My COVID-19 disease modeling findings for Washington State

Image from CDC

I recently fit the Washington state COVID-19 epidemiologic curves to both a SIR and SEIR dynamic disease model to address a few questions that I had about disease control. I developed the modeling in R Shiny, so that the models would be interactive allowing others to vary the assumptions to observe the results themselves. You can find the modeling tool here:

The website explores several questions, including:

  • How many people will be infected?
  • How does social distancing reduce transmission?
  • Does the timing of interventions matter?
  • How long do we need to keep doing interventions?
  • What is R0 vs. what is R?
  • Why we should be conducting seroprevalence testing?
  • What are the impacts of travel restrictions, school closures, social distancing, etc?

There are some interesting findings that I’m still exploring. For instance:

  • The more successfully we are in social distancing, the longer the delay in reaching peak infections.
  • Timing is very important.  It is less important to start early intervention as it is to time it just right to clip the peak of the infections when the force of infection is great.
  • Also with respect to timing, there is a risk of a double peak of infection if we stop interventions when there are still a large proportion of the community that is still susceptible. Based on my R0, the threshold for potentially having a double blip of infection is if there are more that 25% of the population that is still susceptible.

The code is publicly available on my github site: