At 18:49 28/03/2020, Greg Dropkin wrote:
>hi all ... COVID-19 forecast for the UK lockdown
>http://www.labournet.net/other/2003/lockdown1.html
>please fwd this url wherever you think useful. ... and, comments / 
>corrections please.

Greg, I'm sure you will take it in the right spirit when I say that I 
have to agree with the final comment in your conclusions when you say 
that "you hope you are wrong" - and, as I'm sure is true of you and 
many others, I have been trying hard to think of ways in which you 
may be 'wrong'.

I imagine that there is nothing intrinsically wrong with the model 
you are using.  Within my (seriously) limited ability to comment, the 
model looks fair enough and the assumptions on which you have based 
your modelling seem reasonable enough.

The best I can do is ask whether you are 'fairly' modelling what the 
situation actually was.  You are assuming that the 'lockdown' (which 
you hypothesis reduced transmission to 10% or 15%) started suddenly 
on 24th March, but the true situation was not as simple as that.

Over a week before that, on 16th March, following a 'warning' the day 
before that a 'lockdown' was going to come, the PM advised 
'social-distancing'.  Non-essential travel (and going to pubs/clubs 
etc.) was advised against, home working advocated and, in particular, 
he strongly advised the most vulnerable (>70, pre-existing disease of 
pregnant) to 'self-isolate' , with that advice becoming progressively 
stronger over the coming days, coupled with a warning that it would 
soon become 'compulsory'.

There clearly was at least some degree of response to that advice on 
the part of the general public but, for what it's worth, anecdotally 
speaking, the majority of 'vulnerable people I know seemed to comply 
with the self-isolation fairly soon after it was suggested/advised. 
Furthermore, again anecdotally, even in relation to mainly 
non-vulnerable people, by/around Monday 16th March, a high proportion 
of people \i work with were already working from home.

I would therefore suggest that a 'more complete' model might, in 
addition to what you've already got, also include:

1... A period (maybe about a week) prior to 24th March during which 
there was a fairly modest (your guess is as good as mine!) 
progressively increasing reduction in transmission within the general 
population
PLUS, probably more importantly,
2... A period (again maybe about a week) there was a more 
substantial, and progressively increasing, reduction in transmission 
TO those most likely to end up in hospital, ICU or dead if they 
became infected.

Incorporating those into the model would obviously make it 
appreciably more complex (I don't know how easily it can cope with 
progressive, rather than 'step', changes in transmission, nor whether 
it can cope with different transmission in sub-groups with different 
prognoses), but it would seem closer to reality - albeit the actual 
figures you fed in would presumably be little more than blind guesses.

Those additional (earlier) effects, if valid, would presumably have 
had an impact on 'new cases' by now, but there is really  no way we 
can conclude very much from the daily figures we are now being told 
in the UK, which derive from ever-increasing numbers of 
tests.  However, we presumably are fairly close to the time at which 
any such ('early') effects should start being reflected in 
hospitalisations, ICU admissions and deaths.

I think the general principle I'm talking about, of trying to model 
something as close as possible to the actual situation, cannot be 
wrong - but I don't know how easy it would be to implement, and 
certainly haven't got  clue as to how one would guess the relevant 
parameters.  I suppose the interesting thing to do, if such a model 
could be constructed, would be to explore the impact of varying those 
(additional) parameters - essentially to see whether what I'm 
suggesting could/would make any appreciable difference to your forecasts.

Just one other point/question ... your ("15% reduction") forecast of 
150,000 deaths by June 16th presumably implies 7.5 - 15 million 
people infected by then.  When one gets to such a level of (presumed) 
immunity in the population, one is presumably getting into the 
territory where a reduction in "R0" pro-rata to the remaining 
non-immune population will start having a significant effect.  Does 
your model take that into account?

You, and others, might regard my suggestions as being nonsense, but 
it's the best I can currently think of that might possible make your 
forecasts look a bit less frightening!

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.
*******************************************************