Introduction

A public health emergency encompasses a range of sudden, extensive events that significantly affect the health of a population. These can arise from major infectious disease outbreaks, mass unexplained illnesses, or other substantial health threats (Rose et al., 2017). For instance, the 2014 Ebola outbreak in West Africa and the Zika virus epidemic in the Americas are stark examples of how quickly such events can escalate and demand an urgent public health response (Haffajee et al., 2014; Paterson et al., 2018). Such emergencies can have profound effects on patients, as traditional medical care channels are often disrupted due to social distancing measures and isolation policies aimed at controlling the spread of diseases (Anderson et al., 2020; Chinazzi et al., 2020). Natural disasters can further compound this disruption by damaging local health services (Sever et al., 2018). The rapid onset of these emergencies often leaves patients with unanswered questions and an urgent need for health-related information (Beliga et al., 2021; Fleming, 2020; Zheng et al., 2022). Indeed, surveys estimate that over 60% of respondents spent more time searching for health information during the pandemic than before (Fleming, 2020; Mander, 2020).

During public health emergencies, patients not only seek information but may also experience significant psychological and emotional distress. This distress can be intensified by the dissemination of misinformation, which heightens anxiety and fosters negative emotions (Karami et al., 2020; Qiu et al., 2020; Zheng et al., 2022). Additionally, prolonged lockdowns, social isolation, and the psychological impact of traumatic events, such as natural disasters, further contribute to adverse mental health outcomes (Whitaker et al., 2021).

In response to these challenges, patients increasingly seek social support—both informational and emotional—through Online Health Communities (OHCs) (Petrič et al., 2023; Schaefer et al., 1981). OHCs allow patients to connect with medical experts, overcoming geographical barriers and obtaining vital information as a substitute for face-to-face consultations (Sims, 2018; Yang et al., 2020). Additionally, OHCs serve as platforms for patients to access accurate health information, reducing confusion and alleviating anxiety during public health emergencies (Tang and Yang, 2010; Yan et al., 2016). By sharing personal experiences, patients can also find emotional support, easing feelings of isolation and contributing to positive psychological outcomes (Frost and Massagli, 2009; Harmon et al., 2021; Yan et al., 2016).

The specific processes by which individuals interpret emergency-related information and take personal protective actions—such as seeking social support through OHCs—are not yet fully understood. This knowledge gap leads to our primary research question: What factors influence the protective behaviors of patients using OHCs during public health emergencies? Understanding these mechanisms is crucial, as it not only deepens our understanding of individual responses to public health emergencies but also guides the strategic use of OHCs to improve public preparedness and resilience.

Moreover, patients’ protective actions may be shaped by the actions of related stakeholders and the characteristics of the public health emergencies themselves. Stakeholders’ expertise and trustworthiness, especially in a digital environment, play a crucial role in influencing patients’ decisions (Arlikatti et al., 2007; Wang et al., 2018a). Although previous studies have examined the role of traditional local stakeholders in protective action decisions, these studies often overlook the digital context. Given the significant role of doctors and health departments in providing expertise and trustworthiness, this study focuses on doctors’ knowledge-sharing and volunteer medical services provided through OHCs during public health emergencies (Wang et al., 2018a).

Additionally, the moderating role of different types of public health emergencies has been underexplored. Unlike natural disasters such as hurricanes and earthquakes, which can be clearly predicted, tracked and prepared in advance, Black Swan and Gray Rhino events are more closely tied to human systems and require effective risk management and crisis response. These events are likely to benefit from in-depth research to improve response strategies and decision-making (Taleb, 2007; Wucker, 2016). Black Swan events, characterized by their extreme rarity, unpredictability, and profound impact, include examples such as the Ebola Virus Outbreak and the Anthrax Attacks. In contrast, Gray Rhino events are highly probable and impactful but often overlooked despite clear warning signs, such as health issues from air pollution or antibiotic resistance.

Given the significant psychological impact of public health emergencies, which can instill fear of unknown risks (Fischhoff et al., 2016; Lindell and Perry, 2012), this study proposes a second research question: How do the nature of public health emergencies and the actions of stakeholders in OHCs moderate patients’ protective actions? Addressing this question is essential not only for immediate crisis management but also for fostering risk awareness among the public and decision-makers. By adopting these comprehensive approaches, we can better prepare for and mitigate the impacts of Gray Rhino and Black Swan events.

Using the Protective Action Decision Model (PADM), which explains responses to threatening events (Lindell and Perry, 2012), this study constructs a model of individual online protective actions under public health emergency conditions. We empirically analyze data from social media and official sources, including environmental and social cues, as well as consultation dialogues from OHCs reflecting patients’ social support and the actions of related doctors. This study extends the PADM framework to the digital health context, exploring how individual protective actions are shaped during serious public health emergencies and categorizing these emergencies as Gray Rhino or Black Swan events to examine heterogeneous outcomes.

The rest of this paper is organized as follows: The next section discusses the literature on the PADM framework and social support in OHCs, followed by hypothesis development. Subsequent sections detail the experimental data, data analysis, and discussion of the results. Finally, the paper concludes with implications and potential future directions.

Literature review

Classification of public health emergencies: black swans and grey rhinos

The classification of public health emergencies is crucial for understanding and effectively responding to various crises. In recent years, scholars have increasingly categorized these emergencies into two distinct types: Black Swan events and Grey Rhino events (Devarajan et al., 2021; Halliburton, 2020; Lüscher, 2020; Yarovaya et al., 2022). This classification helps to explore the similarities and differences in the factors that influence public behavior in the face of such crises, particularly in the digital health environment.

Black Swan events are characterized by their extreme rarity, unpredictability, and profound impact (Taleb, 2007). These events are entirely unforeseen and unprecedented, making it difficult for individuals and institutions to anticipate or prepare for them in advance. The Ebola Virus Outbreak in West Africa (2014–2016) and the Anthrax Attacks in 2001 serve as prime examples of Black Swan events. Such occurrences typically catch the public and authorities off guard, as they do not consider these possibilities until they actually happen, leading to significant challenges in crisis management and response (Aven, 2013).

The COVID-19 pandemic is a recent and notable example of a Black Swan event. Its emergence was marked by the extreme rarity and far-reaching impact characteristic of Black Swans. The pandemic has resulted in nearly 800 million confirmed cases and over 7 million deaths globally, highlighting its profound influence on public health systems and broader societal structures (WHO, 2023). The rapid mutation of the virus, coupled with its ability to evade immune responses, underscores the unpredictable and extraordinary nature of such events (X. Tang et al., 2020).

In contrast, Grey Rhino events are highly probable and high-impact crises that are often overlooked despite clear warning signs (Wucker, 2016). Unlike Black Swan events, Grey Rhino events are foreseeable and proceed with a series of early warnings and identifiable indicators. However, due to a combination of familiarity with the risk and underestimation of its severity, adequate preventive measures are frequently not taken, leading to significant consequences.

