Social Media as a Catalyst for Sustainable Public Health Practices: A Structural Equation Modeling Analysis of Protective Behaviors in China During the COVID-19 Pandemic

Abstract

Despite the widespread use of crisis communication on social media during the COVID-19 pandemic, the mechanisms with which social media influence crisis-related behavioral coping have been insufficiently explored. By integrating the Social-Mediated Crisis Communication (SMCC) model with the Protective Action Decision Model (PADM), this study investigates how individuals’ protective actions are influenced by crisis-related information disseminated through social media, particularly focusing on the mediating effects of risk perception and emotional responses. This study examines two special periods: the initial outbreak (January–April 2020) and the subsequent period (May–December 2020). The results indicate that the dissemination of crisis-related information on social media platforms has significant positive associations with individuals’ risk perception, information sharing, and protective actions throughout the public health crisis. Notably, information gathering showed a positive relationship with protective actions during the initial outbreak. The study identified a chain mediation effect of emotional response and risk perception in the relationship between information dissemination and information sharing during and after the initial outbreak. Additionally, risk perception emerged as a partial mediator between information dissemination and information gathering during the subsequent period. This study enhances our understanding of the psychological mechanisms through which social media crisis communication influences collective coping responses in China, providing valuable insights for practitioners aiming to optimize crisis information dissemination strategies that promote sustainable protective behaviors on social media platforms.

1. Introduction

The COVID-19 pandemic represents an unprecedented public health crisis, distinguished by its high infectivity and potential lethality, and has emerged as a significant disruptor to global sustainability progress. Its rapid global spread has not only challenged healthcare systems but has fundamentally transformed patterns of communication, work, and daily life. Unlike conventional health communication, the pandemic demands rapid, adaptive, and wide-reaching information dissemination to mitigate immediate threats and guide public behavior amidst evolving circumstances [1,2]. The unprecedented uncertainty surrounding COVID-19’s duration and impact has led individuals in isolation to significantly reduce their exposure to traditional media and face-to-face communication, instead turning to social media platforms for information and social interaction [3]. Consequently, social media platforms have become crucial channels for public information dissemination and exchange during crises.

The social-mediated crisis communication (SMCC) model emphasizes the role of social media messages and user emotions in shaping public responses to crisis events [4,5]. As a long-standing focus of the SMCC model, social media significantly impacts the public’s crisis communication strategies when dealing with natural disasters and organizational crises, demonstrating its usefulness in predicting behavioral response outcomes [6,7]. For instance, research on tornado disasters in the U.S. has shown that interpersonal information seeking and sharing through social media can predict compliance with government directives, such as sheltering in place [8]. While researchers have started to explore crisis communication mediated by social media beyond Western contexts or across countries [9], there is still limited research on the extent to which social media information influences public responses outside Western contexts. China, with its unique and diverse “domestically engineered” platforms like Sina Weibo, WeChat, and Douyin, represents the world’s largest social media market. In 2020, these platforms had attracted 926.8 million highly engaged users, distinct from their Western counterparts [10]. However, comprehensive analysis of how these different types of Chinese social media platforms influence crisis responses during public health emergencies remains understudied.

The Protective Action Decision Model (PADM) provides a basic causal chain from risk information reception to behavioral coping, mediated by psychological processes, offering a useful framework for understanding responses to evolving health threats [11]. PADM posits that socially transmitted warnings trigger preliminary decision-making processes, generating core psychological coping of the external threat and alternative protective measures [11,12]. For instance, psychological responses such as negative emotional reactions, risk perceptions, and hazard intrusiveness are better predictors of taking protective action during hurricanes [13,14]. However, the COVID-19 pandemic presents a distinctly different challenge from geographically confined crises such as hurricanes. Characterized by rapid evolution and pervasive uncertainty, its impact extends far beyond personal health concerns to fundamentally affect societal and global stability [3]. In this unprecedented context, risk perception and emotional response continue to serve as key psychological coping mechanisms, but they appear in increasingly complex and multidimensional ways. Despite the significance of these psychological elements in crisis response, empirical examination of how risk perception and emotional response during COVID-19 influence behavioral outcomes and their dynamic interplay remains limited.

This study aims to examine how social media-based COVID-19 information dissemination influences psychological and behavioral coping mechanisms during different phases of the pandemic in China, with a particular focus on its role in mitigating negative psychological impacts and promote adaptive behavioral responses during public health crises. Building on existing crisis communication frameworks, we conceptualize risk perception as a multidimensional construct encompassing personal health concerns, daily life disruptions, and economic impacts. Similarly, we define emotional response through its manifestation in changing emotional states, cognitive processes, and behavioral patterns [2]. These elements form the psychological coping mechanisms that are essential during crisis events. Our analysis focuses on three critical behavioral coping: protective actions, information sharing, and information gathering. Specifically, we address the following research questions:

RQ1: How does the dissemination of information through social media influence individuals’ risk perception and emotional response?

RQ2: How does the dissemination of information through social media influence individual’s protective actions, information sharing, and information gathering?

RQ3: Do the theoretical relationships among the above-mentioned variables in RQ1 and RQ2 change over different stages of a crisis?

This study makes three principal research contributions. First, it enhances our understanding of social media’s role in information dissemination by encompassing diverse social media platforms. This responds to calls for research on how the public’s use of different social media platforms influences crisis communication [15]. Second, it contributes to the literature on social media use in public crisis management and extends the theoretical framework of the SMCC model by considering the role of social media in affecting peoples’ offline behaviors taken during crises. Third, it contributes to the literature on crisis management by empirically testing the relationship between crisis cognition, specifically the risk perception and emotional response to crises, and the protective actions individuals undertake during crises.

1.1. Literature Review

The SMCC model, originally developed for crisis management, is devoted to understanding how users interpret and share crisis messages on social media, through comprehensively examining message strategies, user types, and the interactions across communication channels [5]. This model guides professional crisis communicators in developing response strategies for influential social media, while acknowledging the remaining significance of traditional media and offline word-of-mouth interactions [16]. The model’s applicability extends beyond crisis communication to health communication, as it elucidates how media content, sources, forms, and key stakeholders influence health information reception [17]. Applying the SMCC model to the pandemic context, we examine how COVID-19-related information disseminated across various Chinese social media platforms influences individuals’ psychological and behavioral coping mechanisms.

