The Influence of Social Media and Institutional Trust on Vaccine Hesitancy in France: Examining Direct and Mediating Processes

Vaccine hesitancy (VH) remains an ongoing challenge in French society. This project explored how institutional trust and preference for information via social media (PISM) drive hesitancy. Across a large, nationally represented population, our findings show that PISM and trust are strongly correlated measures, with both independently predicting VH. Subsequent mediation tests show that social media operates as primarily an indirect contributor to VH through trust. Additional tests involving VH and non-VH typologies revealed that institutional trust consistently predicts greater general support for vaccines and reduced distrust in vaccination. Conversely, PISM directly drives vaccine distrust, with its impact on non-hesitancy fully mediated by institutional trust. Overall, these findings point to the relevance for researchers and public health deciders to address the nature by which people utilize social media information resources and how that interacts with levels of trust for national institutions.


Introduction
In France, the recent emergence of vaccine-related controversies [1,2] has contributed to the long-term persistence of distrust in institutions. Lack of institutional trust, coupled with increased access to misinformation via non-traditional information sources, can further diminish support for vaccination campaigns [2][3][4]. Although substantial attention has focused on social media as a problematic vaccine information source, recent studies also suggest that concerns over the negative effects of social media misinformation are overstated [5]. The current investigation seeks to explore the interrelationship between institutional trust and social media use drive vaccine hesitancy. Drawing from competing theoretical and empirical arguments, we assess how these factors operate sequentially in broader processes leading to vaccine hesitancy. Across four waves of data collected from nationally representative French samples, this exploratory investigation examines pathways linking institutional trust and preference for information via social media (PISM) to vaccine hesitancy (VH) in France. Within this analysis, we examine distinctions among different subgroups/typologies-vaccine hesitant vs. non-vaccine hesitant individuals-to address the importance social media information preference and institutional trust to vaccine judgments. Overall, this investigation seeks to offer a broader understanding of interrelated role between attitudes towards national institutions and alternative media sources on vaccine decision making. In addition, we address the broader significance of social media as a potential driver of vaccine resistance.

Vaccine Hesitancy in France
Conceptually, vaccine hesitancy (VH) can reflect a broad reluctance towards vaccination in general or reticence targeted at some vaccines in particular [6]. Although scholars Vaccines 2023, 11 have called for clarifying this terminology [7,8], a definition commonly utilized refers to vaccine hesitancy as a "delay in acceptance or refusal of vaccines despite the availability of services" [9] (p. 4163). Prior to the COVID-19 outbreak, France was one of the most vaccine hesitant countries in the world (in terms of negative perceptions) [10][11][12]. Rates of VH have ranged from 25-70% depending on measurement and timing of surveys [13]. Although the current study does not focus on COVID-19 vaccination, it is worth acknowledging that at the onset of COVID-19 vaccine distribution (December 2020), France reported the highest level of COVID-19-VH among European nations, with less than 50% of those surveyed intending to get the COVID-19 vaccine [14]. Although more than 90% of French adults have now received a COVID-19 vaccine [15], large sectors of the French population continue to doubt the safety and efficacy of this vaccine for children and adults [16].
Multiple factors contribute to high VH in France, including recommended vaccines being perceived as less important than mandatory ones [17], vaccine-related controversies in France [13], and trust in vaccines and/or institutions [18][19][20]. Furthermore, additional factors linked to VH in other countries-high success rate of previous vaccine campaigns leading to trivilializing virus fears [9] and the spread of misinformation via digital media [21]-may also contribute to French residents being vaccine hesitant. Our investigation centers on the interrelationships between lack of institutional trust-a factor with historical links to VH, and social media-an increasingly utilized information source frequently associated with VH in the literature [6,22]. The current study looks at VH tied to four common vaccines-measles, HPV, hepatitis B, and influenza. Measles and hepatitis B recently (2018) became mandatory in France [1]. HPV is recommended for girls aged 11-21 [23], whereas the seasonal influenza vaccine is recommended for a variety of at-risk populations, including pregnant women, those 65 and older, and obese persons [24].

