Modeling the COVID-19 outbreak in the United States

Travis
Ziebro

Professionally, much of what I do is comb through a bunch of data and sort the wheat from the chaff. From there, my company builds software, models, and processes that help business owners figure out how to run their businesses better. When it comes to COVID-19 and its spread in America, I have seen a lot of chaff. This article and its associated model are an attempt to objectively assess a dangerous situation. Editor’s Note April 22nd: The model has been retired and is no longer accurate.

The model is based on journal studies when possible, a lot of literature, data, and inferences on pandemics. At the heart of the model is a simplified exponential function that decays over time. It uses numerous assumptions as stated below. I believe these assumptions to be the BEST possible assumptions based on empirical data and scholarly articles.

I hope you find the model useful. If you do, please like, comment on, and share this post. If it gets enough traction, I will expand and update the model as this pandemic moves forward.  Below are exhaustive instructions on how to use the model as well as the assumptions that went into building it.

How to Use the Model

How You Behave

AVERAGE NUMBER OF PEOPLE CONTACTED A DAY: This is how many people you contacted via a close proximity conversation or physically touched.

AVERAGE NUMBER OF SURFACES TOUCHED A DAY: This is how many surfaces you touch that day that others have touched in the last week. This is a “non-mitigated” touch, meaning your hands were not washed or sanitized immediately thereafter.

How Society Behaves

AVERAGE NUMBER OF PEOPLE CONTACTED A DAY: This is an estimate of how many people the average person will be in contact with via a close proximity conversation or physical touch. This is capped at 13 as 13 is the average amount of people touched by a person a day and social distancing measures have ensured this number will never be above 13.

AVERAGE NUMBER OF SURFACES TOUCHED A DAY: This is an estimation of how many surfaces the average person touches each day that others have touched in the last week. This is a “non-mitigated” touch, meaning a person’s hands were not washed or sanitized immediately thereafter. This is capped at 40 as 40 is near the average amount of common surfaces touched by a person a day and social distancing measures have ensured  this number will never be above 40.

Assumptions

There are three ways that someone can get COVID-19: zoonotic, direct, and indirect contact. The model rules out zoonotic for obvious reasons. Direct and indirect contacts are per the research studies above.

The model represents a doubling of cases every 6 days at the onset. This draws from four academic studies, conglomerated and referenced by Tomas Pueyo’s article Coronavirus: Why You Must Act Now. More sophisticated compartmental models look at how immunity impacts spread. The model does not include immunity impacts but includes a recovery rate roughly in line with Italy’s. Comparisons to China’s experience are generally discouraged due to vastly different cultures and China’s previous experience with highly infectious viruses.

The model assumes the number of active cases is a factor of the officially recorded cases. This is a difficult number to ascertain and is currently set at 20. That number will decrease if the testing response of America increases.

Several models have projected infection rates of over 100 million people. The model doesn’t see that as being feasible given the fairly extensive social distancing measures that have been put into place. The 100 million + model makes sense in the context of a normal, human contact rich scenario with no mitigation measures. 

Limitations of this Model

  • In a quick-moving, novel scenario such as COVID-19, the number of currently active cases is extremely difficult to pinpoint. The model uses a high multiplication factor based on previous studies and the low amount of tests available in America. This number is slowly decreasing as testing becomes more readily available.
  • The model assumes that the COVID-19 strain will behave like other novel viruses such as SARS and MERS and reach extinction. This is not a consensus opinion.
  • The date where the model “flips” to an exponential decay is a crucial number and is subjectively modeled at this point based on America’s response and the experiences of other nations.
  • Super spreaders are unbelievably effective at spreading viral diseases and impossible to model. In Korea, a single superspreader at one point was linked to 1/8 of the 8,236 cases in the entire country.

Updated – March 26th, 2020 – The second version of the model, released March 25th, 2020, adds the following functionality:

  • the use of published research that determines the epidemic peak based on user behavior.
  • the addition of a graph for easy visibility into the potential peaks of the COVID-19 epidemic.
  • the replacement of the “Days” column with “Official” cases, taken from Worldometers.com.

Be Kind and Love One Another,

Travis