In December 2022, the Chinese State Council issued the Notice on Further Optimizing the Implementation of COVID-19 Prevention and Control Measures, announcing the liberalization of COVID-19 control in China [1]. Following the policy change, there was a rapid and widespread surge in virus transmission. During this period, the health status of older adults became a matter of concern. Older adults, characterized by degenerating body functions, weakened immune systems, and underlying diseases [2], are generally at a disadvantage in fighting viral infections. Survey data showed that the mortality rate of COVID-19 is significantly higher among older adults aged ≥65 years than among younger adults globally. In China, the COVID-19 mortality rate is 2.3%, with >50% of deaths occurring in patients aged ≥50 years [3]. Thus, older adults are considerably more susceptible to the detrimental effects of COVID-19 than the general population. In the postepidemic era, it is crucial to examine the behavioral intentions of older adults to cope with disease risk.

This study aimed to address this research problem by introducing the structural influence model of communication (SIM), a research framework proven to be reliable in the context of large-scale health risks. In addition, this study aimed to enhance the SIM by delving into specific variables in depth, seeking to establish a more comprehensive research perspective on investigating influencing factors. This approach further examined the applicability of the model across a broader range of research contexts.

The SIM suggests that health cognition and behavior should be considered within the elements of the social system based on the social-ecological theory [4]. It builds on the basic idea that communication is controlled by power dynamics, similar to the knowledge gap theory, which suggests that individuals with lower socioeconomic status have limited access to information [5]. Accordingly, the model states that social determinants and the conditions of mobilizable resources greatly shape the information environment, leading to communication inequalities and then differentiated health behaviors and outcomes. However, most previous studies have focused on individual-level factors, such as age [6], education [7], and revenue [8,9], with minimal attention paid to the social resources that individuals may have access to [10]. In fact, a weak ability to mobilize social resources, that is, social capital, could often lead to a disadvantage in health information seeking [11,12] and health behaviors [13].

Furthermore, the communication inequality could manifest in multiple aspects, such as health message search frequency [14], message type [15,16], and media exposure preferences [17]. As media technology advances, disparities in the use of and exposure to traditional and new media, on the one hand, have become visible, resulting in suboptimal health behaviors among certain populations. Several studies support this perspective. For instance, Viswanath and Ackerson [13] explored the phenomenon of mass media exposure in the United States and concluded that social structures, such as race and class, lead to widespread health disparities.

Research on the SIM is an ongoing endeavor. Some scholars believe that more evidence is still needed to establish the relationship among social capital, communication, and health behavior [12,18]. One of the reasons is that current research based on the SIM does not clearly reveal the intrapersonal processes that guide communication decisions [11,19], whereas some other theories state that affective responses to risk, such as anxiety and fear, possibly relate to information seeking [20]. Hovick et al [11] argued that personal perception factors, particularly in the face of risk, can further influence communication inequalities and should be considered. In particular, risk perception can influence information-seeking attitudes and outcomes by shaping affective responses [20]. When individuals perceive threats as serious, it triggers negative emotions, such as fear and anxiety, motivating them to seek and process information to mitigate those emotions [21]. Individuals with a strong sense of risk and self-efficacy are more likely to engage in protective behaviors driven by negative emotions [22]. In this paper, we continue this line of thought and introduce the variable of negative emotions to expand the existing relationship among social capital, media exposure, and protective behavior intention, ultimately forming the theoretical framework.

This study collected data in March 2023. We developed a web-based questionnaire through Wenjuan.com (Zhongyan Technology), a Chinese platform akin to Google Forms (Google LLC). Snowball sampling method was used in this study. We disseminated the web-based survey link on the WeChat (Tencent Inc) platform and encouraged our friends to share it with their older family members. The participants were limited to individuals aged ≥60 years. They were requested to complete the questionnaire, either in person or with the assistance of their family members. Each respondent was rewarded with a monetary incentive (RMB ¥5 [US $0.7]) after completing the survey.

