We conducted a qualitative process evaluation, comprising semistructured interviews with participants following completion of the 13-week RCT [29]. The process evaluation was embedded within the trial and examined participants’ experiences of the Purrble intervention, with analyses conducted independently and before the quantitative RCT findings were shared with the research team, to aid interpretation of the eventual quantitative results [26].

Participants gave informed, written consent at enrollment to the RCT. This encompassed consent to be contacted for interview, be interviewed using audio recording, and for transcription processes. The consent form explained how data would be managed, including that only the King’s College London research team would have access to audio recordings, which were stored in encrypted files. Transcription services were required to sign a confidentiality agreement before receiving any audio data, and all identifiable information was removed during transcription. All instances of audio recordings were deleted once transcription was complete. Participants were required to consent to all of these elements in order to take part in the study.

As compensation, all participants were offered £10 (US $13.27) for their time to participate in the interview. As part of the wider RCT, participants kept their Purrbles after the project was completed.

All study procedures were approved by the ethics committee at King’s College London, UK (RESCM-22/23‐34570) and the trial procedures were prospectively registered ahead of participant enrollment (ClinicalTrials.gov NCT06025942) and published [30].

Members of Sprouting Minds, a youth advisory group with lived experience of mental health difficulties and active involvement in the UK Research and Innovation (UKRI) funded Digital Youth program, played a key role in developing the associated RCT and process evaluation, building on their prior contributions to related projects [22,29]. Advisors reviewed and refined the interview schedule preethical submission, ensuring alignment with study priorities and streamlining inquiry to participant experiences of using Purrble. After data collection, advisors were engaged in reviewing the analysis to confirm that developing codes and themes reflected participants’ perspectives, from their interpretations.

As part of the RCT [29], participants were stratified and randomized to either TAU, receiving only the safety planning intervention, or the Purrble condition, where they received safety planning and the Purrble device. Following the final trial survey, all TAU participants also received a Purrble to keep.

Once participants had finished their final RCT assessment and TAU participants had been sent their Purrbles, all participants were emailed information about the interview with a link to fill in their availability if they wished to take part. The original aim was to collect data from up to ~40% of the Purrble intervention (n=37) and ~20% of the TAU condition (n=18) [30]. This sample would provide a broad, unselected sample maximizing potential experiences captured, and representing a significant proportion of the trial participants [36].

All interested participants were invited to interview (n=55), with priority given to those in the intervention group to ensure control condition members would have received Purrble and had time to engage with the device before they took part in the interview. Eleven participants did not respond when they were contacted for an interview, while 4 were unable to find a suitable time.

Forty participants (21 intervention; 19 TAU) took part in online, semistructured interviews between August-September 2024. Interview participants were all aged between 16‐25 years, representing a range of sexual orientations and gender identities. In the RCT, randomization was stratified by gender identity with consideration given to these differences within analysis. Within interview participants, 21 identified as cisgender and 19 were transgender or gender diverse. This demographic characteristic is denoted by either (TGD) or Cisgender# (cisgender), for clarity. Twenty-nine participants disclosed confirmed or suspected neurodiversity (eg, autism, attention-deficit hyperactivity disorder [ADHD], and dyslexia), 10 stated they had none, and 1 participant did not offer any data on this. Further descriptives are not offered to prevent participant identification, given the physical nature of the intervention and combination of demographics available.

The interviews broadly discussed (1) expectations of the RCT and Purrble; (2) whether their safety plans had been used, and to what degree; (3) engagement and appropriation of Purrble; (4) Purrble’s impact on well-being and mental health, including self-harmful thoughts and behaviors; (5) other coping mechanisms; and (6) potential wider impact for Purrble. The semistructured nature of the interview allowed for reflexivity and flexibility to explore participants’ experiences as they are disclosed [37].

Participants were interviewed by AJW and LP; these were single online interviews with only the researcher and participant present. Both researchers created reflexive diaries to capture their thoughts and observations from the interviews; these were used to support later analysis. The interviews typically lasted 28 minutes (range 12-60 minutes) and were audio recorded. These recordings were transcribed verbatim, with identifying information being removed for confidentiality.