Recognizing that information dissemination’s impact on psychological and behavioral coping likely varies across different pandemic stages, we integrate the PADM to examine these dynamic relationships. The PADM delineates how individuals process and respond to risk information during crises, positing that exposure to crisis-related information from external sources shapes risk perceptions, which ultimately influence protective behaviors [11]. Specifically, this model explicates the process through which environmental and social information influences psychological decision-making and subsequent behavioral responses during disasters [18,19]. The PADM’s unique focus on the psychological processes that mediate between crisis information and protective actions [20] complements the SMCC model’s emphasis on information dissemination channels and user interactions. Risk perception is a central psychological coping factor that predicts the behavioral coping of individuals to environmental hazards and disasters in the PADM [11]. In this study, risk perception denotes the subjective judgment and estimation made by individuals regarding the probability and severity of threats posed by the COVID-19 pandemic, encompassing not only the likelihood of contracting the virus but also the potential adverse physical and social consequences [21]. Emotional response emerges as an increasingly influential psychological coping factor in shaping the public’s perceptions of crises [22], with specific emotional responses, such as fright and anxiety, being evidenced as significant predictors of individuals’ intentions to undertake protective actions [23]. In the context of the COVID-19 pandemic, we operationalize emotional response as the affective evaluation of the crisis, characterized by atypical changes in emotional, cognitive, and behavioral patterns.

1.2. Research Model and Hypotheses

Information dissemination refers to the intentional process of spreading or distributing information to specific audiences through various communication channels and platforms. Understanding which and how crisis-related information is transmitted in the early stage of a public health crisis is crucial, as it influences public psychological responses and subsequent coping behaviors.

In the context of Chinese social media platforms and pandemic health communication, this study operationalizes information dissemination through two dimensions [20]. The first concerns platform types, representing distinct transmission channels reaching varied audience segments, including major categories such as instant interaction platforms, Q&A communities and forums, and short-video platforms. The second dimension examines information content types, including crisis descriptions, potential public health risks, and recommended risk reduction approaches [20,24]. Figure 1 depicts the conceptual research model.

1.2.1. Effect of Information Dissemination on Psychological Coping

Risk perception is a psychological state stemming from the individuals’ judgments about the likelihood of a threat based on crisis-related information [11]. It is often accompanied by feelings of uncertainty and fear, causing adverse physical and social consequences [25]. Social media platforms form public risk perception and intensify emotional responses through the rapid spread of crisis-related information and compelling visuals [8,26]. Emotional response is another critical component of psychological coping. It emerges in the psychological process of evaluating crisis information, which influences individuals’ coping mechanisms and actions [27]. Accordingly, we propose the following hypotheses:

Hypothesis 1a.

The extent to which individuals are exposed to crisis information on social media positively affects their risk perception.

Hypothesis 1b.

The extent to which individuals are exposed to crisis information on social media positively affects their emotional responses.

1.2.2. Effect of Information Dissemination on Behavioral Coping

During the COVID-19 pandemic, social media platforms became a major channel for the dissemination of crisis information [28]. Studies have shown that individuals tend to actively seek information to better understand and cope with crises [29]. A symbiotic relationship exists between the dissemination of crisis information and individual behaviors, individuals are not just passive recipients but active participants in sharing information on social media platforms [30]. Crisis-related information dissemination also motivates protective action [11], which refers to a behavioral response that aims to safeguard oneself and others from the harm of a looming or active crisis.

Drawing from these insights, we argue that the dissemination of crisis information triggers a dual response: it promotes individuals to proactively seek knowledge to better understand and navigate the crisis while also fostering sharing behaviors about collective understanding and action. We also argue that crisis information dissemination positively affects the public’s tendency to engage in protective actions, indicating a direct link between information dissemination and protective behaviors. Accordingly, we propose the following hypotheses:

Hypothesis 2a.

The extent of individuals’ exposure to crisis information on social media is positively affects their information gathering.

Hypothesis 2b.

The extent of individuals’ exposure to crisis information on social media positively affects their information sharing.

Hypothesis 2c.

The extent of individuals’ exposure to crisis information on social media positively affects their protective actions.

1.2.3. Effect of Psychological Coping on Behavioral Coping

Psychological coping factors like cognitive risk perceptions, negative emotional reactions, and hazard intrusiveness are superior predictors of behavioral coping decisions during environmental hazards and pandemics compared to other variables [14]. According to PADM, higher levels of risk perception lead to higher levels of protective behavior [31].

Risk perception usually plays a central role in determining for predicting individual behavioral responses to adjust to various risks, particularly in decisions regarding protective measures. Those who perceive greater risk are more inclined to take preventive actions to reduce potential harm [32,33]. Simultaneously, people with a higher risk perception tend to increase their information gathering, and they need additional information to make accurate risk appraisals and justify appropriate protective actions to undertake [34,35].

During the initial outbreak of the COVID-19 pandemic, the existential unease associated with risk perceptions prompted more rigorous information-seeking behaviors to chart the best course of action in the face of crisis. A previous study concluded that active involvement in information-sharing behaviors regarding vaccinations is linked to risk perception in health-related uncertainty and worry [36].

The drive to share information may stem from individuals’ intention to alleviate perceived communal risks, indicating that a higher perception of risk spurs not just information acquisition but also dissemination, as part of a collective effort to mitigate the crisis’s effects [37]. Given COVID-19’s rapid spread threatening large populations, individuals with a higher risk perception are more likely to increase information sharing to urgently halt the disease’s community-level impact.

Thus, we propose the following hypotheses to encapsulate these relationships:

Hypothesis 3a.

Risk perception has a positive effect on individual’s information gathering.

Hypothesis 3b.

Risk perception has a positive effect on individual’s information sharing.

Hypothesis 3c.

Risk perception has a positive effect on individual’s protective action.

Emotional responses to the dissemination of crisis information are crucial in shaping public processing and reaction in emergency situations. Emotional response, especially negative emotion, is a strong predictor of protective actions [31]. Emotional responses can amplify perceived urgency, particularly those associated with high arousal, like wrath and fear, which in turn prompts protective actions. Previous studies about H7N9 and COVID-19 concluded that emotional response is direct related to protective action adoption [33,38].

