MSM use substances at higher rates than the general population, increasing HIV risk. High rates of drug and alcohol use among MSM relative to the general population have been documented [5-7], and our research has identified higher rates of drug use among YMSM compared with heterosexuals [8-10]. Higher rates of drug use have also been documented among MSM, including YMSM in tandem with sexual activity [11,12]. However, drug-use patterns seem to differ among YMSM with increased rates of cocaine among YMSM, which could have implications for HIV risk given that stimulant use has been linked to condomless anal sex (CAS) [13-16] and higher risk of HIV infection and other STIs [17]. Nearly half of black MSM with newly diagnosed HIV infection (48%) reported substance use during their last anal sex encounter [11]. Substance use has been found to increase sexual risk behavior among MSM [18], placing them at high risk for CAS and HIV seroconversion, exchange sex, and greater number of sexual partners [19]. Our own research using event-level data for the previous 30 days has shown that substance use strongly and significantly predicts the odds of whether YMSM will use a condom [20]. A number of other studies have looked at the impact of substance use on sexual behavior and increased odds of seroconversion [11]. Different substances are associated with sexual risk behavior among specific groups. Among Latino MSM, methamphetamine use [21,22] and among black MSM higher rates of marijuana use have been linked to sexual risk [23]. Therefore, there is a critical need for brief, culturally appropriate, effective behavioral interventions that improve self-management to reduce new HIV infections among substance-using YMSM.

Motivational interviewing (MI) has the potential to improve self-management behaviors in terms of promoting sexual health and reducing substance use among YMSM. There is strong evidence that MI is a culturally appropriate and effective approach for working with racial and ethnic minority populations [24] who are disproportionately affected by HIV. One meta-analysis of MI found a greater effect among minorities [25]. MI has been recommended as particularly effective when working with YMSM [26]. MI promotes increased intrinsic motivation to change and, when paired with information regarding health risk behaviors, reinforces the individuals’ right and capacity to make well-informed health self-management decisions for themselves [27].

YMHP, a manualized structured 4-session intervention using MI and problem-solving skills building, has been listed as a best evidence intervention by the Center for Disease Control and Prevention (CDC) and is the only intervention to reduce both CAS and substance use among YMSM [28]. The CDC’s endorsement of YMHP was informed by results of the National Institutes of Health (NIH)-funded efficacy trial. A total of 143 HIV-negative YMSM (aged 18-29 years) who reported CAS with a high-risk partner and at least 5 days of drug use in the last 3 months were randomized to receive 4 sessions of YMHP or 4 sessions of health education on sexual risk and substance use. The majority [63%(90/145)] of the sample were YMSM of color (black: 21%, Latino: 28.7%, or multiracial: 13.3%). Retention rates were high: 79% at the 12-month follow-up and 88% completed at least 3 of the 4 sessions. Master’s level therapists were trained in MI with ongoing fidelity monitoring using the Motivational Interviewing Treatment Integrity (MITI) coding system. YMSM who received YMHP reduced their cross-time averaged odds of ever using any drug by 67% over the 1-year follow-up. Within-condition analyses showed that YMSM who received YMHP reduced their cross-time averaged odds of ever having CAS by 83% over the 1-year follow-up. YMSM in the YMHP condition were 21% less likely to report CAS on days when they did report drug use relative to men receiving education. This has been the first and only RCT with HIV-negative YMSM to show significant reductions in both CAS and substance use. As such, YMHP has the potential to have a significant impact on YMSM seeking sexual health or HIV counseling and testing services at clinics.

Research on telephone-based MI has consistently found that it produces significant improvements in a wide range of physical health challenges [29-31]. Furthermore, telephone-based MI reduces mental health–related problems [32] and alcohol-related problems [33] as well as sexual risk taking [31,34-36] among HIV-positive individuals. Research comparing the effects of telephone-based MI with face-to-face (clinic-delivered) interventions has produced equivocal results. Across studies examining physical activity, mental health, and substance use outcomes, findings suggest no significant differences in delivery method [34-37]. Carey et al examined the relative efficacy of a telephone versus face-to-face intervention for alcohol use and observed a significant interaction of delivery method with gender [38]. Women on average had better outcome in the face-to-face condition, whereas men responded equally well to both delivery methods.

Notably, the issue of health care access has not yet been examined as a moderator of relative effectiveness. One advantage of telephone-based MI is that it significantly reduces patient burden. It is, therefore, plausible that it will show superior effects among YMSM who experience barriers to health care access. However, it is also possible that remote delivery (via phone or video chat using Skype or FaceTime) will decrease engagement and the quality of the relationship between the clinicians and the participant. This might result in clinic-based delivery being superior among YMSM who have better access to health care. Understanding how health care access intersects with delivery method will substantially inform implementation decisions at clinics and other agencies seeking to utilize YMHP.