Examples of Grey Rhino events include public health issues arising from air pollution and antibiotic resistance, both of which have long been recognized as significant risks but often receive inadequate attention. Floods caused by seasonal weather patterns or river overflows also exemplify Grey Rhino events. Despite accurate weather forecasts and warnings, insufficient preparedness and the absence of precise real-time alerts can exacerbate the impact of these events (Cremen and Galasso, 2020; Rappaport et al., 2009). The predictability of these events, coupled with the neglect of countermeasures, leads to considerable losses, fitting the definition of Grey Rhino events.

The classification of public health emergencies into Black Swan and Grey Rhino events provides a valuable framework for understanding the different types of crises and the challenges they pose. While Black Swan events are rare and unpredictable, requiring flexible and adaptive responses, Grey Rhino events are foreseeable and demand proactive and preventive strategies. Recognizing the nature of these events can help improve preparedness and response efforts, ultimately reducing the impact of future public health emergencies.

Social Support in OHCs during public health emergencies

Online Health Communities (OHCs) have become an essential platform in modern healthcare, particularly during public health emergencies when access to traditional healthcare services may be restricted. These communities provide real-time and accessible virtual spaces for health communication, consultation, and social support, playing a critical role in supporting patients during crises (Yan et al., 2022). The positive impacts of OHCs on patients’ health-related quality of life have been well-documented, including enhancements in healthcare knowledge, decision-making, and the strengthening of doctor-patient relationships across geographical barriers (Chen et al., 2024; Jiang et al., 2022; Wang et al., 2024; Wang et al., 2020; Wu et al., 2021).

Social support within OHCs is a crucial factor in patient health outcomes (Liu et al., 2022; Liu et al., 2024), especially during public health emergencies. Social support typically includes informational support, emotional support, and, to a lesser extent, tangible support. Informational support provides patients with crucial health-related information and guidance, while emotional support offers reassurance, a sense of belonging, and the confidence needed to manage health challenges (Cutrona and Suhr, 1992; Schaefer et al., 1981; Turner and Brown, 2010). Tangible support, involving direct aid such as financial assistance, is less common in OHCs but still relevant in certain contexts.

Numerous studies have highlighted the importance of both informational and emotional support in OHCs (Schaefer et al., 1981; Yan and Tan, 2014). Informational support in these communities is widespread and, when combined with emotional support, plays a significant role in improving patient health outcomes (Petrič et al., 2023; Yan and Tan, 2014). For example, patients in OHCs often receive valuable advice on treatment methods and expert opinions, which are critical for making informed health decisions (Johnston et al., 2013; Song and Xu, 2023). At the same time, emotional support within OHCs provides patients with a sense of attachment and reassurance, which can be particularly beneficial during times of crisis (Zhou et al., 2023).

In specific cases, such as cancer-related OHCs, the exchange of both informational and emotional support is vital for meeting patients’ mental health needs (Love et al., 2012). Similarly, online communities focusing on conditions like depression have been shown to improve participants’ health by providing them with necessary social support at both informational and emotional levels (Lu et al., 2021).

During public health emergencies, social support in OHCs serves as a critical form of protective action. Patients actively seek out and utilize informational and emotional support to manage the challenges posed by the crisis. This behavior reflects a proactive approach to health management, where individuals use the resources available in OHCs to protect themselves and improve their health outcomes during emergencies. Informational support helps patients make informed decisions about protective measures, while emotional support provides the necessary psychological resilience to face the uncertainties of a public health crisis (Jiang et al., 2022).

In conclusion, social support in OHCs plays a pivotal role during public health emergencies by providing essential informational and emotional resources. This support enables patients to take protective actions, improving their ability to cope with and respond to health crises. The integration of social support mechanisms within OHCs is therefore vital for enhancing patient outcomes and strengthening public health responses during emergencies.

Protective actions based on the PADM

The Protective Action Decision Model (PADM) provides a comprehensive framework for understanding how individuals make decisions about protective actions in response to environmental hazards and disasters. Initially developed by Lindell and Perry (2003); Lindell and Perry (2012), the PADM outlines a multi-stage process that begins with the perception of environmental cues, social context, and warning components, which subsequently shape core perceptions of threats and stakeholders (Arlikatti et al., 2007; Lindell and Whitney, 2000). These perceptions form the basis for protective action decisions, leading to behavioral responses such as information search, problem-focused coping, and emotion-focused coping (Lindell and Perry, 2012). It is shown as a flow chart in Fig. 1.

Fig. 1
figure 1

PADM framework.

Full size image

The PADM framework has been widely applied in studies related to natural disasters such as floods, hurricanes, and earthquakes, as well as accidental environmental disasters like nuclear and chemical pollution (Heath et al., 2018; Lindell and Perry, 2012; Strahan et al., 2019). These studies often focus on how environmental and social cues, along with warning components, influence individuals’ protective actions (Heath et al., 2018; Huang et al., 2016; Molan and Weber, 2021; Rasmussen and Wikström, 2022; Zeng et al., 2019). For instance, in the context of natural disasters, rapid evacuation decisions are frequently analyzed through the PADM lens, highlighting the model’s utility in assessing protective behaviors (Heath et al., 2018; Lindell and Perry, 2012).

However, the application of PADM in public health emergencies has been relatively limited. While some studies have investigated behavioral responses to infectious diseases like COVID-19, they have primarily focused on traditional protective actions, such as mask-wearing and social distancing. For instance, research during the second wave of COVID-19 in China highlighted the crucial role of effective communication in shaping protective behaviors by examining the impact of warning messages and risk perception (Guo et al., 2022; Shi et al., 2021).

In the context of digital health, particularly within OHCs, PADM can be adapted to understand how social support influences protective actions during public health emergencies. OHCs provide a unique platform where patients can access real-time informational and emotional support—key components of the PADM framework. Informational support helps individuals comprehend the nature of the threat and make informed protective decisions, while emotional support builds the psychological resilience needed to cope with the stress and anxiety of such emergencies (Zhou et al., 2023).

Studies have shown that both informational and emotional support in OHCs are crucial for improving patient outcomes during crises. These supports are integral to the PADM decision-making process, influencing core perceptions of threats and guiding protective actions (Petrič et al., 2023; Yan and Tan, 2014). For example, during the COVID-19 pandemic, OHC participants benefited from timely and accurate information about protective measures, which was essential in guiding behaviors such as adhering to quarantine guidelines or seeking remote medical advice (Chen et al., 2021).

The PADM serves as a valuable framework for understanding protective actions during public health emergencies (Lindell et al., 2019; Molan and Weber, 2021). When applied to digital health contexts, particularly within OHCs, the PADM highlights the critical role of social support—both informational and emotional—in shaping individuals’ responses to crises. By integrating the insights from studies on social support in OHCs with the PADM framework, future research can better understand the complexities of protective actions in digital health environments, ultimately enhancing the effectiveness of public health interventions.

