Young people are facing a crisis of mental ill-health, with estimates suggesting that 1 in 5 have a mental health disorder [1], making them the most disproportionately affected age group for mental health burden [2], with upward of 50% affected by the age of 25 years [3]. The magnitude of this problem has increased over the course of the global pandemic [4], and rates of mental ill-health have steadily escalated over the last decade, placing unprecedented demands on services that exceed available resources [3,5]. For example, an Australian survey of young people found 44.1% of the general population met the criteria for depression and 48.5% for anxiety [6]. Critically, estimates suggest that only 50% of young people with mental ill-health seek help [7]. Among those who do seek care, many encounter long waiting periods [8] and disengage prematurely, resulting in only a small proportion of people who experience long-term recovery [9,10]. Due to these barriers, as many as 55% of young people worldwide are experiencing unmet needs for mental health care [11]. Smartphone technologies present an important opportunity to address the growing burden of mental ill-health among young people by providing real-time support without reliance on traditional face-to-face services [12,13].

High rates of smartphone ownership, coupled with a strong interest in using digital devices for mental health support among young people and clinicians, provide important opportunities to overcome barriers to mental health care for youth [14]. Smartphones possess a critical capability for the self-management of mental ill-health, enabling the detection and response to daily life experiences through personalized interventions [15]. “Just-in-time adaptive interventions” (JITAIs) represent a novel approach wherein smartphones or other mobile devices are leveraged to tailor momentary intervention strategies based on an individual’s specific context and needs [16]. While research evidence is still developing, a meta-analysis of 33 studies spanning physical and mental health conditions found evidence supporting the effectiveness of JITAIs compared to controls [17].

A unique opportunity to improve treatments for depression and anxiety in young people may involve the use of JITAIs to deliver momentary interventions targeting transdiagnostic mechanisms in real-time, real-world scenarios. Transdiagnostic mechanisms refer to underlying psychological or behavioral processes that contribute to multiple mental health conditions irrespective of their specific diagnosis [18]. These mechanisms potentially offer a more personalized and precise intervention target, applicable across a diverse array of mental health disorders, thereby enhancing their relevance and efficacy [19]. JITAIs vary in their complexity, from basic static algorithms that select a limited number of interventions to more sophisticated systems that are contextually responsive and capable of learning from user data and individual behavior. However, many JITAIs currently in use rely on static algorithms [20]. Repetitive negative thinking (RNT; [21]) is a putative transdiagnostic mechanism with considerable evidence implicating a causal role in driving depression and anxiety symptoms, making it a highly promising intervention target [22]. RNT involves frequently occurring negative thoughts that are passive or hard to control, such as worry (concern for future events) and rumination (concern for past or current events or depressive symptoms; [21]). High levels of RNT are commonly observed among clinical populations of youth with clinical depression and anxiety and are highly correlated with the severity of depression and anxiety symptoms [23]. Inducing RNT in experimental studies results in worsening mood states [24], and heightened RNT predicts the later onset of emotional disorders in youth [25,26]. Meta-analyses have further shown that improvements in RNT are associated with reductions in depression and anxiety following psychological treatment in adults [27,28] and youth [14]. Therefore, targeting a reduction in RNT in treatment appears to be an important component of psychological treatment.

Various psychological treatment approaches, such as rumination-focused cognitive behavioral therapy, metacognitive therapy, and attention training, have been found to improve RNT outcomes [5]. However, traditional treatment delivery methods involving infrequent face-to-face sessions with a clinician suffer not only limitations in accessibility but are also poorly timed to target the process of RNT as it occurs in daily life. RNT is a dynamic process that is influenced by external and internal events over time [29]. Therefore, intervention strategies should be appropriately timed and sensitive to contextual variables that might influence how this process unfolds over time [30]. JITAIs are potentially well suited to targeting RNT in real time by detecting and delivering highly personalized and context-sensitive interventions designed to disrupt the process as it is happening. To test this theoretical intervention model, a JITAI called Mello was developed.

