2.1 Research methods

Based on the attention networks in 31 provincial-level administrative regions (excluding Hong Kong, Macao and Taiwan) in China, an attention network for the public in China was developed using the keyword ’novel coronavirus’ by incorporating information from the Baidu index website from January 9, 2020, to April 15, 2020. Some general domestic research methods, such as the annual variation index, seasonal variation index, and weekly distribution skewness index, are limited by the time period of this study. The coefficient of variation (CV), geographic concentration index (Ge) and Gini coefficient (Gi) were used to analyse the temporal and spatial differences in the novel coronavirus attention network.

1. The geographical concentration index is used mainly to analyse the spatial distribution characteristics of a studied variable, and its formula is: (1)
   In Eq (1), x_i is the search index for the novel coronavirus in the ith province, and T is the sum of the National Novel Coronavirus Baidu Search Index, where n = 31. The larger the value of Ge is, the more concentrated the novel coronavirus attention network, and the smaller the value of Ge is, the more dispersed the novel coronavirus attention network.
2. The Gini coefficient is used in quantitative geography to analyse the spatial distribution differences among variables within a discrete region and is represented by Gi. This index is expressed as follows: (2)
   In formula (2), x_i refers to the proportion of the attention network corresponding to the ith province relative to the overall attention network, N refers to the 31 provinces studied in this paper, and the calculated Gi value range is 0–1. If the Gi value is large, the spatial distribution of the novel coronavirus attention network is unbalanced, and the higher the concentration degree is, the smaller the Gi value and the more dispersed the distribution [37].
3. The coefficient of variation CV is a relative statistic used to measure the discreteness of different samples. This variable is the ratio of the standard deviation of a set of data to the corresponding average value. In this paper, the coefficient of variation is used to measure the degree of discreteness of temporal variations in the attention degree of the novel coronavirus network. The corresponding formula is as follows. (3)
   In formula (3), CV is the coefficient of variation, y_i is the attention in time period i, and is the average value of y_i. The greater the CV value is, the greater the time differences in the novel coronavirus attention network.
4. Moran’s I index, which reflects the similarity of attribute values in regional units that are adjacent in space, is commonly used to measure global and local spatial autocorrelations. The former reflects the overall agglomeration degree of observation indexes in a certain space, and the latter reflects the agglomeration degree of observation indexes between two units within a given space. (4)
   In the formula, I represents the global spatial autocorrelation index and I_i represents the local spatial autocorrelation Moran’s I index. i and j are provinces (plus districts and cities). This paper uses GeoDa software to calculate Moran’s I index.
5. Multivariate stepwise linear regression analysis is based on the principle of introducing all independent variables into the equation. According to the degree of influence of independent variables on dependent variables, the correlation test is used to eliminate the independent variables that are not significant, and the final screened independent variables and dependent variables are established to produce a regression model with correlation [38]. The corresponding formula is as follows: (5)
   Among them, Y is the dependent variable, which represents the attention in the new coronavirus network, X₁ represents the number of Internet users in the field of information, X₂ represents the population at the end of the year, X₃ represents the proportion of the population aged 15–64, X₄ represents the proportion of the population aged over 65, X₅ represents the proportion of the population with a high school education, X₆ represents the proportion of the population with a college education and above, β_i is the regression coefficient, and ε is the residual and obeys the normal distribution.

2.2 Data source

Baidu is the world’s largest Chinese search engine and is committed to providing Internet users with more convenient access to information that they want. Because it is a data sharing platform, we can clearly determine the number of Internet users associated with certain keywords and the development trends of some popular topics over a given period of time. The Baidu index has been applied by many scholars because of its accuracy and convenience of data acquisition [39–41].

Based on the keyword ’novel coronavirus’, the daily average search volume was obtained from January 9, 2020, to April 15, 2020, with the Baidu index platform as the basic dataset. The data collection area included 31 provincial-level administrative regions (excluding Hong Kong, Macao and Taiwan).

The start time of the study was January 9, 2020, because of the outbreak of unexplained pneumonia in Wuhan. The National Health Commission organized multiple units to detect case samples for the first time. On January 8, 2020, the new coronavirus was preliminarily confirmed as the pathogen of the epidemic, and the epidemic dynamics were updated daily. The reasoning behind the deadline selection of April 14, 2020, was that the ’Huoshenshan’ and ’Leishenshan’ hospitals, which played important roles during the development of the epidemic, were officially closed on April 15, 2020, indicating that the epidemic in Wuhan had achieved a phased victory.

