A Look Back at ACC.19: The Impact of Twitter, Hashtag Drift and Confusion
ACC's Annual Scientific Session provides an opportunity for cardiologists to meet and share advances in clinical research and practice. This study describes how information from the 2019 conference was shared on Twitter.
It quantifies the number of contributors posting original content (n=2,161), their posts (n=12,611), and dissemination of that information (n=37,249 retweets). Social media provides a route for clinicians to share advances in cardiology and reflections on these developments of interest to professionals, the public and patients.
However, the volume of information can be overwhelming and realistically requires the curation and filtering described in this study.
- Popular topics included the Apple Watch study, intervention studies and primary prevention guidance.
- Of 2,161 tweeters posting original cardiology content, 294 (13.6%) had 80% of retweets.
- Ten percent of tweeters using #ACC19/#ACC2019 hashtags during the conference period were unrelated to cardiology.
- Adding the #CardioTwitter hashtag could have avoided distractions from noncardiology tweeting.
There was a rich seam of cardiology-related tweets, but sometimes these were mislabelled (#ACC2019) and at other times unrelated tweets on basketball, comics and cricket from events using the same hashtag(s). This might have deterred some contributors.
To avoid hashtag drift and confusion in the future, conferences should include an additional hashtag that clarifies the purpose of the conference – e.g., #ACC20 #CardioTwitter – and should add this to all conference materials, from first announcement, to email correspondence, Twitter advance notifications and conference branding on programs, slide backgrounds and posters.
Science relies on innovation, dialogue, team work, revisions, publication and dissemination. Conferences are an important part of this cycle and social media has a role at each stage. Medical conferences generate considerable interest from professionals, patients and the public.
Professionals, journals, device manufacturers and pharmaceutical companies often seek to launch their work, papers and products at the largest and most prestigious conferences, reaching an international audience of key figures from the discipline.
Social media provides a way for participants at the conference, and observers from outside the conference hall, to hear about the findings, share points of interest, and make their own comments, interpretations and responses.
This potentially feeds back into the cycle of scientific discovery. Understanding the way that information from such events is shared via social media is potentially important, as it helps to understand the main areas of interest from the conference, the key influencers, and the way such information is disseminated within and beyond the conference.
This is particularly relevant for the ACC.20 conference, as the COVID-19 pandemic means that there will not be a physical meeting, but a series of virtual events are going ahead, supported by social media activity.
This study used the tweets and retweets from ACC's Scientific Session held from March 16-18, 2019 in New Orleans (ACC.19) to explore these ideas. The annual event generates a large number of tweets and retweets, and a flow of information can be identified on social media days after the event.
The official hashtag for the event was #ACC19, but there was a substantial number of tweets using #ACC2019, as commented on by tweeters before and during the conference. To complicate matters further, during the same weekend there were other events where tweeters also used these hashtags.
This study examines the cardiology tweets to identify the main influencers and most popular tweets, and to describe the pattern of tweeting and retweeting over time, including the point at which tweeters or retweeters started to engage and the duration of engagement.
Tweets using the #ACC19 and #ACC2019 hashtags were extracted between March 8 and 23 using NodeXL.1 All times are stated in UTC. Individual tweets, mentions and retweets were identified using methods described previously.2 NodeXL can extract 18,000 tweets and retweets in one go, but extracts can be combined.
Daily extracts were taken during the period of the conference, and added to longer periods before and after the conference, giving a total of 13,758 tweets. The analysis focused on the period between March 14 and 22 to allow plotting of hourly data in Excel.
There were 12,969 tweets posted by 2,393 tweeters during the study period, and these tweets were retweeted 37,511 times. The interactions between tweeters, retweeters and the accounts mentioned in tweets are shown in the NodeXL map in Figure 1.
The most retweeted posts are likely to be the most interesting, useful or perhaps controversial. On this basis, the most popular tweets across the days of the conference and subsequent days were identified using rules based on the number of retweets received, to achieve a spread of tweets across the period. The specific rules are recorded in Wakelet summaries that list the top 250 tweets from the conference days, and 100 tweets following the conference.
Tweets on topics other than cardiology using the #ACC19 and #ACC2019 hashtags during the same period included the Atlanta Comic Convention, Alberta Champions Cup, American Craft Council, Asia Crossfit Championship, Attock Cricket Club, and the most popular noncardiology related topic, the Atlantic Coast Conference basketball tournament (held March 16).
