Housing First Combined with Suicide Treatment Education and Prevention (HOME + STEP): study protocol for a randomized controlled trial

Identifying those at risk of suicide is a core objective in national strategies to prevent suicide [18, 19]. Unfortunately, youth who experience homelessness have limited access to medical care or mental health services and are thus unlikely to be screened for suicide in traditional settings. More efforts are needed to identify homeless youth at risk for suicide and link them to appropriate resources. The study will use the SSI-W to assess suicide risk at baseline and 3-month follow-up assessments, as well as past-year attempt at baseline (Fig. 1). In addition, if suicide attempts are reported to advocates at any time during the first 6 months of enrollment youth will be offered CTSP. Protocol to address imminent suicide risk through established emergency procedures (see below) will be followed for any individuals who screen for suicide risk. Those judged not at imminent risk will remain in the project and repeat screening; those with imminent risk will follow emergency procedures for hospitalization when appropriate and be referred to CTSP.

CTSP is based upon the theoretical assumption that people’s thoughts and interpretation of life events determines their emotional and behavioral responses [35]. Therefore, maladaptive cognitions associated with suicidal ideation are the primary focus of the treatment. The intervention protocol was developed to identify cognitive processes relevant to suicidal acts and targets vulnerability factors including: hopelessness, social isolation, poor problem solving, and impaired impulse control. The treatment is designed as a 10-session protocol including weekly or bi-weekly meetings (Table 1) [20, 21, 35]. All subjects with severe suicidal ideation at baseline will receive the therapy from a trained research staff member. The therapy will be delivered by phone or in-person, depending on participant preference and safety precautions for COVID-19. If in person, sessions will occur in a private room set aside for research at the youth drop-in center or, alternatively, in a private room set aside for counseling at a transitional housing unit (if participant randomized to HF). However, youth will be offered up to 9 additional maintenance sessions over a six-month period, tapered in frequency, consistent with treatment offered in prior studies [20, 21]. To increase treatment participation, youth will receive a $5 food gift card (e.g. McDonald’s, Burger King, Wendy’s, or other food gift card) for every CTSP session they attend.

The research staff will receive a 3-day training following the manualized “Cognitive Therapy for Suicidal Patients” [35]. Staff who are trained to serve as youth advocates will receive this in addition to other training for the parent study [16]. All sessions will be digitally recorded, and supervision will be led by the senior author (Dr. Slesnick). The recorded sessions will be rated for quality prior to supervision sessions which occur biweekly.

During research assessments or CTSP therapy sessions, if youth reports any thoughts of suicide at any point in the interview or self-report responses, the assessment of imminent risk will be completed, including 1) whether a plan is in place, 2) whether the means to carry out the plan are available, and 3) whether there is intent to carry out the plan, as well as reasons for wanting to harm themselves. Research staff will then call the senior author to review the youth’s responses and discuss next steps. If imminent risk is determined, the client is asked to accompany the staff member to seek immediate services at a mental health crisis intervention, stabilization, and assessment center. The crisis center is staffed by skilled physicians, nurses, psychologists, social workers, counselors, and technicians and is open 24 h a day. Crisis center staff will coordinate treatment, such as hospitalization, through local agencies when necessary. Services are available to anyone, regardless of the ability to pay. If the subject refuses to accompany the staff to the crisis center, then the police are called to accompany the youth.

The primary outcome for this study is suicidal ideation. Suicidal ideation will be assessed with the 19-item Scale for Suicide Ideation – Worst (SSI-W) [34], an interviewer-administered rating scale. The SSI-W is also used to assess youth’s eligibility for CTSP at baseline or over the 6-month period (as described above). The SSI-W will assess suicidality over the past 90 days, as in previous studies. The SSI-W was developed to identify the intensity of the most severe suicidal ideation experienced by the person. The scale measures the intensity of patients’ attitudes, behaviors, and plans to die by suicide during the time period [34]. Interviewers instruct respondents to recall the time when they had their most intense thoughts and desire to die by suicide. Respondents are then asked to keep this time in mind while the interviewer rates responses regarding how suicidal they were at that this specific time. The SSI-W has moderately high internal consistency (Cronbach alpha = .88) and validity, and associations with other measures of suicidal ideation including the SSI, single items from the Beck Depression Inventory, and Hamilton Rating Scale for Depression [31].

Suicide attempts will be assessed with a single self-report item similar to that used by the Ask Suicide-Screening Questions (ASQ) [36], that has been used in prior studies among homeless youth and is modified to ask about past 12 months at baseline, and past 3 months for follow-up assessments. Depression symptoms will be assessed with the Beck Depression Inventory (BDI), which is collected as a secondary outcome under the parent HOME trial.

