It is estimated that approximately 280 to 320 million people worldwide are coping with depression [1-3]. However, the availability of evidence-based mental health care, such as psychotherapy and psychopharmacology [4], is insufficient owing to high costs and a lack of skilled clinicians [5]. Moreover, persistent symptoms and high recurrence rates reported underline the chronic nature of the illness [6]. The fact that many individuals live long-term with (recurrent and persistent) depression emphasizes the need for recovery-oriented services that focus on emotional support and resilience rather than on symptom reduction [7]. Peer support interventions (PSIs) could provide such a source of support on the longer road to recovery [8], complementing professional treatment [9,10] for depression [11-13]. In particular, online PSIs meet the need for accessible and low-cost interventions [8], and offering the possibility of anonymous engagement helps circumvent the stigmatization associated with depression [14].

Principally, peer support entails giving and receiving help by exchanging personal experiences [15], where the central themes are “respect, shared responsibility, and mutual agreement of what is helpful” [16]. However, owing to the great variety of intervention types, deployment across different user groups and service delivery settings, there are multiple definitions of peer support [17,18]. Considering this heterogeneity, it is difficult to systematically disentangle the principal benefits of these systems.

We recently conducted a comprehensive meta-analysis of 28 randomized controlled trials (RCTs) to assess the efficacy of PSIs across a wide range of mental disorders and intervention types. Compared with control conditions, the PSIs we reviewed were associated with modest but significant positive effects on clinical symptoms and personal recovery (eg, promoting hope [7]) in individuals with mental illness [19]. Specifically, for individuals with serious mental illnesses, including major depressive disorders, peer support was associated with superior outcomes across clinical, personal, and functional recovery variables (eg, quality of life and social support) relative to control conditions.

It needs to be noted that only a limited number of trials included in our meta-analysis focused on online PSIs for depression. Nevertheless, the results of few trials were promising. Specifically, findings of the RCT conducted by Griffiths et al [20] suggested that engaging in a moderated depression internet support group may be clinically effective (ie, reducing depressive symptoms) in the long term, with potential short-term improvements for empowerment as presented in a companion paper of Crisp et al [20,21] reporting on the same trial. In addition to this quantitative evidence from a single trial for a depression PSI, descriptive systematic reviews (ie, a narrative synthesis for the efficacy of PSIs, not including a meta-analysis that systematically assesses the results of previously conducted studies) emphasize the potential of online health-related PSIs in general [22-25] and that of those specifically addressing depression [11,26].

The results of a broad systematic review [27] may help us better understand how these positive outcomes in PSIs may develop. Winsper et al [27] identified four common processes fostering change in recovery across 309 studies on recovery-oriented interventions for mental illness: (1) providing information and skills, (2) promoting a working alliance, (3) role modeling for individual recovery, and (4) increasing choice and opportunities. These processes may best be initiated within nonstigmatized, recovery-focused contexts, such as peer support where psychosocial processes of sharing lived experiences, emotional honesty, strengths-focused social and practical support, and the helper-role are important processes for mental health recovery [28]. The results of our qualitative evaluation study for users of the online peer support community Depression Connect (DC) fit these processes (eg, sense of belonging, self-efficacy, and empowerment) [29].

It remains unclear which PSI engagement processes are associated with these changes. In particular, for online PSIs, a high level of user engagement is considered a crucial factor for recovery [24,30,31]. A systematic review of online health communities showed that several multidimensional factors are relevant when defining user engagement, such as metrics characterizing user networks (eg, the number of people a user has interacted with), content (eg, the nature of posts), and activity (eg, the number of posts and log-in times) [32]. Use of online PSIs is mainly operationalized in terms of frequency of use [32], where the dichotomy between “lurkers” (ie, passive users, generally a substantial group, whose use is mainly restricted to reading others’ posts) and “posters” (ie, active users, generally only 1% of users [33-37]) is widely used. To accurately reflect the larger group of passive users, we need a more nuanced characterization of their engagement [32,33]. Such parameters of inactive engagement are particularly relevant for PSIs for depression, as passive behavior is associated with the condition [38]. Therefore, we conducted a qualitative evaluation of DC users [29], our self-developed online peer support community for individuals struggling with depression [39]. Our qualitative analysis of user experiences of DC revealed 3 successive participation styles (ie, reading, posting, and responding) that individually and together coincided with an increased sense of belonging, emotional growth, self-efficacy, and empowerment [29]. In this quantitative evaluation of DC, we studied engagement patterns as a possible mechanism for recovery more closely by assessing multiple metrics to define engagement as comprehensively as possible [32]. For this study, based on user data logs for DC, we included 3 parameters to operationalize the intensity level of user engagement: the number of posts, the number of page views, and the total time spent on DC. Including both active and passive user modes, it is important to acknowledge that high user engagement was not limited to active users but included users who posted (a substantial number of) messages on the platform. In user data logs for DC, it was not possible to distinguish the 2 active participation styles, posting and responding, that is analyzed in our qualitative evaluation. To include passively engaged users in our sample, we assessed the number of page views and the total time spent on DC per user. However, because users may have been active through sharing posts when viewing various pages on the platform during their time spent on DC, these parameters included, but were not limited to, the passive user mode of reading. Together, our operationalization of user engagement implied that both active users who posted and passive users who spent considerable time on DC and viewed many pages could be categorized as highly engaged users.

