Public Health Messaging on Twitter During the COVID-19 Pandemic: Observational Study

Abstract

Background: Effective communication is crucial during health crises, and social media has become a prominent platform for public health experts (PHEs) to share information and engage with the public. At the same time, social media also provides a platform for pseudoexperts who may spread contrarian views. Despite the importance of social media, key elements of communication, such as the use of moral or emotional language and messaging strategy, particularly during the emergency phase of the COVID-19 pandemic, have not been explored.

Objective: This study aimed to analyze how PHEs and pseudoexperts communicated with the public during the emergency phase of the COVID-19 pandemic. We focused on the emotional and moral language used in their messages on various COVID-19 pandemic-related topics. We also analyzed their interactions with political elites and the public's engagement with PHEs to gain a deeper understanding of their influence on public discourse.

Methods: For this observational study, we gathered a dataset of >539,000 original posts or reposts from 489 PHEs and 356 pseudoexperts on Twitter (subsequently rebranded X) from January 2020 to January 2021, along with the replies to the original posts from the PHEs. We identified the key issues that PHEs and pseudoexperts prioritized. We also determined the emotional and moral language in both the original posts and the replies. This allows us to characterize priorities for PHEs and pseudoexperts as well as differences in messaging strategy between these 2 groups. We also evaluated the influence of PHEs' language and strategy on the public response.

Results: Our analyses revealed that PHEs focused more on masking, health care, education, and vaccines, whereas pseudoexperts discussed therapeutics and lockdowns more frequently (P<.001). PHEs typically used positive emotional language across all issues (P<.001), expressing optimism and joy. Pseudoexperts often used negative emotions of pessimism and disgust, while limiting positive emotional language to origins and therapeutics (P<.001). Along the dimensions of moral language, PHEs and pseudoexperts differed on care versus harm and authority versus subversion across different issues. Negative emotional and moral language tends to boost engagement in COVID-19 discussions across all issues. However, the use of positive language by PHEs increases the use of positive language in the public responses. PHEs act as liberal partisans: they express more positive affect in their posts directed at liberals and more negative affect in their posts directed at conservative elites. In contrast, pseudoexperts act as conservative partisans. These results provide nuanced insights into the elements that have polarized the COVID-19 discourse.

Conclusions: Understanding the nature of the public response to PHEs' messages on social media is essential for refining communication strategies during health crises. Our findings underscore the importance of using moral-emotional language strategically to reduce polarization and build trust.

Keywords: COVID-19; Twitter; emotions; moral foundations; polarization; public health; public health messaging.

Figure 1

Repost interactions. Nodes represent public health experts (PHEs; green) and pseudoexperts (orange) and repost interactions between them. Green edges represent interactions where a PHE was reposted, and orange edges represent interactions where a pseudoexpert was reposted. The size of the node is proportional to the number of times the expert was reposted.

Figure 2

Comparing the activity of public health experts (PHEs) and pseudoexperts. Box plots compare the daily proportion of issue-related posts from PHEs and pseudoexperts. Mann-Whitney U Test with Bonferroni correction was used to assess significance. *P<.05, **P<.01, ***P<.001, and ****P<.0001; ns: not significant.

Figure 3

Daily fraction of original posts by (A) public health experts (PHEs) and (B) pseudoexperts related to each issue. Major events are marked with vertical lines. Major events are marked with vertical dashed lines: Lockdowns (March 15, 2020, purple), health care (March 30, 2020, orange), therapeutics (April 24, 2020, yellow, Trump's bleach proposal), education (July 8, 2020, red, Trump's school reopening call), and vaccines (November 9, 2020, blue, Pfizer's 93% efficacy report).

Figure 4

Box plots compare daily proportion of posts from public health experts (PHEs) and pseudoexperts expressing various emotions. Mann-Whitney U Test with Bonferroni correction is used to assess significance. *P<.05, **P<.01, ***P<.001, and ****P<.0001; ns: not significant.

Figure 5

Dynamics of emotions. Daily fraction of posts from public health experts (PHEs) and pseudoexperts expressing (A and B) positive (optimism, joy, anticipation, and love) and (C and D) negative (disgust, anger, sadness, fear, and pessimism) emotions. Major events are marked with vertical dashed lines: Lockdowns (March 15, 2020, purple), health care (March 30, 2020, orange), therapeutics (April 24, 2020, yellow, Trump's bleach proposal), education (July 8, 2020, red, Trump’s school reopening call), and vaccines (November 9, 2020, blue, Pfizer's 93% efficacy report).

Figure 6

Comparing emotions used by public health experts (PHEs) and pseudoexperts. It compares (A) anger, (B) disgust, (C) sadness, (D) fear, (E) optimism, and (F) joy across various topics. The figure displays the odds ratio of a post’s relevance to a particular issue based on the expression of specific emotions by PHEs.

Figure 7

Comparing use of moral foundations. It compares (A) care, (B) harm, (C) fairness, (D) cheating, (E) authority, (F) subversion, (G) loyalty, and (H) betrayal conveyed by public health experts (PHEs) and pseudoexperts across various topics. The figure displays the odds ratio of a post’s relevance to a particular issue based on the expression of specific moral foundations by PHEs.

Figure 8

Asymmetries in affect toward elites. (A) Public health experts (PHEs) direct more negative emotions (anger and disgust) toward conservative elites and more positive emotions (joy and optimism) in their mentions of liberal elites, which is a hallmark of affective polarization. In contrast, pseudoexperts direct more negativity toward liberal elites. With respect to moral language, (B) PHEs express more subversion in their mentions of conservative elites, in contrast to pseudoexperts. Lib stands for liberal elites and Con stands for Conservative elites.

Figure 9

Engagement with public health experts (PHEs). Dot and whisker plots show the regression coefficients and SEs. The coefficients represent the increase in number of replies when an (A) emotion or (B) moral foundation is used by the PHE in the original post while keeping others constant.

Figure 10

User engagement with public health experts (PHEs) and pseudoexperts. User reactions to (A) emotional and (B) moral appeals from PHEs. The figure demonstrates the odds ratio of users expressing emotions and moral principles in response to those conveyed in the original posts by PHEs. *P<.05, **P<.01, ***P<.001, and ****P<.0001; ns: not significant.