key: cord-0718971-ld0nogfm authors: Zhao, Xinyan; Tsang, Stephanie J. title: Self‐protection by fact‐checking: How pandemic information seeking and verifying affect preventive behaviours date: 2021-08-12 journal: Journal of Contingencies & Crisis Management DOI: 10.1111/1468-5973.12372 sha: 1ffa350683fa6d76dfdf0b84c21813f5be386310 doc_id: 718971 cord_uid: ld0nogfm The COVID‐19 pandemic has witnessed the proliferation of a plethora of (mis)information on various media platforms and inconsistent crisis instructions from different sources. People consume crisis information from multiple channels and sources to better understand the situation and fact‐check COVID‐19 information. This study elucidates how Americans determine their preventive behaviours based on their information seeking and verifying behaviours during the pandemic. Our results were based on a US nationally representative sample (N = 856), and showed that proactive preventive behaviours (e.g., washing hands frequently) were positively affected by information‐seeking through interpersonal channels, news media, and the government, whereas avoidance preventive behaviours (e.g., avoiding social gatherings) were only positively affected by information‐seeking through news media. Crisis information verifying had positive effects on all types of preventive behaviours. Crisis managers are recommended to reach out to the public using appropriate channels and sources and facilitate individual's ability and motivation in verifying pandemic information. communication deals with the change in preventive behaviours during infectious disease outbreaks and how the behavioural shift can be facilitated by media use and socio-psychological factors such as channel-related beliefs (Dunwoody & Griffin, 2013; Oh et al., 2020; Ophir & Jamieson, 2020) . There has also been growing interest regarding the cognitive processing of misinformation and its correction in public health crises (Bode & Vraga, 2018; Krause et al., 2020; Lu & Jin, 2020; van der meer & Jin, 2020) . This study contributes to the thriving literature in two important ways. First, the crisis communication scholarship focuses more on information dissemination and its nuances that are conditioned by information sources and forms (e.g., SMCC; Liu et al., 2019 ). Yet, there has been insufficient knowledge of how people's verification of the sought information (Lu & Jin, 2020) affects their preventive behaviours during crises. Our study considers information seeking and verifying as two important dimensions of information consumption and examines both dimensions' effects on preventive behaviours in a pandemic. This enables us to achieve a more holistic understanding of the nuances and outcomes of crisis information consumption. Second, there has been a paucity of crisis communication research on the socio-psychological mechanisms (e.g., anxiety) underlying the relationships between information consumption and preventive behaviours. Extending the Social Amplification of Risk Framework (SARF; Kasperson & Kasperson, 1996; Pidgeon et al., 2003) , our study also tested the mechanisms accounting for the relationships between information seeking and verifying and health behaviours. To understand how crisis information seeking and verifying are associated with preventive behaviours, we relied on survey data based on a U.S. nationally representative sample (N = 856) during the COVID-19 pandemic. Our results can help crisis managers and public health officers navigate the complex information environment during a pandemic and benefit from a more nuanced understanding of various facets of information consumption and their implications for facilitating preventive behaviours in public health crises. After the first case of COVID-19 was identified in Wuhan, China (WHO, 2020b) , the outbreak dramatically escalated. On March 11, 2020, COVID-19 was officially declared a pandemic (WHO, 2020c). The United States was one of the countries hit hardest by the pandemic, with over 13 million cases by November 2020. The US government and health agencies only started to respond to the pandemic in March, which gave rise to a flood of criticisms regarding the delayed and muffed pandemic response of the Trump administration (Lipton et al., 2020; Yamey & Gonsalves, 2020) . On various occasions, Trump downplayed the threat of COVID-19 and called the criticisms a "hoax" in February during a rally (Garrett, 2020) . There have also been inconsistent or even contradictory health guidelines across different sources, such as Trump and the Centers for Disease Control and Prevention (Kim & Kreps, 2020; Yamey & Gonsalves, 2020). The slow, insufficient and inconsistent communication across various government agencies has resulted in confusion among the people regarding the actual risks (Wise et al., 2020) and uncertainties in relation to the appropriate preventive actions. This study examined how people consumed and reacted to COVID-19 information from different sources and channels in April 2020. A pandemic features enormously high uncertainty and anxiety (Garrett, 2020) . This is mainly because of the lack of scientific consensus and accurate information regarding the route of transmission, clinical symptoms and effective treatment at the start of the pandemic (Reynolds & Seeger, 2005) . To cope with a novel threat, individuals are urged to learn more about the associated risks and potential preventive measures through a multitude of channels and sources (Anthony et al., 2013; Moreno et al., 2020) , particularly on digital platforms . The abundance of health information made possible by ever-changing technologies can help the public stay abreast of the developing situation , keep them informed about the latest