Understanding the Scientific Topics in the Chinese Government’s Communication about COVID-19: An LDA Approach

Abstract

The communication of scientific topics can play a key role in the fight against misinformation and has become an important component of governments’ communication regarding COVID-19. This study reviewed the Chinese government’s COVID-19 information sources and identified the patterns of science communication models within them. A corpus of science-related content was collected and coded from 1521 news briefings announced by the Chinese government. An LDA (latent Dirichlet allocation) topic model, correlation analysis, and ANOVA were used to analyze the framing of the scientific topics and their social environmental characteristics. The major findings showed the following: (1) The frames in the Chinese government’s communication of scientific topics about COVID-19 had three purposes—to disseminate knowledge about prevention and control, epidemiological investigations, and the public’s personal health; to make the public understand scientific R&D in Chinese medicine, enterprises, vaccines, treatment options, and medical resources; and to involve citizens, communities, and enterprises in scientific decision making. (2) The frames were correlated with the public and media concerns. (3) The frames varied with the different levels of officials, different types of government agencies, different income regional governments, and different severity levels of the epidemic. (4) The topics concerning sustainability science were more correlated with public and media concern. In addition, we propose several suggestions for building sustainable communication approaches during the pandemic.

1. Introduction

Since the beginning of 2020, the COVID-19 pandemic has shifted from East Asia to Europe and then to North America [1]. Governments around the world had to make plans and implement immediate communication strategies to explain their response actions and public health policies during the pandemic [2]. However, some dilemmas have occurred with governments’ communication about COVID-19, such as a lack of transparency causing public distrust [3], the failures of spokespersons causing media dissatisfaction [2], and scientists being in the media and public’s focus, facilitating governments to communicate scientific issues during the pandemic. As COVID-19 continues to spread around the world, governments face the challenge of the rapid spread of far-reaching information related to the virus, which is often false or misleading. The communication of scientific topics can play a key role in the fight against misinformation [4].

Scientists [5,6], the media [7,8], and the government [9] all play important roles in the communication of scientific topics, of which the government has more authority and is therefore more likely to be trusted by the public [10]; thus, scientific topics have become an important component of governments’ communications about COVID-19. Research on the purpose of the governments’ communication of scientific issues has also become very meaningful. The popularization of science, public understanding of science, and public participation in science are widely known as the three science communication models in current research [11]. According to those models, there may also be three purposes behind governments’ communication of scientific topics. However, there are few literature sources exploring the application of these models to government communication. The goal of this study was to review the Chinese government’s information on COVID-19 to identify the patterns of their science communication models and the purposes of communicating scientific topics in these models.

An understanding of the Chinese government’s science communication models and purposes can be achieved through two dimensions. The first dimension is the establishment of agendas. The agenda-setting theory, which originated from the pseudo environment theory proposed by Lippmann in 1922 [12] and expanded by McCombs and Shaw in 1972 [13], refers to how the media has affected the attention of the public and policymakers through the setting of issues. The government’s models and purposes are reflected in the setting of scientific topics; that is, the science agenda of the government. The second dimension is the establishment of approaches. Framing is an approach that is often used by the media, governments, or other institutions to communicate an issue; it refers to selecting some aspects of a perceived reality and making them more salient in a communicating text, in such a way as to promote a particular problem definition, casual interpretation, moral evaluation, and/or treatment recommendation for the item described [14]. Dekker and Scholten (2017) [15] suggested that the agenda of a government can be framed. The government’s science agenda (i.e., their science communication models and purposes) is also achieved by the approach of framing in this paper.

The hypotheses of this study were that the framing of scientific topics in the government’s communications about COVID-19 in China fit the three types of science communication models and their functions, and they are modulated by parameters and elements such as sources, audiences, and momentum. The quantitative research data used to verify the hypotheses in this study came from the news briefings announced by the Chinese government, which not only played an important role in the sharing of case numbers, public service announcements, the publication of coronavirus restrictions, etc., but also represented the first means for the government to communicate with the media and public during the pandemic. All the contents of central and provincial governments’ COVID-19 news briefings from 22 January 2020 to 27 August 2021 were selected as a sample to analyze the government’s communication and the scientific topics contained within using the LDA topic model.

2. Literature Review

2.1. Scientific Topics in Government Communications about COVID-19

2.1.1. Government Communications about COVID-19

There are significant differences between government communications and private sector communications [16]. This paper will explore the role of the government in communicating scientific topics regarding COVID-19. Many scholars have explored government communications about COVID-19 in different countries. Through literature studies, it has been found that these topics differ in different political systems. In pluralistic democracies and multi-party political systems, partisan competition is always reflected in communication topics. For example, a study from the United States found that conflicting health messages (whether masks are useful, what the death rates are, whether hydrochloroquine has potential for treatment, and the controversy over the severity of the epidemic) were the main communication topics [17]. Under a political system with one-party rule and centralization, the communication of COVID-19 pandemic information by other institutions outside the government, such as the media, is less influential; instead, it is mainly framed by the structure of government communication, with more consistent mainstream information [18]. Despite the differences in the topics disseminated by governments in different political systems, there are similarities in the communication of scientific topics, as it plays an important role in dispelling rumors and identifying science misinformation.

2.1.2. Government Framing of Scientific Topics Surrounding COVID-19

It is of great importance to understand how a government constructs its crisis frames [19]. Crisis framing is used to show some aspects of a crisis to the audience, and portray an image of the crisis to the public [20]. Although different cultural resonances, moral values, and institutional sources will affect whether the public accepts these crisis frames, it is undeniable that this theory provides for the feasibility of quantitative analysis in the study of communication content [21,22]. In fact, in any public communication, framing is an important tool for the government. Framing analysis allows researchers to elucidate how the government refines the crisis topics they want the public to be aware of.

Many scholars have studied crisis frames. For example, some scholars believe that institutions usually use four frames for public communication during a crisis: responsibility attribution, conflict, economic consequences, and human interests [23,24]. Some scholars have also studied the framing of public health crises. Shih et al. (2008) [25] suggested that during a public health epidemic, the commonly used frames include taking measures, new evidence, reassuring the public, and the uncertainty of the epidemic. Higgins et al. (2006) [26] proposed that there are five frames in public health events: disease detection, disease prevention, medical services, risk factors in lifestyle, and scientific discoveries. Through the existing research on the crisis frames of public health events, we found that many are related to scientific topics, such as disease prevention, disease detection, treatment measures, scientific discovery, etc. In the current study, we found that three types of scientific topics were framed by the government in its communication about the COVID-19 pandemic.

