Modeling COVID-19 Infection Rates Across Counties with Specificity
Using a flow-augmented stochastic SEIR style model to analyze the epidemic
Due to COVID-19’s extended latency period between when an individual is infected and infectious, the susceptible-exposed-infectious-removed (SEIR) style of modeling infectious diseases is best. However, there are still shortcomings in that it assumes homogeneity of the population.
This consistency cannot be expected in real situations, given the various geographic, socioeconomic, and cultural factors at play. Intracounty environments also differ greatly, creating spatial heterogeneity of transmission in different regions.
To account for this, recent research by Xiao Hou, Song Gao, Qin Li, Yuhao Kang, Nan Chen, Kaiping Chen, Jinmeng Rao, Jordan S. Ellenberg, and Jonathan A. Patz developed a human mobility flow-augmented stochastic SEIR-style epidemic modeling framework that allowed them to distinguish different regions and their behavior.
They plot estimated effective reproduction number Rₑ and daily confirmed cases in each region (with a seven day average for smoothing out the possible day-of-week reporting and testing bias). Region 7, which is the downtown Madison and university campus area, has the highest Rₑ, while region 3, the adjourning residential area, has a substantially lower Rₑ, even though it also has a high population density.
Figure 3 shows the flow traffic normalized effective reproduction number in Dane County, which controls for the fact that some regions have higher mobility flows than others. However, in this case, region 7 has significantly higher Rₑ even after being normalized. This suggests that there are social characteristics beyond population mobility that are contributing to rapid spread in this area.
Make sure you check out the full research paper:
Intracounty modeling of COVID-19 infection with human mobility: Assessing spatial heterogeneity with business traffic, age, and race
Xiao Hou, Song Gao, Qin Li, Yuhao Kang, Nan Chen, Kaiping Chen, Jinmeng Rao, Jordan S. Ellenberg, and Jonathan A. Patz
PNAS June 15, 2021 118 (24) e2020524118; doi: https://doi.org/10.1073/pnas.2020524118
To facilitate this research, they relied heavily on SafeGraph point of interest and mobility data. To learn more about how you can use SafeGraph yourself, and how others are already using it, join the Placekey community.