Institutional Trust
Institutional trust is a reflection of healthy, functioning societies [25,26]. Given that public trust impacts the nature by which residents follow informational resources and utilize services, instilling trustworthiness is critical to both preventing and managing epidemics [27]. Van Bavel, Baicker, and Baggio (2021) argue that strategies needed to contain an epidemic are often difficult to enforce [28]. Consequently, institutional trust becomes increasingly more relevant during large-scale public health crises. Prior research shows that while low levels of government trust contribute to less precautionary behavior during virus outbreaks [29], higher levels of institutional trust predicts increased prevention behavior [30].
Recent work has shown that institutional trust directly predicts more negative vaccine attitudes [31]. A systematic review of 25 European countries-including France-showed that trust in health authorities was a central determinant of HPV vaccine hesitancy [32]. The recent history of vaccine-related controversies in France dating back to the late 1990s [13] highlights the struggle to instill public trust. These controversies relate to doubts regarding the safety of adjuvants, the hepatis B vaccine, the HPV vaccine, and the H1N1 vaccine. In particular, during the H1N1 pandemic, the general public pointed to concerns over the marketing of the vaccine and the evaluation of potential side effects [17,33]. Overall, these vaccine-related controversies thrive on a pre-existing lack of institutional trust [34]. Wilson et al. found that French general practitioners (GPs) have themselves expressed distrust in health authorities [34]. This is particularly troubling given the central role GPs play in vaccine implementation/public motivation to vaccinate [35,36].

Social Media
In contrast to traditional media sources, social media is a relatively inexpensive source to share information. Thought-influencers, who potentially lack topic-specific credentials/expertise, can draw large numbers of followers [37,38], with their perspective being perceived as credible as that of the medical establishment [39]. The sheer size of social media audiences can lead to greater exposure to conspiracy beliefs and other misleading information [3,4]. In support of this argument, research by The Centre for Countering Digital Hate (CCDH) showed that within a one-year period from 2019 through 2020, antivaccine social media accounts increased by nearly eight million [40]. In addition, this research showed that those promoting anti-vaccine themes attract substantial followers on both Facebook (roughly 31 million people) and YouTube (17 million). Furthermore, investigations of YouTube and Facebook highlight how vaccine-critical messaging is likely to increase interest/engagement [41][42][43]. Within these platforms, social bots (automated accounts that mimic human beings) operating on social media sites have the capacity to magnify the spread of false information [44][45][46]. The prevalence of these automated accounts-with recent data showing that roughly 9-15% of Twitter accounts are actually social bots [47]-suggests the potential to mislead large numbers of users. Recent analyses assessing French social media environments specifically further support these concerns. Garguilo et al.'s analysis of French-language tweets showed that vaccine critics tend to focus on vaccines perceived to be more controversial [48]. In contrast to pro-vaccine accounts, French vaccine-critical Twitter accounts employ a wider variety of tools to attract interest, including more sources and greater coordination with other hashtags. In an analysis of messaging related to vaccines for pregnant women on social media, Martin et al. found that among sixteen countries, France had the largest percentage of discouraging posts [49].
However, while these studies highlight the potential negative impact of social media accounts on the spread of misinformation and potential reinforcement of false beliefs, there are also data indicating less consequential social media effects. For example, Guess et al. found that visits to fake news websites made up a small share of U.S. web users' Internet diets [5]. These websites were accessed largely by a small subset of U.S. citizens with goals of seeking attitude-confirming content. The researchers argue that effects of fake news exposure may be constrained to increasing support for false claims and consequently that the assumption that fake news has larger effects may be overstated. Several recent studies and commentaries by experts in the field of misinformation studies have also sparked the debate over a possible over-hyping of the effect of fake news on social media [50]. In France, several studies suggest that not only do people have little recourse to social media as a source of information [51], they also tend to be very distrustful of the information circulating on the Internet [52]. Finally, a recent study covering the first year and a half of the COVID-19 epidemic found substantial divergence between the dynamics of the circulation of vaccine-critical discourses on the French Twitter-space and the evolution of intentions to vaccinate against this disease in the French adult population [53].