Mello is a self-guided smartphone intervention designed to reduce RNT (referred to as “stuck thinking” within the app) among young people experiencing depression and anxiety [31]. By using momentary assessments of stuck thinking intensity, mood states, location, and activity, users receive tailored recommendations for a microintervention strategy aimed at disrupting RNT in real time. In a pilot randomized controlled trial (RCT) [31] involving young people who used Mello for 6 weeks compared to a nonactive control group (n=55), the intervention was found to be acceptable, feasible, and potentially effective at reducing RNT, leading to improvements in depression and anxiety with moderate to large effect sizes. Furthermore, 96% of young people would recommend the app to others with similar experiences, with high satisfaction ratings across all domains. Young people also showed sustained use of Mello over time, with average use on 52% of total days and 59% of young people engaging with the app for the full 6-week period. These results demonstrated the highly promising potential of the Mello app and the broader theoretical concept of harnessing JITAs to target transdiagnostic mechanisms using personalized, context-sensitive interventions within real-time, real-world environments.

Building on the results of the RCT evaluating the feasibility and preliminary treatment effects of the Mello app [31], this study aimed to qualitatively explore young people’s experiences with Mello. This qualitative perspective can provide rich insights into how young people engage with and make meaning of their use of Mello, perceived benefits and limitations, and suggestions for refining and enhancing future iterations of the app. A better understanding of young people’s experiences with Mello not only informs potential benefits and areas for improvement but also contributes to the broader understanding of designing mobile real-time digital mental health interventions to enhance outcomes for young people experiencing mental health challenges.

This qualitative study was conducted as part of a broader research project examining Mello—a transdiagnostic smartphone app targeting RNT for young people experiencing anxiety or depression. The feasibility, acceptability, and preliminary clinical outcomes and mechanisms of Mello were examined in a pilot RCT [31]. During the pilot, participants were randomly assigned to either Mello or a nonactive control group for a 6-week period. This qualitative study was conducted with a subset of participants of the Mello intervention group willing to be interviewed following the completion of the 6-week intervention period.

Mello’s use of the JITAI model involved identifying and intervening in moments of need by using real-time assessments to tailor momentary intervention strategies aimed at disrupting RNT (Figure 1). In the pilot RCT, participants allocated to the Mello group received randomized notifications 3 times a day prompting them to complete “check-ins”: a brief 4-item questionnaire completed via the chat interface assessing the level of RNT, mood, current activities, and location. On the basis of the response provided during the check-ins, the app recommended 1 of 12 cognitive behavioral activities (Multimedia Appendix 1). Each activity took approximately 2 to 12 minutes to complete and was delivered through guided text or audio instructions using evidence-based strategies for targeting RNT. The selection of therapeutic activities was informed by a meta-analysis of psychological treatment approaches for RNT [5]. These activities included exercises such as mindfulness, problem-solving, cognitive defusion, worry time, thought challenging, gratitude, and self-compassion [31].

Participants in the pilot RCT were aged 16 to 25 years, experiencing clinical levels of depression and anxiety (≥10 on the Patient Health Questionnaire 8 [32] or Generalized Anxiety Disorder Scale 7 [33]), and elevated RNT levels (≥37) on the Perseverative Thinking Questionnaire [21]. Additional inclusion criteria involved the ability to provide informed consent, sufficient command of the English language, and ownership of a smartphone capable of running the Mello app (Android [Google LLC] or iPhone [Apple Inc]). Exclusion criteria comprised current psychiatric inpatient status or receiving treatment from a mental health crisis management team. There were no additional exclusion or inclusion criteria, beyond being randomized to the Mello arm of the RCT, for participating in the present qualitative study.