Qualitative process evaluation analyses were conducted independently and prior to any quantitative analysis of the RCT data to minimize potential bias and support unbiased interpretation of participants’ experiences [26].

As 90% of the interviews were conducted by AJW, initial analysis began during the data collection and was used to determine that data saturation was achieved before recruitment ceased [38]. This was discussed with LP to ensure similarity across interviews and interviewers. Thematic analysis was used following the steps outlined by Braun and Clarke [39,40], with AJW leading the analysis (100% of transcripts), supported by LP who conducted 10% of the interviews and analyzed 25% of the transcripts.

After data familiarization, both authors (AJW and LP) extensively coded transcripts for topics, experiences, perspectives, and opinions relating to the research questions using line-by-line coding, creating their own initial frameworks. Reflexive diaries were used to document initial impressions of interviews, emerging interpretations, notable participant narratives, and potential assumptions arising during engagement with the data. For example, researchers recorded instances where particular issues appeared especially important to participants, alongside reflections on how their own perspectives may influence interpretation. These reflections were revisited throughout coding and theme development to support ongoing reflexivity and critical examination of analytic decisions. The initial coding frameworks were subsequently compared, with similarities, differences, and areas of overlap discussed in relation to the wider dataset and relevant contextual information. These discussions were used to challenge assumptions, explore alternative interpretations, refine conceptual boundaries, and ensure that themes remained grounded in participants’ accounts. This framework was then reflexively discussed among the team to establish credibility and methodological rigor of the analysis.

Among participants, Purrble was found to support in-the-moment grounding, which allowed them to regulate their emotions in periods of distress. Over half of participants described using Purrble to manage distress related to self-harm, with grounding functioning as a cross-cutting mechanism across the self-harm cycle (Subtheme 1a). Specifically, participants reported using Purrble when recognizing preceding thought patterns to self-harm (Subtheme 1b), to manage the self-harmful thoughts themselves (Subtheme 1c), or to prevent acting on the urge to self-harm (Subtheme 1d).

Purrble was positioned by participants as a relational tool that provided emotionally meaningful yet free from the expectations associated with human interaction. Participants described experiences of Purrble as a comforting presence (Subtheme 2a), a means of sidestepping social demands (Subtheme 2b), and a mechanism for externalizing emotions (Subtheme 2c). Among the participants, loneliness and isolation associated with LGBTQ+ identity or neurodiversity was discussed at length, with youth suggesting that Purrble had the potential to fill a void for some and thereby encourage repeated engagement. This perception was related to Purrble being a constant presence who was undemanding of external needs.

Purrble was designed to support in-situ ER; participants found that if they were able to form a habit with the device, it was consistently useful. However, participants needed to have Purrble close to hand when they were distressed to maintain engagement (Subtheme 3a), with some participants describing it as not accessible enough (Subtheme 3b). This limited how useful Purrble could be in moments of dysregulation.

Some participants found that Purrble was not helpful for them, as it was not suited to their needs, coping strategies, or in a few instances their own acknowledged biases: “I don’t maybe have the patience to have like sat with it and stuff.” (T47, TAU); “just didn’t feel right because I’m a 22-year-old male with a cuddly toy.” (TGD35, intervention). But mainly, this was related to Purrble being unhelpfully stimulating (Subtheme 4a), which made it unsuitable for those with certain experiences or sensitivities (Subtheme 4b).

The purpose of this qualitative process evaluation was to understand how LGBTQ+ young people perceived and experienced Purrble (delivered alongside safety planning) as a novel intervention to support in-situ ER. Although analyses were conducted independently and prior to access to quantitative outcomes, the findings are highly complementary to those of the main RCT [29]. Specifically, the trial demonstrated that participants in the Purrble intervention condition reported significant reductions in ER difficulties, anxiety, and depressive symptoms compared to TAU controls, which aligns with the present qualitative findings, where participants described in detail the ER strategies they developed and deployed through interacting with Purrble to manage distress.

