This study is one element of a larger investigation of UK mental health communities called the Improving Peer Online Forums (iPOF) project. As per the iPOF research protocol [25], this realist evaluation represents the second stage of the project, which builds on a realist synthesis conducted in stage 1 [5]. Realist evaluation involves iteratively developing and refining program theories, or causal statements regarding how interventions work, as data collection progresses [26]. Following realist methodology, program theories from the realist synthesis pertaining to forum safety were used to inform this study. These program theories took the form of context-mechanism-outcome (CMO) configurations. CMOs are a primary unit of analytic output in realist studies and seek to elucidate how a mechanism, defined as a response to resources offered by an intervention, occurs within a particular context [27]. Contexts represent factors impacting the activation, or strength of activation, of a mechanism and exist at a range of levels including individual, organizational, social, and economic factors. Outcomes typically refer to cognitive, emotional, or behavioral consequences that follow from the activation of a mechanism. A total of 8 of the 22 CMOs on mental health forum impacts developed in the prior realist synthesis related to forum safety [5]. CMOs from the previous review described a range of forum features and experiences pertaining to user safety, such as the importance of forum moderation, the value of anonymity, and negative experiences emerging from forum use [25]. These CMOs were used to inform an initial program theory (IPT) framework (Table 1) for this study and influenced data collection, as described below.

Participants were recruited from 3 UK-based mental health forums partnered with the iPOF project. These forums were chosen from the 7 forums in the iPOF study as they represent 3 different service contexts, namely the UK National Health Service, corporate, and voluntary sectors. While only 3 forums were included in this phase of the research, the remaining 4 forums also represent these same sectors. It is hoped that this cross-sector sampling promotes the transferability of the findings of this study to a broad range of service settings. All forums had a mental health focus, could be accessed using anonymous usernames, and had dedicated moderation teams. Key forum characteristics are described in Table 2 and detailed descriptions of each forum partnered with the iPOF project can be found elsewhere [28]. As per an ethical framework developed by the iPOF team [29], forums are described using randomly allocated bird names to protect the anonymity of participating organizations and forum users. Participants were initially invited to participate in the iPOF project via an online survey advertised on forum home pages and through emails from forum hosts to forum users. Those who completed the survey were given the option to consent to be invited to participate in an individual interview.

The survey used in this study was an online questionnaire designed specifically for the iPOF project. It included 95 items across sections investigating forum use, perceptions of forum helpfulness, moderation, user mental health, use of other health services, and forum safety. Most items, including those reported here, were developed for the purpose of this study by the research team in collaboration with a public and patient involvement group to assess concepts and experiences identified by the iPOF realist synthesis. The questionnaire also includes validated measures of anxiety and depression. A full description of the survey is available elsewhere [30]. Only items directly related to forum users’ perceptions of safety or distress are reported here. This included (1) “I have felt distressed by some of the posts I have read on this forum,” (2) “I feel safe to post about my personal experiences in this forum,” (3) “I feel safe to say anything in this forum because no-one knows who I am,” and (4) ‘Talking about my mental health experiences in the forum might make me feel worse.” In this study, cross-sectional survey data were used to assess the prevalence of user-reported distress and feelings of safety, providing context for the program theories presented thereafter.

Participants who had completed the iPOF survey and expressed interest in a qualitative interview were invited to participate. Interviews were conducted via phone or Teams (Microsoft Corp) by PM and ZG, audio recorded, transcribed, and anonymized. The interview procedure followed the principles of realist interviewing [31]. Here, interviewers seek to examine existing program theories by introducing key concepts and ideas that participants are invited to reflect on based on their own experience of the program. Interviews therefore include a combination of deductive questions based on program theories and more exploratory, inductive questions, leaving open the possibility that new concepts not previously captured by the IPT framework may emerge. In addition to the IPTs, interviewers used a semistructured topic guide based on the concepts from these program theories (Multimedia Appendix 1).