All participants assigned to the Mello intervention arm of the pilot RCT (N=29) consented to follow-up after the 6-week intervention period and were invited to participate in a qualitative interview exploring their experience with Mello. As such, at the conclusion of the intervention period, emails were sent to all participants in the Mello intervention arm, inviting them to participate in a qualitative interview. Of the 29 young people invited, 15 (52%) responded and agreed to participate. The gender distribution, mean age, and country of birth within this subgroup were consistent with the overall sample from the pilot RCT.

There was variability in app engagement among the sample: with active use ranging from 6 to 38 days (out of a possible 42) during the 6-week intervention period (mean 21.9, SD 11.1) The engagement levels among this sample were reflective of the broader use patterns observed in the original trial, where participants used the app on 49.2% of days and completed an average of 29.9 check-ins [31]. This variation in engagement is valuable in qualitative research, as it allows a more nuanced understanding of user experiences across the spectrum of engagement, providing insights into both facilitators and barriers to sustained app use across different engagement levels. By including participants with varying degrees of interaction with Mello, this study aimed to identify the factors that either support or hinder ongoing engagement with the app, offering a more nuanced perspective on its efficacy and usability.

The qualitative interviews were conducted between February 8, 2022, and May 2, 2022. On average, participants were interviewed 1.9 weeks (range 0-4.7 wk) following the final date they had access to the Mello app. All interviews were conducted via Zoom (Zoom Video Communications, Inc) by authors CA and EC. CA, a female clinical psychologist with a PhD in clinical psychology, specializes in digital mental health and has extensive experience working clinically with people experiencing mental health challenges. She has led and contributed to multiple qualitative studies from study design to data collection and analysis. EC, also a female researcher, holds a master’s degree in public health and has substantial demonstrated experience in qualitative research, particularly in co-designing and interviewing young people about their experiences with digital mental health products. Both CA and EC had prior relationships with participants, as they were responsible for conducting pre and post-interviews with participants as part of the Mello RCT.

The interview schedule was designed to better understand young people’s experiences and perspectives of the Mello app. Multimedia Appendix 2 provides the full interview schedule. All interviews were conducted and audio recorded via Zoom to facilitate transcription and analysis. The recordings were transcribed verbatim by a third-party service. Both the audio files and transcripts were stored securely on a password-protected server. Identifying information, such as participant consent forms, was stored separately to ensure confidentiality. Access to the recorded interviews and transcripts was restricted to the immediate research team responsible for conducting the analysis.

Thematic analysis, guided by the framework proposed by Braun and Clarke [34], was used for data analysis. To begin, a single transcript was independently coded by authors JN, LV, CA, and IB. Following this, the 4 authors participated in collaborative discussions to review and compare their initially identified codes and coding process. This iterative process allowed for the refinement of an initial coding dictionary, resulting in the development of a robust coding framework.

To further ensure consistency, authors LV and CA independently coded and discussed 2 additional identical transcripts. This additional coding was compared to the established framework to further solidify alignment and consistency in the coding process. With confidence in the coding alignment, author LV proceeded to independently code the remaining 12 interviews. The coded data were then systematically grouped based on similarities, and through an iterative process among authors, overarching themes and subordinate themes were developed.

The systematic thematic analysis approach used in this study established a strong framework for comprehensively capturing and understanding the nuanced experiences and perspectives of participants using Mello.

The study was approved by the University of Melbourne Human Research ethics committee (project 2021-22524-23553-5). Participants provided written informed consent through the REDCap (Research Electronic Data Capture; Vanderbilt University) web application. Before providing consent, participants were given a thorough explanation of the study in simple language, including the study’s aims, procedures, and any potential risks or benefits. Participants were also reminded that they could withdraw from the study at any time, skip questions, or take a break from the interview without any negative consequences. Participants were given a copy of the participant information and consent form and received Aus $45 (US $29.5) as reimbursement for their participation in the research interview. All data were deidentified.

All research questions were intentionally designed to be open-ended to prevent leading participants in predetermined or unconsciously biased directions. In the interview consent process, young people were explicitly assured that their honest opinions were valued, and the research team welcomed negative and constructive feedback as much as positive input.

