Enforced lockdowns and physical distancing measures introduced to slow the COVID-19 infection rate have resulted in disrupted face-to-face connections. Ordinarily, a lack of meaningful social connections through social isolation is associated with poor health outcomes such as sleep problems, loneliness, depression, and anxiety, leading in some cases to suicide. Young people are particularly vulnerable to mental health difficulties such as depression and anxiety because onset usually occurs before the age of 24 years [1], and they are often comorbid globally [2]. Although disrupted social connections and loneliness can have a negative effect on mental health [3,4], feeling socially connected is one of the strongest protective factors for depression [5] and can decrease symptoms of anxiety [6].

Social connection as a concept is multifaceted. It can be described as the quantity of connections, the opposite of loneliness, or as having social support. In our context, social connection is the perceived value of attributes of a meaningful interaction among 2 or more people or a quality social connection (QSC). Such valued attributes can include, for example, feeling listened to, understood, and a sense of belonging. Similarly, a cooperative relationship between client and therapist, comprising a close bond, shared goals, and tasks, is defined as a therapeutic alliance in face-to-face therapy [7]. A therapeutic alliance has been shown to significantly modulate treatment outcomes [8], including in digital settings [9]. Similarly, social prescribing to improve social connection has decreased loneliness and improved health outcomes [10]. However, studies have only subjectively demonstrated the value of strong social networks and social relationships for both physical and mental health [11] and longevity [12]. This suggests a need for well-defined indicators of social connection to enable objective measurement of these effects.

The COVID-19 pandemic has accelerated a rapid shift to digital provision of formal and informal mental health support [13]. Indeed, mental health care is often seen as the best candidate for a digital revolution because prevention and treatment, including talking therapies, are amenable to delivery over screens and remotely [14]. Social media, video consultations, texting, and virtual reality are interventions that can enable social connections [13]. They represent an important intervention for young people with mental health difficulties to strengthen new and existing relationships and facilitate peer-to-peer and formal mental health support [15]. However, digital interventions such as social media use are also associated with negative consequences such as cyberbullying, viewing harmful content, and a greater sense of isolation [16]. This contradiction requires further investigation to identify the ways in which digital interventions may help or hinder QSC.

Young people are the most digitally fluent and most in need of mental health support. However, QSC within digital interventions (D-QSC) has received little attention in relation to outcomes for depression and anxiety in young people. A recent review produced a conceptual framework for social connectedness but positioned it as a solution to loneliness and not as an active ingredient (best bet) for the prevention and treatment of depression and anxiety [17]. It also did not consider digital interventions or young people. A systematic review is needed to help produce a conceptual framework for indicators of D-QSC that can be applied to examine its influence on depression and anxiety outcomes across contexts. Our study aims are to (1) identify indicators of D-QSC and their ability to improve or worsen outcomes for depression and anxiety in young people across contexts, (2) identify digital intervention mechanisms that facilitate QSC, and (3) produce a conceptual framework for indicators of D-QSC.

The research questions are:

We advertised for young people aged 14-24 years with experience of depression or anxiety and digital interventions for mental health to work on a review about social connection in the digital world through The McPin Foundation newsletter, email distribution lists, Twitter, and Instagram. A total of 9 people applied using a simple form, and all joined the Young Persons Advisory Group (YPAG). They represented different genders, ethnicities, ages, and UK locations. We held an initial web-based workshop to help define QSC and D-QSC, inform search terms, and review the protocol. At this stage, we approached the Lancet Commission for Global Mental Health Young Leaders and experts with professional experience (ie, delivery of digital interventions) to ensure a diverse range of experiences, cultural contexts (ie, low- and high-resource settings), and experience of youth interventions for depression or anxiety. We had separate discussions with each group on the web (eg, Zoom). Subsequent changes were made to our definition of D-QSC, selection criteria, and protocol. On the basis of definitions of quality [63] and therapeutic alliance [7] and input from team members (EL, LD, and LR) and experts, QSC was then operationally defined as the perceived value of the attributes of an interaction between two or more people. Key attributes (ie, indicators) of D-QSC that made up the definition were logged across the 3 expert groups and amalgamated with the literature indicators as described previously. Others were changed (eg, changed to plain English) or merged after the YPAG and the Lancet Commission for Global Mental Health Young Leaders were shown the findings and conceptualization framework. A YPAG member (EBH) also screened, extracted, and quality-assessed literature. All were given appropriate support and paid in line with guidance [64].

A total of 5715 records were identified (Figure 1). Of the 5715 articles, 1833 (32.07%) were duplicates and were removed. Substantial agreements were achieved in the screening of the random 9.99% (388/3882) abstracts and titles as well as full-text subsamples (K=0.80 and K=0.70, respectively). Papers were then excluded if they did not match the selection criteria; of the 3882 publications remaining after duplicates were removed, 42 (1.08%) were included (Figure 1). Of these 42 studies, 28 (67%) were of high quality, 13 (31%) were of medium quality, and 1 (2%) was of low quality. High-quality studies largely demonstrated good explanation of aims, methods, and sampling to enable replicability. Medium-quality studies included most of the good study criteria but were lacking in some areas, which reduced their scores. The low-quality study did not provide enough detail across most domains (eg, it did not report ethical considerations, the results were unclear, and the methods were not replicable).

Of the 42 included studies, 25 (60%) were quantitative [18-42], 11 (26%) were qualitative [43-53], and 6 (14%) were mixed methods studies [54-59] (Table 2; Multimedia Appendix 3 [18-59]). The studies mainly used uncontrolled or cross-sectional designs and had questionnaires as their main data collection method (22/42, 52%). All studies took place in high-income countries: the United States [23-28,36,38,40,42,44,47,49,50,53], Australia [22,29-31,43,46,55-58], Ireland [18,45,54], Israel [19,32,48], Taiwan [20,35], Sweden [21,41], the Netherlands [33], Turkey [39], Austria [34], Belgium [37], Canada [59], Cyprus [52], and the United Kingdom [51]. There were 23,319 participants in total (12,825/23,319, 55%, were women; of the 42 studies, 2 (5%) did not report participant gender). Of the 42 studies, 21 (50%) focused on both prevention and treatment [18,20,21,23-29,36-39,41-44,48,50,57], 12 (29%) focused on treatment [22,30,31,45,49,51-56,58], and 9 (21%) focused on prevention [19,32-35,40,46,47,59]. Digital mental health–related intervention types included mental health social networking tools, smartphone apps, self-help cognitive behavioral therapy, telepsychiatry, one-to-one peer mentor support, video gaming, avatars, and internet use for mental health support. Nonspecific informal digital interventions included general social networking and social media (eg, Facebook, Twitter, Tumblr, Snapchat, and Reddit) and general internet use and web browsing. Intervention duration was reported in 48% (20/42) of the studies and ranged from 8 weeks to 1 year.

Indicators and measures used to quantitatively assess D-QSC presence were heterogeneous (Tables 3 and 4). The most common indicator for D-QSC was social support (14/42, 33%; Table 3). Among the 31 quantitative studies assessed, there were 20 different standardized questionnaires used to measure QSC (Tables 3 and 4), with only 4 (13%) studies using the same measure (Multidimensional Scale of Perceived Social Support). Nonstandardized questionnaires were also used in some studies, including single questions (eg, “I hope to gain support through meeting people going through similar experiences, Y/N”). The remaining indicators of QSC were identified from qualitative analysis within 17 studies [43-59].

The relationship between D-QSC indicators and outcomes was mixed. Of the 42 studies, 10 (24%) reported a change in depression symptoms over time after participants experienced a D-QSC, and a pooled analysis demonstrated a significant weighted mean decrease in depression by 25.6% (–0.256, 95% CI –0.352 to –0.160; P<.001), with high heterogeneity (I²= 90.8%; Figure 2) [20,23,24,26,30,36,37,39,41,42]. Of the 42 studies, 5 (12%) reported change over time in anxiety symptoms; there was also a decrease, but it was smaller (15%; –0.151, 95% CI –0.251 to –0.051; P=.003), with high heterogeneity (I²= 83.1%; Figure 3) [29,31,56-58].

