Increasing Sustained Viral Suppression Among Youth Living with HIV: A Randomized Controlled Trial of Stepped Care Intervention

Abstract

This study aimed to decrease viral load (VL) to increase viral suppression (VS) among youth living with HIV (YLH) ages 12–24. This study was a stepped care randomized controlled trial. Sixty-eight YLH with established infection, without VS, and with at least two follow-ups (N = 68) were randomized to a control condition (n = 25) or a stepped care intervention (n = 43), and repeatedly assessed for up to 24 months. Both conditions received referrals for health services and a daily automated text-messaging and monitoring intervention (AMMI). YLH in stepped care who were unsuppressed at 4-month assessments stepped up to peer support and later to coaching. Random effects regressions examined VL trajectories over time as well as trajectories of secondary outcomes. There was significant evidence suggesting a different longitudinal trajectory of VLs for the two conditions. The control condition had improved VLs at about 12 months and then started to return to higher VLs. The stepped-care condition improved over the same time period and remained relatively stable. We estimated that the average VL was lower in the stepped care condition at 24 months, but we cannot claim a statistically significant difference between conditions. Both intervention groups appeared to have positive intervention impacts suggesting some benefits of the AMMI intervention. The improvement in VL at 24 months for stepped care compared to the control condition are suggestive of a viable intervention strategy that warrants further study.

Introduction

Despite national efforts to reduce HIV among young youth in the U.S, HIV remains a significant public health problem for this age group [1, 2]. In 2021, 6,987 youth ages 13–24 were diagnosed with HIV [2]. More than 80% of youth living with HIV (YLH) are gay, bisexual, gender diverse, or transgender, and Black and Latino youth are over-represented from gender and sexual minority subgroups (GSM) [1]. Fortunately, if YLH routinely and consistently receive antiretroviral therapy (ARV), over time viral suppression (VS) is achieved [3]. VS has a significant clinical benefit for both YLH and their HIV negative sexual partners. VS increases the quality and length of life of YLH and reduces rates of HIV transmission [3].

While benefits of VS are clear, there is little evidence that YLH have accessed or consistently adhered to ARV medications or achieved VS. The HIV Treatment Continuum [4] specifies challenges that face persons living with HIV who aim to achieve VS. At each step of the Treatment Cascade, YLH are less likely than older adults to complete the steps towards VS. Only 40% of YLH have actually received an HIV diagnosis [5], compared to 80% of older adults [1]. Typically, a very high percentage of YLH are linked to care and offered ARVs (e.g., 84% [6] 98%) [4]. However, many YLH drop out of care over time; only 34% were retained in care for a year in a recent study [7]. When not retained, adherence to ARVs, which is associated with VS, is difficult. Access to ARVs requires routine prescriptions and typically enrollment in health insurance. Yet, the earliest national study with YLH found only 41% reported consistently adhering, confirmed by a strong association between self-reported adherence and viral load (VL) [8]. Over the last decade, rates of VS among YLH have ranged from 6% [9] to 59% [4] and many points in between (27% [10]; 53%) [11] Rates of VS also vary by race ranging from 65.7 to 80.0% with lowest rates among Black/African Americans and highest among Hawaiian/Pacific Islanders [12]. Yet, rates are far too low to eliminate HIV– a current national goal. Given the low documented rates of VS, this study focused on evaluating a stepped-care approach to decreasing VL to increase VS among YLH [13].

While VS is the most important primary outcome for any intervention for YLH, there are a range of secondary targeted outcomes. Studies have linked VS to adherence to the HIV Treatment Cascade (engagement, treatment, and VS) for YLH, including a recent study conducted at experienced, well-resourced health care sites that found that 98% were prescribed ARVs, and 89% achieved VS [4]. Thus, receipt of medical care is an important pre-requisite to VS and a secondary outcome of this study.

