The Second National Canadian Homeless Youth Survey: Mental Health and Addiction Findings: La Deuxième Enquête Nationale Auprès des Jeunes Sans Abri : Résultats en Matière De Santé Mentale et de Toxicomanie

Abstract

Objective:

Youth experiencing homelessness represent a major social problem in Canada and, as demonstrated in the first national survey of this population conducted in 2015, are experiencing significant mental health challenges. The present study examines the findings of a second national survey completed in 2019. These findings afford the opportunity to examine the reliability of the findings of the first study with another large, representative sample and to attempt to articulate the unique characteristics of youth experiencing the greatest distress among this at-risk population.

Methods:

This study analyzed the mental-health-related data from the 2019 Without a Home–National Youth Homelessness Survey that was administered through convenience sampling at 98 agencies serving homeless youth in 49 communities across Canada. The survey was cross-sectional and self-administered, assessing a range of demographic information, pre- and post-homelessness variables, and mental health indicators. Multinomial logistic regression and linear regression were implemented to evaluate associations with distress level.

Results:

Survey data were obtained from 1,375 youth accessing Canadian homeless services in 9 provinces. Thirty-five percent reported at least 1 suicide attempt, and 33.1% reported a drug overdose requiring hospitalization. The findings of this survey replicated most of the key findings from the 2015 survey. The current findings emphasized, for this high-risk population, the heightened adversity faced by young women, Lesbian, Gay, Bisexual, Transgender, Queer, Two-Spirit (LGBTQ2S), and Indigenous subpopulations, as well as the centrality of violence exposure in determining risk and distress. Sexual violence, in particular, emerged as a key factor in the identification of youth experiencing the greatest distress with risk buffered by contact with family.

Conclusions:

These findings can inform prevention and intervention policies and services and reinforce the importance of attending to violence exposure and trauma as central to the mental health trajectories of youth who have experienced homelessness.

While population estimates must be considered cautiously due to methods limitations, it is estimated that there are over 30,000 youth experiencing homelessness in Canada annually. ^{1 –3} Within this population, there is overrepresentation among subpopulations that experience systematic discrimination such as Indigenous peoples and sexual and gender minorities. ^{4 –6} Canadian studies suggest that at least two-thirds of this population experience childhood abuse or neglect ^2,4,7,8 alongside problematic school trajectories and bullying, family violence and disruption, mental health challenges, and interactions with criminal justice and child protection systems. ^{9 –12} The majority cycle in and out of varying degrees of homelessness, ¹² and homelessness amplifies stressors including the persistent exposure to violence and victimization and a range of deprivations. ^13,14 Outcomes include high levels of mental and physical health problems ^4,12 and high mortality rates, with suicide and drug overdose the primary causes. ^15,16 Efforts to support youth in transitions out of homelessness are routinely thwarted by underfunded and suboptimal services and an underdeveloped intervention evidence base ¹⁷ with a few exceptions (housing and housing stabilization, ^18,19 addictions ²⁰ ).

The 2015 Without a Home National Survey ^4,12 was a step toward generating better, more representative data that might inform systems-level planning for Canadian homeless youth populations. These data were important in highlighting subgroup disparities (e.g., Indigenous youth, sexual and gender minorities, young women, along with key risk indicators such as an earlier age of first homelessness). ^4,6,12 These data emerged in line with other studies showing the buffering effect of social connectedness ²¹ and the centrality of victimization as a driver of risk. ^22,23 Repeating the 2015 survey in 2019 afforded the opportunity to refine metrics and, most importantly, to determine whether the findings of the first survey analyses are replicable, lending greater credence to policy and practice recommendations. It was hypothesized that key pre-homeless and homeless risk factors would be associated with distress, lower resilience, and quality of life (QOL), with victimization during homelessness being particularly detrimental. It was also hypothesized that certain groups would be experiencing these challenges to a greater extent, such as young women, sexual and gender minority groups, and Indigenous-identifying individuals. Novel analyses in this study that were not performed in the first survey, and that were exploratory, included examining both very high and low distress groups dichotomously and distress continuously to explore the utility of articulating an “ultra-high-risk group” and performing moderation analyses for age, gender, and length of time homeless.

