Social media in undergraduate medical education: A systematic review

There are over 3.81 billion worldwide active social media (SoMe) users. SoMe are ubiquitous in medical education, with roles across undergraduate programmes, including professionalism, blended learning, well being and mentoring. Previous systematic reviews took place before recent explosions in SoMe popularity and revealed a paucity of high‐quality empirical studies assessing its effectiveness in medical education. This review aimed to synthesise evidence regarding SoMe interventions in undergraduate medical education, to identify features associated with positive and negative outcomes.


| INTRODUC TI ON
The explosion of Social Media (SoMe) has infiltrated all aspects of modern society. The scale of the phenomenon can be daunting to conceptualise. There are over 3.81 billion worldwide active users, 1 while individuals have an average of seven accounts each. 2 Facebook, the most popular platform, boasts over a billion more users than the population of China, the world's most populous country. 1,3 It is perhaps inevitable that this technology would become ubiquitous within medical education delivery. Our field craves innovation and strives to adapt its future workforce to changing environments. 4,5 There are few arenas as chaotic as SoMe: since its conception, hundreds of platforms have risen in prominence before falling into obscurity. 6 Various platforms have been utilised within undergraduate medical education curricula, with roles in professionalism, 7 blended learning, 8 student well-being 9 and mentoring 10 representing a fraction of their application.
The argument for the integration of SoMe into educational practice is supported by multiple theoretical standpoints, including the development of rapid, accessible communities of practice, 11 based upon constructivist principles, generated during classes, 12 conferences 13 or conversations. 14 Moreover, connectivism is a 'new' educational theory, mirroring constructivism, born out of such technology enhanced learning. 15 Research to date examining the impact of SoMe use on medical students' academic attainment has found no relationship between daily usage, platform preferences and performance on summative assessments. 16 The majority of medical students have SoMe accounts. This has resulted in a powerful tool that can reach virtually every student on platforms already integrated into their daily lives.
As educators, it is no longer a question of whether SoMe has educational applications, or whether undergraduate populations are actively using these platforms for their learning. Rather, we should be asking how best to utilise SoMe, and whether such platforms can facilitate specific outcomes.
Previous systematic reviews of SoMe in medical education have investigated outcomes and efficacy. [17][18][19] However, these took place before the SoMe 'boom' of recent years, revealing a paucity of high-quality empirical studies assessing the effectiveness of SoMe in medical education. An updated review, synthesising contemporary SoMe scholarship, is required for the current landscape.
In the advent of the COVID-19 pandemic, a once-in-a-century event, the medical education community has transformed. 20 There have been calls for a new paradigm of educational delivery through technology-enhanced learning, 21 with SoMe at the forefront of this revolution. 22 It is imperative that, before committing to placing SoMe as a foundation of any new approach, we critically examine the evidence for its efficacy.
This review aims to synthesise the evidence regarding SoMe interventions in undergraduate medical education, in order to identify features associated with positive and negative outcomes.

| ME THODS
This is a systematic review reported in accordance with the STructured apprOach to the Reporting In healthcare education of Evidence Synthesis (STORIES) statement. 23

| Search strategy
We performed an electronic search of 31 terms and their Boolean combinations (illustrative full terms for one database are provided in Appendix S1) through seven databases: Medline, Cumulative Index of Nursing and Allied Health Literature (CINAHL), British Education Index (BEI), Education Resources Information Center (ERIC), Embase, PsycINFO, Applied Social Sciences Index and Abstracts (ASSIA), and the Australian Education Index. Search terms were derived through reviewing keywords of papers identified through a pilot search, and a list of social media platforms. No limits were imposed. Searches were performed between 16th June and 16th July 2020 by one reviewer (ER).
We scrutinised reference sections and performed forwards citation tracking of all included papers using PubMed's 'cited in' feature, and previous pertinent reviews in order to identify further relevant papers. One reviewer (JG) hand searched the full contents of one journal (MedEdPublish) that was considered relevant but not indexed in the above electronic databases. This journal was chosen as studies identified in the reference section of identified papers were published here, and we were aware that it was not indexed in any of the databases we had searched.
All citations were downloaded and imported into web-based systematic review software (DistillerSR, Evidence Partners, Conclusions: Social media is enjoyable for students who may improve short term knowledge retention and can aid communication between learners and educators.
However, higher-quality study is required to identify longer-term impact upon knowledge and skills, provide clarification on professionalism standards and protect against harms.

|
Ottawa, Canada) in order to facilitate screening and data extraction.

