Machine learning analysis of government's public risk communication during COVID-19 lockdown in Wuhan, China

Abstract

The COVID-19 pandemic has caused millions of deaths worldwide since 2020, and has led to sizable health and economic costs. Although COVID-19 cases were first detected in Wuhan, over time the death rate attributable to COVID-19 was one of the lowest globally. Understanding the Chinese government's focuses and strategies can offer insights into early pandemic control in the future. This study aimed to explore the strategies and practices of government risk communication adopted during the COVID-19 lockdown in Wuhan, China; and provide implications for effective health risk communication at the early stage of epidemic response. The 90 government press conference records during the Wuhan lockdown (from January 22, 2020 to May 1, 2020) were divided into three batches and preprocessed. Topic modeling, i.e., the Latent Dirichlet Allocation, was used to computationally extract the topics in each batch. We identified important topics early in the lockdown period, such as “medical team's work”, “assuring supplies for society”, “patients' detection and isolation”; in the middle batch such as “patient treatment and hospitalization”, “enterprises' resumption of work and production”, “epidemic prevention and control”; and later in the lockdown including “policies supporting enterprises”, “ensuring employment”, as well as “blood donation”. We found that communicators from various government sectors provided consistent, concise information during the pandemic. Sectors involved included health, transportation, employment, agriculture, banking, industrial recovery, and resource deployment. Our results suggest that pandemic control requires not only effective public health policies but also collaboration and collective action across diverse societal systems.