participants experienced homelessness during the study period (44, 48, 83). We conducted a meta-analysis for the number of participants who experienced at least one episode of homelessness within one to two years, pooling data from two studies (48, 83). The third study was not included in the analysis due to incomplete reporting of results (baseline and follow-up percentage of participants was not reported, only the pre-post difference in percentage of participants who experienced homelessness during a two year period was reported along with the difference in difference (44).
The pooled analysis, shown in Figure 5, indicates that high intensity case management may lead to little or no difference in whether individuals experience homelessness during a one to two year period compared to usual services. Results, as reported in the original publication, from the third study support this (Bell 2015 (44): OR=0.83, 95%CI=0.60 to 1.17).
Number of participants who experienced at least one episode of homelessness, 12-24 months, high intensity case management vs usual services
Three studies examined the number of participants who reported being homeless at the last follow-up point (12 to 18 months after baseline) (54, 59, 70). We conducted a meta-analysis for the number of participants who were homeless 12 to 18 months after the beginning of the study, pooling available data from three studies (54, 59, 70). One study reported the percentage of participants per group, but not the total number per group (amount of data on participants varied according to outcome), so we calculated the total number of participants per group using the information provided (70). As evident from the forest plot (Figure 6), the pooled analysis indicates that high intensity case management probably leads to fewer individuals who report being homeless at the 12 to 18 month follow-up interview compared to usual services (RR=0.59, 95%CI=0.41 to 0.87).
Number of participants who were homeless at last follow-up point, 18 months, high intensity case management vs usual services
The results and quality assessments for high intensity case management compared to usual services on housing stability and homelessness for adults with mental illness and/or substance abuse problems are summarized in Table 4. The complete GRADE evidence profile is shown in Appendix 8, Table 8.1.1.
Summary of findings table for the effects of high intensity case management compared to usual services (Bell 2012, Bond 199, Cox 1998, Grace 2014, Garety 2006, Killaspy 2006, Nordentoft 2010, Nyamathi 2015, Toro 1997)
Considerable heterogeneity (I2=98%, chi2=186.17).
Risk of performance bias.
Fewer than 300 participants.
Risk of performance bias in all studies. Risk of attrition bias in one study.
Risk of performance bias in four studies, risk of detection bias in two studies, risk of attrition bias in two studies and other risks of bias in two studies. Unclear reporting of selection bias in four studies and detection bias in two studies.
Two studies includd in the pooled analysis (N=515). One study not included in the analysis, but shows a similar result: Bell 2012 (intervention N=567, control N=563) OR=0.83, 95%CI=0.60, 1.17.
Inadequate reporting of randomization, allocation concealment and blinding methods in two studies.
Total number of events is less than 300.
Risk of detection bias and attrition bias. Inadequate reporting of blinding methods for participants and personnel.
Fewer than 400 participants.
Two studies included in the pooled analysis (Garety 2006 (54), Killaspy 2006). Nordentoft 2010 (N=496)showed that the intervention led to fewer homeless participants at 12 month follow-up than the control group (OR=0.53, 95%CI=0.3, 0.9).
Risk of performance bias in four studies, risk of detection bias in two studies, risk of attrition bias in two studies and other risks of bias in two studies. Unclear reporting of selection bias in four studies and detection bias in two studies.
What does the evidence say?
High intensity case management compared to usual services: Probably reduces the number of individuals who are homeless after 12-18 months (moderate certainty evidence). May increase the number of the number of people living in stable housing after 12-18 months (low certainty evidence). May lead to little or no difference in the number of individuals who experience some homelessness during a two year period (low certainty evidence). May reduce the number of days an individual spends homeless (low certainty evidence). It is uncertain whether high intensity case management leads to a difference in the number of days an individual spends in stable housing, the number of days an individual spends in their longest residence, and the number of individuals who do or do not move (very low certainty evidence).
1.A.2. High intensity case management compared to low intensity case management
We identified three studies (40, 52, 53) that examined the effects of integrated high intensity case management compared to standard case management (lower intensity) on housing stability and homelessness. The integrated treatment was based on the assertive community treatment model of case management in all three studies. Integrated treatment differs from standard case management models in that it integrates treatment for substance abuse and mental health issues into one service.
In one study (40), participants were randomized to either assertive community treatment, assertive community treatment with a community worker or brokered case management. The primary authors’ most central hypothesis was that assertive community treatment was better for clients with serious mental health issues than brokered case management. This focus fits with the aim of our review and we therefore attempted to combine results from the two assertive community treatment groups to compare them to the brokered case management group (usual services). For the purpose of this review we are interested only in the assertive community treatment condition and have thus combined the two interventions which employed the assertive community treatment model of case management. In this study the assertive community treatment model was expanded and modified: staff were instructed to visit shelters and were trained in engaging with homeless persons.
In two studies (52, 53),the high intensity case management interventions were based on the assertive community treatment model and were provided by two sites (health centres).
Primary outcome: Stable housing
Three studies (40, 52, 53)examined the effect of assertive community treatment compared to standard clinical case management on the number of days participants reported living in stable housing. In the first study (40),the total number of participants was not reported, and despite contacting the study authors, the information was not available. We therefore only report the results as they are reported in the study: High intensity case management led to more days spent in stable housing compared to low intensity case management (F=3.54, df=2, 129, p<0.032). The assertive community treatment group reported more days in stable housing than participants in the other two groups: at the 18 month follow-up participants in the assertive community treatment group reported a mean of 23.70 days (SD=11.42) in stable housing during the previous month compared to 18.98 (SD=13.89) for the assertive community treatment with community workers group, and 16.02 days (SD=14.77) for the broker case management group. The authors conclude that “[t]he results provide substantial, although not complete, support for the study’ s most central prediction: assertive community treatment is a more effective intervention for people with serious mental illness who are at risk of homelessness than is broker case management” (40), p. 502).
We carried out a meta-analysis for stable housing, pooling available data fromtwo studies (52, 53). The pooled analysis indicates that high intensity case management may make little or no difference to the amount of time spent in stable housing compared to low intensity case management (SMD=0.10 [95%CI -0.10 to 0.29], I2=0%) (Figure 7).
Mean number of days spent in stable housing, 36 months (high intensity case management vs low intensity case management)
The results and quality assessments for high intensity case management compared to low intensity case management are summarized in Table 5. The complete GRADE evidence profile is shown in Appendix 8, Table 8.1.2.
Summary of findings table for the effects of high intensity case management vs low intensity case management (Drake 1998, Essock 2006, Morse 1997)
Risk of detection bias in one study. Inadequate reporting of methods in both studies.
The third study that could not be included in the pooled analysis (Morse 1997). The third study reported that HICM led to more days in stable housing (F=3.54, df=2, 129, p<0.032).
While only two studies are included in the analysis reported here (total population of 401 participants), the outcome is examined in three studies (total population of 458 participants).
What does the evidence say?
High intensity case management compared with low intensity case management for individuals with mental illness and substance abuse problems: Maylead to little or no difference in the number of days people spend in stable housing (low certainty evidence).
1.A.3. High intensity case management compared to other intervention (no case management or housing program)
The study (83)that examined the effect of high intensity case management compared to another intervention that did not include case management or housing on housing stability and homelessness included three trial arms. The first comparison (high intensity case management compared to usual services) is included above. The high intensity case management intervention is described above, and the comparison condition consisted of peer coaching with brief nurse counselling which was identical to the peer coaching component of the intervention program, but lacked the case management component.
Primary outcome: Homelessness
Results from the included study (83)showed that approximately 10% of intervention group participants compared to 11% of comparison group participants reported living on the streets or in shelter during the study period (12 months) (intervention: 17/166; comparison: 20/177), and 50% of the intervention group compared to 41% of the control group reported living in someone else's house. Approximately 40% (66/166) of participants in the intervention group reported living in institutions compared to 47% (83/177) of participants in the comparison group (RR=0.91, 95%CI=0.49 to 1.67).
The results and quality assessments for high intensity case management compared to another intervention with no housing or case management component for recently released criminal offenders are summarized in Table 6. The complete GRADE evidence profile is shown in Appendix 8, Table 8.1.3.
Summary of findings table for effects of high intensity case management vs other intervention (Nyamathi 2015)
Inadequate reporting of methods.
One small study. Wide confidence interval.
What does the evidence say?
It is uncertain whether high intensity case management reduces homelessness for recently released criminal offenders compared to another intervention (very low certainty).
1.A.4. High intensity case management (with consumer case managers) compared to high intensity case management (with non-consumer case managers)
In the study (76) that compared assertive community treatment with consumer case management to assertive community treatment with case management, the assertive community treatment model was similar in both interventions with slight differences in frequency of meetings between the teams. The main difference was that the consumer team had between none and 11 previous psychiatric hospitalizations and the non-consumer team had no hospitalizations. There was no difference in the number of 15-minute time units of services the first year of the program between the two teams, however consumer case managers provided more services in person to their clients and less office-based services. Participants were recruited between 1990 and 1991.
Primary outcomes: Housing stability and homelessness
The results(76) show that 44 of a total of 90 participants lived in the same housing situation during the two year study period. Six participants (not specified from which group) reported being homeless at some point during the study. This study did not report any difference between the groups. There was no more data available and thus no outcomes for which we could assess certainty of the evidence (see GRADE Evidence profile in Appendix 8, Table 6.1.4).
What does the evidence say?
Data on housing and homelessness were not reported apart from the information given above.
Category 1B: Low intensity case management
We identified five studies(26, 47, 64, 74, 77) that examined the effect of low intensity case management compared to usual services (26, 64), another form of low intensity case management (47), or an intervention that included neither housing programs nor case management(74, 77). The studies were conducted in the USA or the UK and participants were recruited between 1991 and 1996 (26, 64, 77)or between 2006 and 2009 (74). Date of recruitment was not reported in one study (47).
The studies varied in terms of how the intervention was described. Studies were included in this category of interventions if the case management was included as part of the intervention, but the case management component was (a) not described as being intensive (e.g. assertive community treatment, intensive case management), or (b) was described as being or using components of brokered case management.