The first type of scientific topic is popularizing scientific knowledge among the public. A study by Ye et al. (2021) concluded that the government wants to let the public know about the causes of diseases, virus transmission routes, and disease prevention [27]. Likewise, Liao et al. (2020) [28] and Li et al. (2021) [29] concluded that the government wants to spread knowledge about the causes of diseases, virus transmission routes, and disease prevention, in addition to information about how to deal with rumors. Górska et al. (2022) [30] suggested that the government also wants to disseminate public health knowledge on social media. The purpose of this kind of topic is in line with the science communication model of popularizing science, which refers to the government’s unilateral desire to spread knowledge to the public [31], with the assumptions that the public is lacking scientific knowledge about the COVID-19 pandemic [32,33] and that the government is needed to improve the public’s level of knowledge.

The second type of scientific topic is making scientific research understandable to the public. The study by Ye et al. (2021) [27] argued that the government wants the public to understand the development of vaccines, specific drugs, and effective treatments for diseases. Li et al. (2021) [29] argued that access to medical resources granted by the government can assist the public in understanding scientific treatment protocols. The purpose of this kind of topic is in line with the public understanding of science communication models, which refers to occasions when the public has doubts about the scientific knowledge disseminated by the government in a one-way communication [34]; hence, the government must have a two-way interaction and dialogue with the public so that the public can fully understand the creation process of science—that is, scientific research—thus increasing the public’s trust in science [35].

The third type of scientific topic is involving the public in scientific decision making. Through the analysis of official postings from government agencies on social platforms, Liao et al. (2020) [28] found that the government called for public and community participation in the prevention and control of the epidemic. Beattie and Priestley (2021) [36] argued that the New Zealand government called on the public to participate in the government’s epidemic prevention actions based on their analysis of the daily news briefing announcements by the prime minister and senior officials of health agencies. Li et al. (2021) [29], who analyzed the government’s social media platform postings, argued that the government called for public participation in the epidemic response. The purpose of this kind of topic is in line with the science communication model of public participation in science, which means that the core of the government’s communication is no longer to popularize scientific knowledge or to make the public understand the correctness of scientific knowledge, but instead is to work with the public on the production of knowledge—that is, to involve the public democratically in scientific decision making [37].

Table 1. Science communication models and scientific topics.

Communication Models	Communication Patterns	Possible Scientific Topics in Current Research	Purposes of Government Communications
Popularization of scientific knowledge	One-way and disposable	Etiology, virus transmission pathways, and disease prevention [27]; knowledge to dispel rumors [28]; public health knowledge [30];	Inculcation and popularization of scientific knowledge.
Public understanding of science	Consultation	vaccines R&D, specific drugs R&D, and effective treatments [27]; medical resources [29].	Explanation of the process of scientific knowledge, namely scientific research.
Public participation in science	Multi-subject participation	Public participation [28,29,36]; community involvement [28].	Co-production of scientific knowledge and participation in scientific decision making.

summarizes the existing literature on science communication models, communication patterns, the framing of scientific topics, and the purposes of government communications.

According to the current research on scientific topics in government communications about COVID-19, we put forward the first research hypothesis:

H1. 

The Chinese government’s frames for its official communications of scientific topics surrounding COVID-19 fits the three types of science communication models and their functions.

2.1.3. Topics of Sustainability Science in Communications about COVID-19

Sustainability science aims to help societies across the globe address environmental and health crises and risks, ranging from poverty to climate change to health pandemics. It is believed that communication makes fundamentally important contributions to the development of sustainability science [38].

Scholars also believe that sustainability science is an important part of scientific topics surrounding COVID-19. Cernicova-Buca and Palea (2021) [39] said that health is a fundamental right of every human being, as well as a major element of sustainability. Bodenheimer and Leidenberger (2020) [40] suggested that the outbreak and spread of COVID-19 was closely related to sustainability: a lack of ecological sustainability contributed to the coronavirus outbreak, a lack of economic sustainability allowed for its rapid and global spread, and a lack of social sustainability added to its severity. The explanation of sustainability science to the public might be an important part of the Chinese government’s communication regarding COVID-19. Thus, our second hypothesis was put forward:

H2. 

Some frames in the Chinese government’s official communications of scientific topics about COVID-19 were related to sustainability science.

2.2. The Social Environment Characteristics of Government Communication

Government communication is one governmental behavior that is associated with organizational behavior. Studies have concluded that the social environment of organizational behavior can be divided into a narrow sense and a generalized sense [41]. The social environment of organizations in the narrow sense refers to the environment in which organizations survive and develop, specifically the network of relationships between organizations and the public, i.e., the public relations of organizations. In a generalized sense, the social environment of organizational behavior includes a wide range of socio-political and economic environments that are also closely related to the development of the organization [42]. Therefore, this paper presents the following literature review from those two senses.

2.2.1. Agenda Setting: Interaction with the Public and Media

Governments want to increase public concern and involvement in scientific topics related to the COVID-19 pandemic, and these topics may be reflected in governments’ agenda setting for science [43,44]. The theory of agenda setting emphasizes the interaction between the media agenda, public agenda, and government agenda [13]. Dai et al. (2021) [45] suggested that the public is not only a simple information provider or receiver, but also that public attention and its emotional changes can affect the government’s agenda setting. Therefore, the social environmental characteristics of government communication in a narrow sense means that the government can interact with the public and media.

Many scholars have studied the relationship between government communications and the public’s reaction. Calderón et al. (2021) [46] analyzed the credibility of government communications among the public and concluded that public trust in government communication messages directly affects the effectiveness of political communication. Zahariadis et al. (2021) [47] also identified a significant correlation between government communication and public sentiment.

Many other scholars have studied the interaction between government communication and the media. Mandl and Reis (2022) [48] concluded that the media has a role in helping the government to improve communication efficiency, monitor and respond to rumors, and promote public participation. Langer and Gruber (2021) [49] suggested that the media has the function of expanding and maintaining attention to events, which helps to attract the attention of the government and make it take action.

The consensus among these studies is that there is a significant relationship between government communication and the concern of the public and the media, which also fits the agenda setting theory. Therefore, we proposed a third research hypothesis:

H3. 

The frames in the Chinese government’s communication of scientific topics surrounding COVID-19 were correlated with the (a) public and (b) media’s concern.

2.2.2. Modulating Factors: Sources and Severity

From the perspective of social environmental characteristics in a generalized sense, government communication may be affected by institutional sources and the severity level of the crisis.

