This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Effective communication during a health crisis can ease public concerns and promote the adoption of important risk-mitigating behaviors. Public health agencies and leaders have served as the primary communicators of information related to COVID-19, and a key part of their public outreach has taken place on social media platforms.
This study examined the content and engagement of COVID-19 tweets authored by Canadian public health agencies and decision makers. We propose ways for public health accounts to adjust their tweeting practices during public health crises to improve risk communication and maximize engagement.
We retrieved data from tweets by Canadian public health agencies and decision makers from January 1, 2020, to June 30, 2020. The Twitter accounts were categorized as belonging to either a public health agency, regional or local health department, provincial health authority, medical health officer, or minister of health. We analyzed trends in COVID-19 tweet engagement and conducted a content analysis on a stratified random sample of 485 tweets to examine the message functions and risk communication strategies used by each account type.
We analyzed 32,737 tweets authored by 118 Canadian public health Twitter accounts, of which 6982 tweets were related to COVID-19. Medical health officers authored the largest percentage of COVID-19–related tweets (n=1337, 35%) relative to their total number of tweets and averaged the highest number of retweets per COVID-19 tweet (112 retweets per tweet). Public health agencies had the highest frequency of daily tweets about COVID-19 throughout the study period. Compared to tweets containing media and user mentions, hashtags and URLs were used in tweets more frequently by all account types, appearing in 69% (n=4798 tweets) and 68% (n=4781 tweets) of COVID-19–related tweets, respectively. Tweets containing hashtags also received the highest average retweets (47 retweets per tweet). Our content analysis revealed that of the three tweet message functions analyzed (information, action, community), tweets providing information were the most commonly used across most account types, constituting 39% (n=181) of all tweets; however, tweets promoting actions from users received higher than average retweets (55 retweets per tweet). When examining tweets that received one or more retweet (n=359), the difference between mean retweets across the message functions was statistically significant (
Public health agencies and decision makers should examine what messaging best meets the needs of their Twitter audiences to maximize sharing of their communications. Public health accounts that do not currently employ risk communication strategies in their tweets may be missing an important opportunity to engage with users about the mitigation of health risks related to COVID-19.
On January 25, 2020, the first case of COVID-19 was reported in Canada in a man who had recently traveled to Wuhan, China, where the virus was first identified [
During this time, public health agencies and officials emerged as the de facto leaders and primary decision makers for setting evidence-based public health policies, practices, and norms. Daily updates from medical officers of health and other public health experts would set the course for how each jurisdiction would respond to COVID-19 and outline the public’s role in “flattening the curve.” Some early research suggests that Canadians listened to these messages and followed public health recommendations [
The roles and responsibilities of Canadian public health institutions and individuals differ across the country and by levels of government. As a result, it is not always clear where the public should go to access and retrieve information during a health crisis. For example, the federal government’s main role is to communicate national case numbers to all Canadians, coordinate control measures across provinces, and provide updates on national issues such as travel and the delivery of medical supplies. These activities are undertaken mostly by the Public Health Agency of Canada (PHAC), which was established as a separate agency of the federal department of health specifically to improve responses to infectious disease outbreaks after the severe acute respiratory syndrome (SARS) outbreak of 2003 [
Conversely, Canadian provinces are responsible for leading the emergency response, whereby ministries of health are tasked with communicating provincial updates on case counts, conducting surveillance and monitoring, providing guidance on infectious disease control measures and policies, and testing and screening practices [
The diversity in public health responses and responsibilities between institutions and individuals led virtually every Canadian province to take a different approach to crisis communications and information dissemination related to COVID-19. For example, provinces such as British Columbia have put their provincial medical officers of health in the spotlight, while others, including Ontario, have opted to have elected officials such as local premiers or health ministers lead some of the response. Occasionally this has resulted in contradictory messaging from multiple spokespeople, leaving the public confused and unsure whose guidance to follow [
One of the ways in which the public has stayed informed on key information and updates on COVID-19 has been through the use of social media applications. Twitter in particular reported its biggest ever annual gain in daily users globally during the pandemic, which was up by 24% year over year during the first 3 months of 2020 [