Emotional responses also tend to drive individuals’ systematic information gathering, as evidenced by studies analyzing Twitter discussions during public health events [36,39]. For example, during the early stage of the H1N1 outbreak in Hong Kong, individuals experiencing stronger emotional responses were more likely to seek out relevant information [40].

SMCC suggested that emotional response may also predict a greater intention to share crisis information on social media during disasters [41]. Research on the Fukushima nuclear disaster found that intense emotional reactions led to greater information sharing about radiation risks on Twitter [42]. The COVID-19 pandemic accompanied by prolonged social isolation has exacerbated emotional response problems, manifested in emotional, cognitive, and behavioral changes [43]. Therefore, we posit the following hypotheses:

Hypothesis 3d.

Emotional response has a positive effect on individual’s information gathering.

Hypothesis 3e.

Emotional response has a positive effect on individual’s information sharing.

Hypothesis 3f.

Emotional response has a positive effect on individual’s protective action.

1.2.4. The Internal Effects of Psychological Coping and Behavioral Coping

Risk perception during epidemic outbreaks correlates with negative emotions like fear, anger, and anxiety. These negative emotions can impact behavior and cognition, influencing individuals’ actions and thoughts [44,45]. Previous pandemics (e.g., SARS, H1N1, Ebola, MERS-CoV) suggest a strong connection between risk perception and emotional responses [46,47,48,49]. Thus, we propose the following hypothesis:

Hypothesis 4a.

Risk perception has a positive effect on emotional response.

Previous studies have confirmed the positive association between information gathering behavior and subsequent protective actions [34,50]. Active engagement with social media for information seeking enhances crisis awareness and supports informed decision-making for protective action [51]. The theory of SMCC provides a possibility to examine the relationship between information sharing behavior and individual protective behavioral responses during public health crises [52]. A review of social media use during public emergencies uncovered that extensive information exchange (i.e., information sharing and gathering) online strengthened collective sense-making, which facilitated greater consensus on appropriate protective actions [53]. Consequently, we propose the following hypotheses to encapsulate these relationships:

Hypothesis 4b.

Information gathering has a positive effect on protective action.

Hypothesis 4c.

Information sharing has a positive effect on protective action.

2. Materials and Methods

This research aims to investigate the evolving public response to the COVID-19 pandemic through social media platforms in China. The study period was divided into two phases based on key events in China’s pandemic management. The initial outbreak phase (January–April 2020) was characterized by high uncertainty and strict containment measures, while the subsequent period phase (May–December 2020) featured downgraded emergency levels and gradually relaxed restrictions.

To investigate how social media usage influenced crisis coping mechanisms, platforms were categorized into three types: instant interaction platforms (e.g., WeChat, Weibo), Q&A communities and online forums (e.g., Zhihu.com, Baidu Tieba), and short video platforms (e.g., Douyin, Kuaishou). This classification, based on interaction styles and user demographics, enables a comprehensive analysis of how different social media types shaped public health communication and individual behavior during the crisis.

2.1. Data Collection

Data collection was conducted in December 2020 via an online survey platform (https://www.wjx.cn/). To ensure broad participation, the questionnaire link was disseminated across various social media-based communities and groups. The research purpose and context were introduced at the beginning of the questionnaire. The survey was designed to collect retrospective data for both phases. Participants were first asked to recall and respond based on their experiences and perceptions during the initial outbreak phase (January–April 2020). Subsequently, they were asked to provide responses based on their more recent experiences during the subsequent period phase (May–December 2020). To minimize recall bias, we introduced clear time periods and prompted contextual materials to facilitate accurate recollection, including references to distinct phases of the pandemic progression and associated public health measures. Although recall data collection presents methodological challenges, evidence from other research shows that people’s memories for events that occurred during disasters are reasonably accurate [54]. And as subject to environmental conditions, such recall data are also used in studies related to extreme disasters, such as hurricanes [55].

A total of 430 questionnaires were distributed, with 69 deemed invalid. Consequently, 361 valid questionnaires were retained, resulting in an effective recovery rate of 83.95%. Demographic information of the participants is summarized in

Table 1. Descriptive statistics of respondents’ characteristics (N = 361).

Item		Frequency	Percent
Gender	Male	190	52.63%
	Female	171	47.37%
Age	<18 years old	4	1.11%
	18–25 years old	119	32.96%
	26–30 years old	69	19.11%
	31–40 years old	125	34.63%
	41–50 years old	32	8.86%
	51–60 years old	10	2.77%
	>60 years old	2	0.55%
Education	High school (below)	42	11.63%
	College	284	78.67%
	Master (above)	35	9.70%
Living Arrangement	alone	53	14.68%
	with family members	308	85.32%

. To strengthen external validity and assess representativeness, we benchmarked our sample demographics against broader population statistics, including the Seventh National Population Census of China and the 47th Statistical Report on Internet Development in China (released by the China Internet Network Information Center in February 2021). Our analysis revealed reasonable alignment on key demographic variables (e.g., gender and age distributions), suggesting adequate representativeness for our research context.

2.2. Measurement

The measurement scale was developed by adapting existing literature. A five-point Likert scale (1 = “strongly disagree”, 2 = “disagree”, 3 = “neither agree nor disagree”, 4 = “agree”, 5 = “strongly agree”) was used for most constructs, while information gathering and sharing were assessed using single-item measures (1 = “very low”, 2 = “low”, 3 = “moderate”, 4 = “high”, 5 = “very high”). The detailed instrument is provided in

Table 2. Measurement Items and Sources in the Literature.

