Vaccine hesitancy and behavior change theory-based social media interventions: a systematic review

Abstract It is widely acknowledged that vaccine hesitancy is a multifaceted problem that cannot be addressed by a single strategy. Behavior change theories and social media tools may together help to guide the design of interventions aimed at improving vaccination uptake. This systematic review aims to identify the breadth and effectiveness of such theories and tools. The systematic review search was performed in PubMed, Scopus, ACM, Cochrane Library, ProQuest, and Web of Science databases for studies between January 2011 and January 2021 that applied social media tools to increase vaccine confidence or improve vaccination uptake. The literature search yielded a total of 3,065 publications. Twenty articles met the eligibility criteria, 12 of which were theory-based interventions. The result shows that the Health Belief Model was the most frequently deployed theory, and the most common social media tool was educational posts, followed by dialogue-based groups, interactive websites, and personal reminders. Theory-based interventions were generally more measurable and comparable and had more evidence to trigger the positive behavior change. Fifteen studies reported the effectiveness in knowledge gain, intention increase, or behavior change. Educational messages were proved to be effective in increasing knowledge but less helpful in triggering behavior change. Dialogue-based social media intervention performed well in improving people’s intention to vaccinate. Interventions informed by behavior change theory and delivered via social media platforms offer an important opportunity for addressing vaccine hesitancy. This review highlights the need to use a multitheory framework and tailoring social media interventions to the specific circumstances and needs of the target audience in future interventions. The results and insights gained from this review will be of assistance to future studies.


INTRODUCTION
Health threats caused by infectious diseases can result in significant and widespread increases in morbidity and mortality and enormous economic losses, and the mortality impact is greater in low-and middleincome countries. Vaccination is the most effective way to save lives from infectious diseases. Since the middle of the 20th century, vaccination campaigns have helped eradicate smallpox and have countered the fear of other deadly infectious diseases, such as polio, measles, and invasive pneumococcal disease [1]. Achieving adequate vaccine coverage can be threatened by people's concerns about a particular vaccine or vaccination in general. Such concerns are known as vaccine hesitancy.
The WHO listed vaccine hesitancy as one of the top 10 threats to world health in 2019. The SAGE Working Group [2] has defined vaccine hesitancy as "the delay in acceptance or refusal of vaccination despite the availability of vaccination services." Vaccine hesitancy has also been described as "the dynamic and challenging period of indecision around accepting a vaccination" [3]. As it is a complex and context-specific problem that varies across time and place and according to the vaccine and is also influenced by factors such as complacency, convenience, and confidence, vaccine hesitancy is not a clear-cut issue but represents a spectrum of beliefs and concerns, as shown in Fig. 1 [4,5]. This paper defines vaccine hesitancy as: "The behavior and psychological concerns that cause people who are able to access vaccination to avoid doing so, either at all or in a timely manner" [2].
Vaccine hesitancy and behavior change theory-based social media interventions: a systematic review Although vaccine hesitancy is not a new phenomenon, the proliferation of antivaccination misinformation through social media has given it a new significance in the context of public health emergencies such as the coronavirus pandemic and hopes for the rapid development of a vaccine [6]. Misinformation can be a contributor to vaccine hesitancy. Although social media platforms help people communicate freely and share information, the rise of fringe opinions, misinformation, and disinformation also spread rapidly, damage public trust, and exacerbate vaccine hesitancy [6]. In particular, the growth of antivaccination groups has led to a sharp polarization of the debate and more people becoming skeptical of vaccination [7].
As well as being a potential source of the kind of (mis)information that causes hesitancy, social media offers two-way communication with target groups and individuals, increases the availability, accessibility, and appropriateness of the information, and provides a platform for low-cost and highly efficient policy advocacy opportunities [8]. It can provide an unprecedented opportunity to directly address public health issues and thus become a significant health promotion tool that is able to support behavior change [9]. Moreover, in the context of vaccine hesitancy, public health promoters have been encouraged to make greater use of social media, to be active rather than reactive regarding vaccination scares [10][11][12], and use network analysis and social computing methods to understand information spread through social media by antivaxxers and to support the development and implementation of social media interventions [13][14][15].
However, because social media is not a static property but a constantly growing and developing technology that could cause a broad and complicated set of behaviors, interactions, and exchanges involving numerous types of interconnected actors, precise definitions are somewhat elusive [16]. This review describes social media not only just as a medium of communication, but also as an online environment that enables individuals to socialize and a platform that features "interaction" as its main function [8]; we define social media interventions as "interventions that use social media as a tool to deliver intervention content with social media components aimed at influencing vaccine attitude or behavior." Vaccine hesitancy is a highly context-specific problem [17]. The extensive literature on the subject seeks to identify the determinants and barriers in people who are hesitant. Studies have found that these factors included contextual influences (such as social norms, religious beliefs, economic or political factors, trust in government and providers), individual and group influences (personal perception of the vaccine, the opinions of family and peers), and issues directly related to the vaccine (such as vaccine safety and ingredients) [2]. As a result, although many behavioral determinants and barriers to different vaccinations have been identified, few existing interventions have been explicitly designed to address vaccine hesitancy, and even fewer studies have quantified the effectiveness of such interventions in terms of increasing the knowledge and improving the attitude or behavior [3,8,18].
Many studies have shown that interventions informed by behavior change theories based on evidence-based principles are more effective than those that lack a theoretical framework [19][20][21]. In the existing literature pertaining to vaccine attitudes and intentions, behavior change theory is more often used to identify the behavioral determinants than to design and evaluate interventions [22]. Research actions informed by theory rather than by researchers' intuition and experience have the potential to increase understanding of behavioral processes, enable the systematic design of evidencebased interventions, and help put behavioral science into practice. Theory-based interventions offer  May 12). Improving communication about vaccination, adapted from Julie Leask's online blog [5].
an opportunity to better understand the factors that drive vaccine hesitancy and to design interventions to address it. To the best of our knowledge, the literature has not yet been systematically examined to explore how behavior change theories have been implemented in interventions that seek to understand and address vaccine hesitancy.
This review sought to identify social media tools and behavior change theories used in vaccine hesitancy studies and assess their implementation and effectiveness in informing the systematic development of novel interventions for improving vaccination rates.

