Browse the corpus

Standard behavioural approaches to encourage vaccination include mandatory vaccination and regulation for healthcare professionals, incentives, public health communication campaigns, and engaging trusted leaders.31 Contemporary methods have started to be implemented on social media,10 31 32 33 including debunking (fact checking specific claims after they have reached social media users) and “pre-bunking,” a behavioural approach in which users are taught about how “fake news” works before exposure.34 Other intervention types include warning (“inoculating”) people about manipulation tactics using non-harmful exposure as a tool to identify misinformation, and using accuracy prompts to trigger people to consider the truthfulness of material they are about to share on social media platforms, without stopping them from posting.35 Recognising the intense effect of social media on vaccine hesitancy,36 the Africa Centres for Disease Control and Prevention developed a toolkit to assist countries in social media strategies aimed at encouraging vaccination.37

ial they are about to share on social media platforms, without stopping them from posting.35 Recognising the intense effect of social media on vaccine hesitancy,36 the Africa Centres for Disease Control and Prevention developed a toolkit to assist countries in social media strategies aimed at encouraging vaccination.37 Such behavioural approaches to misinformation on social media have shown promise in reducing the sharing of disinformation and misinformation (box 1) and in changing people’s beliefs,38 but less clear is their effect on vaccination uptake. Investigating the success of such interventions on uptake is essential because there is an established link between social media exposure and offline beliefs that vaccines are unsafe.39 It is clear, however, that providing fact based probabilistic information alone fails to meaningfully increase uptake40 and might even backfire.41 Factors such as low trust in governments and health institutions are likely to be instrumental in derailing effective immunisation programmmes.42 Multiple drivers and barriers to vaccine uptake must therefore be considered when developing effective tools.

We identified 30 studies evaluating interventions to tackle misinformation on social media that explicitly captured real world behavioural outcomes (see supplementary information on bmj.com). After excluding 19 studies for lack of external validation, only 11 published interventions were left (table 1; box 2; fig 1). There is clearly insufficient evidence from field studies on social media interventions. Only two interventions that met all criteria came from outside the US (Israel43 and Nigeria53), with a recent study conducted in Nigeria producing perhaps the most robust level of evidence. Interventions to reduce vaccine hesitancy through social media that assessed real world vaccine behaviours

We identified 30 studies evaluating interventions to tackle misinformation on social media that explicitly captured real world behavioural outcomes (see supplementary information on bmj.com). After excluding 19 studies for lack of external validation, only 11 published interventions were left (table 1; box 2; fig 1). There is clearly insufficient evidence from field studies on social media interventions. Only two interventions that met all criteria came from outside the US (Israel43 and Nigeria53), with a recent study conducted in Nigeria producing perhaps the most robust level of evidence. Interventions to reduce vaccine hesitancy through social media that assessed real world vaccine behaviours Studies on reducing vaccine hesitancy can inform broader vaccination messaging campaigns. We recommend some critical features that should be included for all interventions (fig 1). Details about the type of vaccine and the target group should be highly specific, as should the form of hesitancy and misinformation,54 the misinformation source, and details of the method (especially for RCTs). Real world effects (ideally direct and broader public health indicators) must be captured. These aspects are relevant for better understanding of what might work and for what reasons and for comparison between methods and social media platforms. The most informative approach would cover the first three levels of evidence under the THEARI system (theoretical, empirical, and real world).55 Over time, these would accumulate to the fourth (replicated) and fifth (well established impact) levels, maximising the potential for predictive validity of applications to public policy (for example, greater confidence in anticipated effects). Doing so would exceed the criteria listed in a recent call for a gold standard for trials tackling vaccine hesitancy.56

rth (replicated) and fifth (well established impact) levels, maximising the potential for predictive validity of applications to public policy (for example, greater confidence in anticipated effects). Doing so would exceed the criteria listed in a recent call for a gold standard for trials tackling vaccine hesitancy.56 Critical aspects of studies that are necessary to better understand and compare possible interventions that tackle vaccine hesitancy on social media (described in box 2). The examples of what specifically to measure and observe are non-exhaustive. All aspects should be included for maximum value, though behavioural and population health outcomes should be prioritised. Boxes with thick outlines and darker shading are considered absolutely necessary, though all boxes are strongly recommended. Some evidence exists from field studies that did not directly target misinformation but provided accurate and useful information about vaccines. These were typically posted on interactive websites or directly in social media feeds. Changes in attitudes, knowledge, and social media engagement were reported for several information campaigns using social media advertisements in before and after designs and designs without control groups.44 45 47 48 But these studies showed no changes in vaccination behaviours. Only two studies, both RCTs, showed that tackling vaccination concerns and providing information in interactive formats on a website increased vaccination46 and reduced days undervaccinated.51