Before the formal distribution of the questionnaire, a small-scale pretest was conducted over the web, and 46 completed questionnaires were collected. On the basis of the feedback received from the respondents, certain questions were modified. The final version of the questionnaire was then distributed on a large scale in March 2023, and 514 completed questionnaires were collected. However, 115 (22.4%) questionnaires were deemed invalid owing to incomplete answers or a completion time of <150 seconds. Therefore, a total of 399 valid questionnaires were included in the analysis, representing a completion rate of 77.6%. The age of the participants varied from 60 to 93 years, with an average age of 64.93 (SD 4.91) years. Table 1 presents the characteristics of the respondents.

The study received ethics approval from the University of Southern California-Shanghai Jiao Tong University Institute of Cultural and Creative Industry Ethics Board. The institutional policy does not require the College to issue a separate approval document with a reference number for this type of humanities and social sciences research currently. Participants were provided with a written description of the study before completing the web-based questionnaire, ensuring the confidentiality of their private information solely for the purpose of this study. The completion of the questionnaire indicated approved consent from each participant.

In this study, SPSS and MPLUS (version 8.3; Muthén & Muthén) were used to test the reliability of the measurement models. Construct reliability was evaluated based on certain conditions, including a Cronbach α value of >.7 and a composite reliability value >0.7 [68,69]. Average variance extracted is considered to be >0.5 [69], but we can also accept 0.4 [70-73]. Then, a confirmatory factor analysis was conducted to test the model fit. The following fit indices, as suggested by Browne and Cudeck [74], were considered for a reasonably fitting model: a root-mean-square error of approximation of <0.08, Tucker-Lewis index and comparative fit index of >0.9, and ratio of chi-square to df (χ²/df) of <5.

To test the research hypotheses, a linear regression analysis was performed. Before conducting the regression analysis, a validated one-way analysis was performed in MPLUS to assess the presence of common method bias. In addition, multicollinearity was evaluated using variance inflation factors, and variance inflation factor values of <10 were considered acceptable, indicating that all predictor variables were available for a multivariate analysis [69]. This analysis was conducted in SPSS. Next, this study tested the mediating and moderating effects using models 4 and 7 of the SPSS macro. Indirect effects were calculated using a bootstrapping technique [75], which involved generating a 95% bias-corrected CI (BC) for indirect effects based on random samples of data. If the BC did not include 0, mediation could be inferred.

Means, SDs, and Pearson r correlations between the main variables are reported in Table 2. It can be seen that the mean value of cognitive social capital was much higher than that of structural social capital. Moreover, the mean value of cell phone exposure exceeded that of the 2 traditional media exposures. Notably, older adults demonstrated a strong overall intention to adopt protective behavior, with negative emotions at a moderate level. In addition, cognitive social capital was significantly correlated with protective intention (P<.01), whereas there was a differential correlation between various types of media exposure and the 2 forms of social capital and between various types of media exposure and protective intention.

In this study, a model consisting of 7 factors was developed (Figure 1). The construct reliability and validity of the model were evaluated based on the following criteria. As shown in Table 3, all the scales’ Cronbach α value and composite reliability value were >.7, indicating good construct reliability. In addition, the average variance extracted was >0.4, suggesting acceptable construct validity. Furthermore, the fit indices of the 7-factor model, which included all the variables in the study, were within the acceptable range after Modification Index correction (χ²/df=2.230, root-mean-square error of approximation=0.056, comparative fit index=0.912, and Tucker-Lewis index=0.901), indicating a good model fit and further supporting the validity of the measurement model.

First, this study examined the relationship between social capital and media exposure (Table 4). Controlling for gender, education, and income, older adults with higher levels of cognitive social capital were more likely to use television (B=0.24; P<.001) and cell phone (B=0.36; P<.001) but not newspaper and radio (B=0.10; P=.16). Structural social capital showed a significant positive association with newspaper and radio exposure (B=0.41; P<.001) and television exposure (B=0.28; P<.001). However, there was no statistically significant association between structural social capital and cell phone exposure (B=0.01; P=.92).

Second, the effect of social capital on protective behavior intention was assessed. The results are presented in Table 5. Cognitive social capital positively predicted protective behavior intention (B=0.52; P<.001; model 7), so hypothesis 1a was supported. However, there was no significant association between structural social capital and protective behavior intention (B=0.07; P=.21; model 8), so hypothesis 1b was not supported.

