Research Article
Free access
Published Online: 11 February 2021

Whose Tweets on COVID-19 Gain the Most Attention: Celebrities, Political, or Scientific Authorities?

Publication: Cyberpsychology, Behavior, and Social Networking
Volume 24, Issue Number 2


Twitter has considerable capacity for health education and proves to be an efficient and accessible communication tool in the coronavirus disease-19 (COVID-19) pandemic. Although many stakeholders saturate Twitter with COVID-19–related information, it remains unknown who disseminates information most efficaciously. COVID-19–related tweets were obtained from Twitter accounts of health agencies, governmental authorities, universities, scientific journals, medical associations, and celebrities. Posts' impact was measured with the nominal and relative (%followers) number of likes and retweets. A sentiment analysis was conducted.We have identified 17,331 COVID-19–related tweets posted by 338 accounts in >4 months since the virus began to spread. The largest number of likes was received by tweets of celebrities (median nominal, relative likes; 14,918, 0.036 percent), politicians (259, 0.174 percent), and health agencies (231, 0.007 percent). Most retweeted messages were also posted by celebrities (2,366, 0.005 percent), health agencies (130, 0.004 percent), and politicians (55, 0.041 percent). Retweets and likes peaked in March 2020, and the overall sentiment of the tweets was growing steadily. Whereas celebrities and politicians posted positive messages, the scientific and health authorities often employed a negative vocabulary. The posts with positive sentiment gained more likes and relative likes than nonpositive. During the pandemic, the tweets of celebrities and politicians related to COVID-19 outperform those coming from health and scientific institutions. Active engagement of Twitter influencers may help key messages go viral.


With 275 million users globally, Twitter is the leading microblog platform. The users may react to a tweet by likes, comments, or share the message by “retweeting.” The account can be followed, which helps in the propagation of the tweets: the messages firstly are displayed to the network of followers. Therefore, both networks of followers and content of a tweet are crucial to reaching wide audience.1 From a psychological perspective, three mechanisms may help to understand the behavior of Twitter users. First, people tend to favor information that confirms their beliefs or action,2 which may result in sharing tweets of their authorities.1 Second, the most popular messages or profiles may gain additional attention through social proof phenomenon.3 Finally, the opinions with which users disagree also can be shared or commented to condemn the author of the tweet and obtain positive feedback from the user's own network.1
The research community has first appreciated Twitter's potential of influencing patients' health-related choices over a decade ago.4 Since then, the iconic 140 character-long tweets have become a clairvoyant of infodemics, capable of predicting flu pandemic dynamics,5 monitoring drug safety,6 and even nonmedical psychostimulant use.7 In the current pandemic, tweets are utilized to analyze various aspects of e-discourse including users' concerns,8 association with the spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2),9,10 views,11,12 and misinformation.13–17 Especially the latter constitutes a challenge to health authorities and governments who have rushed to saturate the Internet with science-based advice and narrative. Nevertheless, in the existing virtual world, the balance of influence is known to be complex, with a leading role played by celebrities. But do the message and the language differ between these actors?
Differences among information streams originating from major groups of Twitter coronavirus disease-19 (COVID-19) stakeholders may indicate the directions of harmonization, which could bolster education and nudge efforts. The awareness of the current state of COVID-19 crowd communication is indispensable for fine-tuning in the near future, for the benefit of the society. Therefore, we have explored the number of reactions to and sentiments of tweets on coronavirus coming from scientific institutions, governmental authorities, and celebrities.

