Effectiveness of online social networking interventions on social isolation and quality of life of people with psychosis: A systematic review

A B S T R


Introduction
Social isolation is prevalent among people with psychosis, with studies reporting that 80 % of participants experience feelings of loneliness and social isolation (Morgan et al., 2012;Stain et al., 2012).This is a social and public health concern.Social isolation has been associated with earlier mortality (Holt-Lunstad et al., 2015;Rico-Uribe et al., 2018), increased rates of depression (Cacioppo et al., 2006;Sibitz et al., 2011), and risk of suicide (Goldsmith et al., 2002), poor mental health outcomes, and reduced quality of life (Leigh-Hunt et al., 2017;Skelly et al., 2015).
Social network size and quality appears important for patients with psychotic disorders, both as a relevant outcome criterion of interventions in its own right and as a factor influencing quality of life (QoL) outcomes (Palumbo et al., 2015).Studies have indicated that social isolation is linked to poorer QoL outcomes (Bengtsson-Tops, 2001;Leigh-Hunt et al., 2017).Additionally, research has shown that individuals with schizophrenia who have more social contacts tend to report higher subjective QoL (Becker et al., 1998;Cechnicki et al., 2008;Sibitz et al., 2011).Hence the concept of QoL is highly linked with social relationships.
Evidence suggests that Online Social Networking (OSN) has qualities that can help people with psychosis to socialise (Highton-Williamson et al., 2015).A systematic review of 11 studies concluded that the use of OSN, such as chat rooms, reduces social isolation through providing spaces to form new relationships, find peers with similar mindsets and reconnect past relationships (Highton-Williamson et al., 2015).
Their review further established that preliminary evidence suggested that people with psychosis utilise OSN at least as frequently and, in some studies, more than comparison groups from the general public (Highton-Williamson et al., 2015).While OSN such as chatrooms are predominantly synchronous, they give the users time and space to conduct and formulate their reply, so they do not make conversational mistakes.This additionally gives those with psychosis time to improve their social skills for use in face-to-face interactions.Furthermore, OSN may help people with psychosis construct support systems without the fear of being stigmatised or labelled incorrectly which would be harder to accomplish face-to-face (Highton-Williamson et al., 2015).OSN may therefore enable those with psychosis to establish social relationships more comfortably, particularly in the early and potentially stressful stages of establishing a relationship.
There are further positives for OSN use for those with psychosis.The internet offers a wide source of information, allowing access to different avenues of help (Fox and Rainie, 2000).There is also a sense of anonymity and a loss of a social hierarchy on the internet which might assist in allowing users to connect with others to whom one may have previously felt socially inferior (Schrank et al., 2010).However, while anonymity can provide various advantages, such as enabling individuals to express themselves and share more openly, it also poses several risks, including cyberbullying, harassment, and the dissemination of misinformation (Rezayi et al., 2018).
Despite the potential advantages of interventions incorporating OSN, research including people with psychosis has been limited.To the best of our knowledge, there has only been one previous review to date that has examined the effectiveness of online interventions at reducing social isolation and improving quality of life for people with psychosis (Välimäki et al., 2016).Given the interest and topicality of this area, a number of studies have recently tested interventions including OSN.
This review examined the emerging evidence regarding the potential of OSN to support psychosis treatments with the objectives of identifying promising treatment opportunities and informing future research.The aims of this study were to assess the effectiveness of internet and mobilebased interventions that incorporate OSN specifically in improving quality of life and social isolation/loneliness related outcomes in people with psychosis.While previous reviews have examined the naturalistic use of mainstream OSN sites (Highton-Williamson et al., 2015), our review will investigate whether they are used as part of a tested intervention.
The specific research questions addressed are: 1) What interventions incorporating Online Social Networking have been used to reduce social isolation of people with psychosis?2) How effective are interventions incorporating Online Social Networking in improving quality of life, social isolation, and loneliness for people with psychosis?

