Effects of online stigma‐reduction programme for people experiencing mental health conditions: A systematic review and meta‐analysis

Abstract Despite the increased awareness of mental health‐related issues, people experiencing mental health conditions have continued to face stigmatization worldwide. The literature on help‐seeking behaviours has frequently highlighted the development of self‐stigma because of public stigma and emphasized the need to address public stigmatization faced by them. Given the increasing acceptance of digital services in recent years, this systematic review aimed to examine the effects of online and face‐to‐face anti‐stigma interventions in reducing public stigma towards people experiencing mental health conditions. A search was conducted on the Cochrane Library, CINAHL, PubMed, Embase, PsycInfo, and ProQuest from inception of the databases to October 2020. Studies were included in this review if they have explored: (i) public stigmatization towards people of all ages with different types of mental health conditions; (ii) online interventions; and (iii) face‐to‐face interventions. Nine studies were included in this review, of which only five were included in the meta‐analysis as the remaining four had incomplete data. The meta‐analysis included an aggregate of 1203 participants while the four excluded studies included 713 participants. Results revealed that online interventions performed favourably with small effect sizes in comparison to face‐to‐face, wait‐list control, and no‐intervention groups. Results from the studies excluded from the meta‐analysis also found a significant reduction of public stigmatization with online interventions. Such findings provide insightful evidence for the effectiveness of online interventions in reducing public stigmatization. Hence, mental health organizations and groups can consider adopting online interventions suitable for their target audience and type of mental health conditions.


INTRODUCTION
The global prevalence of mental health conditions has remained consistently high across the past few decades (Richter et al. 2019), with an estimated 29.2% of the population experiencing at least one form of mental health condition in their lifetime (Steel et al. 2014). In the United States of America (USA) alone, 51.5 million adults (20.6%) had at least one form of mental health condition in 2019 (Substance Abuse & Mental Health Services Administration [SAMHSA] 2020). Yet, only 44.8% (23 million adults) of people experiencing mental health conditions (PMHCs) have received mental health services (SAMHSA 2020). This thus suggests the presence of a significant gap in accessing treatment for mental health conditions. While structural barriers including the socio-economic status, governmental policies, and availability and accessibility of mental health services may underlie the gap, prevalent attituderelated barriers such as stigmatization also discourage the use of existing mental health services, thus exacerbating the gap (Herrera-Ferr a 2020).
Stigma is conceptualized as an attribute that discredits an individual, typically invoking a sense of shame, self-blame, and social exclusion (Goffman 1963;Waqas et al. 2020). Despite the focus on reducing stigma related to mental health conditions over the last few decades, PMHCs have continued to experience much stigmatization worldwide, as reported in China (Young & Ng 2016), Greece (Tzouvara et al. 2016), Canada (Weeks et al. 2017), Nepal (Maharjan & Panthee 2019), and the USA (Bonfils et al. 2018). Dire consequences commonly faced by PMHCs include social isolation (Linz & Sturm 2013), unemployment (Brouwers 2020), reduced access to health care (Fraser et al. 2020), impaired quality of life ( Switaj et al. 2017), and feelings of hopelessness, shame, and guilt (Oexle et al. 2017). These may result in delayed treatmentseeking behaviours (Kular et al. 2018) or even suicidal ideations (Oexle et al. 2017) among the PMHCs.
Stigma related to mental health conditions typically presents in two main forms: public stigma and selfstigma. Public stigma refers to societal stigmatizing reactions and attitudes where PMHCs are perceived to be dangerous or incompetent, or to have a character weakness (Rogers 2018) whereas self-stigma refers to an internalized feeling of shame, guilt, and low self-regard due to societal prejudice (Waqas et al. 2020). Helpseeking literature has thus far underlined self-stigma as a development of anticipated or experienced public stigma to varying degrees, depending on the mental health condition (Hing & Russell 2017;Vogel et al. 2013). This is supported by the modified labelling theory (Link et al. 1989) which postulates that external perceptions of PMHCs and negative labels attached to mental illnesses are internalized by PMHCs which later influence their self-image and lead to self-stigma. Hence, addressing the public stigmatization of PMHCs is paramount in bridging the mental health treatment gap.
Traditional interventions targeting public stigma are often conducted in person at educational institutions, military bases, workplaces, and healthcare organizations, and employ strategies such as psychoeducation workshops, social activism in the form of protests, contactbased approaches, and reading programmes (Corrigan 2012;Moll et al. 2015;Yamaguchi et al. 2013a). Reviews on anti-stigma interventions Griffiths et al. 2014;Yamaguchi et al. 2013b) have found that educational and contact-based approaches were effective in improving knowledge of mental health conditions, reducing stigma, improving attitudes, and reducing feelings of social distancing towards PMHCs. Despite their findings, existing reviews have not examined how the modality of the interventions would affect their effectiveness. Furthermore, given the increasing prevalence of information technologyreflected by an estimated 87% Internet usage rate (International Telecommunication Union 2020) and 98% 3G mobile network coverage in developed countries in 2019an online platform will be ideal for a widespread, low-cost outreach to the population (Rogers 2018). This surge in the interest in and acceptance of digital services has been further catalysed by the COVID-19 pandemic (Wind et al. 2020), due to which safe-distancing measures have been integrated as part of our daily lives. Therefore, an empirical assessment of the effects of online interventions for reducing stigmatization of mental health conditions could offer insights upon which mental health nurses can act in years to come.