In recent years, peer support studies have frequently reported on personal recovery to complement clinical recovery outcomes, with a particular focus on the benefits of online PSIs for empowerment as an important feature in the process of personal recovery that individuals can develop to enable them to live a meaningful life [7,40,41]. Although inconclusive, these findings were promising [21,42-45]. Within online communities, empowerment refers to enabling processes including “becoming better informed, receiving and giving emotional support by sharing relatable experiences of living with the diagnosis, helping others, and networking” [46]. Developing and exploiting self-management strategies can be seen as an active component of empowerment [47,48], and many strategies comprise individual skills “to monitor one’s condition and to affect the cognitive, behavioral, and emotional responses necessary to maintain a satisfactory quality of life” [49]. However, to date, self-management has not been systematically examined as an individual outcome in peer support studies [50]. The same holds true for general well-being (ie, functional recovery, including social functioning and quality of life) [12,42-45,51-53], although both are important parameters for determining the usefulness of recovery-oriented PSIs. Our meta-analysis [19] showed that PSIs may also be effective in terms of clinical recovery (ie, symptom reduction), particularly for individuals with serious mental illness, including major depressive disorder. We also examined whether our online peer support platform helps improve depressive symptoms.

In this longitudinal user survey, we attempted to add to the current literature on online peer support in several ways. First, we examined personal (ie, empowerment and self-management), functional (ie, well-being), and clinical (depressive symptoms) recovery parameters among DC users. Furthermore, we comprehensively explored patterns of user engagement, including parameters reflecting both the intensity and nature of DC use. Following the results of a systematic review of user engagement [32], we clustered user engagement based on the number of posts and page views and total time spent on DC. We aimed to determine whether these user engagement profiles during a 6-month interval were related to changes in empowerment. We expected that user engagement with high frequency—including active (ie, number of posts) and passive (ie, number of page views and total time spent)—would contribute to more improvement in empowerment from baseline to 3 and 6 months compared with user engagement with low frequency. In addition, we explored the use of self-management strategies and changes in depressive symptoms, levels of functioning, and disability over time.

For this longitudinal study, users of our online peer support community for depression, DC, completed an online survey at 3 time points between June 19, 2019, and September 24, 2020.

Launched on June 19, 2019, DC was cocreated with experiential experts, caregivers, and health professionals (therapists, psychiatrists, and psychology researchers) affiliated with the Dutch patient association for depression (The Depression Association), the Centre of Expertise for Depression as part of the Pro Persona Institute for mental health care, and the Radboud University Medical Centre. DC was developed to facilitate the exchange of personal experiences in coping with depression among peers. The online platform was easily and (if preferred) anonymously accessible to anyone dealing with depression. Potential users were not screened for depressive symptoms or other clinical characteristics before they could enter the community. Although no professionals were involved in DC, its moderators, who were all experiential experts, were able to consult a psychiatrist and psychology researchers of our team when feedback was needed. To ensure a constructive and supportive online atmosphere, DC moderators screened all new posts daily. They also generated new content or boosted user activity on the platform, for instance, by posting news items or different viewpoints on coping strategies. The DC team welcomed an average of 90 new members each month. DC members could start a new discussion topic or join an existing topic created by other users or provided by the research team. At the DC’s launch, we created 8 forum topics that referred to the main themes of experiential knowledge in depression, which we identified in our qualitative interview study [29]. Widely used topics included coping with symptoms of depression (eg, concentration problems) and treatment options for depression (eg, medication and mental health care). In addition to reading and posting messages on the forum section of the community, users could read news items (posted by the DC team) and read or post blogs. There was also a function for sending private messages to other DC users. More details about the DC’s development, functionalities, and monitoring procedures are presented in our parallel qualitative evaluation of DC [29].