Search strategy
A systematic review was conducted in line with the protocol submitted to PROSPERO (CRD42022303296).Systematic bibliographic searches were performed to identify relevant English and non-English language peer-reviewed studies from the following databases: PubMed, MEDLINE, EMBASE, PsycINFO, The Cochrane Library; CINAHL+ and Scopus.The literature search was completed on 1st October 2021, and updated on 4th December 2023.
The abstracts, titles and keywords of studies were searched using combinations of the following terms: (Social media OR online social networks OR online interventions OR facebook OR twitter OR instagram OR whatsapp OR snapchat OR bumble OR friend apps OR minecraft OR telegram OR reddit OR quora OR tiktok OR OSN) AND (psychosis OR psychotic OR schizophrenia OR schiz* OR psych*) AND (support groups OR social support OR social isolation OR social phobia OR loneliness OR social isolat*).There were no restrictions on publication period or language.

Eligibility criteria and study selection
This review sought to include all published studies of interventions which included OSN for people with psychosis.OSN was defined for this study as any ability for users to share content with other platform users as part of the intervention, either on the web or via mobile phone applications.This may include the use of mainstream platforms such as that of Facebook, Reddit and Instagram.Studies were excluded if they did not provide quantitative data on quality of life outcomes or either social isolation, social support, social connectedness, social anxiety, social functioning, or loneliness as a primary or secondary outcome; if they investigated only participants' interest in and willingness to receive OSN interventions; or if they did not incorporate the ability for users to share content with other platform users.Studies published as conference abstracts only, case reports and case studies were excluded.
Studies in which interventions were offered in conjunction with faceto-face therapy or treatment as usual were also included.Studies were included if: they involved patients with a diagnosis of schizophrenia, schizoaffective disorder or a related psychotic disorder, or bipolar affective disorder according to DSM-5 or ICD-10 classification; were comparative studies, including randomised trials, comparing interventions including OSN with a control group or no intervention; were controlled before and after studies; cluster randomised trials; or quasiexperiments.Studies were included that used mixed samples of caregivers and service users, although characteristics and outcomes were reported only for the participants from the psychosis user sample.
All references identified through the search were imported into EndNote 20, where duplicates were eliminated using the 'Find Duplicates' function and through subsequent manual scrutiny.All titles and abstracts were screened for inclusion by two reviewers (E.T read 100 % and L.D. read 25 % of the titles).The full texts of the selected studies were subsequently screened for inclusion by both E.T and L.D. A third reviewer (D.G) was available for settling any discrepancies.The rate of discrepancies was minimal, with only one instance identified.A meeting was convened between the two screeners, E.T and L.D, and the adjudicator, D.G, to discuss the relative low number of discrepant screenings of one.

Data extraction
This systematic review was performed based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) 2020 guidelines (Page et al., 2021).After removing duplicated records, primary screening was conducted by assessing the titles and abstracts.
A headed table was used to guide the extraction of information from the texts.Extraction was initially conducted by the primary researcher.To minimise the probability of errors, a second reviewer peer-reviewed the data extraction of all quantitative data.
Several socio-demographic and clinical characteristics were extracted from the eligible studies including: of the intervention at improving quality of life measures; (e) reported effectiveness of the intervention at reducing social isolation and loneliness measures; and (f) variables associated with use of, and engagement with, the intervention.

Assessment of risk of bias
For controlled studies methodological quality was assessed by means of the Cochrane Collaboration 'risk of bias' tool (Higgins and Altman, 2008;RoB tool) to appraise the included studies' quality and potential bias.This was administered in accordance with the Cochrane Handbook (Higgins et al., 2019).The following five domains were considered in relation to each paper: (a) sequence generation; (b) allocation concealment; (c) blinding of participants, personnel and outcome assessors for each outcome; (d) incomplete outcome data; and (e) selective outcome reporting.
Assessing each domain involved the application of several criteria.The ratings produced by the criteria informed an algorithm which led to a risk of bias judgement for each domain at one of three levels: (1) Low risk of bias; (2) Some concern; (3) High risk of bias.The domain ratings were then used to inform the overall risk rating for each paper.
For uncontrolled studies methodological quality was assessed by means of the Downs and Black (1998) quality assessment tool.This tool contains 27 yes/no items across five domains: (a) reporting; (b) external validity; (c) internal validity -study bias; (d) internal validity -confounding; and (e) power of the study.Score ranges were given corresponding methodological quality levels for each included study (Hooper et al., 2008).In accordance with previous research 18 these were classified as excellent (26-28), good (20-25), fair (15-19), or poor (<14).
Both the primary researcher and a second researcher assessed all articles.Any disagreements were resolved by discussion to reach a consensus.