Methods

Design, Recruitment, and Data Collection

This cross-sectional, convenience sampled, self-report survey was administered through 98 agencies serving homeless youth in 49 communities across Canada. Survey sites were all organizations whose service populations are youth who have experienced homelessness and were identified through the Canadian Observatory on Homelessness, A Way Home Canada, and associated networks. Surveys were administered by organization staff on their premises with the criteria for participation being that the individual is actively receiving services. Staff approached clients, providing survey and consent materials, reviewing the purpose and nature of the survey, and completing the consent process. Youth were provided with either paper copies of the survey or a computer or tablet to complete the survey online. Unfortunately, reliable information is not available regarding the consent and survey completion rates against those who declined to participate and how completers compare in number and demographic characteristics with others at a given location. Data collection occurred over a 14-week period between January and April 2019. All participants received US$10 for their participation, and the study was approved by the York University Research Ethics Board.

Measures

Due to the nature of survey administration and a population that is prone to survey fatigue, brief measures were prioritized, and in some instances, key items were derived from validated scales. Wherever item subsets were used, the principals for their selection were to identify items through item to core factor relationships in psychometric analyses where possible, use investigator consensus regarding topic relevance, and obtain feedback from previously homeless youth on draft versions of the survey. Basic demographic information was collected in areas of age, ethnicity, race, gender, sexual identity, and age of first homeless episode. Racialized status was further articulated as racialized non-Indigenous (replied as a person of color but did not endorse Indigenous identity) and racialized broadly (considered self a person of color regardless of Indigenous endorsement). Binary responses were collected for the occurrence of physical and sexual violence in prestreet and street contexts, with street context sexual violence further articulated as unwanted touch and forced activity. Binary responses were also obtained for one or more drug overdoses requiring hospitalization and one or more suicide attempts with the intent to end one’s life.

To assess mental health symptomatology and substance use problems, the GAIN Short Screener (SS) was used, given that it has well-established psychometric properties with adolescent populations. ²⁴ This included the 6-item, internalizing scale (α = .86) and the 5-item subscale tapping substance use (α = .91). To assess subjective QOL, 7 items from the well-validated WHO QOL-Bref ²⁵ were administered. The 7 items were chosen with attention given to covering general life satisfaction and items from physical health, environment, and psychological domains (α = .86). Resilience was assessed using the 10-item version of the Connor-Davidson Resilience Scale ²⁶ which has established psychometrics (α = .93). Social support from friends was assessed with 9 items from the Hemingway Measure of Adolescent Connectedness ²⁷ (α = .95). Three additional general support items were added to tap domains relevant to this population, using the same scale, referring to people who can be counted on in an emergency, for emotional support, and who check in on them.

Data Analysis

In the present study, discrete variables were presented as totals (N) or percentages (%). Continuous data were presented as mean and standard deviations (SD). Chi-square tests were used for the comparison of discrete variables. For continuous variables, Pearson correlation test or two independent-sample T test were conducted for associations’ examination, as appropriate. Following a descriptive analysis of demographic and scale findings and an examination of simple associations, latent profile analysis (LPA) and confirmatory factor analysis (CFA) ²⁸ were used to establish discrete and continuous latent measures of distress, respectively, based on QOL, psychiatric symptoms, substance abuse, and resilience. LPA identify groups of subjects with similar patterns into the same class while CFA assumes that the covariance of the 4 variables is due to a single continuous factor score. The discrete latent distress measure was done to determine whether there are unique associations specific to an ultra-high-risk subgroup (within this generally high-risk population) for whom there may be unique intervention targets. The CFA was completed to assess the validity of considering an ultra-high-risk group as a unique subpopulation (indicating specific responses) versus considering risk as a dimension only. Clusters for highest and lowest distress formed the dependent variable for block-wise multinomial logistic regression in LPA, and a continuous latent measure of distress based on the same group of dependent variables was used as the outcome for linear regression in CFA. The resulting distress outcomes were then regressed on 3 blocks of variables: Step 1 pre-homeless variables (child protection, sexual abuse, physical abuse, neglect, and age of first episode of homelessness), Step 2 street/homeless considerations (time since first episode homeless, physical and sexual violence experienced while homeless), and Step 3 potential protective factors (support of friends, contact with family, and people who provide emotional, emergency, and check in support). Further, potential moderation effects were evaluated for pre-homelessness adversity, in which biological age, sex, and length of homelessness were picked as general demographic moderators a priori based on the findings of the first survey. The measurement models, LPA and CFA, were conducted simultaneously with the corresponding regression model under the structure equation modeling framework to obtain proper estimates of standard errors. To mitigate estimation bias caused by missing data, the full information maximum likelihood ²⁹ method was employed throughout the regression process, so that all available information would be used under the missing at random assumption. Associations and comparisons with significance less than 0.05 are reported acknowledging that multiple-comparison adjustment may render some nonsignificant. All analyses were conducted using SPSS Version 25.0 ³⁰ and Mplus Version 8.3. ³¹