medical findings and practical guidance (Oh et al., 2020) , and mobilize individuals and groups to coordinate relief efforts (Reuter & Kaufhold, 2018) . However, the health information environment in a pandemic is often populated with biased information, false claims or even conspiracy thinking (Bode & Vraga, 2018) , which bias people's pandemic-related beliefs (Kata, 2010; Vraga & Bode, 2017) and impede their adoption of effective actions (Tan et al., 2015) . The prevalence of inconsistent or even false messages circulating can prompt individuals to perform fact-checking (Krause et al., 2020) , which is required to determine the veracity of the coronavirusrelated claims in this case. It is crucial to understand how individuals act upon information verification, as the public may perceive misinformation differently from the consensus among experts and health authorities (e.g., WHO; Vraga & Bode, 2020) . Thus, their information processing and verification may have different behavioural implications. Previous studies suggest that information seeking through different channels and sources is positively associated with preventive behaviours during crises (e.g., Liu, 2020; Liu et al., 2019) . Yet, it is still unknown how people's information verifying affects their preventive behaviours. According to , people's exposure to COVID-19 misinformation reduces the perceived information insufficiency and leads to information avoidance. We argue that the individual conduct of fact-checking can play a key role in predicting preventive behaviours, in addition to traditional information-seeking behaviours. To understand how preventive behaviours can be affected by information seeking and verifying, we introduce the SMCC model and extend it based on the misinformation correction literature below. In the public health context, an infectious disease crisis/outbreak occurs when a disease quickly spreads between people in a region. An outbreak, if not well controlled, turns into a pandemic when a new virus spreads all over the world (WHO, 2010) . In public health crises, governmental agencies are expected to handle the unpredictable incident that poses significant and widespread risks to the public. The SMCC model provides theoretical explanations and predictions regarding people's information seeking and sharing and their subsequent preventive behaviours in crises (e.g., Vijaykumar et al., 2015) . The SMCC mainly discusses three factors affecting people's information dissemination and preventive behaviours, including the following: (1) information forms/channels (e.g., social media, offline interpersonal communication); (2) information sources (e.g., the organization responsible for handling a crisis, a third party); and (3) types of social media publics (i.e., influential social media creators, social media followers, inactives) (e.g., Austin et al., 2012; Jin et al., 2016; Liu et al., 2016; Zhao, Zhan, & Liu, 2018) . The original SMCC studies emphasize crisis information seeking and sharing as two distinct constructs of information dissemination (e.g., Lee & Jin, 2019 ). Yet, the relationship between crisis information verifying and preventive behaviours has not been well understood (Lu & Jin, 2020) . This study examines both information seeking and sharing as effective information consumption strategies for reducing uncertainty in a complex information environment (Bode & Vraga, 2018; Lu & Jin, 2020) . Crisis information seeking through various platforms and sources can mitigate information insufficiency (Griffin et al., 1999) by providing a repertoire of potentially useful information. Crisis information verifying allows individuals to evaluate information quality, increase response efficacy, and perform appropriate preventive actions by comparing the consistency of information from different channels and sources (Anthony et al., 2013) . The following sections discuss crisis information seeking and verifying in a pandemic. The two primary information channels during a pandemic are social media and interpersonal communication. Recent crisis communication research has revealed the significance of social media in information seeking . Social media provide timely, unfiltered and personally relevant crisis information to help people interpret their risks (Chong & Choy, 2018) . However, some scholars argue that individuals may delay their protective actions because they engage in prolonged information seeking and evaluation processes to fully understand the terse information they encounter in social media . Indeed, there has been inconsistent evidence for how social media information seeking affects preventive behaviours in crises (e.g., Liu et al., 2019; Oh et al., 2020) . This inconsistency may be explained by the motivation of social media use and the urgency to decide preventive actions in a certain crisis. During imminent threat crises, such as tornados, people have to make an immediate decision following the initial warning. In this case, information seeking through social media fulfills their immediate information needs and facilitates timely actions (Liu et al., 2019) . Nevertheless, people who perceive lower levels of involvement may delay their protective actions when they continue seeking information on social media. During the COVID-19 pandemic, social media information seeking is anticipated to dampen preventive behaviours. This is because there has been widespread misinformation, disinformation and conspiracy theories surrounding the coronavirus (Nguyen & Catalan, 2020) . As a result, people's perceived risks and needs for preventive actions can vary to a large extent, particularly considering the political implications of the pandemic. On one hand, the Trump administration