In the first study (47), the case management services included an occupational therapist consultant and participants were seen weekly for medication monitoring and money management. In the second study (64), the intervention was described as differing greatly according to the individual case manager in terms of time and services offered. At minimum, each participant received a needs assessment and the assessment with the person's carer (all participants were diagnosed with long-term mental disorders), assistance in meeting the identified needs, and monitoring of the participant's progress. The third study (74) examined the effect of three interventions: community reinforcement approach, motivational enhancement therapy and strengths based case management. We have chosen to focus on the case management intervention as the intervention group for this review. The case management intervention included case managers linking participants with resources in the community, securing needed services, focusing on the clients’ strengths and giving the client high degree of responsibility. The fourth study (77) examined the effect of case management which was a hybrid between brokered case management and full-services models. There was a focus on linking patients with services (medical, psychiatric, social, legal and social), arranging appointments and accompanying participants to appointments. In the fifth study (26), case management was provided for an average of 3 months and included ordinary case management services (not described) and provision of immediate tangible resources (e.g. transport tokens, food vouchers, medical care and rent deposits).
The case management interventions were compared to usual services (26, 64), case management without an occupational therapist(47), brief contact (77), or two other interventions that did not included case management or housing programs(74).
1.B.1. Low intensity case management compared to usual services
We found two studies that compared low intensity case management to usual services (26, 64) in the USA.
Usual services were described as services that are usually provided to individuals with substance abuse disorders after discharge from rehabilitation (26) or services that clientshad been receiving prior to study enrolment(64).
The target populations in the two studies differed (individuals with long term mental illness and individuals with substance abuse disorders), which dictated the type of usual services the comparison groups received.
Primary outcomes: Housing status and homelessness
In the first study (64),participants in the intervention group reported a mean of 44.3 days in better housing during the 14 months prior to follow-up compared to 32.3 days for the control group. The intervention group also reported a mean of 15.1 days in worse housing compared to 33.4 days for the control group for the same time period. There was not enough information to assess the difference between groups.
In the second study (26), participants in the intervention group increased their residential stability by 9 days during the 60 days prior to the 12 month follow-up interview. No information was reported for the control group.
The results and quality assessments for low intensity case management compared to usual services are summarized in Table 7. The complete GRADE evidence profile is shown in Appendix 8, Table 8.1.5.
Summary of findings table for effects of low intensity case management vs usual services (Marshall 1995, Sosin 1995)
Risk of performance bias. Inadequate reporting of methods for dealing with missing data and blinding.
Fewer than 400 participants. Unkown confidence interval.
Risk of selection bias and attrition bias. Inadequate reporting of blinding methods.
What does the evidence say?
It is uncertain whether low intensity case management compared to usual services improves housing stability and/or reduces homelessness (very low certainty evidence).
1.B.2. Low intensity case management with an occupational therapist compared to low intensity case management without an occupational therapist
We found one study (47)that compared low intensity case management to low intensity case management in the USA. In this study the comparison condition was identical to the intervention, but with a regular case manager instead of an Occupational Therapist (OT) as case manager.
Primary outcomes: Housing status
The authors of the study measured and report how the participants’ current housing situation differs from their ideal housing standards according to an unspecified 13-point scale. The mean for the intervention group at 12 months was 1.04 below their ideal and for the control group 1.71 below their ideal housing situation. The authors state that the average variance from ideal housing was lower at 12 months than at baseline for the intervention group (t(24)=-2.16, p=0.04) but there was no difference for the control group from baseline to 12 months.
The results and quality assessments for low intensity case management (with OT) vs low intensity case management (no OT) for homeless adults with mental illness are summarized in Table 8. The complete GRADE evidence profile is shown in Appendix 8, Table 6.1.6.
Summary of findings table for effects of low intensity case management (with Occupational therapist) vs low intensity case management (no occupational therapist) (Chapleau 2012)
Risk of performance bias and reporting bias. Inadequate reporting of randomization and allocation concealment methods.
Fewer than 400 participants.
What does the evidence say?
It is uncertain whether low intensity case management compared to low intensity case management has an effect on the amount of time spent in ideal housing (very low certainty evidence).
1.B.3. Low intensity case management compared to other intervention (no case management or housing component)
We found two studies (74, 77)that compared low intensity case management to other interventions without case management or housing components in the USA.
A total of 460 participants were randomized to either case management (N=183) or another intervention (N=277). The participants were recruited between 1994 and 1996 (77)or between 2006 and 2009 (74).
In the first study (77) the comparison group received brief contact, which is described as one or two sessions with a counsellor with a ratio of approximately 100 participants to one case manager, which involved education about reducing HIV transmission and referrals to other services. The focus of the original study was to investigate brief contact. Case management was used in the control condition. However, we have only reported raw data here, and not the effect size as it was calculated and reported in the original publication. The type of comparison condition thus does not impact the results reported here. In the second study (74), the two comparison interventions were community reinforcement approach (CRA) and motivational enhancement therapy (MET). CRA is described as an operant-based behavioural intervention and focuses on building up skills (anger management, social and recreational counselling, and refusal skills training) within the community to achieve and maintain sobriety. MET is an adaptation of motivational interviewing and was described as lower frequency treatment compared to the other two interventions.
Primary outcome: Homelessness
Both studies reported outcomes related to homelessness. In the first study (77),the authors report the number of participants who reported being homeless at each follow-up point; however, the number of participants included in the analysis for each follow-up point is unclear. At 18 months 11.3% of participants in the intervention group and 13.8% participants in the comparison group reported being homeless.
In the second study (74), participants report the mean percentage of days homeless during the 90 days prior to each follow-up interview. At the 12 month follow-up participants in the intervention group (N=64) reported 20.51 days (SD=35.13) as homeless compared to 20.85 days (SD=34.95) for participants in the community reinforcement approach group (N=70) and 21.89 days (SD=35.31) for participants in the motivational enhancement therapy group (N=69). All three groups reported fewer days spent homeless leading up to the final interview compared to the period before baseline assessment. There was no difference between the low intensity case management group and either the CRA group (MD=-0.34, 95%CI=-12.22 to 11.54) or the MET group (MD=-1.38, 95%CI=-13.36 to 10.60).
It is not possible to report the findings from these studies in forest plots given the lack of information reported in the first study (77), and the comparison with two types of control conditions in the second study (74).
The results and quality assessments for low intensity case management compared to another intervention with no case management or housing component for youth and adults with substance abuse problems are summarized in Table 8. The complete GRADE evidence profile is shown in Appendix 8, Table 8.1.7.
Summary of findings table for effects of low intensity case management vs other intervention (no case management or housing component) (Sorensen 2003, Slesnick 2015)
Risk of performance bias. Inadequate reporting of methods.
Fewer than 300 participants.
Inadequate reporting of blinding of participants and personnel and outcome assessors.
Fewer than 400 participants.
What does the evidence say?
It is uncertain whether low intensity case management compared to another intervention with no case management or housing component has an effect for youth and adults with substance abuse problems.
Category 1C: Critical time intervention
In all three studies that examined the effect of Critical Time Intervention compared to usual services (56, 72, 79), the active part of the Critical Time Intervention was nine months; however the length of follow-up and after care activities in the three studies varied.
1.C.1. Critical time intervention compared to usual services
The three included studies targeted either single mothers living with at least one child between 18 months and 16 years and living in shelters with mental illness and/or substance dependence(72), or adults with severe mental illness who are homeless or at-risk of homelessness (56, 79).
Primary outcome: Homelessness
All three of the included studies examined the effect of critical time intervention compared to usual services on homelessness (56, 72, 79).
Results from these studies could not be pooled due to lack of details in reporting of results. In the first study (56), 58 participants from the intervention group and 59 from the control group were included in analyses. Homelessness was measured in two ways. First, participants reported via The Personal History Form ever versus never being homeless in the 18 weeks prior to the last follow-up interview at 18 months. Fewer participants in the intervention group experienced homelessness during this period (3/58) than in the control group (11/59). The authors controlled for baseline homelessness and used a logistic regression to model the impact of assignment to the intervention group on homelessness during the final 18 weeks of the study and found a that Critical Time Intervention reduced the number of days spent homeless compared to usual services. However, the 95% confidence interval indicates that Critical Time Intervention might increase the number of days homeless (OR=0.22, 95%CI 0.06 to -0.88). Secondly, participants reported total number of days homeless during the 18 weeks prior to the 18 month follow-up interview. Participants in the intervention group reported fewer days homeless (M=6) compared to the control group (M=20) (Poisson regression to control for baseline homelessness, p<0.001). The results from this study also showed that Participants in the Critical Time Intervention experienced fewer days homeless during the study period (1,812 nights) compared to the control group (2,403 nights).
In the second study (79),participants reported the number of nights homeless out of 30 days prior to each monthly interview up to 18 months using the Personal History Form. The authors calculated the mean number of nights across each follow-up period. The intervention group reported approximately one third the number of nights homeless (M=30) as the control group (M=90) (Diff=-61, z=2.8, p=.003), Normal approximation, 95% CI -105 to -19, Nonparametric Bootstrap: 95%CI -110 to -19). Furthermore, the authors reported that the difference between the two groups seemed to widen between after the active part of the intervention ended (i.e. between 9 and 18 months). This study also reports the number of non-homelessness nights during the study period (mean number of days reported each month up to 18 month follow-up). The intervention group reported more nights in housing (not homeless) (M= 508.0, SD=60) than the control group (M=450, SD=139) (MD=58, t=2.64, df=64, p=0.01).
In the third study (72), participants were followed for 15 months. The authors reported the length of time to leave shelter, and the number of days before moving into stable housing. Reports were given using a structured residential follow-back instrument. More families in the intervention group (N=97) left shelter than in the control group (N=113), and the transition from shelter to housing occurred faster with the intervention group. The intervention group used a mean number of 91.25 days (SD=82.3) to first move into stable housing compared to a mean of 199.15 days (SD=125.4) for control group participants. The majority of the intervention group was rehoused after two to three months compared to five months for the control group.
The results and quality assessments for critical time intervention compared to usual services are summarized in Table 10. The complete GRADE evidence profile is shown in Appendix 8, Table 8.1.8.