During a health crisis such as a pandemic, the role of public health agencies and officials as communicators of timely and accurate information is especially crucial in helping the public form accurate perceptions of health risks and adapt their behaviors in ways that are necessary to mitigate risks [
One way to assess the online influence of Twitter accounts is to examine the engagement that their tweets receive. This can indicate how much an account’s communications are being seen, studied, and shared. Tweet engagement can be indicated by various measures including the number of retweets (ie, shares of a tweet), likes (ie, number of times a user has seen and acknowledged or agreed with a tweet), or replies (ie, number of times a tweet has been commented on or responded to). Retweets in particular have been identified as an effective measure of engagement as they can indicate both the level of user agreement with a message and also the level of diffusion that message has undergone based on how many shares it has amassed from the original tweeter [
Our study builds on past research that has examined the use of Twitter specifically by public health departments, agencies, and organizations. Most studies tend to focus on either examining the relationship between tweet features and levels of engagement and/or analyzing the content of the tweets to characterize the tweeting practices of particular accounts. For example, in a study of tweet engagement strategies used by 25 federal health agencies in the United States, it was found that hashtags, URLs, and user mentions were associated with an increased frequency of retweets [
Previous work summarizing best practices in risk communication during broader risk events [
The goal of our study was to characterize the content and level of engagement of COVID-19 tweets made by Canadian public health agencies and decision makers. Further, we propose recommendations for ways through which health agencies and decision makers could adjust their tweeting practices about COVID-19 and other future health crises to improve risk communication and maximize engagement. Our study seeks to answer four primary research questions (RQs):
RQ1: which types of Canadian public health agencies and decision makers tweeted the most about COVID-19 and when?
RQ2: how much engagement did tweets by Canadian public health agencies and decision makers receive during COVID-19? How did engagement change over time by account type?
RQ3: did tweets containing Twitter engagement strategies receive more retweets than those that did not? How did the use of engagement strategies vary by account type?
RQ4: did tweets from Canadian public health agencies and decision makers that employed a particular message function and risk communication strategy receive more retweets than others? How did the use of risk communication strategies in tweets from Canadian public health agencies and decision makers change over time?
A comprehensive list of Canadian public health institutions, agencies, and leaders was compiled after conducting a scoping review of provincial government websites. The resulting list of agencies and decision makers from this initial search was cross-referenced with the “Structural Profile of Public Health in Canada,” a resource published by the National Collaborating Centre for Healthy Public Policy [
Twitter data were downloaded using the Twitter API accessed through R using the
The Twitter data collected between May 23 and July 10 yielded 303,428 tweets from February 2010 to July 10, 2020. The data were then narrowed down to only include tweets authored between January 1, 2020, and June 30, 2020, resulting in 71,014 tweets. This time period was selected as China first reported the outbreak of the novel coronavirus to the WHO on January 1, 2020. Of the 128 Twitter accounts we sampled, 118 had authored tweets between this period. Finally, retweeted tweets were excluded, resulting in 45,310 tweets.
Most tweets in the sample were single standalone tweets within the tweet character limit, but some comprised longer tweet threads. In these cases, tweets were turned into threads using the following procedure. First, all tweets were sorted by publisher and date and time. Next, all tweets that were identified as self-replies (based on the “reply_to_screen_name” variable) were treated as part of a tweet thread. The start of a tweet thread was the tweet that immediately preceded the chain of tweets that were self-replies. This starting tweet, and all subsequent self-replies, were combined into a tweet thread. These tweets and tweet threads (n=32,737) will be referred to simply as tweets throughout the remainder of this paper. The tweets were then classified by whether they were about COVID-19 or not, based on whether they contained any of the following keywords: “covid*,” “coronavirus,” “ncov,” “distanc*,” “pand*,” “tracing,” “testandtrace,” “curve,” “stayhome,” “handwashing,” “mask,” and “masque.” These keywords were identified by scanning tweets within the sample and noting commonly used words in French- and English-language tweets describing COVID-19. During this scan, it was found that most French-language tweets about COVID-19 used #COVID19 to denote the tweet’s topic relevant to COVID-19, which was a less common practice among English-language tweets. Hence, a larger number of English-language keywords were required to classify whether an English-language tweet was about COVID-19. This selection process resulted in 6982 tweets about COVID-19.
