It is essential that health and social care practitioners access contemporary, high-quality research evidence to help them deliver the best evidence-based clinical care and improve patient outcomes [1-5]. Rapid dissemination, by active approaches using specific channels and planned strategies, is recommended [6,7].

Social media may facilitate rapid dissemination to busy practitioners, allowing them to access and interpret research evidence efficiently [8-10]. Because social media are widely used and not limited in space and time [8,11], they have the potential to overcome barriers to dissemination, including reaching practitioners with limited professional opportunities or time constraints and filtering the exponentially increasing volume of research evidence produced every year [9,12-14]. Currently, closed social media channels, such as private and invitation-only groups, are often used by practitioners for day-to-day communications, clinical information sharing, and targeted clinical education, whereas open social media channels that can be accessed by everybody are used for reputation development; public health education; and, increasingly, research dissemination [9,10,15-26].

However, the effectiveness of open social media for the dissemination of research evidence to health and social care practitioners is largely unknown [27]. Existing reviews have narratively synthesized potential uses, benefits and risks, similarities and differences, and qualitative experiences of social media or provided commentaries on the mechanisms of research dissemination by social media [10,22,23,28]. No reviews have conducted a meta-analysis to quantitatively test the effectiveness of social media for the dissemination of research evidence to health and social care practitioners. To inform evidence-based recommendations, the evidence for using social media to disseminate research evidence must be investigated.

The primary research question was as follows: “How effective is open social media as a way to disseminate research evidence to practitioners?” The objective of this systematic review was to quantitatively synthesize and meta-analyze evidence of the effectiveness of social media for the dissemination of research evidence to health and social care practitioners by evaluating both between-group comparisons of social media versus no social media and within-group comparisons of before-after social media campaigns. The social media platforms, formats, and strategies used were also identified, and their effectiveness was compared to understand the most effective social media intervention characteristics for the dissemination of research evidence to practitioners.

The protocol was registered on the International Register of Systematic Review (PROSPERO; CRD42022378793) and published a priori [29]. It was reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidance [30] (Multimedia Appendix 1).

Articles published between January 1, 2010, and January 10, 2023, were eligible for inclusion if they investigated research evidence targeted at health and social care practitioners that was shared using open social media. Articles were included if they quantitatively compared social media versus no social media (either between-group comparisons or before-after social media within-group comparisons) or if they compared social media platforms, formats, or strategies. Eligible study designs included randomized controlled trials (RCTs), case-controlled comparisons, crossover, nonrandomized group comparisons, before-after comparisons, cohort comparisons, and case reports. Eligible outcomes of interest included reach, engagement, direct dissemination, and impact. Definitions of the eligibility criteria terms are shown in Textbox 1.

Articles were not eligible for inclusion if they only compared social media effectiveness in terms of the topic or specialty of the research evidence–related social media post or posts. Excluded study designs included protocols, reviews, studies using only qualitative methods, opinion pieces, and conference abstracts with no linked full-text article. Articles were excluded if they preceded 2010 (refer to the protocol by Roberts-Lewis et al [29]), were not available in English, or did not feature research evidence–related social media of relevance to postregistration health or social care practitioners (eg, only targeting students, service users, or the public or featuring non–health and social care research topics). Articles were also excluded if the social media campaign was targeted at practitioners for purposes other than the dissemination of research evidence (eg, delivering multisource clinical education courses, organizational information, administrative tasks, practical peer support, day-to-day interpersonal clinical communication, professional identity, or reputation promotion). Finally, articles that did not provide sufficient quantitative empirical data on reach, engagement, direct dissemination, or impact were excluded.

Six electronic databases were searched (MEDLINE [Ovid], PsycINFO [Ovid], CINAHL plus [EBSCO], and ERIC [EBSCO] as well as LISTA and OpenGrey). The date of the last search was January 10, 2023. Bibliographies of relevant reviews and included articles were searched for citations and PubMed, Elicit, and Google Scholar were used for reference harvesting.

For full search strategies, refer to Multimedia Appendix 2 and the protocol by Roberts-Lewis et al [29]. Key search terms were grouped as follows:

Records from the electronic and citation searches were exported to EndNote Online (Clarivate) for deduplication and then imported to Rayyan software (Rayyan Systems) for title, abstract, and full-text screening.