Materials and Methods

Data collection

This retrospective infodemiology study did not require ethical committee approval. The research does not violate the terms of the platform.
We collected the names of tweeters' accounts from April 20 to 23, 2020. We distinguished the following categories of accounts: celebrities, politicians, state institutions, health agencies, medical associations, universities, medical, and nonmedical journals. We defined celebrities as individuals who have never been widely recognized for their scientific or political work, but for artistic, media, or sport performance. We collected the account names of 150 top Twitter celebrities according to The celebrities were international group, but most of them came from the United States.
We searched for official and private accounts of heads of institutions. These were, respectively, designated governmental institutions' and politicians' accounts. The following officials were considered: monarchs, presidents, prime ministers, chancellors, and ministers of health or heads of state health departments of European Union countries, Australia, Canada, New Zealand, Norway, Switzerland, the United Kingdom, and the United States. Health agencies included World Health Organization, Centers for Disease Control and Prevention, European Centre for Disease Prevention and Control, National Institute of Health, Food and Drug Administration, European Medicines Agency, Health Canada, The Therapeutic Goods Administration, and The Medicines and Healthcare products Regulatory Agency.
Medical associations comprised associations of physicians (e.g., American Medical Association), with specialists of infectious diseases, pulmonology, intensive care, or epidemiology living in European Union countries, Australia, Canada, New Zealand, Norway, Switzerland, the United Kingdom, and the United States.
The Shanghai Academic Ranking was used to find Twitter profiles of the top 100 universities across the globe.19 We also searched for Twitter accounts representing editorial offices of the top 50 medical and top 50 nonmedical journals, according to Scimago Journal Ranking.20
We obtained tweets from accounts' timelines using rtweet package of R using get_timeline function.21 We included all tweets since January 1, 2020, to date of collection (April 23, 2020) that included terms “corona,” “COVID,” “SARS,” pandemi,” “epidemi,” “CoV,” “coronavirus,” or had at least one of the following hashtags: #coronavirus, #COVID19, #pandemia, #epidemia, #SARS2, #SARS_COV_2, #SARS_CoV2, #SARSCoV2, #Corona, #CoronaVirusDE, #CoronavirusUSA, #pandemic, #epidemie,Epidemia, #VirusCorona, and #viruscoronavirus.

Data analysis

We performed analysis using custom code in R programming language (version 3.6.3., R Foundation for Statistical Computing, Vienna). Relative engagement on posts was measured with two ratios: of likes to followers and retweets to followers. We visualized the data using the ggplot2 package.22 We also performed a sentiment analysis on a subset of tweets in English using the tidytext package.23 The standard procedure included unnesting tokenization to words, removing stop words, and combining with a sentiment dictionary called “Bing.”24 We defined sentiment as the difference between the number of positive and negative words in each tweet. We compared reactions with tweets with positive versus nonpositive (0 or less) sentiment using the Mann–Whitney U test. The R code, data and list of twitter profiles are available on Mendeley.25 Examples of tweets sentiments are presented in Supplementary Table S1.


Among the 114,145 tweets posted by the 337 accounts since January 1, 2020, we detected 17,331 tweets related to the coronavirus. Most of the tweets were posted by universities, official state institutions, and health agencies (Table 1). Celebrities had the largest average number of followers with health agencies coming second (Table 1).
Table 1. Ranking of the Groups of Accounts According to Relative and Nominal Reaction Counts
VariableIncluded posts (n)Followers (n)Likes/followers ratio (%)Retweets/followers ratio (%)Likes (n)Retweets (n)
1Universities 4,796Celebrities: 55,669,226 (13,451,310–55,669,226)Politicians: 0.174 (0.059–0.468)Politicians: 0.041 (0.013–0.105)Celebrities: 14,918 (1,944–38,417)Celebrities: 2,366 (259–5,784)
2Official state institutions 3,410Health agencies: 7,541,142 (1,148,329–7,541,142)Celebrities:0.036 (0.008–0.080)Medical journals: 0.016 (0.006–0.048)Politicians: 259 (64–1,022)Health agencies: 130 (52–332)
3Health agencies 2,576Nonmedical journals: 1,612,708
Official state institutions: 0.027 (0.009–0.067)Official state institutions: 0.015 (0.005–0.037)Health agencies: 231 (94–566)Politicians: 55 (15–232)
4Politicians 2,200Official state institutions: 209,746 (75,386–648,495)Medical journals: 0.027 (0.011–0.073)Universities: 0.007 (0.003–0.021)Nonmedical journals: 86 (33–169)Nonmedical journals: 43 (17–95)
5Medical associations 1,937Politicians: 143,232 (55,368–391,430)Universities: 0.019 (0.008–0.050)Celebrities: 0.005 (0.001–0.011)Official state institutions: 54 (11–195)Official state institutions: 27 (6–108)
6Medical journals 1,659Universities 99,297 (59,574–187,459)Medical associations 0.012 (0.002–0.053)Medical associations: 0.005 (0.001–0.032)Medical journals: 30 (11–78)Medical journals: 19 (6–50)
7Celebrities 391Medical journals: 53,770 (37,159–491,141)Nonmedical journals: 0.009 (0.005–0.026)Nonmedical journals: 0.005 (0.002–0.014)Universities: 19 (8–48)Universities: 7 (3–19)
8Nonmedical journals 359Medical Associations 33,940 (28,847–722,081)Health agencies: 0.007 (0.003–0.025)Health agencies: 0.004 (0.001–0.019)Medical associations: 11 (5–22)Medical associations: 7 (3–14)
Data presented as median (interquartile range).
The groups clearly ranked by the number of likes and retweets relative to follower count and nominally (Table 1). Although tweets of politicians generated the most relative attention, celebrities' posts attracted most engagement overall. The changes in crude and relative attention are presented in Figure 1. The number of reactions to celebrities' tweets decreased over time. Most of the other contributors' posts gained most of the attention in mid-March.
FIG. 1. Time trends of reaction counts on tweets related to the COVID-19. COVID-19, coronavirus disease-19.
A large majority of tweets were in English (13,793; 79.6 percent). Although messages coming from celebrities and politicians displayed positive sentiment, negative words abounded in the output from other groups (Fig. 2). There was a general trend toward a more upbeat tone with time (Fig. 2). A fluctuating pattern of sentiment is seen in health agencies as opposed to medical associations showing a steadily mounting attitude; this reflects the greater granularity of data from health agencies, which were highly active on Twitter. The positive tweets gained significantly higher nominal (50 [14–271] vs. 45 [12–222] likes; p < 0.001), and relative likes (0.02 [0.01–0.07] percent vs. 0.02 [0.01–0.06] percent; p = 0.006). The posts' undertone was not associated with nominal (23 [6–115] vs. 24 [6–120] retweets; p = 0.87) and relative retweets (0.01 [0.00–0.03] percent vs. 0.01 [0.00–0.03] percent; p = 0.08).
FIG. 2. Time trends of the sentiment of tweets related to the COVID-19.