and the Republican party kept downplaying the threat of the pandemic (Halon, 2020) , driving the conservatives to perform less preventive behaviours. By contrast, the Democratic party took the pandemic and preventive measures seriously, trying to prompt liberals to be more alert and adopt more preventive behaviours, such as mask-wearing (Kessel & Quinn, 2020) . Given that social media platforms (e.g., Twitter) were fuelled with President Trump's statements during the pandemic (Jang et al., 2020) , we expect voices downplaying the pandemic to have an upper hand and, in turn, lead to a negative relationship between social media information seeking and preventive behaviours during the COVID-19 pandemic. On the other hand, information seeking through interpersonal channels is foreseen to incite preventive behaviours (Liu et al., , 2019 Stephens et al., 2013) . Two rationales can account for such an expectation. One is that the information exchanged through interpersonal channels can be more personalized and convincing (Dutta-Bergman, 2004) . When personal contacts share COVID-19-related information, people might take it more seriously and stay away from possible avenues of contracting the virus. Another aspect is that interpersonal communication features a higher level of media richness, including more vividness and attachment, which can assist people in interpreting the urgency and the risks associated with the pandemic more easily. Based on the discussion, we propose the first two hypotheses: H1: Social media information seeking negatively affects preventive behaviours during the pandemic. H2: Interpersonal information seeking positively affects preventive behaviours during the pandemic. Seo, 2019; Zhang et al., 2015) . During a crisis, news media constitute the primary gateway for people to seek information and knowledge about a hazard (Reynolds & Seeger, 2005; Zhao, Zhan, & Wong, 2018) . Many consider news media as the most credible sources of crisis information (Jang & Baek, 2019; Utz et al., 2013) , probably because journalists as gatekeepers filter out unverified sources or incredible information in their coverage. During the COVID-19 pandemic, 59% of Americans believed that news media provided them with useful coronavirus information (Pew Research Center, 2020). COVID-19 information seeking through online news media was found to be positively associated with preventive behaviours among the Chinese (Liu, 2020) . Therefore, we anticipate COVID-19 information seeking online to positively predict preventive intentions. Government agencies and organizations, such as the CDC or WHO, have used social media for pandemic preparation, rapid information sharing and public education (Spence et al., 2015) . Individuals typically seek information about imminent threats from governmental agencies (Sjöberg, 2018; Sutton et al., 2018) . During public health crises, people typically perceive official sources (e.g., the CDC) as more credible than unofficial sources and rely more on official sources for decision-making (Austin et al., 2012 ). Yet, during the COVID-19 pandemic, local, national and international agencies have provided inconsistent or even inaccurate information (Kim & Kreps, 2020) , which could impede individuals' preventive behaviours. For example, in March 2020, the CDC recommended that people "wear a facemask if you are sick" (CDC, 2020b, March 14) . In April, the CDC advised individuals to "cover your mouth and nose with a cloth face cover when around others" (CDC, 2020a, April 13). Given the fluidity of the situation and the shift in recommendations, laymen may be confused about whether and when they should use masks. As such, there may be a weak positive association between information seeking through government (i.e., health agencies) and preventive behaviours. Based on the discussion, we propose the following two hypotheses: H3: News media information seeking positively affects preventive behaviours during the pandemic. H4: Government information seeking positively affects preventive behaviours during the pandemic. There has been growing scholarly interest regarding misinformation and its correction (Bode & Vraga, 2018; Krause et al., 2020; Tan et al., 2015) , especially in public health crises. Misinformation in a pandemic typically involves health-related false claims on the virus' origin, outbreak development, preventive measures, and treatment options. The COVID-19 pandemic saw a vast amount of misinformation, especially on social media platforms such as Twitter (Cuan-Baltazar et al., 2020; Rosenberg et al., 2020) . Brennen et al. (2020) found that the number of English fact-checks rose more than 900% from January to March 2020. An emerging direction in SMCC is how people process or verify crisis information before information transmission (Lu & Jin, 2020; van der meer & Jin, 2020) . In an infectious disease outbreak that is characterized by high risk and uncertainty, people need to verify the sought information to determine its accuracy and believability for guidance on preventive behaviours to be adopted (Bode & Vraga, 2018; Vraga & Bode, 2017) . According to Mileti & Sorensen (1990) , during disasters, at-risk individuals engage in a sequential cognitive process involving understanding, believing, and personalizing before deciding on the appropriate action. In this evaluative process, at-risk individuals engage in information seeking, sharing, and verifying to make sense of the situation and decide the protective actions to be carried out. Lu and Jin (2020) defined crisis information vetting as a way to cognitively process information and assess the quality of crisis information. They further discussed two layers of information vetting: information