Comparative Effectiveness of Community-Based vs Clinic-Based Healthy Choices Motivational Intervention to Improve Health Behaviors Among Youth Living With HIV

This randomized clinical trial assesses the comparative effectiveness of the Healthy Choices intervention when delivered in a clinic- vs home-based setting on medication adherence and alcohol use among youths and young adults living with HIV.


Introduction
Youth living with HIV make up more than one-fourth of new infections in the United States 1 and have high rates of risk behaviors, such as alcohol use and nonadherence to medication, 2,3 but are significantly understudied. Alcohol consumption among persons with HIV exacerbates health problems and accelerates HIV disease progression. [3][4][5][6] Optimal adherence to antiretroviral treatment (ART) decreases morbidity and mortality, 4,5,[7][8][9] the potential for the development of drug-resistant strains of HIV, [10][11][12] and HIV infectiousness. [13][14][15][16] Healthy Choices, a 4-session, 10-week intervention based on motivational interviewing, 17 is the only intervention (to our knowledge) to demonstrate improvements in viral load and alcohol trajectories in youths living with HIV in a full-scale, multisite randomized trial when delivered by members of the research team. 18,19 Thus, testing the intervention in a real world clinical setting when delivered by members of the HIV clinical care team, such as community health workers (CHWs) and local supervisors, can provide evidence to inform practitioners on clinically effective and cost-effective treatments that are available prior to dissemination on a wider scale. [20][21][22][23][24] Such effectiveness trials are the next stage on the translational science spectrum. 25 In the original Healthy Choices trial, although most youths attended at least 1 session, less than half completed all 4 sessions. 26 It is possible that short-term viral load improvements may have been sustained if participants received a full exposure of treatment. Many researchers have suggested home-based service delivery to increase access to and engagement in behavioral health services, especially when delivered by CHWs, but this remains untested in youths living with HIV. 27,28 Thus, this comparative effectiveness trial compared Healthy Choices in a home-based vs clinic-based setting, delivered by CHWs and supervised by members of the clinical care teams, on primary outcomes of viral load and alcohol use. We hypothesized that home-based delivery would result in greater improvements in viral load and alcohol use compared with clinic-based services by decreasing barriers to participation and by delivering treatments in the youth's natural ecology.

Participants and Procedures
Participants were recruited from 5 adolescent HIV clinics in Chicago, Illinois; Detroit, Michigan; Memphis, Tennessee; Los Angeles, California; and Philadelphia, Pennsylvania. Inclusion criteria included positive HIV status, age of 16 to 24 years, ability to complete questionnaires in English (96.7% of participants spoke English at home), currently prescribed ART, detectable viral load within the last 4 weeks, and alcohol use in the past 12 weeks. Exclusion criteria included having an active psychosis that resulted in an inability to complete questionnaires. Participants were randomized to intervention delivery in the clinic (n = 93) or at home (n = 90).
This report follows the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials (Figure 1). The trial protocol (Supplement 1) was approved by each site's institutional review board, and a certificate of confidentiality was obtained from the National Institutes of Health. Data were collected between November 1, 2014, and January 31, 2018. Clinicians at each site gave a description of the study to potential participants. If the patient was interested, a researcher obtained verbal consent for screening. When eligible, written informed consent was obtained and a waiver of parental consent was permitted for participants younger than 18 years. Using a Qualtrics survey, participants were then randomized (1:1 stratified by site), and intervention sessions were scheduled immediately after the baseline assessment. Retention strategies included reminder calls at different times of the day and collaboration with clinic outreach staff. Bus tickets or parking reimbursement were given to youths randomized to the clinic as consistent with the sites' standard of care.