If the HOME + STEP intervention is shown to be effective for improving outcomes, the study will explore primary and secondary mediators of the treatment effect. Substance use and housing stability, which have both been shown to improve with exposure to HF in adult samples [25, 26, 37], will be explored as primary mediators. The Form 90 will measure substance use (alcohol or other drug, excluding tobacco) and housing stability [38, 39], which is consistent with the parent trial and other prior studies among homeless youth [16, 40]. For these variables, percent days of any substance use and percent days house will be calculated. Hopelessness will be measured with the Beck Hopelessness Scale (BHS) [41], a self-report instrument that consists of 20 true-false statements designed to assess the extent of positive and negative beliefs about the future during the past week. The BHS is one of the most widely used measures of hopelessness and has demonstrated high internal reliability across diverse clinical and nonclinical populations (Kuder-Richardson reliabilities ranging from .87 to .93) [41]. The BHS has adequate one-week test-retest reliability in a psychiatric outpatient sample (r = .69) and high three-week test-retest reliability in a college student sample (r = .85) [42]. Stress coping will be measured with the Coping Inventory for Stressful Situations-Short Form (CISS-SFC) [43], which is collected as a secondary outcome of the parent trial and has been used in prior studies of homeless youth.

To increase our understanding of the determinants of CTSP implementation including its acceptability, appropriateness, and feasibility, we will use qualitative data collection. Up to 20 youth who receive CTSP, will be interviewed. Qualitative interviews will be conducted by trained research staff. All interviews will be audio recorded and transcribed by a transcription service. Youth will be queried about their experience with CTSP at the end of the treatment period (6 months) and asked to discuss aspects of the CTSP intervention they benefited from and aspects they did not find helpful. This information will give a better understanding of factors that promote engagement among youth as well as factors that may prevent youth from engaging in the intervention. We will also interview research staff who delivered CTSP and community stakeholders (n = 5). Community stakeholders will include representatives from organizations serving homeless youth, such as the city’s homeless youth drop-in center leadership and other service providers. Interviews of therapists and stakeholders will include open-ended questions about what it would require to implement CTSP (e.g., limitations/barriers). In addition, specific probes will be included that focus on domains from a widely used implementation framework, the Consolidated Framework for Implementation Research (CFIR) [44].

CTSP treatment fidelity will be assessed based on the “dose” delivered using study records and therapist ratings. Study records will measure treatment attendance including number of sessions and proportion of youth completing high levels of therapy (80%). The Cognitive Therapy Rating Scale (CTRS) [45], an 11-item scale developed to assess therapist competence, will also be used to assess fidelity. Items are scored on a 7-point scale: 0 (Poor), 1 (Barely Adequate), 2 (Mediocre), 3 (Satisfactory), 4 (Good), 5 (Very Good), and 6 (Excellent). The 11 items are summed to yield a CTRS total score, ranging from 0 to 66 (≥40 considered competent). The CTRS will be completed by the senior author during supervision. CTSP costs will be measured as the advocate time for training, delivering therapy, supervision, and contacting youth for follow-up appointments or rescheduling as well as the labor, equipment, and space costs, based on an activity-based costing method and semi-structured, in-person interviews of therapists and other relevant personnel [46].

The patterns of missing data will be first examined before each analysis. If data are missing at random [47], they will be estimated using full information maximum likelihood (FIML) or multiple imputations (MI) method. When data are missing completely at random (MCAR) or are missing at random (MAR), both FIML and ML produce unbiased results [48]. The main statistical program used for this study, MPlus, can handle missing data by providing FIML estimation. If the missingness cannot be explained by observed data, that is, data are Missing Not at Random (MNAR), data analysis will be conducted using the pattern mixture model framework [49].

The impact of housing on the primary (suicidal ideation) and secondary outcomes (depressive symptoms, suicide attempts) over time will be tested between the two main treatment conditions: housing combined with risk preventive services and risk preventive services alone. For these analyses we will control for CTSP treatment attendance, which has been positively associated with treatment outcomes [50, 51]. Specifically, the primary and secondary outcomes will be analyzed using latent growth models (LGM) to estimate the trajectories of change across five time points (baseline, 3, 6, 9, and 12 months follow-up). LGM can also estimate joint trajectories with multiple outcome variables. Differences between the main 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 housing combined with risk preventive services group will show greater decreases in suicidal ideation and depressive symptoms, and that these improvements will maintain for a longer period of time, compared to those assigned to the prevention services alone group. We anticipate that number of participants reporting suicide attempts in the current sample will be relatively low. Thus, in the LGM analysis, we will dichotomize the measure of suicide attempts (i.e., presence vs. absence of suicide attempt) to test whether the likelihood of suicide attempt decreases faster in the combined housing and services group than in the services alone group. 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.

