BACKGROUND AND OBJECTIVES: "Housing First" has been shown to improve housing stability in homeless individuals with mental illness, but had not been empirically tested in homeless youth. We aimed to evaluate the effect of "Housing First" on housing stability in homeless youth aged 18 to 24 years participating in At Home/Chez Soi, a 24-month randomized trial of "Housing First" in 5 Canadian cities.
METHODS: Homeless individuals with mental illness were randomized to receive "Housing First" (combined with assertive community treatment or intensive case management depending on their level of need) or treatment as usual. We defined our primary outcome, housing stability, as the percent of days stably housed as a proportion of days for which residence data were available.
RESULTS: Of 2148 participants who completed baseline interviews and were randomized, 7% (n = 156) were youth aged 18 to 24 years; 87 received "Housing First" and 69 received treatment as usual. In an adjusted analysis, youth in "Housing First" were stably housed a mean of 437 of 645 (65%) days for which data were available compared with youth in treatment as usual, who were stably housed a mean of 189 of 582 (31%) days for which data were available, resulting in an adjusted mean difference of 34% (95% confidence interval, 24%–45%; P < .001).
CONCLUSIONS: "Housing First" was associated with improved housing stability in homeless youth with mental illness. Future research should explore whether adaptations of the model for youth yield additional improvements in housing stability and other outcomes.
- ACT —
- assertive community treatment
- CI —
- confidence interval
- ICM —
- intensive case management
- QOLI-20 —
- Lehman Quality of Life Interview 20
What’s Known on This Subject:
Despite the burden of homelessness on young people, there is little evidence for interventions targeting their housing. "Housing First," a housing and psychosocial intervention with evidence in adults, is being used in youth without having been tested in this population.
What This Study Adds:
This subgroup analysis of youth aged 18 to 24 years provides experimental evidence that "Housing First" in homeless youth with mental illness improves housing stability relative to usual care. Adaptations of the model may be required to improve other outcomes.
Ten percent of America’s homeless are unaccompanied youth aged 18 to 24 years.1 Homeless youth have rates of mental disorders at least twice those of housed youth.2 On average, homeless youth with mental illness report having been homeless for over 2 years, a substantial portion of their young lives.3 The chronic stress and deprivation associated with homelessness may have lasting effects on cognitive and academic functioning, financial stability, and physical and mental health.2
Despite this, few interventions for homeless youth have been empirically tested,2,4 and even fewer have targeted housing outcomes. Case management, a strategy used widely with homeless youth, has demonstrated improvements in housing stability over time, but did not separate from other active treatments5 or treatment as usual.6 The Community Reinforcement Approach, an operant-based behavior therapy tested in 2 randomized controlled trials in homeless youth, was associated with a reduction of days homeless5 and increased social stability (a measure which included housing),7 but with only the latter trial demonstrating improvement relative to a comparison group, it is not clear that any empirically tested interventions have impacted youth’s housing stability beyond general support and the passage of time.
"Housing First," developed in New York in the 1990s by Pathways to Housing based on the principles of housing as a human right, harm reduction, consumer choice, and recovery, provides immediate access to permanent independent housing in the community.8 In homeless adults with mental illness, it has been consistently shown to increase housing stability, and some studies have also found improvements in other outcomes, including days in the hospital, quality of life, and substance use.9 In 2008, the Canadian government invested $110 million through the Mental Health Commission of Canada for At Home/Chez Soi, the largest trial of "Housing First" to date. At Home/Chez Soi found that "Housing First" improved housing stability in homeless people with mental illness10,11; participants who received more intensive psychosocial support also had improvements in quality of life and community functioning compared with usual care.11 Paralleling its dissemination in homeless adults, "Housing First" is increasingly being applied to homeless youth; the US Department of Housing and Urban Development presents models based on "Housing First" principles as best practice for homeless youth.12 However, "Housing First" has never been experimentally tested in youth, including as a subgroup in previous trials. Given the importance of establishing evidence to improve housing in youth, we conducted a subgroup analysis of youth aged 18 to 24 years in At Home/Chez Soi. Our objectives were to examine the impact of "Housing First" on housing stability in homeless youth, to compare this with the effect in adults, and, guided by findings in adults,9 to explore the effect of "Housing First" on other domains, including quality of life, community functioning, psychological distress, problem substance use, health services use, and arrests in homeless youth.
