A Chronological and Geographical Analysis of Personal Reports of COVID-19 on Twitter

Ari Klein; Arjun Magge; Karen O'Connor; Haitao Cai; Davy Weissenbacher; Graciela Gonzalez-Hernandez

In Figure 1 , the 12 states with the most users who have posted "probable" or "possible" tweets include 11 of the top 12 states reporting COVID-19 cases to the CDC 7 . These personal reports on Twitter began to increase sharply around the beginning of March, as shown in Figure 2 , but not until the middle/end of March for cases reported to the CDC 7 . For many states, we have detected "probable" or "possible" tweets that were posted before the state's first confirmed case. Thus, this study demonstrates that (1) users do report personal information on Twitter that could broadly indicate potential exposure to COVID-19, (2) these personal reports can be understood as signals of COVID-19 cases, and (3) our social media mining approach could help provide an early indication of the spread of COVID-19. Despite these promising preliminary results, many . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.