Understanding the role of social media usage and health self-efficacy in the processing of COVID-19 rumors: A SOR perspective

Apart from the direct health and behavioral influence of the COVID-19 pandemic itself, COVID-19 rumors as an infodemic enormously amplified public anxiety and cause serious outcomes. Although factors influencing such rumors propagation have been widely studied by previous studies, the role of spatial factors (e.g., proximity to the pandemic) on individuals’ response regarding COVID-19 rumors remain largely unexplored. Accordingly, this study, drawing on the stimulus-organism-response (SOR) framework, examined how proximity to the pandemic (stimulus) influences anxiety (organism), which in turn determines rumor beliefs and rumor outcomes (response). Further, the contingent role of social media usage and health self-efficacy were tested. The research model was tested using 1246 samples via an online survey during the COVID-19 pandemic in China. The results indicate that: (1)The proximity closer the public is to the pandemic, the higher their perceived anxiety; (2) Anxiety increases rumor beliefs, which is further positively associated rumor outcomes; (3) When the level of social media usage is high, the relationship between proximity to the pandemic and anxiety is strengthened; (4) When the level of health self-efficacy is high, the effect of anxiety on rumor beliefs is strengthened and the effect of rumor beliefs on rumor outcomes is also strengthened. This study provides a better understanding of the underlying mechanism of the propagation of COVID-19 rumors from a SOR perspective. Additionally, this paper is one of the first that proposes and empirically verifies the contingent role of social media usage and health self-efficacy on the SOR framework. The findings of study can assist the pandemic prevention department in to efficiently manage rumors with the aim of alleviating public anxiety and avoiding negative outcomes cause by rumors.