Hypothesis development

Based on the PADM, this study introduces environmental cues and social cues into the research model and measures patients’ protective actions by their social support-seeking behaviors in OHCs during public health emergencies. However, the warning components in the PADM will not be considered in this study since warnings are issued at different times and at different levels of government depending on the extent of the disaster, which makes it difficult to obtain accurate warning data. Furthermore, receiver characteristics will also not be discussed because such information is unavailable due to the privacy protection in OHCs. Meanwhile, in the context of OHCs, doctors, and platforms can be treated as stakeholders, while their actions can provide extra information to patients that can moderate their process of protective action decision-making. In this study, the COVID-19 pandemic and the 2021 flood disaster in Henan Province represent different types of public health emergency threats, a Black Swan event and a Gray Rhino event, respectively. The constructs and hypotheses will be discussed in the following sections.

Main effects of environmental cues

Environmental cues can be physical factors such as sights, sounds, and smells that directly reflect a disaster threat (Lindell et al., 2006), or they can be defined by the observed behavioral responses of others (Dynes and Quarantelli, 1973). As a result, during public health emergencies, casualties that can reflect developments or forecasts that can reflect its likelihood can be used to represent environmental cues (Huang et al., 2016; Lindell and Perry, 2003; Zheng et al., 2022). After observing the environmental cues related to a public health emergency, individuals become aware of the threat while they may be not competent enough to deal with so that they begin to collect related information to help them make decisions (Lindell and Perry, 2003). In particular, when their health is threatened, people urgently need adequate and reliable informational support to evaluate the situation, ensure their own health, reduce the stress of uncertainty associated with the hazard and respond to it (Gong et al., 2022; Whitcomb et al., 2017; Zheng et al., 2022), thus motivating their search for related information. Therefore, based on the PADM and the discussion above, this study hypothesizes:

H1a: Environmental cues are positively correlated with individuals’ seeking informational support.

Additionally, environmental cues that make individuals aware of hazards may cause a range of negative emotions and even mental health issues in individuals (Janis and Mann, 1977; Kiecolt-Glaser et al., 2002). For example, uncertainty about public health emergencies may trigger fear and anxiety (Bao et al., 2020); the unfortunate experience of casualty losses due to public health emergencies may inspire sadness (Sun et al., 2021). Seeking emotional support may be an effective way for individuals to mitigate the effects of negative emotions. As a result, individuals trigger their need for emotional support to reduce such emotional distress. Therefore, this study hypothesizes:

H1b: Environmental cues are positively correlated with individuals’ seeking emotional support.

3.2 Main effects of social cues

Social cues refer to the transmission of information about risks and protective actions as well as the provision of assistance to reduce risks and assist in protective actions, all through the social context (Lindell and Perry, 2012). Since social media can quickly spread and convey information related to public health emergencies, raise awareness, and guide the public’s attention to protective action (Karami et al., 2020; Panagiotopoulos et al., 2016), social media information is used as social cues in the study, and it has also been widely used in prior studies (Heath et al., 2018; Strahan and Watson, 2019; Strahan et al., 2019). However, the rapid dissemination of social cues in social media in the context of public health emergencies is accompanied by the rapid spread of a large amount of misleading and unreliable information, especially health-related online fake news (Zheng et al., 2022). For most individuals, it is difficult for them to distinguish between fake news and real information, and they may show strong concerns about their health status due to such ambiguous or erroneous information, which prompts them to seek informational support on the Internet to learn how to correctly cope with public health emergencies (Freiling et al., 2023; Zheng et al., 2021). Consequently, social cues that may include conflicting information (Dootson et al., 2022; Lindell and Perry, 2012) and ambiguous information (Zheng et al., 2022) can promote the individual’s need for further information acquisition so they can implement individual protective action. Accordingly, this paper hypothesizes:

H2a: Social cues are positively correlated with individuals’ seeking informational support.

Furthermore, since negative emotions can spread quickly through social networks (Andreassen, 2015; Li et al., 2023), the anxiety caused by fuzzy emergency-related information in the social context, combined with worry from the negative discussion of related events in social media, will push individuals to have negative emotions and then to seek emotional support (Huerta et al., 2021; Li et al., 2023). As a result, it is hypothesized:

H2b: Social cues are positively correlated with individuals’ seeking emotional support.

3.3 Moderating effects of stakeholder actions

According to the PADM model, environmental cues and social cues can elicit stakeholder perceptions that provide the basis for individual protective action (Lindell and Perry, 2012). Stakeholders are perceived to have an obligation to provide protection actions during public health emergencies (Arlikatti et al., 2007; Terpstra and Gutteling, 2008; Wang et al., 2018a), while their expertise and trustworthiness are positively correlated with individuals’ such perceptions (Lindell and Perry, 2012; Lindell and Whitney, 2000). Stakeholder actions can be treated as complementary to an information environment in which individuals can obtain extra informational and social cues about hazards (Wang et al., 2018b). Related stakeholders that have relevant experience, provide basic information and even take actions to help individuals highlight their expertise and trustworthiness under certain emergencies (Lindell and Perry, 1992; Murphy et al., 2018; Taibah et al., 2017), thus influence individuals’ perception towards stakeholders and facilitate their protective action decisions (Gauntlett et al., 2019; Lindell and Perry, 2012; Mileti and Peek, 2000). As a result, stakeholder actions are predicted to be the moderating factor.

In OHCs, the main groups are patients, doctors, and platform administrators, while doctors and administrators can be considered as main stakeholders (Klecun et al., 2019). Based on a preliminary study on the service mechanism and operation mechanism of OHCs, doctors may share professional knowledge with users in the community out of their own sense of responsibility (Hewitt-Taylor and Bond, 2012), while during public health emergencies, such knowledge sharing behavior is to a large extent professional, including the interpretation of environmental and social cues (Murphy et al., 2018). Patients may evaluate the situation of public health emergencies based on the expertise and trustworthiness of stakeholders’ actions and further influence the implementation of protective actions (Wang et al., 2018a). Expertise refers to knowledge about a topic, while trustworthiness is the willingness to convey accurate and complete information about a topic and trust in the role of such stakeholders (Arlikatti et al., 2007; Kruglanski and Stroebe, 2005). Doctors’ knowledge-sharing provides additional interpretation of public health emergencies, reflecting the willingness to provide their patients with as much expertise as possible about public health emergencies. Such strong willingness combined with the authority of doctors increases such expertise and trustworthiness. As a result, patients raise their attention to public health emergencies and implement individual protective behaviors. Therefore, the following research hypotheses are proposed:

H3a: Doctors’ knowledge-sharing behavior positively moderates the effects of environmental cues and social cues on individual protective actions.