Healthy Choices Intervention
The 4-session intervention was identical for the clinic and home locations and has been previously described. 18,29 Community health workers were paraprofessional staff with at least a high school degree or equivalent and no more than a bachelor's degree. Supervisors were clinicians with a master's degree already employed by the clinic. In the original study, participants could focus on sexual risk or other substance use, but for this study the focus was viral suppression via improved adherence and reduced alcohol use. In session 1, participants chose which of the 2 behaviors to discuss first, and, using standard motivational interviewing strategies, the CHW engaged the youth,  collaboratively focused the conversation on the target behavior, evoked motivational statements, and guided the development of an individualized change plan based on readiness to change while utilizing motivational interviewing strategies to support autonomy. The CHW delivered structured personalized feedback on risk behaviors and health status based on the baseline assessment and offered informative handouts to address any knowledge gaps. The second session (week 2) followed the same format but focused on the second target behavior. In the subsequent 2 sessions (weeks 6 and 10), the CHW reviewed the individualized change plan, continued to monitor and encourage progress, and guided the youth to problem-solve barriers and maintain any changes made. Based on randomization, the intervention was delivered in a private room in the clinic or in the home or a community-based location of the youth's choice. Participants received no monetary incentives for attending the intervention sessions.

Intervention Fidelity
Community health workers participated in a 2.

Data Collection Methods
Baseline assessment data were collected in the clinic. The alcohol use timeline followback questions were recorded on paper by an interviewer, and all other data points were collected using audiocomputer-assisted self-interviewing. The same measures were completed at baseline and at 16, 28, and 52 weeks. Participants were paid $50 for each data collection.

Variables Measured
Age, race/ethnicity, biological sex, gender identity, and sexual identity were assessed using audiocomputer-assisted self-interviewing. Blood for quantitative plasma HIV RNA assays was collected within a 2-week window surrounding the data collection time point. Viral load values reported as below the level of detection were set to 20 copies per milliliter. All viral load values were transformed to log10 for the statistical analyses.
We measured 2 dimensions of alcohol use: severity of problems and number of drinks per week.

Severity of problems was measured using the Alcohol, Smoking, and Substance Involvement
Screening Test (ASSIST), a brief screening tool to assess the individual's level of alcohol use and problems related to use. 31 This test yields a summary score for severity of alcohol use ranging from 0 to 33, with higher scores indicating greater severity. Number of drinks per day was measured using the timeline followback procedure. 32 The interviewer asked the participant to use a calendar to The LGCA results are often quite similar to findings observed using multilevel modeling with fixed and random effects. 36,37 However, they have the advantage of being able to capture unique trajectories that may be of interest to the researcher, and they can avoid potential aggregation bias because trajectory changes are captured at the level of the individual. The drawback of using LGCA is that it is based on structural equation models, which are less well known to clinicians, and that LGCA model fitting is complex and requires detailed documentation (eAppendix in Supplement 2). We chose to use LGCA for this study because it allows the reader to more easily compare these effectiveness trial results to the original motivational interviewing study report, 18 and it assures that we used a flexible model in which the unit of analysis is the individual's trajectory over time. The threshold for determining significance was P < .05. We used 2-sided tests of significance in all cases.

Study Power
The planned overall sample size was 500 patients, with 100 patients per site for 5 sites. Assuming We conducted analyses to test for the presence of differential attrition between conditions (eTable 1 in Supplement 2). First, we conducted χ 2 analyses to test for differential attrition by study arm at each follow-up. We also tested for differential attrition at each follow-up across site, as well as several demographic characteristics (race, education, employment status, sexual identity, and gender identity). Finally, we conducted a series of point-biserial correlations to examine attrition at each follow-up point by age and baseline outcome values (viral load and alcohol severity and frequency).

Multivariable Modeling of Effect
A piece-wise LGCA approach, 18,33-35 as discussed previously, was used to assess between-condition differences in viral load and frequency across postintervention time points. Piece-wise LGCA is well suited to situations in which a developmental trajectory has a known pivot point (also called a knot).
In these instances, slopes can be specified that quantify each component of the trajectory. Due to the need to correct for baseline differences in ASSIST scores when calculating follow-up values, we

JAMA Network Open | Infectious Diseases
Effectiveness of Community-vs Clinic-Based Healthy Choices Intervention Among Youth Living With HIV utilized an alternative growth modeling procedure. 38 This analysis modeled only postintervention points corrected for baseline ASSIST values.
Models were estimated using Mplus, and examples of the general coding approach is available through their website. 39 For each outcome, a model was estimated that included 3 latent growth components. These included an intercept (with factor loadings uniformly equal to 1) and 2 slopes.

