1) 17.3 days is Tomas Pueyo's figure for the mean time between infection
and death, for those who die. "We know approximately how long it takes for
that person to go from catching the virus to dying on average (17.3
days)". He also gives a link to the source:
https://github.com/midas-network/COVID-19/tree/master/parameter_estimates/2019_novel_coronavirus

The relevant section appears to be "Time from symptom onset to death"
which quotes 4 studies, whose average time is 17.3. As I'm sure you're
noticing, symptom onset is not the same as "catching the virus". So we
should add the mean incubation time, which is also shown on the github
page, with 9 studies and a mean value of 5.75. So the real figure for time
from infection to death appears to be 23 days.

This will alter my estimates too, as the number of cases on 1 Mar (35)
gets replaced by the number of cases on 24 Feb (13). If real case numbers
are growing by a factor of 1.3 per day, the estimated true case numbers
become:

on 18 Mar: 10,000 * 1.3^23 = 4.2 million if we assume death rate 1% or
around 1 million if we assume death rate 4%.

How plausible is it that real cases are growing by a factor of 1.3 per
day? It's certainly true for the confirmed case numbers since 29 Feb. For
deaths since 29 Feb, the github data (linked in my recent email) shows the
first death on 5 March. From then on, a linear fit to the log Cumulative
Deaths has R2 = 0.97 and the daily growth is 1.37. see attached.

So, assuming that the number of deaths is proportional to the true number
of cases, and that proportion has remained constant in this period, it
seems very plausible that true cases are growing at 1.37, and 1.3 is
conservative.

Finally, whilst the coming strain and possible collapse of hospital care -
we hope not - will affect the deaths, it will not affect the growth rate
of true cases. No vaccine, no therapy, no herd immunity (for some time)
means that only massive changes in R0 by behaviour will affect that.

2) I agree with your second point. UK cases are identified by testing,
which is now primarily in hospital and follows symptoms of respiratory
disease.
Pueyo's point is to work back from the deaths to the true cases, and
although I think 17.3 days is wrong, his idea makes sense to me.

Greg

> At 14:50 19/03/2020, Greg Dropkin wrote:
>>... But I end up close to your rule of thumb, because it's not 10 days,
it's 17
>
> Is that 17 days the estimate of the interval between infection (or maybe
onset of symptoms) and death?
>
> If so, if, in the UK we are (and probably always were) primarily
detecting (and counting/reporting) 'cases' only when they became ill
enough to require hospitalisation, then I imagine that, in the UK, the
interval between that counting/reporting and death could be a lot less
than 17 days, couldn't it?
>
> Kind Regards,
>
>
> John
>
> ---------------------------------------------------------------- Dr John
Whittington,       Voice:    +44 (0) 1296 730225
> Mediscience Services       Fax:      +44 (0) 1296 738893
> Twyford Manor, Twyford,    E-mail:   [log in to unmask]
> Buckingham  MK18 4EL, UK
> ----------------------------------------------------------------
>
>










******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
Disclaimer: The messages sent to this list are the views of the sender and cannot be assumed to be representative of the range of views held by subscribers to the Radical Statistics Group. To find out more about Radical Statistics and its aims and activities and read current and past issues of our newsletter you are invited to visit our web site www.radstats.org.uk.
*******************************************************