Data synthesis and analysis
All studies included in the systematic review were synthesised and summarised narratively.
In selecting a narrative synthesis approach, we considered the substantial variability in how the construct of social isolation was measured across the reviewed studies, leading to diverse outcomes.Meta-analysis was deemed impractical due to this variability in outcome measures.Additionally, narrative synthesis was chosen for its capacity to provide a nuanced exploration of the heterogeneous findings without amalgamating outcomes that differ significantly, thereby enhancing the informativeness of the review process.Furthermore, this decision aligns with established guidelines, including those outlined in the Cochrane handbook chapter 9 Section 9.1.4(Higgins et al., 2019), which caution against meta-analysis when studies are clinically diverse or at risk of bias.
Primary and secondary outcome measures related to loneliness, social isolation and quality of life are, therefore, reported (with effect sizes where possible) along with the characteristics of and usability and engagement data regarding the social networking components of the interventions.
The synthesis of the included studies OSN features followed a structured process informed by guidelines adapted from Popay et al. (2006).This synthesis occurred in two stages, aimed at systematically exploring the identified OSN features.
In the initial stage, we developed a framework of criteria to categorise and analyse the OSN features.These criteria were identified through an inductive process, wherein familiarisation with the descriptions of the OSN features led to the formulation of the criteria.This process was conducted independently by two researchers (E.T and L.D), and the framework of criteria was refined through discussions within the research team.
The second stage involved characterising each OSN feature based on the established framework of criteria.This enabled us to delve deeper into understanding the commonalities and differences between the identified OSN features.Through iterative discussions among the researchers, tables were constructed to systematically capture the essential elements of each OSN feature.Subsequently, the OSN features were categorised into distinct groups based on their characteristics.
This structured approach facilitated a comprehensive analysis of the diverse OSN features identified across the included studies, allowing for a nuanced exploration of their functionalities and implications within the context of our review.

Selection of studies
A total of 9835 studies were identified, with 54 full-text articles selected after the removal of duplicates and the application of inclusion and exclusion criteria to the abstracts.After further assessment of the full text, 11 were eligible for inclusion.The PRISMA flow diagram (Page et al., 2021) provides a visual representation of the screening process and the number of studies included/excluded at each stage (Fig. 1).

Study characteristics
Detailed characteristics of the included articles (N = 11) are provided in Table 1 below.Of the 11 included studies, five were randomised control trials (RCT's) (Alvarez-Jimenez et al., 2021;Dabit et al., 2021;Kaplan et al., 2014Kaplan et al., , 2011;;Rotondi et al., 2010Rotondi et al., , 2005) ) and six utilised a single group pre-post design (Alvarez-Jimenez et al., 2013;Biagianti et al., 2016;de Wit et al., 2015;Lal et al., 2023;Ludwig et al., 2021;McEnery et al., 2021).Mean overall age of participants in the studies varied greatly from 20.3 -47.0 years old.Six of the studies were from the United States, three from Australia, one from Canada and one from the Netherlands.However, although the Dabit et al. (2021) study was conducted in the USA their participants were recruited through the Internet.Participants were primarily located in the United States and Canada with three participants in their study from France, India and the UK.

Characteristics of online social networking features
This review found that all studies including OSN used some form of online peer-to-peer interaction developed specifically for the intervention.Nine of the eleven studies had platforms featuring moderators.Research staff were only available for technical help in the Kaplan et al. study (Kaplan et al., 2011) with both the Listserv and the online bulletin board entirely peer directed and not facilitated by staff.Nine of the eleven studies also featured online group sessions of some format.Additionally, it's noteworthy that just one study (Ludwig et al., 2021) reported to have offered an in-person component as part of their intervention, through optional 'meet-up events'.
There were five different types of OSN features identified in the included studies 1) Newsfeed, where users can share text, video, 'likes', and share links with each other; 2) Forums or Bulletin Boards; 3) Anonymous E-mail lists; 4) Peer-to-peer messaging; 5) Online Groups.See Table 2 below.

