Introduction
At the outset of the COVID-19 pandemic, there was unprecedented social media use, especially from the young and adults conversing opinions through these platforms as their sole communication forums and continued during the pandemic when other personal communication outlets like person-to-person communications were limited due to the lockdown restrictions. This trend continued throughout the vaccination phase to the end of the three prominent waves of the pandemic. The public health governance sector used Twitter (now X) as one of its outreach outlets (Slavik et al., 2021). These rapidly growing communication platforms soon became the sources for researchers and public health governance sectors to uncover public concerns about surveillance efforts. Researchers suggest Covid-19 vaccine uptake data collected from these communication platforms provides real-time insights (Slavik et al., 2021). Social media-based communication has proven to be a valid source of opinions. A US study confirmed a positive significant correlation between vaccine tweet counts and vaccine uptakes. Therein, Tweet opinions is shown to be in alignment with public vaccine uptake (Nelson et al., 2024).
The first COVID-19 vaccine was approved in Canada on December 9, 2020, and vaccination started on 21, December 2020 in British Columbia (BC) and on December 15, 2020, in Ontario (ON)(Health Infobase, 2024), the two study provinces. Throughout the vaccine and delivery processes, the general public and government officials used social media platforms to launch mass media campaigns, exchange public opinions, deliver government announcements, and even to respond to public reactions toward vaccines. Twitter (the platform now known as X) was one of the most widely used social media platforms in Canada and between 2019-2021, 7.4 million Canadians used Twitter and the gender distribution in 2016 showed 48% of female users (Statista, 2023), thus depicting almost equal gender representation. Negative attitudes towards vaccine safety and efficacy played a key role in vaccine hesitancy and social media championed in vaccine-related communication arenas (Cascini et al., 2022). Even further, the ability of social media content to uncover public sentiments towards vaccine communication for social research purposes has been demonstrated (Rotolo et al., 2022a). These previous research-based evidence sets the basis of the current research.
According to the Government of Canada, (Government of Canada, 2024a) the overall fully vaccine rate in Canada is 16.3%. The percentage of people vaccinated with at least one dose is 87%, and Canada falls among the most highly vaccinated countries in the world (World Health Organization, 2024). The percentage of fully vaccinated varies by province and British Columbia (BC) reported the highest (26%) and Ontario the lowest (11.1%) (Government of Canada, 2024a). COVID-19 vaccination began in Canada on December 14, 2020, (Government of Canada, 2024b). The current study period of before and after vaccine analysis includes 01-12-2020 to 23-12-2020.
Demographic characteristics and political attitudes were found to correlate with vaccine attitudes in Canada (Gravelle et al., 2008). Due to the overwhelming social media use, these outlets became a major driver for influencing the attitudes toward vaccine uptake in Canada (Boucher et al., 2021). One of the attitudes that received mixed opinions among global social media users was the proof of vaccine, of which positive attitudes were geared towards ease of traveling and negative attitudes were around depriving human rights and freedom (Khan et al., 2022). To achieve future vaccine campaign success, public health officials need to be aware of the context and patterns of public communication on social media. Canadian Twitter data sentiment analysis on COVID-19 vaccine attitudes revealed a mixture of negativity, expressing possible side effects and positivity depicting optimism, communicating vaccination as the only way to end the pandemic(Jang et al., 2022). However negative attitudes expressed on social media were strongly influential in vaccine uptake decision making. Twitter data analysis of Canadian vaccine hesitancy expressions found, among other factors, a link to a lack of public knowledge and the government’s inability to communicate safety and efficiency information (Griffith et al., 2021). These geographic and temporal variants of vaccine attitudes between and within countries hinder international and national efforts to launch a successful vaccine campaign. One Canadian qualitative study that analyzed vaccine-related newspaper editorial cartoons found an attitude shift from positive to negative, over a span of time, expressing high hopes at the beginning and dissatisfaction at the end (Pelletier et al., 2023) and thus endorses the need to explore temporal variations in research. The above study researchers procrastinated on possible challenges this temporal shift may impose on maintaining a high vaccine uptake (Pelletier et al., 2023).
