Public Medical Appeals and Government Online Responses: Big Data Analysis Based on Chinese Digital Governance Platforms.

Background: In the era of internet-based governance, online public appeals-particularly those related to health care-have emerged as a crucial channel through which citizens articulate their needs and concerns.

Objective: This study aims to investigate the thematic structure, emotional tone, and underlying logic of governmental responses related to public medical appeals in China.

Methods: We collected messages posted on the "Message Board for Leaders" hosted by People's Daily Online between January 2022 and November 2023 to identify valid medical appeals for analysis. (1) Key themes of public appeals were identified using the term frequency-inverse document frequency model for feature word extraction, followed by hierarchical cluster analysis. (2) Sentiment classification was conducted using supervised machine learning, with additional validation through sentiment scores derived from a lexicon-based approach. (3) A binary logistic regression model was employed to examine the influence of textual, transactional, and macro-environmental factors on the likelihood of receiving a government response. Robustness was tested using a Probit model.

Results: From a total of 404,428 online appeals, 8864 valid public medical messages were retained after filtering. These primarily concerned pandemic control, fertility policies, health care institutions, and insurance issues. Negative sentiment predominated across message types, accounting for 3328 out of 3877 (85.84%) complaints/help-seeking messages, 1666 out of 2381 (69.97%) consultation messages, and 1710 out of 2606 (65.62%) suggestions. Regression analysis revealed that textual features, issue complexity, and benefit attribution were not significantly associated with government responsiveness. Specifically, for textual features, taking the epidemic issue as the reference category in the appeal theme, the P values were as follows: fertility issue (P=.63), hospital issue (P=.63), security issue (P=.72), and other issues (P=.34). Other textual features include appeal content (P=.80), appeal sentiment (P=.64), and appeal title (P=.55). Regarding the difficulty of resolving incidents, with low difficulty as the reference category, the P values were moderate difficulty (P=.59) and high difficulty (P=.96). For benefits attribution, using individual interest as the reference, collective interest (P=.25) was not statistically significant. By contrast, macro-level factors-specifically internet penetration, education, economic development, and labor union strength-had significant effects. Compared with areas with lower levels, higher internet penetration (odds ratio 1.577-9.930, P=.004 to <.001), education (odds ratio 2.497, P<.001), and gross domestic product (odds ratio 2.599, P<.001) were associated with increased responsiveness. Conversely, medium (odds ratio 0.565, P<.001) and high (odds ratio 0.116, P<.001) levels of labor union development were linked to lower response odds.

Conclusions: Public medical appeals exhibit 5 defining characteristics: urgency induced by pandemic conditions, connections to fertility policy reforms, tensions between the efficacy and costs of medical services, challenges related to cross-regional insurance coverage, and a predominance of negative sentiment. The findings indicate that textual features and issue-specific content exert limited influence on government responsiveness, likely due to the politically sensitive and complex nature of health care-related topics. Instead, macro-level environmental factors emerge as key determinants. These insights can inform the optimization of response mechanisms on digital health platforms and offer valuable theoretical and empirical contributions to the advancement of health information dissemination and digital governance within the health care sector.