Abstract
This study investigates early COVID-19 communication strategies and content of four public health and four news media organizations across Australia, China, the UK, and the US on X (formerly Twitter) and their public engagement. 15,711 COVID-19-related tweets from the selected accounts posted from January 1 to May 19, 2020, were collected using a web crawler. Public engagement was measured through replies, retweets, and likes. The tweets were grouped into 37 clusters using unsupervised learning and analyzed thematically based on the top 30 tweets per cluster. Descriptive statistics quantified the tweets and their engagement, and a chi-square test compared differences between the two organization types across topics. Six topics were identified: "policies, methods, and action", "case updates", "opinions and responses", "medical research and treatment information", "impacts and consequences", and "health instructions and suggestions", with five communication phases: inception, awareness, panic, spreading, and cohabitation. Analysis revealed that more hashtags and longer texts were associated with lower engagement, except for tweets on "medical research and treatment information" and "health instructions and suggestions". Our study indicated that fewer hashtags and concise language might improve public engagement, while detailed content with more hashtags was effective for specific health instructions. Overall, tweets about treatment progress and health guidance received the most engagement.