Additionally, at the platform level during public health emergencies, administrators will organize free diagnostic activities and encourage doctors to take social responsibility by offering free online consultations and providing help for patients. Since patients may assess the ability and benevolence of the OHC to develop their trustworthiness (Anderson and Griffith, 2022; Yu et al., 2023), free medical services provided by the OHC can show their ability to organize enough experienced doctors as well as their benevolence that willing to minimize healthy barriers for patients in need during public health emergencies. Hence, while patients understand the seriousness of public health emergencies and actively engage in individual protective actions, they also develop sustained behaviors of seeking informational and emotional support due to increased trust in the relevant OHC. Therefore, this paper hypothesizes:

H3b: Stakeholders’ emergency response behavior positively moderates the effects of environmental cues and social cues on individual protective actions.

3.4 Heterogeneity among Black Swan events and Gray Rhino events

Threat perception is likewise an important mediating process in the PADM framework, while the probability and consequences of public health emergencies are generally considered to be essential factors that can influence people’s perceptions of such threats (Lindell and Perry, 2012). Therefore, public health emergencies with varying probabilities of occurrence as well as consequences can affect threat perceptions to varying degrees, and thus moderate the implementation of individuals’ protective actions (Dash and Gladwin, 2007; Lindell and Perry, 2012).

A Black Swan event refers to an unpredictable event with a low probability but a high degree of uncertainty (Taleb, 2007). Information asymmetry is one of the characteristics of a Black Swan event, so there may exist obstacles such as a lack of relevant information, need to deal with information disorder, and the presence of conflicting information when making decisions (Galbraith, 1974; Phillips et al., 2023; Turner and Makhija, 2012). The more uncertainty and ambiguity there is, especially in the case of unpredictable Black Swan events, the greater the need to gather more information (Burns and Wholey, 1993; Dahlmann and Roehrich, 2019). For instance, in the case of COVID-19, which is a Black Swan event, health-related fake news online became a significant social problem due to the lack of information caused by the low-probability nature of the emergency, and this prompted individuals to search for more information (Zheng et al., 2022). In addition, with the rapid development of information technology and social media, the speed of information dissemination is much higher than before, so in the context of poorly understood Black Swan events, panic is exacerbated by the speed of information dissemination (Zheng et al., 2021). In such a case, systematically documenting and understanding the relevant informed explanations can be one way to mitigate such a little-known situation (Bier and Mosleh, 1990; Kunreuther et al., 2004; Paté‐Cornell, 2012). Furthermore, a Black Swan event may cause people ongoing anxiety, as the unknown associated with such events causes them to worry about whether they may become victims in the near future (Mueller and Stewart, 2016). As a result, individuals are more likely to seek both informational and emotional help during such events.

Gray Rhino events are high-probability events that are often predicted but for which the risk is not sufficiently appreciated (Wucker, 2016). Compared with an unpredictable Black Swan event, individuals may have preexisting experience and knowledge relevant to a high-probability Gray Rhino event (Chen et al., 2021). Additionally, people tend to underestimate medium- to high-probability and possible events but value low-probability and unlikely events (de Palma et al., 2014; Fox and Tversky, 1998; Tversky and Fox, 1995). As a result, individuals may opt not to search for additional relevant information. Furthermore, the emotion of anxiety is reduced in a predictable environment, while it is caused by an unpredictable environment (Davis et al., 2016); this reduces the probable need for emotional help in a predictable environment. Therefore, this paper hypothesizes:

H4a: Environmental cues have a greater impact on individuals’ protective actions in a Black Swan event than in a Gray Rhino event.

H4b: Social cues have a greater impact on individuals’ protective actions in a Black Swan event than in a Gray Rhino event.

The theoretical research framework of this paper is shown in Fig. 2.

Fig. 2
figure 2

Theoretical framework.

Full size image

Data and empirical strategy

We selected two representative public health emergencies as study samples for Black Swan and Gray Rhino events. For the Black Swan event, we focused on the COVID-19 pandemic due to its highly contagious nature and substantial global health threat. The World Health Organization (WHO) estimates that the pandemic has resulted in at least 7 million deaths since its emergence at the end of 2019 (Burki, 2023). As a representative of Gray Rhino events, flooding was chosen for its high probability and predictability. Specifically, the massive floods caused by a rare extreme precipitation event in Henan, China, in July 2021 led to hundreds of deaths and impacted tens of millions of people. The flood also triggered contamination-related public health threats, including the spread of infectious diseases (Chen et al., 2023). This event is one of the most severe public health emergencies in China in recent years.

The data for this study were obtained through various methods. The online health consultation data were sourced from a leading Chinese online health community platform, which offers a range of services, including daily consultations to a large user base (Fan et al., 2022; Wu et al., 2020). The data collection process occurred in multiple stages:

First, observation period delineation. Regarding the observation periods for the event samples, we focused on the first wave of the COVID-19 outbreak, which, due to its unpredictability and profound impact, represents the most significant phase for this study (Taleb, 2007). The initial study date was set to January 30, 2020, following the WHO’s declaration of the outbreak as a public health emergency of international concern, marking the highest level of urgency in the WHO’s emergency response framework. The study concludes on March 1, 2020, coinciding with the resumption of schooling in China, which marks a phase of reduced immediate impact and a period of relative recovery. For the Henan flood, the extreme rainfall began on July 18, 2021, causing severe flooding, with flood control and relief efforts completed by July 31, 2021.

Second, keyword identification and data extraction. According to Panagiotopoulos et al. (2016), for the Black Swan event (COVID-19), we used the keywords “COVID-19” and “coronavirus disease” to ensure the relevance of the consultation content. Additionally, we restricted the sample to consultations from the respiratory medicine department to ensure that the data specifically related to the COVID-19 pandemic and its primary symptoms. For the Gray Rhino event (Henan flood), we used the keyword “flood” to identify relevant consultations. Moreover, we employed the IP addresses of the users to confirm that the consultations were from patients located in Henan province, further ensuring the relevance of the data.

Third, data extraction and screening. Keyword detection was performed on the platform, and web scraping software was used to extract the URLs of the webpages containing relevant data. Once the URLs were identified, the data were extracted using the scraping tool. The data were then screened by date, retaining only consultations corresponding to the specific timeframes of the public health emergencies. Figure 3 is the schematic diagram of data extraction and filtering on the website.

Fig. 3
figure 3

Schematic of data crawling options.

Full size image

After processing, the data were integrated by daily consultation dialogues, resulting in a total of 3352 valid consultations across 596 patients. The dataset corresponding to the COVID-19 pandemic spans from January 31, 2020, to March 1, 2020, with 2607 dialogues from 482 patients. The Henan flood data, from July 17, 2021, to July 31, 2021, includes 745 dialogues from 114 patients. In total, 602 dialogue sets were collected during these public health emergencies.

Moreover, environmental cues data related to COVID-19, including the cumulative number of deaths and confirmed cases from January 31 to March 1, 2020, were obtained through a Tencent epidemic data report, reflecting the most severe phase of the first outbreak in China (Xie et al., 2020). For the Henan flood, rainstorm forecast information was obtained from the provincial meteorological department for the period from July 18 to July 31, 2021. Additionally, Weibo hot search data was used as social cues for both emergency events.