METHODOLOGY
The PRISMA approach (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Selection [23]) was used to search for articles from research library databases. Fig. 2 shows the selection process aligned with the PRISMA guideline.
The publications were selected using predefined keywords and selection criteria as follows.

Search strategy
We designed an organized structure of key terms to search literature in electronic databases (for search strategy and search keywords, see Appendix A). As each database works differently, the commands were adapted to obey different rules. Relevant literature published between January 1, 2011, and January 1, 2021, were collected by performing searches in PubMed, Scopus, ACM, Cochrane, ProQuest, and the Web of Science database.

Eligible criteria
Inclusion criteria 1

Exclusion criteria
Any article on the following subjects was filtered out: 1. Vaccine development processes for a particular disease; implementation of a general immunization program in a particular area that focuses on the whole vaccine allocation process, including governance, workforce management, vaccination sessions, and vaccine supply. 2. Theory-based interventions without using social media. 3. Interventions aiming to increase the utilization of immunization services or improve the availability of immunization services by healthcare providers. 4. Face-to-face or school-based educational methods without using social media intervention. 5. Behavior theory used for vaccine adjuvants, rabies, and school-based health promotion programs. 6. Commentary paper on social media and vaccine hesitancy without giving the intervention development process.

Data extraction
All articles that accorded with the search strategy were collated, and duplicates were removed. Next, the authors reviewed the remaining titles and abstracts and excluded the articles that did not meet the inclusion criteria. The full text was evaluated if the title or abstract were inadequate for determining eligibility. The final included articles were divided into four categories based on the three types of social media intervention identified in the European Centre for Disease Prevention and Control (ECDC) social media intervention report [8] and the fourth category identified by the authors: • Educational posts: interventions using social media as a communication channel to post or promote educational information to the public. • Dialogue-based: interventions engaging the public in online discussion groups. • Interactive website: interventions utilizing interactive websites with integrated social media components to deliver information and engage users. • Personal reminder: interventions providing personal reminder information using social media channels.
The authors extracted the key points from the included studies, including study details (study area; targeted groups; sample size; vaccine; intervention type; social media type and theory), study aims, intervention details, outcome, and results.
The validity domain presented the lowest score due to the potential methodological bias. Among these, the issue of small sample size was the most frequent issue which has been found in six studies [25,26,28,36,38,44], followed by the lack of generalization in six studies [25,28,33,37,38,40], and response bias caused by self-reported measures in three studies [26,35,39]. Three studies also reported the inability to measure the outcome due to the study design that leads to the difficulty in distinguishing the impact factors [33,35,40]. Besides, the study also reported the potential social desirability bias and in-group bias [31] and the potential weakness on outcomes biased by baseline vaccine hesitancy [38,41]. Despite the potential biases, the studies elucidated that these methodological issues were minimized by using various strategies.

