World Immunization Week 24-30 April 2020: A rapid review of pro- and anti- vaccination activity at the peak of the COVID-19 pandemic

I wrote this piece at the start of May 2020. It wasn’t accepted by a peer reviewed journal at that time, so I am providing it here as a blog instead. You can read the accompanying Wakelet summary of tweets (pro-vaccination) here.

Abstract

Public attitudes on vaccination are of interest during the COVID-19 pandemic. Social media activity provides a route to unprompted views on vaccination. This study captures global tweeting during the World Immunization Week 24-30 April 2020, with a particular focus on describing the antivaccination content but avoiding identifying the tweeters or the individual tweets to avoid disseminating their views further.

Tweets using the official campaign hashtag #VaccinesWork and related terms including plain English phrases (“World Immunization Week”) and hashtags (e.g. #WorldImmunizationWeek) were extracted using an online social media tool (TAGS). Tweets and retweets were then imported into the NodeXL Excel extension at the end of the week to obtain the most up to date data on number of retweets received, hashtags used and characteristics of tweeters and retweeters. The top 200 tweets based on retweets were identified and antivaccination posts (n=5) used to identify tweeters and retweeters, repeating this six times and performing a further search based on commonly used antivaccination hashtags.

In total there were 10,946 tweets by 6,264 tweeters, with 53,124 retweets by 32,445 retweeters. The great majority of the most popular tweets were pro-vaccination. However, after an analysis of the most popular antivaccination posts, the accounts retweeting these posts, and the specifically antivaccination hashtags used, 218 antivaccination tweets were identified, posted by 89 tweeters, with 724 retweets made by 518 retweeters. These tweets disseminated common antivaccination myths, including material that inverted comments by senior officials to dispute vaccine safety/ efficacy.

This rapid analysis captures and describes some of the most popular pro- and anti-vaccination tweets from the April 2020 World Immunization Week, helping understand public views on vaccination globally at the height of the COVID-19 pandemic.

The full analysis is provided below, including word clouds of hashtags used by the different communities. You can also download this as a PDF file.

Funding: No external funding

No conflict of interest

Dr Graham Mackenzie, MD FRCP(Edinburgh), @gmacscotland on Twitter (4 May 2020, updated 10 April 2021)

Introduction: The COVID-19 pandemic has resulted in unprecedented change in society and communication. Much of the political, public and private discourse about lockdown, social distancing and the eventual recovery has included topics that relate directly to public health, clinical practice and virology. Examples include epidemiology (e.g. reproduction number), the virus (e.g. how long it survives on different surfaces and materials), methods of protection (e.g. masks in healthcare and public settings), and the prospects of a future vaccine. There has often been a political dimension to these discussions, e.g. in the allocation of personal protective equipment, impact of socioeconomic inequalities, and the difference in approaches to lockdown and social distancing internationally. This is the first pandemic during which social media use has provided an unprompted view of public and professional thought at a global level. An attempt has been made to describe the level of tweeting and sentiment analysis using an algorithmic approach during the early stages of the COVID-19 pandemic(1), but a study of social media about immunisation specifically during the pandemic, using content analysis of individual tweets, has not been attempted. Capturing and studying the most popular content posted during the World Immunization Week 2020 (24-30 April each year) is therefore of considerable interest. Twitter is the ideal platform for such study as it is free and open, with most content shared publicly. The official campaign hashtag is #VaccinesWork each year.(2)

I have attempted to collect and study tweets for each year since 2017, learning about the most prominent tweeters – for and against vaccination – in the process. The scale of tweeting, however, represents a challenge due to Twitter limits on the number of tweets that can be collected (18,000 tweets and retweets over the preceding 7-10 days per extract), and the long time required to process such information, both in terms of computing and analytical time. With popular tweets by international organisations (e.g. WHO, UN, UNICEF) and media celebrities receiving tens of thousands of retweets in a matter of hours it was not possible in 2019 to obtain a complete extract, so workarounds were required. That analysis did not capture related tweeting with other hashtags. With this in mind alternatives to data extraction and analysis were explored in advance for the 2020 campaign.

Search strategy and selection criteria: Developments in social network analysis tools have allowed retrospective data collection, without time limit, as long as the unique tweet and retweet identifiers have been recorded. Accordingly, for 2020 an hourly extract using a free web-based tool called TAGS(3) captured tweets and retweets using a range of search terms – the official #VaccinesWork hashtag, variants of #WorldImmunizationWeek (US and UK spellings, adding in 20 and 2020 in case year specific terms had been used), and plain English phrases (“World Immunization Week”, again with US and UK spellings).

