Housing, opportunities, motivation and engagement (HOME) for homeless youth at-risk for opioid use disorder: study protocol for a randomized controlled trial

The study is a randomized controlled trial, whereby 240 youth will be randomized into either housing + opioid and related risk prevention services (n = 120), or opioid and related risk prevention services alone (n = 120). In this study, we seek to determine if adding housing to opioid and related risk prevention services yields a significant benefit above and beyond the benefit yielded by opioid and related risk prevention services alone. Follow-up assessments to collect study outcomes will be conducted at 3, 6, 9 and 12-months post-baseline to evaluate stability of effects.

Youth will be eligible to enroll in the study if they meet the following criteria: (1) youth is between the ages of 18 to 24 years; (2) youth meets the criteria for homelessness as defined by the federal McKinney-Vento Act as “lacking a fixed, regular, stable, and adequate nighttime residence” and includes “living in a publicly or privately operated shelter designed to provide temporary living accommodations, or a public or private place not designed for, or ordinarily used as, regular sleeping accommodations for human beings” [33]; and (3) youth does not meet diagnostic criteria for OUD as assess by the Structural Clinical Interview for DSM-5 (SCID) [34]. Lifetime opioid use is not an exclusionary criterion; however, youth cannot meet diagnostic criteria for OUD at the time of the baseline assessment. The age range was selected because it has been the study team’s experience that landlords will not accept leases signed by youth under the age of 18 years. In addition, “homeless youth” commonly refers to those up to the age of 24 years, and generally reflects the age range of the homeless youth population served by providers across the country [35].

The study will employ a Community Advisory Group (CAG) of key stakeholders including representatives of homeless youth, homeless providers, landlords, substance use treatment experts and policy makers. The CAG will meet throughout the project period, with greater frequency at the beginning and gradually decreasing in frequency. At the beginning of the study, the CAG will review and provide input on study procedures, such as the protocol for identifying youth who are eligible for the study and providing input on any barriers that arise for recruiting landlords willing to rent to participants. This feedback is directly integrated into the study protocol. Moving forward, as the study begins recruitment and follow-up, the CAG will be informed of recruitment progress and challenges, providing suggestions and feedback as needed. Finally, as the study concludes the CAG will give input on best approaches for easing transition of intervention to practice. The CAG is an integral part of the design of the study in order to evaluate both the effectiveness and implementation determinants at an early stage. Too often, interventions do not reach those for whom the interventions were developed [36]. Engaging community stakeholders from the earliest phases of research, as such continuous involvement is critical to dissemination and uptake [36].

The study occurs in a large Midwestern city in the US. The city has one drop-in center for homeless youth between the ages of 14–24 years. Approximately 1000 unduplicated homeless youth are served annually, with 70–95% of homeless youth reporting problem substance use, similar to national samples of homeless youth [37]. The single adult shelters have a total of 457 beds for single men and 97 for single women on any given night. Service-connected youth will be recruited from the drop-in and adult shelters, non-service-connected youth will be recruited through outreach to local soup kitchens, sandwich lines, the streets, parks and libraries. Project staff maintain offices within a local drop-in center serving youth experiencing homelessness. Youth who are engaged on the streets will be transported to the drop-in center. A research assistant will engage and screen youth to determine basic eligibility for the study. After the brief screening and stated interest in the project, written consent will be obtained and the Structured Clinical Interview for DSM-5 Disorders (SCID) section on Opioid Use Disorder will be administered to determine formal eligibility [34]. Upon completion of the baseline assessment interview, youth will be randomly assigned to the intervention conditions using a computerized urn randomization program. At least four advocates will provide services so that advocate effects can be examined [38], and each advocate will provide all opioid and related risk prevention services to the youth on their caseload. For example, even with standardized training and protocols it is possible that unmeasured differences in the personal attributes of advocates could influence youth’s level of engagement with the interventions. Advocates will be crossed by condition to “equate” conditions on advocate characteristics. This approach is preferred over nesting, as advocates in both conditions are trained in the same procedures, which include strengths-based outreach and advocacy (SBOA), HIV prevention, and Motivational Interviewing (MI). Nesting would likely be considered the better choice if treatment philosophies differed among the intervention conditions. RA’s will be blind to condition at all assessment timepoints.

