key: cord-0876576-tdhjfn9h authors: Luo, C.; Li, Y.; Chen, A.; Tang, Y. title: What triggers online help-seeking retransmission during the COVID-19 period? Empirical evidence from Chinese social media date: 2020-06-16 journal: nan DOI: 10.1101/2020.06.13.20130054 sha: 2ba981362578e1b8f9fbd6279c0fd0649d1ac078 doc_id: 876576 cord_uid: tdhjfn9h The past eight months witnessed COVID-19's fast-spreading at the global level. Limited by medical resources shortage and uneven facilities distribution, online help-seeking becomes an essential approach to cope with public health emergencies for many ordinaries. This study explored the driving forces behind the retransmission of online help-seeking posts. We built an analytical framework that emphasized content characteristics, including information completeness, proximity, support seeking type, disease severity, and emotion. Adopting the framework, a quantitative content analysis was conducted with a probability sample of 727 posts. The results illustrate the importance of individual information completeness, high proximity, instrumental support seeking. This study also reveals the severity principle and the power of anger in the dissemination of help-seeking messages. As one of the first online help-seeking diffusion analyses in the COVID-19 period, the theoretical and practical implications of this study are further discussed. 5 (Indian & Grieve, 2014) , Instagram (Andalibi, Ozturk & Forte, 2017) or online forums (Braithwaite, Waldron, & Finn, 1999; Coursaris & Liu, 2009; Pan, Shen, & Feng, 2017) increase the opportunities of receiving and providing social support from all sides. Retransmission, or the so-called retweeting, has always been seen as a crucial indicator of information diffusion in social media platforms (Yang, Tufts, Ungar, Guntuku, & Merchant, 2018) . Some studies treated retransmission as a judgment of the effectiveness of communication (Liu, Lu, & Wang, 2017; Wang et al., 2019) . The retweeting behavior has been demonstrated as a conversational practice on social media platform (boyd, Golder, & Lotan, 2010) ; a way to seek personal benefits from the social network (Recuero, Araujo, & Zago, 2011) ; a kind of prosocial behavior aims to offer help or provide advice to others driven by altruistic and reciprocity motivations (Lee et al., 2015) . Retransmission is critical in online help-seeking, especially in the epidemic context for two reasons. Firstly, retweeting is a typical one-to-many communication describing the degree of viral reach on social media (Liu et al., 2017) . More retransmission means more users receiving the message, thus increasing the chance of getting help. Secondly, from a practical point of view, facing the medical resources shortage and uneven distribution of medical facilities, widely disseminated help-seeking posts played the role of social warning, assisted official institutions in understanding urgent affairs as well as allocating supplies more effectively. What are the driving factors behind online help-seeking messages retransmission? Most existing studies categorized driving forces into two dimensions: the sender factors and the content factors (Briones, Nan, Madden, & Waks, 2012; Yang et al., 2018; Wang et al., 2019) . However, significant driving forces are varied by context. In this study, ordinary people use social media to seek help during a crisis time, which means senders' identities are very close. Thus, it would be more reasonable and significant to explore further in the content factor. There All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 6 is plenty of research providing conceptual references, such as the depth of self-disclosure (Altman & Taylor, 1973; Pan et al., 2018) , different types of support messages (Pan et al., 2018; Wingate, Feng, Kim, Pan, & Jiang, 2020) , physical and emotional proximity to the target (Huang, Starbird, Orand, Stanek, & Pedersen, 2015) , the social capital stock of the help seeker (Pan, Shen, & Feng, 2017) . Enlightened by existing experience, we will summarize the impelling factors of help-seeking information diffusion into five dimensions: completeness, proximity, support typology, disease severity, emotion, and elaborate them in the following sections. Completeness is bound up with credibility. Credibility has always been seen as perceived quality and as a result of intertwined dimensions (Fogg & Tseng, 1999; Fogg et al., 2000) . Previous studies usually concentrated on source credibility and stressed the significant influence of source credibility perception on retweeting behavior in sports news (Boehmer & Tandoc, 2015) and health information diffusion (Lee & Sundar, 2013) , the importance of content credibility was overlooked to some extent. In brief, content with high credibility should be internally consistent and clearly presented. Many scholars have elucidated the connotation of content credibility, including complete, in-depth, precise, reliable, accurate, unbiased, objective, factual, and fair (Sundar, 1999; Bucy, 2003; Hilligoss & Rieh, 2008) . In a systematic literature review, Sbaffi and Rowley (2017) listed dozens of content features influencing trust judgments and credibility perception, which means credibility is a typical compound concept made up of various subdimensions (Self, 1996) . As a result, it is impossible to investigate all elements of credibility in one study. The proper approach is to pick out the critical factor based on the specific research context. