In April 2022, Shenzhen was the first city in China to set up regular nucleic acid testing sites and implement regular nucleic acid testing [25]. By July 2022, most regions across the country had conducted regular nucleic acid testing [26]. Therefore, we used the keyword “nucleic acid testing” to retrieve original Weibo data posted between April 1, 2022, and July 31, 2022. We used Octopus software (Shenzhen Skieer Information and Technology Co Ltd) to obtain data such as blog posts, posting times, and publisher nicknames. After removing duplicate posts, commercial advertisements, and blank data, a total of 463,566 Weibo posts were obtained. Afterward, data cleaning was performed, including removing numbers, punctuation, and special symbols (including URLs, @usernames, spaces, etc), removing non-Chinese content, and converting traditional Chinese characters to simplified Chinese characters. Tokenization was performed by using the Python Jieba package. Because many words on Weibo were internet neologisms and trending terms, and many were related to the COVID-19 pandemic, such as “health code,” “green code,” and even the main focus of this study, “nucleic acid testing,” might not have been correctly segmented. Therefore, we created a custom lexicon to assist with segmentation [27]. The custom lexicon was divided into 2 parts. One part was the cell dictionary of Sogou’s network hot words [28], and the other part consisted of annual hot words and popular phrases collected and deduplicated from various platforms in the first half of 2022, which were selected manually. The stop word list used included Chinese stop word lists, the Harbin Institute of Technology Stop Word List, Baidu stop word lists, and the stop word library from the Machine Intelligence Laboratory of Sichuan University [27]. More details are in Figure 1. At the same time, we obtained the daily number of newly confirmed cases from the official website of the Chinese Center for Disease Control and Prevention [29].

This study used the Latent Dirichlet Allocation (LDA) model for topic modeling. LDA is the most widely used topic model at the time and could be used to identify the most common topics across social media platforms [30]. LDA is an unsupervised learning algorithm that does not require prelabeled datasets, making it suitable for handling large and diverse text data such as Weibo posts [31]. It is a Bayesian probabilistic model with 3 layers: “document-topic-word,” which is highly effective in identifying latent semantic topics in unstructured textual data, such as Weibo texts [32]. Furthermore, the LDA model does not have strict restrictions on the length of text; it is highly efficient in identifying topics in short texts [32]. Before constructing the LDA model, it is necessary to determine the number of topics [33]. Perplexity is a metric for determining whether the number of topics is reasonable, as the appropriate number of topics directly affects the accuracy of topic mining [33]. As the number of topic words increases, the perplexity value first decreases to a minimum point and then gradually increases [33]. The number of topic words when the perplexity reaches its lowest point is the optimal number of topics.

We used a Bidirectional Encoder Representations from Transformers (BERT) to classify the sentiment of the data. The BERT uses a bidirectional transformer architecture, allowing it to understand each word in a sentence from both left to right and right to left simultaneously [34]. This bidirectionality enables BERT to capture the full contextual information, aiding in more accurately understanding complex sentence structures and sentiment expressions in Weibo posts [35]. In addition, Weibo often contains a large amount of unstructured text, emojis, acronyms, and other elements [36]. BERT, with its powerful language modeling and feature extraction capabilities, can effectively handle these characteristics, capturing subtle emotional changes and implicit sentiment information [34]. The model assigned a sentiment score between 0 and 1 to each Weibo post. As the sentiment score increased, positivity increased. Scores between 0 and 0.5 indicated negative sentiment, a score of 0.5 indicated neutrality, and scores between 0.5 and 1 indicated positive sentiment [37]. We tested the model’s accuracy and ensured it was above 85%.

All statistical analyses were conducted using IBM SPSS (version 24.0) and Python software (Python Software Foundation). We performed Spearman correlation analysis to explore the relationship between the daily number of posts with the number of newly confirmed cases per day. We used an autoregressive integrated moving average (ARIMA) model to examine the relationship between sentiment scores and changes over time. The chi-square test was used to compare the proportions of different types of sentiment posts under various topics. The t test was used to compare the sentiment scores at different times within the same topic. A 2-tailed test was conducted, and P<.05 was considered statistically significant.

The public information involved in this study is freely available on Sina Weibo without a password. All data are de-identified. In addition, there is no identifiable information about individual users, IDs, or non-paraphrased posts in the manuscript or any supplementary materials. No individuals were recruited as study subjects and therefore no ethical approval was required for this study.