Results

Participants

A total of 1,375 youth completed the survey, and after cleaning, 1,363 were retained (cases with key demographic data and measures omitted were excluded). Data were collected from all provinces except for Prince Edward Island (see Table 1). The largest amounts of data were collected from Ontario (n = 787; 57.7%), Quebec (n = 174; 12.8%), and Alberta (n = 118; 8.7%). Data were primarily gathered in large urban centers such as Toronto (n = 244; 17.7%), Montreal (n = 143; 10.4%), and Ottawa (n = 113; 8.2%). The complexity attending sexual, gender, ethnic, and racialized identities in this diverse group were readily apparent in multiitem endorsements. Recognizing the limitations of attempting to capture this information categorically, participants were n = 634 (46.5%) cisgender male and n = 491 cisgender female, with n = 793 (58.2%) reporting heterosexual and cisgender identities and n = 407 (29.6%) reporting sexual and gender minority identities. Participants were n = 343 (24.9%) Indigenous-identified, n = 582 (42.7%) identified as White, along with other prominent specific racialized identities that included Black n = 174 (12.8%), and Latin American n = 12 (0.9%) with n = 375 (27.5%) youth identifying as racialized generally. The mean age of participants was 20.5 years (SD = 3.0; range = 12 to 36), with 16.8 years (SD = 3.1) cited as the mean year of first homelessness and 3.6 years (SD = 3.1) the mean years between year of first homeless episode and current age. Mean reported episodes of homelessness in the past 3 years was 1.3 (SD = 2.7) and the most commonly reported places of residence in the month prior included someone else’s home (n = 538; 39.5%), their own place (n = 401; 29.4%), parent/caregiver home (n = 349; 25.6%), and emergency/domestic violence shelter (n = 345; 25.3%).

Table 1.

Participant Demographics by Region.

Variable	British Columbia	Prairie Provinces	Ontario	Quebec	Maritimes	Total
Participants	73	204	787	174	125	1,363
% Female	34.9	41.6	39.6	29.9	48.3	39.2
Mean age	20.5 (2.6)	18.6 (3.3)	20.5 (2.6)	22.0 (2.7)	21.6 (3.6)	20.5 (3.0)
% LGBTQ2S	58.6	31.0	32.2	42.0	26.8	33.9
% Indigenous	57.1	65.1	24.4	17.1	26.1	31.7
% Racialized non-Indigenous	4.5	17.0	38.4	20.2	13.3	30.2
Age first homeless episode	15.8 (2.3)	15.6 (3.0)	17.2 (3.0)	18.3 (2.8)	17.2 (3.1)	17.1 (3.0)
Mean time since first homeless	4.7 (2.9)	2.9 (2.6)	3.2 (2.9)	3.0 (2.9)	4.4 (3.5)	3.3 (3.0)
% Black youth	2.7	3.4	18.8	4.6	7.2	12.8
% Racialized	27.4	29.4	32.4	13.8	12.8	27.5
%White nonracialized	52.1	43.1	51.3	71.3	72.8	54.7

Mean Scores and Correlations

Mean QOL fell in the middle of the 5-point scale (3.3/5; SD = 0.89) with item means ranging from 3.0 (satisfaction with self) to 3.5 (feeling safe in daily life). On the GAIN-SS internalizing scale, 74% of respondents fell in the “high” symptom/distress category with a mean of 2.57 on the 0 to 4 scale (SD = 1.14) with the psychosis item having the lowest endorsement and the sleep/restlessness item the highest. On the GAIN-SS substance scale, 35% of respondents fell in the “high” symptom/distress category with a mean of 1.69 on the 0 to 4 scale (SD = 1.43) with the withdrawal item having the lowest endorsement and the time spent obtaining/recovering item the highest. Suicide attempt rate was defined as the number of participants who attempted suicide over the total number of participants in the corresponding groups. Thirty-five percent of participants reported at least 1 suicide attempt and 33.1% reported having at least 1 drug overdose requiring hospitalization among the entire sample. On the Resilience Scale, the mean score of 3.6 (SD = 1.0) falls in line with moderately low-moderate resilience cutoff (3.6). ²⁶