Integrating implementation science into a comparative effectiveness trial (CET) can minimize the science-practice gap [39]. MI providers need not be clinicians—1 study conducted by our team comparing community health workers (CHWs) with clinicians found both were equally effective in providing high-quality MI and that clients were more likely to be retained in HIV care when working with CHWs [40]. CHWs are commonly integrated into clinics and often play a central role in providing HIV prevention services, including HIV counseling and testing. Training CHWs to deliver evidence-based interventions is a critical step toward realistic and cost-effective implementation [41]. CDC has called for expanded use of CHWs in services for chronic disease [42] with attention to implementation and training [43-50].

Several steps can be taken in a CET of YMHP to promote adoption and sustainability. First, using staff embedded in the existing clinic setting can build capacity for implementation. Second, CHWs have long been the cornerstone of integrating support services into HIV-related prevention efforts [51]. Training CHWs to deliver interventions is a critical step toward realistic and cost-effective implementation. Research has documented the amount of training needed to obtain MI fidelity, concluding that initial training followed by ongoing coaching is required [52-55]. Such training can be costly when relying on outside trainers. Thus, a train the trainer model where expert trainers provide local supervisors with MI coaching skills might be more sustainable [56]. Finally, CET designs help gain information about implementation [57]. In a Hybrid 2 trial, the goal is to dually determine which treatments work in which settings and to simultaneously answer implementation science questions about the potential barriers/facilitators to a treatment’s widespread and continued implementation.

The Center for HIV Educational Studies and Training (CHEST) at Hunter College will use a variety of recruitment strategies to recruit participants for this study. First, as a result of CHEST’s role as the Management Core and Recruitment and Retention Center (REC) for Scale It Up, we will utilize 3 Subject Recruitment Venues (SRVs) in our network (Wayne State University Prevention in Detroit, University of Miami, and Children’s Hospital of Philadelphia) to complete clinic- and field-based recruitment. All 3 SRVs, as well as CHEST, have extensive relations with the gay, lesbian, bisexual, and transgender communities; community service organizations; health service organizations; and providers for MSM. In this aspect, recruitment will occur from routine walk-in visits for HIV testing. Partnering clinics will include offering this study through outreach activities and through their HIV testing services they provide in their clinics. Information about the study will be included in the institutional review board (IRB)-approved palm cards, brochures, and flyers at each clinic site (see Figure 2). This information will be displayed in waiting rooms and exam rooms at the clinic as well as at locations of mobile testing events and outreach shifts to be passed to potential participants to both encourage HIV testing and promote the YMHP study. This method has worked well in the past for numerous studies and takes advantage of when patients have more idle time to learn about the study.

YMSM who test HIV negative at the 3 SRVs in Miami, Detroit, and Philadelphia or through mobile testing efforts provided by community collaborators will be offered the opportunity to participate in YMHP. Those interested will be asked to complete a brief Study Screener on a study iPad to collect demographic and behavior questions related to eligibility criteria. Participant ID numbers are generated and assigned through Qualtrics during the screening process to all potential participants, including those who screen ineligible. If eligible, a screen will be displayed informing the potential participant of his eligibility, and the site study staff person will then schedule the potential participant for a baseline assessment. We anticipate enrollment of 5 participants per month, per SRV, and have allocated staffing resources to ensure this rate.

In addition, CHEST will assist in referring potentially eligible participants to the YMHP study through existing Web-based recruitment efforts. CHEST utilizes the Hunter College IRB-approved online master screener (OMS) used to preliminarily screen individuals who are interested in participating in studies being conducted through CHEST. If an individual is preliminarily eligible for a study, the individual is asked to provide contact information to CHEST for follow-up. For the purposes of this study, the OMS will be used to refer potentially eligible YMSM to HIV testing sites by sending them an email referral informing them about the YMHP study and where to go to determine eligibility after completing the OMS. Potential participants will also be called by CHEST staff to complete the Study Screener over the phone. If they meet the study criterion, they will be scheduled by the study staff to attend the clinic for HIV testing and YMHP enrollment. The OMS, in this instance, will primarily be used as a referral mechanism for the study, directing participants to which study they might be eligible for, including YMHP.

All interested participants are assessed for eligibility by completing a brief Study Screener. Study inclusion criteria include (1) being aged between 15 and 24 years; (2) currently identifying as male (regardless of birth sex); (3) receiving an HIV-negative test result from a study site or mobile HIV testing in the past 90 days; (4) having sex with men in the past 90 days; (5) reporting at least 3 days of illicit drug use or heavy drinking (5 or more drinks) in the past 90 days; (6) reporting at least one episode of CAS with a male partner in the past 90 days or a positive STI test result in the past 90 days; (7) living in the Detroit, Miami, or Philadelphia metropolitan areas; and (8) having the ability to communicate in English. Exclusionary criteria include the following: (1) participants whose mental, physical, or emotional capacity does not permit them to complete the protocol as written; (2) currently taking Truvada as pre-exposure prophylaxis (PrEP); and (3) 5 or more days of injection drug use in the past 90 days.