From April to July 2022, there were a total of 463,566 posts about nucleic acid testing on Weibo, with an average of 3799.72 (SD 1296.06) posts per day. The highest number of posts was 6689 on April 9th, while the lowest number of posts was 1551 on July 23rd. Figure 2 shows the changes in the daily number of posts on Weibo and the daily number of newly confirmed cases from April to July, indicating a positive correlation between the two (ρ=0.656, P<.001).

This paper explored public attitudes and opinions on regular nucleic acid testing on Weibo using sentiment and topic analysis. It analyzed public sentiment expressions on social media in the context of public health emergencies. To our knowledge, this is the first study to explore public attitudes and opinions on regular nucleic acid testing during the COVID-19 pandemic, filling a gap in existing research. The results of this study can help the government promptly understand changes in public sentiment during future emergencies, effectively guide public emotions, manage negative emotions, and maintain social order.

The study results showed a positive correlation between the daily number of posts on the topic of regular nucleic acid testing on Chinese social media and the number of new confirmed COVID-19 cases in China (ρ=0.656, P<.001). As the number of new confirmed cases increased, the government increased the frequency of nucleic acid testing to identify and control the outbreak more quickly. This led to more people participating in testing and sharing their experiences and views on social media. Our findings were consistent with a study on epidemic-related rumors on Weibo from January 20 to May 28, 2020, which found a positive correlation between the daily number of posts about COVID-19 misinformation on Chinese social media and the daily number of newly confirmed and newly suspected COVID-19 cases in China [40]. This indicated that the dissemination of information about COVID-19 in China was synchronized with the spread of the epidemic [40]. The volume of discussions on social media can serve as a supplementary indicator for predicting epidemic trends [41]. By monitoring public opinion on social media in real time, public health departments can promptly understand public concerns and provide early warnings for emerging outbreaks, thereby better formulating and adjusting control measures [41].

This study identified 8 main topics of public interest regarding regular nucleic acid testing. Ranked in order of attention from highest to lowest, they were the Shanghai outbreak, nucleic acid testing price, transportation, epidemic prevention and control measures, regular nucleic acid testing, epidemic notification, and hospital visits. Wang used the Weibo platform to investigate public concerns during the early stage of the pandemic (December 1, 2019, to July 31, 2020), identifying 8 topics, including epidemic notification, prevention measures, and health care [15]. In comparison, this study found that after the implementation of the regular nucleic acid testing policy, public attention shifted more toward topics such as the Shanghai outbreak, nucleic acid testing price, and transportation. Analyzing the reasons, the period focused on in this study coincided with the outbreak of the pandemic in Shanghai, during which multiple rounds of large-scale mass nucleic acid testing were conducted [42]. The frequent testing led to public concerns about the costs of nucleic acid testing, and any government policy adjustments regarding these costs attracted significant attention. Simultaneously, as many regions required travelers to provide proof of a negative nucleic acid test, public transportation and travel faced strict restrictions. Therefore, the Shanghai outbreak, nucleic acid testing price, and transportation have become focal points of public concern due to their close relation to public health and daily life. The government can determine the priority of public concerns based on the volume of posts on various topics on social media and promptly disclose relevant policies and information on epidemic prevention and control.

Sentiment analysis results showed that public emotions toward regular nucleic acid testing were predominantly positive, though there were periods of downturn influenced by certain events. The positive sentiment in Weibo posts indicated that the public generally approved of the regular nucleic acid testing policy, which is consistent with the conclusions of the study by Tan [43]. From June 2021 to August 2021, Tan [43] analyzed the opinions of residents in 4 cities in Guangdong Province on nucleic acid testing using semistructured interviews and social media comments. The results revealed that the public generally held a positive attitude toward nucleic acid testing [43]. Other studies have shown that sudden policy changes provoke heated discussions among the public on social media platforms [14]. If timely explanations and public communication are not provided, this can lead to public concern and anxiety, resulting in a surge of negative emotions [14]. On June 1, the Shanghai government suddenly announced adjustments to its pandemic prevention policies, leading to an increase in negative public emotions [39]. By June 2, the sentiment score had reached its lowest point. Fortunately, public negative emotions decreased afterward, and sentiment gradually stabilized. Support for regular nucleic acid testing increased over time. Decision makers should closely monitor changes in public reactions to address the most critical public needs at different stages of the pandemic [14]. By doing so, they can address public concerns through timely and effective communication [14]. In addition, policy makers can leverage social media and big data analysis to support public health decision-making [44].