At Home/Chez Soi was a randomized controlled trial of "Housing First" across 5 cities in Canada: Vancouver, Winnipeg, Toronto, Montreal, and Moncton. The study protocol has been detailed elsewhere.13 Study recruitment occurred from October 2009 through June 2011. Research ethics boards in each jurisdiction approved the study (11 in total).
Participants were stratified into high- or moderate-needs groups according to their clinical and service use characteristics; high-needs participants were assigned to receive assertive community treatment (ACT) or treatment as usual and moderate-needs participants were assigned to receive intensive case management (ICM) (with some participants also receiving a locally developed psychosocial intervention) or treatment as usual (Supplemental Table 3). This subgroup analysis combined active treatment groups to compare participants assigned to "Housing First" plus ACT or ICM with those who received treatment as usual. The 2-year outcomes associated with "Housing First" plus ACT and "Housing First" plus ICM in the full sample have been reported separately.10,11
Participants were currently homeless or precariously housed individuals ≥18 years in Moncton, Montreal, Toronto, and Winnipeg and ≥19 years in Vancouver with a mental disorder. Homeless was defined as no fixed place to stay for ≥7 nights and little likelihood of obtaining permanent housing in the coming month, and precariously housed was defined as living in a rooming house, single-room occupancy or hotel/motel with ≥2 episodes of homelessness in the past year. Diagnosis of a mental disorder was based on Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition criteria determined by the Mini International Neuropsychiatric Interview administered by researchers under supervision of a clinical psychologist, or written diagnosis (current psychiatric discharge summary or consultation report) at study entry.14 Individuals were excluded if they were already clients of ACT or ICM teams, had illegal immigration status, or did not meet specified criteria for homelessness. Participants were recruited from community agencies that serve homeless people, institutions, including health care facilities and prisons and jails, and directly from the street. Recruitment was targeted at 100 participants per treatment arm per site.
Potential participants provided verbal consent for eligibility screening, and then those who met inclusion criteria were assessed for capacity to consent before providing written informed consent. Participants were randomized to the active treatment or usual care by computer by using adaptive randomization procedures in a 1:1 allocation ratio. The nature of the project did not allow for blinding of participants or interviewers.
Participants assigned to "Housing First" were given access to scattered-site housing of their choice with mobile, off-site mental health services.8 They were required to have weekly contact with a mental health worker, but were not mandated to achieve or maintain sobriety or accept psychiatric treatment. Participants assigned to usual care had access to housing and support services through other programs in their communities postrandomization, including other ACT or ICM programs. None of the study cities offered the Pathways "Housing First" model before this study.
Participants completed comprehensive in-person interviews at baseline and every 6 months and brief measures of housing and vocational status by telephone or in person approximately every 3 months with procedures in place to maximize participant retention and data quality.13 Most participants were followed for 24 months, although those enrolled toward the end of the trial were only followed for 21 months. If interviewers indicated “no confidence” in participant responses, these were treated as missing (accounting for <2% of youth responses). Data were entered remotely on laptops and stored in a central database. Participants received cash incentives for attending interviews ranging from $30 to $50 depending on the site and interview duration. This trial is registered with the International Standard Randomized Control Trial Register (identifier ISRCTN42520374).
The predefined primary outcome for the trial and this subgroup analysis was housing stability, measured with the highly reliable and valid Residential Time-Line Follow-Back Inventory, defined as days stably housed as a proportion of the number of days for which any type of residence data were available over the preceding 6 months.15 The main trial’s secondary outcome was generic quality of life, measured using EuroQoL 5 Dimensions, a standardized health utility and visual analog scale, which is used widely in clinical and economic evaluations.16
As exploratory outcomes, we examined condition-specific quality of life with the Lehman Quality of Life Interview 20 (QOLI-20) index,17 community functioning with the observer-rated Multnomah Community Ability Scale,18 community integration with the psychological integration subscale of the Community Integration Scale,19 recovery with the Recovery Assessment Scale,20 self-rated physical health and mental health symptoms with the physical component summary and mental component summary scores of the Short Form 12 survey,21 past-month mental health symptoms with the Colorado Symptom Index,22 and past-month substance-related problems with the Global Assessment of Individual Needs Short Screener—Substance Problem Scale.23 The Health, Social and Justice Service Use Inventory, developed for the study based on existing measures, assessed health and criminal justice service involvement over the past 6 months: number of emergency department visits, having a regular medical doctor, perceived unmet health care needs, and number of arrests. Service provider visits were self-reported in the past month. Victimization was assessed by using questions from Statistics Canada’s General Social Survey, capturing participants who reported having been robbed by force, hit or attacked, or attempted forced or forced sexual activity in the last 6 months.24 Employment was measured by using the Vocational Time-Line Follow-Back Inventory.25
All outcomes were assessed at baseline. Housing and employment were assessed every 3 months. The remaining outcomes were assessed every 6 months, with the exception of the Short Form-12 (baseline, 12 months, and 24 months), and Recovery Assessment Scale (baseline and 24 months).