Exploring Relationships between Institutional Trust and Social Media
As noted above, social media frequently serves as an outlet for misinformation pertaining to vaccination. Less credible yet widely viewed/shared sources commonly express misleading and/or conspirational views across social media platforms [42,54]. Consequently, although social media content may offer useful insights/information on health-related topics, a greater preference for information exclusively provided via this platform (as opposed to traditional information sources) can be associated with lower levels of institutional trust. Overall, we expect a negative relationship to exist between preference for information via social media (PISM) and institutional trust. In particular, those with greater distrust in institutions are likely to be skeptical of traditional information sources that draw from institutional/mainstream actors. Social media is likely to be an appealing resource as it provides discourses grounded in distrust.
More consequently, we address how the relationship between PISM and institutional trust drives attitudes towards vaccines. The findings above highlight the implications of diminished institutional trust and increased social media information use on VH sentiment in France. While there exists rationale for bi-directional associations between PISM and trust, prior research offers insights into temporal processes driving VH. One argument, drawn from a social cognitive perspective [55,56], suggests that institutional trust operates as an intervening factor between PISM and VH. Researchers posit that institutional trust derives from previous knowledge of "the trustworthiness of the concerned institutions" [56] (p. 4). Knowledge obtained via media sources-including conspiracies tied to institutions-may precede perceptions of trust [56,57].
Empirical research lends support for these arguments. Findings show that institutional trust emerges from information provided by media sources on government activities [58]. Further analyses show that misinformation drives distrust in health professionals, ultimately leading to delays in seeking preventative care [59,60]. Additionally, greater institutional trust directly predicts favorable vaccine attitudes [56], while mistrust in government sources directly predicts greater VH [61]. Drawing from these findings, PISM may indirectly predict VH via institutional trust.
A competing argument explores PISM as a direct antecedent to VH. Theoretical arguments from information-seeking and conspiracy literature highlight how individuals may utilize alternative information sources (such as social media) when faced with uncertainty and ambiguity [62,63]. Ultimately, information seeking via social media may offer clarity when individuals perceive that they cannot trust those in positions of authority. Additional theoretical and empirical research identifies lack of institutional trust as an antecedent to increased risk perceptions [64]. By increasing risk perceptions, individuals are increasingly motivated to seek out information.
Importantly, alternative media channels may be particularly appealing when people distrust institutions. Given French residents' historical distrust for institutions [20,65,66]-which include traditional media-individuals will turn to alternative information resources (such as social media) for information. Distrust in institutions can drive individuals away from mainstream sources, sources that give too much attention and credit to institutional/mainstream actors. Among the alternative sources of information are the forms of social media where people are very likely to encounter discourses/misinformation grounded in skepticism/distrust. Consequently, information obtained via social media could intervene in the relationship between institutional trust and VH. Recent data have linked anti-vaccine beliefs directly to social media use [67]. Overall, a competing mediation model examines how institutional trust indirectly predicts VH through PISM.

Participants
We conducted four cross-sectional online surveys among representative samples of the French mainland population aged 18 and over, with participants randomly selected from an online research panel of more than 750,000 individuals. Overall, a total of 9177 individuals completed the survey spread across four waves: July 2021 (n = 3087), October 2021 (n = 2015), December 2021 (n = 2022), and May 2022 (n = 2053). The researchers employed quota sampling methods to match French census statistics according to age, sex, profession, region of residence, and size of the town. Participants receive points for completing these surveys. Ultimately, participants could exchange points in return for gift cards. Demographic information collected included age, gender, educational level, and income.

Vaccine Hesitancy
Vaccine hesitancy was calculated through five items measured on a four-point Likert scale (1 = strongly favorable, 4 = strongly unfavorable), with participants also having the option to respond with "I don't have an opinion". The vaccine hesitancy measure included one item asking about general favorability towards vaccines and four items addressing favorability towards four specific vaccines: HPV, measles, influenza, and hepatitis B. The five items were summed together, then averaged to create a vaccine hesitancy scale (Cronbach's alpha = 0.84). The "no opinion" responses were inputted to the mean value of each item.

Institutional Trust
The researchers assessed institutional trust through five items all beginning with the following stem: "how much trust do you have in". Participants answered the questions on a 1-4 scale ranging from 1 (no trust at all) to 4 (strong trust). The different institutions included the media, parliament, science, government agencies that control or regulate health and environmental risk, and the government. An institutional trust measure was constructed in similar fashion as the vaccine hesitancy measure (Cronbach's alpha = 0.81).

Preference for Information via Social Media
Finally, we assessed preference for information by asking respondents to indicate their three preferred sources of information out of "television", "radio", "print media", "newspaper websites", "other websites", and "online social networks" (Facebook, Twitter, etc.). To develop the "preference for information via social media" (PISM) measure, we created a dichotomous "yes/no" measure based on whether participants had listed online social networks as one of their three preferred sources of information.

Statistical Analyses
The "R" programming package was employed for all analyses, including bivariate correlations between the central study variables and subsequent multiple regression tests. Mediation tests were performed through the "multilevel" package in R, which combines assumptions from the Sobel test [68] with bootstrapping techniques [69].

Preliminary Analyses: Who Is Vaccine Hesitant and Who Uses Social Media as a Source of Information?
We found great variations in adherence to the four vaccinations (see Table 1). While 77.1% of respondents were favorable to vaccines in general, only 61.4% were favorable to HPV vaccination, whereas 84.6% were favorable to the measles vaccine. We also found differences in the type of hesitancy expressed by respondents (clear negative judgment vs. doubt).