3.1 Time differences in the novel coronavirus attention network

A statistical analysis of the novel coronavirus attention network from January 9, 2020-April 15, 2020, showed that the novel coronavirus attention network index exhibited obvious fluctuations during the study period, as shown in Fig 1. From January 9, 2020, to April 15, 2020, the novel coronavirus attention network displayed a sharp growth trend, and the number of searches rapidly increased from 3674 to 1838957. The first case of unexplained pneumonia occurred in Wuhan on December 8th, and the pathogen was identified as a novel coronavirus on January 9, 2020, by an expert group.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Trend of new coronavirus network attention in China.

https://doi.org/10.1371/journal.pone.0257291.g001

On January 20, Dr. Zhong Nanshan confirmed that the new coronavirus epidemic had the phenomenon of ’human transmission’ and that there had been a medical staff member infection. After the closure of Wuhan on January 23, the first-level response to major public health emergencies in Hubei Province was launched on the eve of January 24. Afterward, the ’Huoshenshan’ and ’Leishenshan’ hospitals opened, and medical and nursing workers were sent from across the country to assist in Wuhan. Within a week, with the introduction of various national anti-epidemic measures and publicity from media such as news media, the public’s attention in the new coronavirus network increased explosively. On January 29, the public’s attention in the new coronavirus network reached its peak. The reasons for the peak include that January 29 was the last day of the Spring Festival holiday, and people paid more attention to the characteristics and dissemination of the epidemic after the holiday. Additionally, because of the series of measures enacted previously, people had a basic understanding of the situation of the epidemic.

One week after the peak on January 29, the attention to the new coronavirus continued to decline. The decline in this stage was because there was little change in people’s attention in the new coronavirus network after the peak. By February 5, it dropped to the attention level before the closure of Wuhan, and the daily search volume remained at approximately 1.2 million times. This attention level continued until February 15. The above change trend was because with the development of the epidemic, the response measures were steadily advancing, and there were no incidents that led to a sharp increase in the public’s attention to the epidemic. However, because the epidemic was far from over, the public’s attention to the epidemic had been relatively stable, indicating that the development of the epidemic was relatively stable in this period of time and did not cause great panic among the people. At the same time, the public did not slacken the prevention and control of the epidemic.

With the help of the ’Huoshenshan’ and ’Leishenshan’ hospitals, Wuhan International Convention and Exhibition Center was transformed into a ’field hospital’, and China achieved remarkable anti-epidemic results. Then, on March 11th, President Xi Jinping visited Hubei Province to investigate the prevention and control of the new coronavirus pneumonia epidemic. On March 25, the control of the roadways from Hubei was gradually relaxed, except for roads to and from Wuhan. Due to the timely adoption of appropriate epidemic prevention measures in China, the epidemic prevention and control effects were positive. After January 30, the novel coronavirus attention network displayed a downward trend. By March 25, the novel coronavirus attention network returned to the form observed before the Wuhan lockdown. From March 25 to April 15, the public’s attention to the novel coronavirus began to subside, suggesting that through the efforts of Chinese citizens and the united response to the epidemic, the epidemic situation in China was continually improving. Notably, orderly work and life resumed, and the attention given to the epidemic remained at a stable level (see Fig 1).

The standard deviation and CV of the novel coronavirus attention network from January 9, 2020, to April 15, 2020, are shown in Table 1. During the 14-week period, the difference in network attention was significant. The CV in the second week (January 16–January 22) was the largest, followed by that in the first week. Although the CV fluctuated each week after the third week, it remained stable, and there was no significant change. This result suggests that the trends during the previous two weeks were due to changes in public awareness and anti-epidemic situations or the effects of news, which resulted in novel coronavirus attention network differences. After the third week, on January 23, the city of Wuhan was closed, the epidemic situation intensified, anti-epidemic policy was introduced, and the epidemic became universal. The public began to pay attention to the epidemic progress each day, and the time differences in the attention network gradually narrowed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. 2020 novel coronavirus network variations from January 9 to April 15.