Many scholars believe that when communication comes from different institutions, their frames are significantly different. The different levels of government are considered as an important reason for the differences in framing. Some scholars have analyzed the official communication framings of different levels of government and found significant differences. Li et al. (2021) [29] concluded that the central government gave more importance to the emotional mobilization frame, while local governments gave more importance to the guiding frame and the adjusting information frame. Some scholars have also explored health agencies as having a special role in government departments in response to an outbreak, suggesting that they are seen as separate from other government departments. For example, Wang et al. (2021) [50] concluded that the state health agencies accounted for a greater share of frames on crisis severity, disposition strategies and guidance information, and inaccurate information management. However, fewer scholars have examined the differences between health agencies and other types of agencies in topics regarding the spread of the COVID-19 epidemic, which is a question to be explored in this paper. There are also scholars who have studied regional differences in communication, as the economic development levels are different between regions. A study on information releases by urban websites in various regions of Romania during the COVID-19 pandemic found that the characteristics of the websites in terms of security, usability, content, services, and citizen participation were different, and this was also related to regional economic development [51].

The severity of the epidemic has also been considered in studies related to political communication, where it has mainly been used to analyze changes in framing. For example, Liao et al. (2020) [28] argued that as an epidemic becomes more severe, there is a greater inclination toward the dissemination of emotionally mobilizing frames.

In order to explore the factors that modulate the government’s framing in the communication of scientific topics surrounding COVID-19, we proposed the following research hypothesis:

H4. 

The frames used by the Chinese government in the communication of scientific topics surrounding COVID-19 varied at different levels, including (a) officials, (b) different types of government agencies, (c) different income regional governments, and (d) severity of the epidemic.

2.3. Research Model

The research model of this paper was established according to the literature review and research hypotheses and is shown in Figure 1.

3. Materials and Methods

Figure 2 shows the research objects and methods of this paper. The boxes with solid lines are research methods, and the boxes with dotted lines are research objects.

3.1. News Briefing Content Collection

News briefing contents were chosen from 21 January 2020 to 27 August 2021, because the first news briefing for the COVID-19 pandemic in China was held by the Guangdong Information Office on 21 January 2020. The last briefing collected was at the end of August 2021—when China had entered a state where the epidemic prevention policy was mature—and the epidemic occasionally repeated in various places but the overall situation tended to be stable.

The materials of 1521 news briefings were obtained from the websites of China’s State Council Information Office and the provincial governments’ information offices. We referenced all news briefings about COVID-19 from the State Council and local governments of 31 provinces and autonomous regions in the mainland from 22 January 2020 to 27 August 2021, having obtained a total of 1521 briefings (1239 were text and 282 were videos that were translated into text). The website of each office and the number of news briefings from each region are shown in Appendix A.

3.2. Establishment of a Corpus Concerning Science-Related Content

The following three steps were carried out to establish a corpus concerning science-related content.

3.2.1. Import Materials into NVivo Software

NVivo is a software that supports qualitative and hybrid research methods and can play an important role in text coding. In our study, the first step of establishing the corpus was to import the information from the 1521 briefings obtained in Section 3.1 into NVivo.

3.2.2. Training Coders

In order to avoid the subjective bias of coders, two trained coders were used to code at the same time. The training content was how to learn about science-related content from the news briefing materials. The specific method is shown in Figure 3. First, manual annotation was carried out for each news briefing, and the science-related contents were extracted into the corpus as analyzed pieces. The content of several news briefings contained multiple analyzed pieces, some of which also did not include any pieces concerning science.

The standard of being science-related was met if the text content contained the following three types of information: describing the various channels through which the government propagates scientific knowledge to the public; describing how the government uses scientific research methods and equipment to make a series of activities, such as investigation and research, experimentation, and trial production, so that objective concepts can be understood further; and describing texts regarding the government’s invitation to the public and other organizations to participate in the scientific prevention and control of the epidemic. A coding sample is shown in Appendix B.

After the training, the two coders were required to complete the coding of 50 materials independently. If the coincidence rate of the two coders reached more than 90%, the training was considered to be successful; if not, the training was continued, and the coders were allowed to code independently again. After two rounds of training, the code coincidence rate reached more than 90%.

3.2.3. Coding of the Science-Related Content

After completing the training, the two coders finished coding the materials from the 1521 news briefings in NVIVO according to the method shown in Figure 2. After the coders completed all the coding, the texts encoded as science-related content by the two coders were selected for the corpus. A total of 1664 analyzed pieces from the Chinese government regarding COVID-19 communications were obtained.

3.3. Measuring the Framing of Scientific Topics: LDA Model Analysis

Two methods are generally applied for framing analysis: content analysis [52] and topic modeling [27,29,53]. Content analysis has to be carried out manually, which makes it difficult to avoid the subjective bias of coders on the one hand, and hard to analyze a large number of samples on the other. Thus, the use of topic modeling for the automatic clustering and determination of the topic identity of texts has emerged as an alternative method [54].

A topic model is a statistical model that clusters the latent semantic structure of a dataset through unsupervised learning. From the 1980s to the early 2000s, scholars achieved many advances in their research on topic models. Salton and Buckley (1988) [55] proposed the TF-IDF model based on the term frequency for the data mining of large-scale document sets. Deerwester et al. [56] subsequently proposed the LSA (latent semantic analysis) model in 1989, which perfected the problem of document dimensions in the TF-IDF model. Blei et al. (2003) [57] proposed the LDA (latent Dirichlet allocation) model; due to it having stable parameters, not being affected by the number of documents, and having strong abilities in potential semantic mining and normalization learning, this model is widely used in topic discovery [58,59,60]. Therefore, this study used the LDA topic model to analyze the framing of scientific topics in the Chinese government’s communications about COVID-19 and to test H1 and H2; the method consisted of the following four steps.

3.3.1. Preprocessing of Text

The text was preprocessed through word segmentation, the creation of a user dictionary, the selection of national words, and the removal of stop words before the formal topic model analysis. First, the NLPIR-ICTCLAS Chinese word segmentation system was used to segment the text, and two types of words were selected: nouns and verbs. Because some words were not included in the segmentation system’s default dictionary, a new user dictionary was built and incorporated into the default dictionary to effectively improve the accuracy of word segmentation. After segmentation, words without actual meaning were added to the default stop word dictionary, and stop words were removed.

3.3.2. Generation of An LDA Topic Model

An LDA topic model is a three-layer Bayesian conceptual model that is used to identify latent topic information in large-scale documents. It is based on the following hypotheses: (1) the number of topics in the corpus collection is K, and all the topics are independent of each other; (2) each corpus consists of a random mixture of K topics, and the topic parameters follow a Dirichlet distribution; (3) each subject has a multinomial distribution on a feature word, and the multinomial distribution’s parameters follow a Dirichlet distribution. Python can automatically generate these topics. After completing the preprocessing of the text, the Python Genism natural language processing software package was used to extract the LDA topics from the text.

