Vaccine hesitancy and skepticism can be triggered by anxiety about possible side effects and concerns related to novel vaccine technologies, such as the messenger RNA (mRNA) which can be used as a reason for not receiving (the COVID-19) vaccine [1]. The University of Lund study: “Intracellular Reverse Transcription of Pfizer BioNTech COVID-19 mRNA Vaccine BNT162b2 In Vitro in Human Liver Cell Line” [2], published on the 25th of February 2022, has been frequently cited since its release, as a confirmation for the reason for vaccine hesitancy, highlighting a potential misconception that the mRNA vaccine alters the human DNA. Furthermore, vaccine skepticism is often taken online on social networking sites. Therefore, our study aims to identify and discuss Swedish mRNA-related social media posts, that are being published during the period after the Lund study. We use structural topic modeling (STM) to explore Swedish tweets and Swedish discussion posts from a popular social media platform (Flashback) about mRNA-related vaccination. Our aim is to give answer to two major questions: what patterns emerge in Swedish social media as a response to the Lund study and how mRNA related discussion topics change over time, to better understand the public perceptions, responses and concerns that arise after the Lund research. We start by applying natural language processing methods to pre-process and classify the mRNA-related narratives and then structural topic modeling to uncover the most prevalent discussion topics and their evolution over time in the Swedish context. STM automatically detects latent topics in the dataset which can be used to investigate the nature of these topics reflected in the mRNA discussions [3] by utilizing an exploratory mixed quantitative-qualitative approach using data collected from Swedish social media. As a methodological approach, STM enable us to identify prevalent topics in the data, followed by a qualitative analysis on the most representative words and posts of each topic which provided us with better and more targeted insights into the pros and cons of public perceptions and concerns about the mRNA vaccine.
Our results could be useful to public health experts and pro-vaccine organizations to formulate even more effective policies and strategies to reduce anti-vaccine reactions and boost vaccine acceptance.
2. Vaccine Hesitancy and Skepticism
According to the World Health Organization [4] vaccine hesitancy, “the reluctance or refusal to vaccinate despite the availability of vaccines”, was one of the top ten threats to global health even before the COVID-19 pandemic. Although vaccinations are considered as one of the most significant interventions to public health, vaccine hesitancy and resistance creates serious concerns for a significant portion of the population in many countries, including Sweden. Skepticism about vaccine effectiveness, adverse effects, personal beliefs and conspiratory claims as well as exposure to misinformation, plays an important role in decreasing rates of vaccination [5,6]. Vaccine hesitancy discussions are often taken online and for an increasing number of people, the use of such platforms has become a major source for information related to health protection and vaccinations [7]. The availability of massive digital content in e.g., Twitter or Reddit enables researchers to rapidly analyze and monitor large amounts of data, to e.g., identify and better understand the vaccine-deniers’ arguments against vaccinations which in turn, can rapidly be spread as rumours to an even wider audience. False and disputed news or misleading information about vaccination keeps emerging and flowing between people in social media [8]. There is a sense of freedom of self-expression in the use of language, indicated by e.g., the magnitude of (negative) ways to refer to vaccine; here are examples from the Swedish data: fejkvaccin ‘fake vaccine’, bluffvaccin ‘hoax vaccine’, fuskvaccin ‘fraud vaccine’, förtryckarvaccin ‘oppressor vaccine’ or försökskaninvaccin ‘guinea pig vaccine’.
3. Data
Swedish tweets were downloaded from Feb., 10, 2022 (two weeks before the Lund study was published) to Nov., 10, 2022. The tweets were collected with the keywords m-?RNA.* (‘?’ the preceding character is optional.; ‘.*’: ≥ 0 characters) or the hashtag #mRNA and lang:sv (Swedish content). The final tweet data set consisted of 1,700 unique tweets from 730 different users. Apart from the previous, we also collected ca 7,600 unique posts from the popular Swedish forum Flashback (
https://www.flashback.org/), from 18 different discussion threads, all related to COVID-19 and mRNA vaccination.[..]
4. Structural Topic Modeling[..]
5. Results and Discussion
To summarize and better understand the public responses and concerns that arose after the Lund study, we took a closer look at the results of the STM, in which several general themes were revealed based on the topics identified. With respect to our first research question, that is what patterns emerge in Swedish social media as a response to the Lund study, the most prevalent theme during the first couple of months after its publication, was, as expected, direct related to the Lund article:
x Oroväckande resultaten från svenska studien: Pfizer-vaccin tar sig in i leverceller - och omvandlas till DNA ‘Alarming results from the Swedish study: Pfizer vaccine enters livercells - and is converted into DNA’.
A major concern was also the future unknown effects of the mRNA vaccine:
x Risken är stor att barnen får svåra skador som kan påverka framtida generationer. ‘The risk is high that children will suffer serious injuries that could affect future generations.’. Related to the previous is the evidence of the conformity with people’s willingness to get
vaccinated, but not with the mRNA vaccine, but rather with a “conventional” one:
x Jag är absolut ingen vaccinmotståndare, så länge de är traditionellt tillverkade på ett avdödat virus. Men dessa mRNA vacciner är jag väldigt skeptisk till. Tar dom helt enkelt inte! ‘I am absolutely not against vaccines, as long as they are conventional made from a dead virus. But I am very skeptical about these mRNA vaccines. Just don't take them!’.
Concerns were raised for possible injuries and/or side effects, e.g., on the male genitalia; on the female ovaries and for the traces of mRNA vaccine in breast milk – these were major themes for the period of August-November 2022, due to new research studies:
x Biverkningar av mRNA-vaccin inkluderar allvarliga skador på penis ‘Side effects of mRNA vaccines include serious damage to the penis’.
Finally, myocarditis risks were prominent during almost the whole examined period:
x Läkare varnar för att mRNA vaccin orsakar myokardit men Twitter stämplar inlägget som falsk information. ‘Doctors warn that mRNA vaccine causes myocarditis but Twitter labels the post as false information.’.[..]
With respect on how discussion topics change over time some topics showed some clear characteristics on the topic prevalence (cf. Fig 1b). The time covariate provides a means for a direct comparison and explanatory power over time which makes it better understand people’s concerns during a time span. Retrospectively, the variation of topics across time, can also be helpful to detect significant events related to e.g., mRNA discussions in social media data over time, as in our case, e.g., through an examination of prominent words within each topic. For instance, figure 1b shows how the prevalence of topic 6 raises towards the end of the examined period, while topic 7 rapidly declines during the middle of the examined period; this is most probably due to the fact that topic 6 relates to the discussion of the potential side effects and risks (e.g., traces of mRNA in breast milk); while topic 7 directly relates or refers to the University of Lund study.
6. Conclusions
This study identifies dominant topics about mRNA vaccine–related issues discussed on Swedish social media. It also examines the changes in these topics over time to better understand the larger trends. Among the selected topics, certain themes, e.g., on myocarditis, remained constant over time. As vaccine development progressed however, other topics became more dominant, driven by e.g., scientific studies introduced to the public. A limitation of the presented work is the search itself which only used a non- exhaustive list of keywords; several relevant posts are probably not included. Consequently, we could expect discussions to be different when using a different keyword set. As a future task, we plan to integrate sentiment analysis. Subsequently, we could use polarity as a covariate to also capture how the sentiment of the mRNA-related events evolve over time. Although it is not the focus of this study, a closer examination of the users might also provide some meaningful information on whether certain users are more likely to post or comment on certain topics (cf. [12]).