Social media platforms have become a free-flowing data-rich source. Uninterrupted by researchers questioning and thereby guiding the thoughts in one direction, these natural flows of rapidly growing ideas on a topic provide opportunities for digital health researchers to uncover existing trends and patterns related to the current topic of public interest. The caveat to this source is the reliability of information and ability to detect prevalence of misinformation and disinformation sharing, (Patrick et al., 2022), which may lead to subsequential influence of public discourse towards undesirable directions. Health misinformation circulation in social media consists mostly of the unverified content by subject experts, that can be categorized as “false, inaccurate, or misleading” (Gottlieb & Dyer, 2020). Next to that is the lack of analytical tools and frameworks to identify and verify the reliability of information to be used in public health decision-making (Fernández-Luque & Bau, 2015). Cognitive bias stemming from social media framing of vaccine (mis)information is well documented in the literature and the need for public health experts to intervene is recommended (Patrick et al., 2022; Skafle et al., 2022).
The aim of this paper is to explore the framing of COVID-19 vaccine communication via tweets in the two provinces with high and low vaccine uptake in Canada. Using the time trends (before and after vaccine arrival) associated with social media health communication polarity, related to negative and positive sentiment expressions to vaccine uptake, we investigate the geographical-level ecological relationship between Twitter sentiments and vaccine uptake at the provincial level. We chose the two Canadian provinces British Columbia (BC) with the highest vaccine uptake and Ontario (ON) with the lowest vaccine uptake for this geographic-level ecological investigation (Health Infobase, 2024). Motivated by our previous research that revealed the association between COVID-19 caseloads and emotional exacerbations measured through sentiments expressed via Twitter data (Weerasinghe et al., 2023a), this paper demonstrates the use of AI-based machine language analytical tools and frameworks to uncover emerging salient topics unfiltered by researchers’ judgments. Our dataset contained 543,630 tweets covering the period of 01-12-2020 to 23-12-2020.
Through the lens of framing theory, we analyzed 20,000 Canadian Tweets related to vaccines to uncover positive and negative emotional sentiments and frame these themes into meaningful actionable subframes to inform future vaccine propagation efforts. The sentiment analysis results illustrated in the current paper will primarily on the negative emotions, to make recommendations to improve vaccine uptake in future vaccine delivery efforts. We will outline pro-vaccine expressions to understand the pattern to provide guidance for future vaccine promotion efforts. This article contributes new knowledge on how social media framing relates to health interventions and public health policy changes, across geographies and periods in Canada. We first illustrate negative sentiment changes over time and geography, and then illustrate counteractive positive sentiments around public health surveillance decisions of vaccine mandate and delivery.
COVID-19 vaccine safety and efficacy were two of the most public health controversies that arose during the pandemic in Canada (Dubé et al., 2022). There were massive public campaigns against vaccine propaganda in Canada and social media has been extensively used to mitigate antivaccine and pro-vaccine messaging. The following literature review provides a summary of research-based evidence on COVID-19 vaccination, primarily focusing on Canada, but also covering global evidence on social media communications. We intend to cover both pro- and anti-vaccine polarized environments. Analytical frameworks and social theories pertaining to the topic are also reviewed.