Vaccine Typologies
As an additional tool to explore descriptive and inferential analyses involving VH, we collapsed this measure into three categories: "non-hesitant", "hesitant but pro-vaccine", and "distrustful of vaccination" categories. "Non-hesitant" individuals represented those favorable to vaccines in general and to all four vaccines listed (measles, hepatitis B, influenza, and HPV; 37.9% of sample; n = 3477). "Hesitant but pro-vaccine" respondents reflected those favorable of vaccines in general but unfavorable or uncertain about at least one of the four vaccines listed (38.8% of the sample; n = 3588). "Distrustful of vaccination" individuals include those respondents not favorable to vaccines in general or who were uncertain about them (22.9% of sample, n = 2102). Notably, only 6.8% of respondents were unfavorable or hesitant to all vaccines. Table 2 provides a descriptive breakdown of social media information preference across different demographic and central study variables (institutional trust, vaccine hesi-Vaccines 2023, 11, 1319 6 of 16 tancy). Among all participants, 23.1% (n = 2120) noted social media as being one of three primary sources for information. Recourse to social media was strongest among those with lower income, those with less institutional trust, and younger participants. Consequently, these social groups were over-represented among the social media users: 61.2% were among the two lower quartiles for institutional trust, 45.6% earned 2000 euros a month or less (vs. 34.8%), and 48.4% were aged less than 35 (vs. 24.9% of the sample). Descriptive statistics provide a first assessment of the relationship between social media use and vaccine hesitancy. Social media users are more likely to be among those with a high vaccine hesitancy score (32.9% are in the highest quartile for vaccine hesitancy) and among the distrustful of vaccines group (34.7% are in this group). However, social media users only compose a minority within each of these groups (31.4% and 35.3% respectively).
Notably, the two variables seem to have a combined effect. Within the first trust quartile, those who use social media seem to be more likely than those who do not to be among the most vaccine-hesitant quartile and among the distrustful of vaccination group (49.9% vs. 39.6% and 50.5% vs. 36.5%, respectively). Also, those who were most likely to be vaccine hesitant were likely to reflect both a combination of preference for information via social media and distrust in institutions. Indeed, among the people in the most vaccine hesitant quartile (Q4) who cite social media as one of their main sources of information, 77.8% were in the two quartiles most distrustful of institutions and only 49.9% were in the most trustful quartile. For the distrustful of vaccination who cite social media as one of their main sources of information, these proportions were 80.1% and 47.2%, respectively.
Overall, these descriptive findings provide frequency comparisons of vaccine hesitancy groupings across PISM and institutional trust categorizations. The subsequent analyses involve formal inferential tests of these relationships.

Central Analyses
We first predicted that a significant negative relationship would exist between PISM and institutional trust. Because the social media measure is dichotomous, for interpretation purposes, we report the multiple regression results predicting institutional trust. After controlling for demographic variables, the regression analyses confirmed a significant, negative relationship between PISM and institutional trust (B = −0.45, p < 0.01, ∆R 2 = 0.01).

Predicting Vaccine Hesitancy
Preliminary correlation tests showed that institutional trust was negatively associated with vaccine hesitancy (r = 0.38, p < 0.01), while PISM positively correlated with vaccine hesitancy (r = 0.13, p < 0.01). Follow-up regression analyses show that both preference for information via social media (B = 0.10, p < 0.05) and institutional trust (B = −0.32, p < 0.01) significantly predicted VH (∆R 2 = 0.12, p < 0.01, see Table 3). PISM was an independent, negative predictor of VH, while institutional trust was an independent, positive predictor of VH. In terms of predictor power, the analyses indicate that social media was a relatively weak contributor to VH, while institutional trust had relatively stronger explanatory power.

Mediation Analyses
The second goal of this study involved testing competing mediating processes involving PISM, institutional trust, and VH. The first test revealed a significant indirect effect of PISM on VH through institutional trust (B = 0.14, p < 0.01; 95% confidence interval: 0.11 to 0.18; proportion mediated is 0.59). Reversing the sequential order of trust and social media use also revealed a significant indirect effect of institutional trust on VH through PISM (B = −0.002, p < 0.05; 95% confidence interval: −0.004 to 0.00; proportion mediated is 0.01). A comparison of these two mediating models highlights the greater support for institutional trust as a mediator between PISM and VH. Figure 1 provides a visual depiction of this relationship.

