Thread by @balajis

Calibrate intuition from math. Assume *roughly*:

2% fatality rate (JAMA)
~4 weeks from infection to death for fatal cases (WHO)
7 day doubling time (NEJM)

Then a report of one death today would mean 50 infected four weeks ago — and 800 cases today.

These numbers are VERY approximate but help calibrate intuition. For every 1 reported death, there can be 100-1000X as many cases out there.

REFERENCES

1) Review of JAMA study: cidrap.umn.edu/news-perspecti…

2) WHO: who.int/docs/default-s…

3) NEJM article: nejm.org/doi/full/10.10…

You can & should plug in your own numbers there if you think fatality rates are lower, doubling rate is different, etc.

But one robust result is that under many reasonable assumptions, if you see one death there is a large right tail on estimate of how many cases are out there.

Btw, one item that doesn’t affect substance of point:
who.int/docs/default-s…

Infection to symptom: avg 5-6 days, range 1-14
Symptom to death *for fatal cases*: 2-8 wks

OP assumes *roughly* ~4 wks from infection to death for fatal cases. But very wide range here.

Looks like @diviacaroline did a very similar calculation last week.

One way of thinking about it is "blast radius" of a reported death. Depending on assumptions, each reported death can mean 100-1000X more cases out there.

https://twitter.com/diviacaroline/status/1231830534029922304