A series of path analysis will be conducted to test the mediation hypotheses. Primary mediators: we will first test whether percent of days housed and percent days of substance use assessed at the 3- or 6-month follow-up mediates the association between intervention (combined house and services vs. services alone) and outcomes at 6-month and 9−/12-month follow-up, respectively. We expect that the intervention condition with the housing component will produce greater improvements in percent day housed and substance use, which in turn will predict reduced suicide ideation, depressive symptoms, and likelihood of suicide attempt. Secondary mediators: we will test the effects of hopelessness and stress coping on the relationship between primary mediators and outcomes. We expect that lower housing instability and substance use at 3 months will be associated with lower hopelessness and greater stress coping at 6, which will in turn be associated with reduced suicidal ideation, depressive symptoms, and likelihood of suicide attempts at 9−/12-months. Following MacKinnon and colleagues [52], 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 [53]. Similar to the primary aim, we will control for CTSP treatment attendance, number of intervention sessions, and contacts with other service providers in the analysis.

We will only include youth at high risk of suicidality in the analyses to provide an estimation of the effect of housing on high levels of suicidality between the two groups: housing combined with services and services alone. The expected sample size is approximately 100. Repeated measures ANOVA (RMANOVA) will be conducted to examine the differences between the groups in continuous outcomes, suicidal ideation, and depressive symptoms. Specifically, a 3 (treatment condition) × 5 (time points) RMANOVA will be conducted with treatment as the between-subject factor and time as the within-subject factor, in order to test the differential change pattern over time between the two groups. For suicidal attempts (both count and dichotomous variable) generalized linear mixed models (GLMM) will be conducted. In the event that the high-risk sample would not provide sufficient power for the GLMM analysis, further exploration analysis will be conducted on the dichotomous suicidal attempts variable to examine the housing effect at the end of treatment (i.e., 6-months follow-up) and the end of the study (i.e., 12-month follow-up), using logistic regression, controlling for the baseline assessment.

To understand how the determinants of CTSP implementation vary across settings, we will use a mixed-methods design. In the quantitative phase, we will examine and compare quantitative data on fidelity and costs across both study arms. Average start-up and ongoing estimated costs for all STEP-related services will be presented overall and across study groups. Qualitative interviews will be iteratively coded using a modified grounded theory approach – modified in that we will orient our initial open coding process around what Patton describes as “sensitizing concepts” including issues related to acceptability, appropriateness and feasibility [54, 55], NVivo 11 software will be used to organize and code the data. Data will be used to identify salient contextual features service delivery that need to be addressed in future implementation efforts and will provide greater explanation about why there are (or are not) differences in implementation outcomes across conditions. Results may also provide insight into which implementation strategies might be needed to deliver CTSP.

The power analyses were conducted using the Monte Carlo simulation method for the Primary Aim and Secondary Aim 1 and using GPower program for Secondary Aim 2. We assumed an overall attrition rate of 15% when calculating the power. The MPI’s ongoing study testing the effect of housing support on homeless mothers reveals medium effect size in reducing depressive symptoms (d = .59), favoring the housing intervention over treatment as usual (TAU)/assessment only. Few studies examined the effect of housing on suicidality; a recent observational study reported a small effect size from baseline to 2-year follow-up [27]. Given the large difference across studies and 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 Primary Aim, with dichotomous predictors (contrasts between intervention conditions) that have regression coefficients of .15 (small-to-medium effect size) for the slopes of growth factors [56], a sample size of 240 can produce a power of .86 to detect the housing effect on the growth rate of outcomes. For Secondary Aim 1, the MPI’s clinical trial of housing support suggested a large effect size (d = 1.26) in improving housing stability and a medium effect size in reducing substance use (d = .49). Similarly, a meta-analysis reported large effect sizes (d = 1.24 on average) of Housing First approaches in increasing housing stability [57]. Thus, we assumed a medium effect size for the intervention-to-mediator paths and a small-to-medium effect size for the mediator-to-outcome paths. Following the model specification suggested by Thoemmes et al. (2010) [58], the proposed sample size could provide a power of .91 to detect mediating effects for models with one mediator and a power of .82 for the serial mediation models with two mediators. For Secondary Aim 2, assuming a small-to-medium effect size (i.e., f = .20), the power to detect a group difference in an RMANOVA with 5 repeated measurements is .84. For the GLMM, to achieve 80% power, the effect size of the group differences in the slope needs to be of medium size (regression coefficient of .20). A smaller than medium effect size will not provide sufficient power to detect housing effect on suicidal attempts. However, our goal is to estimate an effect size that can be used to guide sample-size estimates for a future fully-powered study.