The global outbreak and rapid spread of the COVID-19 pandemic has led the world’s public health and safety systems to face great challenges. As of April 16, 2020, the COVID-19 pandemic has affected at least 211 countries, with 2,362,704 confirmed cases and more than 165,006 deaths globally, while China has confirmed 82,367 cases and the cumulative number of deaths was 3342 [1]. The disease rapidly spread in China because of the massive population migration (ie, Chunyun, also known as the Spring Festival travel rush, which refers to a human migration during this festival) and lack of prevention and control information during the early stage of the pandemic. Moreover, the World Health Organization (WHO) has not only signaled the health risks of COVID-19 but has also labeled the situation as an infodemic, due to the amount of information, both truth and rumors, circulating around this topic [2,3].

Although public health information dissemination represents an exciting combination of broadcasting, sharing, and retrieving relevant health information [4,5], the COVID-19 infodemic did not come as a surprise [6]. The massive growth of health information on the internet was seen to be a real problem [4]. Too much information makes it difficult to find trustworthy sources of information and may harm people’s health. Therefore, the quality of online health information is essential, especially when truth and rumors were still intertwined, which caused the infodemic. However, in the process of fighting COVID-19, the WHO and health authorities worldwide have been working closely with social media platforms, including Facebook, Google, Twitter, and YouTube, to provide evidence-based information to the general public in order to actively counter the rumors that have been circulating [7]. Prevention and treatment information about COVID-19 continues to spread and has an influence on populations.

At the same time, the lack of transparent, timely, and effective risk communication by health authorities regarding this emerging infectious disease in its early stage failed to bring about the appropriate level of public awareness and behavioral responses, such as avoidance of mass gatherings and personal protection in China, Europe, and the United States [8]. Online guidelines providing information for prevention and treatment are inconsistent in adapting to new knowledge, and changing or conflicting information can also confuse the public [9].

Since the beginning of the COVID-19 pandemic, information consumption has increased rapidly and significantly [10]. The new generation of health care consumers consists of an informed public, gleaning truth and rumors about health information with both positive and negative effects on themselves [11]. During periods of SARS spread, most people obtained SARS information from television [12]. By the time the Zika virus emerged in 2015-2016, Google Trends showed a significant increase in public searches on the internet related to the Zika virus [13], and the number of searches on video platforms, such as YouTube, had also increased rapidly [14]. Internet-based platforms that people utilized became diversified. During the COVID-19 pandemic, social distancing and stay-at-home restrictions caused the public to be fully exposed to social media; during this time, people actively searched or passively received a large amount of health-related information to prevent and treat diseases. In China, social media platforms, such as WeChat and Douyin (ie, TikTok), played an essential role in obtaining information after the virus began spreading [15]. However, studies have confirmed that over one-quarter of the most viewed YouTube videos about COVID-19 contained rumors, and these reached millions of viewers worldwide [16]. Therefore, determining whether people can use the prevention and treatment information they find on social media is critical in this pandemic.

Moreover, identifying health information from prevention and control measures is a major blind spot for the public. The public are partly responsible for selecting and filtering trustworthy health information [17]. Research shows that more than half of respondents trust almost all information online [18], and people are more likely to believe health rumors because of basic safety needs [19]. As the core part of health behavior theory, behavior intention is the subjective possibility of engaging in certain behavior. Some studies have confirmed that variables in the health belief model (HBM) can materially affect information adoption intention, and this process will affect subsequent health behaviors [20].

Therefore, if concern about identifying trustworthy information is reflected in the global population, we believe that the influencing factors in the subjective and objective aspects of engaging in that behavior may affect information adoption. Motivated by previous studies, this study was based on the information adoption model (IAM) and the HBM. The aim of this study was to explore the influencing factors and paths during public health events that affect the public’s intention to adopt online prevention and treatment information under the infodemic. We aim to provide a basis for decision making and policy suggestions in order to deal with online health information governance in the internet era. Moreover, this study adds to the sparse literature on information adoption.

Online questionnaires were powered by the survey platform WJX (Changsha Ranxing Information Technology Co), whose web application was embedded into social media platforms from March 24 to April 5, 2020. The electronic version of the questionnaire was uploaded to the WJX web application; respondents (ie, Chinese people in China) could fill in, submit, and share the questionnaire using a Quick Response (QR) code or using a forwarding link issued by the WJX web application. The data were collected using snowball sampling through repetitive one-to-many sharing on social media, a nonprobability sampling method; there were 528 respondents from 30 provinces in China. We gathered data using an online survey because of public space restrictions and because netizens were potentially exposed to the infodemic. After eliminating invalid responses through data filtering, 501 valid questionnaires out of 528 remained (94.9% validity rate). Table 1 shows the demographic characteristics of the sample population.

Out of the 501 valid responses, 216 (43.1%) were from male participants and the other 285 (56.9%) were from female participants. Further, the majority of respondents were between the ages of 20 and 39 years (276/501, 55.1%). Most of the respondents had earned a bachelor’s degree (256/501, 51.1%), which indicated a high level of education among respondents. In terms of occupation, the largest group was students (104/501, 20.8%), followed by professionals (75/501, 15.0%) and enterprise managers (67/501, 13.4%).