While the main RCT did not demonstrate statistically significant effects on self-harm outcomes [29], it is important to note that the trial was not powered for self-harm as a primary outcome. These qualitative findings should not be interpreted as evidence that Purrble prevents self-harm or as an explanation for the RCT’s null self-harm findings. Instead, participants described experiences of using Purrble to identify, manage, and respond to distress, which was often related to their self-harm, providing insights into how they perceived Purrble to support their emotional experiences in real time. Some participants felt that engaging with Purrble helped them recognize and regulate distress before it escalated, aligning with previous findings [22]. In this way, the qualitative findings complement the trial results by generating contextual hypotheses regarding how improvements in proximal outcomes (eg, emotion regulation, anxiety, and depression) may be experienced by participants and how such changes might relate to self-harm risk. These proposed mechanisms remain speculative and should be examined directly in future research.

Furthermore, these qualitative findings also help to theorize why differential effects were observed within the RCT [29]. A small number of participants reported that Purrble could trigger gender dysphoria or felt misaligned with identity-related stereotypes. This provides a potential explanatory mechanism for the observed interaction effects in the RCT and may help to account for why Purrble appeared less supportive for some transgender and gender-diverse participants [29] and highlights the importance of considering identity when interpreting engagement with interventions [6].

A key aspect of these findings was participants’ use of Purrble to manage feelings of loneliness and social isolation, both of which were described as contributing to emotional distress. This is particularly noteworthy given evidence that LGBTQ+ individuals experience high levels of loneliness, social rejection, marginalization, and isolation [5,41-44]. Such experiences are often understood within the minority stress framework [45,46], which proposes that sexual orientation and gender identity minority individuals face additional chronic stressors related to their identity, including stigma, discrimination, and expectations of rejection. These experiences can further harm social connectedness and contribute to poorer mental health outcomes. Whereas strong social relationships and supportive networks are consistently associated with reduced psychological distress and lower risk of self-harm among LGBTQ+ populations [5,46,47]. Therefore, participants’ descriptions of Purrble as companion-like suggest that its benefits may extend beyond emotion regulation alone. For some young people, particularly those with limited social support, those experiencing social withdrawal, or those who are neurodivergent and find interpersonal interactions more demanding, Purrble may provide a temporary source of comfort, connection, and perceived support during periods of distress. While not a substitute for human relationships, participants’ accounts suggest that Purrble may help bridge moments of isolation and provide emotional support, particularly when a young person is already feeling overwhelmed or socially tired.

This study highlights the value of incorporating process evaluations within trials. Process evaluations are increasingly recognized as essential for understanding how complex interventions operate and which components drive observed effects [25-27]. However, few RCTs in LGBTQ+ mental health research have incorporated this methodological approach [14,48-51]. In this study, the process evaluation enabled examination of the relative contributions of Purrble and safety planning within the intervention package. Across conditions, participants rarely referenced their safety plans, and when prompted, many struggled to recall them. Instead, narratives consistently centered on Purrble, suggesting that the observed impacts were largely attributable to the device itself rather than the combined intervention (Purrble+safety planning). It is possible that participants’ limited recall or use of their safety plans reflected a lack of need, as few participants described experiencing the type of suicidal crisis for which the Brown and Stanley safety plan was originally designed [31]. Nevertheless, these findings contribute to ongoing debates regarding the practical utility and implementation of safety plans in practice. While safety planning remains a standard component of interventions for individuals at risk of self-harm or suicide in both research contexts [22,32,33,35,52] and UK clinical practice (eg, Improving Access to Psychological Therapies [IAPT]) [34], previous evidence indicates that engagement is often limited during periods of acute distress [53]. Factors such as ambivalence, reduced confidence in usefulness, or forgetting the plan altogether may contribute to this disengagement [54]. The limited prominence of safety plans within participant accounts in this study aligns with and extends these concerns.