RESULTS
The literature search yielded 3,065 publications from all data sources. Twenty articles met the inclusion criteria. Of these, eight studies used social media to deliver the interventions without using behavior change theory, and twelve studies were informed by behavior change theories. The study characteristics in each group were shown in Appendix C, which displays the study details, including aims, study design, a brief description of the intervention, and results. In this section, we will first present the general description of the included studies, then describe the studies from the perspectives of the social media use and behavior change theory, and present the effectiveness and evaluation of the included studies qualitatively.
The study overview of the included studies is displayed in Fig. 3. Most of the studies were conducted in the United States (n = 11); other in Denmark (n = 2), Germany (n = 2), China (n = 2), Canada (n = 1), Netherland (n = 1), and Italy (n = 1). Twelve interventions addressed HPV vaccination, two on influenza vaccination, one focused on maternal and infant vaccination, and one on Measles-Mumps-Rubella vaccination (MMRs). The remaining four addressed vaccinations in general. In terms of the targeted groups, most of the studies focused on the people aged 12-40, including parents (n = 5), college students (n = 5), adolescents aged 12-18 (n = 3), girls aged 12 (n = 2), and women aged 18-26 (n = 2); the other three studies focused on general online users without giving specific age.

Educational posts
Social media campaigns that post educational messages to the general public were the most common method [25,30,32,33,35,36,38,40,[42][43][44]. People (mostly women) received daily educational messages, including personal stories, medical opinions, facts, and related events, produced by public health agencies through Facebook or YouTube over 1 month to 1 year.
These studies highlighted the effect of different types of contents and ways of posting: Loft et al. investigated the impact of different types of contents on engaging and reaching the audience and concluded that posts containing personal stories performed better than that containing factual information [32]; La Torre et al. found that events were the most popular type of news compared to press releases and publications and Friday is the day when the contents attracted more watches and likes [42]; Robichaud et al. compared the effect of evidencebased and anecdotal video and concluded there is no significant difference in participants' attitude and behavior after watching the videos [36]; Lee and Cho investigated whether different message framing had a various impact on the public's behavior intention to get vaccinated and found participants who received the loss-framed message on Facebook mostly exhibited a higher level of behavioral intention than those in the gain-framing condition [30].
The interventions using educational posts were always combined with other types of interventions. For example, Gerend et al. combined social media posts with posters and yard signs around the campus to promote the HPV vaccine, and it observed a 75% increase in vaccine uptake among the targeted group [40]. Besides, it could also be integrated with other modes of social media. Pedersen et al. carried out a social media campaign to engage Danish girls in communication regarding the HPV vaccine, and the strategy was proved useful [35]. Mohanty et al. used Facebook to reach adolescents for promoting the HPV vaccine while sending reminder letters to the participants [33].

Dialogue-based interventions
Five studies used the dialogue-based social media interventions, three of which were carried out through online forums [26,27,29], and the others were conducted in social media platforms (Facebook (n = 1) [28]; WhatsApp (n = 1) [31]). Participants took part in forums in which they discussed different subjects according to their area of hesitancy. Kimmerle et al. examined the difference of impacts between using factual information versus using personal experience on subsequent communication [27]. The results demonstrated that both kinds of information had a similar effect in engaging users, but the responses to personal experience were significantly more emotional. Hasse et al. attempted to assess the effect of the credibility of the post source on the participant's risk perception toward vaccines, but the result shows the discussions on vaccine adverse events decreased the vaccine intention and increased risk in both neutral and antivaccination online health forums [26].
It can also be combined with other social media modes. For example, Liao et al. promoted the childhood vaccine through discussion groups among parents while sending weekly vaccination reminders, and they identified the most common discussion topics including vaccination decisions, vaccination clinic and cost, vaccine safety and side effects, and vaccine effectiveness, and also found the combined intervention effectively promote mothers' selfefficacy in making vaccine decision [31]. Besides, the dialogue-based interventions can assist the vaccination promotion programs in school settings. Lai et al. designed an intervention for educating students about the HPV vaccination in lectures with two intervention groups in China and concluded the Facebook-assisted discussion method was more effective than the in-person discussion [28].

Interactive websites
Interactive websites provided evidence-based information in an interactive environment through social media. By incorporating social media components such as Instagram, Facebook, Twitter, YouTube, the website could be used to share and interact with other users and receive responses from doctors and other experts. Notably, all three studies were contributed by one research team with a different focus: Glanz et al. developed the interactive social media website that aims to increase childhood immunization by tailoring the content, design, and information architecture toward parents who are seeking to make informed decisions regarding vaccination for their children [37], and they also carried out a series of studies to compare the effectiveness of website without social media, website with social media and usual care on improving parents' vaccine-related attitudes and found pregnant women exposed to the social media were more likely to vaccinate their infants on time than participants receiving usual care [39,41].