3.3.3. Determination of the Optimal Number of Topics

Setting the number of LDA topics is the most pivotal step in the topic analysis process. In this study, the coherence score was used to calculate the optimal number, where a higher score indicated a better number of topics. The coherence score can be expressed by Formula (1) as follows:

$$C\left(z;S^z\right)={{\displaystyle \sum }}_{i=2}^N{{\displaystyle \sum }}_{j=1}^{i-1}\log \frac{P_2\left(v_i^z,v_j^z\right)+1}{P_1\left(v_j^z\right)}$$

(1)

Given a topic thesaurus z, the set of the first N words in z is $S^z=\left\{v_{1,\dots ,}^z{v}_N^z\right\}$; $P_1\left(v_j^z\right)$ is the document frequency of word $v_j^z$; and $P_2\left(v_i^z,v_j^z\right)$ is the co-occurrence document frequency of words $v_i^z$ and $v_j^z$. Combined with the topic coherence score, we generated a series of topic results numbered 1 to 20 to examine the interpretability and representativeness of each group of topics. Based on high-frequency words corresponding to the topic and related documents, two researchers manually interpreted the LDA model [61] and finally selected a group of topics with the highest interpretability and the greatest textual representativeness.

3.3.4. Generation of Word Weights in Each Topic and the Distance of Topics

Based on the assumptions of the LDA model, there were multiple feature words in each topic. After confirming the optimal number of topics and using Python to generate topics, the feature words in each topic and their weights in the topic were first obtained to analyze the internal structure of each topic. The greater the weight of a feature word, the higher the contribution of the word to the topic. In addition, using Python’s visualization toolkit, a topic distance map was drawn to obtain the clusters of each topic. This step was used to validate research hypotheses 1 and 2, which were related to the classification of scientific topics in government communication.

3.3.5. Generation of the Topic Weights of Each Corpus

The manual coding described in the next section was used to obtain the characteristics, such as institutional sources and crisis severity, of each corpus. In order to obtain these topic characteristics, it was necessary to match each topic with each corpus, so as to obtain the weights of the different topics in each corpus. The higher the weight, the greater the relevance of the corpus to a topic. Through the social environment characteristics of the corpus, the social environment characteristics of the topic could be obtained.

3.4. Verifying the Interaction among the Government, Public, and Media

The Baidu search index shows public searches for a keyword obtained by the Baidu company through a big data algorithm (http://index.baidu.com, accessed on 1 May 2022), while the Baidu media index shows the media exposure of a keyword (http://index.baidu.com, accessed on 1 May 2022). These two indices were used to calculate public concern and media concern, respectively, by using seven related keywords, such as “COVID-19”, “COVID-19 outbreak”, “COVID-19 virus”, “Coronavirus”, etc.

Pearson correlation analysis was used to verify H3. In statistics, the Pearson product moment correlation coefficient is used to measure the correlation (linear correlation) between two variables, X and Y; it takes a value between −1 and 1. In the field of natural science, this coefficient is widely used to measure the correlation between two variables. It was invented by Carl Pearson and evolved from a similar but slightly different idea put forward by Francis Galton in the 1880s. This correlation coefficient is also called the “Pearson correlation coefficient R”.

3.5. Verifying the Factors That Modulate Government Communication

The data regarding modulating factors were derived from the manual coding of the corpus concerning science-related content. The institutional sources were measured using three variables according to H4(a), H4(b), and H4(c), i.e., the officials’ level, government agency type, and regions of governments announcing the news briefing, which were all derived from the subject of the corpus. The analysis yielded 1664 corpus subjects sourced from 121 departments or agencies. These 121 departments were categorized to obtain 5 main department categories—health, education, economy, social forces, and the communist party—in addition to other departments. The rank of the corpus subjects was coded as an official rank variable. The geographical region where the news briefing was held was categorized into high-, middle-, and low-income regions according to the World Bank’s classification of income levels and the ranking of China’s GDP per capita.

The crisis severity levels were measured using two variables according to H4(d), i.e., the stage of the epidemic and the number of daily new cases, which reflected the different severity levels of the pandemic; the specific method is shown in

Table 2. Code table.

Factors	Measurable Variables	Coding	Sources
Institutional sources	Types of government agencies	1 = Health 2 = Education 3 = Economy 4 = Social forces 5 = Communist party 6 = Other departments	Artificial recognition
	Officials’ level	1 = No rank (volunteers, doctors, staff, etc.) 2 = Grassroots officers (junior level and below) 3 = Senior officials (deputy director and above)	Artificial recognition
	Region of government holding the news briefing	1 = Low-income regions (ranked 22–31) 2 = Middle-income regions (ranked 12–21) 3 = High-income regions (top 11)	The World Bank’s ranking of income levels China’s GDP per capita ranking
Crisis severity	Periods	1 = Outbreak period (21 January 2020–20 February 2020) 2 = Spread control period (21 February 2020–17 March 2020) 3 = Victory period (18 March 2020–28 April 2020) 4 = Regular epidemic period (29 April 2020–27 August 2021)	A white paper titled China’s actions to fight the COVID-19 pandemic published by Xinhua News in June 2020
	Number of daily new cases	1 = No new cases 2 = The number of new cases ranged from 1 to 10 3 = The number of new cases ranged from 11 to 100 4 = More than 100 new cases	Artificial recognition

.

This paper used analysis of variance (ANOVA) to verify the significant differences in scientific topics between the different political, regional, and pandemic environments. ANOVA was invented by R. A. Fisher and is used to test the significance of the mean difference between two or more samples; it is often used to analyze the relationship between categorical data and quantitative data. Therefore, it was well suited to the analysis of research Hypothesis 4.

4. Results and Discussion

4.1. Scientific Topics in the Chinese Government’s Communications

In order to test the first research hypothesis, the LDA topic model was applied. As can be seen in Figure 4, the number of topics—from 1 to 20—was analyzed to obtain an optimal number of topics at 11, resulting in 11 topics and 10 keywords for each topic, as shown in

Table 3. LDA topic model analysis of the corpus.