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 Completeness is a pivotal part of content credibility, especially in online help-seeking during COVID-19. As the name suggests, it means how much information does an individual disclose and the clarity of the provided information (Stvilia, Mon, & Yi, 2009) . Completeness contributes to the perception of health information quality on the Internet (Bates, Romina, Ahmed, & Hopson, 2006) . Firstly, more complete information means more additional information, which increases the "real world feel" (Freeman & Spyridakis, 2004) . More detailed information implies more communication cues, improves the social presence of communicators in nonverbally environments (Li, Feng, Li, & Tan, 2015) . Secondly, completeness is directly bound up with the amount and depth of self-disclosure. Altman and Taylor (1973) argued that breadth is the amount of disclosed information, while depth is the intimacy of disclosed information. A complete self-disclosure can effectively convey individuals' needs and reveal who they are, also benefits the likelihood of receiving social support (Huang, 2016) . In this study, we extrapolate that online help seekers' willingness to divulge their personal information helps the audience comprehend their identities. Besides, a complete expression of disease development or health condition usually demonstrates the seriousness of the current problem, improves credibility perception, and eventually leads to successful support provision. We propose the following: All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 The construal level theory (CLT) proposed that people live in a variety of intangible "distances," such as spatial-temporal distance, social distance, psychological distance, and so on (Trope & Liberman, 2010; Kwon, Chadha, & Pellizzaro, 2017) . With the extension of distance, people's cognition (that is, the level of interpretation) of related events and characters becomes more abstract and general, which means a high level of interpretation. On the contrary, people tend to have a more specific and concrete conception of things at a close range, namely the low level of interpretation (Fiedler, Semin, Finkenauer, & Berkel, 1995; Trope & Liberman, 2003; Wang et al., 2019; Trope & Liberman, 2010) . This kind of "distance" was interpreted as proximity by many researchers (Trope & Liberman, 2010; Kwon et al., 2017) , consists of social proximity, geographical proximity, etc. The impacts of proximity are reflected in two aspects. On the one hand, proximity plays an irreplaceable role in affecting an individual's empathy and compassion toward others' misfortune. Generally speaking, high proximity increases empathy towards the victims, while low proximity weakens it (Lenstein & Small, 2007) . Numerous studies highlighted the role of psychological distance in eliciting empathy towards intimates (e.g., immediate family members) or strangers, indicating that sympathy and compassion for others are highly associated with psychological proximity (Barnett, Tetreault, Esper, & Bristow, 1986; Batson, Lishner, Cook, & Sawyer, 2005; Loewenstein & Small, 2007) . On the other hand, proximity also affects individual's information processing, such as perceived information trustworthiness (Lee & Sundar, 2013; Shen et al., 2020) , information interpretation (Nan, 2007) , moral judgment (Žeželj & Jokić, 2014) , moral evaluation (Agerström, Björklund, & Carlsson, 2013) , and information sharing (Huang, Starbird, Orand, Stanek, & Pedersen, 2015) . Specifically, a study led by Lee and Sundar (2013) argued that proximal source boosts perceived content credibility significantly than All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint 9 the distal source, which in turn promotes information sharing (Ha & Ahn, 2011), even information adoption (Rabjohn, Cheung, & Lee, 2008) . Huang et al. (2015) also proved the evident influences of physical and emotional proximity on online information seeking and sharing under the crisis context. Based on the above rationale, in supportive communication context, empathy and compassion are closely related to proximity, which further affects information diffusion. Likewise, some studies noted that reciprocal social support behaviors frequently occur in close social proximity circumstances (Rabjohn et al., 1988 ) and a higher social presence environment, which represents psychological proximity (Li et al., 2015) . Thus, we propose the following