It would not be feasible to sort through more than 12,000 tweets by hand, so inclusion and exclusion criteria that could be largely applied using Excel searches were identified. The noncardiology activities tended to use the #ACC2019 hashtag.
Examples of these tweets were identified in the NodeXL Excel extract by reading individual posts that used just the #ACC2019 hashtag across the full period of the extract (n=730 tweets) to identify terms and hashtags that might identify tweets on noncardiology topics.
The exclusion terms included comic, Syracuse, the basketball player Zion Williamson, #ACCTourney hashtag and Duke (when referring to the basketball team rather than the university cardiology department).
These exclusion terms were then applied to tweets using #ACC19 (n=12,885 tweets), or both #ACC19 and #ACC2019 (n=143 tweets), using the Excel search function to sort through the large number of tweets, identifying additional tweets to exclude by noncardiology tweeters, and further hand searching and checks of excluded tweeters.
Drs. Martha Gulati and Dr. Mamas Mamas discuss Social Media use in Cardiology with Dr. Sravya Chirumamilla. https://t.co/17tXk1F8zO— Timothy Joseph (@timjosephmd) March 16, 2019
@mmamas1973 @DrMarthaGulati @ACCinTouch #FITSONTHEGO #ACC2019 #SoMeCVD #ACCWIC #WIC
The earliest instance of a retweet in a NodeXL extract records the original tweeter, allowing the identification of retweets of excluded tweeters, and these retweets were also excluded. Accordingly, accounts that retweeted a mixture of cardiology and noncardiology tweets had relevant material retained in the analysis.
After applying exclusion criteria to the tweets posted during the study period, there were 12,611 tweets (97.2% of total #ACC19 or #ACC2019 tweets posted during the period) by 2,161 tweeters (90.3% of total) with 37,249 retweets (99.3% of total). For these tweets, 294 tweeters (13.6% of total) comprised 80% of retweets, while 476 tweeters (22.0%) had no retweets.
These retweet figures used information recorded in individual tweets. The majority of individual retweets were also recorded in NodeXL (33,514 retweets, 90% of total recorded in individual tweets); the retweet records included the time of the retweet.
This information was used to describe the pattern of retweeting that followed the original tweet. NodeXL records retweets of tweets posted during the period of the extract, but also documents retweets of earlier posts during the period of study.
Both types of retweet were recorded in this analysis.
Figure 1 shows the breakdown of tweeters by type of influencer: tweeter, mentioned account or retweeter. The exclusion criteria applied to tweets and retweets successfully removed all noncardiology tweeters from the top influencer lists.
The majority of accounts engaging with the #ACC19/#ACC2019 hashtag were retweeters rather than active tweeters. The next largest category was accounts mentioned but who did not tweet or retweet themselves.
Two accounts (@ACCInTouch and @DrMarthaGulati) were among the top five for each category of influence (Figure 1).
Wakelet summaries list the individual tweets for the 250 most retweeted tweets, by day, during the period of the conference, and the 100 most retweeted tweets for the four days following the conference.3,4 The conference summary lists the top 20 tweeters based on number of retweets (produced immediately after the conference, thus before applying the exclusion criteria described for this study), and shows the number of followers.
These tweeters ranged from the ACC, top medical journals and clinics, to individual clinicians and researchers with a wide range of number of followers.
The themes emerging from these Wakelet summaries reflect the main topics of the conference (e.g., Apple Watch study, original interventional research, and the new primary prevention guidelines).
They also highlight more general points that reflect delegates' wider experience of cardiology practice (e.g., time spent on computer exceeds time spent with patients, burnout). As well as providing summaries of presentations, publications and posters, many of the tweets incorporate additional reflections and responses. The post-conference summary includes reflections from the conference overall, from social posts to highlights and single tweet summaries of main trials, again with opportunities for reflections and response.
The summaries also included basic headline figures from online social media tools. In both summaries, there were some administrative posts, from invitations to sessions to a pair of lost glasses. Individual tweets can be viewed by clicking them in the Wakelet summaries.3,4
Taking all cardiology-related #ACC19/ #ACC2019 tweeting from the study period, the great majority of tweets and retweets occurred during the three days of the conference (80% of tweets and 73% of retweets) (Figure 2).