AIMS
This review aimed to examine the effects of online anti-stigma interventions when compared with nononline interventions in reducing public stigma towards people experiencing mental health conditions. The specific review question identified were as follows: (1) How effective are online anti-stigma interventions in reducing public stigma when compared with face-toface anti-stigma interventions?

Search strategy
A review protocol was developed and registered on PROSPERO (CRD42020213165) according to recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement (Moher et al. 2015) to ensure methodological rigour of the systematic review (Tam et al. 2019). A search strategy was formulated to identify studies on anti-stigma interventions targeting public stigma towards PMHCs. To maximize the retrieval of relevant studies, help was enlisted from an experienced librarian. All available and relevant primary studies were then identified through the search with specific keywords and Medical Subject Heading (MeSH) terms from inception of the databases to 16 October 2020 in order to ensure the inclusion of updated materials in clinical practices (Tam et al. 2017 Through Boolean search terms, five databases (CINAHL, PubMed, Embase, PsycInfo, and the Cochrane Library) and ProQuest were included, given their coverage of different disciplines including medicine (Pubmed, Embase, and the Cochrane Library), psychology (PsycInfo), nursing, and allied health (CINAHL). End-reference lists of the identified studies were hand-searched to ensure no omission of additional studies (Whittemore & Knafl 2005). The review was limited to articles in English, as the team had no access to interpreters.

Population
The eligible studies had participants of all ages from the general population.

Intervention
The eligible studies examined online interventions targeted at reducing public stigma towards people experiencing mental health conditions.

Comparator
Studies with no comparator, a passive comparator (usual care or wait-list control group) or active comparator (other face-to-face interventions) were included.
Outcomes. The primary outcome for this review was stigmatizing attitudes towards people experiencing mental health conditions. The secondary outcomes were threefold: stigma experienced by people with different mental health conditions; knowledge of mental health conditions; and the desire for social contact with PMHCs. Studies were excluded if they had measured only internalized or self-stigma.

Selection of articles
All retrieved records from database search were uploaded into EndNote X9, and duplicate studies were removed electronically. For the remaining studies, the titles and abstracts were reviewed independently by YSG and JOY; studies not meeting the inclusion and exclusion criteria were removed at this stage. Potential articles deemed suitable by at least one author were then downloaded for further scrutiny. The articles were then reviewed independently by YSG and JOY and disagreements were consensually resolved through discussion with the third reviewer (TWSW).

Data extraction
Data extraction was then performed independently by YSG and JOY for the following aspects: study-related information (author, location, year, research design, and sample size); and primary outcome (stigma attitudes). The authors of the articles were also contacted for missing data on numerical outcomes. To ensure consensus when extracting information from the retrieved studies, a pilot review was conducted independently by YSG and JOY based on a data extraction form adapted from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2020). All disagreement during data extraction was resolved in consultation with a third reviewer (WWT).

