Estimate of COVID-19 Cases in US

Timothy Blumberg
2 min readMar 11, 2020

--

Disclaimer: I am not an epidemiologist or statistician and I have most likely made grave errors in my analysis. Please do not take this post as evidence of anything other than a concerned citizen’s unwarranted attempt to answer the question of how many COVID-19 cases are present in the US.

Date published: March 11th, 2020

I am personally growing worried by the lack of testing for COVID-19 in the US, as it seems that this is a concerted effort to keep the public from panicking.

Things we Know

Median time from disease onset to death: 18.5 days (Table 2 discussion)

Mean incubation period: 5–6 days pg 12

Crude fatality rate: 3.8% pg 12

Calculated crude fatality ratio: 3.67%. This number was calculated using the Johns Hopkin’s COVID-19 dashboard: total deaths / total confirmed cases. At time of writing: 4610(deaths)/125664(confirmed).

Logistic regression of cases outside of china yields a growth factor of ~1.188

Current Situation in US

The afore-linked dashboard from Johns Hopkins currently lists 1135 confirmed cases and 32 confirmed deaths in the US.

The thing is, these 32 deaths died from a population of cases that existed >18.5 days ago. We can use the crude fatality rate, to back calculate how many cases we should have had ~19 days ago in the US. (There are many problems with this approach as not all of these people died today / were infected in the US).

32(deaths) / 3.67%(deaths/cases) => 872 cases (~20 days ago)

We can then (crudely) apply this to our ideal exponential growth function:

y = predicted value; a = starting value; b = growth factor; x = time (more on this)

This yields:

Estimate for March 11th, 2020

Final Notes

There are many problems and errors in this analysis, and I actively invite anyone to call out these errors as you see them.

It has been eye-opening for me to see just how blind we are flying without proper testing in the US. We have no idea how many cases there are and furthermore stand zero chance of containing the spread. It could be way less, it could be way more, we haven’t any idea.

If you do come down with even mild symptoms, please inform those you had contact with over the preceding 5 days so they can self-quarantine and do their part to stop the spread.

Links

Links to smarter people doing this:

--

--

Timothy Blumberg

Eternal Learner and Programmer; Communicates poorly in Chinese; Working on Crypto.