Mediation Analyses
The second goal of this study involved testing competing mediating processes involving PISM, institutional trust, and VH. The first test revealed a significant indirect effect of PISM on VH through institutional trust (B = 0.14, p < 0.01; 95% confidence interval: 0.11 to 0.18; proportion mediated is 0.59). Reversing the sequential order of trust and social media use also revealed a significant indirect effect of institutional trust on VH through PISM (B = −0.002, p < 0.05; 95% confidence interval: −0.004 to 0.00; proportion mediated is 0.01). A comparison of these two mediating models highlights the greater support for institutional trust as a mediator between PISM and VH. Figure 1 provides a visual depiction of this relationship.

Typology Assessments
As noted above, we collapsed the VH measure into three categories. For clarity purposes, the following analyses examine how central study variables predict "non-hesitant" and "distrustful of vaccination" categorization. Results of the first multinomial logistic regression analysis showed that increased institutional trust predicted greater likelihood of being non-hesitant towards vaccination (OR = 1.41, p < 0.01), whereas PISM was not associated with likelihood of being non-hesitant (OR = 0.94 p > 0.05; model χ 2 = 0.08; see Table 4).

Typology Assessments
As noted above, we collapsed the VH measure into three categories. For clarity purposes, the following analyses examine how central study variables predict "non-hesitant" and "distrustful of vaccination" categorization. Results of the first multinomial logistic regression analysis showed that increased institutional trust predicted greater likelihood of being non-hesitant towards vaccination (OR = 1.41, p < 0.01), whereas PISM was not associated with likelihood of being non-hesitant (OR = 0.94 p > 0.05; model χ 2 = 0.08; see Table 4).
Based on these findings, we ran one mediation test to assess whether social media use indirectly predicted likelihood of being non-hesitant via institutional trust. Findings revealed a significant indirect effect of PISM on likelihood of being non-hesitant through institutional trust (OR = −0.03, p < 0.01; 95% confidence interval: −0.04 to −0.03; proportion mediated is: 0.71; see Figure 2). Note: OR = odds ratio. An odds ratio greater than one indicates respondents were more likely to be categorized as vaccine hesitant/non-hesitant. An odds ratio of less than one indicates that respondents were less likely to be categorized as vaccine hesitant/non-hesitant. PISM = preference for information via social media; * = p < 0.05, ** = p ≤ 0.01.
Based on these findings, we ran one mediation test to assess whether social media use indirectly predicted likelihood of being non-hesitant via institutional trust. Findings revealed a significant indirect effect of PISM on likelihood of being non-hesitant through institutional trust (OR = −0.03, p < 0.01; 95% confidence interval: −0.04 to −0.03; proportion mediated is: 0.71; see Figure 2). Separate analyses explored how the central study variables predicted the likelihood of being distrustful of vaccination. Results of multinomial logistic regression analysis showed that increased institutional trust predicted a lower likelihood of being distrustful of vaccination (OR = 0.67, p < 0.01), whereas PISM predicted an increased likelihood of being distrustful of vaccination (OR = 1.33, p < 0.01; model χ 2 = 0.14). In combination with the multinomial regression analyses reported above, we draw two conclusions. First, institutional trust operates to both increase the likelihood of vaccine non-hesitancy and reduce the likelihood of being in the distrustful of vaccination categorization. Second, and in contrast to institutional trust, PISM operates as an independent positive contributor to the distrustful of vaccination categorization, whereas it does not contribute to non-hesitancy categorization. Finally, although the results indicate these two measures predict VH categorization, the examination of odds ratios (OR) suggests that neither variable are Separate analyses explored how the central study variables predicted the likelihood of being distrustful of vaccination. Results of multinomial logistic regression analysis showed that increased institutional trust predicted a lower likelihood of being distrustful of vaccination (OR = 0.67, p < 0.01), whereas PISM predicted an increased likelihood of being distrustful of vaccination (OR = 1.33, p < 0.01; model χ 2 = 0.14). In combination with the multinomial regression analyses reported above, we draw two conclusions. First, institutional trust operates to both increase the likelihood of vaccine non-hesitancy and reduce the likelihood of being in the distrustful of vaccination categorization. Second, and in contrast to institutional trust, PISM operates as an independent positive contributor to the distrustful of vaccination categorization, whereas it does not contribute to nonhesitancy categorization. Finally, although the results indicate these two measures predict VH categorization, the examination of odds ratios (OR) suggests that neither variable are strong contributors to VH typology. Note: OR = odds ratio. An odds ratio greater than one indicates respondents were more likely to be categorized as vaccine hesitant/non-hesitant. An odds ratio of less than one indicates that respondents were less likely to be categorized as vaccine hesitant/non-hesitant. PISM = preference for information via social media; * = p < 0.05, ** = p ≤ 0.01.
To assess indirect effects of these measures on the distrustful of vaccination categorization, we explored two mediation paths. Results of the first test analysis revealed a significant indirect effect of PISM on the likelihood of being distrustful of vaccination through institutional trust (OR = 0.03, p < 0.01; 95% confidence interval: 0.02 to 0.03; proportion mediated is 0.39). The second indirect effect indicated a weaker yet significant mediation effect of institutional trust on the distrustful of vaccination categorization through PISM (OR = −0.001, p < 0.01; 95% confidence interval: −0.001 to 0.00; proportion mediated is 0.02). Overall, the mediation tests again indicate stronger support for institutional trust as the intervening variable in the relationship between social media information use and distrustful of vaccination categorization. Figure 3 provides a visual depiction of this relationship. mediation effect of institutional trust on the distrustful of vaccination categorization through PISM (OR = −0.001, p < 0.01; 95% confidence interval: −0.001 to 0.00; proportion mediated is 0.02). Overall, the mediation tests again indicate stronger support for institutional trust as the intervening variable in the relationship between social media information use and distrustful of vaccination categorization. Figure 3 provides a visual depiction of this relationship. Figure 3. Institutional trust as a mediator between preference for information via social media (PISM) and distrust in vaccination (categorized). note: ** = p < 0.01.