Summary of findings for the effects of critical time intervention vs usual services (Herman 2011, Samuels 2015, Susser 1997)
Risk of selection bias and performance bias.
Fewer than 300 participants.
Risk of selection bias in one study. Risk of performance bias in both studies. Inadequate reporting of randomization and allocation concealment methods in one study.
Fewer than 400 participants.
Risk of performance bias. Inadequate reporting of randomization and allocation concealment methods.
Inadequate reporting of blinding methods. Risk of reporting bias.
What does the evidence say?
Critical time intervention compared to usual services for adults with mental illness: May lead to little or no difference in the number of people who experience homelessness (low certainty evidence). May lead to fewer days spent homeless (low certainty evidence). May lead to more days spent not homeless (low certainty evidence). May reduce the amount of time to leave a shelter (and move to independent housing) (low certainty evidence).
Category 2: Abstinence-contingent housing programs
Description of the included studies
We found six studies with eight comparisons on the effects of abstinence-contingent housing programs(26, 58, 66-68, 75). All of the included studies were conducted in the USA. The data for the included studies were collected between 1991 and 2004. Within the category of abstinence-contingent housing programs, we identified three subcategories (see Table 11).
Overview of abstinence-contingent housing programs comparisons
The above interventions are compared to usual services, or other interventions. That is, abstinence-contingent housing is compared to another active intervention. Table 12 presents an overview of the populations, interventions, comparisons and outcomes in the six included studies. In some studies the duration of the intervention is reported and differs from the longest follow-up point. In these instances we have reported both the duration of the intervention and the longest follow-up point.
Description of studies that evaluated effects of abstinence-contingent housing
Description of the intervention
All of the interventions in the included studies had some component of abstinence-contingent housing. Abstinence-contingent housing in the included studies consisted of program-provided housing for a set period of time (6-8 months) with or without some rent contributed by the participants after the initial phase. Conditionality of tenancy for the participants consisted of a contract agreeing to abstinence and then regular urine testing to screen for substance use. Housing for participants was not segregated (segregated housing is separated from the general public and only for individuals receiving social assistance).
Category 2A: Abstinence-contingent housing with case management
We found one study (26) with two comparisons that examined the effect of abstinence-contingent housing with case management in the USA. Participants were recruited from 1991 to 1992 and randomized to one of three groups: abstinence-contingent housing with the progressive independence model of case management (ACH+CM), the progressive independence model of case management only (CM), or usual services (US).
The abstinence-contingent housing component consisted of supported housing in low-income apartment blocks where tenancy was contingent upon following program rules (26). The case management component in this study was described as a “progressive independence model” with a focus on providing immediate tangible resources while supporting further treatment for substance abuse and other relevant problems. Case management was also contingent on following a contract which participants signed before the start of the intervention.
Participants in the case management condition received an average of three months care, while participants in the housing with case management condition received an average of six months of care.
Abstinence-contingent housing with case management was compared to usual services (26)and case management only (26). Usual services consisted of aftercare services such as referrals to outpatient or inpatient substance abuse agencies or welfare offices.
2.A.1. Abstinence-contingent housing with case management compared to usual services
One study (26)examined the effect of abstinence-contingent housing with case management compared to usual services.
Primary outcome: Housing stability
Results from the included study(26)show that participants in the intervention group reported more days in housing than participants in the control groupat the 12 month follow-up interview (MD=6.4, 95%CI= 6.18 to 6.62). The results for abstinence-contingent housing with case management compared to usual services only are presented in Table 12. The results are controlled for length of time from baseline to the second follow-up interview, which varied due to difficulties arranging meetings with participants and the number of days in the relevant period spent in a controlled environment (e.g. prison or hospital) since they are not truly homeless or housed during this time. Other control variables such as characteristics which were found to vary across the treatment conditions are also controlled for (being recruited from a particular short-term program, reported perception of health problems at baseline, access to an automobile, having ever been married, having foster care experience as a child or having lived with one's mother continuously until 18). Not enough information was provided to present the results in a forest plot.
The results and quality assessments for abstinence-contingent housing with case management compared to usual services are summarized in Table 13. The complete GRADE evidence profile is shown in Appendix 8, Table 8.2.1.
Summary of findings table for the effects of abstinence-contingent housing with case management vs usual services (Sosin 1995)
Risk of selection bias and attrition bias. Inadequate reporting of blinding of participants, personnel and outcome assessors.
Fewer than 400 participants.
What does the evidence say?
It is uncertain whether abstinence-contingent housing with case management compared to usual services leads to a difference in number of days spent in in stable housing (very low certainty evidence).
2.A.2. Abstinence-contingent housing with case management compared to case management
One study (26)compared abstinence-contingent housing with case management to case management only.
Primary outcome: Housing
Results from this study (26)show that participants in the intervention group (N=108) reported a mean increase of 25.6 days housed of the previous 60 days from baseline to 12 months compared to a mean increase of 21.2 days for the comparison group (N=70). Not enough information was reported to determine if there is a difference between groups, or to present the results in a forest plot.
The results and quality assessments for abstinence-contingent housing with case management compared to case management only is summarized in Table 14. The complete evidence profile is presented in Appendix 8, Table 8.2.2.
Summary of findings table for the effects of abstinence-contingent housing with case management services vs case management (Sosin 1995)
Risk of selection bias and attrition bias. Inadequate reporting of blinding of participants, personnel and outcome assessors.
Fewer than 400 participants.
What does the evidence say?
It is uncertain whether abstinence-contingent housing with day treatment compared to case management only leads to a difference in the number of days spent in stable housing (very low certainty evidence).
Category 2B: Abstinence-contingent housing with day treatment
Three studies evaluated the effect of abstinence-contingent housing with day treatment in USA (58, 66, 67).
The abstinence-contingent housing with day treatment intervention consisted of two general components: housing programs in which tenancy is conditional upon maintained sobriety and/or treatment, and day treatment(58, 66, 67).
In one study with two comparisons (58),participants were required to pay to remain in housing (but were not removed if unable to pay). The housing component in this study was only part of treatment and available for a maximum of six months. No information was available regarding segregation of the housing or whether it was individual or group housing.
In the second study (67),participants’ tenancy in program management housing was contingent on abstinence. No information was provided in this study regarding rent payment, or the form of housing provided.
In the third study (66),participants were moved into rent free and furnished housing provided by the program after achieving abstinence. Participants in this study received segregated group or individual housing. After phase I half of the clients remained in this housing arrangement, and half moved to program-managed individual houses.
In these three included studies(58, 66, 67), participants in the intervention group received day treatment in the first phase of a two phase intervention. The second phase of the intervention included abstinence-contingent work therapy with minimum wage which could be used towards rent payments. Some participants also received aftercare (58)(66). Formal treatment ended after six months (58, 67, 68).
Participants in the comparison groups received usual services (67), day treatment only (58, 66), or non-abstinence-contingent housing with day treatment (58).
2.B.1. Abstinence-contingent housing with day treatment compared to usual services
One study compared abstinence-contingent housing with day treatment to usual services (67).
Primary outcome: Homelessness
Results from the included study (67) showed that participants in the intervention group reported a mean of 52 fewer days homeless in the previous 60 days at 12 month follow-up than in the previous 60 days at baseline. There was no change in number of days homeless for the control group.
The results and quality assessments for abstinence-contingent housing with day treatment compared to usual services is summarized in Table 15. A complete GRADE evidence profile is shown in Appendix 8, Table 8.2.3.
Summary of findings table for the effects of abstinence-contingent housing with day treatment vs usual services (Milby 1996)
Risk of performance bias and attrition bias.
Less than 400 participants.
What does the evidence say?
Abstinence-contingent housing with day treatment compared to usual services may lead to fewer days spent homeless (low certainty evidence).
2.B.2. Abstinence-contingent housing with day treatment compared to day treatment
Two studies (58, 66)examined the effect of abstinence-contingent housing with day treatment compared to day treatment.
Participants in the comparison groups received day treatment only which was similar to the day treatment offered to the intervention group for months 1-2 and 3-6(58, 66). These participants were not offered housing.
Primary outcomes: Housing stability and homelessness
Two studies examined the effect of abstinence-contingent housing with day treatment compared to day treatment only on the number of days participants reported being housed during a period of time. In one study (66),participants in both groups reported a greater number days housed of the previous 60 days at 12 months compared to baseline; however, participants in the intervention group showed a greater increase than participants in the control group (MD=18.7 (SE=3.9) compared to MD=16.2 (SE=3.5)) (MD=2.50 [95%CI 1.28 to 3.72]). In the other study (58),participants in the intervention group reported a greater increase in number of days housed from baseline to 12 months (M=17.7 (SD=33.8)) than participants in the control group (M=9.5 (SD=31.0)) (MD=8.20 [95%CI 5.74 to 10.66]). 1
When the results were pooled using SMD, I2=86%. Since this heterogeneity could not be explained, we chose not to pool the results (Figure 8).
Days in stable housing, 12 months, abstinence-contingent housing with day treatment vs day treatment only
The results and quality assessments for abstinence-contingent housing with day treatment compared to day treatment only is summarized in Table 16. The complete GRADE evidence profile is shown in Appendix 8, Table 8.2.4.
Summary of findings table of the effects of abstinence-contingent housing with day treatment vs day treatment (Kertesz 2007; Milby 1996)
Risk of performance bias and attrition bias. Inadequate reporting of allocation concealment in both studies.
Less than 400 participants.
What does the evidence say?
It is uncertain whether abstinence-contingent housing with day treatment compared to day treatment only leads to a difference in number of days spent in stable housing or employed (very low certainty evidence).
2.B.3. Abstinence-contingent housing with day treatment compared to non-abstinence-contingent housing with day treatment
One study examined the effects of abstinence-contingent housing with day treatment compared to non-abstinence contingent housing with day treatment (58). The comparison group consisted of an equivalent intervention as the abstinence-contingent housing group; however, continued tenancy was not dependent on sobriety (i.e. the results of the urine tests). Both groups received the day treatment component.