Operationalization of variables

Environmental cues (Environmental_Cues), one of the independent variables of this study, are observable by individuals and denote the severity of public health emergencies. In the context of COVID-19, patients can usually get information about the pandemic’s development the next day. The ratio of the cumulative number of deaths and the cumulative number of confirmed cases on the day before the focal patient seeks social support in the OHC (e.g., an online consultation) can reflect the knowable severity of the pandemic (Onder et al., 2020). Thus, it can be used as an environmental cues. In the context of the flood in Henan Province, the rainstorm forecast information from the Henan Meteorological Bureau is used as environmental cues. This is reasonable because patients may also take the information of the previous day as a reference when judging the severity of the flood disaster. The environmental cues related to the flood are the ratio of the number of first-level rainstorms predicted and the total number of rainstorms predicted on the day before the doctor-patient consultation in the OHC.

$${Environmental}{\_}{C{ue}{s}{\_}{{{Covid}}_{t}}}=\frac{{Total\; Dead}\,{{Cases}}_{t-1}}{{Total\; Confirmed}\,{{Cases}}_{t-1}}$$
(1)
$${Environmen}{t}{\_}{{Cue}{s}{\_}{{{Flood}}_{t}}}=\frac{{Red}\,{{Alerts}}_{t-1}}{{Total}\,{{Alerts}}_{t-1}}$$
(2)

where t represents the day patients consulted the doctor in the OHC and t-1 represents the day before the doctor-patient consultation dialogue.

Social cues (Social_Cues) refer to information about public health emergencies that individuals can receive in a social context, and they are the second independent variable in this study. Weibo is one of China’s largest online social platforms where users can share information or their feelings and thoughts on specific topics (Cui and Kertész, 2021; Gong et al., 2022). According to Weibo User Development Report release in 2021, it had over 510 million monthly active users (“Weibo User Development Report,” 2021), nearly 40% of the Chinese population and its annual reading flow exceeds 2,400 billion, interactions nearly 7 billion. Weibo hot search provides a ranking of the most popular hashtags based on the volume of Weibo search topics, which reflects the public’s attention to and discussion of specific events. The number of Weibo hot search topics related to certain public health emergencies is used to measure social cues. In public health emergencies such as COVID-19 and the Henan flood disaster, Weibo can provide individuals with rich emergency-related information. The Weibo hot search topics are emerging discussion topics that attract a large number of users to participate in the discussion within a short time. The more times a hot search topic is listed, the greater its exposure opportunity and the more likely it is to be known by the public. Cui and Kertész (2021) use the number of categorized Weibo hot search topics to study the dynamics of public concern under social context during COVID-19. Therefore, the number of public health emergency-related topics listed on Weibo’s hot searches are selected to measure social cues. For the Black Swan event represented by COVID-19, the topics listed on Weibo’s hot searches of keywords “coronavirus,” “pneumonia,” “epidemic,” “confirmed case” and “suspected case” from January 31, 2020 to March 1, 2020 are collected. For the Gray Rhino event represented by the flood disaster in Henan province, the hot search data of keywords including “flood” and “rainstorm” from July 18, 2021 to July 31, 2021 are collected, while manual screening is also used to avoid missing relevant hot search topics and thus ensure the integrity of the data.

To measure the dependent variables as social support seeking in this study, it is essential to focus on the content provided by patients during their consultations with OHCs. Social support seeking can be operationalized through the identification of specific types of support that patients request or express a need for in their messages. These types include emotional support, where patients share their feelings, concerns, or stress; and informational support, where patients ask for specific information related to their health conditions or treatment options. Previous research on doctor-patient interactions has often measured social support by analyzing the frequency and nature of such requests in doctor-patient communications, using content analysis to categorize and quantify different types of support. According to our research target, we concentrated the variable measurements on the inputs from the patients during the consultation process rather than on the responses from the doctors.

Seeking informational support (Seek_for_Info) is one of the dependent variables, representing protective actions that patients perform in OHCs. The example of patients seeking informational support in the OHC is shown in Fig. 4. THU Open Chinese lexicon (THUOCL) from Tsinghua University includes a wide range of thesauri in different fields, which are manually screened in several rounds and continuously updated to include more categories while effectively ensuring the accuracy of the category collection (Han and Gupta, 2022; Han et al., 2016). Therefore, this paper uses the medical terminology thesaurus provided by THUOCL to match the text in the patient consultation, and then count the number of specialized words in the text to measure informational support seeking (Chen et al., 2020).

Fig. 4
figure 4

Timeline of selected public health emergencies.

Full size image

Seeking emotional support (Seek_for_Emo) is another dependent variable that reflects patients’ protective actions in OHCs in this study. The Baidu Sentiment Analysis API is selected to calculate the sentiment value, which is the API with more accurate emotion recognition by considering the word meaning and word order relationship with the conversation scenario comprehensively (Jiang et al., 2022; Tang et al., 2020). The results of the analysis are outputted as a positive likelihood of emotion score, a negative likelihood of emotion score, and a sentence emotion tendency judgment (Fig. 5). The results of the sentence emotion tendency judgments serve as a measure of emotional support, in which a judgment of 0 is negative, 1 is neutral, and 2 is positive. In this study, the sum of the emotion judgment scores of sentences in the text of an online consultation session is chosen as a measure of emotional support seeking (Figs. 6–8).

Fig. 5: Example of patients’ seeking informational support.
figure 5

a Detailed information regarding a patient's condition; b The communication between the physician team and the patient.

Full size image
Fig. 6
figure 6

Example of using Baidu API to analyze patients’ seeking emotional support.

Full size image
Fig. 7
figure 7

Example of doctors’ knowledge sharing.

Full size image
Fig. 8
figure 8

Example of free online consultation-related activities carried out by the OHC.

Full size image

Doctors’ knowledge sharing behavior (Know_Share) in the OHC, which can be treated as the stakeholder action in the context of the OHC, is one of the moderating variables of the study. As the main source of information in OHCs, doctors will provide knowledge sharing for the public out of personal responsibility when they are aware of the public health emergency (Wang et al., 2018a). In the OHC of the study, doctors’ personal homepages contain a section called “popular science articles,” in which doctors can share treatment information and health knowledge in popular fields. Since the popular science articles displayed on doctors’ personal homepages are presented in order of reading volume, users see the articles with the highest reading volume first while browsing, and the publication date of the articles will not affect the presentation order of the articles on the page. Therefore, the number of popular science articles published during and before a public health emergency is selected as the measure of the knowledge-sharing behavior of stakeholders.

OHCs’ emergency response behavior (Emergency_Re) is another moderating variable representing the stakeholder action in the field of digital health, and it is represented in the study by a variable indicating whether the OHC carries out activities related to free online consultation. If the platform carries out activities calling on doctors to provide platform users with free consultation opportunities or establishes a free online consultation option during public health emergencies, the variable value is set to 1 and otherwise to 0.