Title and abstract screening were carried out by 2 independent reviewers (SRL and SQM). There was 92% agreement (κ=0.87) on eligibility decisions and 100% agreement after discussion.

Full-text screening was carried out by at least 2 of 5 independent reviewers (SRL, SQM, LB, HG, and FJL). There was 84% agreement (κ=0.83) on full-text inclusion decisions and 100% agreement after discussion.

Data from the included studies were extracted independently by at least 2 of 5 reviewers (see the Selection Process section) using a data extraction form developed a priori [29]. The accuracy of data extraction was confirmed by comparison between extraction forms, returning to the original article to resolve any disparity.

The variables collected were study characteristics including the number and description of subjects; social media platforms, formats, and strategies; study design; comparisons; and outcomes. For each outcome of interest, means, SDs, and sample sizes were extracted for each comparison. When these data were missing, they were calculated from other reported statistics using recommended methods [31], where possible.

For studies that reported multiple outcome measures, only outcomes of interest were collected (reach, engagement, direct dissemination, and impact). Different measures for the same outcome were prioritized for inclusion in meta-analyses according to the a priori protocol [29]. Subsequent additions were made to the prioritization order to account for heterogeneous data reported in the included studies; these included aggregated total interactions, other types of post interactions, and the measurement time frame according to the most common time frames for each outcome of interest (Textbox 2).

The 34-item (5-domain) Cochrane Risk of Bias 2.0 tool was used to rate the quality of the RCTs as lower risk of bias, some concerns, or higher risk of bias [32]. The Newcastle-Ottawa Scale (score range 0-9) was used to assess the risk of bias in nonrandomized designs [33]. A score of ≤3 was considered high risk of bias, scores between 4 and 6 were considered medium risk of bias, and a score of ≥7 was considered low risk of bias [33]. Risk of bias was assessed independently by at least 2 reviewers and data were checked for accuracy by a third reviewer.

The included studies were summarized narratively in text, tables, and figures. Quantitative comparisons were made using calculated standardized mean differences (SMDs), CIs, and P values for each comparison. SMD effect sizes were calculated using Hedges g to accommodate the heterogeneity of outcomes. Effect sizes of >0.8 were defined as large, ≥0.5 to 0.8 as moderate, and <0.5 as small [34]. Outcome effect sizes were presented as SMD, 95% CIs, z-test, and P value.

For outcomes where group means, SDs, and sample sizes were obtained from at least 2 studies, pooled effect sizes were calculated using random effects models in RevMan (version 5.3; The Cochrane Collaboration). The heterogeneity of pooled data was assessed using I². Funnel plots were assessed visually for each meta-analysis to check for publication bias.

For pooled data with I²>75%, subgroup analyses were planned; however, these were not possible due to the heterogeneous characteristics of the social media interventions in the included studies or an insufficient number of studies to achieve ≥80% statistical power [35]. Therefore, studies were ordered according to effect size, and the common characteristics of social media strategies in the studies with the largest effect sizes were narratively synthesized. Although no sensitivity analysis was planned a priori, evidence from RCTs and studies with low risk of bias was given greater weighting in the narrative synthesis than nonrandomized studies and those with high risk of bias.

For each outcome, the certainty of the evidence base was evaluated based on the GRADE (Grading of Recommendations, Assessment, Development, and Evaluations) approach [36] and categorized as high, moderate, low, or very low [37,38].

A total of 50 studies published between 2013 and 2022 were included; 9 were RCTs [39-48]. In total, 26 studies included nonrandomized comparisons [49-74], 12 studies were before-after comparisons [75-86], and 3 were case studies [87-89]. A total of 10 studies included both between-group comparisons and before-after analyses [40,44,45,47,51,53,60, 63,70,74]. For study descriptions, refer to Multimedia Appendix 3 [39-89].

A total of 36 studies investigated the impact of social media on journal articles, with samples ranging from a single journal article [88] to 15,078 articles from multiple journals [54]. In total, 8 studies focused on research blogs [40,76,87] and microblogs [47,49,57,82,85], 4 studies examined research conference social media posts and hashtags [71,74,86,89], 2 studies investigated clinical guidelines [60,63], and 1 study tested research-related posts linked to health care hashtags [62].