| Inclusion and exclusion criteria
For inclusion, articles had to provide primary data on the use of SoMe amongst undergraduate medical students. Studies describing SoMe interventions and research on exposure to existing SoMe were both included. Only English language papers were included.
Defining SoMe has become increasingly challenging as platforms diversify. We accept in principle the definition of SoMe articulated by Cheston et al, later adopted by Sutherland and Jalali, describing SoMe as "Web-based technologies that facilitate multi-user interaction that goes beyond fact sharing". 17,19 Given the evolution of SoMe since the development of this definition in 2013, more recent examinations should be considered. Chan et al 24 highlight that platforms "facilitate creation and distribution of content", which may be "user-generated or usercurated" to "virtual communities of practice". We would therefore add to Cheston et al's definition that 'such interactions may take place in either public or private domains', stipulate that 'the primary purpose of the platform must be multi-directional interaction' and 'be centred around content creation, curation and community'. Using our adapted definition, we considered true SoMe platforms to place user interaction at their heart. We, therefore, exclude websites or blogs which happen to feature a comments page, as "multi-directional interaction" is a secondary purpose in these instances. We also exclude podcasts on the basis that they are generally unidirectional in nature. YouTube was included as user creation is often driven by community comments, users may curate content and video replies are common.
We defined undergraduate medical education as any educational aspect of the period between the commencement of medical school and graduation.

| Screening of studies
All titles were independently screened by two reviewers (from MU, AA, OB, JA, AO, TC). Reviewers prioritised sensitivity over specificity at this point, so any titles that could potentially have been relevant to our review objective were included. Disagreements progressed to abstract screening in order to enable a more informed decision to be made.
Abstracts of all papers included from title screening, and all papers identified through reference and hand searching were screened against inclusion criteria by two reviewers (from MU, AA, OB, JA, AO, TC). If it was not clear from the abstract whether the paper met the inclusion criteria, it was included for full-text screening.
Disagreements at the outcome or criterion level (ie both reviewers elect to exclude but based on different exclusion criteria) were resolved by a third reviewer (ER or JG).
Papers that met inclusion criteria in abstract screening, or for which insufficient information was presented to enable a judgement, were read in full and reviewed again against the review's inclusion and exclusion criteria. Full-paper screening was conducted independently by two reviewers (ER & JG). Conflicts were resolved through discussion.

| Data extraction
A pilot data extraction form was developed. All reviewers independently reviewed two papers and met to ensure we were extracting consistently and to revise the fields in the coding form. Once the form was finalised, all reviewers independently reviewed a further two papers and met to ensure consistency. Following this pilot phase, all papers were reviewed by two reviewers (one from MU, AA, OB, JA, AO, TC and either ER or JG).
The following data were extracted: • Participants (country, profession, stage of training, number) • Intervention (focus, aim, brief overview, duration, SoMe platforms used) or Exposure (research question, overview of exposure, SoMe platforms investigated) • Evaluation methods (impact of intervention using modified Kirkpatrick's hierarchy, 25 study design, data collection methods) • Results (summary, results for each Kirkpatrick level investigated, key conclusions) We calculated kappa values for full data extraction and quality assessment.

| Assessment of methodological quality
All included papers were assessed for their methodological quality using two or more tools.
Depending on whether papers employed quantitative and/ or qualitative methods, they were assessed for their methodological quality using the Medical Education Research Study Quality Instrument (MERSQI) and/or the Standards for Reporting Qualitative Research (SRQR) instrument, respectively. The MERSQI is a tool containing 10 items in six domains: study design, sampling, type of data, the validity of evaluation instrument, data analysis, and outcomes. 26 The SRQR comprises of 21 reporting standards for highquality qualitative research. 27 Papers describing educational interventions were assessed for risk of reporting bias using the risk of bias measure described by Gordon et al. 28 This assesses the risk of bias due to incomplete reporting of educational interventions. Included papers were assessed against five sources of potential bias using a three-point scale. Papers that provided adequate description were considered low risk of bias, those that provided some but insufficient details were rated as unclear risk  of bias, and those that did not provide any details were rated as high risk of bias. The potential sources of bias considered were the theoretical underpinning of the development, the resources required, the setting, the educational methods employed, and the content.
An overall rating of the strength of the conclusions drawn by the authors was made using the BEME collaboration's five-point scale. 25