Risk of bias
The risk of bias (ROB) of the included studies was evaluated independently by YSG and JOY based on the Cochrane Risk of Bias assessment tool (Higgins 2020). The individual domains of bias in the ROB assessment tool were rated as 'low risk', 'high risk', or 'unclear risk'. Accordingly, a given study was rated as 'high risk' if more than half of its domains were evaluated to be at 'high risk' or 'unclear risk'. Disagreement was consensually resolved through discussion with the third reviewer (TWSW). Review Manager 5.4 software was used to generate the ROB summary graph to illustrate the risk for the domains of each study (RevMan 2020).

Data analysis
Post-programme measurements for each of the outcome measures were extracted from the randomized controlled trials (RCTs) for comparison. The mean difference (MD) and its 95% confidence intervals (CIs) were computed as the effect measure. Heterogeneity was examined by the Chi-square test and I 2 statistics. The statistical significance of heterogeneity was set at P < 0.10, and the degree of variability was estimated through I 2 values, with 75%, 50%, 25%, and 10%, respectively, indicating high, moderate, low, and no heterogeneity (Higgins 2020). The fixed-effects model was used for homogeneous studies (chi-square P > 0.10 & I 2 value < 50%); otherwise, the DerSimonian-Laird random-effects model was used. A funnel plot was created to explore potential publication bias when there were 10 or more studies for the outcome (Sterne et al. 2011). Data were analysed through Review Manager 5.4 (RevMan 2020) according to the modality of the interventions.

RESULTS
Of the 2,381 studies retrieved from the databases, 173 duplicates were removed. Upon screening of the titles and abstracts, another 2117 studies were removed. For the remaining 28 studies, their full texts were retrieved, from which 19 were removed with reasons, leaving nine articles in the review (Figure 1 for PRISMA flowchart).

Characteristics of the included studies
The selected studies were published from 2004 to 2019: five were conducted in Australia (Griffiths et al. 2016;Griffiths et al. 2004;Jorm et al. 2010;Kiropoulos et al. 2011;Shann et al. 2019), one in Russia (Finkelstein et al. 2007), one in the USA (Hamblen et al. 2019), one in the United Kingdom (UK) (Davies et al. 2018), and one in Canada (Moll et al. 2018). Their sample sizes ranged from 55 (Davies et al. 2018) to 525 (Griffiths et al. 2004), with a median of 196 (Table 1).

Risk of bias within the studies
Of the nine included studies (Figure 2), the majority were assessed to be of a low overall risk of bias; in almost each of them, the unclear risk was found to originate chiefly from one category (selective reporting). High risk of bias was found mainly in allocation concealment, blinding of participants and personnel, and outcome assessment.