verifying (i.e., fact-check or confirm the accuracy of the information) and attitude formation (i.e., forming an attitude toward a threat by evaluating the content quality). Given that actual confirmation of the accuracy of the information can be difficult, particularly during the coronavirus pandemic, this study only focuses on the individual act of information verifying rather than on whether people were able to debunk false claims on their own successfully. To understand whether and how crisis information verifying affects preventive behaviour, we propose the following research question: RQ1: How does crisis information verifying affect preventive behaviours during the pandemic? A number of crisis communication scholars have called for more research regarding the socio-psychological mechanisms of individuals' behaviours in crises (e.g., Jin et al., 2016; Liu & Viens, 2020) . To better understand how information consumption affects preventive behaviours through cognitive and affective aspects of risk, we discuss the different aspects of risk and the SARF below. Risk has cognitive and affective aspects (Slovic et al., 2004; Terpstra, 2011) . On one hand, risk perceptions include perceived susceptibility and severity. In an infectious disease outbreak, perceived susceptibility refers to the likelihood to be infected by the virus and perceived severity refers to how severe the consequences of the infection are (Turner et al., 2006) . On the other hand, the risk-as-feeling hypothesis (Slovic et al., 2004 (Slovic et al., , 2005 suggests that negative effects, such as anxiety, help individuals evaluate and respond to risk information rapidly and automatically. Both cognitive and affective aspects of risk affect individuals' preventive behaviours in crises (e.g., Terpstra, 2011). According to the SARF (Kasperson & Kasperson, 1996; Kasperson et al., 1988; Pidgeon et al., 2003) , a threat can interact with social, psychological, and societal processes to amplify or attenuate people's risk evaluations. Risk information can be filtered through various social (e.g., government agencies) and individual (e.g., risk heuristics or prior attitude) amplification stations during transmission. These amplification stations shape the amount and salient aspects of the risk information and alter individuals' risk perceptions. Different communication channels and sources can be understood as social amplification stations (Vijaykumar et al., 2015) . Recent studies have revealed the prominent role of social media (Choi et al., 2017; Chong & Choy, 2018) and news media (Pidgeon et al., 2003; Shih et al., 2008) in amplifying the perceived risk. For example, social media use amplifies negative emotions, such as fear (Chong & Choy, 2018 ) and perceived susceptibility (on a collective level; Choi et al., 2017) , which stimulates preventive behaviours during crises (e.g., Oh et al., 2020) . However, most studies have only examined the consequences of information consumption without considering the complementary use of different channels or sources (cf. Seo, 2019) . People typically use as many available channels and sources as possible for uncertainty reduction in a crisis (Anthony et al., 2013; Liu et al., 2015) . For example, during the MERS crisis, the Korean government withheld key information regarding the list of hospitals affected by MERS, which incited people to turn to social media for information (Seo, 2019) . As such, it is crucial to examine the roles of all relevant channels and sources of crisis information in affecting preventive behaviours in a pandemic. In addition, there has been limited evidence regarding the interrelationships between perceived susceptibility, severity, and anxiety as mediators of the association between information seeking and verifying and preventive behaviours. To better understand how crisis information seeking through different channels and sources affects preventive behaviours through susceptibility, severity, and anxiety, we ask the last research question: RQ2: How do crisis information seeking and verifying affect preventive behaviours through susceptibility, severity, and anxiety? An online panel of survey data hosted and distributed by Qualtrics was collected from April 21-26, 2020 amid the exacerbation of the COVID-19 pandemic in the United States. The panel consisted of a nationally representative (in terms of age, gender, and education) sample of the US population above 18 years old based on quota sampling. As the number of recruited females aged 65 or above and hold less than a high school degree was not sufficient, the quota for less than a high school degree was reduced to 6% from 13% and distributed equally across the other education groups. The final sample size was 856, among whom 441 were females (51.5%) aged 18 to 86 years (M = 46.42, SD = 17.29) . Meanwhile, about 37% had high school or lower education levels, 52% had partial or full college education, and the remainder had a graduate degree. The average household income reported by the sample was between $50,001 to $60,000. We found that most variables assessed by multiple items were unidimensional, except for preventive behaviours. The means and standard deviations of all variables are reported below. For the multidimensional construct, the summary scores of each identified factor are included. Table 1 reports the summary statistics and the correlation matrix of all constructs. As adapted from the SMCC literature Lee & Jin, 2019) , four sets of questions were asked to gauge the subjects' tendency to seek COVID-19 information from different channels and sources on a 7-point scale ranging from 1 "strongly disagree" to 7 "strongly agree." That is, the subjects indicated the extent to which