Primary outcome: Housing stability
Results from this study (58)showed that participants in the intervention group reported a greater increase in the number of days in stable housing in the 60 days prior to follow-up between baseline and follow-up (12 months) (MD=17.7 (SD=33.8)) than participants in the control group (MD=14.2 (SD=31.7)).
The results and quality assessments for abstinence-contingent housing day treatment compared to non-abstinence-contingent housing with day treatment for housing stability and homelessness are summarized in Table 17. A complete GRADE evidence profile is shown in Appendix 8, Table 8.2.5.
Summary of findings table for the effects of abstinence-contingent housing with day treatment vs non-abstinence-contingent housing with day treatment (Kertesz 2007)
Risk of performance bias and attrition bias. Inadequate reporting of allocation concealment.
Less than 400 participants.
What does the evidence say?
It is uncertain whether abstinence-contingent housing with day treatment compared to non-abstinence-contingent housing with day treatment leads to a difference in number of days spent in stable housing (very low certainty evidence).
2.B.4. Abstinence-contingent housing with day treatment compared to abstinence-contingent housing with community reinforcement approach
We found two studies that examined the effect of abstinence-contingent housing with day treatment compared to abstinence-contingent housing with community reinforcement approach (68, 75) in the USA.
Participants in one study (68) were provided with a furnished and rent free apartment and vocational training which was contingent on continued sobriety during phase I (weeks 2-8). In Phase II (weeks 3-24) participants were required to pay a small amount of rent (not specified) from program provided stipends. Participants who maintained abstinence were moved to a transitional housing program. In Phase III (week 25-end) continued tenancy in abstinence-contingent program housing was only available when space was available at a modest rent.
In the other study (75),participants were housed in grant-supported housing for a maximum of three months contingent on sobriety. However, participants who had secured a job and saved a pre-set amount of money could stay one additional month.
In both studies, participants in the comparison groups received the same abstinence-contingent housing, vocational training and work therapy as participants in the intervention group, with the community reinforcement approach in addition.
Primary outcomes: Homelessness and stable housing
Homelessness was reported in one study(75). The rate of homelessness for participants in the intervention group (N=64; 13.7%) was lower than for the control group (N=42; 34%) at four months. There was little or no difference between groups (when reported at all) at the other follow-up points.
Two studies reported outcomes related to stable housing. In one study (75), more participants from the CRA group (62.5%) were paying for housing (rather than staying with friends or in a motel) at the 12 months follow-up than in the day treatment group (44%) (χ2(1, N=80)=2.73, p<0.10).
In the second study (68),a greater proportion of participants in the abstinence-contingent housing with CRA group (N=103; 44.7%) were housed more than 40 of the previous 60 days at 18 months than in the abstinence-contingent housing with day treatment group (N=103; 35.6%). There was also a greater increase in proportion of participants housed 40 of the previous 60 days from baseline to 18 months in the CRA group (36%) than in the day treatment group (25.7%).
Not enough data were reported to assess whether there was a difference in time spent in stable housing between the two groups. Furthermore, the outcomes were reported too differently in the two studies to pool the results.
The results and quality assessments for abstinence-contingent housing with day treatment compared to abstinence-contingent housing with community reinforcement approach is summarized in Table 18. A complete GRADE evidence profile is shown in Appendix 8, Table 8.2.6.
Summary of findings table for the effects of abstinence-contingent housing with day treatment versus abstinence-contingent housing with community reinforcement approach (Smith 1998; Milby 2010)
Risk of detection and selection bias. Inadequate reporting of allocation concealment methods.
Less than 400 participants.
Risk of detection bias, selection bias, and performance bias.
What does the evidence say?
It is uncertain whether abstinence-contingent housing with day treatment compared to abstinence-contingent housing with community reinforcement has an effect on the amount of time spent homeless or in stable housing (very low certainty evidence).
Category 3: Non-abstinence-contingent housing
Description of the included studies
We identified eight studies that evaluated the effect of non-abstinence-contingent housing (24, 42, 43, 55, 58, 65, 73, 78). Most of the included studies were from the USA (N=6); however, the largest study was from Canada (N=1). Data for the included studies were collected between 1997 and 2013. Within the category of non-abstinence-contingent housing programs, we identified two subcategories (see Table 19).
Overview of non-abstinence-contingent housing program comparisons
These interventions are compared to usual services or other interventions. Table 20 presents an overview of the populations, interventions, comparisons and outcomes in the included studies.
Description of studies that evaluated effects of non-abstinence-contingent housing
Description of the intervention
Non-abstinence-contingent housing includes a variety of interventions that provide housing to homeless persons without any conditionality attached to their stays (such as abstinence, treatment attendance, etc.). Four of the included studies examined Housing First (with case management), which encourages early placement in stable housing after staying in transitional housing for a short period of time. The other studies examined supportive housing with assertive community treatment(65), staffed group homes with intensive case management(55), and non-abstinence-contingent housing with day treatment (58).
Category 3A: Housing First
We found four studies that evaluated the effect of Housing First (24, 42, 43, 78). In Housing First treatment and housing domains are considered as being closely linked, but separate domains. In other words, treatment is encouraged, but refusal does not result in removal from housing. The emphasis in Housing First is on consumers’ choice (i.e. the consumer helps to define and plan goals). A central component is that housing is immediately provided if desired, and tenancy is not contingent on adherence to treatment schedules or sobriety. All four studies had two program requirements: tenants had to pay part (30%) of their income (usually Supplemental Security Income) toward the rent by participating in a money management program, and tenants had to meet with a staff member regularly.
One study had three intervention arms and compared two models of the Housing First program (Pathways to housing and Consortium) to usual services (78). As part of the Housing First interventions, participants were offered the ACT model of case management which involves intense case management with a team of professionals that are available 24 hours a day, seven days a week(24, 42, 78). Participants in the second study (24)received the Pathways to Housing model which adds modifications to standard ACT: a nurse practitioner was added to the team to address health problems, and a housing specialist joined the team to coordinate the housing services (24). In the third study(42),participants were divided according to mental health needs (high or moderate) and while the high needs participants received ACT, the moderate needs participants received intensive case management together with Housing First. In the fourth study (43),participants in the intervention group received Housing First with case management (case managers had less than 20 clients each).
Housing in the included studies was provided as group living arrangements or apartments at single and scattered sites (43), or scattered sites only (24, 42, 78).
The intervention was compared to usual services (42, 43, 78), or abstinence-contingent housing (24).
3.A.1. Housing First compared to usual services
Three studies (42, 43, 78) examined the effect of Housing First on housing stability and homelessness compared to usual services in Canada (42) and the USA (43, 78).
In all three studies the intervention was compared to usual services. Usual services included having access to other housing and support services through other programs in their communities. In one study, however, (78)two groups of participants received a version of the Housing First intervention - either the Pathways to Housing model which is a well-established model, but new to this particular community, or the Consortium model, which was made up of a consortium of treatment and housing agencies who had no prior experience of operating Housing First (78). The authors also report differences between these two groups.
The included studies reported number of days homeless, in shelter, in respite care, with family/friends, or in paid housing (43), proportion of time homeless (in shelters or on street) and stably housed (42), housing stability (proportion of time housed) (42),and number of participants in stable housing at end of study (78).
Primary outcomes: Housing stability and homelessness
The first study (42) examined housing stability in two ways: proportion of time during the last 6 months of the study that participants reported being housed all of the time, some of the time or none of the time, and percentage of days spent in stable housing for each three month period of follow-up. Sixty-two percent of participants receiving Housing First reported being housed all of the time compared to 31% of participants who received usual services; 22% of Housing First participants were housed some of the time and 16% none of the time compared to 23% and 46%, respectively, of usual services participants. For the second outcome, Housing First participants were in stable housing an average of 73% of the time compared to an average of 32% the time for participants who received usual services. We were not able to calculate difference between groups due to insufficient reporting of results in the primary study.
This study (42)also reported proportion of time in different types of shelter over the study period: Participants in the Housing First group spent approximately 12% of time in temporary housing, 6% in emergency shelters, 9% in institutions and 3% on the street compared to participants in the usual service group who spent approximately 33% of time in temporary housing, 16% in emergency shelters, 11% in institutions and 8% on the street. We were not able to calculate difference between groups due to insufficient reporting of results in the primary study.
In the second study (43),number of days homelessness was reported at each three month interval follow-up point and accumulated over the 18 month study period. The results were then annualized (converted to a rate for one year). Participants in the Housing First group reported fewer days homeless than participants in the usual services group ((MD=-62.3 (SE=12.4), p<0.05) and more days in paid housing (MD=109.9 (SE=8.7), p<0.05) at 18 month follow-up.
In the third study(78),103 of 209 participants in the Housing First group were placed in permanent housing at the 20 month follow-up compared to 13 of 51 participants in the case management only group.
We were unable to pool results from the included studies due to difference in how the outcomes were reported.
The results are shown in Table 21.
Results for Housing First vs usual services on housing stability and homelessness
The results and quality assessments for Housing First compared to usual services are summarized in Table 22. The complete GRADE evidence profile is shown in Appendix 8, Table 8.3.1.
Summary of findings table for the effects of Housing First with case management compared to usual services (Aubry 2015, Basu 2012, Stefancic 2007)
Risk of performance bias.
Risk of performance bias and detection bias.
Risk of selection bias and attrition bias.
Fewer than 300 participants.
What does the evidence say?
Housing First compared to usual services: Probably reduces the number of days spent homeless (moderate certainty evidence). Probably reduces the proportion of time an individual spends homeless (moderate certainty evidence). Probably increases the number of days in paid housing (moderate certainty evidence). Probably increases the proportion of time in stable housing (moderate certainty evidence). May increase the number of people placed in permanent housing after 20 months (low certainty evidence).
Subgroup analysis
In one study participants were stratified according to mental health needs (42). The authors conducted sub-group analyses where participants with high support needs for mental health services (high needs) and participants with moderate support needs for mental health services (moderate needs) were examined separately(42). All five sites are included in the high needs analysis, but only four sites are included in the moderate needs analysis because one site did not separate participants according to need level.