The HOME interventions integrate independent housing with opioid and related risk prevention services. Prevention services that will be delivered to both study groups include the following: SBOA, HIV prevention and MI [39] (Table 1). SBOA, HIV prevention and MI occur simultaneously in this approach. Each component of the intervention is described below in more detail. Youth who are randomized to the housing condition will receive housing assistance consistent with a Housing First approach. The housing assistance is not contingent on youth’s substance use or attendance of prevention services [19, 20]. To increase treatment participation, youth will receive a $5 McDonald’s, Burger King, Wendy’s, or other food gift card for every advocacy session they attend.

The advocates are hired from the community and may include formerly homeless youth/adults or may include social workers or those from a similar field with experience working with those experiencing homelessness and who are comfortable engaging and working with individuals in non-traditional settings. Research assistants (RAs) include undergraduate/graduate students or staff who will collect data and will be blind to intervention condition. Advocates and the RAs will participate in a two-day training led by the first author on engagement, tracking and assessment procedures, and weekly supervision. Because of the range of backgrounds of advocates, the training will be tailored to each person’s unique needs and skills. Training will include in vivo observation and modeling by the first author to ensure that the outreach workers/advocates are comfortable identifying and engaging homeless youth. MI and HIV prevention training of advocates consists of readings and at least a two-day training including role play exercises and ongoing supervision. Advocates will continue with role play meetings beyond the two days until it is determined that they are both comfortable and competent with employing the proposed procedures. Adherence to the intervention procedures will be independently evaluated by two clinical graduate students through digital recordings of sessions. MI and SBOA fidelity measures have been used successfully to ensure adequate fidelity in prior trials. A random sample of 20% of sessions will be coded to ensure that the procedures are delivered as intended. Percent performance of prescribed behaviors and undesirable behaviors will be used as measures of adherence.

The Ohio State University’s (OSU) Institutional Review Board has reviewed and approved the research study. Adverse events will be monitored, and whether youths show deterioration as a function of the intervention received will be assessed (e.g., increase in substance use, mental health symptoms, etc.). Should a client experience a serious adverse event that was unanticipated and believed to be related to study procedures, the OSU IRB and NIDA will be notified within 48 h. Adverse events that are unexpected and related to the study, but not meeting the definition of a serious adverse event will be reported to the IRB within 10 days of the MPI’s discovery of the event. The OSU IRB reviews the adverse event report and determines if the event is a result of study procedures. If the event is considered a direct result of study procedures, the MPIs and the board will meet within 48 h and will discuss the modifications to the protocol that are needed in order to prevent future adverse events. A data safety monitoring board comprised of three members experienced in clinical trials and/or working with vulnerable populations will be convened twice yearly to review safety and progress across all research projects. If any intervention is found to cause harm to participants, the trial will be stopped. The board’s summary report, any recommended changes, and other required data reports on adverse event cases will be submitted to the IRB and our NIDA project officer.

As a result of the COVID-19 pandemic, RAs’ and research subjects’ temperatures are taken prior to any face-to-face interaction, and all are provided with masks and hand sanitizer for use before and after in-person assessment at the drop-in center or other in-person meetings. Follow-up assessments and the majority of advocacy meetings are conducted by phone.