Data extraction and analysis occurred simultaneously and followed three stages. First, data segments were extracted into a bespoke data extraction form based on the IPT framework. The form was hosted on Excel (Microsoft Corp) online to facilitate collaborative working. Each IPT was written out in full in a separate column. Data segments were copied under the most relevant program theory along with an analytic comment stating how the data segment contributed to the refinement of the respective IPT. Data extraction was conducted by ZG and PM. Second, data segments and analytic comments from across participant interviews were reviewed against the IPT framework. Particular attention was paid to where data supported, refuted, or refined IPTs, and new CMO configurations were formulated to account for the accumulating evidence. For example, IPT 8 refers to how a lack of moderation can contribute to negative outcomes. In interviews, however, participants reflected on the value of proactive moderation, leading to the development of a CMO configuration reflecting this aspect of their experience. IPTs for which there was insufficient evidence in the interview data to support meaningful iteration, such as IPT 1 regarding exposure to self-harm content, were not included in the subsequent analysis. Initial reformulation of the CMOs was conducted by PM. Following this, interim CMOs were reviewed in a full-day analysis meeting held by PM, ZG, FL, and HR, during which the CMO configurations were further refined with respect to illustrative data segments. Third, the analysis was shared with and reviewed by the wider iPOF team comprising academics, clinicians, and experts with experience in mental health. A final draft of this study was shared with an expert by experience whose perspective on this research is captured in a lived experience commentary presented at the end of this study. Lived experience commentaries are short pieces, typically written independently of the research team, that provide the opportunity for critical commentary from those with lived experience of the research topic [32].

This study received ethical approval from the Solihull Research Ethics Committee on June 20, 2022 (IRAS314029). All participants provided written informed consent. Interview data have been anonymized. Participants were compensated with £30 (US $40.31) shopping vouchers for completing interviews and £10 (US $13.44) shopping vouchers for completing surveys.

Semistructured qualitative interviews were conducted with 42 forum users from Magpie (n=16), Dunnock (n=10), and Sparrow (n=16) between August 2023 and October 2024. A total of 504 survey responses were collected from Magpie (n=110), Dunnock (n=287), and Sparrow (n=107) between July 2023 and March 2024. Participant demographics are available in Tables 3 and 4. Quantitative data are reported by survey item (Table 5). A total of 5 CMOs were refined and are reported with reference to illustrative interview data.

Survey responses revealed that over one-fifth of participants felt distressed by posts they had read (n=95, 18.8% agreed and n=18, 3.6% strongly agreed), indicating that even in moderated communities, a notable minority face challenges with exposure to forum content. Similarly, a quarter of respondents agreed (n=113, 22.4%) or strongly agreed (n=17, 3.4%) that discussing their mental health online could make them feel worse. Feelings of safety when posting was comparable to perceptions of safety due to online anonymity, with 42.5% (n=214) agreeing and 13.9% (n=70) strongly agreeing they felt safe to post, and 40.1% (n=202) agreeing and 19.2% (n=97) strongly agreeing they felt safe to share because of anonymity. This underscores the central role of online anonymity in fostering forum engagement.

Anonymity disinhibited users from engaging in online conversations. Several users noted that by not having an online profile linked to their name, they were able to avoid perceived stigma, and therefore, judgment related to their mental health.

By offering users the opportunity to engage anonymously, they are given control over what information they share about themselves, and therefore, how they are perceived online.

Similarly, the following user notes that a forum that does not require their profile to include identifying information provides freedom to decide what to disclose.

Interviewees identified the presence of moderators as an important resource for maintaining a positive community culture and ensuring that distressing content does not discourage users from engaging with the forum.

Issues moderators managed included interpersonal difficulties that could arise in forum conversations. As the following participant identified, their presence also provided reassurance.

Participants also highlighted the importance of moderators’ attentiveness to situations where users feel unsafe due to their mental health. One user valued the ability to request support from the moderation team where this difficulty emerged.

Highlighting the importance of proactivity, situations where moderators missed distressing content were identified as unhelpful.

Participants reflected on how restrictions on discussions around certain topics can be valuable for creating a supportive environment.

However, when implemented in a manner that feels rigid and insensitive, forum users could experience significant distress.