Additionally, we downloaded the publicly available COVID-19 data set from the Government of Canada [
To classify the public health accounts by type, each account was categorized as belonging to either an agency or a decision maker. If the Twitter account belonged to an agency, it was classified as being one of 3 types: public health agency (ie, a federal or provincial public health agency, distinct from a health ministry due to its focus on public health), a regional or local health department (ie, a public health department that offers public health programs or services to communities at a scale smaller than the province), or a provincial health authority (ie, provincial health ministries or health authorities). If the account belonged to a decision maker, it was classified as being one of 2 types: medical officer of health (ie, the chief medical health officer of Canada and of each province, and regional or local medical health officers responsible for public health in smaller communities) or provincial minister of health (ie, elected government official who oversees health and public health agencies).
Tweet engagement was measured using retweet count. For tweet threads that contained multiple tweets, the maximum number of retweets obtained for any single tweet in the thread was used to measure engagement (rather than the average or total number of retweets for all tweets in the thread) to avoid double-counting or high-biased engagement.
Next, accounts were classified based on province, where applicable, or were otherwise identified as a “national” account (eg, PHAC and Canada’s chief medical officer). This was done to select a stratified random sample of 501 tweets about COVID-19 for a qualitative content analysis (see
Before beginning the content analysis, 3 researchers (CS, CB, SS) were trained on a set of 50 tweets randomly selected from the overall sample of 6982 COVID-19 tweets. This was done so that the researchers could familiarize themselves with the various variables for coding and troubleshoot any issues with the definitions of the coded variables. To distribute the 501 tweets for the content analysis equally among the three researchers, the French-language tweets (n=27) were first identified and allocated to the researcher with fluency in French. Among the remaining English-language tweets, 50 were randomly selected and allocated to each of the 3 researchers so that these overlapping tweets could be used to calculate the Krippendorff α value for interrater reliability. This resulted in one researcher coding 201 tweets (including the 27 French tweets), and the other two coding 200 tweets each. To integrate the 50 tweets that all 3 researchers had coded into the overall sample, one researcher’s code was randomly selected from the 3 possible codes for each variable, such that the probability of selection was proportional to the frequency of that answer (eg, if two-thirds of coders agreed on a code, there was a 2 in 3 chance of that code being selected).
There were 10 coding variables in total. The first variable, media, captured the presence of media in the tweet and the type of media present (ie, image, video or document), if applicable. The next variable, message function, was coded using 3 mutually exclusive coding variables: information, action, or community. “Information” tweets included those whose main purpose was to inform, educate, or update the reader on case counts, disease transmission dynamics, policy changes, and COVID-19 symptoms. “Action” tweets included those whose main purpose was intended to prompt changes in the behaviors or actions of readers, such as encouraging social distancing, hygiene practices, or other harm-reducing behaviors. Finally, tweets were coded as “community” if their main purpose was community-building, identifying community supports and programs, or highlighting stories from or about the local community. Since threaded tweets could contain multiple message purposes, coding was based on the most prominent theme for the entire tweet thread. These variables for message function are consistent with those first proposed by Lovejoy and Saxton [
The final variable, use of a risk communication strategy, was coded using 6 nonmutually exclusive coding variables: corrective, risk, efficacy, concern, uncertainty, and experts. Tweets were classified as “corrective” if they corrected some incorrect information about COVID-19 or aimed to prevent the spread of misinformation. Tweets were classified as “risk” if they contained information that would help a reader make a judgment about the risk of contracting COVID-19 or experiencing health complications from COVID-19. This included tweets containing information regarding absolute risks, relative risks, as well as the identification of high-risk subpopulations. Tweets were classified as “efficacy” if they referenced an individual’s or community’s ability to execute an action or activity successfully resulting in some tangible benefit to health or a reduction of harm related to COVID-19. Tweets were classified as “concern” if they acknowledged the fears, concerns, worry, or anxiety associated with COVID-19. Tweets were classified as “uncertainty” if they acknowledged uncertainty, confusion, or a lack of available information about COVID-19. Finally, tweets were classified as “experts” if they implicitly or explicitly mentioned some agreement, coordination or collaboration between public health experts or other credible health organizations or individuals. The presence of any one of these 6 variables was used to indicate the use of any risk communication strategy in the tweet and were based in part on best practices in communication identified by Seeger [
Each of the 3 coders worked independently through the same randomly selected 50 tweets, where each tweet had 10 variables to be codified. Krippendorff α [
To assess whether differences in the mean number of retweets per tweet across each message function category were statistically significant, the nonparametric Kruskal-Wallis one-way ANOVA (analysis of variance) test was used since our data were not normally distributed.