High needs participants received Housing First with Assertive Community treatment while moderate needs participants received Housing First with intensive case management. Both groups were compared to participants who received usual services. For participants with high support needs, those receiving Housing First with assertive community treatment reported a greater mean proportion of time in stable housing over the 24 month study period (71%) than the control group (29%) (adjusted absolute difference AAD=42%, 95%CI 28% to 45%, p<0.01)(42).
For participants with moderate support needs, those receiving Housing First with intensive case management had a higher proportion of days stably housed than the control group across all four included study sites (a summary statistic for the total group of participants across sites was not reported).
Stefancic 2007 (78)also examined the difference between the two models of Housing First included in the study in number of clients placed in permanent housing. Sixty two of 105 participants in the Pathways to Housing group were placed and 52 of 104 in the Consortium group were placed. Housing retention rates were also reported for all participants: at the two-year follow-up point 84% of Housing First participants were housed compared to 88.5% of control group participants and after 47 months 68% were still housed compared to 78.3% of control group participants. Results of housing retention between the two Housing First groups shows that 88.5% of Pathways participants were still in housing compared to 79% of Consortium participants and 88.5% after two years and 78.3% of Pathways participants were in housing, 57% of Consortium participants after 47 months.
3.A.2. Housing First compared to abstinence-contingent housing
One study (24)examined the effect of Housing First compared to abstinence-contingent housing on homelessness in New York, USA.
Primary outcomes: Homelessness and housing stability
As the results indicate, the proportion of time participants spent homeless (public space, on the street or in shelter) was recorded at each 3 month follow-up period over the course of the study. The Housing First group (N=103) reported less time homeless (F(1, 195)=198, p<0.0001) and more time spent stably housed compared to the usual services group (N=103) at all time points. Housing First participants also reported faster decreases in number of days spent homeless (F(4,137)=10.1, p<0.001) and increases in stably-housed status (F(4,137)=27.7,p<0.001) compared to the usual services group.
The results and quality assessments for Housing First compared to abstinence-contingent housing are summarized in Table 23. The complete GRADE evidence profile is shown in Appendix 8, Table 8.3.2.
Summary of findings table for the effects of Housing First vs abstinence-contingent housing (Tsemberis 2004)
Risk of detection bias and attrition bias. Inadequate reporting of randomization, allocation concealment and blinding of participants and personnel.
Fewer than 400 participants.
What does the evidence say?
It is uncertain if Housing First has an effect on homelessness or housing stability when compared with abstinence-contingent housing (very low certainty evidence).
Category 3B: Non-abstinence-contingent housing with treatment
We identified four studies that examined the effect of non-abstinence-contingent housing with some form of treatment (case management or day treatment) (55, 58, 65, 73). The studies were conducted in the USA. The interventions in these studies included provision of housing to participants in the treatment group that was not conditional on maintaining sobriety or attending treatment.
One study (55)compared non-abstinence-contingent housing in the form of group living arrangement versus independent living. Participants in both groups received housing and some form of case management (intensive case management with house staff for those assigned to group living arrangements and assertive community treatment for participants in the independent living group) (55). Participants in the intervention group could be assigned to one seven group homes which accommodated between six and ten participants and had shared amenities but separate bedrooms. The staffing patterns were similar to traditional group homes with live-in staff. The participants had an intensive case manager they met with at least once a week. They paid 30% of their income to cover rent and utilities and were encouraged to attend activities at community mental health centres (55).
In the second study (73),participants in the intervention group were offered temporary program managed shelter as well as intensive case management. Only program participants were housed in the shelter. The research team eventually began to develop their own housing as well. Shelter stay was not contingent on treatment or sobriety; however, a small group of participants were eventually required to enter a payee arrangement due to lack of progress and using their income for drug purchases(73).
In the third study (65),the intervention was described as “parallel housing” where participants are offered housing from “mainstream” (i.e. not segregated) options that were owned and operated by community landlords or housing agencies. Participants lived independently and their tenancy was not conditional on treatment participation. The participants are also offered assertive community treatment with high intensity (low client to case manager ratio and case managers are available 24 hours every day).
In the fourth study (58),participants in the intervention group received non-abstinence contingent housing with day treatment (58). The non-abstinence-contingent housing with day treatment intervention consisted of two components: housing programs in which tenancy is not conditional upon maintained sobriety and/or treatment, and day treatment. Participants were required to pay to remain in housing (but were not removed if unable to pay). The housing component was only part of treatment and available for a maximum of six months. No information was available regarding segregation of the housing or whether it was individual or group housing. Participants in the intervention group also received day treatment in the first phase of a two phase intervention. Day treatment lasted between 6.25 hours daily for the first two months of the study. Phase II of the intervention included abstinence-contingent work therapy with minimum wage. Some participants also received aftercare. Formal treatment ended after six months.
The intervention was compared to usual services (73), non-abstinence-contingent housing in independent apartments (55),“integrated housing” (65), or day treatment (58).
3.B.1. Non-abstinence-contingent housing with high intensity case management compared to usual services
One study (73) evaluated the effect of non-abstinence-contingent housing with high intensity case management compared to usual services on housing stability, homelessness, quality of life and psychological status.
Control group participants were offered usual services provided by the city.
Primary outcome: Homelessness
One study (73)evaluated the effect of non-abstinence-contingent housing on homelessness and housing. The rate of decline in amount of time spent living on the streets over the 24 months study period was almost twice as great for the intervention group (MD=-54.9 (SD=36.9) that the control group (MD=-28.2 (SD=44.5)) (t=4.18, p=0.001). Individuals in the intervention group reported more time in shelters, specifically the program provided respite housing than the control group (MD=23.1 (SD=29.27 compared to MD=2.8 (SD=15.23), p=0.001). While participants in both groups increased the time spent in community housing (including transitional settings, long-term settings), the rate of increase was almost twice as great for the intervention group (MD=21.0 (SD=30.39)) than the control group (MD=9.9 (SD=32.34)) (t=-2.27, p=0.025). At the final follow-up point 38% of the intervention group were in community settings compared to 24% of the control group.
The results and quality assessments for non-abstinence-contingent housing with high intensity case management compared to usual services are summarized in Table 24. The complete GRADE evidence profile is presented in Appendix 8, Table 8.3.3.
Summary of findings table for the effects of non-abstinence-contingent housing with high intensity case management vs usual services (Shern 2000)
Inadequate reporting of allocation concealment measures and blinding.
Fewer than 400 participants.
What does the evidence say?
Non-abstinence-contingent housing with high intensity case management compared to usual services: May lead to greater decrease in proportion of time spent homeless or in shelter (low certainty evidence). May increase the amount of time in community living arrangements (low certainty evidence).
3.B.2. Non-abstinence-contingent group living arrangements with high intensity case management compared to non-abstinence-contingent independent apartments with high intensity case management
One study (55)evaluated the effect of non-abstinence-contingent group living arrangements with high intensity case management (NACHG) compared to non-abstinence-contingent independent apartments with case management (NACHI) on housing stability, homelessness, and satisfaction with life.
Participants in the comparison group were placed in non-abstinence-contingent independent apartments. These apartments were efficiency units operated by the local housing authority and participants were offered voluntary weekly group meetings, but not other programming on-site.
Primary outcomes: Housing stability and homelessness
The included study examined the effect of non-abstinence-contingent group living arrangements on the number of days homeless during the study period and number of days homeless after rehousing (55). A total of 110 participants were included in the analysis for outcomes measured at final follow-up (18 months) (intervention N=61; comparison N=49). There was little or no difference in housing status between groups at 18 months. Participants in the intervention group reported a mean of 43 days homeless over 18 months compared to a mean of 78 days for the control group. We could not calculate the difference between groups due to inadequate reporting in the primary study.
The results and quality assessments for non-abstinence-contingent group living arrangements with high intensity case management compared to non-abstinence-contingent independent apartments with high intensity case management are summarized in Table 25. The complete GRADE evidence profile is shown in Appendix 8, Table 8.3.4.
Summary of findings table for the effects of non-abstinence-contingent group living arrangements with high intensity case management compared to non-abstinence-contingent independent apartments with high intensity case management
Inadequate reporting of randomization, allocation concealment and blinding.
Fewer than 300 participants.
Wide confidence interval.
What does the evidence say?
Non-abstinence-contingent group housing with high intensity case management compared to non-abstinence-contingent independent apartments with high intensity case management Maylead to fewer days homeless after being rehoused and reduce the number of days spent homeless (low certainty evidence). It is uncertain if the intervention has an effect on housing status at 18 months (very low certainty evidence).
3.B.3. Non-abstinence-contingent housing with high intensity case management compared to abstinence-contingent housing with high intensity case management
One study (65) evaluated the effect of non-abstinence-contingent housing with high intensity case management compared to abstinence-contingent housing with high intensity case management on housing stability and homelessness in USA.
In this study (65),the intervention (“parallel housing”) was compared to “integrated housing”. The main difference according to the researchers is (1) housing control: integrated housing is owned or leased by the mental health provider; (2) integration within the community: parallel housing is not segregated housing units while integrated housing is; (3) conditionality: integrated housing is often linked to treatment participation, and (4): live-in staff: integrated housing sometimes contain live-in staff.
Primary outcomes: Homelessness and housing stability
The included study(65)reported proportion of time functionally homeless (a term used by primary authors to describe both time literally homeless and days in temporary or institutional settings that are preceded and followed by days homelessness) and housing stability (stable housing defined by authors as one's own apartment/house, single room occupancy with or without services, family or friends’ house on a long-term basis, boarding house, transitional housing or a group home).
Only 121 participants took part in either the intervention (N=60) or the comparison group (N=61). Participants in both groups reduced the number of days functionally homeless from baseline to 18 months, however there was a greater change in number of days homeless among members of the comparison group over the study period (F=6.07, p<0.05, d=-0.52). At the end of the study 68.1% of participants in the intervention group were in stable housing compared to 85.5 % of comparison group participants (F=5.99, p<0.05, d=0.51).
The results and quality assessments for non-abstinence-contingent housing with high intensity case management vs abstinence-contingent housing with high intensity case management are summarized in Table 26. The complete GRADE evidence profile is shown in Appendix 8, Table 8.3.5.