We conducted quality control through the survey platform and via the investigators. Once on the platform, respondents were invited to fill in the questionnaire voluntarily. Each Internet Protocol (IP) address, computer, or username could only be used once. Also, there were various filtering rules for invalid answers, such as spending too little time on a questionnaire and trap rules to filter out random answers.

After the surveys were submitted, the respondents were screened by the investigators to retain the valid questionnaires. We excluded the following respondents: those who failed the attention check, where the answers to all the questions were the same or cyclical; those with response times of less than 120 seconds; those under 18 years old; and non-Chinese residents. Also, we checked the consistency between the IP address and the selected region, and questionnaires were eliminated if the IP addresses were not consistent.

The study instrument was modified from those in the relevant existing literature. Measurements and scales were translated into the appropriate Chinese versions to ensure the completeness and accuracy of instruments. After the repeated pretest, the final questionnaire was translated back into English, and the main semantics were not changed, indicating a strong correlation with the original English questionnaire (see Table 2 [21,29,34,44-52]). The instruments were measured using a 5-point Likert scale, ranging from 1 (highly disagree) to 5 (highly agree). All dimensions included three items, except the information characteristics construct, which included five items.

This study was approved in writing by the Medical Ethics Committee of Capital Medical University (No. Z2019SY014) and all participants gave informed consent.

After data collection, the two-stage procedure of structural equation modeling (SEM) was applied to conduct data analysis [41]. The first procedure examined scale validity from the measurement model by confirmatory factor analysis (CFA), while the second procedure interpreted hypotheses testing by the structural model. Both SPSS Statistics for Windows, version 19.0 (IBM Corp), and SPSS Amos, version 24.0 (IBM Corp), were adopted as the tools for analyzing the data.

Based on the results of the CFA and modification index of indicator variables, six standard model fit criteria were used to assess the model’s overall goodness of fit: ratio of the chi-square value to the degrees of freedom (χ²/df), goodness-of-fit index (GFI), comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square residual, and root mean square error of approximation (RMSEA).

As shown in Table 5, comparison of all fit indices with their corresponding recommended values provided evidence of a good model fit: χ²/df values were between 1.0 and 3.0; GFI, CFI, and TLI were all greater than 0.9; and RMSEA was smaller than 0.08. This demonstrated that the measurement model exhibited a tolerably good fit with the data collected [56].

Based on the results of SEM, the fit indices of the structural model are shown in Table 5. Under the same criteria, the structure model fits the observed data as well. Meanwhile, the estimated results of the structural model provided the path coefficients shown in Table 6. Among the eight hypotheses, six paths were supported based on the valid data, significant at the P=.01 level, while the remaining two paths were rejected according to SEM (ie, Hypothesis 2 and Hypothesis 5a).

Adoption intention was predicted by information characteristics (β=.266, P<.001) and perceived usefulness (β=.565, P<.001), which jointly explained 66.8% of the adoption intention variance. Information characteristics (β=.244, P<.001), perceived drawbacks (β=–.097, P=.002), perceived benefits (β=.512, P<.001), and self-efficacy (β=.141, P<.001) jointly determined perceived usefulness and explained 81.1% variance of perceived usefulness. In addition, perceived usefulness had the most significant influence on adoption intention. It revealed that perceived usefulness was an important indicator of adoption intention. Perceived benefits had the most significant direct influence on perceived usefulness, followed by information characteristics, while perceived drawbacks had a relatively low path coefficient, which indicated a negative effect at the same time. Since the rejection of Hypothesis 5a, it means that perceived usefulness had a complete mediating effect on self-efficacy to adoption intention. This result confirmed perceived usefulness as an intermediary variable. Surprisingly, Hypothesis 2 was not supported in this study. The path coefficients supported six of all hypothesized relationships (see Table 6, Figure 2, and Multimedia Appendix 1).

To test the indirect effects, bias-corrected bootstrapping with 5000 iterations was implemented to obtain the structural model path significance levels for indirect effects [58]. Bootstrapping, a nonparametric approach, was superior to other approaches in testing mediation models because it does not assume multivariate normality [59,60]. Table 7 shows that perceived usefulness fully mediated the effect of perceived drawbacks, perceived benefits, and self-efficacy on adoption intention, whereas perceived usefulness partially mediated the effect of information characteristics on adoption intention.

This study set out to determine the factors affecting people’s intention of adopting health information on the internet, and the IAM-HBM model was applied. In general, this combined model provided an excellent fit to the data. Our research supports perceived drawbacks, perceived benefits, self-efficacy, perceived usefulness, and information characteristics as factors associated with the intention to adopt online prevention and treatment information to prevent an epidemic in the context of COVID-19.