Throughout the analysis phase, the research team held regular meetings to critically review codes and thematic patterns. These sessions were instrumental in considering how the research teams’ personal experiences, viewpoints, beliefs, and biases may be shaping the interpretation of results and influencing data analysis. The team acknowledges the potential impact of their own beliefs and experiences on the interpretation process.

Ultimately, participants’ experiences were deeply considered, with the research team composing rich and detailed paragraphs linked to illustrative quotes. This approach aims to ensure a transparent and traceable lineage between the reported findings and the actual words and experiences shared by the young people themselves.

The study included 15 participants, with ages ranging from 16 to 25 (mean 20.9, SD 2.9) years. Overall, 60% (9/15) of participants identified as women, 27% (4/15) as men (including 1 transgender man), and 13% (2/15) as nonbinary. Furthermore, 73% (11/15) of the participants were born in Australia, while 27% (4/15) were born overseas. In addition, 53% (9/15) of the sample were receiving mental health treatment while using Mello. The average interview duration was 31.9 minutes, with interview times ranging from 17.0 to 67.2 minutes. Additional demographic details are presented in Table 1.

The analysis identified three superordinate themes reflecting distinct aspects of participants’ experiences with the Mello app: (1) Mello as a tool for intentional reflection; (2) doing therapy your own way; and (3) barriers to engagement during moments of low mood, anxiety, and RNT. These themes, along with representative quotes from participants, are presented in the Results section.

Superordinate theme 1 highlighted how Mello served as a tool for intentional engagement with RNT. Participants described how the app facilitated active self-management of negative thoughts, contrasting with their usual strategies of either resisting support or avoiding difficult emotions. Subordinate themes include habit formation, where automated check-ins helped participants build consistent reflective habits, and enhanced self-awareness, allowing them to recognize emotional patterns through personalized reflection.

Superordinate theme 2 focused on the value of Mello’s self-guided nature, enabling participants to manage their mental health autonomously. Many participants found that the app’s flexibility and accessibility alleviated the anxiety that can be associated with traditional face-to-face therapy, and they appreciated its availability during late hours or when other support systems were inaccessible.

Superordinate theme 3 explored the barriers to engagement, particularly during moments of low mood, anxiety, and RNT. Some participants struggled to engage with the app during emotionally difficult times, feeling “stuck” in their negative thoughts. To address these challenges, participants suggested incorporating gamification elements, such as progress-tracking visuals, to enhance motivation and make the app more engaging and rewarding. Textbox 1 presents a visual representation of these themes.

This study aimed to explore young people’s experience of using Mello, a transdiagnostic intervention designed to alleviate RNT associated with depression and anxiety. A total of 15 participants took part in the study following a 6-week intervention period.