Topics	Topic Words and Weights	Models
Topic 1: Chinese Medicine Treatment and Research and Development	0.014 × ”R&D” + 0.011 × ”Research” + 0.010 × ”Pandemic” + 0.010 × ”Clinical” + 0.009 × ”Science and Research” + 0.008 × ”Treatment” + 0.008 × ”Prevention and control” + 0.008 × ”Chinese Medicine” + 0.008 × ”COVID-19” + 0.008 × ”Testing”	Public understanding of science
Topic 2: Knowledge of Epidemiological Investigation	0.033 × ”Epidemiology” + 0.033 × ”Investigation” + 0.012 × ”Testing” + 0.009 × ”Patients” + 0.008 × ”Cases” + 0.007 × ”Examine” + 0.007 × ”Epidemiological investigation” + 0.007 × ”Virus” + 0.007 × ”Close contact” + 0.006 × ”Nucleic acids”	Popularization of scientific knowledge
Topic 3: Vaccination and R&D of Vaccines	0.026 × ”Vaccines” + 0.018 × ”Inoculation” + 0.014 × ”Pandemic” + 0.007 × ”Prevention and control” + 0.006 × ”R&D” + 0.006 × ”Beijing” + 0.005 × ”COVID-19” + 0.005 × ”Community” + 0.005 × ”Publish” + 0.004 × ”Duration”	Public understanding of science
Topic 4: Enterprise Technology Development	0.031 × ”Pandemic” + 0.027 × ”Prevention and control” + 0.020 × ”Enterprise” + 0.008 × ”Resumption of work” + 0.006 × ”Organization” + 0.005 × ”Technology” + 0.004 × ”Resumption of production” + 0.004 × ”Development” + 0.004 × ”Services” + 0.004 × ”Deployment”	Public understanding of science
Topic 5: Public Participation in Prevention and Control	0.037 × ”Publicity” + 0.028 × ”Health” + 0.015 × ”Prevention and control” + 0.014 × ”Civilization” + 0.014 × ”Knowledge” + 0.012 × ”Pandemic” + 0.009 × ”Guide” + 0.009 × ”Citizens” + 0.009 × ”Activities” + 0.008 × ”Media”	Public participation in science
Topic 6: Enterprise Involvement in Prevention and Control	0.017 × ”Pandemic” + 0.016 × ”Prevention and control + 0.010 × ”Cases” + 0.008 × ”Enterprise” + 0.008 × ”Supplies” + 0.007 × ”Emergency response” + 0.007 × ”Health” + 0.007 × ”Implementation” + 0.007 × ”Location” + 0.006 × ”Safeguard”	Public participation in science
Topic 7: Treatment Options	0.011 × ”Cases of disease” + 0.009 × ”Pandemic” + 0.008 × ”Isolation” + 0.007 × ”Treatment” + 0.007 × ”Patients” + 0.006 × ”Diagnosis” + 0.006 × ”Crowd” + 0.006 × ”Asymptomatic infection” + 0.006 × ”Infected persons” + 0.005 × ”Prevention and control”	Public understanding of science
Topic 8: Medical Resources	0.014 × ”Platform” + 0.013 × ”Pandemic” + 0.008 × ”COVID-19” + 0.007 × ”Prevention and control” + 0.007 × ”COVID” + 0.007 × ”Citizens” + 0.007 × ”Beijing” + 0.007 × ”Enquiries” + 0.007 × ”Doctors” + 0.006 × ”Coronavirus”	Public understanding of science
Topic 9: Knowledge of Prevention and Control	0.035 × ”Pandemic” + 0.030 × ”Prevention and control” + 0.013 × ”Measures” + 0.009 × ”Gathering” + 0.009 × ”Risks” + 0.007 × ”Cases of disease“ + 0.007 × ”Health” + 0.007 × ”Implementation” + 0.007 × ”Testing” + 0.006 × ”Protection”	Popularization of scientific knowledge
Topic 10: Community Participation in Prevention and Control	0.035 × ”Prevention and control” + 0.029 × ”Community” + 0.025 × ”Pandemic” + 0.010 × ”Services” + 0.008 × ”The Crowd” + 0.007 × ”Organization” + 0.007 × ”Crowd” + 0.007 × ”Village” + 0.006 × ”Health” + 0.006 × ”Publicity”	Public participation in science
Topic 11: Knowledge of Personal Health	0.021 × “Mask” + 0.018 × “Disinfection” + 0.010 × “Hour” + 0.010 × “Wear”+ 0.009 × ”Health” + 0.009 × “Psychology” + 0.008 × “Use” + 0.007 × “Ventilation” + 0.007 × ”Protection” + 0.006 × ”Testing”	Popularization of scientific knowledge

.

4.1.1. Science Communication Models

In order to further explore the relationship between topics, a distance map was drawn for each theme, as shown in Figure 5. Research Hypothesis 1 was verified by the determination of the topic distance.

Topics 1, 3, 4, 7, and 8 were relatively close in distance. Therefore, the keywords for each topic were analyzed, and we found that the framing of these topics was to have the public understand the scientific research on the COVID-19 pandemic, which relates to the science communication model of the public understanding of science. Topic 1 described the treatment of Chinese medicine and its research and development. Topic 3 was about vaccines, where the government wanted the public to know that a vaccine was being developed and would be publicly available. Topic 4 regarded enterprise technology development. Topic 7 was a description of the treatment options, including how to carry out scientific isolation and how to treat confirmed cases. Topic 8 was about medical resources, explaining how to consult a doctor and how to use the international medical resource platform.

Topics 2, 9, and 11 were relatively close in distance, and by analyzing the keywords of each topic, it was found that the framing of all these topics was to explain scientific knowledge to the public, which relates to the popularization of the scientific knowledge communication model. Topic 2 mainly included knowledge on epidemiological investigations. Topic 9 comprised knowledge regarding pandemic prevention and control. Topic 11 was about personal health knowledge, such as wearing masks, ventilation, disinfection, and attention to mental health.

Topics 5, 6, and 10 were close, and the framing of these topics was to involve the public in the scientific prevention and control of the pandemic, which relates to the science communication model of public participation in science. Topic 5 was an invitation to the public to participate. Topic 6 was about asking companies to participate in the prevention and control of the pandemic by requiring them to prepare adequate emergency supplies and implement prevention and control measures. Topic 10 was the involvement of grassroots social organizations. Communities and villages, as China’s grassroots governance organizations, were also asked by the government to play a role in the prevention and control of the pandemic in the news briefings.

4.1.2. Topics Concerning Sustainability Science

Sustainability science is an integrated science that studies the dynamic relationship between humans and the environment, the core of which is the relationship between the environment, economy, and society [62]. Through the analysis of the results of the LDA topic model, we found that some of the scientific topics framed by the Chinese government related to the sustainability of the public health environment, economy, and society, verifying research Hypothesis 2.

The topics related to the sustainability of public health were knowledge about epidemic prevention and control, treatment options, and medical resources, all of which provide insight on the public’s awareness of the sustainability of the Chinese government’s public health system. This finding is similar to the results reported by La et al. in 2020 [63]. The topic of knowledge on epidemic prevention and control mainly stressed the scientific method and effective coping strategies, while the topic of medical resources emphasized availability and sufficiency, both of which aimed to help the public reduce their panic behaviors and improve their confidence in the Chinese government’s public health system.