Introduction
At the initial stage of the COVID-19 pandemic outbreak, public anxiety was significantly aroused owing to increasing numbers of infections and areas infected, thus leading to a series of unusual purchasing behaviors (e.g., hoarding toilet papers, food, and masks) and unscientific preventive behaviors (e.g., abuse of virus-related medicine and alcohol) among the public Laato et al., 2020). These irrational response not only distorted the functioning of local society but also induced physical and mental suffering in residents when they were facing the COVID-19 (Gao et al., 2020;Tasnim et al., 2020). Some research indicated that these negative outcomes such as tremendous economic losses and lengthened psychological distance were partially caused by rumors and misinformation circulating on social media (Tasnim et al., 2020;Zhu, 2021), which posed a serious threat to the COVID-19 response. This is because, under information heterogeneity and polarization, individuals who seeking COVID-19 information are difficult to judge the authenticity of the information in social media (Huang et al., 2021). Besides, individuals without pandemic-relevant knowledge may feel anxiety which in turn influence their decision-making . Therefore, to diminish the negative effects of rumors and to make an appropriate response to the pandemic, it is imperative to understand how COVID-19 related factors trigger individuals' anxiety, thus forming trusting beliefs towards rumors and behaving irrationally.
Although previous studies on rumors have extended our understanding of the antecedents that cause individuals' trusting beliefs and irrational behaviors, some significant knowledge gaps remain. First, many scholars have indicated that factors regarding the content of a rumor (e.g., the source reliability of rumor, information without clear source, and argument volume) and rumor receivers' psychological factors (e.g., personal involvement, anxiety, and uncertainty) are critical in causing rumor belief and subsequent behaviors during a crisis (DiFonzo & Bordia, 2007;Pezzo & Beckstead, 2006;Wang et al., 2018). Besides, in the context of the COVID-19, individuals' perceptions of spatial distance to the center of the pandemic may also influence their rumor beliefs. This is because individuals who lived in the risk center area will perceive a higher level of risk and anxiety (Burns & Slovic, 2012), which makes them more prone to trust pandemic-related rumors. Specifically, individuals living closer to the center of the pandemic outbreak may respond to rumors differently from those living farther from the pandemic. However, we have little knowledge regarding how spatial factors (e.g., proximity to the pandemic) influence rumor receivers' rumor beliefs by influencing their anxiety during the COVID-19. Therefore, to better understand the influence mechanism of rumors, we propose the first research question: How do individuals' perceptions of proximity to the pandemic influence their anxiety, which in turn determines their responses such as rumor beliefs and rumor outcomes?
Second, during this special period, social media to some extent enlarges the public's anxiety by providing an ideal platform for the anxious citizen to access pandemic-relevant rumors and propagating them (Luo et al., 2022;Yu et al., 2021), thus sharpening the public's rumor beliefs and causing rumor outcomes such as a series of irrational behaviors. Previous studies regarding rumors mainly have focused on the dark side of social media, for example, its ability to spread rumors and increase the public's anxiety during the COVID-19 outbreak (Gao et al., 2020;Nekovee et al., 2007). Further, platform characteristics and information factors of social media have also been found to induce rumor beliefs and behavioral outcomes (e.g., cyberchondria and verified information sharing) Oh et al., 2018). However, there is also a bright side to social media usage in such situations because of its capacity to provide social support to alleviate the effects of stressful situations (Arora et al., 2007;Oh, Lauckner, et al., 2013). Therefore, another possibility exists-that social media can play a positive role by enabling individuals to evaluate rumors efficiently and avoid unnecessary behavioral outcomes during the COVID-19 pandemic. However, the role of social media usage in shaping individuals' anxiety and response towards the COVID-19 rumors has been significantly under-explored in the current literature. Accordingly, the second research question is as follows: How can social media usage shape the influence mechanism of rumors concerning the COVID-19?
Third, individual differences are closely associated with the ability to discern the validity of rumors circulating on social media (Pennycook et al., 2020). Specifically, individuals with greater cognitive reflection and science knowledge have a stronger ability to discern misinformation posted on social media regarding COVID-19. This finding is consistent with the social cognitive theory proposed by (Bandura, 2006), which indicated that one's self-efficacy refers to her/his competence in coping with tasks or stressors. During the outbreak of the COVID-19, individuals who have high health self-efficacy can effectively engage in preventive behaviors in the face of pandemic and have better mental health (Yıldırım & Güler, 2020). On the other hand, rather than promoting behaviors, self-efficacy not only leads to overconfidence but also lowers their performance (Moores & Chang, 2009), thus shaping rumor beliefs regarding the pandemic much easier. Therefore, we propose that the effect of health self-efficacy on discerning rumors regarding the COVID-19 pandemic may counterintuitively backfire. However, little is known about the role of health self-efficacy in influencing anxiety and discerning rumors. Accordingly, the third question of this study is as follows: Are the effects of individuals' anxiety on their rumor beliefs contingent upon their health self-efficacy?
To answer the research questions provided above, this paper draws upon the stimulus-organism-response (SOR) framework (Mehrabian & Russell, 1974) to conceptualize proximity to the COVID-19 pandemic, anxiety, and rumor beliefs, respectively. Then, a research model with seven hypotheses is tested using data from a large-scale survey conducted during the COVID-19 in China. This study makes multiple contributions to the literature concerning the SOR model and rumors. First, this study fills the research gap in the SOR literature by firstly investigating role of the spatial factor (proximity to the pandemic) on individuals' anxiety which shaping their beliefs and subsequent behaviors towards the COVID-19 rumors. Second, our study extends the rumor literature by examining the contingent effects of social media usage and health self-efficacy on the processing of the COVID-19 rumors. In practice, the epidemic prevention department and relevant organizations in epidemic place can utilize the findings of this study to efficiently manage rumors, therefore alleviating public anxiety and avoiding practicing unusual behaviors caused by rumors.