Three other factors appear important to VS among YLH: mental health, substance use, and sexual risk behaviors. YLH commonly face pervasive mental health challenges [14, 15]. In addition to the biological impact of HIV, the psychosocial burden of living with a disease that may be stigmatizing can be substantial. YLH may experience difficulties related to body image, isolation related to stigma, and stress related to managing their medication regimens, as well as initiating relationships while dealing with disclosure of serostatus. These challenges reduce YLH’s ability to adhere to ARVs and impact quality of life [14, 16]. Conversely, improving mental health among persons living with HIV has been shown to promote better health across 29 studies reviewed with more than 12,000 people living with HIV [17].

Substance use challenges are prevalent among YLH, use which are associated with sexually transmitted infections, criminal justice system involvement, homelessness, and frequent unprotected sexual risks [16, 18–20]. Substance use is negatively correlated with ARV adherence and associated with adherence in qualitative interviews [15, 20]. One reason for substance use is that YLH may their wish to momentarily ignore their HIV status [16, 20], which may be a significant stressor in their lives.

The Undetectable = Untransmittable (U = U) public health campaign has encouraged youth to take ARV medications to improve their health and protect their sexual partners [21]. However, unprotected sexual behavior with seronegative or unknown serostatus peers is still the primary method of HIV transmission [1]. Sexual behaviors are typically initiated in adolescence and are influenced by a complex array of biopsychosocial, environmental, and other factors that increase vulnerability to risk behaviors during this developmental period [22]. If YLH achieve VS, engaging in unprotected sexual acts does not place their sexual partners at risk of acquiring HIV. Given that the YLH in this study were initially virally unsuppressed, unprotected sexual risk was also included as a secondary outcome.

Intervention Strategies to Increase Viral Suppression

A recent review [23] found that four strategies were effective in improving uptake and retention in HIV care and ultimately VS: transportation and accompanying patients to appointments, appointment alerts, psychosocial support, and patient navigation. In their analyses, multiple intervention strategies were typically used to enhance retention in care, a finding similar to that found by Risher and colleagues [24]. Given these data and the importance of basic quality health care [25], the control condition in this study included linking YLH to health care, providing daily messages to encourage healthy behaviors, and weekly monitoring of risk and adherence behaviors– an Automated Messaging and Monitoring Intervention (AMMI). Existing data on adolescents supports this approach. For example, a short message service or text message reminders to improve ARV adherence for YLH using daily reminders were found to be feasible and acceptable, and significantly improved self-reported adherence [26].

Stepped care is a system of delivering and monitoring health and mental health treatments so that the most effective but least resource intensive method of treatment is delivered first - only stepping up a person in treatment to more intensive care as required, depending on the level of need [27–29]. Generally, this method is expected to provide greater cost-effectiveness than either low-intensity only interventions, which minimize the costs/resources needed but possibly reduce efficacy, or high-intensity interventions that may be efficacious but costly. The approach is, in a sense, a push-back against a “one size fits all” model, which as Rapee et al. have noted is neither sufficiently flexible nor acceptable for entire populations of youth, whose treatment needs may be very diverse [29]. Stepped care models have two key features: using the least amount of specialist time expected to be effective for the health outcome of interest and allowing for self-correction through ongoing monitoring when desired health outcomes are not achieved with outcomes varying based on the health issue that is the focus [30]. One benefit of stepped care models for HIV adherence with youth is outcoming monitoring which allows for feedback both to the interventionist and the participant to inform intervention delivery [31]. Monitoring of viral load over time to inform intervention delivery is one such application of a stepped care model in the field of HIV adherence.

Two more intensive interventions identified by the CDC [23] and supported by other research with YLH are peer social support and patient navigation– a strategy defined as “coaching” in this study. Consistent with a stepped care approach, when YLH fail to achieve VS, a more intensive intervention was delivered. While AMMI is a basic, low-cost, structural intervention package, when YLH remained virally unsuppressed, they were stepped up to a more intensive intervention: first peer support and then coaching. This study, therefore, contrasted a structural intervention package with a stepped-care approach in terms of impact on the trajectory of VLs over time and other secondary outcomes.

