Donate SIGN UP

Covid-19 Modelling

Avatar Image
jim360 | 15:53 Tue 05th May 2020 | Science
20 Answers
For the two of you who may care, my ongoing personal quest to understand the latest UK Covid figures and make predictions about the future continues. Things are naturally complicated by the fact that data are continually being updated even weeks after the fact.

My most recent focus was on understanding the figures in the NHS England updates on hospital deaths, as these are at least reliably fixed by actual date of death (see link below). Anyone who wants to analyse the data themselves can do so by using the free model-fitting software fityk, which I also only recently discovered.

Anyway, the latest fit I'm using seems to lead to just shy of 30,000 hospital deaths in England once the epidemic has ended (approximate range 28,000-31,000). Which, considering that the current equivalent total is 21,750, but that the UK-wide total is already in excess of 30,000, suggests that there's a long way to go yet. I've found it sad how consistent the model has been over the last week or so.

But for those who want to distract themselves from the rising death toll by focusing on the mathematical aspects, links below will help them get into it. I should add that my own modelling is based currently on a Split (pseudo-)Voigt distribution, and I haven't yet decided how to interpret this, ie is this merely a mathematical convenience or are there some meaningful conclusions to draw from the goodness of this particular fit as opposed to other models?


1 to 20 of 20rss feed

Best Answer

No best answer has yet been selected by jim360. Once a best answer has been selected, it will be shown here.

For more on marking an answer as the "Best Answer", please visit our FAQ.
Question Author
Please note also that the headlines in various papers about how the UK's toll is now the "highest in Europe"* are misleading: the official death toll is now higher in the UK than anywhere else, but it is almost certainly true that Spain and Italy have under-reported their own deaths. It may be that at the end of it the excess mortality in the UK *is* higher than elsewhere but it will be some time before this becomes clear. What is certain is that this has been brutal. Today saw the ONS figures indicate another week with deaths running at more than double the background rate, and it is likely that the final week of April was also in more or less the same ballpark.
Am I right in thinking that any person who had Covid-19 at their time of death has that listed as their cause of death even if it was not responsible for their death?

You know much more about the figures than me but unless all countries are reporting things in the same way the figures may be misleading.

Question Author
Yes, there is loads still to untangle, and it follows of course that nothing I say about this will be even close to definitive. The picture is roughly this:

1. Spain reported only deaths amongst positive tests, but until recently only those with serious conditions were tested.
2. Until recently, the UK reported only hospital deaths but has now added care homes and a couple of other trackable settings.
3. Belgium reports all deaths where Covid-19 is suspected, which is presumably the main reason the apparent death rate there is the highest in Europe*.
4. Italy reports all deaths amongst positive cases.

Etc. It will probably be months if not years for everything to be untangled.

*excluding San Marino and Andorra.
wolf, I think that answer to that is "maybe" Causes of death on death certificates aren't an exact science and aren't limited to one cause. With the best will in the world, the cause of death can be an opinion, especially where the deceased had multiple co-morbidities.
Question Author
I realise that I didn't fully answer the specific question:

// Am I right in thinking that any person who had Covid-19 at their time of death has that listed as their cause of death even if it was not responsible for their death? //

I think it is listed as *a* cause of death. But what is pretty unambiguous is that something is causing death rates to rocket in multiple countries, and even if Covid-19 itself is not singly responsible (most of the dead had other underlying conditions), then it certainly is helping to shorten the lives of those who were already vulnerable.
I don't suppose that they will do autopsies on all the deceased. Woof, as you say many of the people has multiple serious health problems - it is not always easy to work out which one killed the person.

I'm off to do battle with my flat-mate. He is the boy in my avatar.
PS I think that as Jim has pointed out, different countries have got different "cause of death" "rules" anyway and layered onto this are different kinds of reporting. I know its horrible and smacks of "oh well they would have died anyway" but the important figure is what is being called "excess deaths" because that will make clear how many more people have died because of the pandemic.
//..but it is almost certainly true that Spain and Italy have under-reported their own deaths.//

And it seems fairly obvious to me that the UK is over-reporting ours. I know two people who have died recently. One was a mate's wife. She had a multitude of health problems including pneumonia and pleurisy and was told shortly after Christmas that she would probably not see the early summer. Her death certificate listed those ailments together with heart failure brought on by Covid. She had not been tested for Covid and nobody in the ward she was in had Covid symptoms or was suspected of having it. The other was a mate's mother. She was 95, had early onset dementia and a multitude of health problems (as many people of that age have). Her death certificate states "Dementia associated with Covid-19" as the cause of her death. She had no signs of Covid and nor did any residents or staff at the care home where she lived.

That's why I'm afraid that unless your model sets out to forecast deaths where Covid is mentioned on the death certificate (rather than people who actually died either with or from it) it may not help. I maintain my stance that nobody - from you up to Prof Whitty - has any idea how many people have had this disease, how many they've passed it on to, how many still have it and how many people are yet to contract it.
Question Author
Like I said, I was modelling NHS England's data. The deficiencies in that data are their own.

