Thread by @MichaelWorobey

Very glad you raised this @bencowling88!

First, 9 infections by Dec 1 is the median, but the 95% HPD included 26 cases.

Second, and more importantly, I think this issue here is doubling time. Your paper below estimated a doubling time in Wuhan of 8.7 days.
1/

https://twitter.com/bencowling88/status/1470892299093950467

The fullness of time has shown that is unrealistically slow. Based on Bao et al, we estimated a median 4.1 day doubling time in Wuhan in Jan-Feb.

Nov-Dec, before anyone know what was going on, the epidemic would likely have been doubling even quicker.
2/

nature.com/articles/s4158…

Probably less than every 3 days.

But take 100 cases by Dec 1. Even assuming a 4.1 day doubling time, you're looking at 7-8 doublings in December. That means 12,800 to 25,600 infections. The WHO-China report reports 168 cases in Wuhan with symptom onset before Dec 31.
3/

Let's assume that 7% of them are hospitalized and therefore ascertained: 896-1792. Way too high.

9 cases by Dec 1: 80-161 onset by Dec 31. A little lower than observed.

26 cases by Dec 1 (our upper bound): 232-465 onset by Dec 21.

Clearly our estimates fall are realistic.
4/

And that is not even taking into account that a zero-mitigation Wuhan likely saw doubling time quicker than 4.1 days.

At any rate, 9 and 26 and 100 are not so different. With only about 7% of this cases ending up in hospital some 2 weeks after infection...
5/

... and only a small fraction, say 7%, ending up in hospital and ascertained, the earliest known patient at Dec 10 (~2 weeks from infection), and 168 patients in Wuhan with onset in December, is fully consistent with a mid to late November origin of the pandemic.
6/

In fact, it is *extremely* unlikely, given how contagious even version 1.0 of this virus was, that the first case could have been before mid to late November, or that there could have been more than 10 to 25 infected patients in November.
7/7