Considering bivariate associations (Online Supplemental File 1), youth characteristics were found to be strongly associated with mental health indicators in most domains. This included age-related variables such as younger respondents having greater distress indicators (e.g., more likely to report a suicide attempt), t(914) = 3.15, P = 0.002, and an earlier age of first homeless episode related to distress though, for the amount of time elapsed since the first homelessness episode, only substance abuse emerged as a significantly more problematic. Female gender (cisgender) youth reported greater distress and lower resilience (suicide attempt rate of 58.8% as compared with 46.4%), χ²(1, N = 903) = 13.31, P < 0.001, and, compared to straight and cisgender youth, lesbian, gay, bisexual, transgender, queer, and two spirit youth reported markedly greater challenges (73.8% suicide attempt rate as compared with 39.9% for straight and cisgender participants), χ²(1, N = 872) = 90.41, P < 0.001. Racialized youth generally, and Black-identified youth specifically, reported better mental health and lower suicide attempt rates than other youth. Indigenous youth reported greater mental health challenges including a higher suicide attempt rate (61.0% versus 47.0%), χ²(1, N = 810) = 13.68, P = <0.001.

Across all mental health and well-being indicators, a history of child protection involvement and pre-homelessness physical and sexual abuse and neglect were all associated with greater distress. Similar findings were observed for exposure to physical or sexual violence while homeless. Conversely, being in contact with at least 1 family member was the most strongly associated with mental health indicators including a lower suicide attempt rate (39.5% vs. 60.5%), χ²(1, N = 846) = 35.45, P = <0.001, as were having a support who can help in emergencies, who checks in, and who provides emotional support. Support from peers was less robust across mental health variables, not seeming to assist with mental health symptoms or suicidality.

High and Lower Distress Group Associations

Through LPA, articulating risk subgroups based upon psychiatric and substance abuse symptoms, QOL, and resilience, a 4-clusters solution was adopted due to its having the highest entropy value ³² (0.611) and balanced sample sizes at each cluster. This is consistent with the solution implemented in the previous study. ⁴ Group 1 had 30.8% of the participants who indicated the poorest mental health/wellness, Group 4 with the lowest distress had 20.9%, and the middle 2 groups had 20.9% and 27.4%, respectively. The highest risk group first experienced homelessness at a younger age and had been homeless longer than others and aligned with higher risk demographic group membership except for female gender alongside exposure to violence and fewer social supports (Online Supplemental File 1).

In logistic regression analyses (see Table 2), among pre-homeless variables, Wald criterion indicated that age of first homeless episode was significant (P = 0.008), as was physical and sexual abuse (P = 0.014; P = 0.024), but not neglect nor child protection involvement. For year of first homeless episode, the odds ratio (OR) of 0.92 indicates that for every year of age older at the first homeless episode, youth are 8.0% less likely to be in the high distress group. Youth who reported physical and sexual abuse prior to homelessness are 1.68 and 1.54 times more likely to be in the high distress group, respectively. After adding street/homeless variables at Step 2, Wald criterion indicated that age of first homeless episode and prior sexual abuse became insignificant, with both street sexual abuse touch (P = 0.002; OR = 2.85) and sexual abuse force (P = 0.017; OR = 2.42) increasing the odds of being in the high distress group. Results of adding social supports suggested there are associations between family (P < 0.001; OR = 0.24) and emotional support (P = 0.017; OR = 0.43) with group membership. Prestreet physical violence and street sexual abuse touch remained associated with group membership in the full model. Moderation analysis suggested that there is an interaction between biological age (split by median age of 20) and prestreet sexual violence. For youth whose age was below 20, the OR between prestreet sexual violence and high distress was 151% higher (P = 0.001) than that of those whose age was greater than or equal to 20. A significant interaction was also present between gender and prestreet physical violence. The OR between having prestreet physical violence and high distress for male participants was 5 times higher (P = .004) than that of female participants (see Table 3).

Table 2.

Multinomial Logistic Regression Estimating Associations with Highest Distress Group Membership.