In the study, information characteristics can strengthen perceived usefulness to adopt health information. In addition, the quality and the source of information influence the perceived usefulness and indirectly impact adoption intention. People are inclined to take in evidence-based information rather than misinformation on the internet [61,62]. Therefore, health communication should include more evidence-based information and should meet the public’s health demand [7]. Swamped by information on the internet, if the information is expressed more reliably or if its publisher is more professional and authoritative, the public will have a more robust perception of the usefulness of the information, thus increasing individuals’ willingness to adopt the information. At the same time, in the face of information overload, especially where some information constituted rumors, the public, who lack professional knowledge, will balance the outcomes of adoption information behavior. Many online repositories full of valuable content are underutilized, becoming “information junkyards” [63], and during the COVID-19 pandemic, an infodemic could be triggered by rumors. As information flow has improved, infodemic prevention and management using facts and evidence can mitigate the next infodemic [64,65].

The public’s perceived drawbacks, perceived benefits, and self-efficacy had significant influences on perceived usefulness. Members of the public conduct a cost-benefit analysis before adopting healthy behaviors, in which they weigh the effectiveness of the adoption against the possible cost and risk of time-consuming impediments. As a result, the more benefits and fewer drawbacks one perceives, the more that the health benefits of the adoption behavior outweigh the health risks, resulting in higher perceived usefulness by individuals in considering online health information to prevent COVID-19.

However, only by improving the public’s media literacy and their ability to perceive information can they correctly recognize the obstacles and benefits. At the same time, greater health literacy can improve public health self-efficacy, resulting in increased confidence in adopting healthy behaviors or changing bad behaviors. Many people have limited health literacy [66]; health communication and education are the most cost-effective means to improve health literacy [67]. Therefore, in the release of COVID-19-related health information, attention should be paid to improving public information capacity. For example, officials could actively hold health lectures and disclose health information, and relevant experts could improve the public’s capacity regarding obtaining information.

Simultaneously, with the continuous enrichment of social media, it is difficult for social communication based on social media to achieve full, comprehensive, and balanced transmission of information. We need to avoid falling into information cocoons and confirmation bias [68], and we need to measure the quality of information from an overall perspective.

Surprisingly, our analysis did not support the hypothesis that social influence is positively associated with perceived usefulness of COVID-19 prevention and treatment information. This finding was counterintuitive, and previous research showed that social networks positively affect people by encouraging them to adopt different health behavior intentions [69].

We think that government press conferences and news-based public opinion during the COVID-19 pandemic in China have weakened social influence on people’s perception and acceptance of health information for the following reasons. First, since the SARS epidemic in 2003, the Chinese government has reformed the news release concept and system. In response to the emergencies, the Chinese government issued a series of policy documents and established the State Council Information Office’s three-level news spokesman system for all central ministries and provincial-level people’s governments [70]. Second, governments at all levels use various channels to publicize, in a timely manner, the prevention and treatment information regarding COVID-19, the latest situation regarding the pandemic, and other public concerns, providing the public with a low threshold and low-cost direct information feedback channel [71]. The government also invited medical experts, such as Dr Zhong Nanshan, Head of the National Team for Control of Novel Coronavirus, to communicate with the public, and this strategy gained public trust [72]. Finally, China’s political system practices high-quality and high-efficiency unified decision making, and they have strict controls over content such as social media [73]. For example, Facebook, Twitter, and YouTube are not allowed in mainland China [74], and information monitoring and timely rumor controls are also available within popular social media platforms, such as WeChat and Sina Weibo [75]. Therefore, the government exerted its full influence during the pandemic to position itself as the primary influence [76].

However, social media is both a source of the infodemic and a public health tool [77]. Therefore, it is necessary to include social media platforms in public information dissemination; rethinking the role of public communication will also be necessary to assume corresponding responsibility during the pandemic. The responsibility is not only to delete information but, as much as possible, to ensure the diversity of the information environment to a sufficient degree; this will enable high-quality public content and thereby increase public participation [64].

In a public health emergency, the online infodemic forces the public to negotiate with prevention and treatment information. By integrating IAM and HBM, this study provided the insight and understanding that perceived usefulness and adoption intention of online health information could be influenced by information characteristics, people’s perceptions of drawbacks and benefits, and self-efficacy. Moreover, people also exhibit proactive behavior rather than reactive behavior. Thus, we should consider these factors to help the informed public obtain useful information via two approaches: one is to control the quality of information and the other is to improve the public’s capacity to obtain information, in order to promote trusted information and to fight misinformation. This will, in turn, contribute to saving lives as the pandemic continues to unfold and run its course.

We administered the questionnaire survey during the stage of the pandemic in mainland China when it was under control, which was when the outbreak in China had passed the initial panic stage. People at different stages of the pandemic may have been influenced differently by the influencing factors. Therefore, although we found that social influence had no significant effect on perceived usefulness of information, a more comprehensive future study is suggested to explore whether this is due to social context, stage of the pandemic, or other factors. Moreover, this cross-sectional study was conducted using the WJX web application, and sample populations had a certain amount of experience in filling out online questionnaires and internet use. Therefore, a more comprehensive future study is suggested to include offline and online participants to expand the framework’s application scope.