The following activities occurred in the context of a hybrid effectiveness-implementation study, which first sought to adapt and modernize a previous sexual health intervention called Guy2Guy, originally developed from 2012 to 2013 [26]. Guy2Guy was intended to increase condom use and other HIV prevention and testing behaviors among gay and bisexual teenage boys. The intervention consisted of 8 to 10 daily texts sent to participants over 6 weeks and included various interactive features, including a chatbot and one-on-one anonymous texting buddies. This intervention was associated with a 3-fold increase in HIV testing in the active treatment arm vs an information-only control arm [27,28].

In the new study, which began a decade after Guy2Guy’s original development, we sought to expand the intervention to be inclusive of LGBTQ+ adolescents who could benefit from HIV prevention and testing behaviors, adapt Guy2Guy to include updated HIV and sexually transmitted infection prevention information, modernize the texting delivery (eg, multimedia messaging service rather than SMS), and update other technological components such as the chatbot and texting buddies. This adaptation and modernization process was informed by feedback from a diverse sample of LGBTQ+ youth (aged 13-18 years; n=21) hosted on Discord over a 9-month period.

We recruited participants through paid social media advertisements and an existing institute-wide participant registry. Social media advertisements were developed following the CAN-DO-IT (conceptualize scope, acquire expertise, navigate platform and strategy, develop ad content, optimize recruitment workflow, implement campaign, track, and respond) model [28] and were posted on Meta platforms (Instagram and Facebook). Eligibility criteria were as follows: (1) sexual and/or gender minority identity; (2) age 13 to 18 years; (3) ability to read English at an eighth-grade level; (4) for minors, ability to provide informed assent and pass a capacity to consent assessment; and (5) residence in the United States or US territories.

Interested individuals were directed to complete the eligibility screening, consent, and capacity to consent surveys in Research Electronic Data Capture (REDCap; Vanderbilt University). We purposely oversampled transgender, gender-diverse, and minoritized racial and ethnic participants to ensure diverse feedback. A total of 24 participants were enrolled in the study. Only 12.5% (3/24) of participants formally requested to withdraw before the Discord server officially started its research activities, most commonly citing competing time commitments. Over the active course of the 9-month period of Discord activities (March 2023 to November 2023), no additional participants withdrew. The remaining 87.5% (n=21) of active participants comprised the analytic sample for this paper and participated in the survey reported here, as well as the primary study activities described later.

This study was approved by the Northwestern University Institutional Review Board (IRB; STU00217358 and STU00222652). Informed consent was obtained from all participants prior to participation in the advisory council. A waiver of parental permission was approved by the IRB on the basis of minimal risk and the potential for increased harm if parental consent were required. Previous IRB-approved studies conducted by the investigative team with sexual and gender minority (SGM) adolescents younger than 18 years have demonstrated that parental permission does not consistently function as a protective mechanism for this population [29]. Many SGM adolescents are not out to their parents and do not disclose sexual orientation, gender identity, or sexual behavior due to fear of rejection, victimization, or family conflict [29]. Therefore, requiring parental consent could increase risk by inadvertently “outing” participants or exposing them to harm and would likely result in systematic exclusion of youth most in need of sexual health research. Consistent with previous studies of adolescent SGM sexual health, the IRB determined that requiring parental permission would introduce significant sampling bias and may compromise participant welfare, while the study procedures themselves posed no greater than minimal risk [30-33]. Assent procedures, confidentiality protections, and access to supportive resources were implemented to ensure ethical participation. Through the consent process and subsequent orientation, participants were given information about the activities associated with the advisory council, details concerning privacy and confidentiality, possible risks and benefits, and reminders that participation was voluntary and could be stopped at any time.

We paid participants US $30 per month for active participation, defined as answering 75% or more of the questions asked in each month, delivered via virtual gift cards. Participants also had the opportunity to get paid an additional US $60 every 3 months for completing optional extra activities, such as creating a social media post or providing more in-depth feedback about the intervention. All data were stored on private servers accessible only to study staff, in compliance with the Health Insurance Portability and Accountability Act. Participants were assigned unique identifiers for analysis and reporting requirements. All data were deidentified, and all personally identifying information was removed before data were shared beyond the research team to ensure privacy and confidentiality.

High-level feedback about the acceptability of Discord for these activities was assessed by an online survey completed by all participants (21/21, 100%) hosted through Qualtrics (Qualtrics Inc) 6 months after the start of Discord activities. Questions were developed for the purpose of understanding perceived acceptability in the context of the features of Discord (eg, “I felt comfortable sharing my thoughts and ideas on Discord”) and were aligned with the core principles of HCD, such as reporting iterative changes based on participant feedback (eg, “I felt like the research team echoed or reported our SHER [Sharing Health Educational Resources] feedback back to us after making changes”).