Search term used: ‘#vaccineswork OR “World Immunization Week” OR “World Immunisation Week” OR #WorldImmunizationWeek OR #worldimmunisationweek” OR #worldimmunizationweek2020 OR #worldimmunisationweek2020 OR #worldimmunizationweek20 OR #worldimmunisationweek20’

Twitter does not have a fuzzy search, meaning that only these specific terms will be found using such a search. TAGS collects the activity (tweeting and retweeting) in the previous hour, but does not update the number of retweets in the tweet record subsequently. At the end of the 7-day period 24-30 April 2020 (UTC), allowing 12 hours either side to capture activity in all time zones, the TAGS extract was downloaded and the unique tweet and retweet identifiers (the 19 digit number at the end of the URL for each tweet and retweet) used to repeat the data acquisition in the NodeXL Excel extension.(4) This second extract took almost 24 hours to complete on a powerful gaming computer. There were tweets that were extracted in the hourly TAGS download that were not available at the point of the subsequent NodeXL extract (n=953). Checking a sample of these tweets individually, all had been deleted (checked 3/5/2020). Tweeters will commonly delete a post if they are not happy with the wording, the way that an image is displayed, or on the basis of feedback from other tweeters, so this was not unexpected.

There are advantages to using NodeXL for the updated data acquisition: it records separately the tweets, retweets, and mentions of tweeters in original tweets and replies, and maps these connections. It collects the updated number of retweets at the point of the second data acquisition and records the hashtags used in each tweet in a way that can be easily separated for further analysis. This allowed ranking of the most popular tweets for further inspection, and rapid analysis of the hashtags used for all the tweets. A previous illustration of the application of these methods in infectious diseases conferences has been provided previously(5) and the approach has been refined over time for further conferences and professional networking in infectious diseases (6), surgery (7,8), oncology(9) and cardiology.(10-12)  The approach has been more challenging to apply to public health awareness campaigns due to the scale of tweeting. Tweets were downloaded regardless of language, with German, Japanese, Malay, Indonesian and Spanish language tweets included in the final dataset and translated using the Google translation feature incorporated into Twitter. Hashtags were extracted regardless of language, but only those in the Roman alphabet and Arabic numerals were available for further analysis as hashtags were separate out for individual tweets using the comma separated file format in Excel, which leads the loss of other characters. A comparison with the 2019 campaign has not been attempted as an incomplete extract was obtained for #VaccinesWork tweets in 2019, and it was not possible to extract related hashtags that year due to limitations on the number of extracts that could be performed in the time. Furthermore, the 2019 campaign had a considerable level of retweeting by South Korean pop fans, with the encouragement that each retweet of WHO posts would raise $1 for vaccination work. This led to a large number of tweets and retweets for posts that were unrelated to vaccination but included the campaign hashtag. Similar activity was not evident for the campaign in 2020 after inspecting the most popular content.

Antivaccination tweeters tweeting using the campaign hashtag or related terms were in a quite separate group to the rest of the tweeters, with limited interaction with other tweeters. It would be unfeasible to inspect all the tweets individually, so the most widely disseminated antivaccination tweets were identified rapidly. In order to further study the antivaccination bubble rapidly, so that the findings would be useful during the first wave of the pandemic, a convenience sample was obtained using a set of simple rules. The retweeters of the most widely shared antivaccination posts were identified and their other tweets and retweets were identified from the extract. This process of identifying tweeters and retweeters was repeated for five further cycles. Using the final list of antivaccination tweeters/retweeters, the hashtags most commonly used by this group were identified and those hashtags that were specifically associated with antivaccination messaging were selected for further searching. The hashtags identified in the full dataset and the final antivaccination search were summarised in word clouds using freely available online software that charts aggregate data, as required for the large number of tweets extracted.(13)

Results: In total there were 10,946 tweets by 6,264 tweeters, with 53,124 retweets by 32,445 retweeters during the 8-day period studied (noon on 23/4/2020 until noon on 1/5/2020, providing global coverage for the full week of the campaign). The top tweeters (by number of retweets received) are shown in Table 1, representing a range of vaccination supporting individuals (professional and celebrity) and organisations (international, regional and national).

Table 1. Top 30 tweeters from the campaign, by number of retweets received

The hashtags used in these tweets are summarised in a word cloud (Figure 1). Individual retweet data was obtained for 47,300 retweets (the remainder were either not available via the Twitter API or were retweeted in the period between data extraction and the processing by NodeXL (commenced at midnight on 2 May 2020)). Using this dataset a total of 37,551 accounts tweeted and/or retweeted, with 1,158 doing both. Overall, 166 accounts (2.7% of the total tweeters) received 80% of retweets while 4,135 accounts (66.0% of tweeters) received no retweets.

