Media Data and Vaccine Hesitancy: Scoping Review

"Media studies are important for vaccine hesitancy research, as they analyze how the media shapes risk perceptions and vaccine uptake."

The proliferation of academic studies analysing social media to better understand vaccine hesitancy is due in part to the growing number of media platforms but also to parallel advances in computing and analysis tools that process and handle big data. This review aimed to identify the media platforms and methods used to study vaccine hesitancy and to explore how they contribute to the study of the media's influence on vaccine uptake and public health.

This study followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines. A search was conducted on PubMed and Scopus for any studies that used media data (social media or traditional media, defined as any media before the advent of digital media), had an outcome related to vaccine sentiment (opinion, uptake, hesitancy, acceptance, or stance), were written in English, and were published after 2010. Studies that used surveys or cross-sectional data were excluded.

In total, 125 studies were included, of which 71 (56.8%) used traditional research methods and 54 (43.2%) used computational methods. Of the traditional methods, most used content analysis (43/71, 61%) and sentiment analysis (21/71, 30%) to analyse the texts. The most common platforms were newspapers, print media, and web-based news. A minority explicitly stated a theoretical framework that drives the analysis. For example, Ward and Budarick used a discursive legitimisation strategy and ideological square theories to evaluate the use of anecdote and emotionality by The Daily Telegraph to push provaccine messaging in a campaign to increase vaccination. And a study focusing on discourse used repertoire analysis to understand how parents' repertoires in distrust contribute to a delegitimisation of systems propping up medical services, research, and government authorities. Another study on repertoire echoes those using framing theories to understand how positive or negative framing could coerce behaviour.

All the studies that computational methods (54/54, 100%) were published in or after 2016. Twitter was the most popular platform, with 57% (31/54) of studies using it. The computational methods mostly used sentiment analysis (31/54, 57%), topic modeling (18/54, 33%), and network analysis (17/54, 31%). Fewer studies used projections (2/54, 4%), feature extraction (1/54, 2%), and image analysis (1/54, 2%). Computational studies share a unifying theme with traditional method studies, which is the deficiency of the theoretical focus driving these studies. Of the 6 studies that did (6/54, 11%), only 1 focused on a health behaviour model, and the others used more generalised theories and marketing concepts.

The different platforms used in the included studies span blogging sites, microblogging sites, newspapers, image-based social media platforms (Instagram), video-based social media platforms (YouTube), search engines, and question-answering sites. The growth of live streaming on platforms such as Instagram reels, TikTok, Bilibili (Chinese video streaming platform), and Twitch is likely to pivot analysis methods in the direction of computer vision, and preferences for more advanced methods may follow suit.

The following five major categories of studies arose:

1. A set of studies (39/125, 31.2%) focused on what antivaccination topics arose - e.g., the distrust of institutions, civil liberties, misinformation, conspiracy theories, and vaccine-specific concerns.
2. Studies that focused on provaccination topics that emerge (8/125, 6.4%) centred on ensuring vaccine safety using scientific literature and how having an empathetic connection may lead to perception of vaccines in a more positive light.
3. Framing involves understanding how the media encodes messages through signs and symbols. Who delivers the message, their background, and how they say it are all important in vaccine hesitancy research. One finding from the studies is that personal stories, which are shown to be more engaging, are a tool used by both sides to enforce their viewpoints as correct (such as the use of anecdotes on antivaccination websites or the use of personal stories to encourage positive vaccination dialogue). Another tactic used for framing, especially from the antivaccination side, is the use of shocking images or appeals to emotion through testimony to convince others of the antivaccination agenda. Often, these antivaccination messages misuse scientific evidence and use loss-framed messaging to transmit their ideas. These tactics may allude to a more generalised use of risk-amplifying messages to elicit reactions.
4. With regard to the relative amount of coverage, activity, or engagement in the web of provaccination and antivaccination communities, most studies in this theme (12/125, 9.6%) found that any negative or antivaccination coverage or messages were generally more prevalent and engaged with (shared, viewed, retweeted, and liked). Some studies (8/125, 6.4%) went further to establish an association between coverage - both the type and amount of coverage - and vaccine uptake. One of the studies found that a higher number of tweets, Facebook posts, and internet searches in an area were associated with lower measles, mumps, and rubella vaccine coverage. Another study found that more tailored messages to specific communities would lead to higher proactiveness in certain parts of the population to get vaccinated. One set of studies (9/125, 7.2%) looked at how vaccine-discussing communities engaged with each other. An example of this is the finding that antivaccination groups: discussed vaccination issues much earlier; are deeply fragmented in their beliefs, which spiral into radical communities; and are part of a larger robust network of vaccine-hesitant individuals.
5. Some studies focused on the public reacting to and focusing on certain signals - in particular cases, deaths and scandals - which suggests a more volatile period for the spread of information. For example, Diaz et al. found that there was increased search activity regarding vaccines and infertility following the United States (US) Centers for Disease Control and Prevention (CDC) emergency approval of COVID-19 vaccines. Interactions on Twitter also increased in response to political events. One study found that adverse event reporting produces a more emotional response that leads to a decline in positive sentiments about vaccines. Another study looked at how the public had episodic expressions of distrust toward the Chinese government immediately after a vaccine-related scandal.

In reflecting on the findings, the researcher notes that, "Despite fervor on the growth of this field, a glaring shortcoming misroutes it - a lack of theoretical foundation. Missing a theoretical focus portends the use of methods only for the sake of novelty and not necessarily informativeness....Although this contributes to an overall body of knowledge in vaccine hesitancy research, it disorganizes the trajectory of the field as findings are not built on the cornerstones already set by theories in health behavior, vaccine hesitancy, and public health. Thus, it makes it difficult to draw any conclusive findings on the media's real influence on vaccine hesitancy as measured variables and outcomes differ. Using a theory-driven approach can counter this trend, making the consolidation of findings more cogent."

Areas for further exploration include:

• Understand how trust and distrust toward institutions (government and health care) may influence vaccination. Although media data can aid in the identification and classification of topics and understanding how they spread in networks, more work is needed to clarify the association between trust and adherence to public health measures.
• Understand how misinformation spreads. This field of work will likely involve health psychologists, computer scientists, public health experts, and media researchers, as it involves understanding: how information signals are generated, spread, and processed; what signals are important in shaping risk perception; and how the timing of this matters.
• Explore effective communication of governments and the pharmaceutical industry in addressing vaccine concerns. In this work, "Evaluations of governments on vaccine communication should be performed and benchmarked against WHO [World Health Organization]-prescribed standards such as those laid out in the COVID-19 Vaccine Safety Manual...or the Managing Vaccine-Related Events guide..., with the aim of identifying successful case studies on vaccine communications."

In conclusion: "Researchers inclined to join this field [digital epidemiology or infodemiology] should fully understand that media data analysis methods are meant to supplement - not supplant - current practices in public health research. A way to ensure this understanding is to establish a theoretical focus of the research before method or platform selection. In doing so, the mentality of adopting trending methods is avoided, there is a systematic consolidation in the synthesis of findings, and a coherent paradigm in the subfield of media data research on vaccine hesitancy can be established."