Variables	Step 1: Pre-homelessness			Step 2: Pre–Post Homelessness			Step 3: Protective Factors
	B	SE	OR	B	SE	OR	B		SE	OR
Block 1 (pre-homeless)
Child protection	.15	.18	1.16	0.37	.25	1.45		0.53	.28	1.70
Prestreet physical violence	.52	.21	1.68**	0.80	.29	2.24**		0.80	.33	2.23*
Prestreet sexual violence	.43	.19	1.54**	0.25	.26	1.28		0.31	.31	1.36
Prestreet neglect	.11	.25	1.11	1.28	.42	3.61**		0.73	.48	2.08
Age of first homelessness	−.08	.03	0.92**	−0.01	.05	0.99		−0.03	.06	0.97
Block 2 (homeless)
Years since first homelessness				0.05	.05	1.05		0.04	.05	1.04
Physical violence				0.34	.25	1.41		0.05	.29	1.06
Sexual abuse-touch				1.05	.34	2.85**		1.42	.40	4.13***
Sexual abuse-force				0.88	.37	2.42*		0.68	.42	1.97
Block 3 (supports)
Friend support								−0.01	.08	0.99
Family support								−1.42	.33	0.24***
Help in emergencies								−0.59	.34	0.55
Emotional support								−0.85	.35	0.43*
Person checks in								−0.48	.37	0.62

*P < 0.05. **P < 0.01. ***P < 0.001.

Table 3.

Moderation Analysis for CFA and LPA.

Variables	B	P Value
LPA on Prestreet Sexual Violence × Biological Age by Median
Child protection	0.27	.135
Prestreet physical violence	0.52	.014
Prestreet sexual violence	0.92	.001
Prestreet neglect	0.21	.399
Biological age by median	0.44	.005
Prestreet Sexual Violence × Biological Age by Median	−0.77	.038
LPA on Prestreet Physical Violence × Gender
Child protection	0.27	.308
Prestreet physical violence	1.80	.004
Prestreet sexual violence	0.90	<.001
Prestreet neglect	0.80	.089
Age of first homelessness	−0.10	.015
sex	1.43	.029
Prestreet Physical Violence × Sex	−1.63	.024
CFA on Child Protection × Biological Age by Median
Child protection	−0.10	.229
Prestreet physical violence	0.08	.180
Prestreet sexual violence	0.27	<.001
Prestreet neglect	0.45	<.001
Biological age by median	−0.17	.043
Child Protection × Biological Age by Median	0.24	.026

Note. CFA = confirmatory factor analysis; LPA = latent profile analysis.

Generally, similar results were obtained using CFA and linear regression (see Table 4). At Block 1, prestreet sexual violence and neglect were significantly associated with distress factor scores (P < 0.001) but not child protection involvement, prestreet physical violence, and age of first episode of homelessness. Prestreet sexual violence would incur a 0.26 increase on average on the standardized factor scores with an SD of 1. Neglect prior to homelessness also resulted in a 0.45 higher factor score on average. After adding street/homelessness variables at Step 2, Wald criterion indicated that prior sexual abuse became insignificant, with street physical violence (P = 0.003; B = 0.16), sexual abuse touch (P = 0.002; B = 0.27), and sexual abuse force (P = 0.014; OR = 0.23) significantly increasing the factor score. Adding social supports indicated that there are negative associations between the distress factor score and family (P < 0.001; B = −0.23), emotional support (P = 0.003; B = −0.17), as well as check in (P = 0.025; B = −0.13). Prestreet neglect and sexual abuse touch on the streets remained associated with the distress factor score in the full model. Moderation analysis results indicate there is an interaction between age (split by median) and child protection involvement (P = 0.026). For youth whose age is below 20, the impact of having child protection involvement on the distress factor score was reduced by 0.1 (see Table 3).

Table 4.

Linear Regression Estimating Associations with Distress Factor Score.

Variables	Step 1: Pre-homelessness		Step 2: Post Homelessness		Step 3: Protective Factors
	B	SE	B	SE	B	SE
Block 1 (pre-homeless)
Child protection	.03	.05	.01	.05	.03	.05
Prestreet physical violence	.09	.06	.08	.06	.03	.05
Prestreet sexual violence	.26***	.07	.05	.06	.07	.05
Prestreet neglect	.45***	.08	.40***	.07	.22**	.06
Age of first homelessness	−.02	.01	.00	.01	−.01	.01
Block 2 (homeless)
Years since first homelessness			.02	.01	.02	.01
Physical violence			.16**	.05	.10*	.05
Sexual abuse-touch			.27**	.09	.30***	.08
Sexual abuse-force			.23*	.09	.10	.08
Block 3 (supports)
Friend support					.02	.02
Family support					−.22***	.05
Help in emergencies					−.07	.06
Emotional support					−.17**	.06
Person checks in					−.13*	.06

*P < 0.05. **P < 0.01. ***P < 0.001.