Furthermore, it is necessary to consider that in online consultation, when patients ask doctors questions based on their current disease status, the doctor’s reply will affect the patient’s further consultation due to the information asymmetry between patients and doctors. As a result, the number of questions answered by doctors in the text of an online consultation with the patient (Doc_Answers) is selected as the control variable. We also incorporate comprehensive physician-related and time-related control variables. Specifically, the physician-related controls included city tier, hospital tier, and physician title, while time-related controls categorized online consultation dates into weekdays and weekends. These control variables account for unexplained variations beyond the key independent variables, thereby enhancing the accuracy of the results.

Table 1 provides descriptive statistics for the variables in the paper and Table 2 provides correlations of these variables.

Table 1 Descriptive statistics of the variables.
Full size table
Table 2 Correlation coefficients.
Full size table

Empirical strategy

Based on the above discussion, a multiple regression model is constructed in this section to explore the impact of environmental and social cues on individual protective actions of social support (informational support and emotional support) sought by individuals in the OHC during public health emergencies. The dependent variable in the model is patients’ seeking social support, including emotional support and informational support, so it is necessary to establish separate regression models. The independent variables are environmental cues and social cues related to public health emergencies. The control variable is the number of questions answered by doctors.

The specific form of the model is shown in Equation (3) and Equation (4):

$$\begin{array}{l}{{\mathrm{ln}}\left({Seek}{{\_}}{for}{{\_}}{Info}\right)}_{{it}}={\beta }_{o}+{\beta }_{1}{{\rm{Social}}{\_}{\rm{Cue}}s}_{t}+{\beta }_{2}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}\\\qquad\qquad\qquad\qquad\qquad\quad +\,{\beta }_{3}{{Doc}{{\_}}{Answers}}_{{it}}+{\varepsilon }_{0}\end{array}$$
(3)
$$\begin{array}{lll}{\mathrm{ln}\left({Seek}{{\_}}{for}{{\_}}{Emo}\right)}_{{it}}={\beta }_{o}+{\beta }_{1}{{\rm{Social}}{\_}{\rm{Cue}}s}_{t}+{\beta }_{2}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}\\\qquad\qquad\qquad\qquad\qquad +\,{\beta }_{3}{{Doc}{{\_}}{Answers}}_{{it}}+{\varepsilon }_{0}\end{array}$$
(4)

where \(i\) denotes the doctor and \(t\) denotes the day patients sought social support in the OHC. Parameters of interest are \({\beta }_{1}\) and \({\beta }_{2}\) in both Equation (3) and Equation (4), which measure the impact of social cues and environmental cues, respectively, on patients’ protective actions.

To further explore the moderating effect of stakeholder actions on patients’ seeking social support in the OHC, an interaction term between stakeholder actions and environmental cues and an interaction term between stakeholder actions and social cues are added to extend the regression model:

$$\begin{array}{ll}{\mathrm{ln}\left({Seek}{{\_}}{for}{{\_}}{Info}\right)}_{{it}}={\beta }_{o}+{\beta }_{1}{{\rm{Social}}{\_}{\rm{Cue}}s}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{2}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\beta }_{3}{{\rm{Konw}}{\_}{\rm{Shar}}e}_{i}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{4}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}+{\beta }_{5}{D{oc}{{\_}}{Answers}}_{{it}}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{6}{{\rm{Konw}}{\_}{\rm{Shar}}e}_{i}\times {{\rm{Social}}{\_}{\rm{Cue}}s}_{t}+{\beta }_{7}{{\rm{Konw}}\_{\rm{Shar}}e}_{i}\\\qquad\qquad\qquad\qquad\qquad\times \,{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\beta }_{8}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}\\\qquad\qquad\qquad\qquad\qquad\times \,{{Social}{{\_}}{Cues}}_{t} +{\beta }_{9}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}\\\qquad\qquad\qquad\qquad\qquad\times \,{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\varepsilon }_{0}\end{array}$$
(5)
$$\begin{array}{l}{{\mathrm{ln}}\left({Seek}{{\_}}{for}{{\_}}{Info}\right)}_{{it}}= {\beta }_{o}+{\beta }_{1}{{\rm{Social}}{\_}{\rm{Cue}}s}_{t}\\\qquad\qquad\qquad\qquad\qquad\,\,+\,{\beta }_{2}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\beta }_{3}{{\rm{Konw}}{\_}{\rm{Shar}}e}_{i}\\ \qquad\qquad\qquad\qquad\qquad\,\,+\,{\beta }_{4}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}+{\beta }_{5}{D{oc}{{\_}}{Answers}}_{{it}}\\\qquad\qquad\qquad\qquad\qquad\,\,+\,{\beta }_{6}{{\rm{Konw}}{\_}{\rm{Shar}}e}_{i}\times {{\rm{Social}}{\_}{\rm{Cue}}s}_{t}\\ \qquad\qquad\qquad\qquad\qquad\,\,+\,{\beta }_{7}{{\rm{Konw}}{\_}{\rm{Shar}}e}_{i}\times {{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}\\\qquad\qquad\qquad\qquad\qquad\,\,+\,{\beta }_{8}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}\times {{Social}{{\_}}{Cues}}_{t}\\\qquad\qquad\qquad\qquad\qquad\,\,+\,{\beta }_{9}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}\times {{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\varepsilon }_{0}\end{array}$$
(6)

Parameters of interest are \({\beta }_{6}\), \({\beta }_{7}\), \({\beta }_{8}\) and \({\beta }_{9}\) in both Eqs. (5) and (6), which measure the moderating impact of doctors’ knowledge-sharing behavior and stakeholders’ emergency response behavior.

Results

Table 3 presents the OLS regression results for seeking protective actions in public health emergencies, combining the Black Swan event and the Gray Rhino event and showing both main effects and moderating effects. The independent variable in Model (1) is emotional support seeking, and the independent variable in Model (2) is informational support seeking. In Model (3) and Model (4), the moderating effect of stakeholder actions on the main effect is further explored. Due to space constraints, specific results for the control variables are provided in the Appendix. The results show that both physician-related and time-related control variables significantly influence the analysis, emphasizing the role of individual heterogeneity (e.g., physician characteristics) and temporal factors in shaping the study outcomes.

Table 3 Main effects and moderating effect of environmental cues and social cues on protective actions in the OHC during public health emergencies.
Full size table

The results indicate that environmental cues have no significant effect on patients’ seeking informational and emotional support (p > 0.1), so H1a and H1b are not valid. There is also no significant correlation between social cues and patients’ seeking social support, so both hypotheses H2a and H2b are also not valid. In considering the moderating effects, the platform’s emergency response behavior can positively moderate the relationship between environmental cues and patients’ seeking both informational support and emotional support (0.779 at p < 0.01, 0.562 at p < 0.01); thus, H3b is partly supported. However, no other moderating effects of doctors’ knowledge-sharing behavior are found in the correlation between social cues or environmental cues and patients’ seeking social support in the OHC; thus, H3a is not supported.