Power was not determined a priori for this subgroup analysis of youth, but we calculated that a minimum of 63 participants per group would provide 80% power to detect an effect size of 0.5 for our outcomes, assuming homogeneity across sites and no attrition. We defined youth as ≤24 years old, reflective of the definitions used by the United Nations (which defines youth as 15–24 years) and local governments (most of which define homeless youth as 16–24 years), although we were limited by the original study design to include only people ≥18 years old.
We analyzed the effect of the intervention in youth on the percent of days stably housed over each 6-month time point by fitting an analysis of covariance model that included treatment assignment, study city, indicators of ethnoracial and Aboriginal status, and a treatment × site interaction. We used a repeated measures mixed-effects model to determine the mean difference in housing stability in the "Housing First" and treatment as usual groups between youth ≤24 years and older participants over time by fitting a model with a 3-way interaction (treatment × age group × time). Minimal data on housing stability were missing in the full sample (in the range of 4%).10,11
For analysis of quality of life and exploratory outcomes, we applied a linear mixed models framework to perform repeated measures analysis of longitudinal continuous outcomes. We applied generalized estimating equations to repeated counts assuming the negative binomial distribution when outcomes were overdispersed. We calculated model-estimated differences in mean changes from baseline for continuous outcomes and ratio of rate ratios for counts and ratio of odds ratios for binary outcomes at the 6-, 12-, 18-, and 24-month time points with 95% confidence intervals (CIs). We assessed employment over the 24-month study period. We tested the main fixed effects of treatment group, time, study city, and Aboriginal or ethnoracial status, as well as the overall treatment × time interaction. Because outcomes in our subgroup analysis were exploratory in nature, significance level was maintained at α = .05 and not adjusted for multiple testing. We conducted our analyses by using SAS, version 9.4 (SAS Institute, Inc, Cary, NC).
Of 2148 participants who were included in these analyses, 7% (n = 156) were youth: 87 randomized to "Housing First" (either ACT or ICM) and 69 to treatment as usual (see Fig 1). Baseline characteristics are displayed in Table 1. Although participants had not been explicitly randomized according to age, the youth treatment and control groups were balanced in their distribution of covariates.
In an analysis adjusting for study city and ethnoracial and Aboriginal status, youth in "Housing First" were stably housed a mean of 437 of 645 (65%) days for which data were available compared with youth in treatment as usual, who were stably housed a mean of 189 of 582 (31%) days for which data were available, resulting in an adjusted mean difference of 34% (95% CI, 24%–45%; P < .001).
Figure 2 presents a comparison of youth aged 18 to 24 years and adults aged >24 years on mean percentage of days stably housed. The overall effect of the 3-way interaction of treatment group, age (18–24 years vs >24 years), and time on housing stability was nonsignificant (P = .59). For example, the difference in mean change in housing stability from 3 months to 24 months for 18- to 24-year-olds compared with >24-year-olds was –4% (95% CI, –19% to 12%; P = .62).
Table 2 presents the trial’s secondary outcome (generic quality of life) and exploratory outcomes in youth ≤24 years old. "Housing First" was associated with significant improvements in leisure, a category of the QOLI-20, at 6, 12, and 24 months, and an improvement in the total condition-specific quality of life score at 6 months relative to usual care. However, these improvements in leisure and the total QOLI-20 score did not result in significant differences from treatment as usual in the overall treatment group × time analysis (P = .10 and P = .17, respectively). "Housing First" was associated with lower rates of employment over the 24-month study; 27 (32%) youths who received "Housing First" had at least 1 period of employment compared with 28 (44%) youths who received usual care, resulting in an adjusted odds ratio of 0.48 (95% CI, 0.24–0.99; P = .05). We did not find significant differences between "Housing First" and treatment as usual for other exploratory outcomes.