hypothesis: H3: A high proximity between the online help seeker and the target patient triggers more retransmission than low proximity. People give support to others in multiple ways, while how different types of social support are associated with the seeking-provision remains underexplored. A considerable number of studies summarized distinct social support types and put forward some conceptual frameworks (House, 1981; Malecki & Demaray, 2003) . Some scholars conceptually divided social support into four categories, including emotional support (e.g., expression of empathy, love, trust, and caring), informational support (e.g., advice-giving, providing suggestions, sharing information), appraisal support (i.e., offering information for self-evaluation), and instrumental support (i.e., providing tangible support) (House, 1981; Malecki & Demaray, 2003; Muñoz-Laboy, Severson, Perry, & Guilamo-Ramos, 2014) . Others extended the categories of social support by incorporating All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint emotion, esteem, information, network, and tangible assistance (Cutrona & Suhr, 1992; Coulson, 2005; Coulson, Buchanan, & Aubeeluck, 2007) . Moreover, the social support inventory of UCLA (the University of California at Los Angeles) proposes a classification framework consists of three types: information or advice, tangible assistance or aid, and emotional support (Dunkel-Schetter, Feinstein, & Call, 1986) . For concision and clear, also following the previous operation (Wilson et al., 1999; Shakespeare-Finch & Obst, 2011; Federici & Skaalvik, 2014) , we merged the existing frameworks into two major types: emotional support and instrumental support. The former one typically refers to needs regarding caring, empathy, love, and trust (Wilson et al., 1999; Ko, Wang, & Xu, 2013; Federici & Skaalvik, 2014) , while the latter one denotes the requirements of instrumental resources and practical help (Wilson et al., 1999; Malecki & Demaray, 2003) . The interplay between social support types and support provision is still under-investigated, particularly the comparison of different types' outcomes has not been thoroughly examined in the online context. This study intends to remedy the defects by discussing how distinct types of social support induce different feedbacks. Therefore, we raise a research question: A patient's demands vary by the disease phase because each phase presents a new set of challenges and concomitant opportunities (Luker, Beaver, Leinster, & Owens, 1996; van der Molen, 2000) . COVID-19 is a relatively novel virus with severe clinical manifestation, and it even incurs death (Emanuel et al., 2020) . However, bounded by the shortage of testing capacity in COVID-19's early stage, not all infected patients can get diagnosed formally on time (Shen et All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint 11 al., 2020). As a result, suspected cases and confirmed cases were treated differently, receiving different medical services. Before the establishment of makeshift hospitals, patients who have mild symptoms but not laboratory-confirmed often isolated themselves at home, while those who got confirmed were admitted to hospitals in a relatively short time. Thus it is reasonable to infer that help seekers in different disease phases would get dissimilar attention. And normally, the more serious the illness, the more urgent the help-seeking and the more likely to obtain support provision. Accordingly, we posit: H4: Disease severity expressed in online help-seeking posts is positively associated with retransmission. Because different emotions usually have different effects on information processing and stimulate different behaviors, exploring emotional expression in online support seeking posts is vital to understand supportive communication in virtual spaces. Previous researchers suggested that fear is the most prominent emotion during the pandemic times, which is highly contagious and makes people feel imminent threats easily (Cole, Balcetis, & Dunning, 2012) . In comparison, other types of illnesses often associated with a plethora of emotional types range from hope to fear and humor to sadness (Mukherjee, 2010) . Lazarus (1991) and Dillard et al. (2001) summarized several primary discrete emotions in the health communication field, along with their signal values, functions, action tendencies, and valences (Shown in Table 1 ). As a matter of fact, hope and happiness rarely exist in our research corpus. Consequently, this study selects fear, anger, sadness as three main emotional types, explores the relationship between emotions and retransmission. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. (Lu & Qiu, 2013) . In today's crisis period, Chinese social media exerts the potential to mobilize collective intelligence to overcome difficulties, inspiring other countries to utilize social media to solve health-related problems and cross tough barriers. Sina Weibo, one of the most popular social media platforms in China, was selected as the sample pool. According to Weibo's user report in 2018, this social media service has accumulated more than 0.4 billion monthly active users and nearly 0.2 billion daily active users until Q4 of 2018 (Sina Weibo Data Center, 2019). Weibo has been proved as a vibrant discussion platform and help-seeking space during major social events, especially in the COVID-19 period (Huang et al., 2020; Shen et al., 2020) . All authors went through relevant help-seeking posts on Weibo and picked up pertinent keywords. After screening and merging, three pairs of keyword combinations were determined as search terms, including "pneumonia + ask for help," "pneumonia + seek help," and "pneumonia + help me." Date range starts from Jan. 20th, 2020, which is the date that Nanshan Zhong 1 confirmed human-to-human transmission (Wang & Qu, 2020) , and ends at Mar. 1st, 2020, as the closing date of the first Fang Cang makeshift hospital in the epicenter Wuhan City, indicating the epidemic had been roughly under control in China (Xu, 2020) . A web crawler written in Python programming language was applied to retrieve all qualified posts. 34,088 posts were collected in the first round, and 9,826 posted remained after removing duplicated and unqualified posts. Because of the large amount of data, 727 posts were sampled randomly from the whole corpus for further analysis. Every single post acts as the All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint 14 analysis unit for our study, and each post contains the user ID, post time, retweet number, content, and some other necessary attributes. The measurement scheme was built on the literature review. Emotion type is composed of fear, anger, sadness, and others. Support typology contains emotional support seeking, instrumental support seeking, and no specific kind of support seeking. Proximity was operationalized into reporting self-illness, reporting others' illness, and both. Disease severity consists of three kinds: suspected case, confirmed case, and others. Completeness is further decomposed into the completeness of individual information and disease status. Detailed meanings of those categories are listed in Table 2 . Number of followers, posting frequency and some other indicators are incorporated into analysis as control variables according to existing research (Liu et al., 2012; Zhang, Peng, Zhang, Wang, & Zhu, 2014) . (Shen et al., 2020) . All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. irrevocable failure to meet the goal (Dillard & Nabi, 2006) . Other emotions except for fear, anger, and sadness. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint Coding A pilot study with 100 posts was conducted to test the coding scheme and train the coders. The pilot study validated the effectiveness of the proposed coding scheme and the accuracy of the corresponding classifications. Three coders major in communication studies were recruited to code the posts for inter-coder reliability evaluation. The average Krippendorff's alpha coefficient was 0.753 in the first round, which is slightly below the acceptable level. We then retrained the coders, resolved all discrepancies, and sampled another 114 posts from the corpus randomly for a new round trial coding. The average reliability coefficient was 0.873 in the second round, which means highly consistent among the coders. All coders then performed coding work on the remaining posts independently. Since the dependent variable is the number of retweets, traditional linear regression models are inadequate for modeling this kind of highly skewed count variable. Four count models: Poisson regression, negative binomial regression, zero-inflated Poisson regression, and zero-inflated negative binomial regression were compared to fit the data. Firstly, the conditional variance of the outcome variable far exceeds the conditional mean on most categorical explanatory variables, which violates the underlying assumption of Poisson regression. Secondly, we compared the negative binomial regression with the zero-inflated negative binomial regression. Figure 1 displays the residuals from the two tested models, and small residual distribution indicates the good-fitting of the corresponding model. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint Table 3 shows the descriptive statistical results of all variables. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. Figure 2 summarized the results of negative binomial regression models. Model A contains all control variables, while Model B contains both control variables and principal independent variables. All coefficients are evaluated based on robust standard errors. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 With regard to the completeness, for each one-unit increase in disclosing individual information, the expected log count of the number of retweets increased by 0.588 (p < .001, IRR 2 = 1.800). However, the correlation between completeness of disease status and the number of retweets is insignificant. Thus, H1 was supported but H2 was rejected. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10. 1101 /2020 When it comes to proximity, reporting self-illness (B = 2.408, p < .01, IRR = 11.116) and both self and others' illnesses (B = 3.037, p < .001, IRR = 20.840) are expected to receive more retweets than only reporting others' illnesses. H3 was supported. For support seeking type, the expected log retweet count of instrumental support seeking is significantly more than emotional support seeking (B = 1.929, p < .001, IRR = 6.885), but no specific support seeking is statistically insignificant. RQ1 was answered. H4 discusses the relationship between disease severity and retransmission. Compared with other illness stages, posts about confirmed cases receive more attention (B = 2.079, p < .001, IRR = 7.997), suspected cases afterwards (B = 2.053, p < .001, IRR = 7.792). H4 was supported. RQ2 asks how emotional types affect help-seeking diffusion. The expected log retweet count for fear is lower than the expected log count for anger (B = -1.911, p < .01, IRR = 0.148), followed by others (B = -2.053, p < .01, IRR = 0.128) and sadness (B = -2.343, p < .001, IRR = 0.096). Borrowing former scholars' experience (Wang et al., 2019) , we further conducted a sensitivity test using the MANOVA analysis for cross-validation. The robustness check results (both coefficient size and significance) are close to the negative binomial regression, proving the accuracy of our estimation. First of all, the findings of the completeness part are in concert with one recent research. found that support-seeking posts with peripheral self-disclosure elicit lower perceived anonymity in the support-provider side, thus increasing the trustworthiness of the supportseeking messages and improving the quality of advice. Peripheral self-disclosure denotes All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint 21 biographical data, including name, age, gender, geographical information , which is precisely the "completeness of individual information" defined in our study. From the perspective of uncertainty reduction theory, message without clear source identity is more likely to be perceived as low credibility, further impeding support-providers' engagement in supportive communication (Berger, 1987; Rains, 2007) . In other words, detailed personal information disclosure can effectively diminish perceived anonymity and demonstrate the vulnerability of help-seekers, which is pivotal in text-based online anonymous settings. On the contrary, the completeness of disease status fails to trigger retweet behavior. One possible explanation could be the high threshold of medical knowledge posed a high demand for laypersons, resulting in an invisible "communication gap" (Filho et al., 2020) . The diagnostic report, lung X-ray photo, or medical record requires expertise to understand, which seems impossible to most Weibo users. However, this does not mean that the description of the disease is not important. To improve the communication effect, symptoms and development of the illness should be described in detail through simple words to make it easily understood by ordinary people. This rule could be verified by the positive correlation between the text length and the retransmission (Liu, Shi, Chen, Wu, & Qi, 2014) . The impact of proximity on information transmission is consistent with the previous studies (Huang, Starbird, Orand, Stanek, & Pedersen, 2015; Lee and Sundar, 2013) , which demonstrates that a high level of proximity triggers more retransmission. Proximity was operationalized into reporting self-illness, reporting others' illness, and both in our study. On the one hand, proximity can be interpreted as a type of constructed imaginary relationship with "others," or a perceived connection between individuals (Lenstein & Small, 2007) . Imaginary relationship plays a vital role in eliciting emotional or informational responses to suffering. Reporting self-illness, All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 reporting others' illness, and reporting both represent different levels of perceived psychological distance, lead to varying amounts of information retransmission. On the other hand, the rhetoric school emphasizes the effect of "narrative distance" on audiences' emotional involvement (Andringa, 1996) . Specifically, narrative perspectives (first-person versus third-person) can influence victim blame and supporting intention by affecting the perceived psychological distance (Cao & Decker, 2015) . Follow this thread, reporting self-illness and reporting others' illness can be separately associated with first-person and third-person narrative perspectives, stimulating disparate psychological distances and leads to different responses toward the patients eventually. Although social network services generally offer substantial opportunities for social support transactions, their