Literature Review
Several Canadian vaccine communication studies have used social media platform data. A qualitative study that analyzed contents covered in the Canadian Broadcasting Cooperation news and commentaries, categorized relevant comments into concerns mostly arising along the line of vaccine safety and effectiveness; criticism of the government; sources of information used to support posts, and challenges and misinterpretations (Rotolo et al., 2022b). The analysis led to recommendations towards taking action to communicate facts, quoting evidence-based information to foster vaccine acceptance (Rotolo et al., 2022b). Another Canadian vaccine study that used 605 tweets collected from December 18 to December 23rd, 2020, analyzed the contents using the theoretical domain framework and found issues related to vaccine safety, lack of knowledge, political influence, antivaccine messaging, and lack of liability from vaccine companies (Griffith et al., 2021). A qualitative study that analyzed cartoons that appeared in Canadian newspapers, from January 2020 to August 2022, noted criticism about unvaccinated and the effectiveness, among other themes, in which the criticism was framed as coming from vaccine fatigue (Pelletier et al., 2023). Through the contents of these study findings that analyzed vaccine social media communications at different time frames, safety and effectiveness, government involvement, and issues with information sources were common themes that emerged. However, the discrepancies noted in the Canadian literature were time- and geography-sensitive.
A study that included Twitter and Reddit vaccine sentiments, compared four countries, Canada, the UK, the US and India, and revealed, that in the middle of the year 2021, Canadian Tweets were the most pessimistic, and hesitancy expressions were second highest in Canada (Kaushal et al., 2023). Researchers recommended tracking public attitudes towards vaccines, considering geographic variations to respond to geo-community-specific concerns (Kaushal et al., 2023). Another multicounty study that compared vaccine-negative Tweets across different vaccine types among Canada, the UK and the US found variations over country, time, and vaccine types (Yiannakoulias et al., 2022). The current study fills the gap in Canadian knowledge on temporal and geographic variations in vaccine sentiment communications.
Conspiracy theory and misinformation
A popular conspiracy theory belief underlines the social media influence of (mis)information spread about COVID-19 vaccines worldwide (Farhart et al., 2022), and few studies have confirmed the negative impact on diverse communities in Canada (Burns et al., 2024). Besides the conspiracy theorists have argued a two-pronged contextualization. First contextualization was, one conspiracy has been resulting from or rooted in a belief in a different conspiracy, and the second formulation is that a conspiracy is rooted in the social context wherein stigmatized marginalized groups are becoming subject to the social media and political influences (van Prooijen & Douglas, 2018). Those who believed in or experienced adverse reactions to previous vaccines may spread (mis)information about the COVID-19 vaccine through social media. On the other hand, certain marginalized groups, with diminished institutional trust may cast doubt on the effectiveness of the vaccine. A scoping review of 19 articles reported approximately 96% of Canadians receive misinformation through social media platforms indicating vaccines will be harmful to some racial minorities (Kemei et al., 2022).
Theoretical frameworks
Several theoretical frameworks have been used within the global vaccine research context. The most used was the framing theory. Social media messaging research has used the theory to “identify and label” shared experiences and opinions and applied it to vaccine information exchange via social media platform advertisements (Bradshaw, 2023) and another researcher has framed this as infodemic, a concept used to elaborate the wide use (Mohammadi et al., 2022). Framing theory and there upon the framing analysis, equip communication in to systemized concepts of frames to identify and label perception of events circulating in the social world (Entman, 1993). Within the context of this paper, researchers who reviewed health communication related published articles that included framing theory cited vaccine communication as one of the most areas of the application among others like cancer, nutrition and obesity (Guenther et al., 2021). However, very few studies have applied the theory to social media Tweet analysis for COVID vaccine hesitancy. One study that used Amazon Mechanical Turk social media platform data applied framing analysis to COVID vaccine willingness and they highlighted complexities associated with willingness impact framing and identified dependency on skepticism (Ademu et al., 2023).Research on COVID-related news media coverage applied the framing theory to identify the common problems “framed” in the text data and the major components arose were identified depicting “dominant frames”(Reynolds, 2022). Based on the applications to social media news coverage, the theory has shown flexibility and capability to capture the chronological variations in the frames (Reynolds, 2022) and this applies to our proposed study framework wherein we intend to capture the health communication messaging influence on vaccine uptake temporal variations. The intermediate process of social media coverage of vaccine sentiments and vaccine uptake messaging was the mass advocacy campaigns launched in Canada, inclusive of those who were both pro and against COVID-19 vaccination. The framing theory has been applied to those situations, capturing social movement framing influence identification in social media research during the pandemic (Sorce & Dumitrica, 2023). Further research revealed that the outcomes produced, in the current study context, the choice of vaccine, and the decision to get vaccinated, can be influenced by the way the media frame the problem, in either direction, positively or negatively (Park & Reber, 2010). Two research groups used the psychological principle of the perception changes depending on the way the situation is framed (Park & Reber, 2010; Tversky & Kahneman, 1981). The current study adopted framing theory as the theoretical framework suitable for Tweet data analysis and interpretations.