Rumors and social media
In the social psychology literature, a rumor refers to a story or a statement in general circulation that is lacking confirmation or certainty regarding facts (Allport & Postman, 1947). DiFonzo and Bordia (2007) provided a more detailed and comprehensive definition of rumors relating to three aspects: the context, the content, and the social function. They noted that rumors are unsubstantiated claims that are widespread in vague, dangerous, or potentially threatening situations. In particular, previous research on crisis informatics has shown that social media typically has the advantage of being able to improvise, spread, and disseminate information more easily, faster, and wider than does mainstream media (Guo & Zhang, 2020). With the rise of social media platforms such as Twitter, Facebook, Reddit, Weibo, and WeChat, the spread of rumors has transcended national boundaries and has increased significantly in terms of speed and audience . Although some scholars proposed some methods for recognition of rumor stances in online social media (Luo et al., 2020), social media usage during the pandemic period still increase individuals' anxiety and cause severe mental health problems and cyberchondria (Gao et al., 2020;Laato et al., 2020;Oh & Lee, 2019). The rumor problem has presented a severe challenge to the effective use and scientific management of social media, and even has a significant impact on the real world (Liu et al., 2019).
However, regarding the relationship between social media and rumors, some contradictory findings are revealed (Sahni & Sharma, 2020). According to their review, although social media has brought some problems during COVID-19, it also promotes the well-being of individuals and public health when it is used wisely and prudently. For example, social media can hinder the spreading of rumors by providing and transmitting truthful facts from health experts (Sahni & Sharma, 2020). Antheunis et al. (2013) pointed out that patients' use of social media contributes to their ability to obtain health-related social support and improve their health management. Table 1 presents the relevant research on rumors in relation to social media.
Although some studies on rumors have been conducted during the COVID-19 pandemic, two main deficiencies are apparent in the current studies. First, scholars have not kept abreast of the rampant rumors circulating on social media in the context of the COVID-19. In general, the existing literature has not systematically examined the role of social media on antecedents and consequences of rumors regarding the pandemic. Second, previous studies mainly focused on the negative role of social media in the process of rumor propagation, with little attention paid to the positive role. The role of social media may reverse in discerning the truth against the rumors, which has a significant influence on controlling the COVID-19 pandemic. Based on the above discussion, this paper aims to elucidate the influencing mechanism of rumors during COVID-19 and introduce social media as a significant contingent factor. Next, we will introduce the SOR model that manifests the decision process regarding rumors and another contingent factor, health selfefficacy. Mehrabian and Russell (1974) devised the famous SOR model in the field of behavioral psychology to explain and predict the effects of different environmental stimuli on human cognition, emotion, and behavior. Mehrabian's SOR model is a modification and optimization of Woodworth's stimulus-response (S-R) model (Woodworth & Marquis, 2014), but adding the "O" variable and thereby focusing on the internal consciousness of humans and other organisms. The model assumes that different external stimuli have different effects on the internal state of human beings, and then it determines the decision-making behavior of human beings based on human internal cognitive, and emotional factors (Mehrabian & Russell, 1974).

Stimulus-organism-response model
The SOR model not only has been widely studied in the research fields of marketing and e-commerce to understand consumer behavior (Ettis, 2017;Namkung & Jang, 2010) but also is applicable for exploring the antecedents and consequences of rumors propagation in the context of the pandemic. B€ ohm and Pfister (2005) found that in the process of spreading rumors, the wider the rumor spreads, the more likely will it present to people the illusion that the epidemic situation is not under control. In addition, they indicated that the more the illusion is, cause the higher risk in relation to the epidemic people may perceive, which will, in turn, cause confusion and uncertainty regarding whether the epidemic will affect them personally, leading to anxiety and other negative emotions.
During the COVID-19, the influence mechanism of rumor can also follow the rule of the SOR model. In this study, proximity to the COVID-19 pandemic regarded as an external stimulus can arouse individuals' anxiety, thus leading to a series of psychological and physiological reactions (e.g., rumor beliefs and outcomes). Further, in the context of the COVID-19 pandemic, information spreading has two unique features: (1) social media has played a significant role in rumor spreading and witnessed many unusual behaviors (Tasnim et al., 2020); and (2) the virus affects human health that users with different levels of self-efficacy on their health may react to the COVID-19 rumors differently. Thus, to better fit the human decision during the COVID-19 pandemic and gain a better understanding of COVID-19 rumors, this study incorporates social media usage and health self-efficacy as contingent factors into the SOR model.