Personal reminders
The personal reminders intervention uses a software system to record the data and send the automated reminders through social media to targeted participants. In some cases, personal reminders were always combined with other intervention content and methods to encourage participants to change their behavior [31,33]. Only one study used personal reminders solely to deliver the intervention content in this review, which developed a reminder system by using a mixture of reminder channels-text message, e-mail, phone call, Facebook message, and standard mail to remind the participants to get the HPV vaccine; however, the result shows the intervention did not increase HPV vaccine completion [34].

Behavior change theory
Nine studies reported using behavior change theories to inform the design of social media interventions. The Health Belief Model, Theory of Planned Behavior, and Social Cognitive Theory (SCT) were employed in these studies. Figure 4 presents the use of behavior change theory in each social media mode of delivery.

Health Belief Model
Eight studies used the Health Belief Model (HBM) alone, and the other three combined HBM with the Theory of Planned Behavior. As shown in Fig. 4, six single-theory studies are delivered through educational messages, while two were through dialoguebased social media interventions. The HBM posits four principal drivers of health behaviors: people's knowledge and perception of the severity of disease and their susceptibility to it; people's perception of the barriers to and benefits of being vaccinated; self-efficacy, that is, their belief in their ability to be vaccinated; and the cues to action. The model also assumes that the intention to enact a certain behavior will be followed through [25,45], so it can be used as a framework to evaluate the effectiveness of the intervention on behavior change. Most of the studies that used this model were based on the hypothesis that people with higher perceived susceptibility, higher perceived benefits, and higher self-efficacy can be defined as lower-barrier groups that are more likely to engage in behaviors that promote health [25,46,47].
Of the eight studies, all reported that the HBM was employed as a guiding theory for intervention development: three were based on all components identified by the HBM [33,43,44], four excluded the consideration of cues to action [25,28,30,38], and one study only focused on cues to action [29]. Only four applied the same theoretical framework to evaluate the effectiveness of interventions [25,28,29,44]. In terms of way of incorporation, it was slightly different with different social media modes. For the educational posts, the campaign messages or contents were simply constructed to align with the HBM components [25,38,44], whereas for dialogue-based interventions, the HBM was used as a rationale to design the whole intervention process [28,29]. For example, Allen et al. provided a series of sample tweets for each HBM component, and the messages were sent to the participants to address the targeted barriers [25]; but in Langley et al.'s study, they designed the dialogue-based interventions to prompt external cues to encourage the vaccine behavior [29].

Theory of planned behavior
As mentioned, three studies combined the Theory of Planned Behavior (TPB) with HBM. According to the TPB, the behavior could be directly influenced by intentions, which are determined by attitudes (a function of behavioral beliefs), subjective norms, and perceived behavioral control (a function of control belief) [48]. All the three studies were the intervention delivered through interactive websites by one research team in the same way, that is, the multitheory model was first used in the design of the factual vaccine content to present on the website, and then implemented as indicators in the measurement of effectiveness for assessing the attitude toward perceived benefits and risks of vaccination and self-efficacy regarding vaccine decision-making [37,39,41].
Social cognitive theory SCT considers the six elements in addressing behavior change of individuals, which includes reciprocal determinism (the interaction of individual, environment, and behavior), behavior capability (perceived facilitators and barriers to action), vicarious learning (the source of information cognitive and behavioral development), reinforcements, expectations, and self-efficacy [49]. Compared to the HBM and TPB, SCT focuses more on the social motivation for achieving the behavior change. For example, in Liao et al.'s study, the intervention was developed to promote mothers' self-efficacy through group discussion for taking their children to get the influenza vaccination, and the self-efficacy was served as the main indicator for design and measured effectiveness [31].

