Social media interventions designed to promote cancer awareness and screening were identified using a scoping review guided by the methodology of Arksey and O’Malley [28] and expanded upon by Levac et al [29] and Peters et al [30]. We followed the process outlined in the published protocol and followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) [31,32].

An experienced research librarian helped to develop the search strategy, which included a combination of subject headings and keyword searches. We identified articles by searching five databases: MEDLINE, PsycINFO, Scopus, Web of Science, and CINAHL (Cumulative Index to Nursing and Allied Health Literature). Multimedia Appendix 1 shows the full search strategy used for MEDLINE, which was adapted for the other databases. Additional articles were identified by conducting internet searches for relevant published material and by hand-searching reference lists of included articles. We searched the grey literature for relevant reports not published in peer-reviewed journals. We also contacted the authors of conference abstracts that met inclusion criteria to see if we could include any unpublished results. Organizations and charities related to cancer screening were contacted via email for any relevant published research reports.

We considered peer-reviewed articles and non-peer-reviewed reports from the grey literature, as this was a scoping review that aimed to be inclusive and to explore the breadth of relevant research. The findings will help to inform future systematic reviews of the literature that will also assess the quality of the research. Additionally, non–peer-reviewed articles were included because many evaluations of social media interventions are not published in peer-reviewed journals. Social media is also a rapidly evolving area and current insights might not be captured in peer-reviewed literature, where there is typically a lag time between evaluation and publication. We excluded articles and reports not published in English, as we could not feasibly translate the results into English in a valid and reliable way. Articles and reports that discussed social media platforms, including Facebook, Twitter, Instagram, YouTube, Pinterest, and Snapchat, were eligible for inclusion. Articles relating to social media platforms popular in non–English-speaking countries were not excluded, but as a consequence of limiting our search to articles written in English, the focus of this review is on platforms commonly used in English-speaking countries. We included articles published from 2004—as this was the advent of widespread social media use of these platforms—to June 2019 [33]. The reported findings from mass media campaigns were included if social media was the primary focus of the article. We included articles that discussed both qualitative and quantitative methods. We only included articles about interventions that directly targeted cancer screening and early diagnosis, where the primary message of the intervention focused on raising awareness of cancer symptoms, cancer screening, or promotion of help-seeking for potential cancer symptoms. Articles that reported findings solely on the human papillomavirus vaccine were excluded, as wider issues regarding attitudes toward vaccination and misinformation were considered out of the scope of this review. We excluded articles if the participants were cancer patients, survivors, or health professionals. Articles that evaluated static internet pages, such as blog posts, were also excluded from the review.

All identified articles underwent two stages of screening: title/abstract screening and full-text screening. Three researchers (APK, AC, and SR) divided the articles and screened them against the inclusion criteria. Each article was independently reviewed by another researcher (AK). If the eligibility of any title/abstract was unclear, it was included in the full-text screening and any discrepancies were reviewed by an additional author and resolved in a consensus meeting. Interrater reliability for title/abstract screening was good (Cohen κ=0.69) [34]. Two researchers (RP and AK) piloted the data extraction approach and three researchers (RP, APK, and AC) completed the data extraction. We extracted data using a Microsoft Access database to collect key information on article characteristics, details of the interventions, methodological approaches, outcome measures, and key findings. A second researcher (RP, APK, or AC) verified the data extractions, and discrepancies were resolved in regular meetings with the entire team.

Table 1 describes the outcomes extracted from the included articles. These were categorized using an adapted version of Neiger et al’s [35] key performance indicators and metrics related to social media use in health promotion, which included insights, exposure, and reach, as well as low-, medium- and high-level engagement. Outcome measures capturing cancer knowledge and intention to attend cancer screening were not accounted for using the original framework but were categorized as high-level engagement, as greater knowledge and intentions are precursors to behavior change [36,37]. We also extracted information on the nature of information that was delivered and shared in interventions by users and developers, such as how many posts were related to health issues and fundraising. Additional data extraction was undertaken by one researcher (RP) to understand the mechanisms by which interventions might promote behavior change by using the behavior change technique (BCT) taxonomy developed by Michie et al [38]. BCTs were identified from the articles but also from campaign websites where possible. Following the scoping review methodology [28-30], we consulted with six people with experience developing and evaluating social media interventions for cancer screening, as well as a cancer patient, to validate our findings [39]. We asked them to comment on our preliminary findings, and their input helped to shape the narrative synthesis of the data.

Based on the initial search, 1029 articles were identified after duplicates were removed. We screened a total of 183 full-text articles and included 23 of those articles in this review. Figure 1 outlines the selection process.