https://doi.org/10.1371/journal.pone.0257291.t001

3.2 Spatial differences in the novel coronavirus attention network

3.2.1 Regional differences in the novel coronavirus attention network.

From January 9, 2020, to April 15, 2020, there were significant differences in the novel coronavirus attention network in the 31 provinces. The attention levels in the main regions peaked in the third week of the epidemic (January 23-January 29) and displayed a downward trend afterward (shown in Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Change trend chart of network attention in 31 provinces.

https://doi.org/10.1371/journal.pone.0257291.g002

According to the interpretation of the National Development and Reform Commission, China’s eastern, central and western division is a policy division in which Inner Mongolia and Guangxi enjoy preferential policies for western development formulated in 2000; thus, Inner Mongolia and Guangxi belong to the western region. From the perspective of the novel coronavirus attention networks in the three regions, the network attention in the eastern and central regions decreased over time, and the network users in the eastern region, especially in Guangdong, Shandong, Jiangsu and other provinces, exhibited the highest attention level toward the novel coronavirus. The attention levels in Tibet, Xinjiang, Qinghai and Ningxia in the western region were among the lowest in China.

The novel coronavirus attention networks from January 9, 2020, to April 5, 2020, in 31 provinces were sorted, and the attention levels were ranked from high to low (see Table 2) based on three classes: 1–10 (high), 11–20 (moderate), and 21–31 (low). Correspondingly, the changes among provinces were divided into stable, fluctuating rise, and fluctuating decline classes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Correlation analysis of the factors that influence COVID-19 network attention.

https://doi.org/10.1371/journal.pone.0257291.t002

According to Fig 3, most of the eastern and western regions belong to the stable type, and only four provinces in the central region (Jilin, Jiangxi, Henan, Heilongjiang) belong to the stable type. There are three provinces in the central region (Shanxi, Anhui and Hunan) and only two provinces in the eastern and western regions, namely, Liaoning, Hebei, Yunnan and Guizhou. Beijing and Shanghai in the eastern region, Hubei in the central region and Guangxi in the western region belonged to the type of fluctuation decline. According to the above analysis, we can see that most of the provinces were the stable type in the study period, and the rank changes in several provinces with rising and falling fluctuations are not particularly large, indicating that people’s attention in the new coronavirus network was at a normal level and there were no large fluctuations. This also indirectly shows that China’s various measures in the process of dealing with the epidemic were carried out in an orderly manner and that there was no great panic among the people.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Sequence changes in the network attention level in 31 provinces.

a: Sequence change in network attention in the eastern region. b: Sequence change in network attention in central region. c: Sequence change in network attention in the western region.

https://doi.org/10.1371/journal.pone.0257291.g003

3.2.2 Interprovincial differences in the novel coronavirus attention network.

To further analyse the interprovincial differences in the novel coronavirus attention network, the geographic concentration index (Ge) and Gini coefficient (Gi) were calculated (see Fig 4). The geographic concentration index encompasses attention fluctuations from January 9, 2020, to April 15, 2020. The changes in this index were small, and it remained basically stable, with values ranging from 21–23, which are far less than 100. In general, it can be concluded that the spatial distribution of the novel coronavirus attention network was not concentrated and was generally decentralized. However, the Gini coefficient did not notably vary but was greater than 0.92, which reflects the local spatial distribution of the novel coronavirus attention network. Based on the above analysis, it can be concluded that the novel coronavirus attention network exhibited ’overall dispersion and local agglomeration’ characteristics. ’Overall dispersion’ was analysed from the national perspective, and the novel coronavirus attention network was divided into 31 provinces. ’Local agglomeration’ included high agglomeration in Shandong, Anhui, and Jiangsu and low agglomeration in Xinjiang, Gansu, and Qinghai.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Novel coronavirus network geographic concentration index and Gini coefficient.

https://doi.org/10.1371/journal.pone.0257291.g004

ArcGIS10.2 software was used to draw the provincial spatial distribution map of the network attention of the new coronavirus (Fig 5). Six different time periods were selected to analyse the differences in network attention. These six time periods were the first week, third week, seventh week, ninth week, eleventh week and fourteenth week. The network attention is divided into 1–5 grades according to the depth of colour from low to high. The deeper the colour of the province is, the higher the network attention. The reason for selecting these six periods was that the first week was the initial stage of the epidemic, and the third week to the seventh week were the most serious period during the development of the epidemic. After the ninth week, the epidemic situation was basically stable, and people’s attention to the epidemic was at a relatively stable state afterward.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Spatial distribution characteristics of network attention.