Effectiveness and evaluation
This section comprises the effects of the interventions on improvement in knowledge, intention, behavior, or other outcomes. Due to the heterogeneity of the studies and a lack of suitable data for measuring effectiveness, we were unable to compare the effectiveness quantitatively. Instead, we collected the primary outcomes of the included studies to present the effectiveness qualitatively.
As shown in Table 1, the evaluation approach varies depending on the measurements. Theorybased Likert-type scales are the most used to evaluate knowledge, attitude, and intentions. For example, Lai et al. developed a 20-item self-report scale with a five-point Likert-type rating to assess the attitude and intention of HPV vaccination behavior post the intervention [28], which is similar to the methods used in other studies [25,30,36,41,44]. To evaluate the vaccine uptake rate, six studies used the medical records [33,34,37,40,41,43], whereas the other three collected the data through self-reported surveys [31,38,44].
Fifteen studies reported the effectiveness on knowledge gain, intention, or behavior change;     Medical records (Immunization Registry) Patel et al. [34] Personal reminder None Behavior The reminder system did not increase completion rates, which overall were low at 17.2% in the intervention group and 18.9% in the control group (p = .881).
Medical records one study stated in progress without provided any result [29], and other four studies used other measurements instead, such as engagement rate and click-through rate [32,35,42] and the number of replies [27]. Among the fifteen studies, eleven of them were theory-based interventions, and only four nontheory-based studies reported the effects on intention or behavior change [26,34,36,40], two of which are even no significant change [34,36]. One nontheory-based study reported a 75% increase in HPV vaccine uptake in a school design [40]. Among the other six theory-based studies that reported behavior change outcomes, one indicated no statistically significant difference between intervention and control groups, although the higher completion rate was shown in the intervention group (12% vs. 4%) [38]; two also reported no significant increase for vaccination uptake [31,43]; and the other three studies reported positive effects. One reported 60% of participants received the HPV vaccine due to the intervention [44], another reported a significant difference between control groups that participants in the social media intervention group were two times more likely to receive the MMR vaccine than those in usual care and internet intervention groups [41], and the third reported the number of doses and participants administered for the HPV vaccine (176 doses and 152 unique adolescents) after the intervention [33].
In terms of the mode of delivery, four theorybased education posts reported significant results: one found higher knowledge gain between before and after the intervention, but no significant increase in vaccination rate [43]; one reported the unframing message promoting through Facebook significantly increased the vaccine intention [30]; the other two reported higher vaccination uptake rate due to the intervention [33,44]. Another nontheory-based educational post also reported a higher vaccination rate with a 75% increase among the participants after the intervention [40]. The theory-based interactive website had significantly improved the vaccine knowledge [37], attitudes and intention to vaccinate [39], and vaccine uptake [41]. Three dialogue-based interventions were reported to be effective in improving people's intention to be vaccinated [28,31], one of which is nontheory-based intervention [26]. It also performed well in increasing the vaccine knowledge and awareness [28,38], although two studies reported no significant improvement was found in the vaccine uptake [31,38].

DISCUSSION
Here, we highlight the key findings-the effectiveness, strengths, and limitation of using each social media component and behavior change theories in the interventions to address vaccine hesitancy, then summarize the opportunities and challenges that face future attempts to integrate behavior change theory and social media interventions as a means of addressing the problem of vaccine hesitancy, and eventually identify the benefits and limitations of this research.