The analysis identified three key themes from participants’ experiences with the Mello app: (1) Mello as a tool for intentional reflection; (2) doing therapy your own way; and (3) barriers to engagement during low mood, anxiety, and RNT. In theme 1, participants described how the app could facilitate the active management of negative thoughts and supported the development of reflective habits, contrasting with their typical strategies of avoidance or resistance. Theme 2 highlighted the value of the app’s self-guided nature, with a particular focus on its flexibility and accessibility, particularly when compared to traditional face-to-face therapy. Finally, theme 3 addressed the barriers to engagement, particularly during emotionally difficult times, with participants reporting feeling “stuck” in their negative thoughts. To mitigate these challenges, participants suggested incorporating gamification elements, such as progress-tracking visuals, to enhance motivation and increase engagement with the intervention. Young people’s qualitative experiences spoke to the effectiveness of the JITAI framework [16] in fostering intentional engagement and reflection with thoughts and emotions in times of need, as presented in the first overarching theme. For some, this framework facilitated positive behavioral changes by encouraging active engagement with and reflection on their RNT. It offered a constructive alternative to habitual responses such as resisting support or attempting to avoid engaging with their rumination, with the hope of achieving quicker resolution from difficult feelings. Instead, it encouraged intentional engagement in real time, facilitated meaning-making, and, for some, led to a reduction in RNT. These results may provide explanations for the findings of the pilot clinical trial of Mello [31], which demonstrated moderate to large improvements in depression, anxiety, and RNT in Mello users versus the control group. This suggests that providing brief, smartphone-delivered interventions during moments of RNT offers a promising pathway for disrupting key processes in real time in a way that intentionally engages conscious behaviors and counteracts habitual maladaptive responses. This opportunity is uniquely suited to the JITAI framework and opens opportunities for future research to explore the application of this approach for targeting other transdiagnostic processes with temporal and contextual qualities, for example, emotion regulation, avoidance, and cognitive biases. Future research may benefit from identifying and refining the timing of intervention prompts to optimally disrupt the RNT process, as well as determining which intervention strategy may work best for the individual in any given moment. This can be achieved using artificial intelligence (AI) approaches, such as reinforcement learning [35], which can learn from individual user data to adapt both the timing and type of intervention over time. By understanding the variety of activities and intervention types chosen by the user, these AI systems can better enhance personalized recommendations and potentially improve outcomes. Our data support the efficacy of the JITAI approach in addressing RNT. We aim to advance from a basic JITAI 1.0 model to a more sophisticated intervention. This next-generation JITAI will incorporate a greater variety of activities and use contextual information and inputs, leveraging AI to tailor the timing and delivery of interventions. This enhancement is expected to significantly improve personalization and therapeutic outcomes.

These findings are consistent with meta-analytic review by Wang and Miller [17], which found that JITAIs saw an increased additive effect when tailored to the individuals’ behavioral patterns and current needs. Wang and Miller [17] also found that the effectiveness of the JITAIs was further enhanced by the integration of both a human agent and an algorithm to generate personalized feedback. This mirrors the approach implemented in the Mello intervention in which a static algorithm tailored intervention content specific to the needs and behaviors of the participants. These findings provide support for the effectiveness of tailoring interventions to individuals’ needs in the moment and underscore the importance of combining human and algorithmic support in digital interventions. While Mello is a self-guided app without human support, the current findings suggest future iterations should emphasize personalization. By incorporating personalized notifications and content recommendations, the app can customize the user experience, creating a more engaging and tailored environment that mimics the experience of human support [36].

While some young people acknowledged the benefits of using Mello during difficult moments of rumination, others had difficulty engaging with the app due to low mood, stress, the challenge of RNT, or lack of active follow-up due to the self-guided nature of the intervention, as detailed in theme 3. Challenges with motivation are a qualitatively well-documented barrier to engagement with mental health interventions, which may be particularly notable during periods of low mood, heightened anxiety, or with self-guided interventions [37-41]. This theme aligns with broader findings in the digital mental health literature that demonstrates maintaining engagement, particularly in real-world contexts, remains a considerable challenge, with many interventions experiencing low rates of sustained use [42-45]. This highlights the importance of JITAI’s and future iterations of Mello, which are designed to provide support in the moment and are well-positioned to engage people with personalized reminders and content that is immediately responsive to the thoughts and feelings people are experiencing in the moment.

Advancing digital therapeutic interventions to support mental health likely necessitates further research into the experience of people with low motivation and the development of strategies tailored for this group. One avenue for investigation could involve identifying effective techniques or features within digital interventions specifically designed to engage low- users with low motivation, such as personalized support and encouragement and reminders or motivational messages. Qualitative research by Berry et al [46] with mental health service users emphasized the importance of interventions being perceived as “positive, fun, practical and interactive,” underscoring the necessity for prioritizing engaging and user-centered design in digital mental health interventions to more effectively engage people, particularly those with low motivation, in managing their mental health.