Primary outcome
Public stigma towards PMHCs A meta-analysis was conducted on five studies (Davies et al. 2018;Griffiths et al. 2016;Griffiths et al. 2004;Kiropoulos et al. 2011;Moll et al. 2018) which included 1203 participants in the meta-analysis. Two of the five studies each compared a group with an online intervention with a no-intervention group or a wait-list control (WLC) group (Davies et al. 2018;Griffiths et al. 2016). Two other studies each compared a group with an online intervention with a control group subjected to an interview (Griffiths et al. 2004;Kiropoulos et al. 2011). The last study compared a group with an online intervention with another with a face-to-face intervention (Moll et al. 2018). The combined standardized mean difference (SMD) from the five studies was -0.33 (95% CI -0.60 to -0.05, P = 0.02). The heterogeneity was I 2 = 82% (P < 0.001), alongside significant subgroup differences between three subgroups (P = 0.03) ( Figure 3). As Kiropoulos et al. (2011) exhibited a much larger SMD of -0.89 (95% CI -1.18 to -0.60) than the two intervention groups from Griffiths et al. (2004), sensitivity analysis was conducted by excluding Kiropoulos et al. (2011). After removal, the combined SMD from the four studies stood at -0.22 (95% CI -0.44 to 0.01, P = 0.06). The heterogeneity thus declined to I 2 = 66% (P < 0.001), alongside significant subgroup differences between the three subgroups (P = 0.002) ( Figure 4). This indicates that the group with the online interventions demonstrated a significant reduction of public stigma only when compared to the WLC or no-intervention groups but not when compared to other non-online intervention groups.
Four studies (Finkelstein et al. 2007;Hamblen et al. 2019;Jorm et al. 2010;Shann et al. 2019) could not be included in the meta-analysis and were analysed narratively due to missing summary data (Table 1). A significant stigma-reduction effects were found for affective [F(1, 191) = 14.55, P < 0.001, g 2 = 0.07] and behavioural [F(1, 191) = 10.04, P = 0.002, g 2 = 0.05] public stigma subscales for online interventions (Shann et al. 2019). Finkelstein et al. (2007) found significant stigma-reduction effects for a group with an online intervention and for a reading group, both in comparison to a no-intervention group (BSDS online and reading; P < 0.0001, CAMI online and reading; P < 0.0001) although no comparison was conducted between treatment groups. Likewise, in investigating stigmatization towards people with schizophrenia, Jorm et al. (2010) found significant stigma-reduction effects for a group with an online intervention and for a reading group, both in comparison to the WLC group. Their further analysis found more substantial stigmareduction effects for the group with the online        Hamblen et al. (2019) found no stigma-reduction effects for both the group with an online intervention and the reading group, which might have been due to pre-existing positive attitudes towards mental health treatment at the baseline among all participants. Thus, these findings collectively suggest that online interventions were at least as effective as face-to-face interventions in reducing public stigma towards PMHCs. A meta-analysis was conducted on three studies (Griffiths et al. 2016;Kiropoulos et al. 2011;Moll et al. 2018) which included follow-up assessments and longterm effects of the interventions for stigma towards PMHCs. A total of 428 participants were included in the meta-analysis. The combined SMD from the three studies stood at -0.44 (95% CI -0.60 to -0.28, P < 0.001). The heterogeneity was I 2 = 5% (P = 0.37) ( Figure 5). This small-to-medium effect size suggests longitudinal stigma-reduction effects of the online interventions. However, as the caveat is that only three out of five studies examined longer-term stigmareduction effects in the meta-analysis, this finding should be interpreted with caution.
Three studies (Finkelstein et al. 2007;Jorm et al. 2010;Shann et al. 2019) that included follow-up assessments and long-term effects of the interventions for stigma towards PMHCs were not included in the metaanalysis due to missing summary data (Table 1). All three studies found significant stigma-reduction effects in their online interventions as compared to reading, WLC, and no-intervention groups. These findings collectively suggest potential longitudinal stigma-reduction effects for such online interventions that may be worth further elucidation.

Secondary outcomes
Stigma experienced by people with different mental health conditions A meta-analysis was conducted on five studies (Davies et al. 2018;Griffiths et al. 2016;Griffiths et al. 2004;Kiropoulos et al. 2011;Moll et al. 2018) according to the type of mental health conditions examined. Three studies examined the stigma towards people with depression (Davies et al. 2018;Griffiths et al. 2004;Kiropoulos et al. 2011). One study examined that towards people with any mental health conditions (Moll et al. 2018) and another study examined that towards people with either depression or anxiety (Griffiths et al. 2016). The pooled result for depression was -0.38 (95% CI -0.76 to 0.01, P = 0.06), with the heterogeneity being I 2 = 86% (P < 0.001). The result for anxiety consisting of only scores from Griffiths et al. (2016)was -0.42 (95% CI -0.70 to -0.13, P = 0.004). The result of general mental health conditions consisting of only scores from a sole study (Moll et al. 2018) was -0.00 (95% CI -0.31 to 0.30, P = 0.98). No significant subgroup differences were detected between the three types of mental health conditions (P = 0.12). Similarly, sensitivity analysis was conducted by excluding results from Kiropoulos et al. (2011). The differences regarding stigma reduction between the types of mental health conditions remained non-significant after removal (P = 0.15) ( Figure 6).