Social media use
The included articles were divided into four categories according to the main intervention type, although some had multiple components. Each mode has its advantages and disadvantages.
The chief merits of educational posts are the low cost and the potential for wide coverage [40]; an apparent weakness, however, is the use of nontailored information. Furthermore, with no control group among so many participants, the effects of other factors cannot be ruled out. It is difficult, therefore, to measure the effectiveness of this type of intervention [35]. The alternative would be to narrow down the target group and provide tailored messages for specific populations, such as college students or parents, and offer guidance on how and where to access the vaccination [40]. Such tailoring is always more effective, but the cost is greater.
For the dialogue-based interventions, participants could post their opinions and concerns about the vaccine or disease and freely communicate with other participants about their experiences; therefore, it is much easier to evaluate the effectiveness by collecting the data using repeated surveys than the educational posts. The researchers were able to observe the influence of initial posts and follow-up discussions on changes in participants' attitudes [27-29, 31, 38]. Besides, the results show that information from discussion groups can effectively improve vaccine confidence by enhancing the related knowledge and dispelling the doubts, but this alone may not be enough to solve the vaccination concerns and enable people to make positive vaccination decisions [28,31,38].
Similarly, the questionnaires could be sent to participants to assess their trust in the social media tools and the health information received for the interactive websites. Besides, its multicomponent nature facilitates it to be more functional-participants can receive the educational information from the website and discuss their questions and concerns with other users and receive responses from doctors and other experts. However, its development, engagement, and management cost is much higher than the first two. It is worth noting that the study found the contents using storytelling approaches and answers to questions posed by readers on the website show the highest number of reads and engagement [32].
A personal reminder has the advantage of being easy to measure the effectiveness. Apart from using a self-reported survey, personal medical records were also used to evaluate the effectiveness of the interventions; researchers could observe how many participants delayed or refused vaccination for personal or nonmedical reasons. However, it alone presents the lowest effectiveness compare to other intervention types, and consequently, it should be combined with other intervention components [34].
In terms of effectiveness, we found that the strength of different modes lies in different targeted outcomes. The educational post performed well in improving knowledge, while dialogue-based intervention is better in increasing the intention or change the attitude, but may not be enough on their own to trigger behavior change. Thus, the most effective interventions employed multiple strategies. In this review, we found that the most significant increases in vaccine uptake on the first dose (increases of up to 75%) were observed with interventions that used educational messages with posters and signs [40], particularly those that embedded information about to recommend HPV vaccination to others. Such increases could nonetheless have other causes that are difficult to distinguish. In general, we found the interventions that reported the significant increase in vaccine uptake were those that: (1) targeted specific populations; (2) included tailored content; and (3) used behavior change theory in their design.
Summarizing the benefits and limitations of social media interventions is problematic because the studies were carried out in various countries and cultures and are therefore highly context-specific. In general, evidence-based social media interventions have the potential to significantly improve the knowledge, intention, and behavior about vaccines and clear the behavior barriers, especially when compared to other modes such as mass media [30] and the general internet [37,39,41]. Additionally, there are undeniable benefits to using social media-fast access to a wide audience, low cost-so its future is promising, but more evidence-based strategies and studies are needed.
Three main limitations to these studies can be identified. The first is the difficulty in clarifying the relationship between the effects on the behavior of the social media intervention itself and the effects of other factors [28,33,34]. Most studies measure knowledge and attitudes, risk perception and concerns, and the intention to be vaccinated. Tracking vaccination uptake by accessing participants' medical records could help address this shortcoming, but there are apparent issues here regarding confidentiality. Second, although studies suggest that social media interventions could be effective for people who are already open to vaccination or who already recognize the role of social media in the field of health and wellbeing, it is well understood that social media platforms are vulnerable to exploitation and have become a forum for antivaccine discussions [3]. Third, social media interventions have the potential to exacerbate disparities in healthcare that result from differences between social groups regarding access to the internet and to smartphones [25,35,40]; people without access to social media can become marginalized or completely excluded.

Behavior change theory
Across the studies included in this paper, we identified the behavior change theories used in interventions to influence vaccine uptake, which are the Health Belief Model and the Theory of Planned Behavior, and SCT. However, since only nine studies were identified in this review, additional research is needed to provide more robust evidence for the utility of these theories. Besides, considering only a small number of theories are being implemented in the design of social media interventions, there is scope for expanding the use of theory to provide a greater understanding of vaccine hesitancy and inform the design of interventions aimed at addressing it and improving vaccine uptake.
A clear difference was identified between the studies with and without theory-theory-based studies are more measurable and replicable. The unique role of theoretical methods is their function on bridging the gap between knowledge and behavior, which could expand the function of intervention from knowledge, attitude to intention and behavior. In terms of the application of behavioral theories, we found that in addition to directly guiding the interventions, many studies only used the theories to design parts of the intervention content, such as videos, articles, and stories, or to measure the effects of interventions [33,[37][38][39]. Notably, only three studies have been consistent in their design and evaluation, using the same framework to measure effectiveness [25,28,44].
Through the qualitative evaluation on the effectiveness, we found that the Health Belief Model has great potential in improving vaccine knowledge, intention and uptake, although HBM-based studies also identified no significant improvement in knowledge, intention, or vaccination rate [25,38,43]. Among the HBM components, we found cues to action is the most underdeveloped and least researched element, which is in accordance with the previous studies [50], especially the external cues. However, since vaccination is a social contract wherein getting vaccinated is the morally right behavior [51], which needs endorsement from the social environment. This might be a possible explanation for the low effectiveness of the three low effective studies.
Meanwhile, it highlights the need for a multitheorybased framework that can take all determinants into account. Considering vaccine hesitancy is a behavior in a complex network of related behaviors, changing that one behavior is likely to influence the other components of the system and reconfigure the whole network. In this case, a systematic framework considering a wide range of options would aid the development and implementation of any intervention [52]. Furthermore, although behavior change models such as HBM and TPB have been used to inform the interventions and may have some effect, they may suit only one specific circumstance and have limited value. A framework based on multiple theories, such as Behavior Change Wheel [53], could facilitate the evaluation process by providing tools to diagnose possible loopholes in the development of the intervention and so improve the design and achieve the desired goal.

