Understanding COVID-19-related Discussion and User Engagement on the LIHKG Forum

The COVID-19 pandemic has continued to wreak havoc on public health since the first case of the infection appeared in Wuhan, China, in 2020. In Hong Kong, the fifth wave of COVID-19 claimed more than 6,300 lives. During the pandemic, LIHKG, a popular online forum in Hong Kong, has provided a key venue where Hong Kong residents can discuss the issues related to the pandemic and share their opinions and experiences with others. What are emergent topics regarding COVID-19 in this online forum? What kinds of topics attract greater public attention? To this end, this study aims to identify COVID-19-related topics that are prevalent on LIHKG and investigate to what extent these topics would predict the level of user engagement.

First, we downloaded all 127,339 posts on LIHKG between December 20, 2021, and March 31, 2022, which is the fifth wave of COVID-19. Next, we filtered out COVID-19-related messages using a keyword filter containing 459 keywords related to the pandemic, which generated a final dataset of 5,827 posts. We performed the latent Dirichlet Allocation (LDA) topic modeling and regression analyses to detect COVID-19-related topics and examine the relationships between these topics and the level of user engagement (i.e., the number of comments, dislikes, and likes).

Topic modeling of posts yielded four overarching topics: COVID-19 testing, COVID-19 vaccine, COVID-19 government policy, and COVID-19 prevention measures. Specifically, the most common topic was COVID-19 prevention measures (32.2%), followed by COVID-19 testing (23.9%), COVID-19 vaccine (23.1%), and COVID-19 government policy (20.8%). The topic of COVID-19 testing encompassed information and discussions about testing rules in Hong Kong, confirmed cases, and the symptoms of infection. The topic of the COVID-19 vaccine addressed issues related to vaccination such as vaccine pass, safety, and inoculation rate. The topic of COVID-19 government policy contained discussions of government strategies to control the pandemic in Hong Kong, China, and the U.S., such as the Zero-COVID policy and co-existence with the virus. Last, the topic of COVID-19 prevention measures addressed a range of information and discussions regarding prevention behaviors, such as quarantine rules, staying in isolation facilities, contacting the hotline, and mask-wearing.

Results from the ordinary least squares regression analyses indicated that all four topics were positively associated with the number of comments; the topics of COVID-19 testing (β = .075, p < .001), vaccine (β = .060, p < .001), government policy (β = .075, p < .001), and prevention measures (β = .046, p < .001), predicted a greater number of comments. Also, all the topics were positively associated with the number of dislikes; the topics of COVID-19 testing (β = .032, p = .02), vaccine (β = .043, p = .001), government policy (β = .031, p = .02), and prevention measures (β = .041, p = .002), predicted a greater number of dislikes. Last, we found that the topics of COVID-19 testing (β = .026, p = .05), vaccine (β = .026, p = .05), and government policy (β = .037, p < .001) predicted a greater number of likes, whereas the topic of prevention measures was not associated with it.

In conclusion, this study illustrates the key topics that Hong Kong residents mainly seek and engage in discussions concerning COVID-19 in the online forum. Our findings offer several implications. First, the categorization of COVID-19-related topics can produce useful information about the trends in the public discussion of the pandemic. Particularly, combining topic modeling with regression analysis provides a more nuanced understanding of which topics could drive a higher level of user engagement. Furthermore, our findings can help health agencies and healthcare professionals to understand public conversations and concerns about the pandemic and develop more targeted and evidence-based strategies.