Summary of findings table for the effects of non-abstinence-contingent housing with high intensity case management vs abstinence-contingent housing with high intensity case management (McHugo 2004)
Risk of attrition bias. Inadequate reporting of randomization, allocation concealment and blinding.
Fewer than 400 participants.
What does the evidence say?
It is uncertain whether non-abstinence-contingent housing with high intensity case management compared to abstinence-contingent housing with high intensity case management has an effect on housing stability (very low certainty evidence).
3.B.4. Non-abstinence-contingent housing with day treatment compared to day treatment
One study (58)evaluated the effects of non-abstinence-contingent housing with day treatment compared to day treatment only on housing stability and homelessness in the USA.
Participants in the control condition received day treatment only with no provision of housing.
Primary outcome: Housing stability
The included study (58)reported housing stability as the change in amount of time spent in stable housing from baseline to follow-up. Complete data 12 months post-baseline is only available for 116 participants (intervention N=43, comparison N=34). Although participants in both groups increased the amount of time spent in stable housing from baseline to the final follow-up, the intervention group showed greater gains (MD=14.2 (SD=31.7) than the comparison group (MD=9.5 (SD=31.0)) (MD=4.70 [95%CI-9.38 to 18.78]).
The results and quality assessments for non-abstinence-contingent housing with day treatment compared to day treatment are summarized in Table 27. The complete GRADE evidence profile is presented in Appendix 8, Table 8.3.6.
Summary of findings table for the effects of non-abstinence-contingent housing with day treatment vs day treatment (Kertesz 2007)
Risk of selection bias and attrition bias. Inadequate reporting of allocation concealment methods and blinding.
Fewer than 300 participants.
Wide confidence interval.
What does the evidence say?
It is uncertain whether abstinence-contingent housing with day treatment compared to day treatment only leads to more days in stable housing (very low certainty).
Category 4: Housing vouchers with case management
Description of included studies
We identified four studies with five comparisons that evaluated the effect of housing vouchers with case management (27, 62, 71, 81).
Table 28 presents an overview of the populations, interventions, comparisons and outcomes in the included studies.
Description of studies that evaluated effects of housing vouchers
Description of the intervention
Housing vouchers for the purpose of this review is interventions where the housing component is limited to the provision of financial assistance for housing of the participants choosing. Case management is described above (Category 1).
In the first study (27), 362 participants were assigned to one of four groups: comprehensive case management or traditional case management with or without HUD Section 8 housing certificates (financial assistance). A preliminary analysis of the between group differences showed no correlation between the case management model and housing outcomes, so further analysis was based on the Section 8 housing certificate condition. Therefore the groups were analyzed as following: Comprehensive or traditional case management with HUD Section 8 housing certificates compared to comprehensive or traditional case management without HUD Section 8 housing certificates. Participants in each condition received a range of case management services varying in intensity (time between contact with case managers), case load of case managers (1:22 up to 1:40), and availability (comprehensive case managers were constantly available). The HUD Section 8 housing certificate is a program allowing holders to pay a fixed 30% of their adjusted income for a private rental unit of their choosing. There are no conditions on the tenancy except for that the housing must meet the quality standards of the US Department of Housing and Urban Development and the rent for the unit must be equal or less than fair market rent for the area. The participants in this program received a tailored version of the certificate program with more flexible rules (for example keeping appointments) and with support from housing specialists who assisted with the application process and were sensitive to limitations imposed by severe mental illness.
In the second study (62),participants in the intervention group were enrolled in the Home to Stay program. The Home to Stay model was designed to quickly put families into housing and maintain the housing using a time-limited financial subsidy and temporary support services. At the beginning of the study participants could access 1 year Advantage housing subsidies (three types of locally funded subsidies intended for families with children, clients with disability payments, or employed clients). After three months, clients (participants) were required to contribute 30% of their monthly income and eligibility was restricted to employed (or receiving federal disability payments) adults with children. At the one year mark these subsidies were no longer available for new families and two years after the study began the monthly payments were terminated for all recipients. Initial services in the Home to Stay program was to help families’ secure permanent housing and exist shelter quickly. After they were placed in housing, there was a focus on obtaining employment (income) equal to double the family's rent obligation and/or obtaining a permanent housing subsidy. Participants in this group also received fairly intensive case management services while in shelter. The intervention condition was different than the usual services condition specifically with respect to more frequent case manager contact, smaller caseloads, flexible scheduling, integrated help with financial literacy and continuing the services from shelter into housing.
In the third study (71) the US Department of Housing and Urban Development allocated funds for 1000 vouchers for a program providing housing and case management for literally homeless veterans with mental illness or substance dependence. These participants were offered priority access to the Section 8 housing vouchers (difference between 30% of their adjusted income and the lesser of Fair Market Rent or the unit rent). Case managers put the veterans in contact with the local housing voucher and helped them to locate an apartment, negotiate the lease, furnish and move into the apartment. The case management component was a modified assertive community treatment model (larger caseloads and encouragement of clients to use other Veteran Affairs health services). The intervention was compared to usual services and case management. Participants in the comparison conditions received standard Veteran Affairs homeless services, including short-term brokered case management, or intensive case management.
In the fourth study (81), participants living with HIV/AIDS were provided with long-term rental housing assistance. The amount was determined by The Department of Housing and Urban Development (HUD) annually for each metropolitan area. Each person receiving rental assistance was required to pay 30% of this monthly adjusted income. Study-funded housing referral specialists assisted with finding housing and negotiating leases and participants received referrals to other supportive services.
The interventions were compared to usual services (62, 71, 81), case management (27), or high intensity case management (71).
4.1. Housing vouchers with case management compared to usual services
Two of the three studies that compared housing vouchers with case management to usual services (62, 71, 81)included multiple cities (71, 81). One study included families (62) and one study included adults living with HIV/AIDS.
Primary outcomes: Time to exit shelter, stable housing, homelessness
Three of the included studies evaluated the effect of housing vouchers compared to usual services on housing stability and homelessness(62, 71, 81). The studies measure and report these outcomes in such different ways that we are unable to pool results. The following is a narrative summary of the results from the three studies.
In the first study (62), the authors included work-based subsidies as a covariate in all analyses of differences between the intervention group (N=138) and the control group (N=192). A survival analysis using Cox regression of time to first exist from shelter (at least 30 days away from shelter) shows that the intervention group experienced fewer days to exit shelter (x2 1 = 6.068, 95%CI = 0.589 to 0.942; proportional hazards assumption not violated). The authors also report the time to return to shelter (overnight stay) for those that did return (N=298) and that the intervention group reported longer time to return to shelter than the control group (x2 1 = 6.524, 95% CI = 0.379 to 0.880; proportional hazards assumption not violated).
In the second study (71), data for 182 participants in the intervention group and 188 participants in the control group were reported related to number of days housed during the 90 days prior to each follow-up. We report the longest follow-up at 36 months. The intervention group reported more days housed (M=59.39) compared to the control group (M=47.60) (t=4.88, p<0.001). The intervention group also reported fewer days homeless (M=13.05) than the control group (M=20.45) (t=3.56, p<0.001).
In the third study (81), the authors reported the number of participants in their own home, the number living temporarily with others or in transitional settings, or the number with one or more nights homeless during the 90 days prior to follow-up for the intervention group (N=315) and the control group (N=315). At the 18 month follow-up interview there were more people from the housing vouchers group living in their own home (82.48) than the control group (50.58), fewer people in the housing vouchers group living temporarily with others or in transitional settings (14.96) than the control group (44.40) and half as many who reported being homeless at least once during the previous 90 days (2.55) than the control group (5.02). It is not possible to calculate the effect size due to lack of information reported in the results from the primary study.
The results and quality assessments for housing vouchers with case management compared to usual services are summarized in Tables 29. The complete GRADE evidence profile is shown in Appendix 8, Table 8.4.1.
Summary of findings table for the effects of housing vouchers with case management vs usual services (Levitt 2013, Wolitski 2010, Rosenheck 2003)
Risk of performance bias and detection bias. Inadequate randomization methods.
One small study.
What does the evidence say?
Housing vouchers with case management compared to usual services for homeless families: May reduce the number of days it takes to leave tempoary shelters and increase the number of days before returning to temporary shelters (low certainty evidence). May increase the number of days in stable housing and reduce the number of days spent homeless (low certainty evidence). May increase the proportion of people living in their own house, reduce the proportion of people who experience at least one night of homelessness and reduce the proportion of people who live in transitional settings at 18 month follow-up (low certainty evidence).
4.2. Housing vouchers with case management compared to case management only
We identified two studies that examined the effect of housing vouchers with case management compared to case management (27). The case management component of the intervention varied in intensity. In one study, participants received either comprehensive (high intensity) case management or traditional (low intensity case management) in addition to the housing vouchers while the control group also received one of the two types of case management. Participants in the second study received high intensity case management. We have decided to combine the two studies under a broader heading of case management.
Primary outcomes
The first study(27)reported the type of housing maintained by participants, the number in stable housing and how many participants transitioned early (first six months of study) into independent or community housing (defined in this study as family or friend's home or a boarding/halfway house). Approximately twice as many participants in the intervention group maintained independent housing at the 24 month follow-up (104/181) compared to the comparison group (55/181) (RR=1.89 [95%CI 1.47 to 2.44]). Approximately four times as many participants in the comparison group (44/181) compared to the intervention group (11/181) reported living in community housing at 24 months (RR=0.25 [95% CI 0.13 to 0.47]). More participants in the comparison group were recorded as living in variable housing (unstable, institution, or disengaged from study) (82/181) compared to participants in the intervention group (66/181) (RR=0.80 [95%CI 0.63 to 1.03]). Finally, the authors also measured the proportion of participants who transitioned early into independent and community housing (the first 6 months). The authors reported that participants with housing vouchers stabilized in independent housing faster than participants in the comparison condition and were 8.4 times more likely to obtain independent housing in the first six months of the study (91/115 intervention group participants compared to 25/99 comparison group participants). On the contrary, the comparison group was 3.4 times more likely to obtain other types of community housing in the first six months (28/99 comparison group participants compared to 4/115 intervention group participants).
