The Unique Needs of Homeless Youths With Mental Illness: Baseline Findings From a Housing First Trial

At Home/Chez Soi was a randomized controlled trial of Housing First in five cities in Canada. The study protocol has been described in detail elsewhere (15). Briefly, the trial tested Housing First, which provides access to housing of the individual’s choice and support services without mandating sobriety or psychiatric treatment beyond a weekly visit by the support team. Participants were individuals age 18 or older in Moncton, Montreal, Toronto, and Winnipeg and 19 or older in Vancouver who were currently homeless or precariously housed and had a mental disorder on the basis of DSM-IV criteria determined by the MINI at study entry (6). Individuals were excluded if they were clients of other assertive community treatment or intensive case management teams, had illegal immigration status, or did not meet a strict definition of homelessness. [Additional details about recruitment are reported in an online supplement to this article.]

Participants were recruited from community agencies that serve homeless people; institutions such as health care facilities, prisons, and jails; and directly from the street. Participants provided verbal consent to be screened for eligibility; those who met inclusion criteria were assessed for capacity to consent before written informed consent was obtained. Recruitment occurred from October 2009 to July 2011. Participants were randomly assigned to Housing First or treatment as usual by computer using adaptive randomization procedures. The study was approved by research ethics boards in every jurisdiction (11 in total).

Participants self-reported demographic information (gender was reported as male, female, transgender, transsexual, other, or declined to answer), chronic general medical conditions, diagnosis of a learning problem or learning disability, involvement with health services, and arrests. MINI results obtained at screening were used to determine DSM-IV mental disorders and suicidality (6). The Adverse Childhood Experiences Study (ACE) questionnaire was used to measure childhood trauma (16). Community functioning was determined from the observer-rated Multnomah Community Ability Scale (MCAS) (17,18). Participants completed comprehensive face-to-face interviews at baseline and every six months and brief measures of housing and vocational status by telephone or in person approximately every three months, for a total of 21 to 24 months of follow-up. Data were entered remotely on laptops and stored in a central database. Participants received monetary incentives for attending interviews. The data set for this analysis was drawn from the interviews that occurred at baseline, except for the ACE questionnaire, which was completed at 18 months.

Descriptive statistics were calculated for each demographic, clinical, and service use variable. We defined a “youth” as an unaccompanied individual age 24 and under, reflective of the definitions used by the United Nations and local governments. We compared participants age 24 and younger with those over age 24 by using two-tailed t tests for continuous variables and chi-square tests for binary and categorical variables.

Logistic regression models were used to examine the effect of demographic, clinical, and service use factors on having a regular medical doctor and reporting having had a need for health care and not having received it, which generated odds ratios that indicated the likelihood of the outcome. To model the number of visits to the emergency department, a negative binomial regression model was used to account for any overdispersion in our data. Incidence rate ratios were calculated by exponentiating the estimates, which provided ratios of events compared with the overall rate. To determine the effect of each covariate considered in the analysis, univariate tests were performed to obtain the unadjusted odds ratios or rate ratios. Variables significant at p<.2 were entered into the final multivariate model. All analyses were conducted with SAS, version 9.4.

Mean imputation was performed for the MCAS, an instrument with validated scale scores, for participants who refused or skipped a number of items by adding the mean of remaining items if they provided answers to at least half. Complete-case analysis was used for regression outcomes, because each had less than 4% missing.

In total, 2,255 participants recruited into At Home/Chez Soi completed baseline interviews; 7% (N=164) were age 24 or younger, and 93% (N=2,091) were over age 24.

Participant demographic characteristics are shown in Table 1. Younger participants had a mean age of 21.6 years; most were male, had less than a high school education, and were unemployed. Youths had been homeless for a mean of 26.1 months. Clinical and service use characteristics are shown in Table 2. Most youths had a current diagnosis of substance abuse or dependence and had a TBI. Nearly half had been assaulted in the preceding six months, had four or more adverse childhood experiences, and self-reported a learning disability. Less than half of the youths had a regular medical doctor, and 50% reported a perceived unmet need for health care in the past six months. Over 40% reported two or more psychiatric hospitalizations in the previous five years. Over 60% had visited an emergency department in the past six months. Almost 40% had been arrested in the past six months.

Compared with older participants, youths were more likely to be from an ethnoracial minority group, to have English as a first language, and to have not completed high school (all p<.05). Youths were less likely than the older participants to have a psychotic disorder, more likely to have mania or hypomania or a drug use disorder, and more likely to report having been identified as having a learning disability; youths had lower rates of hepatitis B and C and HIV/AIDS, compared with the older participants (all p<.05). Younger participants were more likely than the older group to have been assaulted and arrested in the past six months. They were less likely to have a regular medical doctor. Youths and adults had similar MCAS scores and psychiatric hospitalization rates.