We used the nonparametric Mann–Whitney–Wilcoxon Test to assess differences in the mean number of retweets per tweet between tweets containing at least one risk communication strategy and tweets containing no risk communication strategy.
Our sample comprised 32,737 tweets, which included individual tweets and threads authored by 118 Canadian public health Twitter accounts (agencies and decision makers) between January 1, 2020, to June 30, 2020. Approximately 21% (n=6982) of all tweets contained content about COVID-19.
Number of tweets and follower counts by account type, January 1, 2020, to June 30, 2020.
Account type | Twitter accounts, n | Total tweetsa, n | Mean tweets per account | Tweets about COVID-19, n (%) | Mean retweets per tweet about COVID-19 | Total follower countb, n (range) |
Public health agencies | 4 | 2272 | 568 | 524 (23) | 60 | 407,546 (10,201-325,112) |
Regional and local health departments | 69 | 19,919 | 289 | 3832 (19) | 10 | 406,108 (194-82,347) |
Provincial health authorities | 15 | 4778 | 319 | 939 (20) | 13 | 170,387 (23-41,779) |
Medical officers of health | 22 | 3859 | 175 | 1337 (35) | 112 | 416,611 (213-206,288) |
Provincial health ministers | 8 | 1909 | 239 | 350 (18) | 52 | 134,019 (908-53,325) |
aEquals the number of tweets and/or tweet threads authored by Twitter accounts per type between January 1 and June 30, 2020.
bCorresponds to the number of followers the accounts had at the end of the study period on June 30, 2020.
Daily rate of COVID-19 tweets by account type, January 1, 2020, to June 30, 2020. WHO: World Health Organization.
Daily retweets per COVID-19 tweet by account type, January 1, 2020, to June 30, 2020. Avg: average; WHO: World Health Organization.
Summed tweet (n=6982) and retweet frequencies and percentages by engagement strategy and account type, January 1, 2020, to June 30, 2020.
Account type and engagement strategy | Number of tweetsa containing each engagement strategy, n (%) | Summed retweets for all tweets containing each strategy, n | Mean retweets per tweet containing each strategy | |
|
||||
|
Media | 350 (67) | 23,291 | 67 |
|
Hashtags | 451 (86) | 29,418 | 65 |
|
URLs | 451 (86) | 24,788 | 55 |
|
User mentions | 139 (27) | 2654 | 19 |
|
||||
|
Media | 2640 (69) | 30,223 | 11 |
|
Hashtags | 2462 (64) | 30,824 | 13 |
|
URLs | 2766 (72) | 27,136 | 10 |
|
User mentions | 753 (20) | 4648 | 6 |
|
||||
|
Media | 477 (51) | 8555 | 18 |
|
Hashtags | 614 (65) | 8333 | 14 |
|
URLs | 669 (71) | 6671 | 10 |
|
User mentions | 326 (35) | 1644 | 5 |
|
||||
|
Media | 342 (26) | 25,924 | 76 |
|
Hashtags | 1058 (79) | 141,841 | 134 |
|
URLs | 681 (51) | 61,697 | 91 |
|
User mentions | 396 (30) | 15,098 | 38 |
|
||||
|
Media | 157 (45) | 5015 | 32 |
|
Hashtags | 213 (61) | 14,261 | 67 |
|
URLs | 214 (61) | 8134 | 38 |
|
User mentions | 135 (39) | 3756 | 28 |
aEquals the number of tweets and/or tweet threads authored by Twitter accounts per type between January 1 and June 30, 2020.
