ABSTRACT
ObjectiveWe sought to understand how U.S. residents responded to COVID-19 as it emerged, and the extent to which spatial-temporal factors impacted response. Materials and MethodsWe mined and reverse-geocoded 269,556 coronavirus-related social media postings on Twitter from January 23rd to March 25th, 2020. We then ranked tweets based on the socioeconomic status of the county they originated from using the Area Deprivation Index (ADI); that we also used to identify areas with high initial disease counts ("hotspots"). We applied topic modeling on the tweets to identify chief concerns and determine their evolution over time. We also investigated how topic proportions varied based on ADI and between hotspots and non-hotspots. ResultsWe identified 45 topics, which shifted from early-outbreak-related content in January, to the presidential election and governmental response in February, to lifestyle changes in March. Highly resourced areas (low ADI) were concerned with stocks, social distancing, and national-level policies, while high ADI areas shared content with negative expression, prayers, and discussion of the CARES Act economic relief package. Within hotspots, these differences stand, with the addition of increased discussion regarding employment in high ADI versus low ADI hotspots. DiscussionTopic modeling captures the major concerns in COVID-19-related discussion on a social media platform in the early months of the pandemic. Our study extends previous studies that utilized topic modeling on COVID-19 related tweets and linked the identified topics to socioeconomic status using ADI. Comparisons between low and high ADI areas indicate differential Twitter discussions, corresponding to greater concern with economic hardship and impacts of the pandemic in less resourced communities, and less focus on general public health messaging. ConclusionThis work demonstrates a novel framework for assessing differential topics of conversation correlating to income, education, and housing disparities. This, with integration of COVID-19 hotspots, offers improved analysis of crisis response on Twitter. Such insight is critical for informed public health messaging campaigns in future waves of the pandemic, which should focus in part specifically on the interests of those who are most vulnerable in the lowest resourced health settings.