In logistic regression analysis, only birth outside Canada and moderate to high suicidality were associated with having a regular medical doctor in a univariate model at p=.2; however, these associations were not significant in the final multivariate model (Table 3). Being nonmale and not being from an ethnoracial minority group or Aboriginal were associated with perceived unmet need for health care in a multivariate model (Table 4). The mean±SD number of emergency department visits among youths was 1.7±2.9, indicating a highly skewed distribution. In a negative binomial regression analysis (Table 5), a learning disability and perceived unmet need for health care and absence of a drug use disorder were associated with having more emergency department visits.

Youths age 24 and younger enrolled in At Home/Chez Soi had a high degree of impairment, as measured by several indicators, including a mean MCAS score suggesting moderate disability, and a substantial rate of multiple psychiatric hospitalizations. Although some differences between youths and older participants are self-explanatory and related to age (for example, lifetime duration of homelessness), and others may be partly explained by an age effect (for example, drug use disorder) or a cohort effect (for example, self-reported learning disability), some of the findings may truly distinguish homeless youths from older homeless individuals, youths in general, and youths with mental illness. These findings may have particularly important implications for prevention and services directed to homeless youths with mental illness.

Youth were less likely than older participants to have completed high school and more likely to report a learning disability, which could represent age and cohort effects. However, rates of not completing high school and a self-reported learning disability in both groups were overwhelmingly higher than national averages (19,20). Given that a high school education has been shown to protect youths from homelessness and predict rehousing and residential stability among homeless youths (21), efforts should be directed to teenagers at risk of dropping out of high school, with the goal of preventing later homelessness. The high rate of self-reported learning disability in this sample, previously undescribed to our knowledge, also suggests that educational programs directed at homeless youths should include screening and corresponding evidence-based interventions for learning disability.

As in previous reports on homeless youths in Canada (22), many youths in our sample identified as being members of ethnoracial minority groups at rates significantly higher than adult participants, and many were Aboriginal. This may be partly explained by a cohort effect, but the ethnoracial differences are more extreme than among older and younger Canadians in the general population (23). Furthermore, although youths from ethnoracial minority groups in the general population tend to report better mental health than white youths, they appear to be overrepresented among homeless youths with mental illness (24). An At Home/Chez Soi subtrial recently found that Housing First with intensive case management enhanced with antiracism and antioppression principles (25) was effective in improving housing stability in an ethnoracially diverse subgroup of participants (26). Further research is needed to guide culturally sensitive interventions for homeless youths. At the Winnipeg site, where over 70% of participants identified as Aboriginal, all active treatment was delivered with an Aboriginal holistic approach by staff who received ongoing cultural training. Practices such as these are likely to increase engagement and improve effectiveness of interventions delivered to Aboriginal homeless youths (27).

In our sample, youths were more likely than adults to have a drug use disorder and had rates of substance use disorder much higher than young people in general (28). In the unadjusted logistic regression model, having a drug use disorder was associated with perceived unmet need for health care among youths. Homeless youths with co-occurring mental and substance use disorders are known to have significant barriers to care (29). More research into the complex interaction between homelessness, drug use, and access to health care among youths is warranted.

A majority of our sample reported a history of TBI, a rate much higher than prevalence estimates in the general population (30). Unfortunately, it was not possible to determine whether TBI preceded or followed homelessness, an important factor in guiding prevention efforts. The literature on homeless adults indicates similarly high rates of TBI and suggests that 70% to 90% of affected individuals experienced their first injury prior to becoming homeless (31). Given the cognitive, emotional, and behavioral sequelae associated with TBI, training of service providers and collaboration with specialized services is recommended.

The prevalence of infectious disease (hepatitis B or C or HIV) in our sample was almost five times as high among adults over age 24 compared with youths, an important finding given that these diseases can lead to chronic morbidity and death and are associated with high health care costs. Although the finding may represent an age or cohort effect, it points to an important opportunity to direct prevention efforts to youths—for example, education about transmission through sex and intravenous drug use, provision of condoms, and needle exchange programs.