they would look for COVID-19 information from (1) This study asked the respondents to indicate how often they factchecked COVID-19 information and news by "searching for more details about a topic and evaluating the sources providing the information" (Bartolomeo, 2020) , including "searching on search engines," "browsing official websites of governmental/health agencies," "reading the news," and "visiting fact-checking sites" on a 7-point scale ranging from 1 "never" to 7 "frequently." The mean of information verifying was 4.03 (SD = 2.13, Cronbach's α = .87). Note: N = 856. Gender: male = 1, female = 2. Male is the reference group. Education: 1 = high school or below, 2 = some college, 3 = Bachelor's degree, 4 = Master's degree or above. Income level: 1 = less than $10,000, 2 = $10,001-$20,000, 3 = $20,001-$30,000, 4 = $30,001-$40,000, 5 = $40,001-$50,000, 6 = $50,001-$60,000, 7 = $60,001-$70,000, 8 = $70,001-$80,000, 9 = $80,001-$90,000, 10 = $90,001-$100,000, 11 = more than $100,000. Crisis involvement was measured by the extent to which respondents considered the pandemic as important and relevant on a 7-point scale (Cronbach's α = .87). Political orientation: 1 = very liberal, 7 = very conservative. *p < .001; **p < .01; ***p < .05. Based on the guidelines by the health authorities, such as CDC (CDC, 2020, March 14), a list of preventive behaviours was generated. The subjects indicated how often they had been engaging in the preventive behaviours on a seven-point scale ranging from 1 "never" to 7 "frequently." These items included: "washing your hands more often," "washing your hands appropriately (about 20 seconds)," "covering coughs and sneezes," "avoiding close contact with people who are sick," "avoiding dining out," "avoiding social gatherings," "avoiding travelling," "wearing a facemask," and "self-quarantine if I was in contact with someone who is sick." The risk measures were adapted from Turner et al. (2006) . Perceived susceptibility was assessed using three questions. On a seven-point scale ranging from 1 "definitely not" to 7 "definitely likely," the subjects indicated the extent of the likelihood that they would get COVID-19 this year, that they were more likely to get COVID-19 than other people their age, and the amount of risk they felt about getting COVID-19 (M = 3.62, SD = 1.48, α = .85). The subjects also indicated their perceived severity by indicating the extent to which they agreed or disagreed with the following statements: "COVID-19 is serious," "COVID-19 can cause death," and "COVID-19 is more severe than most people realize." (M = 6.17, SD = 1.21, Cronbach's α = .87). For anxiety, the subjects indicated the extent to which COVID-19 made them feel "anxious," "worried," and "concerned" (M = 5.09, SD = 1.79, α = .90). Age, gender, education, and income level were used as demographic covariates based on the literature (Choi et al., 2017; Tan et al., 2015) . Issue involvement and political orientation were also covariates (Gadarian et al., 2020) . For issue involvement, respondents indicated the extent to which they perceived the COVID-19 pandemic as (1) 1 "unimportant" to 7 "important" and (2) 1 "irrelevant" to 7 "relevant" on a 7-point scale (M = 6.02, SD = 1.60, Cronbach's α = .87). For political orientation, respondents indicated whether they considered themselves to be 1 "liberal," 7 "conservative," or somewhere in between (M = 4.02, SD = 1.73). To answer H1-H4, as well as RQ1, linear regressions through R were conducted, with different covariates, various information seeking variables, and information verifying used as predictors of two types of preventive behaviours. To answer RQ2, structural equation modelling (SEM) was conducted through the R "Lavaan" Package (Rosseel, 2012) . SEM is a set of multivariate statistical techniques to measure and analyse the relationships between observed and latent variables (Kline, 2015) . Figure 1 shows the main structural model hypothesizing sequential mediators based on the literature (e.g., Seo, 2019 (Little et al., 2002) was used to create one or more composite indicators (for details, see the note of Figure 1 ). Maximum likelihood estimation was used. We also tested an alternative model of parallel mediators to identify the better fitting model. The alternative model is similar to the main model, except that anxiety, perceived susceptibility, and severity simultaneously mediate the effects of different information seeking variables on preventive behaviours. Both models were evaluated based on the standard cut-off values for the model-data fit indices (Hu & Bentler, 1999; MacCallum et al., 1996) . The two models were compared based on the Bayesian information criteria (BIC; Raftery, 1995) . H1 hypothesizes that social media information seeking negatively affects preventive behaviours during the pandemic. Results from the linear regressions (Table 2) RQ2 asks how crisis information seeking and verifying affects preventive behaviours through susceptibility, severity, and anxiety. Moreover, the main model (BIC = 75,716) had a better fit than the F I G U R E 1 Conceptual structural model of sequential mediators. All covariates were regressed on all endogenous variables. They are not shown in the figure for simplicity. For the construct of social media information seeking, the item of Facebook was dropped due to relatively weaker item loadings than the other indicators (see the discussion section for the difference of information seeking across various social media platforms). The parcelling technique was used when a construct had more than three items (Little et al., 2002) . Parcelling was applied for two constructs. First, for information verifying, the third item was created by averaging subjects' responses to "fact-check by visiting fact-check websites" and "fact-check by reading news." Second, for avoidance preventive behaviours, the third item was created by averaging "avoid contact with those who were sick" and "avoid travelling." Refer to the measurement section for the specific items of perceived susceptibility and severity alternative model (BIC = 75,854). As such, the notion of sequential mediators (with anxiety predicting perceived susceptibility and severity) rather than simultaneous mediators was preferred for explaining the association between information seeking and preventive behaviours. 