Methods
Figure 1 shows the methodological steps for acquiring and processing COVID-19 vaccine-related tweets. Permission was obtained from X Corp. (formerly Twitter Inc.) through their Academic Research Track to perform a full-archive tweet search over a long period. For this article, we used a subset of tweets covering the period of 01-12-2020 to 23-12-2020 using geo-tagged tweets from Ontario and British Columbia. The first phase of Twitter data extraction shown in Figure 1, is described in detail elsewhere (Weerasinghe et al., 2023b).
In Phase 1, we extracted tweets and associated metadata (within the period 01-Dec-2020 to 28-Feb-2021) using the following hashtags: #Vaccine, #Vaccines, #CovidVaccine, #ImVaccinated, #IGotVaccinated, #VaccinesSaveLives, #VaccinesWork, #Vaccinated, #VaccineForAll, #GetVaccinated, #vaccination, #vaccineinjury, #VaccineEquity, #covax, #COVIDvaccine, #GetTheShot, and #KeepVaccinating. In addition, keywords related to vaccine were used in the search query, such as vaccine, vaccines, vaccination, immunize, immunise, immunization, immunisation, inoculate, and inoculation. To ensure that extracted tweets are related to COVID-19, we included additional criteria to retain only tweets that contain at least one of the following keywords: coronavirus, covid-19, covid19, and SARS-CoV-2.
Next, we preprocessed the extracted tweets (n=13,433,615) by applying the following natural language processing (NLP) techniques using the Natural Language Toolkit (NLTK) and Python: (1) Removal of hashtags, mentions, and URLs, (2) Removal of punctuation and other special characters, (3) Removal of numbers, (4) Removal of stop words, (5) Lemmatizing words to convert them to their root form, and (6) Removal of duplicates. Afterwards, we geotagged the preprocessed tweets (n=7,638,361) using our GEOTAGGING algorithm described in (Weerasinghe et al., 2023b). Of the 543,630 tweets that originated from Canada, 67,923 and 214,768 tweets emanated from British Columbia (BC) and Ontario (ON), respectively. Of these, we sequentially selected 10000 distinct Tweets related to the COVID vaccine, in each province, which included 5000 tweets covering the before-vaccine arrival period of 01-Dec-2020 to 09-Dec_2020 and another 5000 Tweets covering the after-vaccine arrival period of 20-Dec-2020 to 28-Feb-2021 in BC and in ON. Altogether we analyzed 20,000 tweets in both provinces. Altogether we analyzed 20,000 tweets in both provinces. Rationale for choosing 5000 Tweets is firstly to accommodate analytical capacity of the software that we used for analysis (RapidMiner) and we limited to one week of data acquisition period of the event of interest, which seems to be in par with the time period of other Tweet studies.