One user identified that this could have a muting effect on users’ willingness to engage openly with the forum.

Reflecting on the challenge of balancing restrictions with open dialogue, the following participant recognized the need for an approach that charts a path between forum openness and moderator control in the service of forum safety.

Some forum users recognized rule enforcement to be an essential element of forum safety. Where moderators do intervene, doing so in a way that communicates the intention to support the user may mitigate possible negative reactions.

Some interviewees noted that their experience of safety in an online forum could be undermined by unexpected and particularly emotionally charged content.

Participants noted the central tension that reading others’ experiences could be both challenging and helpful, necessitating a selective approach to what they access.

Examples of forum features that could support this form of self-moderation included content warnings.

Forums set up in such a way that the main text of a post was hidden also supported this decision-making process.

One user recognized that others posting about their lived experiences contributed to a sense of safety that promoted the reciprocal exchange of personal difficulties.

The following participant identified that those sharing authentic accounts of mental health challenges would be less likely to judge those going through similar experiences.

A culture of acceptance further encouraged the following participant to feel able to discuss their mental health experiences online.

This study builds on findings from a previously published realist synthesis by highlighting and expanding on factors impacting users’ perceptions of safety in online mental health forums. Anonymity was found to promote disclosure by mitigating against online posts being linked to users’ offline identities, and by providing users with greater choice regarding what they share online. Perspectives on rule enforcement emphasized the importance of organizations balancing between proactive moderation to ensure that safety issues were addressed quickly and the need to intervene sensitively where conversations on restricted, yet potentially important personal experiences occurred. Participants reflected on the need to choose when and how to use online forums in light of their mental health, recognizing the possibility that accessing forum content could be helpful but also potentially detrimental at times. Factors contributing to a psychologically safe forum environment included the perceived authenticity of others in the community, an atmosphere of nonjudgment, and the value of sharing the space with others experiencing similar mental health difficulties.

Previous research using both qualitative interviews and analyses of forum posting behavior suggests anonymity is a key factor in forum safety [11,34]. The finding that the value of anonymity lies in avoiding the judgment of others aligns with the suggestion that online anonymity mitigates adverse consequences linked to personal disclosure, including reputational harm, social pressure, and exclusion [35]. While in-forum anonymity appears broadly valued by users, variations in forum design highlight important choices regarding its implementation. Some forums emphasize complete anonymity such that users are unknown to each other and the host organization, while other platforms may collect personal information in the interest of safeguarding. Moderation practices also vary. Some forums prioritize online privacy and remove identifying information, while others may allow the sharing of some personal details, which as one participant noted here, can facilitate social connection. In balancing the advantages and challenges associated with anonymity, forum designers may need to consider relevant legal, ethical, and practical issues. These include legal obligations regarding the protection of personal data, the organization’s responsibility for user well-being, community members’ awareness of the implications of sharing their details online, and whether the forum is closed or open to the public [15,36]. Whether differences in the implementation of anonymity impact willingness to register or post to online forums is a potentially valuable avenue for further research.

Participants’ reflections on content moderation emphasized the value of proactivity in managing problematic content, and interpersonal sensitivity when intervening to implement rules. Results here highlight the need for forums to balance between the benefits of open online discussion and the potential for censorship and disengagement. This topic has been the focus of extensive recent debate. Legal frameworks, such as the European Union Digital Services Act [24], have sought to place tighter rules on providers, whose responsibilities now include the need to implement and evidence content moderation and risk management procedures. These developments align with concerns that access to algorithmically generated and poorly managed social media content could rapidly expose users to high volumes of harmful content [37]. However, a recent academic debate on approaches to moderation revealed concerns that greater restriction on discussions of mental health difficulties represents a missed opportunity to facilitate supportive discourse, and creates undue barriers to support for those already experiencing limited access to care [38]. As part of this debate, researchers produced a call to action for more nuanced, evidence-based approaches to moderation. Suggestions included greater emphasis on community moderation, where site members’ knowledge of the community’s culture is leveraged to help make decisions about acceptable content, for example, via specific roles for experienced community members or the use of content labeling to allow users to more effectively flag harmful content [38].

