The novel coronavirus disease (COVID-19) generated significant health concerns worldwide, leading policymakers and health care experts to implement nonpharmaceutical public health interventions to mitigate the spread of the virus. While these interventions played a crucial role in controlling transmission, they also resulted in substantial economic and societal costs, necessitating strategic deployment, particularly during periods of increasing disease activity. In this context, geosocial media posts with explicit georeferences, have emerged as a promising tool for anticipating potential health care crises. However, previous studies investigating the early warning capabilities of geosocial media data have been limited by coarse spatial resolutions or short temporal scopes, with little understanding of how local political beliefs might influence these capabilities.
This research aimed to evaluate how the epidemiological early warning capabilities of geosocial media posts for COVID-19 vary over time and across U.S. counties with differing political beliefs. To achieve this, U.S. counties were classified into three political clusters—democrat, republican, and swing counties—based on voting data from the last six federal election cycles. The early warning capabilities of geosocial media posts were then analyzed across six consecutive COVID-19 waves spanning from February 2020 to April 2022. Specifically, the study examined the temporal lag between geosocial media signals and surges in COVID-19 cases, measuring both the number of days by which geosocial media signals preceded surges in COVID-19 cases and the correlation between their respective time series.
The findings indicate that the early warning capabilities of geosocial media data varied across political clusters and COVID-19 waves. On average, geosocial media posts preceded COVID-19 cases by 21 days in republican counties, compared with 14.6 days in democrat counties and 24.2 days in swing counties. In five out of six waves, geosocial media posts consistently preceded COVID-19 cases across all political clusters. However, a decline over time in the number of days that posts preceded COVID-19 cases was observed, particularly in democrat and republican counties. Additionally, a reduction in signal strength and the influence of trending topics posed challenges to the reliability of early warning signals.
This study provides valuable insights into the strengths and limitations of geosocial media data as an epidemiological early warning tool, particularly by highlighting variations across county-level political clusters. These findings suggest that future geosocial media-based epidemiological early warning systems may benefit from incorporating political beliefs into their models. Furthermore, the observed decline in geosocial media signal strength over time and the influence of trending topics on signal reliability warrant further investigation in future research.
Reference:
Title image: Geospatial distribution of political belief clusters on county level based on the last 6 election cycles.