and after designs and designs without control groups.44 45 47 48 But these studies showed no changes in vaccination behaviours. Only two studies, both RCTs, showed that tackling vaccination concerns and providing information in interactive formats on a website increased vaccination46 and reduced days undervaccinated.51 Two interventions that increased vaccination uptake used targeted messages through Facebook adverts57 and personalised influencer content.58 But each had limitations. In the first, vaccination rates increased only among families of medium-low socioeconomic status, indicating narrow effectiveness. In the second, the measures were primarily attitudinal, and behavioural outcomes were self-reported. Linking online campaigns to true behaviours in field studies is a major challenge. This is partly because running well controlled studies is simpler in laboratory experiments and surveys, and because capturing distal effects (true behaviours) of web based studies is often impossible. But showing some measures of success in increasing uptake is necessary to justify investment to develop interventions at scale. Therefore, it is critical that researchers of vaccine behaviours and associated campaigns engage directly with clinics and public health agencies to improve the ecological and external validity (that is, real world, observed outcomes) of interventions (box 2).

Few interventions among those we evaluated captured evidence about real world behaviours, so there is currently no gold standard toolkit that public health agencies can refer to.56 But the evidence available can inform future tool development. We provide 10 insights based on the existing evidence that should help provide a clearer and more specific, evidence driven toolkit of approaches to reduce vaccine hesitancy. Negative sentiments on social media might increase vaccine hesitancy faster than interventions reduce it 39 59—Although vaccine hesitancy can stem from many sources, including mass media and political rhetoric as well as genuine safety concerns, there is ample evidence of a proliferation of anti-vaccine messages on social media leading to organised offline actions and increased hesitancy.60 This is driven by several social and individual factors, as well as foreign disinformation campaigns and bots.61 62 There is less evidence that efforts to specifically mitigate misinformation have had a reliable effect on real world vaccine uptake. Messaging seems to work best when it is tailored to what groups know and care about—Two relatively large vaccination interventions, one for flu, one for human papillomavirus (HPV), found modest effects through targeting specific groups. Broad campaigns through all forms of media tend to be generally effective. But once misinformation and conflicting views are prevalent, speaking directly to audiences, knowing the reasons for hesitancy, and framing information in a way that matters to individuals are critical.54 63 64

through targeting specific groups. Broad campaigns through all forms of media tend to be generally effective. But once misinformation and conflicting views are prevalent, speaking directly to audiences, knowing the reasons for hesitancy, and framing information in a way that matters to individuals are critical.54 63 64 Simple messaging about benefits and risk based on probabilities is not enough—Filling vaccine knowledge gaps has direct benefits, such as supporting informed decisions,63 but messaging with information about benefits, harms, and associated probabilities is insufficient to resolve vaccine hesitancy.40 Some possible explanations include lack of trust and cultural values, which have a major effect on how scientific data are interpreted and accepted through social media.65 Messages must be conveyed in a way that affirms individual cultural values,66 dealing with topics of importance to individuals, not only health facts, using credible sources of information.67 Visual imagery also helps deliver effective messages,68 and other components should be considered, such as health literacy, simplified language, and context specific features like age or language.

values,66 dealing with topics of importance to individuals, not only health facts, using credible sources of information.67 Visual imagery also helps deliver effective messages,68 and other components should be considered, such as health literacy, simplified language, and context specific features like age or language. Correct misinformation to both parents and their children—Parents, especially mothers, play a major part in child vaccination.58 One large campaign carried out on Facebook directed at mothers of teens had some positive effects in specific income groups, but the overall effect was minimal to null. Addressing parents is clearly of value, but young people also seek out information online for themselves.69 Thus, the information teens are presented with online likely needs similar but unique safeguards against misinformation to the ones discussed. Directly involving parents and young people in the design of messaging might strengthen the effectiveness of these campaigns.70

lso seek out information online for themselves.69 Thus, the information teens are presented with online likely needs similar but unique safeguards against misinformation to the ones discussed. Directly involving parents and young people in the design of messaging might strengthen the effectiveness of these campaigns.70 Trust matters: the message,71 the messenger,63 and the (vaccinated) provider 57—Substantial amounts of anti-vaccine messaging comes from non-medically qualified, or non-expert, voices on social media.72 Thus, medical professionals are under-represented when it comes to legitimate and accurate information about vaccines on social media,73 despite typically being the most trusted sources (especially among parents74). Trust is potentially the most distinct characteristic of successful vaccination campaigns,50 75 and this includes ones delivered on social media.76 Ultimately, the source of the message—whether a healthcare provider, a politician, or a social media influencer—is likely to have a major role in whether individuals and communities deem information to be credible.57 63 These interventions have great potential when delivered to the right population.30 The US state of New York, for example, was a stronghold of anti-vaccine sentiment before 2020, which pivoted after the arrival of covid-19 vaccines. Despite intensifying hesitancy and misinformation waves, the state outpaced the national average for vaccination rates, benefiting both public health77 and economic returns.78 This success is attributed to strategies that, on top of tackling misinformation, also cultivated trust in the source of the vaccine, the messenger, and the provider, involving use of the military (a highly trusted organisation in the US), diverse community messengers (including community health workers and faith leaders), and a broad network of vaccination locations.79 80 81 82 These aspects must be ensured in public campaigns aimed at specific groups, especially those who are vaccine hesitant.64