Recommendations for future interventions
Our review demonstrates that behavioral theorybased social media interventions have the potential to improve vaccine knowledge and confidence, whereas there is insufficient evidence for its function on the effectiveness of increasing vaccine intention or uptake. Most of the findings from the reviewed studies show that these approaches were effective in improving knowledge, attitudes, or belief [30, 31, 33, 37-39, 41, 43, 44, 54], but only a few addressed vaccine concerns sufficiently to achieve a positive vaccination decision [33,40,41,44]. This result may be explained by the fact that most social media interventions tend only to inform and educate participants; taking people on a journey from social media exposure to changes in vaccine behavior is proving a challenge. Based on our findings, we have several recommendations for future studies.
First, future work is required to focus on the behavior change process by adopting the appropriate mode according to the context and targeted outcomes. This review has summarized the strengths and limitations of the different social media components, which could facilitate the better incorporation of behavior theory and social media mode in future interventions.
Second, the targeted outcome should be shifted from knowledge improvement to intention and behavior change. Studies have confirmed that people can remain highly suspicious of vaccines despite the facts and the evidence and that knowledge alone is not enough to initiate behavioral change [55]. Using behavior change theories to better understand multiple factors, such as beliefs, motivations, and socio-cultural influences, and integrating these theories with social media interventions, is seen as one way to move people from merely receiving information to changing their behavior [18].
Third, this review highlights the need to use a multitheory-based framework to inform and evaluate the interventions. It is not only because the multi-theory-based framework has more potential to consider more determinants and trigger the behavior change in the whole system, but also because it can better facilitate the measurement and comparison between different interventions. In this review, we found that although some studies have already used behavior change theories to guide and inform the development of interventions, the effectiveness of those interventions, especially those that use social media campaigns without reporting effectiveness on behavior outcomes, remains difficult to gauge. Furthermore, especially for studies based on behavior change theories, inconsistencies in terminology make comparisons yet more problematic. Interventions based on a single study provide insufficient data for assessing how successful the intervention would be among a broader population or with a different vaccine.
There are undoubtedly some lessons to be learned from the existing studies on implementing theory-based social media interventions. First, social media alone is not an intervention but a channel for facilitating engagement between people who want to interact with others [37]. Therefore, strategies need to be adopted to ensure the accuracy and coherence of the information being provided, and experts in the relevant fields need to be in place to moderate group discussions [37,39,41]. Second, positive behavioral change is more likely to happen if there is a strong trust between the researchers and the target group [37]. Such trust can be fostered by ensuring transparency in all communications, acknowledging participants' concerns, and responding promptly to their questions with tailored information.

Limitations of this systematic review
To the best of our knowledge, this is the first broad-based and systematic review of social media interventions that aim to increase vaccine confidence with respect to all vaccine types and populations. The evidence appears to support and strengthen the feasibility of using theory-based social media interventions to address vaccine hesitancy.
Nonetheless, some limitations in the review process and research design have been identified. First, our review covered all types of vaccines worldwide, but vaccine hesitancy is a highly contextualized problem. The utility of the results is therefore limited. The studies included are at different stages, so the results on effectiveness need to be interpreted with caution. Second, some methodological issues should be mentioned. The articles collected were from only six databases (PubMed, Scopus, ACM, Cochrane, ProQuest, and Web of Science), they were all published between January 1, 2011, and January 1, 2021 in the English language, and no consideration was given to grey literature. The variation in research design, methods, and outcomes further reduce the feasibility of making comparisons between the various findings. The results of the studies are difficult to quantify, but the number of publications can at least be used to reflect the amount of research being carried out in the different disciplines. Finally, although the findings of some of the broader studies could be applicable to a wider population, the overall heterogeneity of the research led to a limited generalization of the conclusions.

CONCLUSION
The purpose of this review was to identify the breadth and effectiveness of behavior change theories and social media tools used in interventions aimed at addressing vaccine hesitancy. The review analyzed 20 studies from 3065 search results according to our search strategy. To the best of our knowledge, this is the first review to examine the use of behavior change theories and social media as tools for delivering vaccination interventions. The review has found that generally theory-based interventions were defined with more rigor allowing for quantifying and comparing outcomes. The most common theory and social media tools used in the previous interventions were Health Belief Model and posting educational messages. Educational posts were proved to be effective in increasing knowledge but less helpful in triggering behavior change. Dialogue-based social media intervention performed well in improving people's intention to vaccinate. It also highlights the importance of using a multitheory framework and of tailoring social media interventions to specific circumstances. However, caution must be applied with a small number of studies included and the heterogeneity between the studies. Despite the limitations, the initial results appear to support the view that interventions informed by theory and delivered via social media platforms offer a significant opportunity and a valuable avenue to address vaccine hesitancy. This review could provide guidance for the future development and proliferation of such interventions.

