Center for Quality, Safety & Value Analytics
Exponential: The exponential model has been widely successful in capturing the increase in COVID-19 cases during the most rapid and difficult-to-mitigate phases. The exponential model takes a simple form and essentially captures the effect of repeated doubling over time (1, 2, 4, 8, …). In our modeling, predicted values of the exponential were determined by linear regression conducted on log(numbers of cases).
Quadratic 2nd order polynomial: A 2nd order polynomial (y ~ x2 + x) captures quadratic growth and is the expected outcome when the growth rate changes and when that rate of change is constant. The rate of increase in this model is initially faster than that of the exponential model. We obtained predicted values from this model using numerical optimization of parameters and curve fitting.
Logistic: When exponential growth slows and tapers-off, the growth curve often becomes logistic, that is, “S” shaped. The rate of increase in this model is initially exponential but slows as an upper limit is approached. We obtained predicted values from this model using numerical optimization of parameters and curve fitting.
SEIR-SD: An epidemiological model that attempts to predict the changes in numbers of people who are susceptible to COVID-19 infection, who have been exposed to COVID-19, who have been infected and are symptomatic, and who have recovered. This SEIR model uses the total population size of your chosen location (source: US Census 2010-2019). It attempts to infer the date of the first COVID-19 case in your chosen location, the average incubation period of COVID-19, average infectious period, and the initial reproductive number. The 'SD' in the model pertains to the inclusion of social distancing. The model assumes that social distancing naturally becomes more prevalent as the percent of the population infected with COVID-19 increases. The model also assumes that testing for COVID-19 was initially low but increased during the weeks following the first reported cases in the US.
Coefficients of determination (r-square values): These pertain to the relationship of observed values vs. predicted values and so, reveal the percent of variation in the observed values explained by the predicted values.Developer
Ken Locey, PhD, Data Science AnalystSite Architect and Administrator
Jawad Khan, AVP, Advanced Analytics & Knowledge ManagementCenter for Quality, Safety & Value Analytics Leadership
Thomas A. Webb, MBA, Associate Vice President
Bala N. Hota, MD, MPH, Vice President, Chief Analytics Officer