f the military (a highly trusted organisation in the US), diverse community messengers (including community health workers and faith leaders), and a broad network of vaccination locations.79 80 81 82 These aspects must be ensured in public campaigns aimed at specific groups, especially those who are vaccine hesitant.64 Debunking efforts have shown mixed effects on social media—Distributing information from public institutions or providing objective information from third parties might help to counter misinformation, reduce the intention to spread misinformation, and promote health behaviours, but the process is not always smooth. The backfire effect is a concerning pattern in which disproving misinformation reinforces it and deepens false beliefs.83 But this phenomenon is not consistently observed in practice. Effects of debunking might be highly dependent on the recipient’s background knowledge and beliefs, methods of presentation, and what other information is viewed.84 85 For now, the immediate priority for this approach is to determine if inoculation based on pre-bunking can reliably reduce the noxious effects of next widespread waves of misinformation and resulting misbeliefs.

pient’s background knowledge and beliefs, methods of presentation, and what other information is viewed.84 85 For now, the immediate priority for this approach is to determine if inoculation based on pre-bunking can reliably reduce the noxious effects of next widespread waves of misinformation and resulting misbeliefs. Raising the quality and visibility of reliable health information can counter misinformation—Providing simple probabilistic information is not likely to completely counteract misinformation on its own, but providing information on how, where, and when to get a vaccine does.76 The high volume of misinformation appearing in online searches, however, can override more reliable sources,86 limiting the effectiveness of high-quality information campaigns.87 To increase campaign effectiveness, interactive designs and visual aids such as posters or videos help target populations see and engage with accurate, accessible information throughout a campaign.51 Making it easier to find such materials is paramount, starting with search engine optimisation to increase visibility of campaigns. Framing of vaccine messages matters—How public campaign messaging is framed affects health decision making. A public campaign can’t cover all vaccines, diseases, populations, and reasons for hesitancy; framing messages to be directly relevant to a populations’ needs (benefits and risks specific to their group) have resulted in significant increases in uptake.88 Positive and negative framing effects are not equivalent,89 however, and this should be factored into messaging decisions.

lations, and reasons for hesitancy; framing messages to be directly relevant to a populations’ needs (benefits and risks specific to their group) have resulted in significant increases in uptake.88 Positive and negative framing effects are not equivalent,89 however, and this should be factored into messaging decisions. Blanket bans can drive groups and activities underground—Broad social media bans of individuals or of specific content can paradoxically result in the spread of misinformation and can galvanise problematic echo chambers by driving discussion into private social media groups or closed forums.90 91 Such closed environments are unlikely to include different viewpoints or corrective information, so misinformation is more likely to be reinforced. Rather than rely on outright bans, policy makers and content managers should explore methods that limit the spread and influence of misinformation.92

or closed forums.90 91 Such closed environments are unlikely to include different viewpoints or corrective information, so misinformation is more likely to be reinforced. Rather than rely on outright bans, policy makers and content managers should explore methods that limit the spread and influence of misinformation.92 Social media platforms need to be part of the solution—If social media platforms are the epicentre of misinformation, then social media companies need to be part of the solution.39 During the covid-19 pandemic, social media platforms took a more interventionist approach to content moderation than before (and, in some cases, removed or limited covid-19 misinformation and conspiracies).93 Some of these approaches are now being rolled back, and social media researchers’ access to data about behaviour on the global platform X (formerly Twitter) is being limited. Content labelling and corrective actions have produced some positive effects,94 but social media companies should be more proactive in dealing with the proliferation of misinformation on their sites.95 We endorse calls39 to make data available and to work with researchers and regulators in all countries to enable developing effective solutions.

Misinformation is not new and its noxious consequences are not insurmountable, but its effect on vaccine hesitancy through social media is an urgent global threat to public health. Increasingly robust evidence has shown the drivers and effects of this phenomenon, but few successful, let alone replicated, interventions exist. One important step towards developing more effective interventions is the close monitoring of public perceptions and opinions about vaccination and services. Digital technologies make it possible to analyse large quantities of “social listening data” in real time.96 Such information would complement the evidence we already have from a variety of study types, improving the design of new, robust, and appropriately targeted interventions. Because most published interventions focused on attitudes and intentions rather than on actual vaccination, however, there remains an urgent need for direct partnerships between behavioural researchers with healthcare clinics and public health agencies.77 Meaningfully developing those partnerships promises direct benefits for more reliable scientific insights that would improve the health and well being of entire populations.