Knowledge of mental health conditions
Only three studies (Griffiths et al. 2016;Kiropoulos et al. 2011;Moll et al. 2018) measured knowledge of mental health conditions. One study measured only the literacy on depression (Kiropoulos et al. 2011), another measured the literacy on both depression and anxiety (Griffiths et al. 2016), and the third measured the knowledge of general mental health (Moll et al. 2018). These differences could account for the substantial heterogeneity

Desire for social contact with PMHCs
Only one study (Jorm et al. 2010) measured the desire for social distancing (contact) with PMHCs. The online intervention was found to produce greater reductions in the desire for social distancing (improvements in desire for social contact) towards PMHCs as compared to both the reading (OR = 0.10, 95%CI 0.02-0.46) and WLC (OR = 0.05, 95% CI 0.01-0.26) groups. However, this effect was not maintained at follow-ups (OR = 0.09, 95% CI 0.02-0.42) ( Table 1).

DISCUSSION
This review aimed to examine the effects of online and face-to-face anti-stigma interventions in reducing public stigma towards people with various mental health conditions. Nine studies comparing an online stigmareduction intervention with a non-online one or a control group were included in the systematic review. However, only five studies were included in the metaanalysis as the rest of the studies had missing or incomplete data. The said five studies included an aggregate of 1203 participants (644 online, 559 comparison) while the four excluded ones had an aggregate of 713 participants (305 online, 410 comparison). Overall, results from pooled values of post-programme data and studies excluded from the meta-analysis revealed that participants in the online intervention groups demonstrated a significant reduction of public stigma in comparison to non-online, WLC, and no-intervention groups. The small effect sizes observed are consistent with those for other public stigma-reduction interventions ) and subgroup differences concerning the modality of interventions remained significant for the pooled values. Findings from our study are supported by other evidence-based online anti-stigma interventions, which have proved a promising avenue for removing barriers to accessing help for mental health symptoms among PMHCs (Nickerson et al. 2020).
Furthermore, according to the social-cognitive model (Corrigan et al. 2001), the reduction of stigma could be a result of the modification of erroneous social beliefs such as stereotypes and attitudes towards PMHCs (Pedersen et al. 2011). Therefore, such findings suggest that online interventions were as effective as nononline ones in reducing public stigma towards PMHCs.
One noteworthy finding concerns the absence of between-group differences in the effectiveness of the interventions for the various mental health conditions studied. This suggests that online interventions targeting stigma towards any form of such conditions will likely exhibit similar effectiveness (Hanisch et al. 2016). As aforesaid, no meta-analysis could be performed for the effects of the interventions on the knowledge of mental health conditions and the desire for social contact with PMHCs, given the heterogeneity of the studies and their missing data. However, individual results from the few studies examining these outcomes indicate possible benefits of online interventions and warrant further research. This review contributes to the findings of previous systematic reviews on the effectiveness of general stigma interventions Griffiths et al. 2014;Yamaguchi et al. 2013b) in its examination of interventional modality in public stigma research. Traditional interventions typically use educational (delivering content on mental health conditions and challenging myths) and social contact-based approaches (involving personal contact or presentations by PMHCs) that have been found effective in reducing public stigma Griffiths et al. 2014;Yamaguchi et al. 2013b). Based on the intergroup contact theory (Allport 1955), the premise for social contact-based approaches involves invoking greater familiarity and a more positive appraisal of PMHCs (Holmes et al. 1999), therefore raising the possibility that online interventions may diminish the effectiveness that these in-person interventions offer.
However, social contact-based interventions through video recordings have been found equally effective in FIGURE 7 Knowledge of mental health conditions. achieving de-stigmatization effects (Brown 2020;Janou skov a et al. 2017). As all nine studies in this review have integrated either one or both approaches in their online interventions, it is unsurprising that a significant stigma-reduction result was found when compared to WLC or no-intervention groups. This may indicate that traditional stigma interventions can be successfully transferred to online platforms while retaining their clinical utility (Sampogna et al. 2017). Nonetheless, online interventions come with challenges: high drop-out rates have been observed, especially for interventions spanning substantial periods of time (Donkin & Glozier 2012). Pedersen et al. (2019) have suggested that the highest risk of drop-outs from any programme would happen in the beginning when attrition resulted from a reduction of online activities from participants over time. Hence, programme coordinators should discern abnormal decreases in online activities as an indicator reflective of potential attrition and pre-empt it by initiating re-engagement. Furthermore, helping participants to identify and incorporate intrinsic motivations such as noticing their own improvement has been found to foster a sense of duty to self and personal empowerment: this component in the programme can further encourage adherence (Donkin & Glozier 2012).
The dearth of research comparing online and nononline public stigma interventions in conjunction with the missing summary data for multiple secondary outcomes prevented further conclusive remarks regarding the effectiveness of online and face-to-face interventions. However, our findings suggest that the online interventions afforded public stigma-reduction effects (Roslee & Goh 2020) that may last well beyond their initial completion. This bears critical implications for the future delivery of such interventions: the rapid proliferation of internet users (International Telecommunication Union 2020) and the evolving COVID-19 pandemic may increase the acceptability of online interventions (Wind et al. 2020). It may follow that such interventions may be delivered to a larger population at a lower cost (Rogers 2018) without compromising their quality and effectiveness.
However, mental health practitioners need to be vigilant in noting any potential unintended effects from the interventions that could eventually reinforce the public stigma which they are designed to address. Although the use of online anti-stigma interventions has overcome geographical, mobility, and time constraints, potential costs on the users and their social networks have to be considered. PMHCs already facing public stigma may feel further stigmatized when excluded from the mainstream healthcare delivery system (Corrigan & Rao 2012;R€ usch et al. 2005). It is, therefore, important for researchers to include theoretical underpinnings that can provide an empirical evaluation on their chosen online anti-stigma interventions in place of non-online ones (Griffiths et al. 2006). Finally, having randomized control trials with nononline anti-stigma interventions will improve the evaluation of Internet-based interventions as an effective prospective intervention.