Health self-efficacy
Self-efficacy refers to "people's beliefs about their capabilities to produce designated levels of performance that exercise influence over events that affect their lives" (Bandura, 1977). Accordingly, in this paper, health self-efficacy is defined as people's beliefs about their capabilities to keep healthy during the COVID-19 pandemic (Lee et al., 2008). When faced with some fuzzy information or unverified news regarding the pandemic, individuals' discerning abilities significantly depend on their health self-efficacy. Therefore, health self-efficacy is a valid and legible instrument for assessing how people avoid becoming infected by, recognize the symptoms of, and home-manage COVID-19 (Hern andez--Padilla et al., 2020).
Health self-efficacy has been investigated as a significant factor influencing individuals' behaviors and emotions during the pandemic. For example, individuals with a higher level of health self-efficacy are more willing to engage in preventive health measures such as complying with government-advised preventive measures and taking vaccines (Bults et al., 2011;Liao et al., 2011). Further, high self-efficacy can increase individuals' elaboration of news and knowledge about swine flu (Lo et al., 2013). On the other hand, health self-efficacy can be served as an efficient means of alleviating individuals' negative emotions such as anxiety regarding the COVID-19 (Petzold et al., 2020). However, some scholars indicated that high self-efficacy may negatively influence subsequent performance (Vancouver et al. 2001(Vancouver et al. , 2002. Similarly, individuals with a high level of self-efficacy became overconfident in themselves, thus leading to poor performance (Moores & Chang, 2009). These findings can be explained by the perceptual control theory (Powers, 2005), which illustrated that individuals with high self-efficacy may become overconfident when they perceived a relatively small difference between the perceived external state and an inner desired state. Thus, when facing rumors regarding COVID-19, individuals with high health self-efficacy may have a perception that they have a high discerning ability, thus readily placing trust in rumors without seeking further confirmation. Although current literature has investigated the positive role of health self-efficacy during the COIVD-19, its role in the influence mechanism of rumors has rarely been empirically investigated. By testing the moderating role of health self-efficacy, this study fills the gap in the literature regarding the role of health self-efficacy on rumor propagation during the period of the COVID-19 pandemic.

Research model and hypotheses
Drawing on the SOR model (Mehrabian & Russell, 1974), our research model (see Fig. 1) is to explain the influence mechanism of rumors regarding the COVID-19 pandemic. Specifically, this framework suggests that proximity to the pandemic influences rumor receivers' anxiety, which intervenes in their rumor beliefs and rumors outcomes. Moreover, the moderating effects of social media usage, as well as health Rumor spread Perceived importance positively affects rumor propagation. Tanaka et al. (2012) and Oh and Lee (2019) Anxiety, personal involvement, and ambiguity of information sources positively influence rumor propagation.
Oh, Agrawal, and Rao (2013) Public emotions such as anger, fear, sadness, and happiness positively influence rumor spread in the context of COVID-19. Dong et al. (2020) Rumor trust Both personal involvement and rumor fear positively affect rumor trust, while the presence of counterrumors negatively influences rumor trust.

Rumor spreaders
A lower ratio of following-to-follower is more likely to spark rumors. Bodaghi  self-efficacy on the SOR framework, are tested. According to the basic tenet of the SOR framework (Mehrabian & Russell, 1974), in the context of this study, the stimulus (S) is defined as an external stimulus that is the proximity to the pandemic, while the organism (O) refers to the emotional response of individuals which refers to individuals' anxiety, which is a common emotional response to disasters (Bergeron & Sanchez, 2005;Kim & Kim, 2017). Jacoby and Jacob (2002) indicated that an external stimulus will trigger organism to feel emotions. Previous scholars have pointed out that panic caused by risk events produces a ripple effect, which means that when individuals are located in the risk center area, their perception level of risk and anxiety is highest; the farther away people are from the center, the lower the risk (Slovic, 1987) and negative emotion they perceive (Burns & Slovic, 2012). For example, the closer people are to nuclear pollution, the more worried they are about the impact of nuclear leakage on their lives, resulting in stronger negative emotions and evasive behavior (Semenova et al., 2019;Suzuki et al., 2018). In the case of our study, rumor receivers who are living nearby the place of COVID-19 outbreak will perceive a higher level of anxiety. In other words, proximity to the pandemic as a stimulus can increase rumor receivers' anxiety. This is because people perceive they have a high probability of being affected by COVID-19 if there is an outbreak nearby. Therefore, based on the above discussion, we propose the following hypothesis.
H1. Proximity to a pandemic is positively associated with anxiety.
In the SOR framework, the response (R) refers to the organism's behavioral decision-making, which integrates external environmental stimulus and internal psychological attitude, including avoidance and approach behavior (Mehrabian & Russell, 1974). Accordingly, response in this study refers to rumor receivers' rumor beliefs aroused by anxiety during the COVID-19. Since anxiety is particularly likely to prevent people from forming an objective evaluation of circumstances (Hannabuss, 2008), an individual is more likely to trust a rumor which is congruent with her/his emotional state (Na et al., 2018). Besides, anxiety serves as an important link between disasters and rumor propagation (Allport & Postman, 1947;Askarizadeh et al., 2019;Oh, Agrawal, & Rao, 2013). Under the pandemic health crisis, pandemic diseases aroused a high level of anxiety (Bergeron & Sanchez, 2005), thus leading anxious individuals to form their beliefs that believing in pandemic-relevant rumors. Consequently, the higher level of anxiety a rumor receiver perceives, the more likely she or he is to trust rumors regarding the COVID-19. Based on the above arguments, we hypothesize that.