| Synthesis of evidence
Extracted data were synthesised using narrative synthesis. This involves synthesising the findings from primary studies textually, without conducting meta analyses. 29 This approach enabled this review to synthesise findings from both qualitative and quantitative studies to provide a comprehensive synthesis of the research literature in this field. 30 The review group met virtually every week throughout the data extraction and synthesis phases to discuss evolving findings. We constructed overview findings for the subgroups of intervention studies and exposure studies. We undertook conceptual mapping to identify themes within which to synthesise and present the findings of primary studies. 31

| RE SULTS
Database searches yielded a total of 1,442 papers. A further 321 papers were identified through reference and citation searches, and 44 through hand searching. Deduplication identified 720 duplicates, leaving 1087 papers for screening. Title and abstract screening excluded 327 and 441 papers, respectively. After reviewing the full texts of 319 full papers, 112 met our inclusion criteria and were included in the final review. An illustration of record flow can be found in Figure 1. Weighted overall Kappas for data extraction and quality assessment using MERSQI and SRQR were 0.98, 0.96 and 1.00, respectively.

| Details of included papers
The studies included participants from 26 different countries across six continents: Asia (37), North America (33)
Fifty-eight out of the 112 included studies (52%) involved the development of novel SoMe teaching (intervention studies) whilst the other 54 (48%) focussed on how students currently utilise SoMe platforms in medical education (exposure studies).

| SoMe Interventions
Novel SoMe interventions were created with the aims of: (a) improving knowledge and skill development; (b) supporting curricular activities; and (c) assessing the acceptability of SoMe.

| Exposures to SoMe
Of the 54 included studies addressing exposures to SoMemediated medical education, the most prevalent focus was assessing usage patterns of SoMe platforms by students, seen in 38 studies (70%). Projects looked to identify SoMe platforms and services most commonly used [89][90][91][92] and the duration or purpose of such use. [93][94][95] The second major theme was that of professionalism (17/54, 31%). In particular, student and faculty behaviour online was analysed for whether these behaviours met regulator standards. 96,97 Questions were raised as to whether SoMe makes students and doctors more likely to have lapses in professional behaviour, 98 whether guidelines in professional SoMe use are necessary, 99 and who should be responsible for producing and leading them. [100][101][102] Several studies explored professionalism in the context of acceptable patient interaction. 103,104 The findings from included studies have been grouped thematically into six groups: usage, acceptability, educational activity, academic performance, professionalism, and health risks.

| Usage
SoMe use by undergraduate medical students is widespread. 105 Facebook was identified as the most popularly used platform amongst this group, 90,94 although YouTube and WhatsApp were the most favoured platforms for educational content. 89 Between a third and a half of students used social media for educational purposes on a daily basis. 89,90 Frequency of social media usage was not associated with gender or academic year. 106 Stopping the use of Facebook was associated with a sense of loss. 107

| Acceptability
SoMe were generally perceived favourably by medical students.
Enjoyment and engagement with various SoMe platforms were described. Only two studies differed: one found students' opinion on SoMe efficacy to be 'divisive', 90 whilst another described less than half of their student population finding a SoMe intervention useful in their studies. 58 Usability was key to effective SoMe use in medical education. 14,36,63,77,108,109,110 Familiarity with SoMe platforms was a mediator of success. Indeed, unfavourable student reviews were reported when unfamiliarity was highlighted. 58 One study with a bespoke SoMe platform reported that students required greater accessibility and a more user-friendly interface. 53 Technological issues and poor digital literacy may contribute to inconsistencies in the effectiveness of SoMe interventions. 62,66 The amount of information presented via a SoMe platform contributed to its effectiveness. Students preferred short content 44,77,81,85 with the "time-consuming" nature of searching SoMe for relevant information contributing to ineffectiveness. 111,112 It is suggested that the need to regularly check SoMe to keep up to date with content contributes to such inefficiencies; however, push notifications may improve this. 38,63 Type of educational information was also important, with many studies reporting a preference for SoMe posts that inform and test, for example, quizzes 109,113 and images with missing labels. 64 Visual posts with images 85 and videos 114 were suggested to be effective methods of disseminating information on SoMe, with one study reporting that highly visual platforms such as Instagram and Snapchat are the most popular amongst students for medical education. 115 One study argued that SoMe platforms may be perceived as more user friendly and less academic than institutionally designed solutions. This study used Facebook and YouTube to deliver peer developed resources to alleviate stress and depression amongst recent entrants to medical school. 70 Contributions were monitored by a trained mental health professional.

