The proliferation of vaccine misinformation on social media has seriously corrupted the public’s confidence in vaccination [1-3]. Extreme antivaxers create vaccine misinformation online [4-6], and ordinary users often spread vaccine misinformation voluntarily on social media [7,8] because of the emotional and attractive nature of misinformation. Fortunately, social media platforms are increasingly implementing positive interventions to resist vaccination misinformation [9,10].

These endeavors mainly focus on 2 different approaches. The first approach relies on information technology, such as fact-checking labels [9] or debiasing strategies [11], to reduce the impact of existing vaccine misinformation on its audience. The other approach provides more vaccination knowledge on social media to suppress the diffusion of vaccine misinformation [4,8,12,13].

However, additional strategies to mobilize the public to debunk vaccine misinformation on social media are needed to successfully achieve public health goals. Just as vaccine misinformation can drive sharing intentions, provaccination information, if presented properly, can motivate the public’s willingness to share on social media. Inspiring the public’s willingness to share provaccination information can improve the current situation in which antivaccination messages occupy a more controversial space than provaccination messages on social media [8,14], and prevent the adverse effects of misinformation [15]. Hence, the main goal of this study is to explore strategies to present effective messages to improve the public’s willingness to share provaccination messages on social media.

Previous studies have confirmed that emphasizing the collective benefits of vaccination, such as the long-term benefits of herd immunity for society, could positively affect participants’ vaccination intention [8,16,17]. In particular, according to the “sharing for the community” model [18], community interest serves as a more influential motivation for the online sharing of health-related knowledge than personal gain does [19], and information-sharing behaviors driven by community interest are based on reciprocity norms [20]. Therefore, it is logical to hypothesize that provaccination messages that emphasize collective value can promote the willingness to share provaccination messages better than those that emphasize individual value.

In addition, message framing plays a key role in health communication [21]. Based on gain versus loss message framing [22,23], the value of vaccination can be framed by highlighting either desirable consequences (eg, health benefits) or undesirable consequences (eg, disease risk). Although a meta-analytic review showed that gain-framing messages are significantly more likely to encourage illness prevention behaviors compared to loss-framing messages [22], another study showed that framing effects are significant only when individuals perceive the issue to have high personal relevance [24]. Therefore, we assume that the framing effects may be less obvious when messages are described with collective value, since personal relevance is diminished when messages are described with collective value compared to individual value. Thus, we used message framing as an independent variable to explore whether message framing and value type synergistically affect vaccination message-sharing intention.

We conducted 2 survey experiments during the COVID-19 pandemic outbreak in mainland China. Experiment 1 explored whether a message that emphasized collective value improved vaccination intention more than one that emphasized individual value during the COVID-19 pandemic outbreak. Experiment 2 tested our primary hypothesis of whether collective value significantly facilitates the dissemination of provaccination messages online. We also measured the perceived importance and responsibility of message sharing as mediating variables in experiment 2 to examine the mechanism of collective value on the willingness to share messages.

Two 2 (value type: collective vs individual) × 2 (message framing: gain vs loss) between-participant experiments were conducted using Sojump [25], an online study portal. Experiment 1 explored the main effect of value type and message framing on vaccination intention for the influenza vaccine (including a no-message control group), while experiment 2 focused on message-sharing willingness.

The sample size was estimated using G*Power 3.1 [26], assuming a statistical power of 90%, a significance level of .05, and an effect size of 0.25. The estimated minimum sample size for 2-factor ANOVA was 172 (n=43, 25%, for each condition). Accordingly, we recruited 450 volunteers in total (experiment 1: n=250, 55.6%; experiment 2: n=200, 44.4%) for monetary compensation.

All participants were native Chinese readers; none had a medical education background or were from Wuhan, China. Manipulation checks in the survey resulted in a total of 384 valid participants in the experiments (experiment 1: n=213, 85.2%; experiment 2: n=171, 85.5%). Table 1 presents the descriptive information about the participants.

The system flowchart can be seen in Figure 2. After obtaining informed consent, the survey gathered information about the participants’ demographics (gender, age, education, and marital status) and control variables. The participants were randomly assigned to 1 of the conditions using the online survey platform’s randomization function. Those assigned to 1 of the 4 message conditions proceeded to a page that showed them a message and required them to answer 2 manipulation check questions. They were then instructed to respond to questions related to the variables described in the following section. Those in the control condition in experiment 1 proceeded directly to a similar survey without reading a message.

All control and dependent variables used in our experiments are presented in Table 2. The details of all measures used in experiments are provided in Multimedia Appendix 1.

Vaccine confidence, assessed by modifying the 4-item vaccine confidence scale to 3 items by removing the item about religion [30], was measured as a control variable in experiments 1 and 2. The propensity to share on social media [31] served as a control variable in experiment 2.