Usage and engagement
Five studies reported on participant engagement with social features.Participants were shown to have engaged with social networking features markedly more than they did with non-social features in four studies.
Two studies reported the social features engagement for the Horyzons intervention which features the moderated newsfeed and online group characteristics.Alvarez-Jimenez et al. (2013) pilot study found high levels of OSN use (95 % of the participants used the newsfeed features, and a median range of 192 page views per participant) in comparison with therapy modules of which 95 % completed at least full therapy module which dropped to 15 % completing all 7 over the weeks (with a median of 65 therapy-based page views per participant).However, Lal et al. (2023) reported limited usage of the social networking features, such as newsfeed comments among participants engaged in the peer-to-peer web-based social networking platform (the Café), with a mean of 1.85 interactions per active participant across weeks compared to a mean of 4.69 steps completed.
Three studies reported on the engagement with the online group social feature.Rotondi et al. (2010Rotondi et al. ( , 2005) ) found patients spent significantly more time on the forums of the SOAR site (72 % of total site visits) compared with the psycho-educational activities (18 % of total site visits) over the 12-month study.Biagianti et al. (2016) found that participants attendance for group therapy sessions was higher than for social cognition training in the CLIMB intervention, with 9704 visits to group sessions compared to 755 visits to educational components over the course of the 3-month study.Likewise, the engagement with group sessions was reported to be high in a later RCT of CLIMB with participants attending 80 % of sessions across the 9 week study (Dabit et al., 2021).Comparatively, engagement with the computerised Social Cognition Training and General Cognitive Training exercises was reported as "lower than anticipated", although specific percentages were not provided.
Newsfeed and Online Group Characteristics: For those studies featuring newsfeed and online group characteristics, all five studies reported relevant outcomes.2021) also found modest improvements in participants perceived social support scores at mid-treatment and post-treatment, though these were not maintained at follow-up.However, Horyzons was tested in an RCT study (Alvarez-Jimenez et al., 2021), where social functioning was a primary outcome, and in both the treatment and control groups social functioning results remained high and stable from baseline to 18-month follow-up.There was no evidence of significance between-group differences (PSP mean difference: -0.29, 95 % CI: -4.20 to 3.63, p = 0.77).There was also no evidence for the effectiveness on other secondary outcome variables such as social support and loneliness.Additionally  Primary Outcomes: • UCLA loneliness scale (Russell et al., 1978).

Secondary Outcomes:
• A brief version of the UCLA Loneliness scale (Steptoe et al., 2013).
(continued on next page) PSP: b = 6.50, 95 % CI = [− 3.55, 16.56], p = 0.20) from baseline to follow-up.Moreover, secondary outcomes indicated a nonsignificant upward trend in participants perceived social support from baseline to follow-up (b = 0.63, 95 % CI = [− 0.31, 1.58], p = 0.18).Forums and Online Group Characteristics: Three studies that included of both forums and online group characteristics reported relevant outcomes.Rotondi et al. (2010Rotondi et al. ( , 2005) ) did not find statistically significant improvements in perceived social support (p = 0.06) at the 3-month follow-up compared with treatment as usual (TAU) for SOAR.Neither did both Listserv RCT studies (Kaplan et al., 2014(Kaplan et al., , 2011)), which additionally incorporated e-mail list social characteristics, did not yield any significant findings in social support between the experimental and control groups (F(1298) = 0.08; p = 0.93) (Kaplan et al., 2011).
Peer-to-peer and Online Group Characteristics: One study (de Wit et al., 2015) which featured only the peer-to-peer messaging characteristic found no significant results for measures of social support (0.1, 95 % CI: -0.8 to 0.27, p = 0.27) or measures of social networks (-0.78, 95 % CI: -3.33 to 1.77, p = 0.54).Another study featuring peer-to-peer messaging, in combination with online groups characteristics, demonstrated significant improvements in social functioning (p = .046)as an overall in the CLIMB RCT (Dabit et al., 2021) however there were no between-group differences.