Given the dearth of recent research on what types of interventions and delivery models (e.g., stepped care) have positive impacts on adherence among youth, it important to address these unanswered questions. Another study using a coaching intervention for YLH in the US, which also included dose monitoring and outreach, documented impacts on adherence but not on viral suppression [32]. Similarly, Taiwo and colleagues found the same outcomes when examining the effects of peer navigation and text messaging with YLH in Nigeria [33]. To date, only one recent study (using a stepped care design) with YLH has found positive impacts on both behavioral (i.e., adherence) and viral load outcomes using text messages (Step 1) and a five-session counseling intervention (Step 2) [34]. This study aims to fill the gap in the existing literature on both interventions and delivery models among a group with low rates of viral suppression and multiple factors that can impact adherence.

Methods

This clinical trial is part of the Adolescent Medicine Trials Network (ATN) study (ClinicalTrials.gov NCT03109431) aimed at increasing VS by reducing VL among YLH with established HIV infection. Participants were recruited from homeless shelters, clinics, and community-based organizations working with sexual and gender diverse youth in Los Angeles, California, and New Orleans, Louisiana, from May 2017 to May 2020 and through social media and dating applications. The study sites were identified because both cities had high rates of youth living with HIV within diverse communities and established infrastructures for conducting research with this population. All procedures were approved by the Institutional Review Board (IRB) of (blinded), which served as the single IRB for all sites.

Potential participants participated in a screening to identify youth living with HIV (see consort diagram). Our study was powered on a target sample size of 220. However, due to recruitment challenges that occurred during the COVID-19 pandemic, the target sample size was not obtained. Thus, the study is not powered to detect statistically significant differences between groups. Of the eligible youth, 103 were determined to be acutely infected and were referred to another study associated with this project, and 170 youth with established HIV infection enrolled in this study. The Fiebig staging system, developed in 2003 [35], a method of assessing the stage of HIV infection, was used to assess whether participants where acutely infected (stages 1–5) or had established (stage 6) HIV. Study interviewers who did the study assessments were blind to the study condition. Participants were assigned to their interventions after the baseline assessments. The care providers provided HIV standard of care treatment to all study participants. The treatment for HIV was not part of the study protocol.

Of these youth, 69 were initially virally unsuppressed, and 101 were suppressed. Of the 101 initially suppressed youth, 15 became unsuppressed during the study and were randomized in addition to three youth from two other associated study protocols. Thus, of the 85 randomized youth, the final analytic sample (n = 68) excludes those lost to follow-up who received no assessments after the baseline (n = 17). Older youth (ages 15–24) provided verbal consent for the screening, and younger youth (ages 12–14) gave written informed consent for the screening. A waiver for parental consent was obtained for those who were eligible and written informed consent to participate was obtained. The waiver for parental consent was obtained due to concerns that asking for parental consent could create additional risks for participants and would require disclosure of their HIV status, sexual identity, or other sensitive information.

To be eligible, youth had to test seropositive for HIV. All youth were tested for HIV antigen and antibody screening with a rapid-HIV test using the Clinical Laboratory Improvement Amendments waived and FDA approved Alere (Waltham, MA) Determine HIV-1/2Ag/Ab Combo finger stick and the Cepheid (Sunnyvale, CA) Xpert HIV-1 Qual Assay. The test is a point-of-care lateral flow strip that detects both HIV-1/2 antibodies and the HIV-1 p24 antigen using 50 µl of fingerstick whole blood. The window period is 12–26 days, and results are ready in 20–40 min. A second RNA HIV test was also administered and batch tested to identify YLH who were acutely infected who were then referred to one of the other associated studies.

Study Interventions

As demonstrated in Fig. 1, YLH who met the study criteria (N = 68) were randomized to one of two study conditions: (1) enhanced standard care (control; n = 25) or (2) a stepped care intervention (intervention; n = 43). Youth in the control condition received referrals for health services and AMMI focused on increasing adherence and other related topics. Youth also received a weekly self-monitoring survey that included questions about a variety of health-related topics and behaviors. Text messages focused areas of youth’s lives that can impact adherence including health care, wellness, medication reminders/adherence, substance use, and sexual health. The focus was on empowering youth and providing health-related information.

Fig. 1.