During the content analysis of 501 tweets, 16 tweets were identified by the 3 researchers as not being directly related to content about COVID-19, resulting in a data set of 485 tweets about COVID-19. When coding the tweets for message function, 21 tweets were found to not have a classifiable purpose and were omitted from this part of the analysis.
Summed tweet (n=464) and retweet frequencies and percentages by message function and account type, January 1, 2020, to June 30, 2020.
Account type and message function | Number of tweetsa with each message function, n (%) | Summed retweets for all tweets of each function, n | Mean retweets per tweet of each function | |
|
||||
|
Information | 17 (52) | 957 | 56 |
|
Action | 13 (39) | 554 | 43 |
|
Community | 3 (9) | 36 | 12 |
|
||||
|
Information | 51 (24) | 325 | 6 |
|
Action | 101 (47) | 1043 | 10 |
|
Community | 64 (30) | 223 | 3 |
|
||||
|
Information | 47 (47) | 393 | 7 |
|
Action | 33 (33) | 388 | 12 |
|
Community | 19 (19) | 142 | 7 |
|
||||
|
Information | 56 (58) | 6172 | 103 |
|
Action | 30 (31) | 7765 | 259 |
|
Community | 11 (11) | 223 | 20 |
|
||||
|
Information | 10 (53) | 293 | 27 |
|
Action | 5 (26) | 221 | 44 |
|
Community | 4 (21) | 90 | 23 |
aEquals the number of tweets and/or tweet threads authored by Twitter accounts per type between January 1 and June 30, 2020.
Summed risk communication strategies (n=334) and percentages and retweet frequencies by strategy and account type, January 1, 2020, to June 30, 2020.
Account type and risk communication strategya | Number of risk communication strategiesb used by account type, n (%) | Summed retweets for all tweets containing each risk communication strategy, n | Mean retweets per tweet containing each strategy | |
|
||||
|
Corrective | 1 (4) | 134 | 134 |
|
Risk | 8 (31) | 112 | 14 |
|
Efficacy | 12 (46) | 845 | 70 |
|
Concern | 2 (8) | 154 | 77 |
|
Uncertainty | 1 (4) | 67 | 67 |
|
Experts | 2 (8) | 18 | 9 |
|
||||
|
Corrective | 4 (3) | 119 | 30 |
|
Risk | 5 (4) | 20 | 4 |
|
Efficacy | 81 (60) | 711 | 9 |
|
Concern | 24 (18) | 236 | 10 |
|
Uncertainty | 10 (7) | 90 | 9 |
|
Experts | 12 (9) | 35 | 3 |
|
||||
|
Corrective | 1 (2) | 0 | 0 |
|
Risk | 12 (23) | 38 | 3 |
|
Efficacy | 18 (34) | 129 | 7 |
|
Concern | 10 (19) | 220 | 22 |
|
Uncertainty | 3 (6) | 5 | 2 |
|
Experts | 9 (17) | 38 | 4 |
|
||||
|
Corrective | 4 (4) | 767 | 192 |
|
Risk | 24 (22) | 2709 | 113 |
|
Efficacy | 49 (45) | 12,413 | 253 |
|
Concern | 13 (12) | 3928 | 302 |
|
Uncertainty | 4 (4) | 1433 | 358 |
|
Experts | 15 (14) | 1695 | 113 |
|
||||
|
Corrective | 0 (0) | N/Ac | N/A |
|
Risk | 2 (20) | 123 | 62 |
|
Efficacy | 3 (30) | 175 | 58 |
|
Concern | 1 (10) | 43 | 43 |
|
Uncertainty | 0 (0) | N/A | N/A |
|
Experts | 4 (40) | 53 | 13 |
aRisk communication strategies were not mutually exclusive; therefore, a single tweet could contain multiple strategies.
bEquals the number of tweets and/or tweet threads authored by Twitter accounts per type between January 1 and June 30, 2020.
cN/A: not applicable.
Weekly frequency of COVID-19 tweets containing any risk communication strategy, across all account types, January 1, 2020, to June 30, 2020. WHO: World Health Organization.