| 1231
One concern with SoMe in medical education was the lack of critical appraisal performed by students on the information presented. 75,84,108,116 Content creation by faculty members 61,76,108 may be a solution to this but adds to the workload of the educators. 114 Many studies argue that training is required for both staff and students to capitalise on the benefits of SoMe interventions in medical education. 8,49,62,75,79,101,117,118

| Educational activity
Social media platforms were used in myriad ways. Educators used them to share resources, establish a dialogue with students, and facilitate classroom activities. Students used them for informal conversation, for accessing and sharing educational resources, to arrange educational and social events, to discuss opinions, and to participate in surveys, quizzes and educational games. 90,109,110,113 Several studies delivered structured teaching using social media platforms. 6,36,48,65,85 Use of social media within structured learning activities was found to improve communication and participation, 82,88 and facilitate teamwork. 62,82 Several studies reported that SoMe initiatives made students feel more able to ask questions to their peers, compared to asking questions in a clinical setting, 87 in part driven by an expectation that they would generally receive answers to questions more quickly from a multitude of voices, especially international professionals. 74,119 Students reported a change in the student-educator relationship. 14,37,120 They perceived SoMe as having "flattened hierarchies", allowing students to feel more comfortable interacting with educators. 14,121 Similar improvements furthered peer communication and working, 62,88 with facilitators in one study noticing greater student collaboration with SoMe. 62 Students reported newfound appreciation and interest for the subject material when presented with novel SoMe options for learning for example videos on YouTube 59 and discussion groups on Facebook. 64,88 SoMe were also commonly used to supplement classroom teaching. 62,67,72,73,74 One study utilised SoMe to create a 'flipped classroom' in which a case was reviewed by participants on Twitter before classroom sessions. 67 Classroom discussions were described as more efficient, and Twitter lessened the educator burden. SoMe was found to enhance communication and collaboration between students and educators 69,72,74,106,112 as well as students and their peers. 62,69,72,83,106,112,122 Students reported SoME facilitated them being more productive in their studies, 16 and enabling them to work faster. 120 Video-based SoMe platforms were considered to be particularly useful. 95,114,116 SoME were also considered useful in supporting reflective practice. 46 Finally, SoMe were considered useful in learners' professional identity formation. 53,118 3.6 | Academic performance Subjective improvements in students' self-reported performance 56,60,63,123,124 and objective increases in assessment outcomes 33,35,38,40,69,82,125 were reported, including in two studies with controlled designs. 32,80 However, it is unclear whether the differences found in the latter two studies result from the use of SoMe or other confounding variables such as weekly assessment 32 or earlier exposure to educational material. 80 Some studies showed that the rate of utilisation of SoMe positively correlated with test score, 37,39,69 however, others suggested time spent on SoMe had no impact on knowledge improvement. 126 Furthermore, most interventions tested participants' knowledge and skills acquisition in the short term, with limited findings on long-term effectiveness. 44 Knowledge retention in the long term may be poorer with SoMe learning compared to lecture-based learning. 39 Three studies demonstrated no effect from SoMe on academic performance. 34,36,61 However, one studied 'technology enhanced learning' (including SoMe), making it difficult to identify any SoMe specific results 61 and another did not assess baseline knowledge between the experimental and control group prior to intervention. 34 In one study, students linked social media use to subjectively worse test performance. 127 The studies reporting higher academic performance tended to be dialogue-focussed, using push notifications, poll quizzes and closed communities of practice. They generally used WhatsApp, Facebook Groups and WeChat.

| Professionalism
One concern with the use of SoMe was maintaining professionalism. One study assessed students' Facebook activity levels and perceptions of guidelines for professional behaviour, 128 while another tested their response to simulated medical professionalism scenarios. 129 The latter demonstrated an inverse correlation between having a personal board on PTT, Taiwan's largest SoMe platform and humanism, as well as a similar correlation between SoMe use and medical professionalism scores. 128 This was despite the fact that SoMe use was actually associated with increased awareness of the need for professional behaviour on SoMe. 128 Additionally, despite this negative correlation, some students highlighted that SoMe, specifically Twitter, allowed them to develop empathy and understand previously unknown elements of patient experience. 119 One study indicated that simply by surveying students on contentious behaviours (in this case patient-targeted googling, PTG) the act of surveying may reduce the incidence of such behaviours. 130 Students involved favoured more explicit teaching around PTG and such surveys may represent an opportunity to improve levels of professionalism.
Multiple studies found that students' behaviours were subject to change with the knowledge that their peers would be able to view the content they posted. 45