Quality of life
This review includes the analysis of eleven studies.Only five reported quality of life outcomes and found no demonstrable improvement.
Peer-to-peer characteristics: Three studies that utilised peer-to-peer messaging did not report notable improvements.Biagianti et al. (2016) only reported "trend-level improvements in aspects of quality of life" using the SQLS in CLIMB.A later RCT of the CLIMB platform (Dabit et al., 2021) found a significant group-by-time effect in aQLS (p = .025),with intent-to-treat analysis indicating the active control group showed greater improvements.No significant results were found at follow up with the WHO-QOL-Bref for MPC (− 2.53, 95 % CI:− 7.87 to 2.80, p = 0.34) (de Wit et al., 2015).
Newsfeed characteristics: Similarly, one study with newsfeed social features found no evidence for the effectiveness of Horyzons on quality of life found in the RCT (Alvarez-Jimenez et al., 2021).
Forum, e-mail list and online groups characteristics: There was also no significant difference identified between experimental and control groups for quality of life in the Kaplan et al. (2011) Listserv RCT (F (1298) = 0.47; p = 0.62) which incorporated both forum, e-mail list and online groups characteristics.

Discussion
This review findings suggest that the promising results of early smallscale studies on the effects of OSN interventions on reducing social isolation and loneliness, as well as improving quality of life, were not confirmed in larger adequately powered effectiveness studies.This review identified reportedly higher engagement with social features in comparison with non-social networking features such as education or therapy modules.Consequently, we cannot conclude that the lack of effectiveness on social isolation of the interventions which were tested so far was related to a lack of engagement.OSN appears to be attractive to people with psychosis and they seem to be engaging well with these platforms.Hence, interventions which use OSN may still be able to engage people with psychosis.Yet, the intervention approaches may need to be reconsidered if the aim of the intervention is to reduce social isolation.
This review did not solely focus on purpose-designed platforms; however, it did not identify any studies that used pre-existing mainstream OSN platform (such as Facebook and Instagram) to reduce social isolation in people with psychosis.Instead, the studies included in this review have all utilised some form of purposefully designed platforms featuring peer-to-peer interaction.It is possible that interventions that increase access to current social media platforms may be more effective in improving social isolation related outcomes compared to those that attempt to build social contacts via new, purpose-built platforms.
Mainstream social media platforms have a large user base and therefore provide a better opportunity to connect with more people and build a larger social network, outside of clinical environments.These larger preexisting networks can enable users to become connected to new people or re-connect or maintain their relationships with acquaintances, colleagues, and former classmates.These relationships can be defined as weak ties (Granovetter, 1973).Weak ties possess information that an individual's close connections might not have and would otherwise not receive (Bakshy et al., 2012;Granovetter, 1973).In comparison with other types of contact, these weak ties are more likely to be different from the user regarding lifestyle, experience, culture, hobbies, or occupation.The inclusion of those contacts, and the fact that they do not drop out of sight as they might in the 'real world', increases the diversity of the network (Papacharissi, 2010).Weak ties were also found to be more helpful than strong ties for the general public (family and close friends) when responding to participants online posts requesting help (Gray et al., 2013) and have been found to provide more supportive responses than strong ties (Rozzell et al., 2014).This diversity among weak ties, which typically comprise acquaintances rather than close friends, proves advantageous.These connections often stem from social networks and groups beyond an individual's usual circles.Such diversity exposes individuals to a broader range of perspectives, information, and opportunities that may not be readily available within their close-knit social circles (Bakshy et al., 2012;Granovetter, 1973).These diverse weak ties serve as valuable sources of novel information and support, contributing to the overall richness and effectiveness of the social network.
Interventions directly targeting measures relating to social isolation and quality of life, such as CLIMB (Biagianti et al., 2016;Dabit et al., 2021), EMBRACE (McEnery et al., 2021), and Horyzons (Alvarez-Jimenez et al., 2021, 2013;Lal et al., 2023;Ludwig et al., 2021), exhibited varying degrees of success.While EMBRACE showed improvements in social anxiety symptoms but not loneliness, CLIMB and Horyzons did not yield statistically significant improvements in their RCT studies.On the other hand, interventions with more indirect purposes, like Listserv (Kaplan et al., 2014(Kaplan et al., , 2011)), SOAR (Rotondi et al., 2010(Rotondi et al., , 2005) ) and MPC (de Wit et al., 2015), also did not demonstrate significant improvements.This might suggest that the relationship between intervention purpose and improvement in outcomes may not be straightforward and could be influenced by various factors such as intervention design or participant characteristics.