Figure 1. Hashtags from all tweets in the extract (n=31,031 hashtags from 10,604 tweets)

The tweeters and content were explored using an approach selected for speed and reproducibility. Selecting the most retweeted posts provides a quick view of the most popular and/or controversial material. The top 200 tweets were identified, each of which had received at least 36 retweets by the point of the NodeXL extract (median 78 retweets, interquartile range 49-137), and were posted by 100 organisations or individuals with a median of 102,057 followers (IQR 27,669-372,458). Just 5 of the top 200 posts were clearly antivaccination tweets, posted from 4 accounts. There were no tweets in the top 200 posts that expressed views related to vaccine hesitancy.

Overall 195/200 of the most shared tweets were pro-vaccination tweets, and these have been captured in a Wakelet summary.(14) COVID-19 related terms featured in 88/200 posts. These most popular tweets used a range of hashtags: 158 used #VaccinesWork, 63 used #WorldImmunizationWeek or related terms, and 5 had no hashtag. These posts covered a wide range of topics related to the development, testing, delivery, rationale and impact of vaccination programmes, with material tailored for different countries and age groups, with additional messages related to animal vaccination. Some tweets referenced the wide range of infectious diseases and related medical conditions against which vaccination provides protection. Other tweets provided details about the success of vaccination in protecting against specific pathogens including polio and smallpox. They also addressed antivaccination myths in clear and accessible language. Tweets relating to COVID-19 explained the impact of the pandemic lockdown on routine vaccine delivery to key age groups, the ongoing efforts in developing a COVID-19 vaccine, and the place that this might take in the pandemic response. For many of the tweets, for both routine vaccination and a future COVID-19 vaccine, the focus was on vaccination for all.

The antivaccination material has not been shared here, to avoid giving the tweeters and their ideas a wider audience. After an extended search for tweeters and retweeters and relevant hashtag a total of 218 tweets posted by 89 tweeters were identified, with 724 retweets made by 518 retweeters. In total there were 593 tweeters and/or retweeters, with 14 performing both activities. Overall, these tweeters and retweeters had a median of 370 followers (IQR 88-1,334 followers). The main themes in the antivaccination tweets can be summarised by representing the hashtags used in these tweets in a word cloud (Figure 2).

Figure 2. Hashtags in antivaccination tweets. (n=891 hashtags from 218 tweets)

This provides insights into the current concerns, preoccupations and delusions of the antivaccination movement. These posts were made by individuals, alternative media outlets and organisations, typically with limited information in their user profile. The tweeters return to familiar topics, some tailored to raise concerns about the COVID-19 response, warping topical news stories in an attempt to substantiate dubious points. The reasoning behind the ideas posted is typically very obviously wrong, often at multiple levels, but the repetition, framing, imagery and use of scientific language is of concern. Long busted myths are perpetuated, including false claims relating to the constituents of vaccinations, links to neurodevelopmental disorders, and the involvement and motive of pro-vaccination philanthropists, citing links to the pharmaceutical industry. Opinions of senior US government professionals are ridiculed, with opposing opinions of the US President supported and amplified. In one example, the disastrous comments by Donald Trump on use of detergent were inverted to make inaccurate claims about the manufacture of vaccines. In another, the ongoing investigation of immunity following COVID-19 infection was used to claim that immunisation would not be effective. The wholly discredited idea that telecom signals causes COVID-19 infection was also aired in tweets, as identified in the word cloud, though these posts were not widely disseminated, with more prominent antivaccination tweeters perhaps wishing to distance themselves from these views. Other widely shared posts made comments about vaccination more generally. In some cases, observations reported in peer reviewed papers and official reports were reported without context, linked to unrelated claims, and amplified by other tweeters, attempting to add scientific rigour to conspiracy theories. Alt right media outlets carried interviews with discredited individuals with medical or scientific qualifications, providing direct links to content that has been removed by YouTube and other mainstream outlets. These unsubstantiated and dangerous claims ignore the much more widely circulated materials by public health and other healthcare organisations and individuals across the world that support the use of vaccination, at no financial gain to themselves.

Discussion: This analysis has provided a review of the tweeting – for and against vaccination – during World Immunization Week 2020. It explains methods that can be used rapidly to capture tweeting during a large global public health campaign, and ways to explore and describe opposing views. These approaches are particularly relevant to a period of massive change, as with the COVID-19 pandemic, but they have wider application to healthcare tweeting more generally.