The topics related to economic sustainability were enterprise R&D and enterprise participation. During the epidemic, governments stressed the resumption of work and production and engagement in independent R&D to cope with the epidemic by recovering the sustainability of the economy.

The topics related to social sustainability were public participation and community participation in prevention and control. Social sustainability focuses on issues such as poverty reduction and equity. Public and community participation in epidemic-related decision making is an effective way to achieve social equity.

4.2. Social Environmental Characteristics of the Scientific Topics

4.2.1. Interaction with the Public and Media

To test the third research hypothesis, the topic weight of each corpus was obtained by analyzing the LDA topics. Pearson correlation analysis was carried out between the topic weights of each corpus and the public or media concern on the date the corpus was generated. The results are shown in

Table 4. Interaction between the public and the media.

	Public Understanding of Science					Scientific Knowledge Popularization			Public Participation in Science
	1	3	4	7	8	2	9	11	5	6	10
Topic	Chinese Medicine	Vaccines	Enterprise R&D	Treatment Options	Medical Resources	Investigation	Prevention and Control	Personal Health	Public	Enterprise	Community
Public concern	−0.047	−0.020	0.067 **	−0.036	0.080 **	−0.013	−0.107 **	−0.089 **	−0.016	−0.004	0.175 **
Media concern	0.109 **	0.028	0.088 **	0.038	0.000	−0.08 **	−0.066 **	0.030	−0.103 **	−0.025	0.025

** Significant correlation at the level of 0.01 (two-sided); there were no significant correlations at the level of 0.05 (two-sided).

.

The scientific topics communicated by the government that showed a positive correlation with public attention included enterprise R&D (topic 4), medical resources (topic 8), and community participation (topic 10). The topics that had a negative correlation included prevention and control knowledge (topic 9) and personal health knowledge (topic 11). Thus, research Hypothesis 3a was validated. In the topics related to the public understanding of science, traditional Chinese medicine R&D (topic 1), vaccine R&D (topic 3), and treatment options (topic 7)—which the government subjectively and actively wanted to explain to the public—did not change whether the public was concerned with them or not. Enterprise R&D (topic 4) and medical resources (topic 8) were things that the public needed to know. In particular, the topic of medical resources was important to inform the public about how to obtain these resources. Therefore, those topics needed to change according to public attention. In the topics related to scientific knowledge popularization, epidemic investigation knowledge (topic 2) was also something that the government wanted the public to know but was not affected by public concern. Additionally, prevention and control knowledge (topic 9) and personal health knowledge (topic 11) needed more consideration from the government when the public’s attention declined and relaxed their vigilance towards epidemic prevention and control. Thus, a negative correlation was found. Regarding the topic of public participation in science (topic 5), how grassroots organizations, such as communities, could carry out epidemic prevention and control was the topic that most concerned the public; thus, it showed a positive correlation.

The scientific topics communicated by the government that had a positive correlation with media attention included traditional Chinese medicine R&D (topic 1) and enterprise technology development (topic 4). Thus, research Hypothesis 3b was validated. The topics that had a negative correlation included epidemic investigation knowledge (topic 2), prevention and control knowledge (topic 9), and public participation (topic 5). For the topics regarding the public understanding of science, vaccine R&D (topic 3), treatment options (topic 7), and medical resources (topic 8) were not the focus of the media. However, traditional Chinese medicine R&D (topic 1) and enterprise R&D (topic 4) were reported on and propagated by the media; thus, they were positively correlated with media attention. For the topics regarding scientific knowledge popularization, epidemic investigation knowledge (topic 2) and prevention and control knowledge (topic 9) needed to be widely publicized by the media. Therefore, when media attention declined, the government needed to increase this type of corpus. For the topics regarding public participation in science, inviting the public (topic 5) to actively participate in epidemic prevention and control through the media was the main intention of this type of corpus. It was necessary to increase the use of such a type of corpus when media attention declined. Thus, negative correlations were presented.

4.2.2. Institutional Sources

The topic weights of the corpus were analyzed by ANOVA with the officials’ levels, sectors, and regions to obtain the results in

Table 5. Scientific topic framing by different institutional sources.

		Public Understanding of Science					Scientific Knowledge Popularization			Public Participation in Science
		1	3	4	7	8	2	9	11	5	6	10
		Chinese Medicine	Vaccine R&D	Enterprise R&D	Treat-Ment Options	Medical Resources	Investigation	Prevention and Control	Personal Health	Public	Enterprise	Community
Officials’ level	1	0.0791	0.0584	0.0830	0.0773	0.0746	0.0989	0.0803	0.1388	0.1554	0.0500	0.0825
	2	0.0682	0.0809	0.0546	0.0492	0.0345	0.0646	0.1319	0.0700	0.1983	0.0639	0.1631
	3	0.0885	0.0496	0.1147	0.0578	0.0257	0.0692	0.1260	0.0586	0.1866	0.0619	0.1389
	F	1.225	4.344 *	9.455 ***	1.467	12.318 ***	1.891	2.715	14.504 ***	1.266	0.413	5.413 **
Types of agencies	1	0.0706	0.0679	0.0616	0.0689	0.0332	0.0910	0.1645	0.0915	0.1858	0.0414	0.1014
	2	0.0690	0.0219	0.0880	0.0626	0.0464	0.1001	0.0480	0.1621	0.1488	0.0821	0.1526
	3	0.0887	0.0518	0.2683	0.0663	0.0090	0.0182	0.1194	0.0123	0.1617	0.1001	0.0746
	4	0.0745	0.0189	0.1573	0.0530	0.1087	0.0359	0.0380	0.0448	0.1954	0.0776	0.1758
	5	0.0401	0.0321	0.0939	0.0419	0.0353	0.0414	0.0744	0.0308	0.2538	0.0778	0.2560
	6	0.1145	0.0581	0.1294	0.0467	0.0220	0.0583	0.0978	0.0460	0.1735	0.0752	0.1566
	F	4.248 **	2.467 *	13.062 ***	1.362	7.007 ***	4.129 **	9.258 ***	10.498 ***	2.151	3.978 **	10.76 ***
Regions of different income	1	0.0235	0.0386	0.1518	0.0483	0.0328	0.0987	0.1403	0.0723	0.1616	0.0685	0.1399
	2	0.0820	0.0167	0.1185	0.0553	0.0463	0.0581	0.0943	0.0338	0.2398	0.0554	0.1718
	3	0.0904	0.0565	0.0846	0.0561	0.0334	0.0789	0.1032	0.0626	0.2155	0.0547	0.1425
	F	5.013 **	11.041 ***	4.697 **	0.8	2.301	3.743 *	11.054 ***	9.755 ***	24.111 ***	1.95	3.808 **

* p < 0.05; ** p < 0.01; *** p < 0.000. (Officials’ level: 1—no rank; 2—grassroots officers; 3—senior officials. Types of agencies: 1—health; 2—education; 3—economy; 4—social forces; 5—communist party; 6—other departments. Regions of different income: 1—low-income regions; 2—middle-income regions; 3—high-income regions.).