| 1233
Two articles indicated that in response to participating in the study, students would actively change their SoMe profiles to be less publicly accessible. 52,54 Even though students understood the need for professionalism when using SoMe for medical education, 49,93,107,125,132,133,134 they appreciated educational interventions with regards to this. 40,56,135,136 Studies suggested privacy concerns were a barrier to effective SoMe use 47,64,69 but closed SoMe groups or bespoke networks may alleviate such concerns. 53,74 This does, however, limit global interactions 53 and only three studies identified accessing expertise from around the world as an advantage. 74

| Methodological quality of included papers
Eighty-four (75%) papers utilised quantitative (including mixed methods) designs. The mean (SD, range) MERSQI score for these papers was 9.1 (2.3, 5-14) out of 18 (Table 1). These studies predominantly employed a single group design with a single data collection point (63, 75%), at a single institution (73, 87%), using subjective data (60, 71% Overall, the risk of bias in reporting educational interventions was reasonable (Figure 2). There were no sources of bias where over 50% of papers were considered to be at low risk of bias. The highest sources of potential bias were in reporting details regarding the settings in which interventions took place and in the educational methods used.

| Methodological quality of primary studies
It is telling that our review has included 98 more studies than a re- Our review identified a mean between those of these two previous reviews, and a lower maximum MERSQI score. While ours is the first SoMe review to use the SRQR to assess methodological qual-

|
suggests that while much of the literature on SoMe remains of relatively poor quality, research on SoMe in graduate medical education appears to be of higher quality.
Several previous studies have examined the impact of SoMe upon student outcomes, with focuses on the impact upon knowledge. Whilst Cheston's review concluded that their intervention studies led to equivalent test scores for students who did and did not use SoMe tools, 17 another found Facebook to be beneficial to the affective aspects of learning and YouTube to be an inadequate source of information for medical education. 19 However, the latter review describes its study outcomes as weak, poorly transferrable and relying on self-reporting. Our results conversely demonstrate subjective and objective knowledge improvement across several platforms, however, cautions against the relatively strong likelihood of confounders, suggesting that the benefits are likely limited to the short term. We echo the findings of Chan et al, 24 where the major-

| Strengths and limitations
While this is not the first systematic review of SoMe use in undergraduate medical education, we have identified significantly more papers than previous reviews. We believe this review has benefited from a rigorous and sensitive search strategy including multiple databases, reference, citation, and hand searching. We have benefit-

| Implications for practice and research
For the educator designing social media education tools, the best practice appears to be grounded in familiarity and features intensive induction for staff and learners. We must also urgently reflect on how we undertake research in this field. Whilst we have examined research encompassing a diverse range of rapidly emerging platforms, we suggest a reactionary approach to SoMe research is inappropriate. Future work should focus instead on the common factors across the spectrum of social media in order to address the significant gaps raised by this review.
Specifically, we recommend the study of the impact of a range of SoMe platforms upon long-term knowledge retention, largely absent from the current literature. Additionally, a robust assessment of measures to prevent the SoMe harms highlighted in this review is essential.
When considering commonalities between platforms, one must consider the underpinning theory on SoMe. SoMe platforms have their functional differences but educationally the principles are the same: these are rapid, often short-lived 6

| CON CLUS IONS
Despite an explosion of research surrounding social media in medical education, understanding this social phenomenon has not significantly progressed in almost a decade. We have established that social media is enjoyable for students, may improve short term knowledge retention and can aid communication between learners and educators. However, students and educators alike remain wary of professionalism concerns and warnings against potential SoMe harms remain.
We suggest that rather than attempting to undertake a superficial evaluation of the latest SoMe trend, the community should instead consider longer-term, higher quality research, rooted in the underpinning educational theories which unite these diverse platforms.

ACK N OWLED G EM ENTS
None.

CO N FLI C T O F I NTE R E S T
We have no declarations of interest to declare.

AUTH O R CO NTR I B UTI O N S
JG designed the study, extracted data, synthesised the results and contributed to writing the first draft of the manuscript and subsequent drafts. MU extracted data, synthesised results, and contributed to all drafts of the manuscript. AA extracted data, synthesised results, and contributed to all drafts of the manuscript. OB extracted data, synthesised results, and contributed to all drafts of the manuscript. AO extracted data, synthesised results, and contributed to all drafts of the manuscript. TC extracted data, synthesised results, and contributed to all drafts of the manuscript. ELR designed the study, extracted data, synthesised the results and contributed to writing the first draft of the manuscript and subsequent drafts. All authors approve of the final manuscript for submission.

E TH I C A L A PPROVA L
Ethical approval not sought as this is secondary research not involving human participants.