Still, as a matter of fact, you cannot really look at the data on weekly reported deaths in England and Wales and conclude that Covid-19 isn't having a major impact. People don't die in the UK in these volumes unless there is (a) a war on, or (b) a new disease to which there is no natural immunity.

I agree of course that the data will always be difficult to interpret -- how much is a death down to Covid-19 and how much is because the patients were sick anyway? -- but what is not open to interpretation is that there has been a clear spike, that something new is contributing to that spike, and that Covid-19 is something new.

I'm not sure, however, that you can conclude from your two cases that deaths are being over-reported. It's likely to go both ways, at the very least: some people are recorded as having had Covid (or suspected Covid) when it may have been at most a coincidence, others will have died of Covid without it being recognised; others still may have died before their time not because they had Covid-19 but because for one reason or another they didn't get the support they needed for another health risk.

What Whitty et al can say for certain is that the number of deaths because of or related to Covid-19 is greater than 0 and less than 120,000, and probably accounts for something approaching 30% +/-5% of deaths in the four weeks ending April 24th. There will always be an uncertainty because death and life are too complicated to be otherwise, but it remains a mistake to turn that uncertainty into "no idea", or (educated) guesswork.
NJ, that is why the excess deaths figure is so important.
On comparisons with other countries and why the UK has now overtaken Italy in the number of deaths, this is interesting. Many factors to take into consideration.

Also, is there readily available ONS data on excess deaths and comparisons from the last couple of years?
Cheers Jim.
Many thanks Jim
I will have a look - not done it yet - does the slope of the decrease give you an Ro and is it useful ? - that is if the curve goes clunk, is the Ro nearer zero and are we winning
and if the curve seems to cling to the asymptote y=1, then does that show we are really achieving diddly squat?

getting a biological signal and modelling it accurately - is an admirable achievement - well done

then saying - yeah OK does it correspond to anything biologic ? is more fraught. I suppose in Physics you can say
" there is a gap there, and I think there is something there - which I will call epsilon-minus - and very soon someone will find it!"
Mendeleev set the standard in 1867 - but in biology it is much more fraught

Oh Boris in Hospital - he will have had a reduced circulation ( first space ) and the cells would be a bit swollen in the base of his lung ( second space ) and there will be a huge gap where the fluid has gone somewhere - (third space loss) which was causing his lung problems. But the Third space is only physiological and you cant see under a microscope.

currently the hospitals are running the patients 'dry' so that these losses dont er drown the patient. But as you can imagine a dry patient has his own problems. -
( when I was in hospital I had a third space loss of - 10 li - yeah I was 10kg heavy than I should have been. oo-er mrs!)

many thanks for your efforts
\\ I maintain my stance that nobody - from you up to Prof Whitty - has any idea ....... how many people are yet to contract it.//

yes we do actually - more than zero and less than 30 000.

Ignore him Jim - he does not like and never has liked models and those that churn them out - and if ever you need to know why the govt decision making is a disaster, read some of his other posts. Classics grads (Boris) and lawyers ( Starmer) and media grads ( the whole of the Beeb) can handle a major epidemics MUCH better than doctors or MOHs.

One thing about interpretation is that Merkel is a scientist ( nuclear chemist actually ) and may have paid greater attention to her scientific advisers - or not we dont know ....

I think it is very unlikely that the excess deaths are accounted for by doctors liking to write "Covid 19" when it isnt
and so you are left with - "there is a lot of death around and we have NO idea why" -

causes of death from death certs are well known to be - - erm .....
however - you know - the graveyards are full and there is something killing the old people
hi Jim
have you given a ref to your own effort >
I cant quite see it -thx
Question Author
I can upload my data sheet but since it's just a case of taking a five-day rolling average, putting it into the fityk software, and selecting the Split Pseudo Voigt option, I wasn't sure it was worth it. I can do, I suppose, and it might even help to show how the projections are changing over time.
well breaking news
neil Ferguson - the Ferge - has left SAGE

blood on the carpet there boys - Prof has a knife out for him - and Neil got it badly wrong on BSE - knives are out because there werent a 100 000 deaths from vCJD. and Neil said there would be

as a result of us butchering all the cows in the country as far as I am concerned but others say it was because he was a Bad Modeller. - and it was never gonna be like that

his models are OK
but he was caught out on a social distancing gaff

cd still give advice foo a window I suppose
or thro a member who remains in Sage

like Trump - so what if he stuffs his hand down a ladys doo dah - can he run a country ?
actually I say the answer is no - but you know
horses for courses ..some say yes
three ( a lot ) doctors die from window falls in Russia whilst er treating covid patients .....

they wanted to get away from it all but taking the nearest way out .....

1 to 20 of 20rss feed

Do you know the answer?

Covid-19 Modelling

Answer Question >>

Related Questions

Sorry, we can't find any related questions. Try using the search bar at the top of the page to search for some keywords, or choose a topic and submit your own question.