Study flow consort diagram. It presents the flow of the study recruitment, enrollment, intervention, and follow-up visits

Viral load was used to determine the “step up” process. Participants in stepped care also received the same AMMI intervention but “stepped up” to a higher level of intervention at four-month intervals if they had not achieved vs. (< 200 copies/mL), which was changed during the trial to be to > 20 (i.e., detectable). Participants’ viral status was monitored via blood draws to assess viral load or reports of these values from treatment records where participants were receiving care. Twenty-six youth stepped up to the second level, an online peer support group. The peer support was conducted on a private discussion board hosted by the study team. Participants had anonymous user profiles and engaged in discussion of HIV-related and other youth-oriented topics. These discussions were monitored daily by study team members with no private messaging allowed. The third and highest level was strengths-based coaching delivered by trained coaches who had similar backgrounds and lived experiences. The coaching model was based on the Strengths Based Case Management intervention developed by Rapp [36] and the later adaptation for use in HIV prevention and care [37, 38]. The coaching intervention aimed to address multiple known barriers for viral suppression among YLH, including structural, social, and mental health [39], by providing support, skills, and referrals for needed services. Coaching consisted of a strengths’ assessment conducted by the coach that focused on the youth’s current status regarding a variety of life domains, including daily living, health, healthcare, social support, mental health, and risk behaviors (substance use and sexual behavior). Next, personalized goals were identified by the youth with follow-up coaching sessions focusing on cognitive and behavioral skills training targeting goal attainment. Goals were revised as needed based on goal attainment. Thirteen YLH stepped up twice to coaching (at least two assessments virally unsuppressed); two YLH were evaluated by the clinical team as needing immediate step-up to coaching due to their clinical needs.

Study Outcomes

The primary outcome was VL assessed over time (up to 24 months). Secondary outcomes were as follows:

Retention in Medical Care.

This was defined as two visits from 0 to 12 months and two visits from 12 to 24 months.

Depressive Symptoms.

Depressive symptoms were assessed using the 9-item Patient Health Questionnaire (PHQ-9), a checklist of symptoms of a major depressive episode [40].

Anxiety:

The Generalized Anxiety Disorder 7 (GAD-7) [41] was used to assess anxiety.

Substance Use.

Substance use was coded as “yes” if the participant self-reported use of a specific substance or had a positive RDT for any substance other than alcohol or marijuana.

Unprotected Sex with HIV Negative Partner.

This variable was assessed using the following question: In the past 4 months, did you have vaginal or anal sex WITHOUT a condom with a partner you knew was HIV negative or with a partner whose HIV status you didn’t know?

Data Management and Quality Control

All study data were managed by the study’s data analytic team who managed the data for all three associated studies that were part of this project. This team used a secure electronic data capture system (EDC) which allowed for the use of mobile technology to collect data from participants in near real-time. The EDC also allowed the team to monitor all of the study functions and generate reports that were used to carefully monitor and support data quality (see Comulada et al. 2018 for a more detailed description) [42]. Intensive strategies were used to engage participants, including collecting multiple types of contact information, building connections with participants, having interviewers go into the field at agencies where youth were recruited (when possible to do so), and using ongoing text messages surveys to maintain contact with participants between study assessments. Given that the study took place during the COVID-19 pandemic, unanticipated challenges occurred during times when university research activities where shut down or limited in scope.

Analytic Methods

We used mixed effects models to evaluate the impact of the interventions on our primary outcome of VL over time and secondary outcomes. Following standard practice, we applied the logarithmic transformation to VL using a base of 10 (log₁₀) [43]. Models included main effects for time, the intervention, and a time by intervention interaction. We used a random, participant-specific intercept to account for individual correlation over time. To measure time for VL, we used the months between the date the VL was assessed and when the participant was randomized to the intervention or control condition. Most participants were assessed on their randomization date, but a few were assessed a couple of days before and not on the exact randomization date. For these participants, we used this previous measurement as their baseline (months = 0) VL. We considered models with linear, quadratic, and cubic time trends (Table 1). We selected the best fitting model as the one with the lowest BIC fit statistic and selected the quadratic trend for VL. This means we had main effects for the intervention, months since randomization, and months since randomization squared. We also had interaction effects between the intervention and both time variables. To fit a mixed-effects Tobit model [44] we used the R censReg package [45] to log10 VL as a primary outcome, left-censored at log10 (20). We used the lme4 package [46] to fit linear mixed effect models for the GAD-7, the PHQ-9, unprotected sex with HIV negative or unknown partners, and substance use as secondary outcomes. We selected models with linear time trends based on Bayesian Information Criterion (BIC) fit statistics which involves selecting the model with the lowest score as the decision criteria. We evaluated intervention effects on the outcomes by estimating differences in mean outcome levels between conditions at 12 and 24 months and by visually examining graphical displays of the estimated outcome trajectories over time. Because medical adherence was only measured at 12 and 24 months, we examined whether rates of medical adherence differed between these two time points by conducting two Fisher’s Exact tests to compare medical adherence proportions.