, which show the institutional sources of the scientific topics.

The scientific topics of medical resources (F = 12.318 ***), personal health (F = 14.504 ***), vaccine R&D (F = 4.344 *), community participation (F = 5.413 **), and enterprise R&D (F = 9.455 ***) were all framed differently according to the different levels of officials in the Chinese government’s communications about COVID-19. Through this process, Hypothesis 4a is verified. The topics of medical resources and personal health were mostly framed by a non-ranking official, which is a non-governmental official who is invited to make a speech at the press release conference; for example, doctors prefer to communicate personal health knowledge to the public and help them understand medical resources. The frames of the vaccine R&D topic and the community participation topic were mostly used by grassroots officers, which refer to officers at the junior level and below, such as the subdistrict office director, who are invited to make a speech; their duties are to manage and control vaccination and the community epidemic. Therefore, they are more eager to let the public understand vaccine R&D and promote their local communities to become involved in science. The frame of the enterprise R&D topic was used more by senior officials, whose concerns for epidemic control and economic development in an epidemic were from a macroscopic-level perspective; they preferred the public to understand more about enterprise R&D so as to improve their confidence and promote the resumption of work and production and thus economic recovery. However, there was no significant difference in the other scientific topics framed by governmental officials at the three different levels.

The scientific topics of Chinese medicine (F = 4.248 **), vaccine R&D (F = 2.467 *), enterprise R&D (F = 13.062 ***), medical resources (F = 7.007 ***), epidemiological investigation (F = 4.129 **), prevention and control (F = 9.258 ***), personal health (F = 10.498 ***), enterprise participation (F = 3.978 **), and community participation (F = 10.76 ***) were framed differently by different types of agencies in the Chinese government’s communications about COVID-19. Through this process, Hypothesis 4b was verified. It was also shown that every department tended to use a frame that was identical to their own departmental functions; for example, the topics of vaccine R&D and prevention and control were framed most often by the health department. The topics of enterprise R&D and enterprise participation were framed mostly by the economy department. Additionally, governmental departments also tended to select the scientific communication model identical to their own functions; for example, the topics of epidemiological investigation and personal health related to the scientific knowledge popularization model were mostly framed by education departments, which related closely to their functions. Community participation was mostly framed by party organizations, as party building leads community governance in China.

The scientific topics of Chinese medicine (F = 5.013 **), vaccine R&D (F = 11.041 ***), enterprise R&D (F = 4.697 **), epidemiological investigation (F = 3.743*), prevention and control (F = 11.054 ***), personal health (F = 9.755 ***), public participation (F = 24.111 ***), and community participation (F = 3.808 **) were framed differently in regions with different incomes. It was easy to find a significant difference in the scientific communication models used by governments in different regions with different revenues. Thus, Hypothesis 4c was also verified. Topics related to the scientific communication model of public participation in science were mostly framed by governments in the middle-income regions (public participation = 0.2398, community participation = 0.1718); however, topics in the scientific knowledge popularization model were framed most often by governments in low-income regions (epidemic prevention and control = 0.0987, prevention and control = 0.1403, personal health = 0.0723), while the topics of traditional Chinese medicine (0.00904) and vaccine R&D (0.05656) related to the public understanding of science model were mostly framed by governments in high-income regions.

4.2.3. Crisis Severity

The ANOVA analysis of the topic weights of the corpus using the stage of the epidemic and the number of daily new cases on the day the corpus was generated yielded the values listed in

Table 6. Scientific topic framing in different crisis severity levels.

		Public Understanding of Science					Scientific Knowledge Popularization			Public Participation in Science
Topic		1	3	4	7	8	2	9	11	5	6	7
		Chinese Medicine	Vaccine R&D	Enterprise R&D	Treatment	Medical Resources	Investigation	Prevention and Control	Personal Health	Public	Enterprise	Community
Period	1	0.0593	0.0600	0.1312	0.0578	0.0419	0.0624	0.1012	0.0778	0.1347	0.0586	0.1927
	2	0.1047	0.0484	0.1194	0.0620	0.0260	0.0475	0.1401	0.0506	0.1684	0.0716	0.1395
	3	0.0971	0.0505	0.0710	0.0423	0.0283	0.0493	0.1451	0.0784	0.2551	0.0580	0.1020
	4	0.0865	0.0675	0.0596	0.0731	0.0283	0.1343	0.1145	0.0619	0.2054	0.0588	0.0891
	F	4.104 **	0.949	10.08 ***	2.026	1.764	15.523 **	3.086 *	2.096	15.073 ***	0.517	15.260 ***
Cases	1	0.0966	0.0442	0.0868	0.0494	0.0324	0.0698	0.1316	0.0723	0.2240	0.0532	0.1181
	2	0.0632	0.0789	0.0957	0.0805	0.0314	0.0702	0.1021	0.0757	0.1671	0.0732	0.1397
	3	0.0483	0.0626	0.1323	0.0379	0.0383	0.0870	0.1421	0.0585	0.1251	0.0472	0.1975
	4	0.2457	0.0172	0.1461	0.0835	0.0162	0.0250	0.0746	0.0176	0.0909	0.1387	0.1235
	F	15.969 ***	5.287 **	3.489 *	5.093 **	0.554	1.578	2.653 *	2.070	11.739 ***	5.836 **	6.467 ***

* p < 0.05; ** p < 0.01; *** p < 0.000. (Period: 1—outbreak period; 2—spread control period; 3—victory period; 4—regular epidemic period. Cases: 1—no new cases; 2—number of new cases ranging from 1 to 10; 3—number of new cases ranging from 11 to 100; 4—more than 100 new cases).

, which show the crisis severity levels of the day those scientific topics were generated. This helps to document how research Hypothesis 4d was validated.