Results from the second study (71) show that the intervention group reported more (16.9%) days housed (M=59.39) compared to the control group (M=50.81) (t=2.90, <p=0.004) at 36 months. The intervention group also reported fewer days homeless (M=13.05) than the control group (M=20.33) (t=2.87; p=0.004) at 36 months.
The results and quality assessments for housing vouchers with case management compared to case management only are summarized in Table 30, and the complete GRADE evidence profile is shown in Appendix 8, Table 8.4.2.
Summary of findings table for the effects of housing vouchers with case management vs case management only (Hurlburt 1996, Rosenheck 2003)
Inadequate reporting of methods.
Risk of performance bias. Inadequate reporting of methods for blinding of outcome assessors.
Fewer than 400 participants.
What does the evidence say?
Housing vouchers with case management compared to case management only May increase the number of people living in independent housing and reduce the number of people living in community housing (low certainty evidence). May increases the number of days spent in stable housing and reduces the number of days spent homeless (low certainty evidence). May lead to no difference in the number of people living variable housing situations (low certainty evidence).
Category 5: Residential treatment
Description of included studies
We identified two studies that evaluated the effect of residential treatment (49, 63). Both studies were conducted in the USA. Table 31 presents an overview of the populations, interventions, comparisons and outcomes in the included studies.
Description of studies that evaluated effects of residential treatment
Description of the intervention
The two studies that evaluated the effect of residential care on homelessness and housing stability (49, 63). The interventions in the included studies are different due to the different populations which they target. In the first study (49), the intervention was divided into two phases: the residential phase (0-6 months) and the community phase (7-12 months). During the residential phase participants received case management services, treatment planning and service referral, counselling, and material assistance. During the community phase participants were placed in community living with continued case management and cognitive behavioural therapy and self-help groups such as Alcoholics and Narcotics Anonymous. Participants were followed up to 24 months, even though the active part of the intervention only lasted 12 months.
In the second study (63), participants were placed in a non-profit supportive housing program which used single rooms in an urban hotel. This permanent residence provided services such as a furnished room, case management, coordination of public assistance, medication and money management, meals, therapy and referrals to appropriate services. Both the treatment and the longest follow-up time was 12 months.
5.1. Residential treatment compared to usual services
We found two studies that evaluated the effect of residential care compared to usual services (49, 63).
While both studies compared the intervention to usual services, these services differed due to the different target populations in the studies. In the first study (49), the usual services was inpatient treatment in hospital wards for two to three weeks and included substance abuse education, therapy, self-help services, medical care, material assistance and referral to appropriate services. Customary community care was provided up to 12 months and included services as needed, half-way houses and mental health treatment for post-traumatic stress disorder.
In the second study (63), participants in the usual services condition received standard post-discharge care, of which one quarter of participants refused. No further information was provided on what this care entailed.
Due to the difference in population, intervention and comparison group characteristics we have not pooled the results. We present a narrative summary of the results from each study below.
Primary outcomes: Homelessness and stable housing
Both of the included studies reported the proportion of nights spent homeless ((49, 63). In the first study (49), participants in the intervention group (N=178) reported less homelessness than the control group (N=180) during the 60 days prior to the 24 month follow-up interview (11% compared to 2% for the control group) (Random effects regression estimate=0.104 (SE=0.037), Z=2.846, p=0.004). In the second study (63), participants in the intervention group reported less time homeless over the 12 month study period (6% SD=22 compared to 46% SD=51; t2=2.62, df=31, p=0.019). Furthermore, the authors report that during the study period, participants in the intervention group had a 13% chance of having 30 or more consecutive nights homeless compared to 39% for the control group (x2=87.46, df=1, p=0.01).
The first study (63) also reported the proportion of time participants reported being housed. Participants in the intervention group (N=26; 79%, SD=26) reported being in permanent housing more than twice as much as the control group (N=23; 33% SD=36) during the study year (t2=4.32, df=32, p=0.0001). Furthermore more than twice as many participants from the intervention group reported being in permanent housing at the 12 month follow-up interview (69% compared to 30%). Data was not reported for number of nights spent in shelter (63).
The results and quality assessments for residential treatment with case management vs usual services is summarized in Table 32. The complete GRADE evidence profile is shown in Appendix 8, Table 8.5.1.
Summary of findings table for the effects of residential treatment with case management vs usual services (Conrad 1998, Lipton 1988)
Risk of attrition bias, reporting bias in one study. Inadequate reporting of methods in both studies.
Inadequate reporting of methods.
Fewer than 400 participants.
What does the evidence say?
Residential treatment with case management compared to usual services: May reduce the proportion of nights spent homeless (low certainty evidence). May increase the proportion of time spent in stable housing (low certainty evidence). May increase the number of participants who are in stable housing after one year (low certainty evidence).
Discussion
In this systematic review we aimed to summarize empirical research assessing the effect of housing programs and case management on improving housing stability and reducing homelessness for individuals who are homeless, or are at-risk of becoming homeless. We included 43 randomized controlled trials with a total of approximately 10,570 participants. The majority of the studies included adult participants with mental illness and/or substance abuse. All of the studies were assessed as having high risk of bias. Five main groups of interventions were identified: case management, abstinence-contingent housing, non-abstinence-contingent housing, housing vouchers, and residential treatment. The interventions were compared to usual services or another intervention. In practice, this means that all participants received or had access to some type of service.
Within these groups, a total of 28 comparisons assessed housing stability and/or homelessness. In addition, many of the included studies also addressed secondary outcomes such as employment, physical or mental health, quality of life, social support networks, substance abuse and criminal activity.
Overall, the findings suggest that case management and housing programs are consistently more effective than usual services in reducing homelessness and increasing the amount of time spent in stable housing. It is difficult to conclude whether interventions which combine housing with case management are more effective than case management only since only one study included that comparison and this evidence was assessed as having very low certainty.
Discussion of main results
We included 24 studies that evaluated the effect of case management on housing stability and/or homelessness. Eligibility criteria in the majority of the studies included homeless adults or those at-risk of becoming homeless, with mental illness and/or substance abuse issues. Three studies included other populations (disadvantaged youth, recently released criminal offenders, and homeless adults with families). Case management is a broad term and includes an array of interventions. For the purpose of this review, we therefore categorized them into either high intensity, where the intervention was described as assertive community treatment or intensive case management, or low intensity, where the intensity was not specified, or where case managers met with participants less than weekly. These interventions were compared with either usual services, another type of case management (of varying intensity), or an intervention that included neither a case management nor a housing component (for example motivational enhancement therapy). Importantly, even comparison group participants who received usual services were offered some type of service, support or treatment. This means that all interventions were, in reality, compared to an active comparison group to some degree.
Case management
High intensity case management probably reduces by almost one-third the number of individuals with mental illness and/or substance abuse problems who report being homeless, and increases by about 25% the number in stable housing 12-18 months after services are initiated compared to individuals who are offered usual services. It probably leads to little or no difference in the number of people (with mental illness and/or substance abuse, or recently released criminal offenders) who experience some homelessness during a two year period. Furthermore, high intensity case management may lead to a lower mean number of days spent homeless compared to usual services for both adults with mental illness and/or substance abuse problems and homeless adults with families. Taken together these findings suggest that although individuals who receive high intensity case management are probably just as likely to experience some homelessness, overall it may be fewer days total. For this reason, at any given point in time (e.g. follow-up interview), individuals who receive high intensity case management are less likely to be homeless and more likely to be in stable housing, compared to individuals who are offered usual services.
When compared to low intensity case management, high intensity case management may lead to little or no difference in the number of days spent in stable housing or the number of participants who experience some homelessness.
For many of the outcomes, both the quantity and quality of available evidence was too limited to draw conclusions. Many of these outcomes are related to mean number of days in stable housing or homeless, longest residence, number of moves, number of people who report not moving, and the number of days in better or worse housing.
In summary, it appears as though high intensity case management is better than usual services, but not better than low intensity case management in improving housing stability and reducing homelessness for adults with mental illness and/or substance abuse problems and homeless adults with families. This is perhaps not surprising given the variation in how the case management interventions are designed and implemented. It may indicate that in practice there is not much difference with respect to intensity, for example, between high intensity (ACT and ICM) and low intensity case management interventions. Alternatively, it may suggest that having at least one individual (case manager) guiding and supporting a participant through the number of disjointed services may be more important than the degree of intensity of the intervention.
For the two comparisons which included young people or youth, the results showed that case management (high or low) compared to usual services or another intervention with no housing or case management component may lead to little or no difference in number of days spent homeless, the number who were homeless at follow-up or the number of moves experienced during a 12-month period. These results differ slightly from the comparisons which only included adults. Chamberlain and MacKenzie (2004) described the stages which youth go through before they are identified as homeless and argued for prevention and interventions which target these stages: 1) at-risk as identified by school counsellors, 2) runaways, 3) no longer belonging to the family, and 4) transition to chronicity where there are longer periods of homelessness(84).Chamberlain and MacKenzie (2004) argued that in the later stages, interventions with community placement components are necessary. Participants from both of the included studies (comparing high or low intensity case management to usual services or another intervention with no housing or case management component) included youth in the last stage (homeless or history of homelessness). In one study, the case management condition did not seem to include the community placement component, while in the other study, the comparison groups appeared to include equal or greater community placement components (CRA and MET). This could explain why there were no differences between the groups on housing stability or homelessness for this particular population. Alternatively, youth are often considered much more vulnerable and may just require more intensive case management services than even the high intensity case management models such as ICM and ACT which are intended for adults, currently provide.
Critical time intervention (CTI) may be more effective than usual services at improving housing stability and reducing homelessness for adults with mental illness. Even though individuals who receive CTI may be just as likely to experience some homelessness as individuals who receive usual services, they may spend fewer days homeless in total, and take half as long to leave shelter for stable or community housing.
Our findings are largely consistent with those from other reviews of case management for homeless populations(18, 20, 28). Coldwell and Bender (2007) also found that assertive community treatment reduced homelessness among populations with severe mental illness (28). Nelson and colleagues (2007) also found ACT and ICM to be superior to standard care for achieving housing stability among individuals with mental illness (20). Most recently de Vet and colleagues concluded that case management has a positive effect on homeless populations compared to standard care (18). Slesnick and colleagues (2009) summarized the research on youth homelessness and also concluded that comprehensive interventions that address youth and families, rather than single-issue interventions (such as case management), may be more successful with this particular population (30). This review included a wide variety of study designs and provided an overview of the studies rather than a synthesis of results.