Table 1.

Description of participants

	Control (N = 25)	Intervention (N = 43)	Overall (N = 68)
Months since HIV diagnosis
Mean (SD)	65.5 (80.6)	44.5 (74.6)	52.3 (77.0)
Median [min, max]	27.6 [0, 277]	16.9 [0, 289]	24.0 [0, 289]
Missing	0 (0%)	1 (2.3%)	1 (1.5%)
Site
LA	7 (28.0%)	16 (37.2%)	23 (338%)
NOLA	18 (72.0%)	27 (62.8%)	45 (66.2%)
Age
Mean (SD)	21.7 (2.25)	21.5 (2.31)	21.5 (2.28)
Median [min, max]	22.0 [17.0, 24.0]	22.0 [15.0, 24.0]	22.0 [15.0, 24.0]
Sex at birth
Female	5 (20.0%)	4 (9.3%)	9 (13.2%)
Male	20 (80.0%)	39 (90.7%)	59 (86.8%)
Gender identity
Cisgender	22 (88.0%)	38 (88.4%)	60 (88.2%)
Transgender	3 (12.0%)	4 (9.3%)	7 (10.3%)
Gender nonconforming	0 (0%)	1 (23%)	1 (1.5%)
Sexual identity
Bisexual	5 (20.0%)	5 (11.6%)	10 (14.7%)
Gay/SGL	16 (64.0%)	27 (62.8%)	43 (63.2%)
Heterosexual	4 (16.0%)	8 (18.6%)	12 (17.6%)
Pansexual	0 (0%)	3 (7.0%)	3 (4.4%)
Race/ethnicity
Black: Hispanic	4 (16.0%)	3 (7.0%)	7 (10.3%)
Black: Non-Hispanic	17 (68.0%)	25 (58.1%)	42 (61.8%)
Other Hispanic	1 (4.0%)	9 (20.9%)	10 (14.7%)
Other Non-Hispanic	1 (4.0%)	1 (23%)	2 (29%)
White: Non-Hispanic	2 (8.9%)	5 (11.6%)	7 (10.3%)
Highest level of education
< high school	5 (20.0%)	11 (25.6%)	16 (23.6%)
High school graduate	10 (40.0%)	14 (32.6%)	24 (35.3%)
Some college	10 (40.0%)	18 (41.9%)	18 (41.9%)
Employment
Employed	12 (48.0%)	21 (48.8%)	33 (48.6%)
Student	5 (20.0%)	6 (14.0%)	11 (16.2%)
Unemployed	8 (32.0%)	16 (37.2%)	24 (35.3%)
Income > S1063.33/month
No	22 (88.0%)	31 (72.1%)	53 (77.9%)
Yes	3 (12.0%)	12 (27.9%)	15 (22.1%)
Lifetime homelessness
No	15 (60.0%)	25 (58.1%)	40 (58.8%)
Yes	10 (40.0%)	18 (41.9%)	28 (41.2%)
Lifetime incarceration
No	14 (56.0%)	34 (79.1%)	48 (70.6%)
Yes	10 (40.0%)	9 (20.9%)	19 (27.9%)
Missing	1 (4.0%)	0 (0%)	1 (1.5%)
Lifetime interpersonal violence
No	15 (60.0%)	26 (60.6%)	41 (60.3%)
Yes	10 (40.0%)	14 (32.6%)	24 (35.3%)
Missing	0 (0%)	3 (7.0%)	3 (4.4%)
Lifetime substance use treatment
No	21 (84.0%)	35 (81.4%)	56 (82.4%)
Yes	4 (16.0%)	8 (18.6%)	12 (17.6%)
Insurance
Insured	17 (68.0%)	29 (67.4%)	46 (67.6%)
Uninsured	4 (16.0%)	8 (18.6%)	12 (17.6%)
Unsure	4 (16.0%)	6 (14.0%)	10 (14.7%)
Healthcare provider (current)
No	3 (12.0%)	6 (14.0%)	9 (13.2%)
Yes	22 (88.0%)	37 (86.0%)	59 (86.8%)
Take ARVs (recent)
No	8 (32.0%)	8 (18.6%)	16 (23.6%)
Yes	16 (64.0%)	34 (79.1%)	50 (73.6%)
Missing	1 (4.0%)	1 (2.3%)	2 (2.9%)
ARV adherence (1–6)
Mean (SD)	4.44 (1.46)	4.68 (1.27)	4.60 (1.32)
Median [min, max]	5.00 [1.00, 6.00]	5.00 [2.00, 6.00]	5.00 [1.00, 6.00]
Missing	9 (36.0%)	9 (20.9%)	18 (26.6%)
Secondary outcomes
GAD-7 score
Mean (SD)	3.24 (1.33)	6.00 (135)	4.99 (10 8)
Median [min, max]	3.00 [1.00, 5.00]	4.00 [0, 66.0]	3.00 [0, 66.0]
PHQ-9 score
Mean (SD)	6.08 (4.57)	5.83 (4.76)	5.92 (4.65)
Median [min, max]	5.00 [0, 15.0]	6.00 [0, 20.0]	5.00 [0, 20.0]
Missing	0 (0%)	3 (7.0%)	3 (4.4%)
Substance use
No	12 (48.0%)	24 (55.8%)	36 (52.9%)
Yes	13 (52.0%)	19 (44.2%)	32 (47.1%)
Unprotected sex with HIV negative partner
No	12 (48.0%)	18 (41.9%)	30 (44.1%)
Yes	7 (28.0%)	12 (27.9%)	19 (27.9%)
Missing	6 (24.0%)	13 (30.2%)	19 (27.9%)