H2. Anxiety positively affects rumor beliefs.
Based on the Theory of Reasoned Action (Fishbein & Ajzen, 1975), individuals' beliefs affect their attitudes toward a behavior, thus leading to the intention to perform this behavior. In our study, rumor beliefs and rumor outcomes are conceptualized as trusting beliefs regarding COVID-19 related rumors and irrational behavioral intentions respectively. When individuals perceive rumors highly trustworthy in an anxious situation, their usual behaviors will be significantly distorted. For example, previous studies indicated that individuals' perceived relevance and importance of a rumor is positively associated with their trusting beliefs, thus verifying and sharing rumors (Chua & Banerjee, 2018;Oh & Lee, 2019). In other words, when people perceive that rumors regarding the pandemic are trustworthy, they are more likely to perpetuate the rumors and take relevant actions such as purchasing unnecessary preventative items. Therefore, we propose the following hypothesis.
Social media as an emerging interactive communication channel empowers the public to actively search for information. Therefore, through using social media, the public is no longer information recipients of mass media but turns to be active information seekers who wish to receive information instantly (Stephens & Malone, 2009). During the COVID-19, compared to traditional information sources such as mass media, print media, and online official websites, information on social media is easier to arouse the public's information anxiety (Soroya et al., 2021). This is because news regarding the COVID-19 pandemic on social media leads people to understand that the pandemic is highly contagious and deadly (Lee et al., 2020). As such, the public may take more time using social media to search for relevant information when they realize they are getting closer to the center of a pandemic. Accordingly, the positive association between proximity to pandemics and anxiety can be strengthened by the frequent use of social media. Therefore, individuals who are close to the pandemic may perceive a higher level of anxiety when they use social media frequently to search for relevant information about the COVID-19. Based on the above arguments, we propose the following assumption.
H4. Social media usage strengthens the relationship between proximity to the pandemic and anxiety.
When people are in a state of anxiety, they are more likely to trust rumors because anxiety prevents their judgment ability (Hannabuss, 2008). However, anxiety also increases people's demand for information during a pandemic. For example, public anxiety leads to an increase in information seeking during the outbreak of influenza H1N1 (Tausczik et al., 2012). This is because anxious individuals prefer to collect more information to ease emotional pressure (Rosnow 1991). In this vein, anxious individuals will increase their information seeking on social media during the COVID-19. By doing so, they acquire much professional and social support from social media (Arora et al., 2007;Oh, Lauckner, et al., 2013), which may help ease the public's anxiety. Accordingly, under the condition of a high level of social media usage, individuals' anxious mood may be eased, and they will be more cautious about forming trusting beliefs regarding a rumor. In other words, the association between anxiety and rumor beliefs can be weakened by using social media for social support. Based on the above arguments, we propose the following assumption.
H5. Social media usage weakens the relationship between anxiety and rumor beliefs.
Health self-efficacy refers to individuals' beliefs that they are capable to manage their health (Lee et al., 2008). Although some scholars indicated that health self-efficacy plays a significant role in engaging in preventive actions regarding a pandemic (Bults et al., 2011;Liao et al., 2011) and reducing negative emotions (Petzold et al., 2020), individuals with a high level of health self-efficacy may become over-confident in their capabilities of managing their health during the COVID-19 (Moores & Chang, 2009). This is because that people with strong self-efficacy tend to believe in themselves sufficiently to extricate themselves from a threatening situation (Bandura, 1977). In such a case, when individuals have a higher level of health self-efficacy, anxious ones are likely to prevent themselves from forming an objective judgment, which has led to a higher level of rumor belief during the COVID-19 pandemic. In other words, anxious individuals with a high level of health self-efficacy may be more firmly believe in their subjective judgments regarding rumors of COVID-19. Accordingly, the positive effect of anxiety on rumor beliefs will be strengthened by health self-efficacy. On this basis, we hypothesize that.