Additionally, we allowed participants to provide open-ended answers to the following question: “Would you use Discord again for future research projects like this? Please tell us why.” Results of this survey and the open-ended answers are presented together in thematic groupings and include the themes of psychological safety and ease of communication. We used a rapid, inductive qualitative approach to analyze open-ended Discord feedback [34,35]. Two team members independently reviewed and grouped responses, refining categories through discussion in weekly meetings. Because the purpose was exploratory and aligned with HCD processes, we resolved differences by consensus rather than by computing interrater reliability. Coding and data management were conducted in Microsoft Excel.

The study’s Discord server consisted of multiple channels organized into 3 categories: “Information,” “Activities,” and “Creative and Fun.” “Information” channels included server guidelines, resources, a channel to directly contact staff, and a copy of the consent form. “Activities” channels consisted of weekly questions that engaged participants in the HCD process. Feedback from the HCD activities was collected through poll responses and open-ended threads on Discord. When responding to poll questions, participants were prompted to react to staff messages by selecting the emoji corresponding to their chosen vote, and open-ended questions used the “thread” feature to organize responses in one location within the channel.

All weekly questions, extra activities, and retention activities on Discord were developed during weekly collaborative staff meetings with the full study team. Staff identified “themes” for weekly questions and key areas for participant feedback based on the needs of the project (eg, new intervention content, adaptation of existing intervention content, and software design and development). The questions and relevant materials were organized in an Excel document. Discord moderation team members then reviewed the questions proposed by the research team to ensure consistency with the language and tone of the Discord server. After internal approval, questions were posted to Discord by research staff. Participant feedback was copied and pasted into a separate Excel document to facilitate easier visualization and data analysis. Key findings from participants were presented during weekly research team meetings and content development meetings, where the research team decided which feedback to implement. This process included qualitatively interpreting Discord posts through these team discussions to contextualize participant comments and clarify emerging insights. Any clarifying questions were also proposed back to the participants.

Full sociodemographic information of the participants is presented in Table 2. Participants had a mean age of 17.4 (SD 0.98) years. The sample was diverse across multiple identities, including sexual orientation, gender, and location, with participants representing 16 different states. Results of the survey informed our best practices when engaging in HCD on Discord for digital intervention design and adaptations. For the full results of the survey, refer to Multimedia Appendix 2.

One benefit of using Discord for HCD activities was the ease of participation and the ability to ensure safety and anonymity. Participants consistently described feeling at ease participating in design activities within Discord compared to more formal research settings, which likely contributed to high engagement. One reason for this was that participants largely reported no concerns about their privacy on Discord; about 90% (19/21) responded that they had no privacy concerns on the platform. The flexibility to engage asynchronously, use chosen names or pronouns, and opt into different modes of participation contributed to high engagement across our HCD activities. In total, 80% (n=17) of participants responded that they were comfortable visiting Discord to engage in the HCD activities and posting their opinions in threads and activities online. Additionally, participants reported that their familiarity with Discord in their personal lives made them likely to use the platform again for a similar project. Two participants mentioned this factor:

A critical component of the HCD process is maintaining effective, ongoing communication between design teams and participants across the EDIPT phases. We found that Discord facilitated highly efficient communication among participants, moderators, and the research team. Most participants (18/21, 85%) reported feeling that the research team listened to their suggestions for improving the intervention, and 90% (n=19) noted that the team communicated how their feedback directly influenced design changes. This feedback loop was essential; participants overwhelmingly indicated that seeing their ideas reflected in the evolving intervention strengthened their sense of contribution and improved their overall experience with the HCD process. Participants reiterated this qualitatively:

Improved communication between the research team and participants may have also supported conflict resolution. Although this was not assessed in the survey, a participant reported a conflict directly to the research team, suggesting a level of trust in the moderation process.

Lightweight tools such as emoji reactions and informal polls proved to be more engaging and generative than structured surveys or longer text-based threads. Participants (19/21, 90%) agreed that it was easy to share feedback on Discord using our methods, which included using preselected emojis to respond to questions (Figure 1). Additionally, a majority of participants (n=16, 76%) mentioned that they felt they were able to share all their suggestions for the prototypes and interventions, which suggests that our lightweight approach to data collection was acceptable and capable of capturing their feedback. For instance, one participant mentioned the following:

Another participant commented about the tone and the use of emojis in the activities:

To establish best practices for using Discord for HCD design purposes, the moderation team described effective, casual study procedures that improved engagement and psychological safety. Refer to Textbox 1 for a brief report on the best practices reported by the moderation team.

Our team developed a novel research participation hub on Discord to engage LGBTQ+ youth in HCD for the adaptation of a digital health intervention. Our methods encouraged high engagement and, most importantly, were considered acceptable by participants. Hosting this process on Discord allowed direct and instantaneous access to LGBTQ+ youth for supporting the successful adaptation of a digital health intervention component. Our findings support the acceptability of Discord as a tool for engaging in HCD activities in a community-based participatory research in a digital setting. We demonstrated how this digital environment can facilitate the co-design of digital health interventions.