1 This study examined whether and how crisis information seeking and verifying affected preventive behaviours during the COVID-19 pandemic. Our results based on US nationally representative data showed that pandemic information seeking through interpersonal channels, news media, and the government was positively associated with proactive preventive behaviours, whereas only information seeking through news media was positively associated with avoidance preventive behaviours. Information verifying was consistently associated with all types of preventive behaviours. The results are discussed in detail below. First, news media information seeking was strongly associated with all kinds of preventive behaviours during the COVID-19 pandemic. These results support the pivotal role of news media in crisis communication (e.g., Utz et al., 2013) and show that people rely more on news media (compared with other sources or channels) for selfprotection during the pandemic. Information seeking through news media amplified people's "felt" and subsequently "perceived" risks, which, in turn, increased preventive behaviours. These results imply that even in the era of social media and communication technologies, news media are still the most important sources amplifying risk with regard to a crisis, thus supporting the original SARF (Kasperson & Kasperson, 1996) . Our results also demonstrate a process of sequential risk amplification, in which the affective heuristic helps individuals assess the likelihood of infection and how severe the situation is (Slovic et al., 2004) . The inclusion of both cognitive and affective components of risk provides stronger explanations and predictions for preventive behaviours (as evidenced by an increase in the explained variance of preventive behaviours). Second, social media information seeking was negatively associated with proactive and avoidance preventive behaviours. Social media information seeking had no effects on anxiety or facemask wearing. These results do not support the literature on the positive influence of social media on crisis-related perceptions, emotions, and behaviours (e.g., Chong & Choy, 2018; Oh et al., 2020 selective and purposeful use of digital platforms during the pandemic, that is, people might use Instagram/Snapchat for entertainment and relaxation during the pandemic. Altogether, these results suggest the importance of examining individuals' motivation and use of specific social media platforms during the pandemic. Next, interpersonal information seeking was positively associated with proactive preventive behaviours. These results support the literature regarding the pivotal roles of interpersonal channels (e.g., Liu et al., 2016) in accessing crisis information and shaping preventive behaviours. People rely more on coronavirus information from their personal contacts in deciding whether to adopt proactive behaviours, such as washing hands frequently, probably because this information contains personal implications and increase efficacy beliefs. Government information seeking was also positively associated with proactive preventive behaviours (and facemask wearing). Consistent with the literature, government agencies play a key role (e.g., Sutton et al., 2018) Given the consistent and positive relationships between crisis information verifying and various preventive behaviours, it is crucial for researchers to further investigate the antecedents, processes, and perceptual outcomes of crisis information verification. As the relationship between information verifying and preventive behaviours was not explained by any risk-related mediators (e.g., anxiety), future studies should test additional mechanisms, such as response efficacy or uncertainty reduction. For example, verifying crisis information may increase one's perceived effectiveness of the adopted action and thus facilitate preventive behaviours. Furthermore, answering the call for more public-driven research on the socio-psychological mechanisms of individuals' behaviours during crises (Liu & Viens, 2020) , our findings support a sequential process of risk amplification accounting for the associations between information seeking through various sources and channels and preventive behaviours during a pandemic, thereby advancing the literature on crisis communication and information behaviours (e.g., SMCC; Austin et al., 2012) . Future studies should also examine the mechanisms underlying the attitudinal and behavioural outcomes of information sharing and verifying by examining other important sociopsychological factors, such as construal level or cognitive elaboration. Given the heterogeneity of the crisis information environment on different social media platforms (e.g., Lee & Jin, 2019) , future research should systematically examine platform characteristics (e.g., affordances) and their impacts on preventive behaviours for a more refined understanding of information consumption on social media. The COVID-19 pandemic is an ongoing public health crisis worldwide. During this pandemic, people are not passive receivers of crisis information. Instead, they fact-check