Next, this study evaluated the mediating effects of 3 types of media exposure between social capital and protective behavior intention (Table 5). According to model 9, the positive effect of cognitive social capital on protective behavior intention remained significant after the mediating variable cell phone exposure was added (B=0.43, P<.001). By contrast, after the mediating variables were added to the regression model of structural social capital and protective behavior intention, newspaper and radio exposure showed a positive effect on protective behavior intention (B=0.14; P=.01), television exposure (B=0.12; P=.01), and cell phone exposure (B=0.23; P<.001; model 10).

Bootstrap repeated sampling analysis (Table 6) examined the indirect effects and provided further insights into the mediating effect of media exposure between social capital and protective behavior intention. Mediated by cell phone exposure, the effect value of cognitive social capital on protective behavior intention was 0.05, with a 95% BC of 0.01-0.11, which does not contain 0, so hypothesis 2c was verified. By contrast, when newspaper and radio exposure acted as a mediator, the indirect effect value was 0.01, with a 95% BC of –0.003 to 0.04, so hypothesis 2a was not supported. Similarly, television exposure was not confirmed as a mediating variable between cognitive social capital and protective behavior intention, so hypothesis 2b was rejected.

In addition, when newspaper and radio exposure and television exposure were seen as mediating variables for the effect of structural social capital on protective behavior intention, the upper and lower limits of bootstrap 95% CIs did not contain 0, indicating that structural social capital was significantly associated with protective behavior intention through both mediating paths of newspaper and radio exposure and television exposure; thus, hypotheses 2d and 2e were supported. By contrast, cell phone exposure was not proved as a mediator between structural social capital and protective behavior intention, so hypothesis 2f was not supported.

Finally, this study tested the moderating effect of negative emotions between social capital and media exposure (Table 7). According to model 11, the interaction term between cognitive social capital and negative emotions had a significant positive effect on cell phone exposure (B=0.14; P=.03). However, according to models 12 and 13, the interaction term of structural social capital and negative emotions was not statistically significant for newspaper and radio exposure or television exposure. A further simple slope analysis of the moderating effect of negative emotions between social capital and media exposure (Figure 2) showed that, on the one hand, for research participants with higher levels of negative emotions (Mean+1SD), cognitive social capital had a significant positive predictive effect on cell phone exposure (simple slope=0.52; t₃₉₅=4.36; P<.001). On the other hand, for participants with lower levels of negative emotions, cognitive social capital positively influenced cell phone exposure to a lesser extent (simple slope=0.37; t₃₉₅=5.95; P<.001). In addition, the mediating effect of cell phone exposure on the relationship between cognitive social capital and protective behavior intention tended to increase at all 3 levels of negative emotions. Thus, cognitive social capital was more likely to enhance older adults’ protective behavior intention by increasing their cell phone exposure as older adults’ levels of negative emotions rose. Therefore, hypothesis 3 was partially supported.

Figure 3 illustrates the correlations and effect values among the variables used in this study. Notably, cognitive social capital exhibited a significant correlation with protective behavior intention (P<.001), with the mediating effect of cell phone exposure established between the 2 variables (P<.001). In addition, the mediating effect of newspaper and radio exposure and television exposure between structural social capital and protective behavior intention was established (P<.001), whereas negative emotions played a moderating role in the relationship between cognitive social capital and protective behavior intention (P<.001).

This study aimed to investigate the applicability of the SIM in the social context of the post–COVID-19 epidemic era in China. The study focused on assessing the reliability of social capital in influencing health protection intentions through media exposure, with a specific emphasis on the older Chinese population. This study categorized social capital into cognitive and structural dimensions and examined media exposure from the perspective of 3 main media types: newspaper and radio, television, and cell phone. Considering the risk characteristics of COVID-19 messages and the influence of selective exposure, this study also tested whether negative emotions play a moderating role between social capital and media exposure.

The results of this study support the SIM by indicating that potential factors at the personal, interpersonal, and social levels can explain individuals’ intention to adopt health behaviors. First, cognitive social capital was shown to promote health behavior intention, whereas structural social capital did not have a significant effect on health behavior intention, which is consistent with the findings of De Silva et al [40] on social capital and mental health outcomes. However, the underlying reasons for these differences were not fully elucidated and should continue to be explored in depth in future research.