Construct		Item		Refs.
Information dissemination (INDI)	Social media type (SMET)	SMT1	I rely on information related to the COVID-19 pandemic from instant interaction platforms (e.g., WeChat and Weibo).	[20]
		SMT2	I rely on information related to the COVID-19 pandemic from Q&A communities and forums (e.g., Zhihu.com and Baidu Tieba).
		SMT3	I rely on information related to the COVID-19 pandemic from short video platforms (e.g., Douyin, Kuaishou).
	Information content type (ICOT)	IC1	I pay attention to the information content of updates on the COVID-19 pandemic, which encompasses information on symptoms, modes of transmission, infection numbers, and details about infected individuals.
		IC2	I pay attention to the information content of knowledge about COVID-19 prevention measures, including protective measures and healthy habits.
		IC3	I pay attention to regional information regarding COVID-19, including the current situation in affected regions and detailed insights about the pandemic within my own region.
Risk perception (RISP)		RP1	The severity of the pandemic, considering factors such as mortality rate, infectiousness, duration, and total number of infections, is high.	[40]
		RP2	The severity of the pandemic’s impact on my health, including the risk of self-infection with COVID-19 and potential adverse consequences, is high.
		RP3	The disruptions caused by the pandemic in my daily life, such as potential upheaval in daily activities or significant property damage, are high.
		RP4	The impact of the pandemic on the economy (e.g., concerns about potential effects on national or global economic conditions) is high.
Emotional response (EMRE)		ER1	I have experienced emotional distress, such as anxiety, fear, tension, sadness, despair, irritability, and helplessness.	[44,45]
		ER2	I have experienced abnormal behavior, such as fidgeting, social withdrawal, avoidance, and appetite changes.
		ER3	I have experienced cognitive impairment, such as blunted perception, poor concentration, memory decline, errors in judgment, and extreme thinking.
Information gathering (IGAB)		IG	Overall, my level of engagement in gathering pandemic-related information through social media can be rated as: _____	[30]
Information sharing (ISHB)		IS	Overall, my level of engagement in sharing pandemic-related information through social media can be rated as: _____
Protective action (PRAC)		PA1	I maintain good hygiene habits, such as frequent handwashing, avoiding facial contact with unwashed hands, and using tissues or elbows when coughing or sneezing.	[40,56]
		PA2	I take personal protective measures, such as avoiding crowded places, limiting outdoor activities, adhering to nucleic acid testing protocols, maintaining a one-meter social distance, and consistently using face masks in public spaces.
		PA3	I adhere to health requirements, such as advocating for eating separately with utensils, adhering to a nutritious diet, and strictly following quarantine policies.

.

The research model incorporates both first-order and second-order constructs, with information dissemination conceptualized as a reflective–formative higher-order construct. Figure 2 illustrates the hierarchical model, which includes gender, age, education, and living arrangements as control variables [57].

3. Results

3.1. Common Method Bias (CMB)

This study controlled for CMB by ensuring the anonymity of the questionnaires and reassuring respondents that there were no standard answers to alleviate their evaluation apprehension [58]. Additionally, Harman’s single-factor test was conducted to identify potential CMB [59]. The first extracted factor for each phase accounted for less than the threshold value of 40% of the variance (initial outbreak: 27.2%, subsequent period: 23.5%), indicating a lack of substantial CMB.

3.2. Demographic Variables

This study examined the impact of key demographic variables, including gender, age, education level, and living arrangements, on individuals’ reliance on social media platforms and their protective actions during the pandemic.

A t-test reveals that gender significantly impacted the preference for social media type. Specifically, females exhibit a greater preference for instant interaction platforms (e.g., WeChat and Weibo) (M = 3.63, SD = 1.18) compared to males (M = 3.13, SD = 1.14).

A one-way ANOVA shows that age has a significant effect on protective actions. Post hoc comparisons suggest that participants aged 31–40 years (M = 4.38, SD = 0.57) engage in markedly more hygiene-maintenance actions than those aged 18–25 years (M = 4.05, SD = 0.67, p < 0.001) and 41–50 years (M = 4.02, SD = 0.69, p < 0.01). This suggests that protective action involvement peaked among middle-age groups during the pandemic.

Education level also impacts social media platform preference. The ANOVA results show significant differences for instant interaction platforms (e.g., WeChat and Weibo), F(2,358) = 6.181, p < 0.01, and Q&A communities and forums (e.g., Zhihu.com and Baidu Tieba), F(2,358) = 3.672, p < 0.05. Participants with a college education (M = 3.91, SD = 0.67) show a stronger preference for instant interaction platforms (e.g., WeChat and Weibo) compared to those with high school or lower education (M = 3.51, SD = 0.72, p < 0.01). Additionally, college-educated participants (M = 2.88, SD = 0.85) prefer Q&A communities and forums (e.g., Zhihu.com and Baidu Tieba) more than those with high school or lower education levels (M = 2.51, SD = 0.97, p < 0.05).

Finally, living status significantly influences protective actions per the t-test results, t(359) = 3.433, p < 0.01. Those residing with family (M = 4.26, SD = 0.65) display more hygiene-maintenance habits than individuals living alone (M = 3.92, SD = 0.67).

3.3. Social Media Types on Crisis Coping

This study examined the impact of different social media types on people’s coping during the crisis (see

Table 3. Results of ANOVA Analyses.

Construct	Significance (F Test)
	Instant Interaction Platforms	Q&A Communities and Forums	Short Video Platforms
Initial Outbreak
Risk perception	4.231 ***	1.895 ns	5.029 **
Emotional response	2.137 *	2.621 **	2.269 ns
Information gathering	2.111 *	0.511 ns	0.358 ns
Information sharing	3.35 **	1.179 ns	2.103 ns
Protective action	5.115 ***	1.405 ns	3.443 **
Subsequent Period
Risk perception	0.597 ns	4.277 ***	4.171 **
Emotional response	1.134 ns	4.696 ***	1.149 ns
Information gathering	0.783 ns	1.492 ns	2.622 *
Information sharing	1.758 ns	1.389 ns	2.065 ns
Protective action	2.409 *	1.621 ns	4.451 **

*** Sig. < 0.001; ** Sig. < 0.01; * Sig. < 0.05; ns Sig. > 0.05.

). Short video platforms significantly affected risk perception in both phases. In contrast, instant interaction platforms exhibited significant effects only during the initial outbreak, with their influence on risk perception faded in the subsequent period. Q&A communities and forums, on the other hand, had no significant impact on risk perception in the initial outbreak phase but became significant in the subsequent period.

Regarding emotional response, Q&A communities and forums were more influential than instant interaction platforms and short video platforms across both phases. Information gathering and sharing exhibited interesting dynamics. During the initial outbreak, instant interaction platforms had a greater impact on information gathering and sharing compared to Q&A communities and forums, as well as short video platforms. In the subsequent period, short video platforms increasingly influenced information gathering, while the effects of other platforms diminished. Finally, protective actions were significantly influenced by instant interaction platforms and short video platforms, whereas Q&A communities and forums had minimal impact across both phases.

3.4. Assessment of the Measurement Model

Considering the hierarchical structure of our research model and the inclusion of the reflective–formative construct (i.e., information dissemination), we adopted the established validation procedures to assess the measurement model [60,61].