One consideration relates to the characteristics of the process evaluation sample. Participants were recruited from those who had taken part in the related trial [29] and consented at enrollment to being contacted for a qualitative interview. While this approach enabled the capture of experiences from individuals with direct exposure to the intervention and minimized participant burden, it may also have introduced self-selection bias. Participants who agreed to be interviewed may have been more engaged with the intervention, more willing to discuss their emotional experiences, or more motivated to share particularly positive or negative views than those who declined participation. In addition, individuals who may have become interested in contributing their perspectives at a later stage were not captured through this recruitment approach. As reasons for declining participation were not collected, it was not possible to determine whether systematic differences existed between interview participants and the wider trial cohort. This should be addressed in future trials investigating the impact of Purrble.

Neurodiversity was also highly represented within the sample and may have influenced participants’ engagement with Purrble. Many participants described the device’s tactile and sensory features as supporting grounding and ER during periods of distress, consistent with previous work highlighting sensory engagement as a key contributor to Purrble’s acceptability [22,23]. However, sensory features were not universally experienced as beneficial, with some participants reporting that the vibrations could be overstimulating and contribute to disengagement. Given that sensory preferences and sensory regulation are important considerations for many neurodivergent individuals, these factors may have shaped both positive and negative experiences of the device [55]. As neurodivergence was not examined systematically within the analysis, it was not possible to determine whether engagement or perceived benefits differed between neurodivergent and neurotypical participants. Consequently, although these findings provide valuable insight into how LGBTQ+ young people may engage with and benefit from Purrble, they are not intended to be representative of all trial participants, but rather to offer potentially transferable insights for similar populations and contexts. From a clinical and policy perspective, these findings have important implications. Demand for psychological support among young people in the UK is substantial: approximately 1 in 5 meet criteria for a probable mental disorder, most commonly anxiety and depression, and anxiety is the leading reason for referral to National Health Service (NHS) child and adolescent mental health services [56,57]. In this context, the cost of providing a Purrble device is relatively low (up to £50 plus £6 shipping [US $66.34+$7.96]) compared with the estimated cost of 1 hour of NHS clinical psychologist time (approximately £69–£75 [$91.55-$99.51]) [58], highlighting its potential as a scalable, low-intensity support option. Purrble may therefore represent a cost-effective intervention, particularly for LGBTQ+ young people who face persistent structural and interpersonal barriers to accessing formal mental health support [9,10,12], although this remains to be established through formal economic evaluation. As an independent, self-directed tool, Purrble can provide support at moments when young people might otherwise disengage from services. Participants’ narratives suggest that it functions as a source of perceived social support, a well-established protective factor against self-harm [59,60], by offering a consistent, nonjudgmental presence that promotes ER strategies.

Crucially, Purrble’s mechanisms of change are transdiagnostic [20], targeting ER processes that underpin diverse mental health and well-being outcomes. This process-level focus, rather than a diagnosis-bound approach, means that Purrble may be suitable across clinical services, educational environments, and medical settings. For example, Purrble may be particularly valuable for young people on waiting lists for mental health services, by promoting foundational ER skills and scaffolding engagement with later therapy. This is a particularly important application context because young people awaiting services are at elevated risk of suicide attempt [61] and frequently report feelings of disengagement and disenchantment prior to treatment [62], whereas the provision of waiting list interventions may help to mitigate these risks [63]. More broadly, socially assistive robots have been shown to be feasible and acceptable within educational and healthcare settings, supporting learning, confidence, and reductions in anxiety [64-66]. Purrble may therefore offer a means of extending such benefits to older adolescents and young adults, including in emotionally demanding situations such as difficult conversations with education staff or managing medical anxiety. Across these settings, the provision of a Purrble device for private, self-directed use represents a relatively low-cost approach with the potential to reduce acute distress.