This study examines 4 coprimary outcomes related to substance use (1 outcome) and sexual health management (3 outcomes). These are measured during baseline; immediate postintervention; and 3-, 6-, 9-, and 12-months follow-up assessments. Specifically, sexual health management is measured by 3 outcomes: (1) decreased STIs, (2) decreased CAS, and (3) increased PrEP uptake/adherence. HIV/STI testing occurs after immediate postintervention at 3- and 9-month follow-up assessments.

Substance use is measured using the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) [60]. ASSIST is often utilized as a screener in primary care settings for substance abuse. Specifically, ASSIST assesses participants’ use of alcohol, tobacco, cannabis, sedative, hallucinogens, inhalants, opioids, and other drugs. The primary substance use outcome will be substance use days during the past 90 days.

To measure CAS, participants complete a series of questions pertaining to their sexual behavior with main and casual partners [61]. Participants estimate their total number of sex partners in the past 3 months. Participants also indicate if they had unprotected, receptive or insertive, anal sex with partners who were HIV positive or of unknown HIV status.

The Motivational PrEP Cascade is a series of 21 questions that are designed to assess PrEP treatment uptake and adherence [62-64]. Participants report their familiarity with PrEP; experiences and acceptability; as well as PrEP contemplation, preparation, action, and maintenance utilizing the Transtheoretical Model of Change framework. This includes assessing willingness and intentions for PrEP uptake. The Motivational PrEP Cascade also assesses when participants begin and stop taking PrEP [65].

Moderators and mediators of intervention effects will be assessed with health care access and other self-management constructs for decision making, problem solving, self-regulation, and provider and health care relationship. Health care access is a series of 9 questions adapted from Williams and Chapman’s unmet health and mental health need [66]. Items included whether youth were not able to get health care services (ie, routine physical examination, see a provider for sexual health care or get access to PrEP, psychological or emotional counseling, and counseling for drug or alcohol use) in the past year when they felt they needed those services.

Self-management constructs included the Behavior Rating Inventory of Executive Function (BRIEF-A) [67] and the Patient Activation Measure [68]. BRIEF-A is a series of 22 questions asking patients to rate whether certain problems (eg, bothered by having to deal with changes, do not plan ahead for tasks, and problems completing my work) occurred in the past month. Responses included never, sometimes, and often. For patient activation measure, participants were asked on a 4-point scale, ranging from strongly disagree to strongly agree or not applicable, a series of 10 questions related to their health. Examples of some of the questions include, “When all is said and done, I am the person who is responsible for taking care of my health” and “I am confident that I can follow through on medical treatments I may need to do at home.”

The primary hypothesis is that, although the main effect of YMHP delivery modality will be nonsignificant, there will be a significant interaction between access to health care and delivery modality. Specifically, it is hypothesized that remote-based YMHP will demonstrate greater improvements in sexual health management (as measured by decreased STIs, CAS, and increased PrEP uptake/adherence) as well as greater reductions in substance use, compared with clinic-based YMHP, among YMSM who do report barriers to health care access. In contrast, it is hypothesized that clinic-based YMHP will demonstrate greater improvements in sexual health management and reduced substance use among YMSM who do not report barriers to health care access.

The protocol lead and analytic team will test the effect of clinic-based versus remote-based delivery on STI rates, alcohol and drug use behavior, CAS, and PrEP uptake through multilevel growth mixture modeling (GMM). A separate model will be run for each outcome. Each model will be a 2-level model in which individuals (level I) are nested in clinics (level II). This approach controls for the nonindependence of individuals within clinics. As 3 sites provide extremely limited predictive power at level II, no site covariates are included in the model. The models are empty at level II.

In a GMM, a latent growth curve with an intercept and linear slope factor is specified. Latent class analysis is applied to these 2 growth components (intercept and slope) to identify groups of individuals who share trajectories. For example, with regard to drug or alcohol use, immediate and sustained responders might have the lowest postintervention intercept and a flat slope. Meanwhile, non-responders might have the highest postintervention intercept and a flat slope. In contrast, delayed responders might have a high postintervention intercept but a significant negative slope, indicating reductions in missed medication over the follow-up period.

If modeling results indicate that discrete classes are not present, we will proceed with analyses in which the growth factors (intercept and slope) are predicted directly by intervention condition (and demographic factors found to be associated with condition after randomization or with attrition over follow-up). GMMs can subsequently incorporate predictors of class membership. These analyses can be conceptualized as a multinominal logistic regression with the latent trajectory-class membership constituting the outcome. The predictors of primary interest will be intervention condition, the presence of any barriers to health care access, and the interaction between condition and barriers to access. We will include as covariates any demographic variables that were associated with condition after randomization or with attrition over the follow-up period.