3.4.1. Assessment of the First-Order Constructs Measurement Model

This study examined the measurement model by reliability, convergent validity, and discriminant validity in each phase. Reliability assessment involved calculating internal consistency, the item reliability was assessed by the factor loadings, and the construct reliability was assessed using composite reliability (CR).

As exhibited in

Table 4. Descriptive statistics of the measure.

Construct	Indicator	Mean	S.D.	Loading	T-Statistic	C.R.
Initial Outbreak
Social media type (SMET)	SMT1	3.859	0.690	0.823	19.060 ***	0.731
	SMT2	2.821	0.878	0.585	6.205 ***
	SMT3	3.640	1.184	0.650	9.290 ***
Information content type (ICOT)	IC1	4.420	0.775	0.826	30.776 ***	0.848
	IC2	4.240	0.875	0.776	27.338 ***
	IC3	4.290	0.841	0.818	28.859 ***
Risk perception (RISP)	RP1	4.070	0.870	0.768	28.367 ***	0.814
	RP2	3.540	1.093	0.730	19.970 ***
	RP3	3.700	1.044	0.757	23.679 ***
	RP4	3.760	1.006	0.632	12.709 ***
Emotional response (EMRE)	ER1	3.470	0.916	0.887	34.645 ***	0.769
	ER2	3.180	0.614	0.622	8.596 ***
	ER3	3.220	0.708	0.652	10.956 ***
Information gathering (IGAB)	IG	3.930	1.087	1.000	-	1.000
Information sharing (ISHB)	IS	3.680	1.235	1.000	-	1.000
Protective action (PRAC)	PA1	4.300	0.814	0.830	37.784 ***	0.837
	PA2	4.300	0.915	0.810	30.787 ***
	PA3	4.170	0.843	0.741	19.427 ***
Subsequent Period
Social media type (SMET)	SMT1	3.859	0.690	0.579	5.836 ***	0.738
	SMT2	2.821	0.878	0.740	8.139 ***
	SMT3	3.640	1.184	0.763	11.482 ***
Information content type (ICOT)	IC1	4.040	0.849	0.798	29.509 ***	0.807
	IC2	3.910	0.910	0.742	19.359 ***
	IC3	4.100	0.888	0.748	20.261 ***
Risk perception (RISP)	RP1	3.590	0.912	0.728	21.225 ***	0.839
	RP2	3.040	1.165	0.803	35.451 ***
	RP3	3.210	1.171	0.812	36.424 ***
	RP4	3.390	0.975	0.659	15.190 ***
Emotional response (EMRE)	ER1	2.660	0.893	0.823	20.574 ***	0.765
	ER2	2.940	0.537	0.628	8.135 ***
	ER3	2.870	0.687	0.706	13.015 ***
Information gathering (IGAB)	IG	3.520	1.252	1.000	-	1.000
Information sharing (ISHB)	IS	3.090	1.352	1.000	-	1.000
Protective action (PRAC)	PA1	4.120	0.797	0.760	19.866 ***	0.814
	PA2	4.020	0.928	0.750	20.180 ***
	PA3	4.020	0.908	0.800	25.707 ***

*** Sig. < 0.001.

, the factor loadings of all the items in both phases exceed 0.5 [62,63,64] and are significant (p < 0.001) [65]. The CRs exceed the suggested cut-off value of 0.7 [66]. Thus, these results demonstrate the adequate internal consistency of our measurement model, indicating a well-constructed reliability.

To evaluate discriminant and convergent validity, this study employed the Fornell–Larcker criterion and Heterotrait–Monotrait Ratio (HTMT). Each method demonstrates satisfactory reliability and validity for the first-order constructs, fitting the data well, and being appropriately specified in both phases. The relevant statistical results are presented in

Table 5. Correlation matrix.

	AVE	SMET	ICOT	RISP	EMRE	IGAB	ISHB	PRAC
Initial Outbreak
SMET	0.480	0.693
ICOT	0.651	0.248	0.807
RISP	0.524	0.299	0.419	0.724
EMRE	0.533	0.244	0.172	0.349	0.730
IGAB	1.000	0.101	0.110	0.115	0.138	1.000
ISHB	1.000	0.211	0.145	0.192	0.211	0.168	1.000
PRAC	0.631	0.280	0.571	0.326	0.222	0.174	0.158	0.795
Subsequent Period
SMET	0.488	0.699
ICOT	0.583	0.252	0.763
RISP	0.567	0.269	0.367	0.753
EMRE	0.523	0.176	0.169	0.412	0.723
IGAB	1.000	0.163	0.172	0.256	0.148	1.000
ISHB	1.000	0.143	0.232	0.244	0.254	0.336	1.000
PRAC	0.593	0.170	0.448	0.223	0.081	0.119	0.108	0.770

Note: AVE = average variance extracted; the bold diagonally presented data refer to the square roots of AVEs. SMET = social media type; ICOT = information content type; RISP = risk perception; EMRE = emotional response; PRAC = protective action; ISHB = information sharing; IGAB = information gathering.

and

Table 6. HTMT criterion test.

	SMET	ICOT	RISP	EMRE	IGAB	ISHB	PRAC
Initial Outbreak
SMET	-
ICOT	0.373	-
RISP	0.471	0.568	-
EMRE	0.489	0.229	0.506	-
IGAB	0.128	0.129	0.142	0.179	-
ISHB	0.285	0.169	0.224	0.254	0.168	-
PRAC	0.411	0.792	0.445	0.309	0.210	0.185	-
Subsequent Period
SMET	-
ICOT	0.448	-
RISP	0.418	0.537	-
EMRE	0.308	0.285	0.600	-
IGAB	0.226	0.214	0.294	0.178	-
ISHB	0.193	0.287	0.277	0.304	0.336	-
PRAC	0.340	0.690	0.329	0.224	0.144	0.132	-

Note: SMET = social media type; ICOT = information content type; RISP = risk perception; EMRE = emotional response; PRAC = protective action; ISHB = information sharing; IGAB = information gathering.

.

3.4.2. Assessment of the Second-Order Constructs Measurement Model

Following the established procedures [67], three diagnostic criteria of second-order construct were examined: indicator collinearity, statistical significance, and relevance of indicator weights. The results are summarized in

Table 7. Results of the second-order constructs measurement model assessment.