All individuals who registered with DC were invited to participate in our study when the website was launched. There were no strict conditions to participate in the study regarding demographic and clinical characteristics, either for the minimum or maximum level of engagement with DC. All new DC members, and thus potential study participants, received an email to welcome them to the community, including information about our quantitative evaluation study and a link to the survey. An email address of the research team was also displayed to provide users the opportunity to ask questions about study participation. Participation was on a voluntary basis without any financial or other compensation. Interested users were invited to complete the online survey 1 or 2 days after registering and at 3 and 6 months after joining DC. Of the 1374 new members who joined DC during the recruitment period, 317 (23.07%) users completed the baseline survey. Subsequently, 5% (16/317) of participants deleted their accounts, including their user data logs, leaving 21.91% (301/1374) of participants for the final sample. The data sets of participants who completed only the baseline assessment (179/301, 59.5%) were not included in the outcome analyses.

Participants completed the following measures at baseline and at 3 and 6 months after joining DC.

After evaluation, the local ethics committee (Commissie Mensgebonden Onderzoek Arnhem-Nijmegen) determined that no ethics approval was required, given the minimal burden to the study participants. The users provided passive consent to log and analyze their user data.

Of the 301 DC users who had provided their consent and completed the baseline measurement, 179 (59.5%) individuals did not complete the survey at the 3- and 6-month time points. In total, 15.9% (48/301) of DC users completed the 3-month and 24.5% (74/301) of users the 6-month survey. There were no missing data for the 4 main outcome measures at any of the 3 time points. For age and current depression, we noted 2 and 6 missing variables, respectively. A total of 496 observations from 301 participants were included in the mixed modeling analyses.

The participants’ demographic and clinical characteristics as well as the means and SDs for the outcome variables at baseline are shown in Table 1. Our sample of 301 DC users included individuals with self-reported depression and a mean age of 50.2 (SD 13.12) years, 66.1% (199/301) of them were female and most of the respondents (216/301, 71.8%) had completed some form of secondary education or training. More than half of the respondents (166/301, 55.1%) reported having severe depressive symptoms (mean BDI score of 38.7, SD 6.57) and almost one-quarter of the population (72/301, 23.9%) had moderate to severe symptoms (mean BDI score of 23.8, SD 2.53). Of the remaining respondents (63/301, 20.9%), 13% (8/63) reported mild symptoms, and 8% (3/63) had minimal symptoms. The overall mean baseline BDI score for the entire sample was 29.84 (SD 11.85). Most DC users (241/301, 80.1%) received current treatment or some form of support or care from a mental health service and 69.8% (210/301) reported current use of antidepressants.

Completers, that is those respondents that had completed the baseline and at least one second assessment, were on average 4.95 years older (SD 11.4; 2-tailed t₂₉₇=3.25; P=.001) and reported significantly higher levels of empowerment, self-management, and less severe depressive symptoms and disability in major life domains compared with DC users who had only completed the baseline assessment.

Our cluster analysis of the forum uses parameters as described in the Methods section and resulted in 4 user engagement profiles: very low (profile 1), low (profile 2), medium (profile 3), and high (profile 4). The user parameters for the total study period (6 months) are listed in Table 2. Baseline engagement profiles did not significantly differ for age; sex; current depression; current treatment or medication; or baseline scores on empowerment, self-management, depressive symptoms, and disability. However, results did show significant differences between participants completing the baseline assessment only and participants that completed one or 2 assessments for the “very low” engagement profile (177/301, 58.8%), of which 33.9% (60/177) participated in a second assessment and 66.1% (117/177) completed baseline only (χ²₃=27.1; P<.001; Multimedia Appendix 1).

Figure 1 depicts the (changes in) outcomes and session duration for the participants who completed only the baseline assessment (panel 1, n=179) and for those who completed at least 2 or all 3 assessments (panel 2, n=122). The graphs present data modeled using a longitudinal mixed model regression analysis for session duration for each individual session (black dots), empowerment (NEL), self-management (ASAD), depressive symptoms (BDI-II), and functioning and disability (WHODAS 2.0) over time, including the baseline means for each outcome. They show an increase in empowerment and self-management and a decrease in depressive symptoms and disability over time (days of engagement with the online peer support community—DC).