Phase 2 analysis (Figure 1) involves machine language-based analysis which was carried out using RapidMiner software (Mierswa, Ingo, 2018). Data preprocessing was carried out first subsetting BC and ON tweets from the Canadian Tweets and then filtering out vaccine-unrelated terms. We used the remove duplicate operator in RapidMiner, to remove duplicates and selected the first 5000 tweets (from 01-Dec-2020 to 09-Dec-2020) and the last 5000 tweets (from 20-Dec-2020 to 28-Feb-2021) covering the two study periods in BC and ON. We used the natural language processing-based sentiment classification system, VADER (Valence Aware Dictionary for sentiment Reasoning) (Hutto C.J. and Gilbert, 2014) to uncover positive and negative sentiments in the Tweets using the extract sentiment operator in RapidMiner. The VADER assigned a score between -4 and +4 based on the sentiment expressed in the Tweets; higher numbers represent greater sentiment strength, and the sign represents the direction of the valence, a negative sign indicates a negative sentiment, and a positive sign indicates positive sentiments. Zero was assigned to neutral sentiments. Then we removed neutral sentiments and positive and negative sentiments were used for Latent Dirichlet Allocation (LDA) topic modeling.
Topic modeling has been widely used in social media health communication research data analysis. Two of such applications are early COVID pandemic assessment (Chipidza et al., 2022) and a special application to Tweet data analysis that compares several media coverage within a given period and found high level of coherence(Pinto et al., 2019). LDA is a widely used method for social media data topic modeling (Blei et al., 2003), in which documents are automatically classified into salient topic groupings, using a method called TF-IDF, which stands for term frequency-inverse document frequency(Ljajić et al., 2022). TF-IDF is a widely used method in natural language processing to uncover the importance of a topic in relation to all the topics, in the document. Salient groupings that emerged, containing topics with higher weights (TF-IDF values) were obtained at each period for each province. After obtaining automated groupings we used framing theory to provide meaningful interpretations to the groups. LDA remains a valuable tool for topic modelling where topics are distributions over words and documents are distributions over topics, thereby making the results interpretable (Blei et al., 2003). However, LDA treats documents as unordered collections of words, ignoring word order and context, which can limit semantic understanding. BERT-based models address this gap by incorporating semantic context through deep learning in that they use transformer architectures to capture word context and semantics, considering word order and surrounding words. Yet, BERT-based models are not only resource-intensive but also black boxes, which makes it harder to interpret how topics are formed compared to LDA’s probabilistic distributions. Therefore, the choice between LDA and BERT-based models depends on the specific requirements of the task, such as the importance of interpretability versus the need for capturing nuanced semantic relationships.
Results
Before vaccine arrival in BC, the province with the highest vaccine uptake in Canada, there were 1313 (26%) negative, 2294 (46%) positive and the rest were neutral sentiments. During the same period in ON, the province with the lowest vaccine uptake, there were 1158 (23%) negative and 2663 (53%) positive sentiments, before vaccine arrival. In ON negative sentiments increased after vaccine arrival to 34% and positive sentiments decreased to 41%, after vaccine arrival. In BC negative and positive sentiments were roughly the same after vaccine arrival (28%) and (44%). Figure 2 shows topic modeling results and we noticed appearance of some recurrent topics across geographies and periods. Dominant negative sentimental topics that emerged were categorized into four intertwined frames: Misinformation and conspiracy beliefs; Antivaccine and hesitancy; Lack of faith and skepticism and Fear, death and worry.
In BC, misinformation and conspiracy sentiments declined by 20%, after vaccine arrival, and antivaccine and hesitancy communications dropped by 9% (Table 1). However, after the vaccine arrival, lack of faith and skepticism grew by 19% in BC. Though fear, death and worry sentiment prevalence increased in BC by 41%, 11% of those sentiments took a positive spin stating “not to worry” about the vaccine. After removing double negating expressions, the after-vaccine negative sentiments in this category rose to 57%, still showing a remarkable increase from before-vaccine arrival sentiment percentage. Following the same trend as in BC, in ON, the misinformation and conspiracy belief communications declined by 16% after vaccine arrival. So did the antivaccine and hesitancy sentiments in ON dropped by 18%. Unlike in BC, lack of faith and skepticism dropped down in ON by 6%. However, fear, death and worry increased by 22% of which worry was mostly associated with negating (38%) in Ontario, stating “not to worry” about the vaccine. Taking this into account “fear, death and worry” increased by 22% in ON after vaccine arrival and was stable in BC (Table 1).