Source: The primitive base map from the website of the Ministry of Natural Resources of China (http://bzdt.ch.mnr.gov.cn/) was plotted by ArcGIS 10.2 software (ESRI®), and its Drawing Review Number is GS (2016) 2893.

https://doi.org/10.1371/journal.pone.0257291.g005

The following conclusions can be drawn from the figure. First, Hubei and Guangdong were the provinces with the deepest colour in the first week, and the provinces with the deepest colour became more numerous in the third and seventh weeks. In the ninth week, only Guangdong belonged to the type with the deepest colour. This change shows that since the epidemic first occurred in Wuhan City of Hubei Province, Hubei Province was at the fifth level of network attention in the first week. Guangdong Province was at the fifth level of network attention because of its large population, developed economy and rapid information dissemination. With the aggravation of the epidemic situation, in the third week and the seventh week, the number of provinces at the fifth level increased significantly. However, after the ninth week, with China’s positive and effective response measures, the epidemic situation was basically stable, and Guangdong was the only province with the fifth level of network attention.

Second, it can be seen from the graph that most of the eastern provinces and central provinces were at the fourth and third levels of network attention, and this level included only Sichuan Province among the western provinces. The reason for this phenomenon is that these provinces are relatively developed, and the speed of information dissemination is fast. According to the analysis of Fig 5, the level of new coronavirus network attention was highly consistent with the Hu Huanyong line, which is China’s population boundary. The area east of Hu Huanyong line is densely populated, and the network attention was high, mostly in the third to fifth grades. The provinces west of the Hu Huanyong line had low levels of network attention, mostly in the first and second grades.

There was also a phenomenon in Hubei Province. This province experienced the first outbreak of the new coronavirus, and its epidemic situation was the most serious. In the first week, attention in Hubei Province was at the fifth level, which is the highest degree of network attention. Beginning in the third week, in Hubei Province, the level of network attention began to decline compared with the first week. This phenomenon did not indicate that people’s attention in the new coronavirus network decreased in Hubei Province, instead indicating that with the aggravation of the epidemic situation, nationally, people’s attention to the epidemic became generally strengthened, which shows that in the process of epidemic prevention and control, with the participation of the whole population, the collective consciousness is enhanced.

3.3 Spatial correlations and clustering in the novel coronavirus attention network

Moran’s I index was used to test the spatial correlations and agglomeration in the novel coronavirus attention network. The research period was half of a month after the lockdown of Wuhan on January 23, and the novel coronavirus attention network of 31 provinces in China was considered. This period was selected because of the explosive growth of the epidemic in the half-month from January 23 to February, and the public’s attention to the epidemic also peaked during this period; therefore, the network attention level in this period was representative. First, the global Moran’s I index was used to test the spatial correlations in the novel coronavirus attention network with GeoDa software, and the global Moran’s I value was 0.146, with a P value of 0.067. These results indicate that the novel coronavirus attention network displayed few positive spatial correlations, mainly because the spatial differences in network attention among different provinces were large, and the agglomeration degree was not high.

Through LISA agglomeration analysis, the spatial agglomeration characteristics of the attention level of the novel coronavirus network were obtained (see Fig 6). It is obvious that these characteristics were not significant, although the agglomeration characteristics of some provinces in the eastern region are obvious. High-high concentration areas (H-H) of the novel coronavirus network were mainly distributed in Shandong, Jiangsu, Anhui and Jiangxi Provinces. The regions with low-low network attention (L-L) were mainly distributed in Xinjiang and Gansu. Sichuan Province was located in the ’V’ type cluster, and it displayed a high network attention level, although the levels of the surrounding provinces were low. The LH-type agglomeration area, namely, the ’A’ type agglomeration area, was mainly distributed in Hainan, Fujian and Shanghai. In contrast to the trend in Sichuan Province, these areas gave limited attention to the novel coronavirus, while the surrounding provinces had high network attention levels. Combined with the analysis results of Moran’s I index, it can be seen that the attention in China’s new coronavirus network presented ’overall dispersion and local agglomeration’ in terms of spatial correlations. Local agglomeration is mainly reflected in the high-high agglomeration areas in the figure, including Shandong, Jiangsu, Anhui and Jiangxi. The regions with low-low agglomeration in the figure are Gansu and Xinjiang.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. LISA concentration of novel coronavirus network concern.