We focused on vaccination intention for the 2020-2021 influenza vaccine. A self-administered scale was used. Specifically, participants were provided with medical news. The news read, “Once COVID-19 infection occurs together with influenza, the difficulty of diagnosis will be increased…” In addition, the intention to receive the influenza vaccination was measured by how likely the participants were to accept the influenza vaccination for themselves or recommend it to their families in the winter. The secondary dependent variable in experiment 1 was vaccine acceptance, which captured the entire conceptual domain of vaccine acceptance [32].

Experiment 2 comprised 3 dependent variables: the perceived importance of messages [33], the perceived responsibility for message sharing [34], and message-sharing willingness [33].

The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of Nankai University (#NKUIRB2020023, approval date February 25, 2020).

The descriptive and inferential statistics of experiment 1 are shown in Table 3. Demographic variables were also included in the analysis as control variables.

The descriptive statistics of experiment 2 are shown in Table 4.

In experiment 1, we found that individual value had a more substantial effect than collective value on influenza vaccination intention. Interestingly, the results from experiment 2 indicated that the message describing collective value promoted message-sharing willingness better than the message related to individual value; the perceived responsibility for message sharing mediated the effect of collective value on message-sharing willingness but not on perceived importance.

In general, individual-focused messages increased vaccination intention, but collective-focused messages increased the likelihood of sharing. These results seem to produce conundrums for message designers and reveal that designers should finely construct specific messages for different targets. For example, if the goal is to make provaccination messages occupy more space on social media, then collective value should be prioritized to describe messages. In another circumstance, if the improvement of vaccination rates is more important and urgent, especially under circumstances with high risks, individual value should be considered first. Moreover, the message of collective value described as gain framing also enhanced vaccination intention compared to the control group. Therefore, collective-gain messages effectively increase vaccination intention, while enhancing message-sharing willingness.

This study also obtained some other results. First, there were no differences between collective and individual value or gain and loss framing on vaccine acceptance; however, vaccine acceptance improved in all intervention groups compared to the control group. Second, gain framing was more effective than loss framing in influenza vaccination intention. Finally, we did not find an interaction effect between value type and message framing on any dependent variables.

These results contribute to the literature on online health knowledge sharing and provide implications for practice.

There are some limitations of this study. First, we did not design a condition that combined collective and individual value and that tested whether such messages can promote both vaccination intention and message-sharing willingness. This limitation can be addressed through follow-up research.

Second, the external validity of this study is limited. In a realistic network environment, other types of information with high attractiveness or interesting themes, such as stories about celebrities or sports news, may cause the public to ignore provaccination messages [36]. Therefore, it is necessary to simulate the real network context to verify the robustness of the results of this study.

Third, we conducted the research in the early stages of COVID-19. During this period, the epidemic in Wuhan, China, was more serious, while people in other regions were mostly in a state of unknown risk. As the form of the epidemic continues to change in China, the public’s perception of the risk will also change dynamically. This may further lead to changes in the way messages function in different severe situations. Subsequent studies could consider collecting data at different time points and drawing on studies that use time series models [37,38] in an attempt to optimize health strategies to resist vaccine misinformation and boost vaccination.

Finally, the participants were mainly female, younger, and single and had a college degree or above. Although we controlled for these demographic variables in the analysis, the external validity of the experimental results is limited. This study can be further improved to examine whether the function of the messages varies among populations.

Despite the limitations, the study complements and extends previous studies that focused on collective benefits in personal vaccination decisions.

The effect of value type on vaccination intention contradicts a previous study that found that collective benefit promotes vaccination intention more than individual benefit does [16]. Notably, a study supports our finding, showing that individuals self-categorized as being in a high-risk group are more likely to adopt vaccination behavior following self-benefit messages than social benefit messages [39]. Our study was conducted during the COVID-19 outbreak, and high risk perception might have resulted in the public’s heightened sensitivity to individual value messages.

In contrast, message-sharing willingness was promoted by collective value more than individual value, and perceived responsibility played a mediatory role. These findings are consistent with a similar study that found that social integrative benefits significantly and positively influence knowledge-sharing intention in virtual communities [40].

Finally, the results of message framing coincide with a previous study that suggested that gain-framing messages are more likely to encourage illness prevention behaviors compared to loss-framing messages [22].

Online messages emphasizing individual value promote the intention to vaccinate more strongly than those emphasizing collective value, but this effect is reversed in message-sharing willingness. Furthermore, perceived responsibility for message sharing plays a mediating role between collective value and message-sharing willingness. Our findings have practical implications for constructing and providing effective, targeted provaccination messages.