Half of the studies explored multiple social media platforms and the other half of the studies examined a single platform. Twitter was used in all but 1 study [85], Facebook was used in 23 studies [40,42-44,47,48,53,57,60,63,64,67-70,76,77, 80-82,85,89], LinkedIn was used in 8 studies [40,48,60,63,64,68,87,89], Instagram was used in 6 studies [57,62,67,68,76,87], and YouTube was used in 5 studies [58,60,63,64,87], whereas TikTok [87], Weibo [39], Google+ [64], Tumblr [81], and Spotify [87] were each used in 1 study. The most common media formats examined were text posts, which usually included links and images. Journal clubs or tweet chats were included in 9 studies [51,60,62,74,75,77,78,80,81], and video media were used in 6 studies [52,60,62,63,73,78]. The social media campaign duration ranged from 1 hour [51] to 5.5 years [54].

Outcomes measurement duration ranged from 3 days [86,89] to 4 years [64]. Typically, outcomes were measured after 1 month [42-44,48,52,53,56,60,63,67,75-81,83-85,87] or 1 to 2 weeks [39-41,45,47,51,68,74,82,88]. Citations were measured after ≥1 year, except in 1 study that reported citations after 6 months [61].

In total, 5 RCTs had low risk of bias [39,41-43,45,46], 3 RCTs had some concerns [40,44,48], and 1 RCT had high risk of bias [47] (Multimedia Appendix 3). The most common reasons for risk of bias included insufficient information provided about the allocation sequence, handling of missing data, or prioritization of multiple eligible outcome measurement time points. In total, 11 nonrandomized studies had low risk of bias [49,50,55-57,63,65,70,72,79,81]; 25 nonrandomized studies had moderate risk of bias [51-54,58,59,61,62,64,66-69,71,73-78, 80,82-85]; and 5 nonrandomized studies had high risk of bias [60,86-89] (Multimedia Appendix 3). The most common reasons for risk of bias included targeted selection of studies for social media sharing and incomplete reporting of data handling. Funnel plots did not indicate a high risk of publication bias in pooled data.

A total of 21 studies (including 3 RCTs [41,45,47]) evaluated engagement (6 studies examined likes [45,47,49,51,83,88]; 11 studies investigated total engagements, including shares, comments, and other interactions [41,45,49,51,53,54, 67,71-73,78]; 7 studies assessed only post shares [47,49,52,56,68,83,84]; and 2 studies reported on other post interactions only [63,75]). No included studies evaluated comments or reposts alone.

In total, 23 studies (including 8 RCTs [39-45,48]) evaluated direct dissemination (10 in link clicks [41,45,49,51,53,56,73,79,80,83], 13 studies reported article views [39,40,42-44,48,63,69,70,74,81,84,87], 8 measured PDF downloads [40,44,48,50,63,69,79,81], and 9 assessed the altmetric score [44,45,51,69,74,77,79,83,87]).

A total of 13 studies (including 3 RCTs [46-48]) evaluated impact (7 assessed article citations [46,48,54,55,58,59,65], 3 investigated impact factors [54,61,64], and 4 examined changes in thinking or behavior [47,57,60,82]).

Considering all the evidence, there was high certainty that social media is effective for the dissemination of research to health care practitioners. The outcome with the highest certainty across all comparisons was direct dissemination. Evidence was insufficient to determine the size and certainty of between-group effects of social media compared to no social media on reach and engagement; however, there was moderate certainty of large and moderate effects of social media interventions on direct dissemination and impact, respectively. After social media exposure, there was moderate certainty of large, positive within-group effects on reach, engagement, and direct dissemination, whereas there was low certainty of a small effect on impact. Certainty was generally low regarding the size of the effects of platforms, formats, and strategies on each outcome. However, the direction of effects was consistently in favor of using multiple platforms (particularly Twitter and Facebook), using images (particularly infographics), and involving influential others in social media campaigns.

The level of certainty about the size of the effects in favor of social media was different depending on the outcome of interest and study characteristics. There was a tendency for smaller effect sizes in RCTs and studies with less-intensive social media interventions. The certainty ratings were lowered for all comparisons due to the variability of the included study designs, many of which were descriptive and not designed for rigorous quantitative evaluations, meaning most included studies had moderate risk of bias. Consistency and precision of effect size estimates were also threatened by the low number and heterogeneity of studies included in some comparisons. However, for some comparisons, certainty was uprated due to the large magnitude of estimated effects and potential dose-response gradients between the intensity of social media interventions and their effectiveness.