potential to provide emotional and instrumental support may differ (Trepte, Dienlin, & Reinecke, 2015) . In this study, more than half of posts contain instrumental support seeking intention, far exceed emotional support seeking, and no specific kind of support seeking. Instrumental support seeking posts receive more retransmission. This result suggests that in the health communication field, especially during a severe pandemic, it is inevitable to face the explosive growth of information. Under this circumstance, attention becomes a scarce but valuable resource (World Health Organization, 2018) . Compared with nihilistic emotional support seeking, instrumental support seeking focused on improving one's health condition, demonstrating direct material needs, and showing the willingness for immediate help. Besides, based on Vitak and Ellison's work (2013) , many people are reluctant to express emotional needs online because they do not want to appear "needy." Emotional support also articulates with intimacy, which is a prerequisite for emotional communication between interactive partners (Stokers, 1983) . The anonymity feature of online communication hinders help seekers' desire for All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 emotional support, impedes the occurrence of in-depth emotional communication (Newman, Lauterbach, Munson, Resnick, & Morris, 2011) . Regarding severity, extant studies revealed that giving priority to the worst-case when allocating health resources in a pandemic is a fundamental principle (Emanuel & Wertheimer, 2006; Persad, Wertheimer, & Emanuel, 2009; Rosenbaum et al., 2011; Biddison et al., 2014; Zucker et al., 2015) . This study proves that how to treat patients is in accordance with this principle. The severity of illness positively correlates with retransmission. One reason is that confirmed patients need timely medical care than other patients. Senses of urgency and scarce medical resources drive other social media users to participate in the retweeting process to help confirmed cases find adequate medicare in time. When it comes to emotion, posts expressing anger received more retransmission than fear, sadness, and other kinds of emotions. Based on our observation, anger mostly stems from hospitals' unfair treatment or official institutions' delayed responses. It is reasonable for social media users to stand on the "just side" to support the unfortunate patients. By retweeting the angry posts, retweeters vented their feelings, intended to get more attention, and urged relevant departments to take effective measures. In our research corpus, sadness is the second common emotion. However, it received the least retransmission. This can be attributed to sadness's intrinsic characteristics: posts with sadness mainly express disappointment toward reality but resist to advocate practical attempts to change reality. This finding implies that when seeking help online, one should stress the principal problem and avoid simple catharsis. Similarly, ingeniously using prevailing emotion contributes to receiving sympathy and pragmatic feedback from others. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10. 1101 /2020 Conclusion poses enormous threats to the whole world. Due to the unbalanced distribution of medical resources and inadequate response at the early breakout stage, many patients (both suspected and confirmed), along with their relatives or friends, turned to social media for support seeking. This study explores the driving forces behind online help-seeking post transmission in a global pandemic period by emphasizing the content characteristics. Completeness, proximity, support typology, disease severity, and emotion composed the analytical framework. By employing content analysis, 727 randomly sampled help-seeking posts were analyzed based on a coding scheme derived from the literature review. Negative binomial regression reveals that posts release anger, express instrumental support seeking intention, report self-illness, expound confirmed cases' conditions, and have detailed individual information disclosure are likely to have more retransmission. Coronavirus is still spreading fast around the world. As one of the first countries to restrain the spread of the epidemic, China provides a significant experience to other countries stuck in a dilemma. As the first online help-seeking analysis research in the COVID-19 period, this study offers some insights about health communication via social media, especially how to develop a potent help-seeking post in public health emergencies. However, a couple of limitations need to be mentioned. First, retransmission of post, as the core dependent variable in our study, is a multidimensional concept. For example, Wang et al. (2019) decomposed retransmission into "scale" and "structural virality." Limited by research resources and time, we failed to elucidate information diffusion comprehensively. Second, although China's experience is representative, whether this pattern can be generalized into other contexts worth carefully examining. Future studies should conduct more explorations on other social media platforms (e.g., Twitter) to verify All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint 25 the reliability and validity of our results and summarize effective online help-seeking strategies in diversified environments. 