Our study of Canadian public health Twitter accounts revealed that tweeting practices and tweet engagement differed between various agencies and decision makers during the COVID-19 pandemic. Of the five account types that we examined, public health agencies and medical officers of health stood out for their tweeting frequency and the high engagement that their tweets received. These two account types had the largest percentage of tweets about COVID-19 relative to all their total number of tweets, as well as the highest daily rate of COVID-19 tweets. This finding is consistent with their roles as the primary agencies and individuals responsible for implementing and communicating a public health response during health crises in Canada. Our findings showed that these two account types also received the highest average engagement (retweets) during the study period, suggesting that Twitter users also recognize these agencies and individuals as the primary information sources to share with their peers during a public health crisis like COVID-19. These findings are positive in that if public health agencies and medical health officers continue to establish a consistent Twitter presence, attract followers, and engage their Twitter audiences, their communications may continue to be shared widely.
We observed that trends in the daily frequency of tweets by account type over time were generally consistent with changes in public concern and engagement over the course of the pandemic [
It is also interesting to note that daily rates of COVID-19 tweets authored by public health agencies seemed to peak around the same time that key moments related to COVID-19 took place in Canada (ie, after Canada’s first case, just before the first peak in COVID-19 cases, and shortly after the second peak in COVID-19 cases), which suggests that trends in case counts may have at least partially informed this account type’s tweeting practices. This trend was not seen for medical officers of health, whose daily rates of COVID-19 tweets trended downward after the WHO declared COVID-19 a pandemic, despite subsequent peaks in Canadian COVID-19 cases. This is important because increasing COVID-19 case counts partially inform the public about their risk of contracting COVID-19 and thus may lead to increased information seeking. As a result, more public health accounts should respond to increasing case numbers with increased daily tweeting about COVID-19, not less. Unfortunately, information seeking cannot be indicated by engagement metrics such as retweet counts; therefore, our results do not reflect which tweets were most seen or read by Twitter users. However, other research has noted that frequently updated social media feeds are perceived by audiences as more relevant during crises and should be updated enough so as to be noticed by users, regardless of whether they receive engagement [
Despite having lower daily COVID-19 tweet rates compared to public health agencies, medical officers of health and provincial health ministers received the highest daily retweets per tweet on day 80, 10 days after the WHO declared COVID-19 a pandemic. This finding was somewhat surprising given that past research has identified health agencies (especially at the federal level) as being able to garner more retweets than other public health accounts during past pandemics, largely due to their recognition and influence over online information sharing [
One reason that medical officers of health may have been able to maintain a higher average retweet count for the remainder of the study period is that Twitter users may have perceived their content as more medically or scientifically relevant during the pandemic, which other studies have found can lead to more retweets [
The frequent use of hashtags and URLs in the majority of COVID-19 tweets that we analyzed is consistent with other studies examining engagement to tweets authored by health agencies [
In our content analysis of a stratified random sample of tweets by Canadian public health agencies and decision makers, we found that of the three message functions that we examined, information tweets were most common across all account types, except regional and local public health departments, which used action tweets more frequently. This finding was consistent with other studies that have found tweets conveying information to be the most frequently tweeted by health organizations during other pandemics [
Furthermore, the results of our content analysis demonstrated that the risk communication strategies that we examined were not very widely used by any account type, appearing in just over half of the tweets that we analyzed. For example, our study found very few tweets provided corrective information that could be used to tackle misinformation about COVID-19, which is consistent with work by Guidry et al [
Despite the overall lack of risk communication strategies employed in the tweets in our sample, our findings demonstrate that including one or more strategies was associated with more engagement on average compared to tweets that did not contain any risk communication strategies (61 retweets per tweet versus 13 retweets per tweet, respectively). We also found that risk communication strategies tended to be used at key moments during the COVID-19 pandemic. The use of risk communication strategies appeared to peak in the weeks just after the WHO declared COVID-19 a pandemic and trended slightly upward after each of the two peaks of COVID-19 cases in Canada. This finding is consistent with other studies that have found that risk communication becomes less prevalent over the course of a crisis [
With so much discussion of the pandemic online, supplying information users with high-quality, consistent messaging on the health risks associated with COVID-19 is critical for improving health literacy in the population [
One methodological advantage that sets our study apart from others is the use of tweet threads, in addition to individual tweets, as the unit of analysis, rather than analyzing each tweet from a thread on its own. Analyzing threads allowed for a more accurate examination of tweeting practices by recreating the message as it would have been viewed originally to a Twitter user. This was a major strength of our content analysis as it allowed for the entire message to be coded and analyzed, rather than a small segment of it. Since threads are commonly used to craft messages that would otherwise be impossible to fit within a tweet’s character limit, analyzing them individually would have provided an incomplete picture.