Most of this sample of homeless youths reported one or more adverse childhood experiences, and almost half reported four or more adverse childhood experiences, the cutoff associated with an increased risk of poor mental health, behavioral, and general medical outcomes (16). This adds to the body of evidence from longitudinal studies that early victimization can predict homelessness in young adulthood and that preventive interventions should be targeted to child victims of abuse—for example, through child protective services (1). Our findings also suggest that homeless youths continue to be victimized at rates that exceed those of youths in the general population (32). The phenomenon of trauma both precipitating and resulting from homelessness among youths has been well described (13). Not surprisingly, over one-third of youths met criteria for current posttraumatic stress disorder, underscoring the need for trauma-informed and trauma-specific services directed to homeless youths.

Over one-third of the youths in our sample had been arrested in the past six months, a rate significantly higher than among adult participants and youths in the general population but similar to rates among other youths with psychiatric diagnoses, such as drug use disorders (33). Given the established relationship between incarceration and homelessness (34), our findings lend further support for optimizing the criminal justice system as an entry point to housing and supports.

Less than half the youths in the sample reported having a regular medical doctor, a significantly lower proportion than among adult participants and Canadian youths in general (35), and most had visited an emergency department in the past six months (medical visits are covered by universal health care in the study setting). This finding suggests that homeless youths are receiving less preventive health care and more acute (and costlier) care. Half the youths reported unmet need for health care, which was associated with being nonmale and not being from an ethnoracial minority group or Aboriginal. The variable entails both the perception of having needed care and not having received it. Nonwhite homeless youths may be less likely than white homeless youths to perceive that they require care or may feel that their needs are better met.

The number of recent emergency department visits in this sample was associated with a self-reported learning disability and perceived unmet health care need. Because the reason for the visit was not assessed, it remains unclear whether these were true emergencies or whether the youths encountered barriers to using less acute services; the precise nature of the interaction between dissatisfaction with care, cognitive ability, and emergency department use also remains unclear. The association between perceived unmet need for health care and frequent emergency department use is consistent with findings from homeless adults (36). Having a drug use disorder was associated with fewer emergency department visits, a surprising finding that may indicate that youths with these disorders prefer receiving services elsewhere (for example, community withdrawal services) or that they may use fewer medical services in general, perhaps out of fear of criminal charges or discrimination (37). The mean number of emergency department visits was highly skewed in this sample, suggesting that a small number of individuals may have accounted for most of the visits. Studying the most frequent users of the emergency department among homeless youths may help identify those whose care and outcomes could be improved by interventions while reducing health care costs.

Several limitations must be considered in interpreting the results of our study. The primary purpose of the trial was to test Housing First among homeless adults with mental illness. Participants were recruited on the basis of study inclusion criteria and not necessarily to represent homeless youths. Our findings are similar to those from the general literature on homeless youths, which is based mostly on recruitment from local agencies (13). We were limited to analyzing data collected in the trial, which omitted some factors particularly important to youths, including sexual orientation. Furthermore, in our regression analyses, we chose to group nonwhite participants to determine the impact of race-ethnicity on service use, which may have obscured important differences between participants from ethnoracial minority groups and Aboriginal participants. This study was not designed to identify differences between younger and older homeless adults; given the discrepancy in sample sizes between the group of participants ages 18 to 24 and the remaining participants, differences should be interpreted with caution. The instruments used in this study were mostly validated (and pretested for the trial) in the general adult population and not necessarily among young adults, which may have resulted in less than optimal precision. Finally, our analysis was cross-sectional, and we are unable to make conclusions about causative or even temporal effects of measured variables on homeless youths with mental illness. Forthcoming studies will examine the longitudinal effects of Housing First and other factors on youths enrolled in this trial.

In this study of homeless individuals from five cities in Canada, homeless youths differed from homeless adults. Youths had higher rates of high school noncompletion and learning disabilities, and a larger proportion were from ethnoracial minority groups, had a drug use disorder, had been victimized or arrested recently, and did not have a regular medical doctor. Both youths and adults had similar rates of overall impairment, but the findings suggest that youths have unique needs. These differences may also reflect diverse root causes of homelessness for youths and adults and should be explored in future research to more effectively guide prevention efforts. Future At Home/Chez Soi analyses will focus on determining the effectiveness of Housing First for homeless youths with mental illness. Findings such as those described here may be used to guide the adaptation of future interventions to more effectively target youths in order to reduce the individual and societal costs associated with youth homelessness.

The authors thank Andree Schuler, Ph.D., and Agnes Gozdzik, Ph.D., for their assistance and expertise. The authors also thank Jayne Barker, Ph.D. (2008–2011), Cameron Keller (2011–2012), and Catharine Hume (2012–present), who served as national project leaders for At Home/Chez Soi; the five site research teams; the site coordinators; the numerous service and housing providers; and the individuals with lived experience who contributed to this project and the research.