Our findings provide several directions for additional research. First, studies should examine potential differences in Purrble use and perceived impact across sexual orientations and gender identities, with particular attention to whether specific design features may inadvertently evoke discomfort for young people experiencing gender dysphoria. Such work would inform refinements to ensure the intervention remains affirming and psychologically safe. Where appropriate, future research should also adopt coproduction approaches to ensure that interventions are developed in partnership with those who may use them [67], in this case TGD young people. Indeed, it is unlikely that a single intervention will be effective or acceptable for all, particularly in the context of mental health and self-harm. Individuals differ in their needs, preferences, and circumstances, and interventions should reflect this diversity. The aim, therefore, should be to develop and evaluate a range of evidence-based approaches that can be tailored to different needs and contexts. Such flexibility is likely to maximize engagement, acceptability, and effectiveness across diverse groups. Second, participants often forgot about the safety plan component of the intervention, a common challenge reported in the safety planning literature [53]. Future intervention designs may benefit from incorporating multiple opportunities to revisit and review safety plans in order to enhance engagement and encourage their use during periods of distress. These intervention nudges may promote active interaction with safety plans, requiring individuals to review, update, and personalize content over time rather than passively receiving information. Active choice engagement is typically associated with greater behavioral impact than information provision alone [68]. Third, many participants highlighted neurodiversity, noting that the tactile and sensory aspects of Purrble were especially beneficial during periods of distress. Given the overlap between neurodiverse and LGBTQ+ identities, with evidence indicating that LGBTQ+ youth are approximately twice as likely to be neurodiverse [69,70], that neurodiverse young people are also underserved [71], and the substantial representation of neurodiverse participants within the present study, future research should systematically investigate whether neurodiversity influences uptake, engagement, or perceived benefit of Purrble. This could help disentangle whether the benefits observed are associated with LGBTQ+ experiences specifically, neurodiversity, or the intersection of both identities. Such research may clarify whether participant labels alone mask important individual differences. Finally, findings from the RCT indicate that Purrble improves ER and broader clinical outcomes over a 13-week deployment [29], while qualitative data further suggest young people can use Purrble in order to respond to and manage their distress associated with self-harm. A fully powered trial specifically designed to examine the impact of Purrble on self-harm outcomes is therefore warranted.

Several limitations should also be considered. The findings may be subject to self-selection bias; as although some participants articulated ambivalence toward Purrble and identified contexts in which it was less suitable, it is likely that those who agreed to be interviewed were more positively disposed toward the intervention than the wider trial cohort. Furthermore, as information regarding why RCT participants declined interview participation was not collected, it is not possible to assess whether interview participants differed from nonparticipants. Assessment of those who took part in these process evaluation interviews and those who did not has not been conducted. In addition, while the trial recruited participants with a broad range of sexual orientations and recruitment to interviews aimed to balance gender identity broadly (cisgender vs transgender or gender diverse), sexual orientation was not explicitly incorporated into interview sampling or analytic decisions. Consequently, potential differences in how sexual orientation may have shaped perceptions of Purrble’s utility, appropriation, or engagement were not examined in depth.

This qualitative process evaluation provides in-depth insight into how LGBTQ+ young people engaged with and interpreted the Purrble intervention following participation in an RCT [29]. Across 4 themes, participants described varied but meaningful ways in which Purrble was appropriated to support in-the-moment emotion regulation, manage distress, and navigate difficult emotional states. Importantly, these findings help to contextualize and theorize why significant improvements were observed in the trial in emotion regulation, anxiety, and depressive symptoms [29]. Participants’ accounts illustrate how Purrble facilitated the development and deployment of practical ER strategies in real-world settings, offering a plausible mechanism through which these outcomes were achieved. The qualitative findings also provide insight into how participants understood the relationship between Purrble, emotional distress, and self-harm-related experiences, offering perceptions of how Purrble might support the management of distress and self-harm urges as described in participants’ individual narratives. However, the findings should not be used to explain the absence of statistically significant effects on self-harm outcomes observed in the RCT, given the subjective nature of individual narratives and qualitative analysis. Finally, the variability captured across themes underscores that Purrble is not universally acceptable or beneficial, and that individual differences, such as gender identity, preferences, or past experiences, can shape engagement with the intervention, suggesting further research is needed to understand the value of Purrble with such populations. Overall, this study highlights the potential of low-intensity, accessible interventions like Purrble to support ER among LGBTQ+ young people, while also emphasizing the importance of understanding for whom and under what conditions such interventions are most effective.