The frames of Chinese medicine (F = 4.104 **), enterprise R&D (F = 10.08 ***), epidemiological investigation knowledge (F = 15.523 **), knowledge about epidemic prevention and control (F = 3.086 *), public participation (F = 15.073 ***), and community participation (F = 15.260 ***) varied according to the different stages in the development of the epidemic. Enterprise R&D and community participation were mostly framed in the outbreak period of the epidemic, signifying that the two topics played an important role in coping with the epidemic in its preliminary stages. Traditional Chinese medicine R&D was mostly framed in the control period of disease spread; as it is a therapeutic regimen with Chinese characteristics, its greater spread at the stage of epidemic control may have improved the public’s self-confidence. Epidemic prevention and control and public participation were mostly framed at the stage of victory, signifying that the two topics played an important role in the control of the epidemic by the Chinese government. Epidemiological investigation knowledge was mostly framed at the regular epidemic stage, signifying that the government was eager to let the public know when the epidemic was essentially brought under control.

The governmental frames of Chinese medicine R&D (F = 15.969 ***), vaccine R&D (F = 5.287 **), enterprise R&D (F = 3.489 *), treatment options (F = 5.093 **), prevention and control knowledge (F = 2.653 *), and all three topics in the science communication model of public participation in science (public participation = 11.739 ***, enterprise participation = 5.836 **, and community participation = 6.467 ***) varied between different new case counts. Thus, it was easily found that most of the frames were used by the government when there were too many cases (11~100 and 100 or above). However, only two topics were used more by the government when the patient cases were not severe. One was the topic of public participation, which was mostly framed when there was no patient caseload since the public was required to extend their cooperation to epidemic control and prevention when the epidemic was not severe. The other was the topic of vaccine R&D, which was mostly framed when the new patient cases were between 1 and 10 since the government needed the public to know about vaccine R&D to improve vaccination rates when there were few cases.

5. Conclusions

5.1. Summary of Major Findings and Contributions

This research showed that the lack of information disclosure during the COVID-19 pandemic was the main reason for ineffective government communication [64]; however, information overload can also have adverse consequences for the public [65]. Therefore, government communication should reasonably consider the needs of the public. Sustainable communication approaches should be established by the government [39]. Our suggestions for sustainable communication approaches can be found in the major findings of this paper.

5.1.1. Identification of the Patterns of Science Communication Models in the Chinese Government’s Communication about COVID-19

The present research verified that the popularization of scientific knowledge [31], public understanding of science [34,35], and public participation in science [37] are three development stages and communication models in science communication. This study viewed the Chinese government’s information sharing policies regarding COVID-19 to identify patterns of science communication models and found that the Chinese government’s frames for its official communications of scientific topics surrounding COVID-19 fit the three types of science communication models and their functions, which was consistent with the findings of existing studies.

There are few studies that have explored how the three modes of science communication can be achieved through government framing. Our findings further explored how the Chinese government framed scientific topics about COVID-19 according to different communication models. The Chinese government framed the scientific topics of the knowledge of prevention and control, epidemiological investigation, and personal health according to the popularization of scientific knowledge models; Chinese medicine R&D, enterprise R&D, vaccine R&D, and medical resources according to the public understanding of science model; and scientific topics of citizen participation, community participation, and enterprise participation according to the public participation in science model. These findings not only summarize the strategies of the Chinese government to communicate scientific topics, but also are informative about the spread of scientific topics in other countries in the world; additionally, they provide sustainable communication examples from China for the practical application of science communication theories, such as science popularization, public participation in science, and public understanding of science.

5.1.2. Verification of the Interaction among Government Frames, Public Concern, and Media Concern about COVID-19

According to the agenda-setting theory and existing studies, effective interactions can be achieved between the governmental agenda, public agenda, and media agenda [13]. In this paper, the theory was again verified by the Chinese government’s communications on scientific topics about COVID-19. On its basis, a further finding from this paper was that the frames of the Chinese government were mostly correlated with public concern and media concern and the positivity or negativity of such correlations. The frames of enterprise R&D, medical resources, and community participation were positively correlated with public concern, while knowledge about epidemic prevention and control and public health were negatively correlated with public concern. Traditional Chinese medicine R&D and enterprise R&D were positively correlated with media concern; however, epidemiological investigation knowledge, epidemic prevention and control knowledge, and citizen participation were negatively correlated with media concern. All of these findings will be helpful for the Chinese government to address the concerns of the public and media. One approach to build sustainable communication is as follows: if aspiring to arouse the concern of the public and media, more communication with the frames that have a positive correlation might be preferred; if aspiring to reduce concern, more communication with the frames that have a negative correlation might be preferred.

5.1.3. Identification of Factors That Modulate Government Framing of Scientific Topics about COVID-19

According to existing studies, the source of institutions and severity levels of the epidemic might be the main factors affecting a government’s framing of scientific topics about COVID-19. Current studies focus mostly on the difference in government level [29], income [51], and epidemic severity [28] in framing, but less on the framing difference among the different types of governmental agencies. In this paper, the latter was explored. In addition, we found the influence of various modulating factors on governmental framing. The main findings from our study include: (1) high-level officials framed the topics of enterprise R&D more often, while low-level officials framed the topics of vaccination and epidemic prevention and control more, and non-ranking officials framed medical resources and personal health knowledge more. (2) Every department was more likely to use the frame identical to their own departmental function and also tended to select the science communication model identical to their own function. (3) The topics of traditional Chinese medicine and vaccine R&D were framed more in the high-income region, while the topics related to the public participation in science model were framed more in the middle-income region, and the topics related to the scientific popularization model were framed more in the low-income region. (4) During the time when the epidemic became severe, the Chinese government framed more topics of community participation and topics related to the public understanding model; however, when the epidemic was not severe, governments framed more topics of citizen participation and topics related to the scientific popularization model. From the perspective of the government’s sustainable communication, all findings herein are helpful guides for governmental agencies and officials at different levels, as well as regions with different economic development levels, in using effective frames to achieve effective communication when the epidemic situation changes.

5.1.4. The Topics Concerning Sustainability Science were More Correlated with Public Concern and Media Concern

An interesting finding of this study is that the scientific topics related to sustainability science were more correlated with public concern and media concern. The first topic that correlated mutually with public concern and media concern was enterprise R&D, which is related to economic sustainability. The second topic that was mutually correlated with public concern and media concern was knowledge about epidemic prevention and control, which is related to the sustainability of the public health environment. This finding can be applied to the government’s communication of sustainability issues during the COVID-19 pandemic. When other governments want to spread sustainability issues with more public concern during COVID-19, they can also frame similar issues.

5.2. Limitations and Suggestions for Future Research

The main limitation of this study is the native limitation of the LDA topic model. The clustered words formed by the LDA topic model must be chosen by subjective judgment and classification by humans; thus, there may be deviations. In order to solve this problem, in this study, two researchers manually interpreted the LDA model based on high-frequency words corresponding to the topic and related documents and finally selected a group of topics with the highest interpretability and the most textual representativeness. Further studies in the future may explore the use of other big data analysis methods so as to achieve more impersonal subject classifications.