Dear Pandemic: A topic modeling analysis of COVID-19 information needs among readers of an online science communication campaign

Abstract

Background The COVID-19 pandemic was accompanied by an “infodemic”–an overwhelming excess of accurate, inaccurate, and uncertain information. The social media-based science communication campaign Dear Pandemic was established to address the COVID-19 infodemic, in part by soliciting submissions from readers to an online question box. Our study characterized the information needs of Dear Pandemic’s readers by identifying themes and longitudinal trends among question box submissions. Methods We conducted a retrospective analysis of questions submitted from August 24, 2020, to August 24, 2021. We used Latent Dirichlet Allocation topic modeling to identify 25 topics among the submissions, then used thematic analysis to interpret the topics based on their top words and submissions. We used t-Distributed Stochastic Neighbor Embedding to visualize the relationship between topics, and we used generalized additive models to describe trends in topic prevalence over time. Results We analyzed 3839 submissions, 90% from United States-based readers. We classified the 25 topics into 6 overarching themes: ‘Scientific and Medical Basis of COVID-19,’ ‘COVID-19 Vaccine,’ ‘COVID-19 Mitigation Strategies,’ ‘Society and Institutions,’ ‘Family and Personal Relationships,’ and ‘Navigating the COVID-19 Infodemic.’ Trends in topics about viral variants, vaccination, COVID-19 mitigation strategies, and children aligned with the news cycle and reflected the anticipation of future events. Over time, vaccine-related submissions became increasingly related to those surrounding social interaction. Conclusions Question box submissions represented distinct themes that varied in prominence over time. Dear Pandemic’s readers sought information that would not only clarify novel scientific concepts, but would also be timely and practical to their personal lives. Our question box format and topic modeling approach offers science communicators a robust methodology for tracking, understanding, and responding to the information needs of online audiences.