Research Letter
September 29, 2023
Randomized Clinical Trial Visual Abstract Display and Social Media–Driven Website Traffic
N. Seth Trueger, Eman Aly, Sebastien Haneuse, et al
JAMA. Published online September 29, 2023. doi:10.1001/jama.2023.16839
Since 2016, medical journals have been using visual abstracts (VAs) to promote research on social media.1 VAs use images and text to efficiently convey a study’s methods and findings. Prior small randomized trials of individual specialty journals have reported VAs increase engagement on Twitter (now called “X”).2-5 This study assessed whether VA use across multiple journals and specialties with different readerships within the JAMA Network, including the flagship journal JAMA, was associated with measures of social media engagement.
Methods
This study focused on all randomized clinical trials (RCTs) published with VAs in JAMA Network journals from September 21, 2021, through May 15, 2022. Each journal had its own Twitter and Facebook accounts. We developed social media posts comprising a text summary of the RCT conclusions, an article link, and 1 of 3 images (eFigure in Supplement 1): a linking VA (clicking the image opens the article in a web browser), an expandable VA (clicking the image expands the image to full screen), or an article figure or table (Results figure if present, ie, usual practice for non-RCTs and journals that do not use VAs). The 3 identical posts varying by image were published once to each platform 1 minute apart in random sequence 1 hour after RCT publication, remaining online indefinitely. The prespecified primary outcome was link clicks to the article because they are discrete measures of active engagement and social media–driven site traffic was a JAMA Network priority. Secondary outcomes were platform-reported impressions and reach (measures of how many people saw each post on their timeline); and “all other engagement” (Twitter: sum of replies, retweets, detail expansions, likes, profile clicks, hashtag clicks, and follows; Facebook: sum of comments, shares, reactions, and clicks on “see more,” profile page, or profile photo). All outcomes were cumulative measures reported by Twitter and Facebook 7 days after publication. Subgroup analyses included outcomes by platform and journal category (non-JAMA journals, specialty journals [all Network journals except JAMA and JAMA Network Open], individual journals). Differences were analyzed with nonparametric Kruskal-Wallis testing in R version 4.1.1; statistical significance was set at 2-sided P < .05.
Results
Among 205 RCTs with VAs published across 12 JAMA Network journals, link clicks were statistically significantly higher with linking VAs (median, 18 [IQR, 4-64]) than with expandable VAs (median, 11 [IQR, 3-38]) or article figures (median, 9 [IQR, 2-40]); P = .02) (Table 1). Impressions/reach and engagement were statistically significantly higher with expandable VAs than with linking VAs or article figures (impressions/reach median [IQR]: 3344 [1102-17 008] vs 1874 [685-6874] vs 2082 [765-7957]; P < .001; engagement median [IQR]: 43 [12-162] vs 24 [7-76] vs 21 [6-77]; P = .001).
In prespecified subgroup analyses, link clicks were statistically significantly higher with linking VAs (vs other display types) on Twitter but not Facebook; the increases in impressions/reach and engagement with expandable VAs were evident on Facebook but not Twitter (Table 2). JAMA was the only journal with a statistically significant increase in link clicks with linking VAs; some specialty journals, notably JAMA Oncology and JAMA Cardiology, had increases in impressions/reach and engagement with expandable VAs.
Discussion
Linking VAs were associated with increased link clicks compared with expandable VAs and article figures. Expandable VAs were associated with increased impressions and engagement compared with linking VAs and article figures. The increases differed within platform and journal categories. Their magnitudes were comparable to those in previous RCTs of VAs.2-5Limitations include the artificiality of 3 near-simultaneous posts; use of only Twitter and Facebook; the platforms’ opaque algorithms, which may treat links or expandable images differently; and use of platform-reported (rather than website) metrics.6 Interpretation of subgroup analyses is limited by multiple comparisons and small subgroup sizes. Although user engagement with RCT social media posts likely benefits journals and authors, the editorial importance of the available measures and the differences observed are unclear, and the benefits must be balanced against the investments needed to produce accurate high-quality VAs.