Multimedia Appendix 2 outlines the characteristics of each included article [40-62]. Only 2 of 23 (9%) articles were from the grey literature and were charity campaign reports; the rest of the articles were published in academic journals. First author institutions represented 10 countries, most commonly the United States (11/23, 48%), Canada (3/23, 13%), and the United Kingdom (2/23, 9%).

Multimedia Appendix 5 provides a breakdown of the methods used in each article. In most of the included articles (20/23, 87%), quantitative research designs were used; one-quarter (5/20, 25%) of those were experimental research designs. A single-group posttest design was the most commonly used experimental design (3/5, 60%) [40,43,44]. Two articles (2/5, 40%) that used experimental designs used two-group pre- and posttest research designs [47,54]. Most (15/20, 75%) of the quantitative articles used observational research designs, where the researchers observed the impact of previously developed interventions. Just over one-half (8/15, 53%) of these were longitudinal studies. One-half (4/8, 50%) of the longitudinal studies measured outcomes over a period of 1 to 2 years [50-53]. Two (2/23, 8.7%) articles used qualitative research designs [55,57] and thematically analyzed social media messages. One (1/23, 4%) article used mixed methods by conducting a single-group pre- and posttest design and thematic analysis of social media messages [61].

Multimedia Appendix 5 shows the outcomes measured for each included article. Most (18/23, 78%) articles measured at least one of Neiger et al’s [35] key performance indicators related to social media use in health promotion. Two (2/23, 9%) articles assessed insights by measuring user opinions [40,43]. Six (6/23, 26%) articles measured exposure, the most common indicator being impressions (the number of times a post was viewed; 4/6, 67%) [40,41,49,62]. Three (3/23, 13%) articles measured reach and the demographics of followers [40,49,52]. Five (5/23, 22%) articles measured low-level engagement by measuring the number of likes on a post [40,46,49,50,53]. Many (11/23, 48%) of the included articles measured medium-level engagement, with the most common indicator being the number of posts/tweets by users (8/11, 73%) [42,45,48,51,52,59,62]. Five articles (5/23, 22%) measured indicators of high-level engagement. Two (2/5, 40%) of these articles measured knowledge of cancer symptoms and screening [47,54] and one (1/5, 20%) article measured intention to get screened [40]. One (1/5, 20%) article measured participation in an offline advocacy event as a volunteer [41] and one (1/5, 20%) measured the number of people who attended screening [61].

Most studies of social media interventions have evaluated national cancer awareness month campaigns, using observational studies to measure exposure, reach, and low- to medium-level engagement with a campaign. A small number of studies suggested that regional cancer awareness month campaigns and targeted interventions might improve cancer awareness, as well as screening intentions and uptake. There was evidence that exposure, reach, and engagement with the interventions varied by age, gender, and ethnicity of users, and also by cancer type.

This scoping review was the first to focus on social media evaluations of interventions to improve cancer screening and early diagnosis. It added to the literature by highlighting the limited number of robust evaluations that captured high-level engagement/behavior change, such as attendance at cancer screening. Evaluating high-level engagement is challenging because timely observational data on cancer screening attendance can be difficult to access and link with social media data [24]. Evaluation is also challenging because social media interventions are often designed without evaluation in mind [3]. During the consultation for this review, experts commented that campaigns are set up so quickly that there is not always time to consider evaluation. Improving cancer screening and early diagnosis is seen as a long-term goal that will take many years to realize, so immediate changes might not be expected or measured. A comprehensive evaluation framework that incorporates elements from behavior change theories and social media engagement frameworks could foster more robust evaluations that capture outcomes that demonstrate impact on behavior change and engagement [17,63,64]. However, as noted during the consultation for this review, there are further challenges to evaluation, such as the difficulty of demonstrating that a specific campaign caused a change in outcomes, as well as the limited time and resources of organizations to conduct evaluations.

This review was the first to use the BCT taxonomy to identify a variety of BCTs that social media interventions used to change health behaviors, including social support and providing information about health consequences [38]. One article reported restructuring the environment to provide better access to cancer screening [61]. As noted during our consultation, governance and data protection issues often limit the ability of health providers to use social media to improve access to care. These issues need to be addressed if interventions are to be tested or implemented on a larger scale. Many articles also reported that information delivered and shared during national cancer awareness month campaigns consisted of more non-health messages relating to fundraising than health messages relating to cancer. For many of these campaigns, the theory of change may be that fundraising messages increase cancer awareness, thereby increasing help-seeking or uptake of screening. Behavior change theory suggests that providing actionable health messages, such as information on cancer symptoms, could influence behavior change more directly than fundraising messages [12,38,63,65,66]. Future evaluations are needed to test our assumptions about how national campaigns might lead to behavior change and what messages would be most effective.

