Promote citizen engagement with warnings - an empirical examination of government social media accounts during public health crises.

BACKGROUND: Effective warnings are important for preventing the spread of disease during the early stages of outbreaks. Social media serves as a valuable platform for disseminating warning messages. The success of warnings issued through government social media accounts (GSMAs) depends on citizen engagement. However, an incomplete understanding of the relationship between warning messages and audience responses has hindered the design of crisis communication strategies. METHODS: We investigated the factors affecting citizen engagement with warnings on GSMAs during public health crises. Drawing on the Elaboration Likelihood Model (ELM) and the Crisis and Emergency Risk Communication (CERC) framework, model was developed to analyze the effects of central routes (content features) and peripheral routes (microstructural and source features) on citizen engagement, as well as the moderating effect of disease type. Data were collected from 38 Sina Weibo accounts of government agencies in China during two public health crises: COVID-19 and H1N1. Logit regression analysis was conducted to test the hypothesized relationships. RESULTS: The results indicate that (1) positive emotional tendencies and more warning elements are associated with citizen engagement; (2) the relationship between message length and citizen engagement follows an inverted U-shape; (3) media richness and information style variety significantly enhance citizen engagement; and (4) disease type (emerging vs. reemerging infectious diseases) moderates the relationships between media richness, information style variety, source influence, and citizen engagement. CONCLUSIONS: Given that issuing warnings is critical to emergency management, our findings provide significant theoretical and practical insights, particularly for improving early government-public communication through social media platforms. TRIAL REGISTRATION: Not applicable.