Furthermore, co-design in the mental health context and the involvement of end users in the design of interventions have been associated with a generally enhanced effectiveness [47]. Therefore, actively engaging users with low motivation in the design and implementation process holds great promise for gaining valuable insights into their unique experience of low motivation and for effective solutions. Personalization, in particular, tailors the app’s features and content to align with the individual preferences and needs of each user, creating a more relevant, autonomous, and engaging experience [36,48]. Gamification introduces game-like elements and incentives to make interacting with the app more compelling and motivating [36]. In a 2024 study, Lieder et al [49] examined the optimization of gamification as a method to facilitate positive behavior change; for instance, in an app aimed at increasing water intake, users who received feedback based on optimized gamification were more likely to enact the desired behavior more frequently compared to control groups. Overall, further research in this area can help identify strategies to motivate individuals more effectively during times of emotional difficulty, ultimately maximizing the potential of digital therapeutic interventions to support young people’s mental health.

The self-guided aspect of Mello was recognized for offering a nonjudgmental environment, empowering users to address their mental health needs without the pressure that can be felt in face-to-face therapy sessions. Beyond this, it offered users the flexibility to access support and manage their mental health beyond the constraints of traditional clinical hours, particularly when seeking support from family and friends was not feasible or preferred. This adaptability is further explored in theme 2, where participants described Mello as a self-guided tool that offers a sense of control and autonomy over their therapy experience. It provides a nonjudgmental outlet for expression, offers flexibility in therapeutic approaches, and ensures continuous accessibility, ultimately empowering individuals to independently and effectively manage their mental health challenges. This flexibility allows young people to engage, adjust, and switch therapeutic activities in real time, tailoring their experience to their needs and preferences. Thus, although research suggests that dropout rates for self-guided interventions can be as high as 80% [50], the experiences of participants in the Mello trial highlight the significant benefits that self-guided mental health interventions can offer. These include promoting autonomy, flexibility, and reducing anxieties related to therapy. Therefore, understanding how to sustain engagement while maintaining the positive impact of self-guided interventions is crucial for the ongoing development of digital mental health interventions for young people.

While all young people assigned to the Mello intervention group were offered the opportunity to participate, it is plausible that those with lower motivation may have been less inclined to participate, potentially skewing our findings. While our themes do capture insights from participants who expressed having low motivation, the results may not fully encapsulate the experiences of this subgroup. This study was also conducted in the context of using Mello within a clinical trial; therefore, the experience of real-world users may be different.

Although members of the youth advisory committee of Orygen provided feedback on the interview questions, the analysis phase lacked direct involvement from a young person with lived experience of mental health challenges. Moving forward, incorporating such perspectives would enrich our research endeavors. Young people with lived experience may identify nuances within experiences that those without such background might overlook, underscoring the importance of their inclusion in the analysis process.

Through qualitative exploration, our findings identified the value of Mello’s JITAI framework in fostering intentional engagement and reflection on RNT during times of need. This framework reportedly facilitated positive behavioral changes, steering young people away from habitual responses such as resisting support and toward active engagement and reflection on their RNT. These findings align with prior research by Wang and Miller [17], emphasizing the value of tailoring interventions to individuals’ needs and combining human and algorithmic support in digital interventions.

However, some participants faced challenges engaging with Mello due to low mood, stress, and avoidance related to the experience on RNT, which highlights the importance of addressing motivation barriers and avoidance in JITAI. Looking ahead, further research into strategies tailored for individuals with low motivation is imperative. This could involve exploring techniques such as co-design to identify effective personalized support and encouragement as well as furthering the incorporation of user-centered design principles such as personalization and gamification. Such efforts aim to enhance engagement and address the dropout rates associated with self-guided interventions.

Overall, while Mello’s self-guided nature provided a nonjudgmental space and flexibility for users to manage their mental health, sustaining engagement remains a significant challenge. As such, understanding and mitigating these challenges is crucial for maximizing the potential of digital therapeutic interventions to support young people’s mental health in the future.