Power analysis was conducted based on the 4 outcomes of STIs, PrEP uptake, alcohol and drug use, and CAS. First, we analyzed power assuming independence of participant observations (assuming that the nesting of people within clinic was irrelevant). With regard to STIs, we utilized the repeated measures module of PASS 13.0 [NCSS Statistical Software] to examine power to detect odds ratio differences in a repeated measures design. Specifying compound symmetry, we allowed ρ to vary between .2 and .5. Assuming the prevalence of STIs in remote delivery condition varies between .05 and .15, the proposed design (N=270) has power .80 to detect an odds ratio of 0.20 to 0.50. Similarly, with regard to the odds of PrEP uptake, allowing the rate of uptake in the remote delivery condition to vary between .05 and .20 and allowing ρ to vary between .4 and .7, the proposed design has power of .80 to detect an odds ratio of 1.9 to 2.3. With regard to number of alcohol and drug use days, we utilized the Tests for Two Poisson Means module in PASS 13.0. On the basis of the data from our previous studies, we allowed the rate of heavy drinking in the remote delivery condition to vary between 7 and 9 days during a 30-day assessment. The study is adequately powered to detect a 3% reduction in the number of heavy drinking days in the in-person condition at any single follow-up point. Allowing the rate of substance use in the remote delivery condition to vary between 3 and 5 days during a 30-day assessment period, the study is adequately powered to detect a 4% reduction in drug use in the in-person condition. A similar analysis was conducted with respect to CAS. On the basis of our previous research, we allowed the rate of CAS in the remote delivery condition to vary between 2 and 4 acts in a 30-day assessment period. The study is powered to detect a difference as small as 5% between the remote delivery and in-person YMHP conditions. The nesting of individuals within sites has the potential to reduce power because substantial variability in outcome across sites can obscure level-II treatment effects [69,70]. The design effect can be used to tailor power analyses calculated under assumptions of independence. In the case of a level-I predictor with a fixed effect that is uncorrelated with other covariates in the model, the design effect is equal to the 1−ρ, where ρ is the intraclass correlation or the percentage of variance accounted for by variability between sites. In previous ATN intervention trials, between clinic site variability in HIV-related outcomes did not differ significantly from 0. We anticipate an absence of variability across clinics, suggesting that the design effect would result in a negligible reduction in power. Finally, a sample size of 270 is sufficient to detect a moderation effect with an f² of 0.02. Cohen [71] designates this as a small effect; however, recent work has characterized an effect of this size as moderate to large as applied to moderation [72].

To enhance the likelihood of uptake if effective, the cost-effectiveness of 2 delivery models of YMHP in reducing sexual risk and substance use will be assessed utilizing CDC’s guidelines for cost-effectiveness analysis on HIV infections averted. The economic analysis will have 2 components: (1) a cost analysis of the YMHP intervention and (2) an incremental cost-effectiveness analysis that compares the value of clinic-delivery of YMHP over remote delivery. We will first estimate the marginal costs of delivering the 2 formats of YMHP. Using data from the modified Drug Abuse Treatment Cost Analysis Program [73,74] and study contact and expenditure records, key statistics from the cost evaluation will include the total annual economic cost for each program, weekly economic cost per client, and total economic cost per intervention session [75-77]. To highlight the relative contribution of the various cost components and necessary future budgeting, we will also perform a descriptive analysis of the cost accounted for by resource category.

The mean aggregate cost of the interventions will be used as inputs in the cost-effectiveness model. Cost-effectiveness will be modeled for clinic-delivery of YMHP as compared with remote-delivery for the differences in CAS and predicted through Markov modeling for 5 and 10 years and over a lifetime using varying assumptions about decay of the effect of the intervention over time. Modeling will be performed from the perspectives of (1) a third-party payer, (2) the medical care system, and (3) society.

Using GMM models similar to those described above in primary outcome analyses, we will examine the cross-time effects of intervention on the 5 dimensions of self-management. Where a significant between-condition difference in self-management is detected, we will explore mediation by examining whether intercept and/or growth factors for that dimension of self-management in turn predict outcomes. Indirect effects in GMM will be evaluated using bootstrapping estimation where possible. When this is not possible, a constraint approach will be employed. This approach involves comparing the fit of 2 models: one model in which the product of constituent pathways is constrained to be 0 and another in which the product of the direct effects is unconstrained. A significant Chi-square test associated with this comparison indicates that the constraint significantly diminishes model fit and constitutes evidence of the significance of the indirect pathway [78].