H6. Health self-efficacy strengthens the positive relationship between anxiety and rumor beliefs.
When faced with a threat, people with high-level self-efficacy tend to engage in higher task intention and less procrastination, which means they are more likely to accept challenges, execute available strategies, and persist regardless of setbacks and risks (Haycock et al., 1998;Wieber et al., 2010). In the realm of healthcare, those with a high level of health self-efficacy are more prone to engage in health-promoting and health-impairing behaviors (Bandura, 1986). During the COVID-19, health self-efficacy has been found to have positive effects on a series of preventative behaviors such as self-isolation . When faced with rumors regarding the COVID-19, people with stronger health self-efficacy have greater intention to perpetuate rumors they accept and are more confident in their capabilities to complete the task of protecting themselves, for example, executing an irrational purchase. Under the condition of a high level of health self-efficacy, individuals may have a stronger willingness to turn their rumor beliefs into outcomes. In this vein, the association between rumor beliefs and rumor outcomes is strengthened by health self-efficacy. Therefore, we put forward the following hypothesis.
H7. Health self-efficacy strengthens the positive relationship between rumor beliefs and rumor outcomes.

Measurements
An online cross-sectional survey was conducted to test the proposed research model and hypotheses. We evaluated all the measures based on a five-point Likert scale ranging from 1 to 5, representing "strongly disagree" to "strongly agree." Some of the constructs were adapted from a pre-designed research instrument from prior studies. The measures for the degree of anxiety were adapted from Elhai et al. (2020), while the measures for social media usage and health self-efficacy were adapted from Oh, Lauckner, et al. (2013). In addition, the measures for rumor outcomes were adapted from Laato et al. (2020). Furthermore, some of the measures were modified to fit our research context. Proximity to the pandemic was measured by the number of newly confirmed COVID-19 cases in areas where samples lived in. The measures of rumor beliefs were self-defined for our research purpose. All the measures are shown in Appendix.

Data collection
To collect the data for this study, we chose February 2020 to distribute this survey online for three reasons: (1) our research topic is about the mechanism of proximity to the pandemic, anxiety, beliefs, and outcomes of COVID-19 rumors on social media; (2) in this February, China was suffering heavy levels of morbidity and mortality caused by COVID-19, whereas pandemic issues in other global regions remained relatively low (Qiu et al., 2020); and (3) under the circumstance of the pandemic, it was safer and more convenient to survey our participants online. Therefore, our samples for the online survey were representative to a certain degree. Further, to guarantee the quality of the questionnaire, we firstly translated the English version of the questionnaire into Chinese and then asked three postgraduate students to check the translation accuracy. A pretest with 50 participants has been conducted before formally distributed to make sure the readability and easy-to-follow of each question. Our questionnaire received 4628 responses in total. In the data cleaning process, we removed incomplete questionnaires and invalid questionnaires in which the answers were the same and obviously contradictory. 1246 responses were eventually valid after the quality control checks, with a response rate of 27%. The demographic information is provided in Table 2.

Data analyses
Structural equation modeling (SEM) consists of covariance-based SEM and variance-based SEM (e.g., partial least squares-SEM) (Cepeda-Carrion et al., 2019). The partial least squares SEM (PLS-SEM) technique is used to analyze our research model due to the following two advantages. First, compared to covariance-based SEM, PLS-SEM is more suitable for an exploratory study for theory development and the structural model consists of one or more formative constructs (Gefen et al., 2011;Hair et al., 2019;Khan et al., 2019). Second, since multiple regression and ANOVA (analysis of variance) may underestimate the interaction effect, PLS-SEM is a better choice to test the interaction effect in the model (Chin et al., 2003). Below, both the measurement model and the structural model are assessed to test the appropriateness of this research model and related hypotheses.