Second, the mediating role of media exposure between social capital and health behavior intention was also emphasized, validating the SIM pathway. Interestingly, the results revealed that the 2 types of social capital influenced protective behavior intention through different patterns of media exposure. Cognitive social capital positively influenced cell phone exposure, indicating that older adults with higher levels of trust and reciprocity were more inclined to use new media for health information seeking and adopt protective behavior. By contrast, structural social capital predicted health behavior intention through newspaper and radio exposure and television exposure, with a stronger mediating effect observed for newspaper and radio exposure. This implies that older adults with more social engagement and larger social networks prefer traditional media, engaging in more offline information activities. Previous studies have suggested that people tend to communicate information across all (or many) media channels [76]. Therefore, most studies have not categorized media channels when examining the relationship between social capital and health information seeking. However, according to the findings of this study, it may be advantageous for health communicators to target different characteristics of the population using different media channels when disseminating messages rather than solely focusing on increasing people’s social capital through welfare measures.

In addition to the association among social capital, media exposure, and protective behavior intention, the results of this study illustrate the importance of negative emotions. The higher the level of negative emotions, the more pronounced the positive effect of cognitive social capital on cell phone exposure, validating the findings of previous studies that negative emotions have a positive effect on information seeking [77,78]. By contrast, the results of this study indicate that negative emotions did not cause interference during the influence of traditional media exposure, such as newspaper and radio exposure and television exposure, among older adults. As numerous studies have pointed out, the intensity of negative emotions during pandemics interacts with the impact of efficiently disseminated media [79,80]. Consistent with this, this study also suggests that new media exhibit a more pronounced correlation with negative emotions than traditional media. However, the specific direction of influence between negative emotions and media exposure remains unclarified in this study. Although some studies indicate that negative emotions may prompt individuals to seek information from new media [52,81,82], a substantial body of evidence also suggests that an overload of information can significantly induce panic, anxiety, and frustration [83-85].

Although the threshold of emotion varies from person to person [86] and it is difficult to define an optimal level that can be generalized, future research should explore the differences in the effects of negative emotions in specific social and media environments, as well as the reasons that trigger such differences. As a result, public sentiment can be better appeased during health crises, thus encouraging healthy behavior. Furthermore, the findings also suggest the need to explore other factors influencing health communication and behavior to extend the SIM to improve its applicability in relevant research.

First, this study used an web-based questionnaire to recruit the sample, which may have introduced a bias in the sample. It is possible that a significant portion of the sample had access to and preferred using the internet for information seeking. Although efforts were made to involve relatives in completing the questionnaire on behalf of older adults who did not use the internet, the effectiveness of this approach in capturing the perspectives of non–internet users could not be fully controlled or measured. Thus, the generalizability of this study may be limited to this sampling method.

Second, the cross-sectional nature of this study, in which participants’ responses were recorded at a single point in time, restricted the ability to draw causal conclusions.

The relationships observed between the variables may be subject to reverse causality or the influence of unmeasured confounding variables. However, the SIM assumes a directional relationship among social factors, information behavior, and health outcomes [12], and this study did demonstrate a pattern of association in this direction. Future research can use longitudinal designs or experimental approaches to better understand the temporal and causal dynamics among social capital, media exposure, negative emotions, and health behavior intention.

Third, this study attempted to include the element of negative affect to extend the SIM. In fact, there may be additional variables that may influence information behavior and health outcomes. Future research should consider incorporating additional relevant factors to enhance the comprehensiveness and applicability of the SIM.

This study partially supports the SIM by indicating that social capital can directly influence health-protective behavior intention. In addition, the results suggest that media exposure can act as an additional bridge in this pathway of influence and that negative emotions can contribute to increased older adults’ engagement in epidemic information communication via cell phones. Therefore, in the context of the post–COVID-19 epidemic era, health communicators need to focus not only on message delivery but also on developing targeted health communication strategies. Consequently, communication effectiveness could be enhanced for older adults who are in health vulnerability as well as information vulnerability. Future research should continue to explore theories and applications of the SIM to discover more intervention points to promote universal health behaviors.