Results

A description of the sample is presented in Table 2. Participants were an average age of 21 years old; 87% were male, and 11% were transgender or gender non-conforming. Most (82.4%) identified as a sexual identity other than heterosexual. The majority were Black or Hispanic (86.8%). About 42% had some college education. About half were employed, and most lived below the poverty line. A substantial minority had been previously homeless (28%), incarcerated (28%), in substance abuse treatment (17.6%), or had experienced interpersonal violence (35%). Most (82%) had health care insurance. Similarly, the vast majority had a health care provider (87%). Most (74%) had taken ARVs recently and reported relatively high daily adherence. In Los Angeles, 20 youth were recruited through clinics and three youth were recruited from via social media. In New Orleans, 33 youth were recruited via clinics, 10 via social media or a dating app, and two were referred to the study by someone they knew.

Table 2.

Analysis of variance table for the Tobit mixed-effects model

Variable	Estimate (SE)	95% confidence interval
Intercept	3.576 (0.322)	(2.945, 4.208)***
Intervention	− 0.559 (0.427)	(− 1.296, 0.277)
Months since randomization	− 0.228 (0.052)	(− 0.330, − 0.125)***
Monthŝ2	0.009 (0.002)	(0.005, 0.013)***
Intervention × months	0.131 (0.067)	(− 0.001, 0.263)*
Intervention × monthŝ2	− 0.006 (0.003)	(− 0.011, 0.000)**