Results
Demographic data for the total sample are presented in Table 1 t 181 = −2.29). In subsequent analyses of primary outcomes, demographic covariates were excluded from the growth curve models. Models for ASSIST scores included baseline scores as a covariate in the estimation of latent growth factors to adjust for baseline differences between delivery condition. χ 2 Analyses did not reveal differential attrition based on condition. When examining demographic characteristics, we did find differential attrition based on gender identity at 52 weeks (χ 2 4 = 10.03; P = .04). Specifically, transgender women were less likely to be retained in the 52-week follow-up compared with cisgender male and cisgender female participants. t Tests provided no evidence of differential attrition by age or baseline outcome values (viral load and alcohol severity and frequency). With respect to intervention dose receipt, 35.5% of the sample completed 0 sessions, 26.8% completed at least 2 sessions, and 25.1% completed all 4 sessions. There were no differences between groups in dose received (χ 2 4 = 3.39; P = .495).

Primary Outcomes Viral Load
In the viral load model (Figure 2), there were no between-condition differences in change from  Figure 3 shows percent undetectable at follow-up points, with 0% undetectable at baseline based on eligibility requirements.   attenuate to a mean difference of 2.5 drinks at 52 weeks. Given that responses ranged from 0 to 39 drinks, these differences are modest in magnitude relative to the variability in drinking.

Discussion
In this randomized clinical trial, Healthy Choices resulted in improvements in viral load and alcohol use (severity and frequency) during 12 months of follow-up when delivered by CHWs to a primarily racial/ethnic-minority sample of youth living with HIV in the clinic or at home or in the community.
Unlike in the original Healthy Choices trial, viral load improvements were sustained over time, possibly owing to the improved regimens in the current trial compared with the state of antiretroviral therapy in the original trial a decade ago. Contrary to our hypothesis, the clinic-delivered intervention outperformed the home-or community-based delivery with regard to viral suppression and alcohol severity; it may also be less costly and easier to implement. However, most youth did not receive a full dose of treatment, and intent-to-treat analysis suggested improvements even with small doses of intervention. Future studies could consider an adaptive design to test the effect of a single session, with additional sessions offered to nonresponders. 40 Furthermore, because home-and communitybased delivery did not result in increased session attendance, other modes of delivery, such as videoconferencing and mobile apps, should be considered to increase session attendance.
More than half of participants did not achieve viral suppression, regardless of condition.
However, youths who still have unsuppressed viral load in the current era of simplified regimens and effectiveness of antiretroviral therapy, even with less-than-perfect adherence, 41 may be faced with many more psychosocial barriers than youths who struggled with viral suppression a decade ago.
Thus, improving viral suppression in one-quarter to one-third of such a high-risk group has significant public health implications. While brief interventions may be more easily implemented in clinic settings, a more intensive intervention may be needed to sufficiently halt viral replication among youths at highest risk. 42,43 Some youths may benefit from motivational interviewing alone, and some may benefit more from motivational interviewing combined with cognitive-behavioral skillsbuilding interventions. 44

Limitations
Limitations include lower-than-optimal study retention, although this is not unusual for a real-world effectiveness trial. In addition, generalizability of the sample, recruited from academic medical settings, may be limited.

Conclusions
This randomized clinical trial found that the Healthy Choices intervention resulted in greater improvements in viral load and alcohol use when delivered in a clinic vs a home setting. Given these findings, in addition to considering designs to develop an adaptive intervention and technologybased delivery, future directions may include conducting a cost-effectiveness analysis of clinic vs community-based delivery.