coronavirus information through various means and determine their preventive behaviours for staying safe during the pandemic. Therefore, crisis managers, particularly those who work with government agencies or health authorities that are responsible for the crisis response, should not only provide accurate, timely and targeted health information during the pandemic but also educate people in navigating the complex information environment. Furthermore, they should supply them with evidence-based methods for identifying false claims and combat the negative effects of unverified biased information. For example, government agencies, such as the CDC, can partner with influencers to share tips on COVID-19 fact-checking and direct users to credible sites of fact-checks on social media platforms. Public health officers can also collaborate with social media companies to develop various algorithmic solutions to address the infodemic, such as adding misinformation identification and verification functions or providing a built-in real-time fact-check service that is highly accessible. Furthermore, as news media information seeking (but not government information seeking) was associated with all kinds of preventive behaviours, public health information officers should still reach out and build relationships with local and national news media, and constantly provide media outlets with accurate, timely, and evidence-based crisis information. Using social media to raise people's perceived risk and preventive behaviours may be futile during the pandemic, perhaps due to the complex social media information environment, where people struggle to sift facts from misinformation. This study has several limitations. First, survey data do not enable the test of causal relationships among variables, although they provide more ecologically valid results compared to experimental data. Future studies should perform experiments to formally assess the causal influence of information verification on preventive behaviours and its underlying mechanisms. Second, numerous studies have shown that cultural background strongly affects perceived risk and preventive behaviours (e.g., . Future studies should examine the behavioural outcomes of information seeking and verifying in another cultural context or conduct cross-cultural comparisons. Third, our study is based on cross-sectional self-report data. Digital traces over time that track the changes in pandemic information consumption will be valuable to understand the dynamics of individuals' information processing goals (e.g., verification or confirming prior beliefs) and their effects on preventive behaviours during an epidemic. Despite the limitations above, our study highlights the importance of information verifying in driving preventive behaviours during the COVID-19 pandemic and clarifies the mechanism underlying the effect of information seeking on preventive behaviours, thereby opening new opportunities for crisis communication and management theory building. The data that support the findings of this study are available upon request from the corresponding author. ENDNOTE 1 As the model-data fit was not ideal, we also conducted theory-informed model modification based on the modification indices. Mainly by allowing cross-loading among information consumption constructs and their indicators (e.g., CDC information seeking cross-loaded on both social media and government information seeking), the model-data fit demonstrated a minor increase (χ 2 = 1596.51, relative χ 2 (i.e., χ 2 /df) = 3.56, SRMR = 0.09, RMSEA = 0.055, 90% CI RMSEA = [0.052, 0.058], p < .01, and CFI = 0.93). The modified model provided similar results. Message convergence as a message-centered approach to analyzing and improving risk communication How audiences seek out crisis information: Exploring the social-mediated crisis communication model How to fact-check the Internet See something, say something: Correction of global health misinformation on social media Types, sources, and claims of COVID-19 misinformation How to protect yourself & others Legacy and social media respectively influence risk perceptions and protective behaviors during emerging health threats: A multi-wave analysis of communications on Zika virus cases The impact of social media on risk perceptions during the MERS outbreak in South Korea The social amplification of haze-related risks on the Internet Misinformation of COVID-19 on the Internet: Infodemiology study The role of channel beliefs in risk information seeking Interpersonal communication after 9/11 via telephone and internet: A theory of channel complementarity Social media use during disasters: A research synthesis and road map Toward more audience-oriented approaches to crisis communication and social media research Partisanship, health behavior, and policy attitudes in the early stages of the COVID-19 pandemic Americans struggle to navigate COVID-19 "infodemic COVID-19: The medium is the message Coronavirus research March 2020 Series 3: Multi-market research Proposed model of the relationship of risk information seeking and processing to the development of preventive behaviors Stephanie Grisham: Democrats using coronavirus 'as a tool to politicize things and scare people Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives Exploratory analysis of COVID-19 related tweets in North America to inform public health institutes When information from public health officials is untrustworthy: The use of online news, interpersonal networks, and social media during the MERS outbreak in South Korea The scared, the outraged, and the anxious: How crisis emotions, involvement, and demographics predict people' conative coping The social amplification and