Post Hoc Analyses
We performed additional analyses to assess whether any interactions emerged between institutional trust and PISM in predicting VH. Results yielded no significant interaction effects.

Main Findings
This study examined how institutional trust and social media information preferences contribute to vaccine hesitancy in France. Across four waves among nationally representative survey populations, our findings indicate that lower levels of institutional trust act as a key contributor to increased VH. In addition, preference for information via social media (PISM) emerged as a significant, albeit weaker, positive predictor of VH. When examining mediation pathways, our results strongly support trust as a direct antecedent to VH, while PISM indirectly contributes to VH via trust. Typology classifications of "vaccine-hesitant" and "non-hesitant" groups offer additional insights on the role of these two predictors. Preliminary descriptive analyses show that respondents who display PISM only represent a minority among the vaccine hesitant and that most vaccine-hesitant respondents who display PISM also display lower than average trust in institutions. Inferential tests show that PISM had no direct relationship to "non-hesitant" classification; rather, this relationship was fully mediated through institutional trust. Conversely, PISM emerged as a strong, direct contributor to being part of the distrustful vaccination group.

Implications
Our study contributes to the current debate over how the circulation of false information on social media affects public attitudes to vaccines and science more generally. Our findings offer some support for arguments that the more direct relationships between social media exposure and VH may be overstated [70] and that too much emphasis might be put on social media use as an explanation for anti-science attitudes [50,71]. We found that few people turn primarily to social media for information, and that even among them, only a minority becomes particularly vaccine hesitant. This can be explained in several complementary ways. In France, as in many other countries, most people-even heavy social media users-are particularly distrustful of the information that circulates on social media [52]. Also, the circulation of information on social media tends to be segmented; meaning that different users will come across different types of information [72]. In the case of vaccination in France, a recent study of the spread of vaccine critical contents on Figure 3. Institutional trust as a mediator between preference for information via social media (PISM) and distrust in vaccination (categorized). note: ** = p < 0.01.

Post Hoc Analyses
We performed additional analyses to assess whether any interactions emerged between institutional trust and PISM in predicting VH. Results yielded no significant interaction effects.

Main Findings
This study examined how institutional trust and social media information preferences contribute to vaccine hesitancy in France. Across four waves among nationally representative survey populations, our findings indicate that lower levels of institutional trust act as a key contributor to increased VH. In addition, preference for information via social media (PISM) emerged as a significant, albeit weaker, positive predictor of VH. When examining mediation pathways, our results strongly support trust as a direct antecedent to VH, while PISM indirectly contributes to VH via trust. Typology classifications of "vaccine-hesitant" and "non-hesitant" groups offer additional insights on the role of these two predictors. Preliminary descriptive analyses show that respondents who display PISM only represent a minority among the vaccine hesitant and that most vaccine-hesitant respondents who display PISM also display lower than average trust in institutions. Inferential tests show that PISM had no direct relationship to "non-hesitant" classification; rather, this relationship was fully mediated through institutional trust. Conversely, PISM emerged as a strong, direct contributor to being part of the distrustful vaccination group.