The indicators of D-QSC associated with improved depression or anxiety symptoms included social support [23,24,30,36,37,39,41,42,56], social connectedness [22,28,29,31,52,57,58], loneliness (reduced) [30,31,40,54,59], relatedness [56], sense of belonging [22,31], being able to share [21,52,54], less alone in one’s feelings [25,54], feeling normalized [52,59], feeling close to peer [57,59], feeling you are not a burden [22,31], feeling accepted [29,59], shared understanding [54], and trust [59] (Table 3). For example, depression outcomes improved after good social support for those abused on the web [36] and for adolescents with high social anxiety [21]. In contrast, negative interactions [20,26], negative experiences of social support [49,51,52], and frequent social media use [27] were associated with worsened outcomes (Table 4). For example, young people aged 17-19 years using social media (eg, Facebook, Instagram, and Twitter) to connect with others were more likely to have anxiety but not depression at 3 years’ follow-up [27]. A similar but older study found a worsening of depression outcomes with social networking interactions (ie, Facebook, Myspace, and texting) [26]. This relationship between negative interactions on social media and worsened outcomes was evident in both men and women, particularly in those also receiving low face-to-face social support [20]. Additional indicators of D-QSC that did not explicitly indicate effect on depression or anxiety were feeling validated [47], rapport [45], empathy [27,34], and emotional connection [48,50]. The indicators also improved well-being outcomes (Multimedia Appendix 4 [19,30-33,35,48,49,53]).

Digital intervention mechanisms mainly helped facilitate QSC [18,20,25,30,35,43,44,46,50,51,55,56,58,59]. Forum moderation, confidentiality, ease of access, and anonymity supported by digital interventions were cited as valuable [18,43,50,56,58,59], facilitated open sharing in digital environments, and could lower inhibitions compared with face-to-face engagement [50,51]. The usual signals received during face-to-face interactions, such as body language or facial expression, were lost during digital interactions [52] and could impair the quality of interactions [26].

D-QSC was deemed more valuable when digital interactions were blended with face-to-face interactions [20,35,43]. For example, digital interactions were convenient and accessible, whereas face-to-face meetings helped maintain the connection. Higher levels of web-based social support were associated with increased symptoms of depression, specifically in both men and women and those who had little in-person social support [20]. One study found that participants were disconnected from the real world through high levels of web-based engagement [43]. Indeed, disconnection can have an impact on the interaction between peers and family and result in increased loneliness [19,32]. Other studies indicated that participants felt ignored [51], misunderstood [51], and had hurt feelings [43,49].

Participants also valued opportunities to support others [41,55], to connect with peers, and compare similar mental health experiences [25,54]. Some participants considered the networking component as the most helpful aspect of a moderated web-based social therapy tool, more helpful than the therapy itself [46]. Harassment was also identified as occurring frequently on the web. For example, a study reported this frequently among transgender adolescents [32].

To our knowledge, this is the first systematic scoping review with meta-analysis to examine D-QSC as an active ingredient for depression and anxiety outcomes in young people. Usually conflated with quantity of social connections, QSC has now been comprehensively examined for its relevance to the mental health outcomes of digital interventions. We coproduced a conceptual framework of D-QSC for young people experiencing depression or anxiety that summarizes current understanding of component attributes or indicators. The RIVER framework comprises indicators relevant for establishing and assessing the presence of D-QSC. This can be characterized across 5 components. These components are interconnected and may not be exhaustive, but they provide a foundation for further work in this field to establish appropriate metrics for D-QSC.

D-QSC seems to help improve depression outcomes across most digital interventions. However, there is weaker evidence that D-QSC improves anxiety and well-being. There was also limited evidence of gaming, which was surprising considering that the participative nature with other users is at its core. D-QSC also worsened depression and anxiety outcomes in some instances, but this was often a result of negative interactions through social networking sites, which could be construed as a poor D-QSC. Few studies examined individual factors, contextual factors, or digital mechanisms that may modulate the impact of digital QSC on mental health outcomes. However, in the few studies that did report on mechanisms, a face-to-face connection before web-based support was an important consideration for improving outcomes. Furthermore, the impact of web-based support was modulated by the strength of offline connections.