	Weights	T Statistics	VIF
Initial Outbreak
SMET → INDI	0.421	5.198 ***	1.066
ICOT → INDI	0.808	14.477 ***	1.066
Subsequent Period
SMET → INDI	0.360	4.230 ***	1.067
ICOT → INDI	0.847	15.820 ***	1.067

*** Sig. < 0.001.

.

The variance inflation factor (VIF) values are all under 3, indicating no critical collinearity problems. The weights for all formative indicators in both phases are significant at the 0.001 level, demonstrating their statistically meaningful contribution to forming the higher-order information dissemination construct. Furthermore, the weights in both phases are ranging from 0.36 to 0.85, signifying all first-order constructs substantially contribute to information dissemination based on their empirical weights. The validity of the higher order construct is therefore warranted.

3.5. Assessment of the Structural Model

To evaluate the structural model, we employed a disjoint two-stage approach [61], which allows for accurate estimation of complex hierarchical structures. This approach first obtaining latent variable scores from first-order constructs and then uses these scores as indicators in a higher-order model. The results of this analysis are presented in Figure 3.

The explanatory and predictive power of the research model was further evaluated using R² values and effect sizes (f²). The R² values indicate the variance explained in the endogenous constructs, measuring the model’s predictive accuracy [68]. Effect sizes f² values of 0.02, 0.15, and 0.35 indicate small, medium, and large effects, respectively [69,70]. The results of predictive power and effect size analyses are shown in

Table 8. Structural parameter estimates.

Hypothesized Path	Initial Outbreak			Subsequent Period
	Path Coefficients	p-Value	Effect Size	Path Coefficients	p-Value	Effect Size
H1a: INDI → RISP	0.465	0.000	0.275 (medium to large)	0.408	0.000	0.200 (medium to large)
H1b: INDI → EMRE	0.102	0.092	0.009 (not significant)	0.047	0.403	0.002 (not significant)
H2a: INDI → IGAB	0.089	0.196	0.006 (not significant)	0.117	0.030	0.012 (very small)
H2b: INDI → ISHB	0.133	0.043	0.015 (small)	0.169	0.003	0.027 (small)
H2c: INDI → PRAC	0.522	0.000	0.329 (medium to large)	0.410	0.000	0.164 (medium)
H3a: RISP → IGAB	0.037	0.586	0.001 (not significant)	0.190	0.001	0.028 (small)
H3b: RISP → ISHB	0.076	0.248	0.005 (not significant)	0.102	0.084	0.008 (not significant)
H3c: RISP → PRAC	0.025	0.664	0.001 (not significant)	0.065	0.292	0.004 (not significant)
H3d: EMRE → IGAB	0.103	0.099	0.009 (not significant)	0.045	0.427	0.002 (not significant)
H3e: EMRE → ISHB	0.154	0.004	0.022 (small)	0.177	0.001	0.029 (small)
H3f: EMRE → PRAC	0.062	0.177	0.005 (not significant)	−0.027	0.682	0.001 (not significant)
H4a: RISP → EMRE	0.304	0.000	0.084 (small to medium)	0.393	0.000	0.155 (medium)
H4b: IGAB → PRAC	0.103	0.017	0.016 (small)	0.040	0.477	0.002 (not significant)
H4c: ISHB → PRAC	0.007	0.889	0.000 (not significant)	−0.009	0.852	0.000 (not significant)
R-square
RISP	0.216			0.167
EMRE	0.132			0.171
IGAB	0.030			0.079
ISHB	0.075			0.111
PRAC	0.380			0.207

, providing additional empirical support for the hypothesized relationships in the research model.

The results provide support for several hypothesized relationships in the structural model. First, information dissemination positively influences risk perception (H1a), information sharing (H2b), and protective action (H2c) across both phases. However, its direct effect on emotional response (H1b) is not significant, and its effect on information gathering (H2a) is significant only in the subsequent period. Second, risk perception positively affects emotional response (H4a) across both phases. Its effect on information gathering (H3a) is only significant in the subsequent period, while the impacts on information sharing (H3b) and protective action (H3c) are not statistically significant. Third, emotional response positively influences information sharing (H3e) at both phases, but does not significantly affect information gathering (H3d) or protective action (H3f). Finally, information gathering significantly predicts protective action (H4b) only in the initial outbreak phase, while the effect of information sharing on protective action (H4c) is nonsignificant.

A multigroup analysis was conducted to test changes in effects between the two phases [63]. This analysis reveals two significant differences in path coefficients between the initial outbreak and subsequent period (see

Table 9. Multigroup analysis.

Hypothesized Path	Path Coefficients		Difference
	Initial Outbreak	Subsequent Period
H1a: INDI → RISP	0.465	0.408	−0.057
H1b: INDI → EMRE	0.102	0.047	−0.055
H2a: INDI → IGAB	0.089	0.117	0.028
H2b: INDI → ISHB	0.133	0.169	0.036
H2c: INDI → PRAC	0.522	0.410	−0.112
H3a: RISP → IGAB	0.037	0.190	0.153 *
H3b: RISP → ISHB	0.076	0.102	0.026
H3c: RISP → PRAC	0.025	0.065	0.040
H3d: EMRE → IGAB	0.103	0.045	−0.058
H3e: EMRE → ISHB	0.154	0.177	0.023
H3f: EMRE → PRAC	0.062	−0.027	−0.089
H4a: RISP → EMRE	0.304	0.393	0.089
H4b: IGAB → PRAC	0.103	0.040	−0.063
H4c: ISHB → PRAC	0.007	−0.009	−0.016

Note: * p < 0.05. INDI = information dissemination; RISP = risk perception; EMRE = emotional response; PRAC = protective action; ISHB = information sharing; IGAB = information gathering.

). First, the positive effect of risk perception on information gathering (H3a) becomes significantly stronger over time, with the path coefficient increasing by 0.153. This suggests an enhanced role of risk perception on information gathering as the crisis evolves. Second, the significant paths (H1a, H2b, H2c, H4a, H3e) remain consistent across both phases, indicating the stability of these hypothesized relationships.

3.6. Mediation Analysis

To further elucidate the mechanisms underlying the relationships in our theoretical model, this study examined the potential mediating roles of individual psychological coping factors using 5000 bias-corrected bootstrap samples at the 95% BC confidence level [60,71]. The results provide evidence of partial mediation for several hypothesized relationships (see

Table 10. Results of the mediating effect based on bootstrapping test.