The approach followed here can be repeated with readily available tools and only limited analytical knowledge. Social media activity involving separate and opposing groups with a range of hashtags, both generic and specific to the groups involved – e.g. around vaping, global warming, gun control – could potentially be studied using the same approaches. Twitter is used by a smaller number of people globally than other social media platforms including Instagram and Facebook.(15) However, Twitter remains the only tool that is openly accessible to such an analysis. It is possible that antivaccination views posted in private groups (e.g. Facebook, WhatsApp) would be different to the public views analysed here. While it is likely that there were other antivaccination tweets among the 10,964 tweets posted during the week, this analysis allowed the rapid identification of a posts representing a wide range of antivaccination sentiment for further analysis.

The antivaccination lobby used a range of techniques in these tweets that are familiar from the populist copybook. These antivaccination posts represent a minority of tweets posted overall during the week long global campaign, disseminated within a limited number of tweeters and retweeters. The public nature of the information, and the responses made by members of the antivaccination and pro-vaccination communities, is helpful in understanding the range of opinions, and starting to counter disinformation and protect vulnerable individuals and communities that may be susceptible to their messaging. The much more commonly shared posts tweeted by WHO, UNICEF and other prominent public health organisations and individuals provide a useful counterpoint to these antivaccination posts, but in themselves are unlikely to change the opinions and vaccination choices of the most hardened antivaccination campaigners and their families.

References

  1. Abd-Alrazaq A, Alhuwail D, Househ M, Hamdi M, Shah Z. Top concerns of tweeters during the COVID-19 pandemic: A surveillance study. J Med Internet Res 2020;22(4):e19016. https://www.jmir.org/2020/4/e19016/
  2. WHO webpage. World Immunization Week 2020. https://www.who.int/news-room/campaigns/world-immunization-week/world-immunization-week-2020 accessed 3/5/2020.
  3. Hawksey M. TAGS website. https://tags.hawksey.info/get-tags/ accessed 3/5/2020/.
  4. Social media research foundation. NodeXL website. https://www.smrfoundation.org/nodexl/ accessed 3/5/2020.
  5. G Mackenzie. Twitter big data and infectious disease conferences. Lancet Infectious Diseases 2018 https://doi.org/10.1016/S1473-3099(18)30011-2.
  6. Cevik M, Ong DSY, Graham Mackenzie. How scientists and physicians use Twitter during a medical congress. Clinical Microbiology and Infection 2019. https://doi.org/10.1016/j.cmi.2019.04.030.
  7. Kjetil Søreide, Graham Mackenzie, Karol Polom, Laura Lorenzon, Helen Mohan, Julio Mayol. Tweeting the meeting: Quantitative and qualitative twitter activity during the 38th ESSO conference. EJSO February 2019. DOI: https://doi.org/10.1016/j.ejso.2018.11.020
  8. R Grossman, D Mackenzie, D Keller, N Dames, P Grewal, et al. #SoMe4Surgery: from inception to impact. BMJ Innovations 2020. http://dx.doi.org/10.1136/bmjinnov-2019-000356  
  9. Antonio Passaro, Graham Mackenzie, Matteo Lambertini, Gilberto Morgan, Stefan Zimmerman, Pilar Garrido, Giuseppe Curigliano & Dario Trapani. European Society for Medical Oncology (ESMO) 2018 Congress Twitter analysis: from ethics to results through the understanding of communication and interaction flows. ESMO Open 2020. http://dx.doi.org/10.1136/esmoopen-2019-000598
  10. Sarah Hudson and Graham Mackenzie. “Not your daughters Facebook”: Twitter Use at the European Society of Cardiology Conference 2018. Heart (British Cardiac Society) October 2018. http://dx.doi.org/10.1136/heartjnl-2018-314163
  11. D. Graham Mackenzie, Sarah Hudson, Martha Gulati. Who influences tweeting at international cardiology conferences? European Heart Journal 2020. https://doi.org/10.1093/eurheartj/ehaa162
  12. Mackenzie G, Gulati M. ACC.20: Impact of Social Media at the Virtual Scientific Sessions During the COVID-19 Pandemic. Clinical Cardiology 2020. https://doi.org/10.1002/clc.23387
  13. WordArt website. https://wordart.com/ accessed 3/5/2020.
  14. Wakelet summary of the campaign in top 200 tweets (will be 195 for final version) https://wakelet.com/wake/_zp80VEljA57TUoZrp83I accessed 3/5/2020.
  15. Pew Research. Share of U.S. adults using social media, including Facebook, is mostly unchanged since 2018. https://www.pewresearch.org/fact-tank/2019/04/10/share-of-u-s-adults-using-social-media-including-facebook-is-mostly-unchanged-since-2018/ accessed 4 May 2020.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s