Measurement model
The reliability and validity of the measurement model were tested to indicate the goodness of fit. The Cronbach's α value of each variable was above 0.70, and all other values were greater than 0.80, except health self-efficacy (as shown in Table 3). Composite reliabilities were above 0.85 (as shown in Table 5), indicating that the research data had high reliability (Chin et al., 2003;Urbach & Ahlemann, 2010). Further, the loadings of each item were above 0.65 (as shown in Table 4; values in bold) and the average variances extracted (AVEs) were all greater than 0.60 (as shown in Table 5), which indicates that convergent validity was good (Chin et al., 2003). For discriminant validity, the items of each construct loaded greater on themselves than on other constructs (as shown in Table 4) and the square root of the AVE of each variable (as shown in Table 5) was greater than the correlations with the other variable, thus indicating acceptable discriminant validity (Chin et al., 2003). The results are shown in Tables 3-5 4.3.2. Structural model Based on the hypotheses, the structural model was tested in two stages. During the first stage, we examined the direct effects of the model. The relationship between proximity to the pandemic and anxiety was found to be positive and significant (β ¼ 0.043; p < 0.001). The results further reveal that anxiety had a positive effect on rumor belief (β ¼ 0.261; p < 0.001). And the effect of rumor beliefs on rumor outcomes was positive and significant (β ¼ 0.393; p < 0.001). Thus, hypotheses H1-H3 are all supported.
In the second stage, we tested the moderating role of social media usage and health self-efficacy. The results show that social media usage had a positive effect on the relationship between proximity to the pandemic and anxiety (β ¼ 0.092, p < 0.01), while having no significant effect on the relationship between anxiety and rumor belief (β ¼ À0.045, p > 0.10). Meanwhile, health self-efficacy had a positive effect both on the relationship between anxiety and rumor belief (β ¼ 0.085, p < 0.01) and the relationship between rumor belief and rumor outcome (β ¼ 0.112, p < 0.01). Thus, H4, H6, and H7 are supported, but not H5. The results of model estimations are presented in Fig. 2.

Key findings
This study investigated the influence mechanism of COVID-19 rumors drawing upon the SOR framework and the contingent roles of social media usage and health self-efficacy in this mechanism. Several main findings emerge from this study. First, proximity to the pandemic as a stimulus is positively associated with anxiety as an organism, which has a positive effect on rumor beliefs as a response. This finding is consistent with the basic tenet of the SOR framework (Mehrabian & Russell, 1974). During COVID-19, the closer rumor receivers are to the pandemic, the more likely they are to feel anxious and the higher possibility they will choose to trust rumors regarding COVID-19.
Second, social media usage strengthens the relationship between proximity to the pandemic and anxiety but has no significant effect on the relationship between anxiety and rumor beliefs. This implies that in the initial stage of the COVID-19 pandemic, under a high level of social media usage, individuals living near the pandemic will become more anxious. This finding is supported by previous research  which indicates that more frequent social media exposure can lead to cyberchondria. However, regarding the unsupported hypothesis 5, one explanation is grounded in the fact that frequent exposure to social media during a global health crisis strengthened the effects of information overload and information anxiety, which in turn leads to information avoidance (Soroya et al., 2021;Yavetz et al., 2022). In such a situation, anxious individuals will not continue using social media for seeking information before they are forming rumor beliefs. Therefore, social media usage has a non-significant effect on the association between anxiety and rumor beliefs.
Third, health self-efficacy strengthens (1) the positive relationship between anxiety and rumor beliefs, and (2) the positive relationship between rumor beliefs and rumor outcomes. This finding implies that under the condition of the high level of health self-efficacy, anxious rumor receivers are more likely to form their rumor beliefs, thus putting their beliefs into practice. As shown by Moores and Chang (2009), this result confirmed that self-efficacy is positively associated with overconfidence in individuals' judge ability, thus indicating the negative effect of health self-efficacy on performance.