Additionally, 'closed' Listservs, Bulletin Boards and group email lists were identified as a medium of OSN in several of the studies (Kaplan et al., 2014(Kaplan et al., , 2011;;Rotondi et al., 2010Rotondi et al., , 2005)).While such methods of OSN are useful, they are arguably outdated, especially considering the emergence of platforms like MySpace in 2003 and Facebook in 2004, which were already gaining popularity.These methods are dissimilar to the multimedia OSN tools currently most frequented and referred to in the literature describing OSN use by people with psychosis (Highton-Williamson et al., 2015).Even studies which have incorporated Facebook style elements such as that of Horyzons (Alvarez-Jimenez et al., 2021, 2013;Lal et al., 2023;Ludwig et al., 2021) and EMBRACE (McEnery et al., 2021) are arguably starting to age.Increasingly fewer young people are choosing Facebook as they grow older, dropping from 94 % of young people having an account in 2012 (Madden et al., 2013) to just 27 % in 2021 (Piper Sandler, 2021).Despite this, reports of higher engagement with OSN social features compared with non-social networking features indicate a continued utilisation of such platform elements.It is therefore important to incorporate innovative multimedia OSN to enhance their acceptability, particularly when working with younger populations who have been shown to prefer mixed media such as a combination of text, graphics, and videos (Lal et al., 2015;Naslund et al., 2014).
It is important to consider the heterogeneity of outcome measures used by the studies in this review, and the findings should be approached with this in mind.Overall, there were nine different measures relating to social isolation and loneliness, and five different measures for quality of life which prevented direct comparison of different studies and the ability to complete a meta-analysis.An examination of the quality assessment of the RCTs demonstrated that the 'measurement of the outcome domain' was rated as having 'some concerns' in multiple studies.Therefore, implying that the heterogeneity of outcome measures may have impacted the reliability of the results.Future studies should aim to use standardised and validated methods of assessment to be able to reduce the impact of measurement artefacts that could impact the interpretation of results.
The varied structures of control groups across studies are noteworthy, as they may have influenced intervention effectiveness.For example, Kaplan et al. (2011) used a waitlist control, while Kaplan et al. (2014) implemented an online course.Alvarez-Jimenez et al. (2021) Horyzons and Rotondi et al. (2010Rotondi et al. ( , 2005)).SOAR control groups received TAU, and Dabit et al. (2021) CLIMB's control involved cognitive training without social functioning targeting.Despite these variations, none of the studies employing a control group demonstrated significant results in measures relating to social isolation or quality of life outcomes.Furthermore, most of the studies in this review were only moderate quality, with only two of the studies ranking as high quality.Only one of the studies included in this review was double blinded, none of the other studies were blinded, nor was an attempt made at blinding because of the nature of the interventions.Further analysis of the results of the quality assessment revealed that most of the single-group studies scored poorly on the 'internal validity bias' and 'internal validity confounding' domains of the Downs & Black checklist (Downs and Black, 1998).Therefore, the relationship between the delivered interventions and their effectiveness could have been influenced by confounding variables, suggesting a higher risk of bias.While the risk of bias and confounding variables may suggest potential influences on intervention effectiveness, the lack of significant findings in studies ranked as high quality suggests this might be the most reliable finding.

Strengths and limitations
One of the strengths of this study is the thorough and extensive search strategy to ensure the capture of a large number of studies and confidence in the inclusion of all relevant evidence, providing a comprehensive and robust evidence base for the analysis.
As a further strength no language restrictions were adopted for this study.However, those studies that were identified as meeting the search criteria were all conducted in Western countries.Most studies originated in Australia and the United States, and it is unclear whether the same findings would be reflected in other countries or cultures.
A limitation of this review was that it relied on a narrative synthesis approach to compare the results of the included studies, owing to inconsistency of outcome measures across studies, and the literature in the field is still scarce.
A further limitation of this study is that it only included studies with a confirmed diagnosis of psychosis.While this approach was taken to guarantee homogeneity of the findings and specificity of our results to clinical populations, it also resulted in the exclusion of studies that did not have formal diagnostic criteria as access criteria.Consequently, relevant evidence from studies that may have included individuals with psychosis but did not require a confirmed diagnosis was not captured in our review.This limitation may have led to overlooking valuable insights from studies approaching psychosis from different diagnostic perspectives or including individuals with subclinical presentations.
Additionally, this study did not review the grey literature, which may have contained promising studies, but instead chose to focus exclusively on peer-reviewed studies to ensure that effectiveness data were drawn from studies whose quality was checked.