There was evidence of the effectiveness of social media for the dissemination of research evidence to health care practitioners. All the included studies reported some findings in favor of social media, although there was considerable heterogeneity in effect sizes, and study quality was mixed. Effect sizes of social media effectiveness were influenced by the frequency, intensity, and composition of social media interventions. Effectiveness was enhanced by the use of multiple social media platforms (including Twitter and Facebook); multiple social media accounts; ≥1 social media post per day; appealing formats (including infographics or other visual media, blogs, and links to articles); professional social media marketing and scheduling tools and involving relevant and influential people, organizations, and events in social media campaigns.

Our findings that social media was beneficial for the dissemination of research evidence to practitioners concurred with existing literature about the largely positive impact of social media on dissemination [10,21-23,28,65,90-93]. Quantitative analyses in other studies have revealed positive correlations between social media use and the dissemination and impact of health research evidence [65,91-93]. Narrative and qualitative reviews have highlighted benefits of social media for clinicians including connectedness and network accessibility to all (particularly with increasing use of communication technology and mobile apps in practice and day-to-day life), the large audience of practitioners and policy makers that uses social media (particularly for information acquisition and educational purposes), and the effectiveness of social media to deliver clinical guidelines and research evidence–based information that could be implemented in practice [21-23,28,90]. Reviews also highlighted the challenges of synthesizing social media effectiveness for the dissemination of information due to the heterogeneity of how social media are studied and used [22,28,90].

Social media campaigns to disseminate research evidence to health care practitioners should consider target outcomes because specific features enhanced the 4 outcomes of interest differently (Table 1).

Our quantitative findings in favor of using multiple social media platforms, including Twitter (for engagement in particular) and Facebook, concurred with existing literature that recommends using a range of social media platforms [23,28,90,94] and highlights the prominence of Twitter for research dissemination [28] and Twitter and Facebook for e-professionalism [22]. New knowledge from our systematic review includes the potentially beneficial effect of using multiple social media accounts on the same platform. This may also link with the apparent dose-response relationship between effectiveness and the average number of social media posts per day.

There was consistent evidence to support the use of visual media, particularly infographics for better reach, engagement, and direct dissemination; our findings suggest that social media effectiveness may also be enhanced by other post formats including podcasts, blogs, questions, and practice points to disseminate research evidence on social media. This adds to previous literature, which recommended posting a range of appealing multimedia that is accessible, useful, relevant, authentic, and credible [10,23,28,90]. Our review also highlighted the importance of including links to original research; comprehensive infographics or practice points posted on social media might reduce the likelihood of viewing the original article by link click [49]. Nevertheless, including links offers the viewer the opportunity to check the authenticity and credibility of the information in a post. Including links may also facilitate the delivery of simple, clear, and practice-relevant messages without scientific language [88] because the main message from the research can be easily understood and accessible to all practitioners, while further details can be sought by accessing the link.

Our findings corroborate existing recommendations to identify and involve key influencers, organizations, events, communities, #hashtags, and @mentions and to use professional tools to plan sustained, scheduled, and regular posts to overcome the transient nature of social media [10,23,28,90]. Our systematic review extends existing recommendations by identifying that posting at least once a day on average (often achieved using multiple accounts, platforms or both) was more effective than less-intensive social media strategies, suggesting a dose-response relationship between post frequency and social media effectiveness. This dose-response relationship may explain the negligible effect of one journal’s social media campaign tested in 2 RCTs included in this systematic review [42,43].

The optimal timing of social media posts and campaigns for the dissemination of research evidence to practitioners is contentious in existing literature [53,95,96]. Optimal timing may be outcome dependent; in this review, different times of the day and different days of the week were more effective depending on the outcome measured (Table 1). Similarly, while reach and engagement may be enhanced by established social media initiatives extending over months or years, impact may be best achieved in shorter, targeted social media campaigns over days or weeks.

