COVID-Net: A deep learning based and interpretable predication model for the county-wise trajectories of COVID-19 in the United States (preprint)

The confirmed cases of novel coronavirus disease (COVID-19) have been reported in the United States since late January 2020. There were over 4.8 million confirmed cases and about 320,000 deaths as of May 19, 2020 in the world. We examined the characteristics of the confirmed cases and deaths of COVID-19 in all affected counties of the United States. We proposed a COVID-Net combining the architecture of both Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) by using the trajectories of COVID-19 during different periods until May 19, 2020, as the training data. The validation of the COVID-Net was performed by predicting the numbers of confirmed cases and deaths in subsequent 3-day, 5-day, and 7-day periods. The COVID-Net produced relatively smaller Mean Relative Errors (MREs) for the 10 counties with the most severe epidemic as of May 19, 2020. On average, MREs were 0.01 for the number of confirmed cases in all validation periods, and 0.01, 0.01, and 0.03 for the number of deaths in the 3-day, 5-day, and 7-day periods, respectively. The COVID-Net incorporated five risk factors of COVID-19 and was used to predict the trajectories of COVID-19 in Hudson County, New Jersey and New York County, New York until June 28, 2020. The risk factors include the percentage of the population with access to exercise opportunities, average daily PM2.5, population size, preventable hospitalization rate, and violent crime rate. The expected number of cumulative confirmed cases and deaths depends on the dynamics of these five risk factors.

Risk factors associated with mortality of COVID-19 in 2692 counties of the United States (preprint)

Background: The number of cumulative confirmed cases of COVID-19 in the United States has risen sharply since March 2020. A county health ranking and roadmaps program has been established to identify factors associated with disparity in mobility and mortality of COVID-19 in all counties in the United States. Methods: To find out the risk factors associated with county-level mortality of COVID-19 with various levels of prevalence, a negative binomial design was applied to the county-level mortality counts of COVID-19 as of April 15, 2020 in the United States. In this design, the infected counties were categorized into three levels of infections using clustering analysis based on time-varying cumulative confirmed cases from March 1 to April 15, 2020. COVID-19 patients were not analyzed individually but were aggregated at the county-level, where the county-level deaths of COVID-19 confirmed by the local health agencies. Findings: 2692 infected counties were assigned into three classes corresponding to low, medium, and high prevalence levels of infection. Several risk factors were significantly associated with the mortality counts of COVID-19, where elder (0.221, P=0.001) individuals were more vulnerable and higher level of air pollution (0.186, P=0.005) increased the mortality in the metropolis areas. The segregation between non-Whites and Whites had higher likelihood of risk of the deaths in all infected counties. Interpretation: The mortality of COVID-19 depended on sex, race/ethnicity, and outdoor environment. The increasing awareness of the impact of these significant factors may lead to the reduction in the mortality of COVID-19.

Data-driven analysis on the simulations of the spread of COVID-19 under different interventions of China (preprint)

Since February 2020, COVID-19 has spread rapidly to more than 200 countries in the world. During the pandemic, local governments in China have implemented different interventions to efficiently control the spread of the epidemic. Characterizing transmission of COVID-19 under some typical interventions is essential to help countries develop appropriate interventions. Based on the pre-symptomatic transmission patterns of COVID-19, we established a novel compartmental model: Baysian SIHR model with latent Markov structure, which treated the numbers of infected and infectious individuals without isolation to be the latent variables and allowed the effective reproduction number to change over time, thus the effects of policies could be reasonably estimated. By using the epidemic data of Wuhan, Wenzhou and Shenzhen, we migrated the corresponding estimated policy modes to South Korea, Italy, and the United States and simulated the potential outcomes for these countries when they adopted similar policy strategies of three cities in China. We found that the mild interventions implemented in Shenzhen were effective to control the epidemic in the early stage, while more stringent policies which were issued in Wuhan and Wenzhou were necessary if the epidemic was more severe and needed to be controlled in a short time.