How do Behaviors Impact Epidemics?

The Goal

Our goal is to leverage the ICC perspective on epidemics to give us new insight on how behavioral/policy/other changes impact the spread. The goal is to show a correlation between major behavioral events (such as holidays) as well as government orders (such as stay-at-home orders) and the subsequent change in projected case values. 

A Quick Overview

When considering an epidemic, typically this does not occur in one single wave. Rather, it occurs in multiple waves, with individuals reacting, changing their behaviors and the virus mutating. To best account for this, we need to use a method which identifies when things change. As a quick exercise, look at the images below and see if you can identify when the epidemic is changing. 

Which curve makes identification easier? The ICC Curve or the Epi-Curve?

The ICC Curve eliminates the exponential growth and decay which can be hard to analyze and pin-point exactly when things change. This task is much easier for us to do on the ICC Curve. And, as it turns out, can be leveraged for a computer to do this as well. 

With these newly identified potential points, we can segment our curve and use an inference method motivated by previous research to fit ICC curves to each segment. The results are below

With this we can identify the exact dates that the algorithm says things changed and report the found parameters describing exactly how much they changed by! We plot these results in both the case and time domain below.