SUPPLEMENTARY MATERIAL
Supplementary material is available at Translational Behavioral Medicine online.

Compliance with Ethical Standards
Conflict of Interest: The authors declare that the research was conducted in the absence of any commercial or financial.
Authors' Contributions: All review authors contributed to the conduct of the research. L.L. contributed to the selection of studies, data extraction and assessment of the risk of bias of included studies. L.L. drafted the manuscript and C.W. and P.K. provided critical comment on the drafts. All authors contributed to the interpretation of review findings.
Transparency statements: This review aligns with the reporting guidelines specified within the 2009 PRISMA checklist for systematic reviews and utilized Cochrane methodology. The completed 2009 PRISMA checklist for systematic reviews has been provided in Supplementary File A.
Ethical Approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed Consent: Informed consent was obtained from all individual participants included in the study. This article does not contain any studies with animals performed by any of the authors.

Appendix A | Keywords
Keywords Results TITLE-ABS-KEY: (twitter OR facebook OR "online social network" OR "social media" OR sns OR Instagram) AND (vaccin* OR immuniz* OR immunis* OR jab* OR Shot*) AND (intervention OR strateg*) OR TI(vaccin* OR immuniz* OR immunis* OR shot? OR jab?) AND AB,TI,IF(((behavio?r* OR psychology OR psychological* OR attitude?) AND (chang* OR alter* OR modif* OR diagnosis OR determinant? OR framework? OR intervention* OR theor* OR model?)) OR (theor* adj2 ("reasoned action" OR "normative conduct" OR "social cognitive" OR "self efficacy" OR "habit goal" OR "health belief"))) AND    The study findings confirmed that the campaign has been successful in engaging the FB users. The descriptive analyses show higher engagement, higher reach and more positive sentiments for the posts that contained personal stories compared to those containing factual information. Parents engaged with different types of content; however, the engagement was highest when the content was personal and less factual, even if the factual information related to the safety of the vaccine. This is not in accordance with the opinions obtained from focus groups where the participants requested more in-depth information, including the safety of the vaccine. In order to assess the performance of different types of content, They categorized the posts on the FB page into six subgroups: 1. To present the results of one year of this activity in terms of how the contents of the page were received, liked, and shared among its followers.
Events were the most popular type of news, followed by press releases, and scientific publications. Institutional Videos and documents are forms of communication less considered or appreciated by users. Examine the prior attitudes of first year medical students to seasonal influenza immunization (their risk-benefit calculation, their sense of vulnerability to seasonal influenza, their overall attitudes towards immunization and their immunization history) and assess the impact of the most popular vaccine-critical YouTube videos on their attitudes towards seasonal influenza vaccine The study did not find a significant difference in the responses to the questions asked before and after watching the videos. This study randomly assigned medical students to watch one of two YouTube videos with different rhetorical styles (evidence-based versus anecdotal) and measured any change in attitudes and behaviors before and after watching the video Allen et al. [25] Country: USA Researchers aimed to understand if the credibility of the source of a forum post influenced the readers' perception of vaccination risk. They did this by presenting identical narratives, but with different introductory texts (one from a neutral online health forum and the other from a known anti-vaccination website). Parents took part in discussion forums, where they were exposed to the following conditions: Participants randomly assigned to the conditions of a 2 (source peer vs governmental organization) x4 (influence strategy: source credibility, self-belief, direct challenge, indirect challenge) between-subjects design. Online discussion group to discuss raising adolescents and related issues. Parents were asked to log in on a specified date and time and to be available to participate in the experiment for 50 minutes. Some messages were sent by the study team to look like they came from peers or government officials and were followed by a pro-vaccination comment with different persuasion strategies.  This study compared the differences between using factual information versus personal experience in a post on health-related Internet forums. The researchers observed how each condition of the initial post affected subsequent communication on the forum. They considered the number of responses each post received and how emotional the language/content of the posts were. At the time of the enrollment visit, women in the intervention arm also selected their preferred method for reminders -text message, e-mail, phone call, private Facebook message, or standard mail. The "Staying on Track" software system, designed for this study, recorded subject data, and sent the automated reminders. Each intervention participant received four messages (one if she selected standard mail), sent three days apart prior to doses two and three. The reminder schedule mirrored the recommended dosing schedule. Reminders for dose two were sent six weeks after the initial visit. Timing of reminders for dose three was dependent on when the participant returned for dose two; reminders were sent either 12 weeks after the second dose or 24 weeks after the first dose, whichever was sooner.