Our findings suggested a tendency for greater impact on thinking and practice from social media featuring targeted topics or specialist areas of health care. This concurred with previous reviews that found that the dissemination of clinical guidelines or clear, evidence-based behavior change messages, such as practice points, may increase impact with health care practitioners [90,94]. This also resonates with recommendations in the existing literature to consider the target audience when selecting platforms, formats, and strategies that optimize content for the dissemination of research evidence [23,88,90,94]. For instance, the choice of platform and content about pediatric colorectal cancer should be guided by the understanding that, in this field, Twitter is typically used to share research evidence, Facebook is used for support offered by nonprofit organizations, and Instagram is used for sharing personal stories [67].

Our findings suggested a tendency that social media was more effective on impact (in citations) when used to share research articles about broad topics. Thus, there may be interactions between the content of research evidence posts and the effectiveness of dissemination on social media, regardless of platform, format, or strategy. Indeed, there are indications in the wider literature that research evidence source, topic, and post content may influence how effectively it can be disseminated on social media. For example, geographically, compared to the United States, authors from Europe, and UK clinical guidelines, achieve better altmetric attention scores and citation rates [58,97]. Published articles may achieve greater reach, research conference posts may receive greater engagement [73], and clinical guidelines may better influence thinking and practice [90]. Research evidence that aligns with “hot topics” on social media may also achieve better reach and engagement [68,98]. Engagement, direct dissemination, and citation rates may be better for open-access articles (especially reviews) that are recently published with shorter titles (which are provocative, interrogative, or declarative and free from methodological description), with a greater number of authors, and by higher impact factor journals (often with a larger social media presence in terms of followers and number of tweets per month) [54,58,59,65,99-101]. Depending on the health topic, posts with humor, shock value, inaccuracies, rumors, or emotional content might achieve better reach [66,102,103]; posts with practical guidance may be more likely to be shared [66,102,104,105]; and short videos with positive titles might receive more likes and comments [104].

The key strengths of this systematic review are the quantitative meta-analytical methods used that allow robust conclusions based on cumulative evidence about the effective use of social media for the dissemination of research evidence to practitioners. Methodological limitations have been discussed in the published protocol [29]. The search strategy was comprehensive, using multiple reviewers to ensure reliability and comprising a range of study designs including RCTs. However, no studies were identified investigating the effectiveness of social media for the dissemination of research evidence to social care practitioners. There were also relatively few RCTs, and the mixed quality of the included studies reduced the certainty of evidence about effect sizes for some outcomes. A wide range of social media interventions and research evidence content were represented by the included studies; this reduced the risk of confounding by topic, source, or content. However, time may have reduced the consistency of findings between included studies because the research and social media landscapes are rapidly evolving and have changed significantly from 2010 to 2023 [11,14]. Furthermore, the time frame of outcome measurement can influence potentially transient social media effectiveness [55]. Effectiveness was evaluated thoroughly in 4 outcome domains, comprising 9 outcomes of interest and multiple measures; this helped to accommodate the variability of reporting and design in the included studies. Although the diversity of the included studies threatened the consistency of effect sizes, the direction of effect in favor of social media was consistent. The heterogeneity of findings was ameliorated using a random effects model for more conservative estimates of effect size than a fixed effects model. Using 4 outcome domains added nuance to the existing understanding and facilitated the development of clear suggestions about how to optimize social media effectiveness for the dissemination of research evidence to health care practitioners. However, a cautious interpretation of a causal relationship between dissemination effectiveness and specific social media tactics is required. Effects may have been inflated by other confounders; for example, larger organizations with greater resources for public relations not only can post more frequently on social media but also may have greater reputational influence and share higher-quality research.

In conclusion, social media were effective for disseminating research evidence to health care practitioners. Large effects of and after social media interventions on measures of direct dissemination were particularly evident. There may be a dose-response relationship between the intensity of the social media campaign and its effectiveness. Selected social media intervention characteristics including platforms, formats, and strategies may enhance reach, engagement, direct dissemination, and impact of research evidence for practitioners. Future research directions include repetition of this review to keep up with the rapidly evolving use of social media for research dissemination; quantitative testing of the potential dose-response relationship between dissemination effectiveness and social media frequency and intensity; and further evaluation and exploration of how different practitioner groups, particularly social care practitioners, use social media to access research evidence.