attenuation of risk The social amplification of risk: A conceptual framework A postmodern Pandora's box: Anti-vaccination misinformation on the internet Both Republicans and Democrats cite masks as a negative effect of COVID-19, but for very different reasons An analysis of government communication in the United States during the COVID-19 pandemic: Recommendations for effective government health risk communication Effects of COVID-19 misinformation on information seeking, avoidance, and processing: A multicountry comparative study Principles and practice of structural equation modelling Fact-checking as risk communication: the multi-layered risk of misinformation in times of COVID-19 Crisis information seeking and sharing (CISS): Scale development for measuring people' communicative behavior in social-mediated public health crises He could have seen what was coming: Behind trump's failure on the virus. The New York Times To parcel or not to parcel: Exploring the question, weighing the merits How disaster information form, source, type, and prior disaster exposure affect public outcomes: Jumping on the social media bandwagon Social media use during disasters: How information form and source influence intended behavioral responses Crisis and risk communication scholarship of the future: Reflections on research gaps How people' active and passive communicative behaviors affect their tornado responses: An integration of STOPS and SMCC COVID-19 information seeking on digital media and preventive behaviors: The mediation role of worry Information vetting as a key component in socialmediated crisis communication: An exploratory study to examine the initial conceptualization Power analysis and determination of sample size for covariance structure modeling Communication of emergency public warnings: A social science perspective and state-of-the-art assessment Covid-19 communication management in Spain: Exploring the effect of information-seeking behavior and message reception in public's evaluation. El profesional de la información Digital mis/disinformation and public engagment with health and science controversies: Fresh perspectives from Covid-19. Media and Communication The effects of social media use on preventive behaviors during infectious disease outbreaks: The mediating role of self-relevant emotions and public risk perception. Health Communication. Advance online publication The effects of Zika virus risk coverage on familiarity, knowledge and behavior in the US-A time series analysis combining content analysis and a nationally representative survey Americans turn to technology during COVID-19 outbreak, say an outage would be a problem The social amplification of risk Bayesian model selection in social research Fifteen years of social media in emergencies: A retrospective review and future directions for crisis informatics Crisis and emergency risk communication as an integrative model The Twitter pandemic: The critical role of Twitter in the dissemination of medical information and misinformation during the COVID-19 pandemic Lavaan: An R package for structural equation modeling and more (Version 0.4-9 beta) Amplifying panic and facilitating prevention: Multifaceted effects of traditional and social media use during the 2015 MERS crisis in South Korea Media coverage of public health epidemics: Linking framing and issue attention cycle toward an integrated theory of print news coverage of epidemics It is not about facts-It is about framing. The App Generation's information-seeking tactics: Proactive online crisis communication Risk as analysis and risk as feelings: Some thoughts about affect, reason, risk, and rationality Affect, risk, and decision making Variability in Twitter content across the stages of a natural disaster: Implications for crisis communication Organizational communication in emergencies: Using multiple channels and sources to combat noise and capture attention Willingness to click: Risk information seeking during imminent threats Exposure to health (mis) information: Lagged effects on young adults' health behaviors and potential pathways Emotions, trust, and perceived risk: Affective and cognitive routes to flood preparedness behavior The role of anxiety in seeking and retaining risk information: Testing the risk perception attitude framework in two studies Crisis communication online: How medium, crisis type and emotions affected public reactions in the Fukushima Daiichi nuclear disaster Seeking formula for misinformation treatment in public health crises: The effects of corrective information type and source Social media and the virality of risk: The risk amplification through media spread (RAMS) model I do not believe you: How providing a source corrects health misperceptions across social media platforms. Information Defining misinformation and understanding its bounded nature: Using expertise and evidence for describing misinformation Changes in risk perception and protective behavior during the first week of the COVID-19 pandemic in the United States Coronavirus disease 2019 (COVID-19) Situation Report-94 WHO Director-General's opening remarks at the media briefing on COVID-19-11 COVID-19 cases, recoveries, deaths among most impacted countries as of Donald Trump: A political determinant of COVID-19 Media use and health behavior in H1N1 flu crisis: The mediating role of perceived knowledge and fear Disentangling social media influence in crises: Testing a four-factor model of social media influence with large data Segmenting and understanding publics in a social media information sharing network: An interactional and dynamic approach Selfprotection by fact-checking: How pandemic information seeking and verifying affect preventive behaviours