We characterize main groups of COVID-19 opinion leaders and provide data that indicate their key strengths. First, politicians are profoundly influential and surpass the official state institutions, thus exposing the value of personal versus official communication channels. Second, celebrities' rare but positive message attracts by far the most public attention. Third, the sentiment of tweets on coronavirus increases over time, and more optimistic messages got more likes. This is the first study to comprehensively compare the reactions to tweets of various groups of stakeholders and their sentiment in the course of the COVID-19 pandemic.
Tendencies in the spread of SARS-CoV-2 shifted in time with the spread in the Far East in January, which was followed by a temporary period of calm and a subsequent surge of infections around the world in March and April.26,27 COVID-19 began to attract attention in February as reflected by the increase of health agencies' post retweets and has peaked in mid-March with average retweets decreasing gradually afterward.
Overall, political, health, and scientific authorities had far less reaction to their posts than top celebrities. However, even the tweets of celebrities have lost some of their appeal as the virus spread worldwide and followers became accustomed to it. Nevertheless, celebrities were still able to reach many more Twitter users than institutions, although with a modest retweet-to-follower ratio. The private accounts of politicians gained the most relative attention, which did not decay with the further development of the pandemic. Similarly, official state accounts and medical journals were retweeted relatively often. We speculate that the health crisis has caused political and scientific authorities to gain more attention. Not only the retweet ratio, but also a further expansion of the followers' network is necessary to reach more users and spread key message. It may be imagined that celebrities could help by convincing their fans to follow institutional accounts providing valuable health information.
Interestingly, the politicians' and celebrities' tweets tended to have more positive undertones than tweets of the other groups. It may be speculated that since they direct their messages to a wide and generally nonprofessional audience, they prefer to be more joyful and supporting, and thus more engaging. This tone may also be used to ensure more attention and support in the future. The sentiment of the tweets increased over time. We would tend to think that initially the news related to new outbreaks and inadequate understanding of the virus caused a general negative sentiment of the tweets. In further months, when hopeful news and initiatives appeared, the messages became more optimistic.