The main outcomes of the study are (1) time to OUD, and (2) percent days of opioid use in the past 90 days. The Structured Clinical Interview for DSM-5 Disorders (SCID) will be used to determine whether youth meet criteria for OUD at assessments. The SCID is a semi-structured diagnostic interview for DSM-5 diagnoses, which has good reliability, with kappas ranging from 0.65 to 1.0 [34]. The primary measure of substance use quantity and frequency will be the interviewer-administered Form 90 Substance Use Interview [50]. This interview yields total number of days in the 90 days prior to last use of all alcohol/drug use, total number of drugs used, age at first use, lifetime weeks of use and level of use, and has been used successfully with runaway and homeless youth [51]. As further validation of self-reported drug use, urine toxicology screens (tox-cups) will be collected for youth at baseline, 6- and 12-month follow-up. Urine screens are completed onsite with one-step BMC ToxCup® Test Kit (Branan Medical Corp., Irvine, CA), which provides instant readings for the detection of cannabinoids, amphetamines, methamphetamines, phencyclidine (PCP), cocaine/crack, and opiates.

Secondary outcome measures will be collected in order to evaluate the additional outcomes hypothesized to be positively associated with the housing intervention, including mental and physical health outcomes, additional measures of alcohol and other drug use, housing stability, employment or education, and HIV risk. The percent of days of other alcohol/drug use will measured based on the Form 90 (described above), as will measures of housing and employment (e.g. percent of days housed, percent of days employed, percent of days in school/training). Drug use consequences will be measured using the Shortened Inventory of Problems—Alcohol and Drugs (SIP-AD) [52]. The HIV risk behaviors measured will include condom use, intravenous drug use, history of STDs, and sexual activity with multiple and/or high risk partners [53, 54]. The survey items measuring these behaviors are adapted from prior surveys among homeless youth [55]. The responses to individual items are combined to yield a total score ranging from 0 from 7 points, consistent with prior studies [41, 56‐58]. Self-regulation will be measured with the Short Self-Regulation Scale, a 31-item instrument of cognitive competence that is highly correlated with the longer 63-item survey (alpha = 0.92) [59, 60]. The Short Self-Regulation Scale assesses youth’s ability to regulate cognitions and behaviors related to achieving their goals. The Beck Depression Inventory II (BDI-II) will be collected to measure mood, cognitive and somatic aspects of depression [61]. Resilience will be measured with the Brief Resilience Scale (BRS), a Likert-type 5-item instrument that measures the ability to bounce back or recover from stress [62]. The BRS is recommended for use in stressful contexts and has shown good internal consistency reliability (0.80–0.91) [62, 63]. Cognitive distortions will be measured with the Cognitive Distortions Questionnaire (CD-QUEST), a 15-item measure of cognitive distortions that assesses the frequency and intensity of common cognitive distortions over the past week [64]. The CD-QUEST has demonstrated good internal consistency (alphas = 0.80–0.91) and convergent validity across a variety of clinical and non-clinical samples [64, 65]. The Coping Inventory for Stressful Situations-Short Form (CISS-SFC) is a 21-item measure that yields 3 subscale scores: Task-, emotion-, and avoidance-oriented coping [66]. Reliability for the three subscales is 0.90, 0.88 and 0.83, respectively [67]. Health status will be assessed with the Short Form-6 is a standardized, internationally used instrument that provides a general measure of health status [68]. Its construct validity has been evaluated specifically with adult users of homeless day shelters [69]. Finally, the client satisfaction questionnaire (CSQ-8) will be collected, which is one of a limited number of standardized satisfaction measures that have been used widely across mental health services and has sound psychometric properties for substance users [70, 71].