Discussion

As with the first national survey, ⁴ the findings of the second survey reported here afford access to more representative data in a context where data are typically from small, local studies. This second survey was able to achieve slightly better recruitment though with poorer representation from British Columbia and with a slightly older participant group with fewer homeless episodes. Nonetheless, they struggled with psychological distress and frequent reports of suicide attempts and drug overdoses at rates similar to previous findings and with a similar profile of risk associations as the first national survey and other smaller studies. ^7,8 This confirms the study hypotheses and helps to validate, with 2 large and likely independent samples, these findings as being stable within this population over time and across contexts geographic and otherwise.

Also similar to the first survey, ⁴ there was some indication that racialized, non-Indigenous youth and, more specifically, Black-identified youth, appear to be reporting less distress and greater resilience than other youth. This finding is difficult to interpret. In general, some studies of racialized populations have found lower rates of suicidality, for example, ³³ and racialized youth facing intersecting forms of discrimination can engage social supports in a manner supportive of resilience. ³⁴ However, the broad metrics of the present study can do little more than flag this observation as needing further inquiry. Social supports were also related to lower distress greater well-being, particularly family support and instrumental forms of support (e.g., someone to go to in emergencies) as has been observed previously. ^4,35

The second objective of this study was to attempt to articulate the unique characteristics within the most distressed youth in the sample. Considering both sets of regression, it would appear that sexual violence exposure, in particular, would seem to be important to consider in the identification of those at greatest risk, with prestreet physical violence exposure likewise having a sustained influence. Prestreet sexual violence exposure may be particularly impactful as a risk signal for younger youth while prestreet physical violence may be a stronger risk factor for males. Other challenges such as prestreet neglect and street physical violence exposure had greater relevance when distress was considered on a continuum. As a whole, these findings validate previous work that has indicated that violence and, by proxy, trauma before and during homelessness are key to the distress experienced by this population. ^4,23,36,37 These findings also highlight the particularly damaging impacts of sexual violence. Considering potential moderators of risk, as observed previously, ²¹ both regressions indicated the likely benefit of youth having some contact with family (or benefits that such contact indicates) with support from friends more equivocal. Support from friends can be challenged in homeless contexts due to the stressors all youth are facing. ³⁸

This study had several limitations. These included a cross-sectional design and a convenience sample, the use of abbreviated self-report measures, and questions of population representativeness. As well, the influence of multiple comparisons is a consideration that if adjusted for, would change the significance level to .004 making some of the findings reported here lose significance. National representation was also challenged by a response rate that was not balanced across potential sites (e.g., British Columbia). Finally, youth who do not or rarely utilize the kinds of services that participated as sites this study, or services generally, would not be well-represented in these findings. While these are not trivial limitations, these surveys have established a time and cost-effective method for collecting data from a much larger sample of young people who have experienced homelessness than typical surveys. Having now validated what would appear to be a stable set of findings, next steps would be to enhance representativeness through purposeful and better tracked sampling and establishing a cohort design. The historical challenges that longitudinal cohort designs have represented in this population are increasingly mitigated through digital platform use and access.

Youth homelessness is a wicked social problem ³⁹ with variable definitions, multiple determinants, corollaries, and outcomes. Accordingly, interventions from service to policy need to be carefully targeted to increase their impact. ⁴⁰ The findings of the current study validate those of the first national survey and inform the tailoring of interventions to the unique needs of specific subgroups—such as girls and women, Indigenous, and LGBTQ youth. In particular, a close attention to assessing violence exposure and intervening to reduce violence exposure and address trauma are essential in the effort to better serve those in greatest risk and distress. ³⁷ There may also be value in using the report of pre-homelessness sexual violence as a signal that higher intensity supports may be needed and how interventions may benefit from tailoring as a function of gender and age. Conversely, the robust nature of having family support found here provides a further rationale for the growing implementation of family reconnection initiatives internationally. ⁴¹