In the first experimental study of "Housing First" for homeless youth with mental illness, "Housing First" significantly improved housing stability. This effect was not significantly different from the effect in older adults and was consistent with findings from previous studies of "Housing First" in adults,9 as well as a case study that demonstrated high rates of housing retention in youth who received "Housing First".26
"Housing First" plus ICM or ACT did not appear to have a statistically significant effect on other outcomes in our youth sample. This is consistent with findings from the larger At Home/Chez Soi trial,10 in which "Housing First" plus ICM improved housing stability in moderate-needs participants, but for other outcomes did not separate from usual care. "Housing First" plus ACT also improved housing stability and resulted in small but statistically significant improvements in quality of life and community functioning relative to treatment as usual, but there were no other differences.11
There are several potential explanations for our findings that "Housing First" only improved housing stability in homeless youth with mental illness. First, the sample size of youth aged 18 to 24 years was not determined a priori, and the study may have been underpowered to detect between-group differences. Second, combining ACT and ICM recipients may have obscured differences in outcomes by level of need. As noted above, analyses of the larger sample suggested that high-needs participants receiving the more intensive ACT intervention showed improvements in additional domains, but our sample size did not allow us to examine outcomes by intervention type.11 Third, participants in both the intervention and control groups may have exhibited regression to the mean on these outcomes. Those receiving usual care had access to other psychosocial services, which, perhaps for youth especially, may be fairly comprehensive, thus decreasing between-group differences. We found that "Housing First" and usual care recipients did not significantly differ in their service contacts. Fourth, our follow-up period may have been long enough to capture differences in housing outcomes, but not some of the secondary benefits that may occur as a result of stable housing. Based on earlier work, we know that our youngest participants had been homeless for years, with high rates of childhood trauma, chronic conditions like traumatic brain injury, and ongoing victimization, and therefore may take time after their housing is stabilized to recover in other areas and shift their use of acute medical services.3 Lastly, although ICM and ACT teams were regularly evaluated for model fidelity, the psychosocial interventions delivered may have varied across sites or service providers, or may not have been as well-suited to youth. Although the current "Housing First" plus ACT or ICM model improved housing stability in homeless youth, adaptations may be needed to target other outcomes. A detailed comparison of youth and adults in this sample suggested that youth were more likely to be an ethnoracial minority, have not completed high school, have a learning disorder or a drug use disorder, and to have been arrested recently, and were less likely to have a regular medical doctor, all of which may warrant tailored approaches to serving them.3
One exploratory outcome that did show some responsiveness to "Housing First" was satisfaction with leisure activities. This echoes the idea in Mullainathan and Shafir’s Scarcity that great stress diminishes the ability to engage in other tasks.27 Restoring housing stability might allow young people to enjoy leisure activities even if it does not result in a measurable impact on other areas of functioning. Youth who received "Housing First" were less likely to have worked than those who received usual care. This is consistent with findings from the larger At Home/Chez Soi trial, which showed that "Housing First" recipients had lower odds of obtaining competitive employment compared with those who received treatment as usual.28 The authors hypothesized that receiving rent subsidies and government benefits may have reduced the financial burden of unemployment and decreased the incentive to work, particularly in provinces with lower earning exemptions for people receiving benefits (where participants had even lower odds of obtaining employment). We only examined a crude measure of employment in youth, but not the suitability or duration of employment. Additional research is needed to better understand pathways to meaningful employment for homeless youth with mental illness.
The parent study has many strengths, including its randomized controlled design, extensive outcome measures, and delivery in a range of local contexts. However, we acknowledge some limitations in our work. First, this is a subgroup analysis of a larger study on "Housing First"; the intervention was not adapted to youth, nor were the instruments specifically selected with youth in mind. We were limited by study inclusion criteria to examining homeless youth aged 18 years and older. Although including youth aged 16 to 17 years would have more accurately reflected existing services for homeless youth, differences in legal status for tenancy and, in some jurisdictions, informed consent make "Housing First" more complicated to study and implement in youth <18 years. These findings may not be generalizable to young people <18 years. We conservatively used 24 years of age as the upper limit for “youth,” and the resulting smaller sample may have been underpowered to detect differences in the intervention and control groups. Because of the nature of the study, participants and researchers were not blinded to treatment allocation. Because outcomes were exploratory in nature, we did not correct for multiple comparisons. Although sites were assessed for model fidelity, we could not control for all possible differences in services (eg, access to psychosocial interventions). Most of the outcomes analyzed in this study were based on self-report, including health services use and arrests, which may benefit from corroboration with administrative data.