Relationship	Initial Outbreak			Subsequent Period
	Indirect Effect	Direct Effect with Mediation	Direct Effect Without Mediation	Indirect Effect	Direct Effect with Mediation	Direct Effect Without Mediation
INDI → RISP → PRAC	0.012	0.522 ***	0.551 ***	0.027	0.410 ***	0.437 ***
INDI → RISP → ISHB	0.036	0.133 *	0.195 ***	0.042	0.169 **	0.246 ***
INDI → RISP → IGAB	0.017	0.089	0.129 *	0.078 **	0.117 *	0.196 *** (Partial mediation)
INDI → RISP → EMRE → PRAC	0.009	0.522 ***	0.551 ***	−0.004	0.410 ***	0.437 ***
INDI → RISP → EMRE → ISHB	0.022 *	0.133 *	0.195 *** (Partial mediation)	0.028 **	0.169 **	0.246 *** (Partial mediation)
INDI → RISP → EMRE → IGAB	0.015	0.089	0.129 *	0.007	0.117 *	0.196 ***
INDI → EMRE → PRAC	0.006	0.522 ***	0.551 ***	−0.001	0.410 ***	0.437 ***
INDI → EMRE → ISHB	0.016	0.133 *	0.195 ***	0.008	0.169 **	0.246 ***
INDI → EMRE → IGAB	0.010	0.089	0.129 *	0.002	0.117 *	0.196 ***

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. Mediation types are marked below their values of direct effect without mediation. INDI = information dissemination; RISP = risk perception; EMRE = emotional response; PRAC = protective action; ISHB = information sharing; IGAB = information gathering.

).

4. Discussion

Our findings suggest that social media information dissemination strongly influences risk perception at each phase (H1a), with no significant differences in this effect across phases. Consistent with PADM literature, obtaining crisis-related information from external sources contributes to risk perception formation and increases likelihood of protective behaviors [12]. Interestingly, social media dissemination had no influence on emotional response at any phase (H1b), while risk perception strongly triggered emotional responses across phases (H2a), with this relationship strengthening over time. Previous PADM studies have highlighted emotional responses during environmental hazards [14]. Unlike previous studies focusing on specific emotions [6,23], this study investigates emotional response from emotional, behavioral, and cognitive alterations to capture its features, providing a more comprehensive understanding.

Our results demonstrate that information dissemination through social media platforms strongly impacts the behavioral coping factor of protective action at each phase (H1e). Addressing prior calls to expand the SMCC model by examining protective actions as behavioral outcomes [41], this study empirically supports the relationship existing during and after the initial outbreak of COVID-19. Information sharing and information gathering are key public communicative actions in the SMCC model [8]. Our findings show that information dissemination through social media strongly influences information sharing across phases (H1d), echoing previous SMCC literature. In contrast, its effect on information gathering is only significant in the subsequent period (H1c). Notably, risk perception also directly impacts information gathering in the subsequent period (H2b), partially mediating the relationship between information dissemination and information gathering. Moreover, information gathering significantly predicts protective actions during the initial outbreak (H4a) but not in the subsequent period. This may reflect the early uncertainty, when limited public knowledge prompted active information seeking, leading to protective actions. These findings complement the PADM, proving that social media platforms as information sources significantly influence behavioral reactions to some extent.

Previous studies about SMCC indicated that emotional response can be a predictor of individuals’ consequent behavioral response in crisis [46], such as protective action. Our findings extend this by showing that emotional responses do not directly influence protective actions (H3c) but significantly affect information-sharing behaviors across phases (H3b). We further examined the mediating role of emotional responses in the relationship between information dissemination and behavioral coping.

Results demonstrate that a combined effect of risk perception and emotional responses mediates the link between information dissemination and information sharing. Negative emotions during crises stem not only from concerns about personal health and safety but also from worries about others’ well-being, particularly as perceived risk increases [37]. During COVID-19, characterized by widespread infection, prolonged uncertainty, and severe health impacts, information sharing functioned as a key coping strategy to manage these emotions. These findings advance SMCC research by clarifying the role of emotional responses in shaping public behavior during and after health crises.

4.1. Theoretical Implications

This study offers significant theoretical contributions. First, this research extends the theoretical boundary of PADM by examining its application to social media-based health communication. Our study investigates the mechanisms through which social media information dissemination influences information sharing and gathering, and subsequently drives protective actions. Our analysis reveals phase-specific behavioral coping patterns: while information dissemination via social media maintains a direct association with protective actions throughout the pandemic, information sharing and information gathering exhibit distinct temporal variations. Specifically, during the initial outbreak period, individuals predominantly engaged in information sharing. This pattern evolved during the subsequent period to encompass both information sharing and information gathering. These findings significantly advance PADM’s theoretical framework by demonstrating how behavioral responses to social media information dissemination evolve across different phases of a prolonged public health crisis, suggesting that temporal dimensions should be incorporated in health communication when explaining behavioral responses, particularly in extended crises like COVID-19.

This study contributes to the health communication literature, particularly in the domain of crisis communication during global health emergencies. While previous research has typically examined information sharing and gathering as discrete phenomena or subsumed them under the broader concept of information exchange [72], our study provides novel insights by simultaneously investigating these interrelated information behaviors across different phases of the pandemic. Notably, information gathering demonstrates a significant association with protective actions during the initial outbreak but not significant during the subsequent period. In response to calls for research on the link between information sharing and protective behaviors [36], our findings further show that information sharing consistently lacked a significant relationship with protective actions throughout the pandemic.

This study advances understanding of emotional response mechanisms in public health crises, particularly regarding protective actions. While previous studies have focused primarily on specific emotions such as fear and anxiety [6,23], We examine emotional responses generated during the pandemic through a multidimensional lens, encompassing emotional, behavioral, and cognitive alterations. This holistic perspective allows us to capture the complex nature of emotional responses in crisis situations more accurately. Our findings provide empirical evidence that emotional response significantly influences information sharing through a chain-mediation effect with risk perception. Additionally, we demonstrate that during public health crises, information dissemination directly affects protective actions rather than operating indirectly through risk perception, while risk perception exhibits an increasingly strengthened positive effect on emotional response over time, partially extending the PADM literature.