1. Nanshan Zhong is a well-known Chinese pulmonologist, he was one of the leading experts in managing COVID-19 in China. 2. IRR is the abbreviation of Incident Rate Ratios, which means the conditional incidence rate compared with baselines. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint 29 Fiedler, K., Semin, G.R., Finkenauer, C. & Berkel, I. (1995) . Actor-Observer Bias in Close Relationships: The Role of Self-Knowledge and 21(5) , 525-538. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10. 1101 /2020 Look at yourself! Visual perspective influences moral judgment by level of mental construal Social penetration: The development of interpersonal relationships Sensitive Self-disclosures, Responses, and Social Support on Instagram: the case of# depression Effects of 'narrative distance on readers' emotional involvement and response Similarity and empathy: The experience of rape No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted The application of Internet-based sources for public health surveillance (Infoveillance): Systematic review The effect of source credibility on consumers' perceptions of the quality of health information on the Internet Similarity and nurturance: Two possible sources of empathy for strangers Communicating under uncertainty Uncertainty and information exchange in developing relationships Handbook of personal relationships: Theory, research and interventions Ethical considerations: care of the critically ill and injured during pandemics and disasters: CHEST consensus statement Why we retweet: Factors influencing intentions to share sport news on Twitter No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted Tweet, tweet, retweet: Conversational aspects of retweeting on Twitter. 43rd Hawaii International Conference on System Sciences Communication of social support in computer-mediated groups for people with disabilities When vaccines go viral: An analysis of HPV vaccine coverage on YouTube Media credibility reconsidered: Synergy effects between on-air and online news Psychological distancing: the effects of narrative perspectives and levels of access to a victim's inner world on victim blame and helping intention Statistical Report on Internet Development in China Affective signals of threat increase perceived proximity Receiving social support online: an analysis of a computer-mediated support group for individuals living with irritable bowel syndrome No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted Social support in cyberspace: a content analysis of communication within a Huntington's disease online support group An analysis of social support exchanges in online HIV/AIDS self-help groups Controllability of stressful events and satisfaction with spouse support behaviors The persuasive influence of emotion in cancer prevention and detection messages Helping behavior and altruism: An empirical and conceptual overview UCLA social support inventory. Unpublished manuscript Fair allocation of scarce medical resources in the time of Covid-19 Public health: who should get influenza vaccine when not all can? Conference on Information Systems Developing a unifying framework of credibility assessment: Construct, heuristics, and interaction in context Work stress and social support (Unpublished master's thesis) Mining the characteristics of COVID-19 patients in China: Analysis of social media posts Examining the beneficial effects of individual's self-disclosure on the social network site Connected through crisis: Emotional proximity and the spread of misinformation online Proceedings of the 18th ACM conference on computer supported cooperative work & social computing When Facebook is easier than face-to-face: Social support derived from Facebook in socially anxious individuals Understanding the different types of social support offered by audience to A-list diary-like and informative bloggers Proximity and terrorism news in social media: A construal-level theoretical approach to networked framing of terrorism in Twitter Emotion and adaptation To tweet or to retweet? That is the question for health professionals on Twitter Why do people retweet a tweet?: Altruistic, egoistic, and reciprocity motivations for retweeting Interplay of negative emotion and health self-efficacy on the use of health information and its outcomes Data mining and content analysis of the Chinese social media platform Weibo during the early COVID-19 outbreak: Retrospective observational infoveillance study Characterizing the propagation of situational information in social media during COVID-19 epidemic: A