This study has significant implications as the first to present experimental data examining "Housing First" for homeless youth with mental illness. It suggests that "Housing First" is a viable intervention to promote housing stability in homeless youth with mental illness and is as effective for young people as it is for adults in general. This real-world trial is expected to be reproducible given fidelity to a well-described model that has been replicated internationally.8 However, given that other outcomes did not appear to respond to the intervention, we suggest considering modifications of "Housing First" to maintain fidelity to core principles while better meeting the needs of youth. This may include attention to issues such as peer/family relationships, sexual health, education and job skills, culture, life skills, substance use, and crime avoidance, and should engage youth in all stages of implementation and evaluation. The results presented here are an important step to developing effective interventions to decrease the long-term consequences of homelessness in youth.
The authors wish to thank Jayne Barker, PhD (2008–2011), Cameron Keller (2011–2012), and Catharine Hume (2012 to present) (Mental Health Commission of Canada At Home/Chez Soi National Project Leads), the National Research Team, the 5 site research teams, the site coordinators, and the numerous service and housing providers, as well as persons with lived experience, who have contributed to this project and the research. The above individuals have received compensation from the Mental Health Commission of Canada for their role in the study. The authors wish to dedicate this paper to the memory of Paula Goering, RN, PhD, lead investigator of the At Home/Chez Soi trial, who drove the design, implementation, analysis, and dissemination of the original trial. We thank Dr Goering for her mentorship and for sharing her immense expertise with our research team, guiding our analysis and interpretation, and providing feedback on earlier drafts of this manuscript. We also thank Agnes Gozdzik, PhD, for her assistance and expertise and Amanda Noble, MSW, and Michael Little, PhD, for providing feedback on earlier drafts. The views expressed herein solely represent those of the authors.
- Accepted July 25, 2016.
- Address correspondence to Nicole Kozloff, MD, Centre for Urban Health Solutions, Li Ka Shing Knowledge Institute, St. Michael’s Hospital, 30 Bond St, Toronto, ON M5B 1W8, Canada. E-mail:
Dr Kozloff guided the conceptualization of the analysis, selection of variables, and interpretation of findings, drafted the manuscript, and made substantive edits to the draft; Dr Adair was a member of the National Research Team and contributed to all aspects, including the design, implementation, analysis, and interpretation of the findings of the original trial, and was involved in the conceptualization of the analysis, the selection of variables, interpretation of findings, and providing feedback on the current manuscript; Mr Palma Lazgare contributed to the conceptualization and selection of variables and led the analysis of the current study and provided feedback on the manuscript; Drs Poremski and Cheung and Ms Sandu were involved in the conceptualization of the analysis, the selection of variables, interpretation of findings, and providing feedback on the current manuscript; Dr Stergiopoulos was a site coprincipal investigator and member of the National Research Team and therefore contributed to the design, implementation, analysis, and interpretation of the findings of the original trial, and she guided the conceptualization of the current study and the analysis and provided extensive feedback on the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This research was made possible through a financial contribution from Health Canada provided to the Mental Health Commission of Canada. Dr Kozloff’s work was supported by a postgraduate research award (Faculty of Medicine, University of Toronto), a Canadian Institutes of Health Research fellowship, and a NARSAD Young Investigator award from the Brain & Behavior Research Foundation.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Henry M,
- Cortes A,
- Shivji A,
- Buck K; Office of Community Planning and Development, U.S. Department of Housing and Urban Development
- Tsemberis S
- Aubry T,
- Nelson G,
- Tsemberis S
- Stergiopoulos V,
- Hwang SW,
- Gozdzik A, et al; At Home/Chez Soi Investigators
- US Department of Housing and Urban Development
- Goering PN,
- Streiner DL,
- Adair C, et al
- Giffort D,
- Schmook A,
- Woody C,
- Vollendorf C,
- Gervain M
- Statistics Canada
- Latimer EA,
- Lecomte T,
- Becker DR, et al
- Scott F,
- Harrison SJ
- Mullainathan S,
- Shafir E
- Copyright © 2016 by the American Academy of Pediatrics