Theoretical implications
This research yields several contributions. First, this paper not only expands the scope of use of the SOR model but also promotes the systematization of rumor research by linking several scattered variables (e.g., social media usage and health self-efficacy) into an integrated  framework. In the previous literature, the SOR framework has been mostly applied in the field of offline and online consumer behavior (Gatautis et al., 2016), and there was a gap in the field of rumor exploration (Pal et al., 2019). The pandemic environment (proximity to the pandemic) is an external stimulus that guides the public into a specific state (anxiety), who take related action (rumor beliefs and rumor outcomes), which is consistent with the mechanism of SOR relations applied in marketing (Eroglu et al., 2001). Second, this paper, to the best of our knowledge, is one of the first that proposes and empirically verifies both the bright and dark side of social media on the influence mechanism of rumors during COVID-19. Previous studies have mainly focused on the dark side of social media in spreading rumors, increasing public anxiety, causing rumor beliefs and rumor outcomes during the COVID-19 pandemic (Gao et al., 2020;Laato et al., 2020), however, few studies have systematically examined the bright role of social media on the influence mechanism of rumors regarding COVID-19. This study addresses this research gap by investigating the contingent role of social media on (1) the association between stimulus and organism, and (2) the association between organism and response. Although some scholars indicated that social media can provide useful social support to people to mitigate the negative effects of stressful situations on health (Arora et al., 2007;Oh, Lauckner, et al., 2013), the results show that the negative role of social media still dominates arousal of anxiety and beliefs in pandemic rumors during the COVID-19.
Third, this study highlights the significant role of health self-efficacy in the influence mechanism of rumors. Previous studies have pointed out that health self-efficacy is positively related to individuals' health preventive measures (Bults et al., 2011;Liao et al., 2011;Tang & Wong, 2003) and mental health during COVID-19 (Yıldırım & Güler, 2020). However, the negative role of health self-efficacy has been significantly ignored. These results indicate that a high level of health self-efficacy may make individuals being blindly confident in their abilities to manage their health, in turn easily forming rumor beliefs and rumor outcomes. This study enriches the rumor literature by highlighting the negative role of health self-efficacy in the spreading of rumors regarding a pandemic.

Practical implications
This empirical study provides some significant implications in practice. First, the results show that proximity to the pandemic will increase rumor receivers' anxiety, which in turn shapes their rumor beliefs.
The pandemic prevention department should pay more attention to the emotional states of individuals living nearby the pandemic center and take measures to alleviate their anxiety and to avoid forming rumor beliefs.
Second, our results indicate that social media usage positively moderates the positive relationship between proximity to the pandemic and anxiety. The Cyber Security Institute needs to pay more attention to the dark side of social media on rumor propagation and publish some regulations to effectively manage information and news on social media to relieve public anxiety during the pandemic.
Third, our study shows that health self-efficacy strengthens the link between anxiety, rumor beliefs, and outcomes. Health self-efficacy, to some extent, can create negative effects on the influence mechanism of rumors regarding a pandemic. Therefore, forming correct health selfefficacy under scientific and medical guidance is also vital for rumor control. The pandemic prevention department should take advantage of both mass media and social media to educate the public to form the right health self-efficacy, thus taking accurate measures to cope with the COVID-19.

Limitations and future directions
This study has some limitations. First, this research was conducted in China. Therefore, the generalizability of the results in other countries should be questioned. Second, we have only collected data through an online survey, lacking samples of infrequent Internet users. In future research, samples are expected to be enriched by collecting more offline data. According to the work of Zhao et al. (2021), future research should collect the data set from real social media to build a rumors detection model. Finally, rumor propagation is the result of the combined action of event importance and information fuzziness (Allport & Postman, 1947;Rosnow & Ralph, 1991). Other factors such as interpersonal relationships and social influence could also possibly affect rumor belief (Brock et al., 2012;Kelman, 1958;Neal & Chartrand, 2011), which can serve as directions for future research.

Conclusion
The purpose of this study is to understand the influence mechanism of rumors regarding the COVID-19 to alleviate public anxiety and reduce negative outcomes caused by rumors. Therefore, this study draws upon the SOR framework to explore how proximity to the pandemic influences anxiety, which in turn affects rumor beliefs and rumor outcomes. Further, the contingent role of social media usage and health self-efficacy on the SOR framework was examined. Based on an online survey, the research model and seven hypotheses were tested. The results indicated that proximity to the pandemic is positively associated with anxiety, which leads to rumor beliefs and rumor outcomes. In addition, social media usage strengthened the effects of proximity to the pandemic on anxiety. Health self-efficacy positively moderated the relationships (1) between anxiety and rumor beliefs, and (2) between rumor beliefs and rumor outcomes. This study enriches the literature on the SOR framework and rumor by highlighting the influence mechanism of rumors regarding the COVID-19. Practically, the findings of this study will assist the pandemic prevention department to efficiently manage rumor propagation to relieve public anxiety and alleviate negative outcomes cause by rumors during the COVID-19.

Conflict of interest
We have no conflicts of interest to declare.