Mediators were selected based on the biopsychosocial model and Bronfenbrenner’s Ecological Systems model. The current study hypothesizes that increased access to social resources and social support and decreased exposure to violence will promote positive changes in outcomes and that the effect will be strongest for those individuals randomized to receive HF with risk prevention services. Several measures will be collected as possible primary mediators of the treatment effects. First, the number of service contacts will be based on the study’s service contact log. The log identifies the youth’s frequency of contact with several different services (e.g. employment, medical, shelters/drop-ins). Second, the Working Alliance Inventory (WAI) will be used to measure the perceived level of connection and shared goals between the outreach worker and client and yields 3 subscale scores: task agreement, goal agreement, and bond development [72]. Third, the Social Network Inventory (SNI) will be collected as a measure of social support and has been used in multiple studies with homeless populations and high risk adolescents [73‐75], with test–retest reliabilities of 0.74 to 0.82 for the key SNI variables [76]. Satisfaction, substance use and illegal behaviors of network members is queried. This study will utilize a support contact measure based on the mean frequency of contact across all network members who the respondent indicated as having provided emotional, tangible, or other support. Finally, number of violence exposures (such as robbery or assault) will be assessed with self-report items from the HEQ.

Secondary mediators are assessed to evaluate whether individual resources are activated by the positive changes in the primary mediators described above. Measures of stress response (self-report and physiological) and self-efficacy will be collected as possible secondary mediators of the treatment effect. Individual stress will be measured using Cohen’s Perceived Stress Scale (PSS) [77], which is one of the most widely used instruments for measuring an individual’s perception of their stress levels and reports both high validity and reliability [78]. The PSS has 10 items, each scored on a scale of 0–4, and the total score ranges from 0 to 56, with higher scores indicating greater perceived stress. We measure the chronic physiologic stress response through cumulative cortisol concentration collected and assayed from hair. Hair grows approximately 1 cm a month, thus each 1 cm of hair growth approximates the mean cortisol level for the corresponding month. We will assay up to 3 cm of hair growth from the scalp at each 3-month data collection time point to capture the mean cortisol concentration over the preceding 3 months. Prior research found interventions can improve salivary cortisol regulation [79], and although limited in number due to the novelty of using hair to measure cortisol, the evidence suggests improvements in hair cortisol regulation as well [80‐82]. To collect the hair samples, approximately 10–75 mg (approximate width of shoelace tip when bunched) of hair is cut with professional shears from the posterior vertex region of the scalp as close to the scalp as possible [83, 84]. The hair will be assayed in duplicate for the mean cortisol value at the Ford lab via adapted protocol by Meyer et al. [83]. Finally, self-efficacy will be assessed as a mediator with respect to three health behaviors. First, specific self-efficacy for alcohol and other drug use will be assessed using the 8-item Drug-Taking Confidence Questionnaire-8 (DTCQ-8) [85]. Second, the 9-item HIV Behaviors Self-Efficacy Scale measures specific self-efficacy in regard to HIV risk behaviors [86]. It has shown high internal consistency (alpha = 0.77) and support for its construct validity [86]. Third, personal control self-efficacy will be assessed using Pearlin and Schooler’s 7-item Mastery Scale [87]. The instrument has robust psychometric properties and has proven validity with those experiencing homelessness [87‐89].

The impact of the two intervention conditions (i.e., 6-months housing + opioid and related risk preventive services and opioid and related risk preventive services alone) on the primary and secondary outcomes will be tested in Aim 1. Specifically, the primary outcome, time to the onset of OUD, will be analyzed using the discrete-time survival analysis. In this analysis, whether the onset of OUD has occurred at a specific time point is included as the indicator of a latent factor, the proportional odds for the hazard of OUD, which is predicted by the contrasts between conditions. Under the proportional hazards assumption, it is expected that youth assigned to the housing + preventive services condition will have longer delay in the onset of OUD than those in the preventive services alone condition. Time varying covariate will be explored if the proportional hazard assumption is violated. For the primary outcome (% days opioid use) and secondary outcomes that are continuous variables, latent growth models (LGM) will be conducted to estimate the trajectories of change across five time points (baseline, 3, 6, 9 and 12 months follow-up) to estimate the trajectories of change over time. Differences between treatment conditions on estimated growth parameters, including intercepts (i.e., initial status) and slopes (i.e., the rates of change), will be tested. It is expected that those assigned to the 6-month housing + opioid and related risk prevention services group will show greater decreases in opioid use (for those using opioids), as well as greater improvement in secondary outcomes, and that these improvements will maintain for a longer period of time, compared to those assigned to prevention services alone. To increase validity of the conclusions drawn from the analysis, the number of intervention sessions and contacts with other service providers will be covariates in the LGM analyses.