Conclusion
At present the interventions including OSN are not effective in improving social isolation and quality of life.However high rates of engagement suggest that OSN is attractive to people with psychosis.
Given these findings, researchers who are interested in utilising OSN for reducing social isolation in people with psychosis should consider several key recommendations in future research.
First, interventions that use pre-existing mainstream OSN platforms may be tried if the aim is to reduce social isolation.The OSN features in the studies reviewed were closed patient groups, meaning that only those taking part in the studies could interact with one another.Another opportunity might be to use online meetings as a gateway towards full engagement with offline social opportunities, such as those provided by social prescribing programmes.Increasing access to current and widely used OSN platforms may also be beneficial as they have a larger user base that can be leveraged to build broader social connections.However, people with psychosis may also need help to develop the skills and competencies to build resilience and maximise benefits of OSN for building social networks.It is also important to note that mainstream OSN is not without risks, as previously mentioned, such as cyberbullying, harassment, and the dissemination of misinformation (Rezayi et al., 2018).
Secondly, it is important to use updated technology in interventions.The systematic review found that some of the studies used outdated OSN technologies.Therefore, future research could focus on developing and testing interventions that use updated technology to reach and interact with people with psychosis.
Finally, future research should use more homogeneous evaluation measures.Standardisation of the outcomes would be valuable to develop a clearer evidence base.For instance, measures such as the Social Media Use Integration Scale (SMUIS; Jenkins-Guarnieri et al., 2013), which has been reviewed by Sigerson and Cheng (2018).Additionally, the ongoing SCENE study (Giacco et al., 2021), is assessing social contacts through both the Social Contacts Assessment (SCA; Giacco et al., 2016) and an adapted version of the SCA for online contacts.The SCA defines social contacts as named individuals with whom participants engage in meaningful conversations, collecting information on relationship type, communication method, and contact frequency.

Declaration of competing interest
The Authors have declared that there are no conflicts of interest in relation to the subject of this study.
(a) mean participant age; (b) country; (c) sample size; and (d) diagnosis.Further information was extracted in relation to the OSN intervention: (a) study design; (b) intervention format; (c) type of control group; (d) reported effectiveness McEnery et al. (2021) identified improvements in levels of experiencing pre-post social anxiety symptoms with the EMBRACE platform, as measured by the Liebowitz Social Anxiety Scale (d= − 1.35, p = 0.002) and the SIAS (d= − 1.70, p = 0.0005).However, statistically significant pre-post improvements in loneliness symptoms, a secondary outcome of this study, were not found (McEnery et al., 2021).After one month of using the Horyzons platform, Alvarez-Jimenez et al. (2013) found that 60 % of patients reported increased perceived social connectedness, thus suggesting Horyzons to have preliminary efficacy.A follow-up study found similarly promising results on reducing loneliness scores (Ludwig et al., 2021) with large improvements shown from baseline to mid-treatment.Ludwig et al. ( Fig. 1.PRISMA Flow diagram.

Funding
Emily Thelwell is funded by the National Institute for Health and Care Research (NIHR) Applied Research Collaboration (ARC) West Midlands.Professor Domenico Giacco is supported by the NIHR ARC West Midlands.The views expressed are those of the author and not necessarily those of the NIHR.CRediT authorship contribution statement Emily L.R. Thelwell: Writingoriginal draft, Software, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization.Laura Dunkerley: Writingreview & editing, Methodology, Formal analysis.Robin Goodwin: Writingreview & editing, Supervision.Domenico Giacco: Writingreview & editing, Supervision, Conceptualization.

Table 1
Study characteristics.

Table 1
(continued ) a Study included data from caregivers or from patients with a different condition; results from these participants are not reported in this review.E.L.R.Thelwell et al.

Table 2
OSN features in the included studies.