MIDAS Webinar: Stochastic modeling of SARS-CoV-2 transmission in the US

July 31, 2020


July 31, 2020


The MIDAS Webinar Series features research by MIDAS members, and is open to the public. 

Date/time: Friday July 31, 12:00 – 1:00pm, EDT

Topic: Stochastic dynamical model of SARS-CoV-2 transmission in the US 

Speaker: Dr. John Drake, Distinguished Research Professor, Center for the Ecology of Infectious Diseases, University of Georgia 

John M. Drake is Director of the Center for the Ecology of Infectious Diseases at the University of Georgia and Distinguished Research Professor of Ecology in the Odum School of Ecology.  His research seeks to understand the dynamics of biological populations and epidemics, focusing on bringing experimental and observational data together with mathematical theory. Current projects concern the dynamics of Ebola virus in West Africa, spread of White-nose syndrome in bats, proliferation of the SARS-CoV-2 virus in the United States and globally, the development of the theory of infectious disease early warning systems, and infectious disease intelligence. Dr. Drake holds a PhD from the University of Notre Dame, and has been with the University of Georgia since 2006. He was Leverhulme Visiting Professor in the Department of Zoology at Oxford University in 2012.

We present a stochastic model for the transmission of the SARS-CoV-2 virus from March 2020 through the present in all states of the United States, for inference, forecasting, and scenario analysis. The model is a modified SEIR type stochastic model, with compartments for susceptible, latent, asymptomatic, undetected and detected symptomatic, diagnosed, hospitalized, recovered, and deceased. Compartments are split into multiple subcompartments using the linear chain trick to allow for more realistic distributions of movement through compartments. Model features include realistic interval distributions for presymptomatic and symptomatic periods; independent rates of transmission for asymptomatic, presymptomatic, and symptomatic individuals; time varying rates of case detection, isolation, and case notification; and realistic intervention scenarios. In order to model the effect of human behaviors on transmission, we include a relative human mobility covariate based on state-level cellular phone data, as well as a state-specific latent trend process, modeled using a fitted spline, that allows transmission to vary over time due to environmental factors and other behaviors that can reduce transmission but are difficult to measure (e.g. wearing a facemask). Fixed model parameters are defined using clinical outcome reports, and the model is fit to case incidence case and death reports for each state. We currently consider three scenarios for each state based on a change in relative human mobility; namely, return to pre-outbreak baseline mobility (i.e. normal mobility), maintenance of last measured mobility (i.e. status-quo mobility); and further reduction to 30% of baseline (i.e. shelter-in-place mobility). As of July 10, our model was forecasting continued strong exponential growth for the six weeks following July 10 in several states under both status quo and return-to-normal mobility scenarios, and suggests a return to shelter-in-place mobility could reverse the trend in cases and deaths in most locations.


No testimonials available.



There are no photos available.
This site is registered on as a development site.