In the figure, the scale values were transformed to a range from 0 to 4. To avoid overlap, self-management was raised by 0.4 and functioning and disability was lowered by 0.25. The session duration is transformed to the square root of the duration in minutes and divided by 5, so that most values are in the figure. These lines are based on the values predicted by the model. Values at day 0 are actual means with the confidence levels.

Table 3 lists the results of the models that investigated changes in outcomes over time (days of DC use). We computed a significant increase in empowerment (NEL) over time in days (β=.00078; SE 0.00022; P=.001) with a small effect size (Cohen d=0.36, 95% CI 0.15-0.57). Self-management also increased over time (β=.0222; SE 0.011; P=.046), again with a small effect size (Cohen d=0.22, 95% CI 0-0.43). Depressive symptoms (BDI) significantly decreased over time with a small effect size (β=−0.0244, SE 0.00612; P<.001; Cohen d=0.44, 95% CI 0.21-0.66). In addition, disability (WHODAS 2.0) significantly decreased over time with a small effect size (β=−0.0212, SE 0.00693; P=.001; Cohen d=0.29, 95% CI 0.10-0.47).

Engagement profiles were not significantly associated with changes in any of the outcomes at 3 or 6 months as indicated by the nonsignificant effects of the dependent variable by engagement profile: empowerment (F_3,176=0.07; P=.98); self-management (F_3,169=0.1; P=.96); depressive symptoms (F_3,184=0.14; P=.94); and functioning and disability (F_3,181=0.2; P=.90); and nonsignificant time by profile interactions: (empowerment (F_3,131=0.55; P=.65); self-management (F_3,126=.93; P=.43); depressive symptoms (F_3,140=0.09; P=.97) and functioning and disability (F_3,158=0.09; P=.96).

Although the potential benefits of engaging peer support for people with severe mental illness (SMI) are widely acknowledged [25,26], peer support studies are limited for online intervention types targeting depression [20,21] and the processes for user engagement remain unclear [32,70]. In this longitudinal user survey of the online peer support community—DC, we explored patterns of user engagement and examined whether user profiles were associated with recovery-oriented outcomes. To quantify baseline to 6-month changes in empowerment, self-management, depressive symptoms, and functioning and disability in the users of DC and considering the complex interplay of relevant aspects of user engagement in PSIs, we entered the user data logs of 3 parameters (ie, total session duration, page views, and number of posts) into a cluster analysis, resulting in 4 engagement profiles. Most of the survey respondents (177/301, 58.8%) had very low or low engagement levels (87/301, 28.9%), with 9.9% (30/301) having medium and 2.3% (7/301) high user profiles. However, none of the profiles showed significant differences for age; sex; having current depression; receiving treatment at the time of assessment; or with regard to the baseline scores for empowerment, self-management, depression, and functioning and disability. All recovery-oriented outcomes improved over time; however, contrary to our hypothesis, the nature and intensity of DC user engagement were not significantly associated with any of these improvements.

The number of user surveys and RCTs for online PSIs for depression is limited; however, the results are promising. Although our results did not show a significant relationship between the level of user engagement and recovery, Griffiths et al [20,21] reported positive results for engaging a online PSI for depression in their trial. They found that depressive symptoms reduced in the long-term period (6 and 12 months) and empowerment may improve in the short-term (after the intervention or at 3 months). Furthermore, reviews with and without meta-analyses of PSIs that include a heterogeneous population, which includes primarily individuals with SMI, report positive changes in psychosocial outcomes [24,25,43,45], more specifically for self-efficacy and hope [12,42,44,51-53]. We confirmed this in our new and updated meta-analysis, which included PSIs for mental illness [19]. However, research on peer support is associated with methodological issues (eg, establishing model fidelity is not possible at this point [25]). Therefore, the results of this longitudinal user survey as well as the results from the other PSI studies should be interpreted with caution.

Considering the level of user engagement, it is generally known that online communities [37,70] are associated with low engagement rates. This is often referred to as the 1% rule [36,37] and is in line with our results. However, our study adds to the current literature, as the results improve insight into the intensity and nature of user engagement for online PSIs for depression. Further research is needed to better understand the relationship between levels of user engagement and positive changes in recovery.

Finally, other recovery-oriented outcomes may be more relevant for evaluating an unstructured online peer support community such as DC. Empowerment and self-management may serve as those attributes that would characterize more advanced stages of recovery from mental illness, such as depression, as they take time to develop and generally require guidance from a nonpeer (ie, a paraprofessional) [18,46], face-to-face PSI format [42,43,77], or a wider supportive context involving family or friends [78,79]. Moreover, considering the informal nature and flexibility, free use of our platform, and the fact that our sample mostly consisted of individuals with moderate to severe depressive symptoms, smaller goals such as an increased sense of being (emotionally) supported or finding new hope are probably more feasible [80].