The period of peak retweeting each day extended beyond the peak of retweeting. While hourly tweeting dropped to close to zero during the early hours of the morning (using Central Time Zone for New Orleans), retweeting rarely dropped below 100 per hour during the conference days.
While tweeting activity dropped off rapidly on March 19 (the first day after the conference), there remained considerable numbers of retweets for the two days after the conference, and some tweeting and retweeting activity continued through March 22 (and beyond, data not shown). There were new tweeters and retweeters throughout this period (Figure 3).
The time spent tweeting/retweeting during the study period (a total of 216 hours) is shown in Figure 4, based on the time of the first and last tweet posted/retweeted for each contributor. Accounts that just retweeted – the largest category – tended to have short engagement (73% retweeted just once).
Accounts that just tweeted – the smallest category – typically engaged a little longer, but 65% tweeted just once. The final category – tweeters who also retweeted – tended to engage for much longer, with 50% of this category tweeting and/or retweeting over a period spanning 74 hours or longer from first to last tweet or retweet, and 10% engaged for 158 hours or longer.
This study has identified some of the methodological challenges in organizing and studying social media activity. While conference organizers select and promote hashtags before and during a conference, commonly there will be "drift" away from the official hashtag as delegates mistype, forget or make up their own.
Including #ACC2019 in the NodeXL extracts identified hundreds of tweets posted about the conference, a small number of which ended up in the top tweet summaries of the conference.
The more pressing issue for this analysis, however, was the confusion from other events, from comic conventions to sports events. Several cardiology tweeters commented on this problem in tweets posted during the conference and this is likely to have been a cause of frustration more generally and may have led to disengagement.
This study provides an estimate of the scale of the issue: 10% of tweeters using #ACC19 or #ACC2019 hashtags during the period were not related to cardiology, posting 3% of tweets and making fewer than 1% of retweets.
The process of sifting, setting rules and checking outputs, was time-consuming but was required to have a true assessment of the impact and reach of this meeting.
Cardiology tweeting dominated these hashtags, particularly the official #ACC19 hashtag. Twitter can be an efficient way of identifying tweeters with a shared interest and sorting their posts, but conversely, competition around the same hashtag can make it a frustrating experience.
Excluding the noncardiology tweeters was important prior to studying the pattern of cardiology tweeting, particularly in studying the contributions of the included accounts, which was a main focus of this study.
A short hashtag such as #ACC19 is simple to remember and brevity would have been particularly important when the character limit of a tweet was 140. However, since November 2017 the Twitter character limit has increased to 280.
Combining the conference hashtag (#ACC20) with a more specific term (e.g., #cardiotwitter or even simply #cardiology) would be much simpler for delegates to search during the conference and study after the event. Some conferences have chosen to use the same hashtag annually, which may help with awareness and consistency when used.
There are limitations to social media analysis, in particular the reliance on a specific search term (in this case a hashtag). Posts that did not include the hashtag were not extracted by NodeXL. That also applies to tweets posted in reply to a post containing the conference hashtag.
An inspection of the top 10 tweets (by number of retweets) showed that none of the replies had included the conference hashtag. Some of these replies included additional information or reflections that would have been of interest to the #ACC19 audience. Including these replies, and exploring their meaning, would however require Twitter searches by hand and application of sociological methods.
Their omission means that the social media reach of the conference will have been even larger than identified in this study. Additionally, while NodeXL extracts the majority of retweets, there were an estimated 10% or more missing from the extract. Although some of the missing retweets will have been from tweeters included elsewhere in the analysis, there will have been some impact on the data presented in Figures 1-4.
Conversely, some noncardiology tweets and retweets may have slipped through the manual and Excel based search of tweets, though these tweets/tweeters were not sufficiently popular to meet criteria for the Wakelet summaries or the Venn diagram of influencers (Figure 1).
This study provides novel and useful data and insights into the way that information is produced and disseminated at a conference. Some of the findings are consistent with other studies. This includes the dominant category being retweeters, who share other people's content rather than generating their own.2,5
However, other findings are new either methodologically or in their application to a conference of this size. ACC's Annual Scientific Meeting is a huge event. ACC.18 had more than 16,600 delegates.6 The number of contributors, the quality of new work presented and the press interest make this one of the world's largest medical conferences.