A novel contribution of this study relates to users’ reflections on the need to self-moderate their access to online mental health forums. The recognition that forum content can be helpful and at times detrimental aligns with a large body of social media–related research highlighting both the advantages of social connection and the impact of distressing content, social comparison, and harmful advice [39]. Interviewees valued forum design features that supported their decisions to access or avoid potentially distressing content, such as content warnings, and forum providers may wish to consider additional strategies to support the user’s self-management of their services. Social media websites have in recent years integrated features including customizable update feeds and tools that filter content containing certain words or phrases, links to emotional support resources from trusted partners, and the use of content such as blogs to directly address the challenges of reading about distressing topics online [40]. To date, however, it is unclear to what extent these strategies are valued by users or effective in achieving their intended aims. Further research focusing on how services can support effective self-moderation of online communities is therefore warranted.

This paper emphasizes that moderators’ actions often underpin safe and effective online mental health forums. Existing research makes clear that intervening to support users and implement rules are highly skilled and demanding tasks. For example, the online context requires moderators to engage in a range of nuanced text-based practices to build rapport, understand users’ challenges accurately, and offer corresponding social support, including in situations where users are experiencing high levels of distress [41]. However, access to a large volume of distressing content, pressure to manage interpersonally sensitive and high-risk situations, and issues such as high workload, represent challenges to moderators’ workplace well-being [42,43]. Forum safety is therefore likely to be enhanced by sufficient support for moderators. This may include appropriate initial and ongoing training in the mental health difficulties likely to be discussed on the forum, support from other moderators, including during shifts, to share challenges and strategies for managing difficult situations, and supervision from senior colleagues [19].

Online forums rely on a regularly updated stream of user-generated content, yet only a minority of those viewing forums post to them [44]. It is therefore vital to understand factors that may promote engagement with online conversations. Results here suggest that positive perceptions of the interpersonal safety of online forums may promote engagement and that this relies on a community embodying nonjudgment and being populated by people with similar mental health experiences whom forum users may intuitively trust with their own narratives. As previously noted, the concept of psychological safety, typically defined as a culture in which people feel able to take interpersonal risks, has parallels with this aspect of forum user experience. Consistent with the findings reported here, trust has been identified as an antecedent to psychological safety in online communities [20]. While largely conducted within in-person and nonclinical groups, psychological safety research has demonstrated that its presence is associated with knowledge sharing and “speaking up” behaviors, and its absence with “silencing behaviors” rooted in fear of negative social consequences [45]. Given that the psychological safety construct has been fruitfully applied in occupational research to identify individual and group-level factors linked to a range of important outcomes, including learning, well-being, and workplace performance, it may be advantageous for future research to apply this conceptual lens in developing a more thorough understanding of forum safety.

This study has several limitations. First, the sampling strategy recruited current users of online mental health forums. This study therefore does not capture the perspectives of those who no longer use these services, some of whom may not do so because of fears over safety, and as such, this sample may provide more insight into safety-promoting measures rather than actions forums hosts should avoid. Second, participants were recruited from UK-based online mental health services focused on particular mental health challenges, potentially limiting the transferability of our findings to other forum contexts. A third related issue is that while all of the experiences described in the IPT framework were probed as part of the realist interviews, some, such as the negative impacts of anonymity and the normalization of harmful behavior, did not emerge as prominent issues in interviews with the participants in this study and were therefore not reflected in the refined program theories. This may be because forums used for recruitment to this study were actively moderated services, limiting exposure to these experiences.

Understanding how to deliver safe online mental health support is a topic of growing international significance. The findings of this realist evaluation highlight how users’ perceptions of safety depend both on features of forum design—such as anonymity and content moderation—but also on the interpersonal interactions that emerge from their implementation. Delivering safe and effective online mental health forums therefore depends on understanding how, in practice, steps taken to reduce harm and promote engagement are experienced, highlighting the need for ongoing research in this domain.

A lived experience commentary by an expert is presented in Textbox 1.