This study has several limitations. The first one lies in the operationalization of user engagement. Rather than opting for (more frequently used) self-report measures, we tried to objectively quantify forum engagement using logged user data (total time spent on the platform and the number of page views and posts); however, there are other potentially relevant indicators of engagement, such as the number of posts the user reads [34], the length of threads [74], and the number of replies received [74,81]. Particularly for individuals with depression, these activities and interactions that reflect recognition and support may reduce stress and negative emotions [82]. Unfortunately, we were unable to extract these parameters from the user data logs. Second, the results of our previous qualitative exploration of DC user experiences were primarily related to users with an active engagement style. In this quantitative study, the number of highly frequent and actively engaged users—those posting significantly more than their peers with other engagement profiles—was too small to detect any reliable effects on empowerment. Third, the lack of a comparison group in this longitudinal user survey precluded exploration of causal relationships between DC use and recovery; however, the effect sizes of RCTs comparing PSIs for mental illness with a control group that we pooled in our meta-analysis were significant both for clinical and personal recovery indices.

Finally, the generalizability of our findings is limited as we evaluated self-selected samples, where the decision to participate may contain some inherent positive bias toward engaging in (online) peer support. It is possible that users with a low engagement profile were not motivated to complete the follow-up assessments in our evaluation study because they lacked commitment to DC or may not have experienced any benefits of engagement. However, our study has an explorative character, with a naturalistic sample that informed us of the general and heterogeneous population of individuals with depression who engage in peer support. Given the observational character of our study, its internal validity is limited. We do not know whether the improved outcomes are related to DC use and to what extent other types of support or the many other variables that are part of the real-world setting (eg, the level of offline social support, self-stigma, and societal participation) influence these results. Regardless, considering the free and informal nature of our online peer-to-peer support environment that allows users to tune their use of the forum to their personal needs and the improvements observed, our survey expands the current literature by focusing on an online PSI for depression. The results underscore that this type of peer support appears to be beneficial and promotes recovery among individuals with self-reported depressive symptoms. These promising results are not only reflected in our survey but also in previously conducted user surveys in PSIs, underscoring the benefits of peer support for clinical [14] and personal recovery [33].

As the various engagement profiles that we identified indicate that DC users appear to prefer a flexible use of the platform, insight into the content of their posts would foster the interpretation of our findings. Therefore, we recommend assessing the perceived quality of interactions (eg, “Is the content helping you to cope with your depression?”) in future research on online PSIs. The quantitative variables such as thread length and the number of posts and responses or comments might indicate how effectively a topic was explored [74]. Synthesizing qualitative data (content analysis) and quantitative data (metrics of use) of peer support user engagement would enhance our understanding of its implications for recovery. In addition to the clinical characteristics and (treatment) history of depression, it may be informative to describe the societal context of individual users. Possibly, the availability and quality of social support from family or friends may predict users’ need for online peer support and explain their low or high engagement. As peer support is considered adjunctive to formal mental health care [9] and it has been suggested that peer support encourages users to engage more actively in their professional treatment [14], it is worthwhile to investigate the usefulness and benefits of (online) peer support for concurrent professional therapy. Finally, recovery is a multidimensional concept; however, the various factors and processes involved are difficult to disentangle. Including comprehensive measurements in which the umbrella concepts of clinical, personal, and functional recovery-related indices are assessed separately and in depth, such as in the case of the Recovery Assessment Scale, might improve the validity of the findings.

This longitudinal user survey provides insights into the characteristics of user engagement in DC, an online peer support community for depression. Active engagement was limited to a small group of DC users and was not significantly associated with superior improvements in empowerment and secondary recovery-oriented outcomes. Users appear to attune the intensity and nature of their forum use to their personal recovery pathway and current needs, where their engagement levels may shift from low to high and from passive to active. Corresponding to the self-determination theory, the autonomy to choose the level of engagement might be one of the most valued and effective features of intervention types, such as DC, whereas in other more formal supportive environments for depression, a certain level of engagement is predetermined. Future online PSI studies should explore the content and quality of user interactions to determine what constitutes optimal user engagement, where flexibility and usefulness match users’ clinical needs and motives to seek and offer online peer support.