We identified a need for more social media interventions targeted toward colorectal and lung cancer, as most of the articles in this review were evaluations of social media interventions for breast and prostate cancer awareness and screening. Survival rates for lung and colorectal cancer are poorer than for breast and prostate cancer, which is partly because of poor uptake of cancer screening and delayed help-seeking that can lead to a delayed diagnosis [67]. Previous studies [68,69] have shown that there is more stigma around lung cancer than other cancers, and higher perceptions of cancer stigma are associated with delays in seeking medical care. Social media interventions could play a key role in changing social norms and stigma around help-seeking and screening for these cancers [70]. BCTs used by current breast cancer campaigns, such as sharing personal stories, could help to create social support and influence how people view these cancers, which in turn could encourage help-seeking behavior, increase screening uptake, and improve health outcomes.

This review also added to the literature by exploring to what extent inequalities in cancer screening and early diagnosis were measured and potentially addressed by social media interventions. We found some evidence that social media interventions have poorer reach and engagement with ethnic minorities, but there was no information on engagement with other minority groups [52]. Individuals from ethnic minorities might have less interaction with social media campaigns and not seek out cancer screening information because they have less access to cancer screening and higher cancer stigma [68,71-74]. The written and visual communication in social media interventions may also exclude ethnic minorities if the information is only available in English, presented in inaccessible language, or framed in a way that is unrelatable [74]. A lack of social media influencers or campaign role models that resonate with ethnic minorities may also make it less likely that they undertake a behavior, as suggested by social cognitive theory [75]. Future evaluations of social media interventions should measure inequalities in exposure, reach, and engagement and consider their success directly in relation to the groups that they were seeking to target. Targeting interventions toward those with a disproportionate disease burden could help to improve health inequalities seen in cancer screening and early diagnosis. However, social media interventions will have little impact on those who do not use social media, who may also be those in greatest need of information on cancer screening.

Some differences in the use of social media platforms by gender and age were found in this review [49,62]. Facebook content reached more women than men, and reached older-aged adults, and YouTube content reached more men than women [40,43,49]. This is consistent with recent data on social media use in the United Kingdom by age and gender [76]. This highlights the importance of identifying which platforms target users are currently more likely to engage with when designing social media interventions. However, as noted during our consultation, social media is a constantly changing landscape, so messaging needs to be continually updated and transferred to different platforms. Targeting messages also has potential risks, such as the threat to privacy and ethical issues, and it often requires payment and significant time and resources [17,24]. Additionally, data may not be available for particular target users; for example, we only found evidence to show how best to target those who identify as male and female. Previous research has shown that there are gender identity disparities in cancer screening, and trans and nonbinary individuals could benefit from more information regarding cancer screening and early diagnosis [77,78]. Further research is needed to understand the most effective way of targeting social media interventions toward these individuals.

We acknowledge that scoping reviews have several limitations, but a scoping review allowed us to gain a wide-ranging understanding of the role of social media in cancer screening. Research into social media is rapidly growing and this scoping review is a snapshot of evidence for social media interventions at a particular time [26]. Furthermore, as we only selected studies written in English that did not include information about some popular social networking sites in non–English-speaking countries, the findings of this review might not be generalizable. Many of the social media campaigns in this review were also part of multimedia campaigns. Therefore, it is difficult to know whether changes in engagement with social media or health behaviors were due to the social media element of those campaigns and how social media interacts with other aspects of the campaign. Additionally, the coding of BCTs in this review was dependent on reported content and online sources, so there was insufficient detail to identify all techniques used. The review was also dependent on what outcomes the evaluations chose to measure and report. There is currently no protocol for how to report evaluations of social media interventions, so there may be some degree of reporting bias in included articles. Future research would benefit from the development of a reporting protocol based on current frameworks for evaluating social media research [35].

This review found that most evaluations of social media interventions to improve cancer screening and early diagnosis did not report behavior change outcomes. The limited available evidence suggests some types of social media interventions may improve cancer awareness and intended and actual uptake of screening. Use of evaluation frameworks and reporting guidelines could help future researchers to plan robust evaluations of social media interventions that capture outcomes of behavior change and explore how these interventions work. Future evaluations could also measure who engaged with these interventions to assess whether social media interventions for cancer screening and early diagnosis can address some health inequalities. Interventions focusing on cancers that have received less social media attention, such as colorectal and lung cancer, could help to influence social norms around help-seeking and screening uptake for these cancers, which could improve health outcomes for patients.