Health information mavenism-Misinformation and conspiracy beliefs
Rooted in the market information exchange, mavenism has been introduced to health communication pattern exploration among the public via interpersonal connections (Hayashi et al., 2019). In that venue, mavens, defined as those who spread information, are considered to have limited subject expertise (Hayashi et al., 2019) and a recent study defined as opinion leaders in a healthcare community who share up-to-date health knowledge (Smith et al., 2022a).
Even before vaccine arrival, social media users in Canada expressed concerns about the spread of misinformation and conspiracy beliefs around the COVID-19 vaccine. A few examples of misinformation circulated among tweeters are listed in Table 2. These Tweets came as warning messages to the government. Tweeters pointed out the consequences of misinformation, as hindering government efforts to combat the spread of the virus. Social media users framed misinformation as “fake news” and stated this became a “second pandemic” and quoted reliable sources of world organization news pages pleading with government organizations to take necessary steps to stop the spread of this so-called “second pandemic”. Tweeters not only express their own opinions, and they are always backed by sources of information link sharing. The conspiracy theory is framed in two ways, by “a weird pseudo” theory and by the QAnon theory (Table 2). Though no strong evidence is available to show that the social media platform has been used to spread misinformation, there are some news flashings without concrete evidence. Conversely, the myths were also busted with scientific evidence by some tweeters.
After vaccine arrival and delivery social media users continued communicating devastating consequences of misinformation but on a lighter tone. They also added, “Misinformation online can have devastating consequences”. Like BC Tweeters, Ontarians articulated the responsibility of social media. Furthermore, the evolving types of conspiracy beliefs were also identified, in ON and one conspiracy theory has led to another as Ontario tweets related the conspiracies moved from United States such as QAnon, a widespread unfounded theory (Table 2).
Constellation of insecurities - Antivaccine and hesitancy sentiments
Before vaccine arrival in both BC and ON, antivaccine movement was noted as a big threat to government campaigns in BC and this was seen as hostile propaganda coming from social media outlets such as Facebook in Ontario. Labeling as an “anti-vaxxer” is seen as an insult in some cases and racially motivated to label some marginalized groups who are hesitant to receive the vaccine (Table 3).
At the planning stage of vaccine delivery, the priority groups were identified. The healthcare professionals were at the forefront of the priority list, due to their high level of exposure to COVID-19 patients and to avoid transmission through them to other patients. As BC Tweeter said healthcare professionals’ hesitancy came from the undesirable notion of becoming experimental subjects, another misunderstanding (Table 3). Based on an Ontario tweets there were expressions related to government taking precautionary measures to mitigate reassurance among the skeptic healthcare workers. Pro-vaccine groups identified anti-vaxxer as an insult used by hostile propaganda and misinformed movement.
After vaccine arrival, the hesitance continued in Ontario as one Tweeter put it is a constellation of multiple negative personal emotional sentiments (Table 3). There is a shift of sentiments from being more objective before vaccine arrival to becoming more subjective after, perhaps because misinformation that was circulated has been replaced by scientific data later. The need to overcome vaccine hesitancy over the issue of antivaccine and the issue of historically marginalized groups becoming experimental subjects of vaccine testing in the past were emphasized as fundamental issues to be addressed. Information sources for mitigating the challenges of anti-vaccine misinformation were also highlighted.
Pandemic of mistrust-Lack of faith and skepticism
The Tweet communications revealed mistrust and lack of faith in the public health systems, and this is framed as creating skepticism about vaccine delivery and administering mechanisms that hinder safety and efficiency (Table 4). Tweeters indicated these lingering sentiments are not about vaccines but related to systemic issues. In Ontario, lack of faith was tied to mistrust over the authorities who share information which is seen as hampering vaccine distribution efforts. The tweeters framed it as a pandemic of mistrust. Vaccine skepticism is another concept that is centered around the public health system. In Ontario, skepticism came from other countries such as the UK being one of the first countries to approve the vaccine without proper regulatory approval.