^*p < 0.10

^**p < 0.05

^***p < 0.01

For the primary outcome (VL), we found modest evidence that the intervention impacted VL trajectory over time. Figure 2 shows the trajectory of all participants as well as the estimated mean trajectory in both conditions. In the control condition, average baseline VLs were higher than intervention VLs, then decreased for the first 12 months. Afterwards, control condition VLs increased again, and at 24 months the average VL was close to its baseline level. In the intervention condition, the trajectory was more stable. There was a slower decrease in VL from baseline to about 16 months, and then a slight increase from 16 to 24 months. We estimated the 24-month mean log₁₀ VL was 3.164 (SE = 0.37) in the control and 2.562 (SE = 0.37) in the intervention condition; however, we cannot claim statistical significance for this difference (p = 0.258). In general, there is some certainty that the intervention impacted the trajectory of VLs. Both linear and quadratic interaction effects had relatively small p-values: 0.053 for the linear interaction (est. = 0.131, SE = 0.067) and 0.048 for the quadratic interaction (est. = − 0.006, SE = 0.003; Table 3).

Fig. 2.

Plot of the viral load trajectories. The smaller, less-bold lines are the individual trajectories for all study participants. The larger, bold lines are the estimated mean trajectories from the Tobit model

Table 3.

Secondary outcomes

Outcome	Month	Control mean (SD) N (%)	Intervention mean (SD) N (%)	Total mean (SD) N (%)	Overall p-value
GAD-7	12	5.1 (6.2)	5.6 (6.1)	5.4 (6.0)	0.463
	24	3.0 (3.6)	5.7 (6.6)	4.6 (5.7)
PHQ-9	12	5.1 (4.7)	4.8 (5.3)	4.9 (5.1)	0.315
	24	3.2 (4.1)	4.9 (6.5)	4.2 (5.7)
≥ 2 medical visits	12	9/14 (64.3%)	20/26 (76.9%)	29/40 (72.5%)	0.469
	24	11/15 (73.3%)	18/21 (85.7%)	29/36 (80.6%)	0.418
Unprotected sex HIV negative or unknown partner	12	5/14 (35.7%)	9/26 (34.6%)	14/40 (35.0%)	0.845
	24	4/15 (26.7%)	9/21 (42.9%)	13/36 (36.1%)
Substance use (non-alcohol/marijuana)	12	6/15 (40.0%)	10/26 (38.5%)	16/41 (39.0%)	0.462
	24	4/15 (26.7%)	9/21 (42.9%)	13/306 (36.1%)

The p-values for all outcomes other than medical visits are from the time by intervention interaction in the mixed effects model. The two p-values for medical visits are from fisher’s exact test using only the 12 or 24 month visit data

We found no significant intervention effects for any of the secondary outcomes (Table 3). Additionally, we did not consistently estimate that the secondary outcome was more favorable in the intervention condition. In general, there is no evidence to suggest an intervention effect on any of the secondary outcomes.

Discussion

In terms of the primary goal of this study—to determine if the more intensive stepped care intervention appeared to impact VL trajectories VLs over time as compared to the enhanced standard care control condition—we found some evidence that the intervention had positive impacts. There was significant evidence suggesting the trajectory of VLs over time was different and more favorable in the intervention condition. However, there are limitations associated with the overall sample size and the number of youth assigned to each condition. Consistent with other HIV stepped care studies conducted during the COVID-19 pandemic [47], we did not reach our target sample size which impacted our statistical power. Thus, no definitive conclusions can be drawn about intervention effectiveness. As noted above, our study was powered based on a target sample size of 220 (n = 110 in each condition), and we had anticipated 80% power to detect time-averaged percent differences in viral suppression. Nonetheless, both groups appeared to have initial positive changes in VL over the first year that were maintained for some time before slightly increasing towards the end of the middle of the study period. Additionally, at the end of the study, both groups had overall improved VLs as compared to baseline suggesting benefits of both intervention conditions.

Efforts aimed at VS among youth is a complex undertaking that was even more challenging in the context of the COVID-19 pandemic. At times during the study it was difficult for youth to access care, medication, and labs to get VLs checked. As noted by others, the challenges of the pandemic also impacted our ability to collect some study outcome data and the ability of participants to accessed needed HIV care [47], which likely had an indirect impact on viral suppression in this study. A recent review of psychosocial interventions for YLH suggests that some, but not all, interventions studied in recent years have had positive impacts on VS [48]. However, the review suggests that the theorized mechanisms to increase adherence, leading to VS, should address individual level factors, tailor delivery strategies to specific needs, and provide supportive resources. Interventions targeting YLH who are not virally suppressed must also consider the structural, social, and mental health challenges faced by these youth, particularly during the transition from adolescence to adulthood [39].