4.2. Practical Implications

Our study yields three practical implications and contributes meaningfully to several Sustainable Development Goals (SDGs) established by the United Nations. Firstly, it informs strategic health communication design via social media during global crises. Findings underscore the importance of tailored messaging, as demographics significantly influence crisis responses and coping behaviors. Women tend to prefer instant interaction platforms, whereas college-educated individuals show a preference for both instant interaction platforms and Q&A communities and forums. Middle-aged and family-residing individuals exhibit stronger hygiene practices. Practitioners can leverage these platform preferences and behavioral tendencies to optimize health-related message dissemination across diverse population segments. Such social media platforms act as an innovative digital infrastructure supporting social resilience during crises, especially in the context of China’s vast social media ecosystem, directly aligning with SDG3 (Good Health and Well-being) and SDG9 (Industry, Innovation, and Infrastructure). For example, instant interaction platforms, especially those equipped with short video capabilities, can effectively reach targeted demographics while facilitating the dissemination of hygiene practices and real-time updates during emergencies. This technological approach enhances community resilience and operational continuity during public health crises. Strategic use of various social media platforms, aligned with demographic factors, can enhance the effectiveness of public health interventions, effectively communicating essential information and promoting protective actions among distinct populations.

Secondly, our research highlights social media’s crucial role in health crisis communication by examining the dynamic changes in individuals’ psychological and behavioral coping mechanisms. Findings reveal that social media-disseminated information directly impacts risk perception, information sharing, and protective actions throughout a health crisis. Consequently, we emphasize the importance of ensuring accurate and timely crisis-related information on social media platforms, which is aligning to SDG 11 (Sustainable Cities and Communities). This alignment manifests through the enhancement of urban resilience and community preparedness, as reliable information systems serve as fundamental infrastructure for disaster risk reduction and emergency management. Accurate social media communications enable communities to make informed decisions during crises, potentially reducing casualties and economic losses while strengthening social cohesion. We recommend that governmental health crisis management strategies prioritize the consistent delivery of reliable information through social media channels to effectively shape public response and promote adaptive behaviors.

Thirdly, our study illuminates the mechanism of emotional responses in health communication through social media. Contrary to previous speculation [22,73], our empirical findings demonstrate that individuals’ emotional responses directly influence information-sharing behaviors but not protective actions. This insight emphasizes the critical need for government officials to monitor emotional sentiment in user-generated content and strategically modulate official crisis communications, which is aligning to SDG 16 (Peace, Justice and Strong Institutions). This connection is particularly significant as effective institutional communication fosters public trust—a cornerstone of legitimate governance during crises. By acknowledging and addressing emotional responses within digital discourse, authorities can develop more transparent, responsive, and accountable communication strategies that reduce misinformation proliferation and promote factual content sharing. In practice, such an approach enables officials to identify at-risk populations, and guide targeted emotional support provision during health crises, ultimately enhancing the effectiveness and reach of public health interventions.

4.3. Limitations

Despite this study’s potential contributions, there are still some limitations that can be addressed by future research. Firstly, our exploration of information dissemination is restricted to social media type and information content type. Future studies could incorporate additional dimensions of information dissemination, such as information form type (e.g., text, graph, video) [20], to provide crisis communication managers with more comprehensive guidance.

Secondly, this study aims to present a universal mechanism of information dissemination on individuals’ psychological and behavioral responses, ignoring a specific category of information. During crises, however, accurate information often becomes intermingled with unverified claims and misinformation. To deepen our understanding, future research could categorize these information types and longitudinally track their differential effects on subsequent public perception and behavior.

We believe that a longitudinal study design provides comprehensive insights and evidence on individuals’ entire decision-making process: from initial crisis-related information dissemination on social media platforms, through early psychological coping (risk perception and emotional response), to subsequent behavioral coping (information gathering, information sharing, and protective action), and finally to the longer-term psychological-behavioral adaptations. Such an approach would better capture the temporal dynamics of these relationships and potentially reveal causal mechanisms that cross-sectional designs cannot identify.

Furthermore, future research should distinguish between passive social media engagement (reading posts) and active engagement (creating/sharing content), as these different modes of interaction may differentially influence individuals’ psychological and behavioral responses. Additionally, examining how various platform features and algorithmic recommendations moderate these effects would contribute valuable insights to both theory and practice in crisis communication.

Thirdly, it is important to acknowledge the inherent limitations of survey-based research regarding the declarative nature of respondents’ answers. As with all self-reported data, participants’ responses may be influenced by social desirability bias, where individuals tend to present themselves favorably rather than reporting their true attitudes or behaviors. Future research could address this limitation by triangulating survey data with observational methods or digital trace data to provide more robust validation of self-reported information consumption and behavioral response patterns.

Finally, this study exclusively focuses on the impact of diverse social media platforms to enhance our understanding of SMCC during the COVID-19 pandemic, disregarding other information channels. However, cultural differences may influence preferred information channels (e.g., TV, and newspapers) [74]. Thus, we call for future research to expand on the current study by conducting cross-cultural comparisons in the context of public health crises.

5. Conclusions

This study investigated the interplay between psychological coping (risk perception and emotional response) and behavioral coping (information gathering, information sharing, and protective action) in the context of crisis-related information dissemination on social media platforms. Based on a survey of 361 participants, our findings show that social media information dissemination significantly and directly influenced public risk perception, information sharing, and protective actions during the COVID-19 crisis. Importantly, psychological coping factors mediated the relationship between information dissemination and information sharing across all phases of the crisis.

These insights enhance our understanding of health communication dynamics during public health emergencies and highlight the critical role of social media in shaping public responses to crises, with significant implications for multiple SDGs. Our findings support SDG 3 by improving crisis communication strategies that can enhance public health outcomes during emergencies. The demonstrated importance of social media platforms as crisis communication infrastructure aligns with SDG 9, particularly in developing resilient digital systems for information dissemination. Furthermore, our results contribute to SDG 11 by providing insights on how accurate and timely crisis information enhances community resilience and emergency preparedness. The emotional response mechanisms identified in our study support SDG 16 by offering guidance for government officials to build more transparent, accountable, and responsive communication frameworks that foster public trust during crises. Collectively, these contributions advance our understanding of how effective crisis communication through social media can support multiple dimensions of sustainable development while enhancing societal resilience to public health challenges.