However, our review differs from previous systematic reviews in five main ways: 1) we have included only randomized controlled trials, which are considered the best method for examining the effectiveness of an intervention; 2) we have only included studies which follow participants for at least one year; 3) we have grouped interventions according to low and high intensity and thus we have results for a larger group of interventions rather than individual models of case management (e.g. ACT, ICM); 4) we have pooled the results (continuous and dichotomous separately) where possible which has allowed us to look at the evidence across studies and not conclude based on small sample sizes from individual studies, and; 5) we have applied GRADE to the outcomes and thus provided a more concrete indication of our certainty in the evidence.
Abstinence-contingent housing
Abstinence-contingent housing combined with day treatment may reduce the number of days spent homeless when compared with usual services; however, we are uncertain of its effects on housing stability and homelessness when compared with other interventions due to very low certainty evidence. Furthermore, we are uncertain of the effects of abstinence-contingent housing with case management.
Non-abstinence-contingent housing
We identified two categories of non-abstinence-contingent housing: Housing First, and other programs that did not explicitly use the Housing First model.
The Housing First model probably improves housing stability and reduces homelessness compared to usual services. There are no previous systematic reviews that we are aware of that have specifically looked at the effects of Housing First on housing and homelessness. The results from this review indicate 1) that Housing First probably reduces homelessness and increases the number of days in stable housing among adults with mental or chronic medical illness; and 2) may double the number of participants placed in permanent housing within two years.
We are uncertain of the effects of Housing First when compared with abstinence-contingent housing due to very low certainty evidence. However, there are no indications that Housing First is less effective in reducing homelessness or improving housing stability.
The results discussed here are from studies conducted in the USA and in Canada. The consistency of the above results, which include multiple settings with diverse social welfare, political and economic settings, supports the idea that Housing First can work in a variety of settings.
Non-abstinence contingent housing programs that did not explicitly employ the Housing First model may also reduce the amount of time spent homeless or living in shelters and increase the amount of time in stable housing compared to usual services. Furthermore, group homes where tenancy is not contingent on treatment adherence or sobriety may reduce the amount of time homeless compared to independent apartments with similar non-abstinence contingent tenancy.
However, when compared with abstinence-contingent housing (integrated housing), non-abstinence contingent housing may be less effective at reducing homelessness and improving housing stability.
We are uncertain of the effect of non-abstinence contingent housing combined with day treatment compared with day treatment only due to very low certainty evidence.
Housing vouchers
All of the included studies were conducted in the USA and thus used Section 8 Housing Vouchers provided by the Department of Housing and Urban Development. These housing vouchers combined with case management are probably more effective in reducing homelessness and improving the amount of time in stable housing than usual services or case management alone for adults with mental illness or HIV. Housing vouchers may help homeless families leave temporary shelters more quickly and stay out of shelters for longer periods of time.
Residential treatment with case management
Residential treatment with case management for adults with mental illness and/or substance abuse may be more effective at reducing the amount of time people spend homeless after leaving treatment, and increase both the amount of time spent in stable housing and the proportion of participants who are in stable housing one year after beginning treatment.
Overall completeness and applicability of the evidence
Completeness of the evidence
The identified studies include a fairly good representation of the typical populations which struggle with housing stability (adults with mental illness and/or substance abuse) as well as some relatively smaller portions of the homeless population (families, youth, recently released criminal offenders). The included studies also examined, altogether, all of the interventions which were identified in the protocol for the project. They were compared to both usual services and other interventions. As specified in the inclusion criteria, all of the studies addressed the primary outcomes (homelessness and housing stability) and many of the studies also examined secondary outcomes.
There are, however, three legitimate concerns regarding applicability of the review findings to other contexts. Firstly, usual services may differ substantially from context to context (e.g. between Denmark and the USA, or between states within the USA). Relatively better usual services in a given context may reduce the difference in outcomes between intervention and usual services groups. Secondly, there is a concern regarding the definition of homelessness. In some countries, “homeless” includes “literally homeless,” or people with no shelter (living on the streets). In contexts where homelessness is defined more broadly (anyone in transitional or unstable housing) there may be less of a difference between intervention and control groups for some outcomes.
Quality of the evidence
Although all 43 of the included studies were randomized controlled trials, all studies with the exception of one were assessed as having high risk of bias. This high risk of bias is due to: risk of selection bias, particularly poor randomization (N=4) or poor allocation concealment procedures (N=4); performance bias (N=21); detection bias (N=12); attrition bias (N=15); or reporting bias (N=2). In 12 studies other risks of bias were also identified, including addition of new participants halfway through the study period without providing details regarding demographics or background, self-selection of participants during pre-treatment assessment period or discretionary approval of individuals’ participation in the study by the implementing institutions, participants moving between intervention and control conditions, and treatment diffusion, introduction of new policies which resulted in media attention or impacted “usual services” during the study period, and varying degrees of treatment fidelity as discussed by the primary authors. However, the most common issue across studies was poor reporting of methods, including inadequate reporting of randomization, allocation and blinding methods. In many studies it was not possible to ascertain whether attempts were made to blind participants, personnel or outcome assessors to the assigned intervention condition. It can be assumed, due to the nature of the intervention, that blinding was neither possible nor attempted in most of these studies, and thus we often interpreted unclear reporting for these domains as high risk of bias. We attempted to assess risk of bias separately for subjective and objective outcomes due to the lack of or unclear blinding of participants and personnel, as performance bias is more likely to influence subjective outcomes than objective outcomes. However, there were very few objective outcomes included in the study. When number of days spent homeless or in different housing situations was reported, it was either explicitly indicated that these were self-report measures using an interview form, or the data collection methods were not described (i.e. no mention of use of administrative records) and we assumed self-report measures were employed. Some of the secondary outcomes reported in the individual studies used objective measures such as urine analysis; however, we have not graded evidence for any secondary outcomes.
Strengths and limitations of this review
This review has numerous strengths. Firstly, the findings of this review are based on a rigorous and systematic search of the published and grey literature. Furthermore, identification and selection of relevant studies and publications were carried out by at least two reviewers and based on a priori defined criteria. This was also the case for data extraction, appraisal of the risk of bias in the included studies and grading of the evidence for all outcomes. The published protocol is available at kunnskapssenteret.no. Secondly, we only included randomized controlled trials, thereby including evidence from only the most appropriate study design to answer this review of effectiveness. Thirdly, many of the included studies presented enough data on the difference between groups so that it was possible to statistically estimate the effect of case management or housing programs on housing stability and homelessness. Fourthly, by appraising the methodological quality of the included studies and grading the evidence, we are able to point out clear areas where future research can be improved in terms of design, conduct and reporting. Finally, by including both housing programs and case management interventions, we have provided a comprehensive overview of what is known about the effect of most types of interventions available to prevent or reduce homelessness among homeless or at-risk groups and a comparison of their relative effectiveness where possible.
However, this review is not without limitations. Firstly, the complex nature of the interventions included in this review have three important consequences: 1) we may have missed relevant interventions in the literature search that were labelled as something else butincluded many or all of the same components of the included interventions; 2) we have grouped interventions together in an attempt to provide the end user with a more clear overview of types of interventions that work – this unavoidably leads to less detail regarding individual interventions, and; 3) the included interventions are likely to have varied greatly in how they were implemented, between study sites and across studies, even where they were reported as having followed a specific model (e.g. Housing First). We have not reported treatment fidelity for the included programs. Treatment fidelity was not systematically reported in the included studies, and was thus left out of our analysis. Secondly, due to archiving problems, we are unable to provide a complete list of reasons for exclusion for studies excluded after being read in full-text in the first search. Thirdly, for resource reasons, we have not attempted to synthesize, narratively or through meta-analysis, results for secondary outcomes. Finally, we did not extract data on, or include, cost-effectiveness data, which is important in making decisions on implementing such large social interventions, nor did we include qualitative research, which is used to examine participants’ perceptions, preferences and/or experiences with interventions.
Conclusion
In this comprehensive systematic review of 43 randomized controlled trials, we aimed at determining the effect of interventions to improve residential stability and reduce homelessness. We found that housing programs and case management interventions appear to improve housing stability and reduce homelessness compared to usual services. There was no evidence that housing programs or case management resulted in poorer outcomes for homeless or at-risk individuals than usual services.
Research gaps
There is a great deal of research available on interventions to improve housing stability and reduce homelessness, as demonstrated by the large number of randomized controlled trials included in this review (and the large number of quasi-experimental studies excluded). However, the majority of the existing research has been judged to have high risk of bias, mostly due to poor reporting of methods, and lack of blinding of participants, personnel and outcome assessors. Although it is impossible to blind personnel and participants due to the nature of the interventions, the outcome assessors could be blinded. Furthermore, there has been no clear improvement in reporting between the year the first included study was published (1992) and 2015 (the most recent publication). Specifically, details are lacking regarding comparison group conditions, and the reporting of effect estimates within primary studies isinadequate.
Aside from a general need for better conducted and reported studies, there are specific gaps in the research: Case management for specific sub-groups, specifically families and disadvantaged youth Abstinence-contingent housing with case management or day treatment Non-abstinence contingent housing, specifically different living arrangements (group vs independent living) Housing First compared to interventions other than usual services (e.g. abstinence-contingent housing, case management only, housing vouchers) All interventions from contexts other than the USA
Footnotes
Information about this review
1
Kertesz 2007 58. Kertesz SG, Mullins AN, Schumacher JE, Wallace D, Kirk K, Milby JB. Long-term housing and work outcomes among treated cocaine-dependent homeless persons. Journal of Behavioral Health Services & Research. 2007;34(1):17-33. reported standardized deviations, so the review authors converted this number to standard errors since this is a more correct statistic for the data. The data included in this analysis was not available from the publication, but was sent by the study authors to the review team upon request.