In this study, one possible explanation for the lack of differences between groups was that both conditions received some form of intervention– referrals for services and AMMI. Similar to other stepped care research with PLH [31], our results suggest benefits of treatment as usual that likely impacted both study conditions. Furthermore, this intervention is likely more than youth will receive in “real world” settings as text messaging adherence interventions are not universally implemented as part of HIV care. The level of intervention across groups is a limitation of the study but was necessary given the need to ensure that YLH received the support needed to reduce their VL. Offering text messages to youth may provide sufficient support to increase adherence, but these interventions may need to be more interactive and personalized than automated and standardized. Data from a single-arm prospective study [49] suggest that a two-way text message intervention can impact both VS and mental health among youth and young adults living with HIV with a dose response based on text message engagement.

Many of the YLH had VLs that improved during the first 12 months of the study (Fig. 2). In the intervention condition, however, about 60% of YLH required one or more step ups as they did not achieve VS with AMMI alone. This finding is similar to results from a stepped care study of PLH with alcohol use disorder where 52% of participants stepped up during the trial [50]. Our findings also suggest that while there may be benefits to text messaging interventions, not all youth will get sufficient support to achieve VS. Furthermore, it is possible that a meaningful proportion of youth may require only a low-intensity intervention to achieve optimal benefit, allowing resources and interventions to be used for those needing more intensive efforts. Although a small trial, our outcomes suggest that the interventions to which youth stepped up were insufficient to meet their needs given the complex array of factors that can impact adherence.

Interestingly, there was no evidence that the intervention affected any of the secondary outcomes. There are a number of reasons this may have occurred. Two of the five secondary outcomes (sexual risk, substance use) were assessed utilizing a dichotomous scale, and this may be a limitation of the study. For retention in health care, the vast majority of YLH in both conditions had a healthcare provider across the entire study (approximately 87%) so retention in care may not have been a challenge for most of the sample. While other recent studies [34, 39], have examined similar characteristics, including substance use and mental health, no other known recent studies have included these variables as published secondary study outcomes.

The most surprising finding is the lack of apparent intervention effects on mental health outcomes. In fact, the intervention condition had higher (but not statistically different) mean scores on the PHQ-9 and GAD-7 at both study time points. While we used actual scores, examination of the percentages of youth who met the clinical cut score for both measures at each time point provides some further insight into these findings. For depression, the percentage of participants with a baseline PHQ-9 score of ≥ 10 was 24% for the control and 23% for the intervention condition. The percentages varied over time with the lowest being 9% and 14% by condition respectively across 24 months. For anxiety, GAD-7 rates were slightly higher at baseline with 20% of control and 16% of intervention participants reporting scores of ≥ 10 indicative of moderate to severe anxiety. Over time, the lowest percentage of participants reporting clinically significant anxiety at any follow-up point was 0% for the control (at 24-months), and 12% for the intervention condition (at four-months). Thus, given that only about one-fifth to one-fourth of participants appeared to have clinically significant mental health symptoms at time of study entry, it is less unusual that positive impacts on these symptoms were not found over time. By contrast, in Amico and colleagues’ recent study of YLH, 53% of participants appeared to have depressive symptoms at study entry [39].

Decreasing VL to increase VS among YLH is very important from individual and public health perspectives. YLH who achieve VS will likely have improved quality of life and will not transmit the virus to their sexual partners. However, finding strategies to effectively and consistently improve VS have proved challenging, and there is currently no gold standard for these intervention efforts. Our results, with a relatively small sample size but with longitudinal data, suggest that text messaging interventions that remind youth to take their medications and support other positive behaviors may be sufficient for some youth to achieve VS. Other YLH may need more intensive or personalized interventions to achieve VS. The improvement in VS at 24 months in stepped care compared to the control condition are suggestive of a viable intervention strategy that warrants further study.