The moderation team also assessed satisfaction, Discord acceptability, and Discord server recommendations with participants quarterly. For instance, when we assessed participant suggestions for improving data collection on Discord, participants mentioned that it was difficult to respond to specific prompts (Multimedia Appendix 3). We changed our Discord structure so that all posts contained all possible response emojis. This way, participants could click on the emoji reaction to “vote” without manually searching for the specific emoji. By asking for feedback surrounding the design of data collection itself, participants can cocreate and democratize the server environment to meet their own needs. This approach also ensures that researchers can improve participant retention, satisfaction, and participation.

There are many benefits to using Discord in the HCD process for digital interventions. Discord allowed for substantial anonymity (eg, not requiring real names or profile pictures), does not require transit, and does not require parental or guardian approval for participation among users older than 13 years. In addition, the platform is familiar and easy to navigate for many young people in the United States, particularly LGBTQ+ youth, who are often highly connected in digital communities. Our participants reported that the HCD activities conducted on Discord were widely acceptable. Additionally, Discord allowed affordances that followed the principles of community-based participatory research, specifically shared decision-making, mutual ownership of the research process and its outcomes, colearning, and recognition of the community as a unit of identity [37,39].

We also found that the digital affordances and structure of Discord were useful for this research. We were able to sort different conversations into specific channels for specific topics. For example, we had separate channels for more HCD-focused discussions, and others focused on building a sense of community. This fostered a more holistic and inclusive research environment for participants, where they could be present on Discord not only as study participants helping to design a health intervention but as individuals with diverse interests who were invested in being part of a community [39-41]. The use of Discord for engaging in HCD activities is effective at lowering barriers between academic professionals and LGBTQ+ youth [42] and offers avenues for direct and clear communication between the groups [21]. By centering the design of interactions around meeting youth on platforms they already use, such as Discord, researchers are able to develop bonds with and learn more efficiently from community youth, leading to improved intervention design. Liem et al [43] also demonstrated the process of how the feedback from this Discord youth advisory board informed the adaptation of a separate digital health intervention.

Social media platforms may fall in and out of favor over time, so adapting digital HCD methods as existing platforms evolve and new platforms arise will be necessary. However, the best practices we shared are generalizable to many current social media platforms that researchers may use for the HCD of digital interventions with youth, LGBTQ+ youth, and others. While an HCD approach to digital interventions is valuable, there are clear limitations. Our own study relied on a small number of end users, which, while common in HCD work, does hamper our ability to generalize to other end users [44]. Sample adequacy in HCD is instead evaluated based on the diversity of perspectives, sufficiency of idea generation, and whether additional participants yield novel design insights. Across sessions, we observed recurring themes and did not observe substantively new insights emerging, indicating that the sample size was appropriate for the aims of this study.

Our purposeful sampling approach also helped to ensure a diversity of perspectives by intentional recruitment efforts focused on racial and ethnic minority persons who are LGBTQ+. Consequently, nearly half of the sample identified as belonging to racial or ethnic groups other than White. In addition to sample size, participation required access to Discord and a baseline level of digital literacy, which may have deterred some youth and introduced platform-specific selection bias. These factors limit the generalizability of our accessibility model. However, participants in this study demonstrated high familiarity with Discord, consistent with previous research showing higher Discord use among LGBTQ+ youth than among heterosexual peers and broader patterns of youth social media engagement [20,45]. Future work should assess variation in digital literacy and examine whether these methods translate to other platforms.

In addition to sample size, participation required access to Discord and a basic level of digital literacy, which may have excluded or deterred some youth and introduced platform-specific bias. These barriers limit the generalizability of our “accessible” model. Understanding the extent of youth digital literacy and whether the reported methods are usable on other platforms should be addressed in future work.

Furthermore, as HCD is an iterative process, it can be infeasible for some intervention developers due to the cost and time required for maintaining an ongoing Discord server for the purpose of intervention design and adaptation. Our 9-month Discord engagement period was moderately resource intensive and may not be easily scalable. However, while certain routine tasks could be automated (eg, onboarding or reminder bots), maintaining participant trust required active human facilitation. Moderation from research staff would have to be responsive to a higher number of participants in any one server. Future research should identify which components can be automated and explore the most effective amount of moderation. Despite this potential limitation, this online methodology was cost-effective and potentially more feasible for youth participation than intensive in-person alternatives, such as iterative focus groups, which require travel expenses and logistics. While other studies have used Discord for similar research purposes, they often used already existing Discord servers or communities [25]. Our approach allowed for the bottom-up and community-forward creation of a Discord server that fostered participatory benefits and improved engagement throughout the HCD process.

This study highlights the innovative use of Discord as a dynamic, youth-friendly platform for engaging LGBTQ+ youth in HCD activities. By leveraging a space where participants already feel comfortable and connected, we demonstrate how this digital environment can facilitate more authentic, accessible, and sustained involvement in the co-design of digital health interventions. Our findings underscore the potential of integrating familiar social media technologies into participatory research practices, ultimately advancing inclusive and acceptable digital intervention design. This work contributes valuable insights for researchers and designers who engage in HCD to create more equitable and responsive interventions for marginalized youth communities.