Health Communication Through News Media During the Early Stage of the COVID-19 Outbreak in China: A Digital Topic Modeling Approach

Background : In December 2019, some COVID-19 cases were first reported and soon the disease broke out. As this dreadful disease spreads rapidly, the mass media has been active in community education on COVID-19 by delivering health information about this novel coronavirus. | Methods: We adopted the Huike database to extract news articles about coronavirus from major press media, between January 1st, 2020, to February 20th, 2020. The data were sorted and analyzed by Python software and Python package Jieba. We sought a suitable topic number using the coherence number. We operated Latent Dirichlet Allocation (LDA) topic modeling with the suitable topic number and generated corresponding keywords and topic names. We divided these topics into different themes by plotting them into two-dimensional plane via multidimensional scaling. | Findings: After removing duplicates, 7791 relevant news reports were identified. We listed the number of articles published per day. According to the coherence value, we chose 20 as our number of topics and obtained their names and keywords. These topics were categorized into nine primary themes based on the topic visualization figure. The top three popular themes were prevention and control procedures, medical treatment and research, global/local social/economic influences, accounting for 32.6%, 16.6%, 11.8% of the collected reports respectively. | Interpretation: The Chinese mass media news reports lag behind the COVID-19 outbreak development. The major themes accounted for around half the content and tended to focus on the larger society than on individuals. The COVID-19 crisis has become a global issue, and society has also become concerned about donation and support as well as mental health. We recommend that future work should address the mass media9s actual impact on readers during the COVID-19 crisis through sentiment analysis of news data.