Strengths and practical implications

To our best knowledge, this is the first kind of study comparing the reaction to tweets coming from a different group of opinion leaders. The social media dynamics during the crisis are under-researched, and this article makes a step forward. During the pandemic, the tweets of top celebrities related to COVID-19 outperform messages coming from health and scientific institutions. To be more effective, authorities' accounts should collect and maintain a high number of followers. Previous studies suggested that the health behavior of celebrities has a strong impact on community interest.28–31 Therefore, it may be helpful to engage celebrities for spreading key messages.32,33 More positive tweets from institutions may also help, but this incurs the risk of underrating the seriousness of the message by the recipient. These intriguing hypotheses require further studies.


Three limitations should be considered. First, the results concern leaders coming from western Europe, the United States, Canada, Australia, and New Zealand; thus, they should not be too easily extrapolated. Second, the rtweet package allows to obtain a maximum of 3,200 tweets (including retweets) from the timelines. Therefore, we might have omitted some tweets from very active accounts. Finally, the data did not include changes of followers over time; thus, the relative reaction counts refer to the followers' number on the day of collection.


During the pandemic, the tweets of celebrities and politicians related to COVID-19 outperform those coming from health and scientific institutions. Active engagement of Twitter influencers may help key messages go viral.

Ethical Committee

This retrospective infodemiology study did not require ethical committee approval.

Supplementary Material

File (supp_tables1.doc)


1. Casarin R, Correa JC, Camargo JE, et al. What makes a tweet be retweeted? A Bayesian trigram analysis of tweet propagation during the 2015 Colombian political campaign. Journal of Information Science 2019:016555151988605.
2. Klayman J, Ha Y. Confirmation, disconfirmation, and information in hypothesis testing. Psychological Review 1987; 94:211–228.
3. Cialdini RB. (1993) Influence: the psychology of persuasion. Rev. ed. New York: Morrow.
4. Sinnenberg L, Buttenheim AM, Padrez K, et al. Twitter as a tool for health research: a systematic review. American Journal of Public Health 2017; 107:e1–e8.
5. Achrekar H, Gandhe A, Lazarus R, et al. (2011) Predicting flu trends using Twitter data. In: 2011 IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). Shanghai: IEEE, pp. 702–707.
6. Freifeld CC, Brownstein JS, Menone CM, et al. Digital drug safety surveillance: monitoring pharmaceutical products in Twitter. Drug Safety 2014; 37:343–350.
7. Hanson CL, Burton SH, Giraud-Carrier C, et al. Tweaking and tweeting: exploring Twitter for nonmedical use of a psychostimulant drug (Adderall) among college students. Journal of Medical Internet Research 2013; 15:e62.
8. Abd-Alrazaq A, Alhuwail D, Househ M, et al. Top concerns of Tweeters during the COVID-19 pandemic: infoveillance study. Journal of Medical Internet Research 2020; 22:e19016.
9. Mavragani A. Tracking COVID-19 in Europe: infodemiology approach. JMIR Public Health and Surveillance 2020; 6:e18941.
10. Su Y, Xue J, Liu X, et al. Examining the impact of COVID-19 lockdown in Wuhan and Lombardy: a psycholinguistic analysis on Weibo and Twitter. International Journal of Environmental Research and Public Health 2020; 17:4552.
11. Budhwani H, Sun R. Creating COVID-19 stigma by referencing the novel coronavirus as the “Chinese virus” on Twitter: quantitative analysis of social media data. Journal of Medical Internet Research 2020; 22:e19301.
12. Jain S, Sinha A. Identification of influential users on Twitter: a novel weighted correlated influence measure for Covid-19. Chaos, Solitons & Fractals 2020; 139:110037.
13. Cuan-Baltazar JY, Muñoz-Perez MJ, Robledo-Vega C, et al. Misinformation of COVID-19 on the Internet: infodemiology Study. JMIR Public Health and Surveillance 2020; 6:e18444.
14. Kouzy R, Abi Jaoude J, Kraitem A, et al. Coronavirus goes viral: quantifying the COVID-19 misinformation epidemic on Twitter. Cureus 2020; 12:e7255.
15. Rosenberg H, Syed S, Rezaie S. The Twitter pandemic: the critical role of Twitter in the dissemination of medical information and misinformation during the COVID-19 pandemic. CJEM 2020; 22:418–421.
16. Pulido CM, Villarejo-Carballido B, Redondo-Sama G, et al. COVID-19 infodemic: more retweets for science-based information on coronavirus than for false information. International Sociology 2020; 35:377–392.
17. Ahmed W, Vidal-Alaball J, Downing J, et al. COVID-19 and the 5G conspiracy theory: social network analysis of Twitter data. Journal of Medical Internet Research 2020; 22:e19458.
18. Anon. Top 500 Twitter users by followers. Social Blade.
19. Anon. Academic Ranking of World Universities 2019.
20. Anon. SCImago Journal & Country Rank.
21. Kearney M, Heiss A, Briatte F. (2020) Package ‘rtweet’: collecting Twitter data. (accessed Sep. 21, 2020).
22. Anon. (2018) ggplot2—Elegant graphics for data analysis | Hadley Wickham | Springer. (accessed Feb. 15, 2018).
23. De Queiroz G, Fay C, Hvitfeldt E, et al. (2020) Package ‘tidytext’: text mining using “dplyr,” “ggplot2,” and other tidy tools. (accessed Sep. 21, 2020).
24. Hu M, Liu B. (2004) Mining and summarizing customer reviews. In Proceedings of the 2004 ACM SIGKDD International Conference on Knowledge Discovery and Data Mining—KDD’04. Seattle, WA, USA: ACM Press, p. 168.
25. Kamiński M, Szymańska C, Nowak JK. (2020) Whose tweets on COVID-19 gain the most attention: celebrities, politcal or scientific authorities?. (accessed Sep. 21, 2020).
26. Anon. Timeline of the 2019–20 coronavirus pandemic. Wikipedia.
27. Anon. (2020) Timeline of WHO's response to COVID-19. 2020.
28. Ueda M, Mori K, Matsubayashi T, et al. Tweeting celebrity suicides: users' reaction to prominent suicide deaths on Twitter and subsequent increases in actual suicides. Social Science & Medicine 2017; 189:158–166.
29. Kamiński M, Nowak JK, Skonieczna-Żydecka K, et al. Gluten-free diet yesterday, today and tomorrow: forecasting using Google Trends data. Arab Journal of Gastroenterology 2020; 21:67–68.
30. Kaleem T, Malouff TD, Stross WC, et al. Google search trends in oncology and the impact of celebrity cancer awareness. Cureus 2019; 11:e5360.
31. Sillence E, Martin R. Talking about decisions: the facilitating effect of a celebrity health announcement on the communication of online personal experiences around decision-making. Health Communication 2020; 35:1447–1454.
32. Albalawi Y, Sixsmith J. Identifying Twitter influencer profiles for health promotion in Saudi Arabia. Health Promotion International 2017; 32:456–463.
33. Dalglish SL. COVID-19 gives the lie to global health expertise. The Lancet 2020; 395:1189.