It was observed that the main effects (social cues and environmental cues) on seeking social support were not significant. However, as mentioned earlier, the differing nature of Black Swan and Gray Rhino events may lead to differentiated effects. Specifically, the nature of this effect stems from the structural differences between the Black Swan and Gray Rhino event groups, which may cause the overall regression coefficients to become statistically insignificant or even change direction. This phenomenon is known as Simpson’s Paradox (Simpson, 1951). To verify this, we introduced an interaction term between social cues and a dummy variable for event type (Flood = 0, COVID-19 = 1), denoted as Event Type, into the model.

$$\begin{array}{ll}{\mathrm{ln}\left({Seek}{{\_}}{for}{{\_}}{Emo}\right)}_{{it}}={\beta }_{o}+{\beta }_{1}{{\rm{Social}}{\_}{\rm{Cue}}s}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{2}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\beta }_{3}{{Event}{{\_}}{Type}}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{4}{{Social}{{\_}}{Cues}}_{t}\times {{Event\; Type}}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{5}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{{Event}{{\_}}{Type}}_{t}+{Controls}+{\varepsilon }_{0}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{9}{{Emergency}{{\_}}{\mathrm{Re}}}_{i}\times {{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\varepsilon }_{0}\end{array}$$
(7)
$$\begin{array}{l}{\mathrm{ln}\left({Seek}{{\_}}{for}{{\_}}{Info}\right)}_{{it}}={\beta }_{o}+{\beta }_{1}{{\rm{Social}}{\_}{\rm{Cue}}s}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{2}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}+{\beta }_{3}{{Event}{{\_}}{Type}}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{4}{{Social}{{\_}}{Cues}}_{t}\times {{Event\; Type}}_{t}\\\qquad\qquad\qquad\qquad\qquad+\,{\beta }_{5}{{Environmental}{{\_}}{Cues}}_{\left(t-1\right)}\\\qquad\qquad\qquad\qquad\qquad+\,{{Event}{{\_}}{Type}}_{t}+{Controls}+{\varepsilon }_{0}\end{array}$$
(8)

Table 4 presents several significant interaction effects, including the impact of environmental cues on seeking emotional support under different event contexts (3.551 at p < 0.01), as well as the influence of both social and environmental cues on seeking informational support across various event backgrounds. Such results indicate that the impact of social and environment cues on seeking social support differs significantly between Black Swan and Gray Rhino events, thereby validating the presence of Simpson’s Paradox. This finding underscores the importance of considering contextual heterogeneity in behavioral responses to public emergencies.

Table 4 Simpson Paradox validation.
Full size table

To further examine the heterogeneity between Black Swan and Gray Rhino events as two distinct types of public health emergencies, we selected COVID-19 and the Henan flood disaster as the backgrounds for regression analysis. Since the data analysis revealed that no free online consultation was offered on the OHC platform during the flood disaster, the variable related to platform emergency response behavior was not included in the data analysis for the Gray Rhino event. The main effects of COVID-19 and the flood are presented separately in Tables 5 and 6.

Table 5 Heterogeneity analysis for COVID-19.
Full size table
Table 6 Heterogeneity analysis for Henan Province flood disaster.
Full size table

Table 5 shows that in the context of the Black Swan event, environmental cues are positively correlated with both emotional support and informational support sought by patients in the OHC (3.528 at p < 0.01, 4.546 at p < 0.01), which supports H1a and H1b. Moreover, social cues are positively correlated with patients’ seeking informational support in the OHC (1.507 at p < 0.05), while no significant positive correlation is displayed between social cues and patients’ seeking emotional support, thus only supporting H2a.

Table 6 shows that there is only a significant positive correlation between social cues and patients’ seeking informational support (0.499 at p < 0.05). As a result, H2a is partly supported, while the remaining hypotheses are not supported in the Gray Rhino event.

Comparing the results in Table 5 and Table 6, the coefficient representing the impact of environmental cues on both informational and emotional support sought by patients in the Black Swan event is significantly more positive than in the Gray Rhino event. This suggests that environmental cues have a stronger influence on individuals’ protective behaviors during Black Swan events compared to Gray Rhino events, which supports H4a. Furthermore, the impact of social cues on the informational support sought by patients in the Black Swan event is significantly more positive than in the Gray Rhino event, providing partial support for H4b. Moreover, Table 5 indicates that in the context of the Black Swan event, doctors’ knowledge-sharing behavior positively moderates the relationship between social cues and patients’ seeking emotional and informational support (5.701 at p < 0.1, 7.856 at p < 0.05). However, Table 6 shows no such significant moderating effect during the flood disaster, highlighting the heterogeneity of the effects across different events.

Robustness tests

The level of informational and emotional support in a patient’s text may be influenced by the physician’s responses throughout the online consultation. To address this potential bias, a robustness test was conducted using only the patient’s initial message in the OHC to assess the extent to which they sought emotional and informational support. The results (shown in Tables 7, 8, and 9) were largely consistent with the main findings, confirming the reliability of the primary analysis.

Table 7 Results based on the patient’s initial text in OHCs during public health emergencies.
Full size table
Table 8 Results based on the patient’s initial text in OHCs during COVID-19.
Full size table
Table 9 Results based on the patient’s initial text in OHCs during the Henan Province flood disaster.
Full size table

Moreover, to ensure the robustness of our findings, we extended the data collection period for the Black Swan event to a broader timeframe, from January 23, 2020 (when Wuhan announced the city lockdown) to March 28, 2020 (when China imposed entry restrictions for foreign nationals), resulting in a total of 4555 patient online medical consultation records. Using the same model, we reanalyzed the data for this extended period, and the results remained consistent with the main findings, further reinforcing the validity of our conclusions (see Table 10).

Table 10 Main results with the extended dataset.
Full size table

Discussion

Principal findings

This study examines the influencing factors of individual protective actions in a scenario where patients are seeking social support through an OHC during public health emergencies. In view of the different characteristics of two public health emergency types, the Black Swan event and the Gray Rhino event, this paper analyzes the heterogeneity of related hypotheses. Our empirical model shows that social and environmental cues can to some extent influence patients’ seeking informational and social support in the OHC and that heterogeneity exists in public health emergencies of different natures.

First, the study shows that social cues can have a certain influence on patients’ seeking informational support. Although no significant relationship is found between social cues and patients’ seeking social support in general during these two types of public health emergencies, heterogeneity analysis shows that social cues promote individual protective actions of seeking informational support within the OHC in both the Black Swan event and the Gray Rhino event. To some extent, this finding verifies our hypothesis that patients have a need for more relevant information after receiving social cues, so they need to exchange more information for an online medical consultation to be satisfying. In addition, heterogeneity analysis finds that social cues have a greater positive effect on informational support seeking in the Black Swan event than in the Gray Rhino event. This effect is similar to findings in previous research showing that people tend to underestimate medium- to high-probability events but value low-probability and unlikely events (de Palma et al., 2014; Fox and Tversky, 1998; Tversky and Fox, 1995); thus, they are more likely to value-seeking informational support for unpredictable low-probability events like a Black Swan event.