After vaccine arrival, Tweeters noted escalating death tolls and while the public was losing faith public health officials were putting faith in vaccines by showing them receiving vaccines on mass media.
Opinions and perceptions - Fear, death and worry
Fearing was related to rising COVID-19 deaths and doubts about vaccine effectiveness and these perceptions were more prominent before than after vaccine arrival (Table 5). Lingering conspiracy theories of genetically engineered vaccines and people dying during vaccine trials added to the fear and worry. Contrary to Ontarian Tweeters expressing negative sentiments about their worries, BC communications were geared toward minimizing worrying about vaccine efficacy instead of prevention efforts of the virus spreading and accessibility to vaccines.
After the vaccine’s arrival, a sense of relief was expressed by those medically vulnerable, and “fear of needles” was also expressed by some. There was added fear and worry about mandating to carry vaccine cards, what was framed as “immunity passports”. So fear and worry take different facets at different stages of the vaccination campaign. In Ontario, vaccine delay was seen as comforting to those who fear needles. Also, vaccination is seen as a government effort to lift lockdown, and fear of death due to rising COVID was expressed.
Discussion
Our social media study on vaccine Tweet communication exchange text mining and framing analysis using topic modeling revealed time and geographic varying patterns of negative and positive sentiment exchange, before and after vaccine arrival in two geographies in Canada with the highest (BC) and lowest (ON) vaccine uptake. Our machine learning-based sentiment analysis and topic modeling suggest geographic differences in antivaccine sentiments and hesitancy expressions before vaccine arrival, with the lowest uptake province showing higher negative sentiments, across the period. The four negative sentiment topics that emerged were framed as: health information mavenism; the constellation of insecurities; the pandemic of mistrust and opinions and perceptions.
Three types of misinformation found in a review of 45 articles based on 7 countries’ social media misinformation exchange were vaccine development, medical information, and conspiracy claims (Skafle et al., 2022) coincides with the misinformation that we uncovered within two Canadian provinces. A study that analyzed masks and vaccine-related Tweets during the first 6 months of the pandemic found tweets with misinformation on vaccines were mostly on political nature (Trotochaud et al., 2023). The current study Tweets were collected during December 2020, six months after the aforementioned study and similar misinformation were lingering in tweets. Conspiracy beliefs and misinformation spread framed as health information mavenism in the current study suggested two types of vaccine-related health mavenism: misinformation circulation from the public and retweeting by subject experts and government officials demystifying with scientific evidence. These two types align with the literature-suggested definition of health mavens as opinion leaders of health communication (Smith et al., 2022b). There is evidence in the literature that conspiracy and misinformation or disinformation spread on social media threaten compliance and correlates with vaccine hesitancy and uptake (Skafle et al., 2022). Though the same individuals were not followed over time, we found a decline in misinformation circulation in both provinces, more so in BC with the highest vaccine uptake than in ON (the capital city) with the lowest uptake. Ontario’s misinformation came from opposition political parties showing the government’s inability to guarantee vaccine safety. The health mavens who demystify the misinformation were stronger communicators in the two study provinces than the public who communicated vaccine side effects misinformation.