The data available from the NodeXL extract allowed the detailed study of the patterns of social media activity, including the timing of the retweets that followed the original tweets (Figure 2), the number of new contributors as they joined in (Figure 3), and the duration of tweeting (Figure 4).
This information is not available from other widely quoted social media summaries – e.g., those produced via Symplur health care hashtags, a number of which are included in the Wakelet summaries.
In contrast to our analysis, the automated Symplur reports did not clean up the data to exclude noncardiology tweeting, and did not combine the official #ACC19 hashtag with the unofficial #ACC2019 hashtag. One online tool – Twitonomy - provides an interactive geographical map, that gives an indication of location of tweeters and retweeters based on information in the user profile.
A sample of #ACC19 tweets is mapped using this tool for March 18, 2019.7 This reveals tweeting from around the world, and may explain why retweeting did not dip as much as tweeting overnight, with retweeting spreading across the world depending on time zone. However, clicking into the map reveals that the mapping is not always entirely accurate, largely because of ambiguous information in user profiles.
The ability to study individual tweet and retweet data from the NodeXL extract, combining extracts to allow for the sheer scale of tweeting around the event, provides insights previously not possible. The different contributions made during and after the conference all play a role in exploring, discussing and disseminating new findings.
A relatively small number of tweeters (central zone of Figure 1) were the main contributors of new social media content, and many of them posted across the entire period of the conference and beyond. Their views can be as important as those presented from the main stage, so it is important for a conference organizer and the wider scientific community to know what is being written and shared.
The reflections emerging from this highly informed audience went well beyond the simple sharing of photos of conference slides that typified early medical conference tweeting when character limits were more restrictive. The much larger number of "just retweeters," though typically contributing a small number of retweets individually, are important in spreading the message to a wider audience.
There are also areas of opportunity for future events. For example, involving some of the "just mentioned" accounts in Figure 1 in tweeting and retweeting at future events: some of these accounts are opinion formers and disseminators in their own right, some with a considerable number of followers.
This summary of a large cardiology conference with international reach describes the dissemination of information relating to the conference via tweets, demonstrating the scale of engagement during and beyond the conference period. It provides information relevant to delegates and conference organizers at future conferences: know the hashtag and try to avoid hashtag drift; add the hashtag and another identifier (e.g., #cardiotwitter) to all tweets and replies.
This information should be added to all conference materials, from the first announcement of the conference, to email correspondence, conference branding on programs, slide backgrounds and posters, and on screens during the conference. Creating links from the conference app to tweet from the conference might be considered with an automatic generation of the appropriate hashtag.
The hashtag drift observed regularly in conferences and health campaigns is evidence of the need for repeated reminders. This study also identifies some of the limitations to social media analysis that require further development and resource, including the fact that replies that omit the hashtag are not extracted.
These findings help understand the use of Twitter in the scientific cycle of developing, debating and disseminating new ideas in cardiology.
Graham Mackenzie, MD, NHS Education for Scotland; Martha Gulati, MD, MS, FACC, University of Arizona; C. Michael Gibson, MD, FACC, Beth Israel Deaconess Medical Center, Boston, MA.
- NodeXL report for #ACC19 and #ACC2019, 8-23 April 2019. Accessed here on April 16, 2019.
- Søreide K, Mackenzie G, Polom K, Lorenzon L, Mohan H, Mayol J. Tweeting the meeting: Quantitative and qualitative twitter activity during the 38th ESSO conference. DOI.
- Conference summary of #ACC19/ #ACC2019 cardiology tweets. Accessed here. March 16-18, 2019. Accessed here on April 21, 2019.
- Post-conference Summary of #ACC19 / #ACC2019 cardiology tweets March 19-22, 2019. Accessed here. Accessed April 21, 2019.
- Social media activity at ISPAH 2018: running to keep up. Blog posted on BJSM site Nov. 26, 2018. Accessed here on April 22, 2019.
- Conroy, E. Conference report: the 67th American College of Cardiology Annual Scientific Meeting. Therapeutic Advances in Cardiovascular Disease. Accessed here.
- Twitonomy map of #ACC19 tweets and retweets, March 18 from 13:05 to 16:57. Accessed here on April 22, 2019.
< Back to Listings