Youth-Provider Relationships Move to the Center of HIV/STI Prevention Research
A new secondary analysis protocol will pool data from six HIV and STI prevention trials to examine a question that has long hovered at the edge of adolescent prevention science: why do some high-risk young people respond to intervention while others do not, even when programs are delivered with similar fidelity? The planned work focuses on the youth:provider relationship as a predictor of prevention response, with special attention to alcohol and cannabis use as part of a broader syndemic picture.
The study is grounded in a practical clinical problem. Standard HIV/STI prevention programs have produced only modest gains in youth, especially among young people of color and sexual and gender minority youth. Intervention format alone has not explained much of the variation in response. Relationship quality may be one of the more consequential missing variables, particularly in adolescents whose risk behavior is shaped by trust, disclosure, engagement, and the willingness to return for follow-up care.
A Harmonized Cohort Built From Six Behavioral Trials
Investigators will harmonize six sizeable behavioral prevention studies into a single longitudinal dataset containing 1,136 independent young people. Baseline age averaged 17 years, with participants ranging from 13 to 24 years old. The cohort was 43% female, 21% sexual and gender minority, 54% Hispanic, 9% African American, and 7% Native American or Alaska Native. Those numbers matter because the population most affected by HIV and STI disparities is often the least represented in prevention science, and the analysis is being designed around that reality rather than around a single homogeneous sample.
Participants in the parent trials received prevention programming and were followed at 1, 3, 6, and/or 12 months. That repeated follow-up structure gives the planned analysis room to examine whether relational effects persist, intensify, or fade over time. A one-time survey could not capture that evolution. Mixed-effects modeling can, in principle, account for repeated measurements within individuals while also preserving between-person variation, which is essential in youth prevention research where behavior often shifts over short intervals.
One sentence captures the central wager of the protocol: relationship quality may shape whether prevention content ever translates into behavior change.
Alcohol, Cannabis, and HIV/STI Risk Will Be Studied as a Syndemic
The protocol does not treat HIV/STI risk in isolation. It will also assess syndemic outcomes, specifically co-occurring HIV/STI risk behaviors, alcohol use, and cannabis use. That choice reflects the clinical reality that these behaviors do not always travel separately in adolescents and young adults. Alcohol and cannabis can reduce inhibition, complicate negotiation of condom use, impair judgment around sexual partners, and weaken adherence to safer-sex plans established during counseling sessions.
By placing these behaviors into one analytic frame, the study may clarify whether the same relational factors that support HIV/STI prevention also influence broader health behavior regulation. If the youth-provider relationship is a durable predictor across multiple domains, prevention programs may need to invest more heavily in relational engagement rather than relying primarily on content delivery. If the relationship predicts only certain outcomes, the field will need to identify which parts of risk behavior are most sensitive to trust, rapport, or provider responsiveness.
The protocol also plans to include participant factors, provider factors, and their interaction. That design is important because the quality of the relationship is unlikely to reside entirely in the youth or entirely in the clinician. Adolescents arrive with different histories, social pressures, and levels of readiness; providers bring their own style, training, and communication patterns. A model that ignores either side would miss the interplay that often determines whether brief prevention encounters are experienced as credible, respectful, and worth acting on.
Mixed-Effects Machine Learning Will Probe Common and Unique Pathways
To analyze the harmonized cohort, the team will build a longitudinal mixed-effects machine learning model. The intent is not simply to describe associations, but to identify common factors and unique pathways linking prevention response with syndemic behaviors. That analytic strategy is well suited to heterogeneous youth data, where a single linear explanation rarely fits everyone.
In practical terms, this kind of modeling can help separate shared signals from subgroup-specific patterns. A youth who enters care with elevated alcohol use, for example, may respond differently than a peer whose main concern is sexual risk in the absence of substance use. Likewise, a sexual and gender minority youth may experience a provider relationship through a different social and psychological lens than a cisgender heterosexual peer. The model is designed to let those differences emerge rather than flatten them into an average effect.
The protocol does not claim that relational factors will replace all other explanations. Instead, it treats the youth-provider relationship as a plausible mechanism that may help explain why otherwise similar interventions yield different outcomes across real-world adolescents. In that sense, the study sits at the intersection of behavioral science and implementation science, with direct relevance to clinic-based prevention programs.
Why This Protocol Matters for High-Risk Youth Care
The planned analysis is especially relevant for young people who face layered vulnerability: HIV and STI exposure, substance use, and structural inequities tied to race, ethnicity, sexual orientation, or gender identity. In such settings, the clinical encounter is not merely a venue for delivering information. It is often the main opportunity to establish trust, assess co-occurring risk, and offer a prevention plan that feels credible enough to follow.
If the analysis confirms that the youth-provider relationship predicts better outcomes, prevention programs may need to train providers in communication strategies that strengthen engagement early and consistently. If provider characteristics interact with participant characteristics, future interventions may need more tailored matching or adaptation. Either way, the work aims to move the field beyond a one-size-fits-all view of adolescent prevention.
Ethics, Dissemination, and the Limits of a Secondary Analysis
This secondary analysis is exempt from human subjects regulations under category 4(iii) as determined by the Institutional Review Board at UConn Health. Findings will be shared through scientific conferences, peer-reviewed journals, mental health and substance use providers, and community programs serving youth at high risk for HIV, STI, and substance use.
Because this is a protocol for secondary analysis, the study cannot test a new intervention or establish causality on its own. Its value lies in re-reading an unusually rich body of existing trial data through a more integrated lens. The harmonization process will also have to contend with differences across the original studies in intervention content, follow-up schedules, and measurement approaches. Those constraints are built into the design, and they will shape how confidently any common pathway can be generalized across settings.
Even so, the question is clinically concrete: when adolescents are asked to change sexual and substance-use behaviors, does the relationship with the provider help determine whether prevention takes hold? This analysis is designed to answer that question with more precision than previous single-trial comparisons have allowed.