Implications
Our study contributes to the current debate over how the circulation of false information on social media affects public attitudes to vaccines and science more generally. Our findings offer some support for arguments that the more direct relationships between social media exposure and VH may be overstated [70] and that too much emphasis might be put on social media use as an explanation for anti-science attitudes [50,71]. We found that few people turn primarily to social media for information, and that even among them, only a minority becomes particularly vaccine hesitant. This can be explained in several complementary ways. In France, as in many other countries, most people-even heavy social media users-are particularly distrustful of the information that circulates on social media [52]. Also, the circulation of information on social media tends to be segmented; meaning that different users will come across different types of information [72]. In the case of vaccination in France, a recent study of the spread of vaccine critical contents on Twitter during the COVID-19 epidemic found that vaccine critical activists had a limited ability to reach a wide audience, despite their intense activity [53]. A similar study conducted before the epidemic found similar results, as well as that defenders of vaccines tend to be retweeted by a greater number of users and that mainstream media occupy an important share of discussions [48]. Several recent studies on social media content suggest that the balance between defenders and critics of vaccines might have tipped in favor of the former [48,73,74]. Part of the explanation for the limited association between preference for social media and VH could therefore be a surge in pro-vaccination mobilization on social media and changes made by platforms to limit the spread of misinformation.
Our findings-suggesting that preference for social media offers a limited explanation of vaccine hesitancy-are consistent with research documenting that France is among the most vaccine hesitant western nations while ranking low in social media use [75]. In the case of France, the prevalence of VH would mostly be explained by the combination of a structural distrust of public authorities and the recent emergence of public controversies surrounding vaccination which have made vaccines an object of this distrust [2]. However, the debate over the effects of the recent shift in public preferences for information from non-traditional media sources should not be reduced to the question of whether these effects are exaggerated versus understated. By focusing on those who have a preference for social media as a source of information, our study provides some elements to this debate, albeit in a limited manner. As several qualitative studies conducted in France have shown, even occasional social media use can lead to some parents encountering vaccine-critical information. This in turn triggers information-seeking processes that result in doubt or rejection of one or more vaccines [17,76]. Given how vaccine critics regularly appear in mainstream media-as has been the case in France in the past ten to twenty years [77]-people connect information on social media to information from other sources as well as interpret it against the backdrop of their worldviews and personal experience. Research on the role of social media in VH and public attitudes to science more generally is increasingly shifting attention to these questions pertaining to the way social media use acts as a cog in a set of complex mechanisms combining media practices, beliefs, worldviews, and structural factors [78,79]. Our study contributes to this movement by examining the complex inter-relations between institutional trust, preference for social media as a source of information, and vaccine hesitancy. In particular, one of the crucial findings highlights how these different phenomena operate in unison. The vaccine hesitant who prefer social media also distrust institutions.
Our study also suggests pathways to advance understanding of the relationship between these complex phenomena, such as further investigation of generational differences in social media use [80]. Indeed, we found that the young are more vaccine hesitant and have a much higher PISM. While we did not find a strong generational gap when it comes to distrust in institutions in general, many studies focusing on trust in political actors, politicization, and participation to politics in France have found that people currently aged less than 30-40 are particularly wary of politicians and public institutions [65]. In France, the generation that grew up with the rise of social media also grew up with the economic crisis of 2008, rising inequalities, the rise of the Far Right and the Far Left, the demise of the traditional opposition between Center Right and Center Left parties, and the multiplication of vaccine-related controversies. Investigating the generational differences in social media practices and how they intersect with issues of trust, ideology, public participation, and health trajectories could help better understand exactly how social media use can affect public attitudes to vaccines and science more generally.
In France, vaccine-related controversies over the past 25 years have contributed to this distrust in vaccines. Consequently, institutional trust-impacted itself by faults in vaccine-related campaigns-drives VH. While France is not unique in these relationships, the fact that French respondents frequently report higher VH levels than other comparable nations [10][11][12] highlights the importance of clear, transparent messaging from French government/public health entities.
The concern over how social media use impacts VH reflects a more recent shift in public preferences for information from non-traditional media sources. Our findings suggest that the association between social media and VH can at least partly be explained by how social media contributes to decreased institutional trust. Whereas traditional media sources are inclined to reference and/or support information provided by national institutions, social media may offer an outlet to those dissatisfied with these centralized sources. Consequently, social media may reinforce existing distrust and trigger new doubts in the actions of national institutions.