Source: The primitive base map from the website of the Ministry of Natural Resources of China (http://bzdt.ch.mnr.gov.cn/) was plotted by ArcGIS 10.2 software (ESRI®), and its Drawing Review Number is GS (2016) 2893.

https://doi.org/10.1371/journal.pone.0257291.g006

(1) Economic development level

Table 2 shows that although there was a positive correlation between the per capita GDP and network attention, the relation did not pass the significance test. Therefore, economic factors had no significant impact on the attention network in different regions. From the outbreak of the epidemic to the Wuhan lockdown on January 23, various news items were released about the novel coronavirus. In addition, each region started to respond to major public health emergencies, making people realize the urgency and danger of the epidemic. Consequently, the novel coronavirus became a national focus within a very short period of time. Economic factors had a limited influence on the outbreak of new coronavirus pneumonia because the particularity of this outbreak is that it is related to people’s health and safety, and everyone has the responsibility and obligation to understand the development of the epidemic.

(2) The degree of network development

The degree of network development is one of the indispensable factors related to network attention. This paper studies the degree of network development from two aspects: the proportion of Internet users in the field of information and the Internet penetration rate. From Table 2, it is concluded that the positive distribution of the Internet penetration rate in the novel coronavirus attention network was not obvious, and the proportion of Internet users in the field of information was significantly positively correlated with network attention (r = 0.941, P < 0.05).

These results show that the popularity rate of the Internet in a region could not significantly indicate the level of network attention, and the network behaviour of Internet users in the information field could better reflect people’s search needs. Provinces with a more accurate population and a larger proportion of Internet users in the information field exhibited higher network attention to the new coronavirus. During the epidemic period, the state and the government updated the relevant information and provided the latest progress of the epidemic in a timely manner. The release of information was more accurate and transparent, which was helpful in allowing Internet users to obtain the latest epidemic information.

(3) Regional population statistics

Since the epidemic of neo-coronal pneumonia is a national public health event and the infectiousness of neo-coronal pneumonia is very strong, the larger the population size of a region is, the greater the risk of epidemic transmission and the more attention people generally pay to events such as epidemics. Table 2 shows that the novel coronavirus attention network was significantly positively correlated with the population at the end of the year (r = 0.918, P < 0.05). The larger the population at the end of the year in a province was, the higher the network attention level. The research data also suggest that Guangdong and Shandong, which were the two provinces with the highest populations at the end of the year, exhibited the highest network attention levels. The more densely populated areas are, the higher the risk of cross-infection is. Therefore, for areas with large populations and high population densities, the risks and challenges in epidemic prevention are also great.

(4) Social demographic characteristics

This paper analyses the impact of sociodemographic characteristics on network attention from two perspectives: education level and age structure. This paper divides the educational level of the population into four grades: primary school and below; junior high school; high school; and college and above. Table 2 shows that the four levels of education were significantly correlated with network attention, and the higher the education level was, the more significant the correlation was. The significance of junior college and above education (r = 0.939, P < 0.05) was the highest, exhibiting a strong correlation and indicating that the level of education was one of the important factors that influenced the novel coronavirus attention network. The higher the education level, the higher the utilization rate of the Internet, and the higher the dependence on the network. To assess the potential impacts of age, this paper divides the age structure of the population into three levels: 0–14 years old, 15–64 years, and over 65 years old.

From the analysis results, it was concluded that the population distribution in the three age groups was significantly correlated with network attention. The correlation coefficient (r = 0.932, P < 0.05) between the 15-64-year-old age group and network attention was the largest, and the correlation was strong. The reason for this phenomenon is that the population base of this age group is relatively large, thus leading to a high network attention level. The population of this age group is the main group of Internet users. Additionally, the age group over 65 years old (r = 0.879, P < 0.05) was highly threatened by the novel coronavirus. The novel coronavirus is a new virus that was discovered in 2019 and has never previously been found in the human body. There were no antibodies in the population, so all humans were susceptible. The most susceptible population is the elderly population, especially individuals with low immunity and chronic basic diseases. Thus, many people over 65 years old have paid special attention to novel coronavirus news.