Reviews assessing the efficacy of digital mental health interventions for young people have found digital interventions to be as efficacious as, or sometimes more efficacious than, similar interventions delivered in person [65]. The strongest review to date that most closely relates to QSC collated measures of social connectedness to produce a conceptual framework of social connectedness in mental disorders (CIVIC) [17]. However, the review positioned social connectedness as the solution to loneliness and not as an active ingredient for the prevention and treatment of mental disorders (ie, depression and anxiety). Our work extended this framework to ensure that QSC was considered for digital interventions, for young people, and across different contexts. This process substantially expanded the elements indicating the development or presence of D-QSC and required redefining the CIVIC framework components to form the RIVER framework of D-QSC indicators for young people. Interestingly, the component of the CIVIC framework found to be most frequently assessed in current QSC metrics was Identity and common bond; this was the only component of our adapted framework not selected as a top priority in the context of digital interventions for anxiety and depression during review of the framework by 9 young people. This highlighted the need to develop improved metrics that are uniformly applied.

Young people with lived experience were involved at all review stages, including screening and interpretation. Dual independent review of the literature with people with direct experience of the review area helped us to overcome some limitations inherent in the current literature to gain better understanding of QSC and ensure accuracy of the screening. The insight from the young people during the data synthesis and interpretation stages helped to retain the data integrity. We have subsequently added to the limited evidence base for the impact of patient and public involvement throughout all stages of systematic and scoping reviews. Our adapted RIVER framework provides the foundation for future work to develop measures that would enable a developer, evaluator, or practitioner using digital interventions for mental health to assess the presence and degree of the QSC established.

The main limitation was that the studies did not control for a previous established connection offline before the D-QSC. Of the 42 studies, only 12 (29%) could be included in the meta-analysis because of a lack of measured effect sizes in previous work and heterogeneity across approaches, suggesting that the results should be interpreted with caution. Because of the scoping nature of the review, there was also statistical and methodological variability in the meta-analysis. Only manuscripts that were in English were included, which enables cultural bias. However, this was mitigated to an extent by working with young people and professionals from a variety of countries and cultural contexts to interpret the findings.

QSC should be considered in the development and application of most digital interventions, particularly for depression. However, more research is needed to examine its impact within gaming platforms. In general, digital interventions mostly helped facilitate QSC; therefore, developers should consider factors such as user preference, anonymity, delivery medium, and content moderation. Initially, they should consider whether D-QSC is appropriate, depending on the target audience, and whether it will be important for engagement, or efficacy, or both. Further research is required to establish which individuals, conditions, and therapeutic mechanisms respond most strongly to D-QSC and what format is most appropriate. Clinical trials of any new digital intervention for mental health should control for previous face-to-face connections.

Future research should build upon our RIVER framework to further examine relationships among individual indicators of QSC, variations across different digital interventions, and the impact on outcomes across different user groups, particularly those in low- and middle-income countries. Factors that may mediate any causative relationships between QSC and mental health outcomes also deserve further attention. This work will inform the creation of standardized measures for D-QSC to evaluate its presence across different social settings. New measures should be developed to assess (1) attributes of a digital intervention that help or hinder good QSC and (2) the perceived value of a particular QSC for an individual and its relationship to outcomes within digital interventions. This work has value for development, regulation, and evaluation of digital mental health interventions, as well as delineating helpful and harmful web-based interactions for young people, including social media. It will be vital to expand digital mental health care provision during the COVID-19 pandemic.

As the COVID-19 pandemic accelerates the shift to digital delivery of traditional therapy [13], clinicians should be trained in how to incorporate techniques for developing or maintaining D-QSC. Guidelines should be developed to ensure that moving face-to-face therapies to web-based spaces does not affect the QSC in the practitioner-patient dyad, and they should include strategies to improve connection on the web. Further clinical recommendations include a prioritization of video communication for web-based therapy to allow body language to be observed. However, anonymity can be beneficial to some users when first divulging sensitive mental health information. Blended care should enable patients to first meet their therapist in person, if desired, to facilitate QSC that can be translated to digital follow-ups.

In conclusion, D-QSC is important and an underconsidered component supporting engagement and efficacy for young people with depression and anxiety. In the wake of the COVID-19 pandemic, our work holds relevance as mental health needs rise and support will increasingly be provided on the web.