The second finding is that environmental cues are positively correlated with patients’ seeking social support, including both emotional support and informational support, only in the Black Swan event. In other words, environmental cues are more likely than social cues to prompt patients to seek emotional support, and environmental cues influence patients’ protective actions only in the Black Swan event and not the Gray Rhino event.

Furthermore, regardless of differences in the nature of public health emergencies, platform emergency response behavior positively moderates the impact of environmental cues on patients’ seeking social support within the OHC. However, when differentiating public health emergencies with different characteristics, only in the Black Swan event does doctors’ knowledge-sharing behavior positively moderate the impact of social cues on patients’ seeking social support within the OHC. These results reflect the fact that OHC stakeholder actions can positively regulate the impact of environmental and social cues on individual protective actions to some extent, while different OHC-related actions may have different impacts due to the nature of different public health emergency types.

Theoretical contributions

This paper makes a unique and significant contribution to the research on OHCs, the PADM and public health emergencies. First, this study facilitates an understanding of individual protective actions based within OHCs during major disasters and extends previous studies focusing on traditional protective behaviors, such as emergency evacuation in natural disasters and social distancing in infectious disease (Guo et al., 2022; Shi et al., 2021), to the online context. Especially in a case where offline channels are limited, such as in public health emergencies, which have been frequently proposed to cause a shortage of medical resources (Anderson et al., 2020; Chinazzi et al., 2020; Sever et al., 2018), this research is very necessary. Previous studies on individual protective actions during public health emergencies have basically observed the positive influence of environmental and social cues on individual protective actions (Heath et al., 2018; Huang et al., 2016), which is similar to our research conclusion. However, these studies often focus on protective actions that directly respond to or help individuals avoid related events, such as evacuation during natural disasters (Huang et al., 2016) or social distancing and wearing masks during COVID-19 (Shi et al., 2021). In other words, extant research generally overlooks the informational and emotional support that OHCs may provide to patients during public health emergencies. Therefore, this study is necessary to expand the understanding and applications of OHCs in public health emergencies.

Second, this study extends the PADM to the field of digital health, especially OHCs. Considering the importance of informational support and emotional support, which are two main forms of social support widely studied in OHCs, in improving the health of OHC patients (Johnston et al., 2013; Yan and Tan, 2014), the individual protective actions in the original PADM are replaced by patients’ seeking social support in OHCs. The empirical results show a significant correlation between environmental cues as well as social cues and seeking social support, including informational support and emotional support. These findings validate the effectiveness of replacing individual protective action with seeking social support in the context of OHCs. In addition, since the actions of stakeholders can be regarded as a supplement to environmental and social cues (Wang et al., 2018b), the moderating effects of stakeholders are considered. The results indicate that both knowledge sharing from doctors and emergency responses from the platform produce effective moderating effects, thus complementing the basic PADM. Meanwhile, the empirical method is used to measure variables of PADM in the study to make up for the deviation that usually arises from subjective, self-reported data.

Third, this study compares the Black Swan event and the Gray Rhino event in the same analyses, identifying their differences in individual protective actions based on their distinctive sets of characteristics: unpredictable events with low probability and high uncertainty (Black Swan) and high-probability events that are often predicted but not fully paid attention to (Gray Rhino). This empirical study verifies that the impact of both environmental cues and social cues on individual protective behavior in the Black Swan event is greater than in the Gray Rhino event, adding to the literature on the application of PADM.

Practical implications

By understanding how patients obtain social support through OHCs in public health emergencies, our study provides important insights into how relevant departments can provide useful guidance to the public in emergencies and reduce the negative impact on individuals. Considering the positive influence of social cues on patients’ seeking informational support, platforms that release social event information shoulder important social responsibilities. For example, social media platforms like Weibo that exchange social event information need to strengthen their content management and verify published content in detail so as to avoid the negative social influence caused by false information. Also, due to the correlation between environmental cues and patients’ seeking social support (including both emotional and informational support), especially during Black Swan event like COVID-19, relevant institutions should reasonably interpret the information conveyed by relevant early environmental cues and provide appropriate guidance to the public. This would to help the public to respond rationally to public health emergencies.

In addition, the research results in this paper, combined with the moderating effects of OHC stakeholder actions, are very valuable for OHC platforms. These findings can help them further improve incentive mechanisms and policies, encourage doctors to provide free online consultation services during public health emergencies, and balance the deficit of medical resources caused by public health emergencies. Such actions would relieve patients’ anxiety, offset the information asymmetry caused by public health emergencies, and attract more users for the OHC. Meanwhile, platforms can also further improve doctors’ popular science articles. This study finds that in the OHC data, some popular science articles written by doctors have more valuable information but a relatively low number of page views from users. The platform can use this knowledge to present the most relevant popular science knowledge to the public in conjunction with major social events.

Limitations and future directions

This study has several limitations. First, while the study primarily examines text consultations between healthcare providers and patients within OHCs, which constitute a significant portion of the social support interactions, it acknowledges the presence of other support modalities, including telephone consultations. The platform’s privacy policies limit the accessibility of data on these alternative methods, thereby constraining the comprehensiveness of the study’s findings. In addition, the study employs a quantitative approach to measure informational and emotional support by analyzing the frequency of specialized terms and emotional tone in consultation texts. Although this method proves effective, it suggests that a more nuanced qualitative analysis with in-depth text analysis methods could reveal deeper insights into the nature of support provided. Moreover, the study’s findings are based on two specific public health events: the Black Swan event and the Gray Rhino event. While these cases yield significant and robust comparative insights, the study recommends that future research should include a broader range of events, varying in geographical scope (e.g., regional vs. international), the temporal dimension (e.g., short-term events vs. long-term crises), and cultural context, to enhance the model’s applicability.

Conclusions

In conclusion, this study provides new insights into how patients seek social support through OHCs as a form of individual protective action during public health emergencies. Based on the PADM’s attention to the impact of environmental cues and social cues on individual protective actions, this paper investigates the mediating effects of doctors’ knowledge-sharing behavior and platform emergency response behavior in the OHC. In addition, according to their characteristics, this paper identifies two types of emergencies, the Black Swan event, represented by COVID-19, and the Gray Rhino event, represented by the 2021 flood disaster in Henan Province. This study finds that different mechanisms of public health emergencies have different effects on patients’ seeking social support in online consultations on OHCs. The main difference is that a Black Swan event, such as COVID-19, will have a stronger effect on patients’ seeking informational support in the consultation. Meanwhile, a Black Swan event will also prompt obvious emotional support seeking, which is not seen in a Gray Rhino event like a flood disaster. Moreover, in the Black Swan event, the knowledge-sharing behavior of doctors plays an important role in the information obtained by patients. This has major significance for the future development of OHCs, as these platforms need to assume important social responsibilities while providing medical help.