Our study data framed both anti-vaccine sentiments and vaccine hesitancy, as arising from a constellation of insecurities, all of which decreased after vaccine arrival (Table 1). The contributing factors to the constellation were, historical exposure of Black people as vaccine experimental subjects, fear of needles, side effects and anti-vaccine conspiracy theories, which was noted as the strongest contributor of vaccine hesitancy through online sample surveys (Farhart et al., 2022). The COVID vaccine hesitancy cross-sectional Canadian study results, found ethnicity, political affiliation, and beliefs in vaccine conspiracies were associated with hesitancy (Burns et al., 2024). Another online survey among participants from UK, US and Canada found Canadian and US participants’ COVID vaccine hesitancy is more related to political affiliation than in the UK (Gravelle et al., 2008). Our data also suggest vaccine supporters condemning the propaganda against anti-vaccine movements and this is similar to the findings of two studies on Canadian vaccine tweet analysis (Gravelle et al., 2008; Griffith et al., 2021). Congruence of these survey study findings add to the credibility of our findings on social media vaccine hesitancy.
Pre-vaccine arrival lack of faith and skepticism expressions percentage, which was higher in ON than BC, declined post-vaccine arrival as opposed to a post-vaccine increase shown in BC. The tweet expressions of lost faith and skepticism were related to how government officials were handling vaccine delivery mechanisms and regulatory approval processes. A four-country twitter data analysis noted the highest rate of time-varying attitudes towards vaccines in Canada compared to UK, US and India (Kaushal et al., 2023). Lack of faith, sprouted from mistrust around safety concerns were the lowest rated in our study contrasting another study on Canadian tweets which noted trust as the most dominant emotion tweet (Jang et al., 2022). Another Canadian Tweet study noted vaccine manufacturers release of effectiveness information diminished lack of faith and skepticism (Lyu et al., 2021). Health mavens may have played a role to increase faith and decrease skepticism, as noted in the literature (Smith et al., 2022a). Though our study was the first Canadian study that pointed out a lack of faith and skepticism due to system issues contributing to hesitancy, a study in India found regulatory agencies’ lack of transparency and inappropriate expedited processes generating skepticism (Sarkar et al., 2021).
Our last frame of fear of death and worry about the COVID-19 vaccine side effects received the lowest ranking in both study provinces, which increased in Ontario, after vaccine arrival. Our data suggest time-varying contexts of fear and worry, started from conspiracy theories and misinformation about death due to vaccine, worrying about vaccine efficacy, fear of needles, worrying about vaccine passport mandates and possibilities of rising death tolls due to post-vaccine lifting of lockdown measures. A longitudinal study conducted in UK confirmed continuing fear and worry about COVID to the level that impairs peoples’ sleep quality (Quigley et al., 2023). There was no other Canadian study that explored effects of COVID vaccine-related fears and worries, but confined to COVID death (Taylor et al., 2020).
Conclusion
The temporal and geographic patterns expressed through social media platform Twitter resonate with public opinion on vaccine hesitancy and uptake and provide a basis for future public health surveillance program planning. The current study finding of misinformation and conspiracy claims was comparable to the other social media studies in different countries and periods. However, no study has shown patterns of variations before and after public health intervention of vaccination across distinct frames and themes. The current study adds to the growing literature in this area about the potential of social media in public health surveillance and possibilities of subject experts and government officials demystifying vaccine myths using scientific evidence to reduce the influence of misinformation mavenism. Additionally, our study adds to the body of knowledge about within-country geographic differences in the vaccine hesitancy movements and some frontline healthcare professionals feeling hesitant due to being in the vaccine priority group because there is a lack of long-term effects studied. Even further, our study was the first Canadian study that pointed out a lack of faith and skepticism due to system issues impeding vaccine uptake and also uncovered fears and worries arising due to lifting lockdown measures.
Limitation
The study on social media communication data analysis has several inherent limitations that are common to other similar studies. VADER sentiment analysis was not sensitive to capture sarcasm and manual coding of salient topics that combine string of words is strongly advised. Foremost there are no demographic diversity captured in the tweets and research suggest low sensitivity (68%) in geotagging for location detection (Compton, R., Jurgens, D., & Allen, 2014). Though tweets mention about racial minority and gender specific information the demographic diversity of individuals who expressed vaccine sentiments were not captured. Further research may need to confirm the social media study findings using online sample survey data to better understand demographic diversity.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