Information & Authors


Published In

cover image Cyberpsychology, Behavior, and Social Networking
Cyberpsychology, Behavior, and Social Networking
Volume 24Issue Number 2February 2021
Pages: 123 - 128
PubMed: 32986469


Published online: 11 February 2021
Published in print: February 2021
Published ahead of print: 23 September 2020


Request permissions for this article.




Mikołaj Kamiński [email protected]
Department of the Treatment of Obesity and Metabolic Disorders, and of Clinical Dietetics, Poznań University of Medical Sciences, Poznań, Poland.
Cyntia Szymańska
Faculty of Medicine I, Poznań University of Medical Sciences, Poznan, Poland.
Jan Krzysztof Nowak
Department of the Pediatric Gastroenterology and Metabolic Diseases, Poznań University of Medical Sciences, Poznań, Poland.


Address correspondence to: Dr. Mikołaj Kamiński, Department of the Treatment of Obesity and Metabolic Disorders, and of Clinical Dietetics, Poznań University of Medical Sciences, Szamarzewskiego 84, Poznań 60-569, Poland [email protected]

Author Disclosure Statement

No competing financial interests exist.

Funding Information

No funding was received for this article.

Metrics & Citations



Export citation

Select the format you want to export the citations of this publication.

View Options

View options


View PDF/ePub

Get Access

Access content

To read the fulltext, please use one of the options below to sign in or purchase access.

Society Access

If you are a member of a society that has access to this content please log in via your society website and then return to this publication.

Restore your content access

Enter your email address to restore your content access:

Note: This functionality works only for purchases done as a guest. If you already have an account, log in to access the content to which you are entitled.







Copy the content Link

Share on social media

Back to Top