Typologies
By creating "distrustful of vaccination" and "non-hesitant" categories, we offer a distinct assessment of how institutional trust and PISM contribute to VH identifications. Whereas PISM modestly predicts "degree" of vaccine hesitancy, a clearer reflection of its role on vaccine-related attitudes emerges with the link between PISM and the likelihood of individuals "grouped" as distrustful of vaccination, representing those with concerns over vaccines in general. Individuals in this category do not discriminate across specific vaccines, but rather are likely to express hesitancy to any vaccine. In this regard, social media can potentially have a more generalizable impact on vaccine hesitancy beyond any specific vaccine concerns (e.g., HPV, measles). Conversely, levels of institutional trust increase the likelihood of non-hesitancy categorization (no vaccine-related concerns overall or to any specific vaccine) as well as decreasing distrustful of vaccination designation.
Arguably, the most pronounced representation of the interrelationship between PISM and institutional trust on VH emerges when examining these typologies. PISM alone does not contribute to non-vaccine hesitant categorization, but rather, its impact is fully indirect through institutional trust. Thus, while PISM does not directly reduce non-vaccine categorization, it indirectly operates to decrease this designation by reducing trust. When investigating those who are non-hesitant, researchers should explore how often, and in what selective ways, these individuals utilize information provided by social media sources. Consistent with prior research on the limited spread of social media misinformation [5], these individuals may be more inclined to seek out, consume, and ultimately believe legitimate/credible vaccine-related information. In terms of "vaccine hesitancy" grouping, PISM holds up as a direct factor increasing VH, while also having indirect effects via trust. Among these individuals, identifying the most problematic social media that acts to reinforce blanket concerns over vaccination is central to ongoing campaign efforts.

Limitations
Our analysis presents the usual limitations of survey-based studies. The literature on fake news and mis-/dis-information has underlined the gap between respondents' declarations regarding social media use and the time they actually spend on social media [50,51]. To reflect this, we focused on preference for social media information rather than declaration regarding the frequency of social media consultation. The timing of the study can also bear heavily on responses to survey-based questionnaires. Social media use and preferences for sources of information are in constant evolution, especially in France [52]. This study sheds light on the relationship between preference for social media as sources of information, institutional trust, and vaccine hesitancy at a particular, but crucial moment in time both for vaccines and for social media: the second year of the COVID-19 epidemic. Similar analyses on data pertaining to the period pre-COVID-19 and to future periods are necessary to build a more fine-grained and comprehensive understanding of the dynamic relationship between these factors. Vaccine hesitancy is context dependent, and previous pandemics have seen dramatic shifts in attitudes to vaccines in general [81]. Notably, COVID-19 does not seem to have dramatically changed attitudes to vaccines. In our study, the share of respondents favorable to vaccines in general is modestly higher than studies conducted pre-pandemic [82]. In sum, our results indicate that (a) a majority of French people can be defined as hesitant, (b) a minority reject all vaccines, and (c) most hesitant people do not distrust vaccines in general. These findings are consistent with studies conducted before and during the COVID-19 epidemic [2,20].

Conclusions
Vaccine hesitancy (VH) continues to pose significant public health challenges in French society. This project explored how institutional trust and preference for information via social media (PISM) drive hesitancy. Across a large, nationally represented population, our findings show that PISM and trust are strongly correlated measures, with both independently predicting VH. Subsequent mediation tests show that social media operates as primarily an indirect impact on VH through trust. Additional tests involving VH and non-VH typologies revealed that institutional trust consistently predicts greater general support for vaccines and reduced distrust in vaccination. Conversely, PISM directly drives vaccine distrust, with its impact on non-hesitancy fully mediated by institutional trust. Overall, these findings point to the relevance for researchers and public health deciders to address the nature by which people utilize social media information resources and how that interacts with levels of trust for national institutions. Funding: This study was part of a project labeled as a National Research Priority by the National Orientation Committee for Therapeutic Trials and other researches on COVID-19 (CAPNET). The investigators would like to acknowledge ANRS | Emerging infectious diseases for their scientific support, the French Ministry of Health and Prevention and the French Ministry of Higher Education, Research and Innovation for their funding and support. It has also benefited from funding from the Agence Nationale de la Recherche (ANR-20-COVI-0035-01). The funding sources had no role in the design of the study, analysis of the data or writing of the paper.

Institutional Review Board Statement:
The methodology of the study was reviewed and approved by the ethical committee of the Institut national de la santé et de la recherche médicale (INSERM) (#21-770).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The descriptive data/analyses for this study are available at https://doi.org/10.6084/m9.figshare.23681031. All other data is available upon request.

Conflicts of Interest:
The authors declare no conflict of interest.