case study on Weibo How social context cues in online support-seeking influence self-disclosure in support provision A Two-Phase Model for Retweet Number Prediction When health information meets social media: Exploring virality on Sina Weibo Determinants of information retweeting in microblogging The scarecrow and the tin man: The vicissitudes of human sympathy and caring Microblogging and social change in China Information needs and sources of information for women with breast cancer: A follow-up study What type of support do they need? Investigating student adjustment as related to emotional, informational, appraisal, and instrumental support The emperor of all maladies: A biography of cancer The role of social media for collective behavior development in response to natural disasters No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted Differential impact of types of social support in the mental health of formerly incarcerated Latino men #Stupidcancer: Exploring a Typology of Social Support and the Role of Emotional Expression in a Social Distance, Framing, and Judgment: A Construal Level Perspective It's not that I don't have problems, I'm just not putting them on Facebook: Challenges and opportunities in using online social networks for health Proceedings of the ACM 2011 Conference on Computer Supported Cooperative Work What you say is what you get: how selfdisclosure in support seeking affects language use in support provision in online support forums What to say when seeking support online: A comparison among different levels of self-disclosure You get what you give: understanding reply reciprocity and social capital in online health support forums Principles for allocation of scarce medical interventions No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted Examining the perceived credibility of online opinions: information adoption in the online environment Annual Hawaii International Conference on System Sciences The impact of anonymity on perceptions of source credibility and influence in computer-mediated group communication: A test of two competing hypotheses Communicating social support in computer-mediated contexts: A meta-analytic review of content analyses examining support messages shared online among individuals coping with illness How does social capital affect retweets? Report: China' s fight against COVID-19 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity Health organizations providing and seeking social support: a Twitter-based content analysis Trust and credibility in web-based health information: A review and agenda for future research Credibility The development of the 2-way social support scale: A measure of giving and receiving emotional and instrumental support Using reports of symptoms and diagnoses on social media to predict COVID-19 case counts in mainland China: Observational infoveillance study Predicting satisfaction with social support from social network structure A model for online consumer health information quality Exploring receivers' criteria for perception of print and online news Influence of social support received in online and offline contexts on satisfaction with social support and satisfaction with life: A longitudinal study The social side of gaming: How playing online computer games creates online and offline social support Construal-Level Theory of Psychological Distance Relating information needs to the cancer experience: Themes from six cancer narratives There's a network out there you might as well tap'': Exploring the benefits of and barriers to exchanging information and support-based resources on Facebook The development of novel coronavirus infection: Human-to-human transmission confirmed and medical staff infected What makes cancer information viral on social media? The influence of gender and emotional versus instrumental support on cardiovascular reactivity in African-American adolescents The Influence of Self-Disclosure in Online Support Seeking on Quality of Received Advice Managing epidemics: key facts about major deadly diseases. Geneva: World Health Organization Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: A modelling study Closure of Wuhan's 11 makeshift hospitals reflects epidemic under control. Global Times To retweet or not to retweet: Understanding what features of cardiovascular tweets influence their retransmission Replication of experiments evaluating impact of psychological distance on moral judgment Facts or stories? How to use social media for cervical cancer prevention: A multi-method study of the effects of sender type and content type on increased message sharing All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted . . https://doi.org/10.1101 /2020 All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10.1101/2020.06.13.20130054 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted June 16, 2020. All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted June 16, 2020. . https://doi.org/10. 1101 /2020