To test the proposed behavioral change mechanisms (i.e., social and individual resources), a series of path analysis will be conducted. Specifically, we will first test whether the primary mediator (e.g., service connections, social support) or secondary mediator (e.g., self-efficacy, stress) assessed at the 3- or 6-month follow-up mediates the association between intervention and outcome at 6-month and 9-/12-month follow-up, respectively. We expect that the intervention condition with the housing component will produce greater improvement in social/individual resources, which in turn will predict better outcomes. We will also estimate serial mediation models, in which both primary and secondary mediators are included, to test whether intervention first affects social resources, which then lead to changes in individual resources and subsequent outcomes. In these models, the baseline assessment of the mediators and outcomes will be controlled for in the analysis. Following MacKinnon and colleagues [92], the product of the coefficient of the path from the independent variable to the mediator(s) and the coefficient of the path from the mediator(s) to the outcomes will be computed as the indirect (mediation) effect between the intervention and outcomes. The strength and significance of the mediation will be estimated using a bootstrap sampling method [93].

The power analyses were conducted using the Monte Carlo simulation method. The MPI’s ongoing study testing the effect of housing support on homeless mothers reveals medium effect sizes in reducing substance use (d = 0.49) and depressive symptoms (d = 0.59), and a large effect (d = 1.26) in improving housing stability, favoring the housing intervention over TAU/assessment only. Given that the proposed study tests housing support against an active intervention, and the differences between the two conditions may be smaller than that between housing support and TAU, we assumed small-to-medium effect sizes when conducting the power analysis. For the latent growth analyses, with dichotomous predictors (contrasts between intervention conditions) that have regression coefficients of 0.15 (small-to-medium effect size) for the slopes of growth factors [94], a sample size of 240, with an overall attrition rate of 15%, can produce a power of 0.86 to detect the intervention effect on the growth rate of outcomes and a power of 0.94 for the survival analysis of OUD. For Aim 2, in the simulated mediation analysis, we again assumed a small-to-medium effect size for the intervention effect. Data from the MPI’s ongoing clinical trial show that the associations between social/individual resources and outcomes are of medium effect size or larger (e.g., personal control is associated with depressive symptoms at r = -0.46 and with general mental health at r = 0.43). Thus, we assumed medium effect sizes for the mediator to outcome pathways. Following the model specification suggested by Thoemmes et al. [95], the proposed sample size could provide a power of 0.93 to detect mediating effects for the one mediator model and a power of 0.82 for the serial mediation model with two mediators.

We will quantify the economic cost of both interventions, the housing + prevention services and prevention services alone. The cost comparison will lead to a cost-effectiveness analysis of the added benefit of housing. Resources used and their associated dollar cost are collected using activity–based costing method [96]. The steps for conducting cost analysis include (i) inventorying resources consumed, (ii) counting the number of units consumed in each resource category, (iii) estimating cost per unit of each resource type, and (iv) calculating total costs of the intervention and expressing this cost per-youth. Pre-implementation and implementation costs will be collected separately since the implementation strategies used during preparation phases are different than the implementation strategies used once an intervention is deployed [97]. Pre-implementation strategies include planning meetings, training of advocates, hiring, development of policies and procedures, and management. We will count the number of units used via weekly time sheets and other entries. Implementation activities include time spent recruiting participants, screening for eligibility, time of trained professionals delivering housing delivering SBOA, supervision, monthly housing and utilities, and other resources including mileage and office supplies. Time spent for each of these activities are recorded via Qualtrics survey. Housing and utilities are monthly recorded on an excel sheet. For the effectiveness analysis, the change in initiation of opioid use between assessments, as well as the change in escalation of opioid use, will be utilized.