Strength and limitations
Through our timely evaluation of the effectiveness of online anti-stigma interventions, healthcare providers, not-for-profit organizations, and mental health advocacy groups looking to implement such interventions have been provided with an overview of the advantages and disadvantages. This allows them to make an informed decision when choosing the most appropriate modality for their specific needs. However, some limitations are of note, such as the small number of studies comparing online and non-online interventions addressing public stigma; substantial heterogeneity in some analyses due to the variations of the contents of the interventions, the comparator (or usual care) and the instruments used in measuring the outcome; and missing summary data in many studies. As a result, only a small number of trials were included in the meta-analysis (four trials), thus limiting the ability for further conclusions to be drawn. Despite that, this review still provides a valuable foundation for future research examining the impact of the modality on interventions addressing public stigma. In this review, the pre-defined protocol ) was adhered to with some minor alterations to the secondary outcomes. The longitudinal effect of the stigma interventions was considered as an extension of the primary outcome while the predefined subgroup comparing population groups was modified to examine different mental health conditions.

CONCLUSIONS
The current review of nine studies provides evidence that online interventions are comparatively effective in reducing public stigma towards PMHCs as non-online interventions. Future replications and research on this topic are required to establish conclusive evidence for their effectiveness. Additionally, future research may consider examining the effectiveness of online interventions on not only stigma-related areas (such as knowledge and desire for social contact) but also the type of stigmatized mental health conditions; these aspects have remained unclear due to the limited studies on this area of research.

RELEVANCE TO CLINICAL PRACTICE
This review presents some evidence that the effectiveness of online interventions in reducing public stigma towards PMHCs would persist even after the conclusion of the intervention. The effectiveness of online interventions did not differ based on the types of mental health conditions. As the interventions reviewed focused on stigma towards PMHCs, organizations, and groups with an interest in mental health may consider adopting online interventions suitable for their target audience. Due to the small effect sizes found in the studies in this review and other interventions addressing public stigma , further developments of stigma interventions, regardless of the modality, may also be warranted for greater efficacy in this field.