Due to the economic burden on the public induced by regular nucleic acid testing, the government adjusted the prices of nucleic acid tests multiple times. On May 22, 2022, the National Healthcare Security Administration issued a notice to reduce the prices of nucleic acid testing [38]. However, after the policy was implemented, the sentiment score in Weibo posts, while still positive, was lower than before the policy was issued. Analyzing the reasons, first, although the cost of regular nucleic acid testing was primarily subsidized by the government, the requirement to undergo testing every few days disrupted the public’s daily life [45]. Second, when the price reduction policy was first announced, media reports sometimes highlighted inconsistencies between local and central policies or issues with policy implementation. Such negative reports could raise public doubts and increase the proportion of negative emotions. Studies have shown that after the introduction of new policies, insufficient government explanation or failures in public communication can greatly affect public perception and trust due to misinformation [46]. Therefore, when introducing new policies, the government should release information in a timely and transparent manner, including specific implementation details and progress, to reduce information asymmetry. During the policy implementation process, the government should establish diverse channels for public feedback, gather public opinions and suggestions, and promptly respond to and address public concerns to enhance the effectiveness of policy implementation.

Among all topics, the proportion of negative emotions regarding the “Shanghai outbreak” reached 36.60%, and the public sentiment score declined after June 1. Since the outbreak of the pandemic in March 2022, the number of infections in Shanghai had grown exponentially [47]. On April 1, the Shanghai government decided to implement city-wide static management, initiating a lockdown that lasted for 3 months [42]. Studies have shown that quarantine can lead to negative psychological effects such as posttraumatic stress symptoms, confusion, and anger [21,22]. On June 1, Shanghai fully resumed normal life [39]. However, during the actual process of resuming work and production, various issues were encountered, leading to strong negative emotions expressed by people on Weibo, which caused a decline in public sentiment scores. Furthermore, although the restoration of production and daily life was underway, the risk of a resurgence of the pandemic remained. This caused public concern about the future development of the pandemic, leading to a decline in sentiment scores. In addition, numerous rumors about the Shanghai outbreak circulated on the internet, contributing to negative emotions [23]. Previous research has found that public sentiment typically shifts from negative to positive after rumors are debunked [24]. Guo investigated public attitudes toward the Shanghai outbreak on the Weibo platform and found that public sentiment scores consistently declined during the outbreak, falling into negative territory in the second and third quarters [48]. The severe negative emotions among the public significantly affected Shanghai’s image [48]. Therefore, after announcing policies, the government needs to promptly track trends in public sentiment, identify potential negative emotions and hot issues, and respond effectively to control negative emotions within reasonable limits. In addition, the government should proactively guide public opinion by using authoritative media and expert interpretations to effectively disseminate policies related to public health emergencies, thereby reducing the spread of rumors. Furthermore, the public needs to enhance their ability to identify and not believe or spread rumors. Joint efforts from the government, media platforms, and the public are essential to create a positive and uplifting public opinion environment.

This study has several limitations. First, the Weibo data collected used only 1 keyword, which may include only a portion of posts related to nucleic acid testing rather than all relevant posts. Therefore, the summary of the main emotions and topics of Weibo users may not be comprehensive. Second, most Weibo users are young people, so the analysis results may be more biased toward the perspectives of young people [49]. Third, we collected data from only 1 social media platform. Although Weibo has a relatively large user base in China, it is hasty to conclude that the opinions expressed on Weibo represent the views of the public [50]. Finally, due to the limitations of the Weibo platform and its functionalities, most information regarding user geographical location and other details collected in this study was missing [51]. Consequently, the study did not explore whether users from different geographic locations had different sentiments and topics of concern regarding the regular nucleic acid testing policy.

During public health emergencies, the topics of public concern were multifaceted. Public sentiment toward regular nucleic acid testing policies was generally positive, but fluctuations occurred following key policy announcements. The government needs to monitor information posted by the public on social media platforms to understand the main areas of concern. By promptly disclosing information on media platforms and engaging in effective communication, the government can bridge the information gap between public needs and government responses to create a favorable public opinion environment.