This research has several limitations related, in large part, to its reliance on Twitter data. First, we analyzed the tweets of public health agencies and decision makers in Canada who had tweeted between January 1 and June 30, 2020; however, there are numerous other authoritative health-related Twitter accounts and other official government accounts that tweeted about COVID-19 during this period (eg, physicians, health nongovernmental organizations, etc), which the public may have also engaged with. Moreover, not all members of the public use Twitter; therefore, engagement as measured by counts of retweets does not offer insights into which public health agencies or decision makers the broader Canadian public may consider most important. Additionally, our results may not reflect today or tomorrow’s Twitter landscape, and therefore only indicate tweeting practices at a snapshot in time. However, our study can offer a glimpse into trends on information sharing during the first 6 months of the COVID-19 pandemic based on what Canadian public health agencies and decision makers were tweeting and how Twitter users engaged with this content. Our study’s findings can be useful to those public health accounts that we included in our analysis as well as to other health organizations or individuals that may be looking for ways to better utilize Twitter to engage with users seeking health information on this platform.
The findings from our study could be improved through additional analysis of content authored by public health accounts on other social media platforms (eg, Instagram, Facebook, etc) or additional refinement of the categories we used for classifying the Twitter accounts and tweet content. For example, future work examining public health communications in Canada could build on our work by contrasting tweeting practices by province or other geographic elements, which could uncover more trends in information sharing and engagement given the region-specific administration of Canada’s public health policies. In addition, future work in this area could explore patterns in retweeted tweets and perform a network analysis to examine the Twitter interactions between various public health accounts.
This study analyzed tweets by Canadian public health agencies and decision makers between January 1, 2020, and June 30, 2020, to examine their tweeting practices during the early stages of the pandemic. We also aimed to identify ways that tweets could be improved to effectively communicate risk and maximize engagement on this platform. Using a mixed methods approach, we conducted Twitter analytics and a qualitative content analysis to characterize the level of engagement that COVID-19 tweets authored by Canadian public health accounts received, as well as the purpose of their tweets and their use of key risk communication strategies. Our research findings suggest that while public health agencies authored more daily COVID-19–related tweets than other account types, engagement tended to be higher for tweets authored by medical health officers, particularly during key moments of the pandemic. Overall, most account types appeared to focus on disseminating information, with the exception of regional and local health departments, which tended to promote more action from users. Our results also point to the need for public health agencies and decision makers to monitor Twitter analytics in order to understand their audience and leverage whatever Twitter engagement strategies help maximize shares of their communications.
During the beginning of any crisis, the use of risk communication strategies by organizations and officials leading the response is critical to help inform the public about an often highly uncertain and rapidly evolving situation, address concerns, and instill trust in those leaders. Our study found that risk communication strategies were not widely used in tweets by any account type, even though these strategies were associated with more engagement. These findings suggest that Canadian public health agencies and decision makers may be missing an important opportunity to engage with information users about the mitigation of health risks related to COVID-19. Finally, our study builds on other literature that has explored differences in engagement to communications authored by individuals and institutions, and suggests that any medical officer of health or other expert individual currently not on Twitter should consider the platform as a means to disseminate information that Twitter users appear to be interested in sharing.
Flow chart of Tweet exclusion process.
Dataset of manually coded Tweets (n=501) from content analysis.
analysis of variance
Public Health Agency of Canada
research question
severe acute respiratory syndrome
World Health Organization
This project is supported in part by funding from the Canadian Social Sciences and Humanities Research Council (grant: 435-2020-0257).
None declared.