Donate to Remove ads

Got a credit card? use our Credit Card & Finance Calculators

Thanks to johnstevens77,Bhoddhisatva,scotia,Anonymous,Cornytiv34, for Donating to support the site

Coronavirus - Modelling Aspects Only

The home for all non-political Coronavirus (Covid-19) discussions on The Lemon Fool
Forum rules
This is the home for all non-political Coronavirus (Covid-19) discussions on The Lemon Fool
johnhemming
Lemon Quarter
Posts: 3858
Joined: November 8th, 2016, 7:13 pm
Has thanked: 9 times
Been thanked: 609 times

Re: Coronavirus - Modelling Aspects Only

#377795

Postby johnhemming » January 16th, 2021, 2:07 pm

jfgw wrote:It appears that North East and Yorkshire NHS (Top one, mid-blue) have nabbed a bit of Cumbria,

One other obvious difference is that East of England NHS (dark blue) includes a bit of the South East (bottom right, pink) although this is unlikely to be significant as long as you consider per capita statistics.

Thanks for doing this. Obviously it means that figures will be slightly wrong, but there may still be useful information.

funduffer
Lemon Quarter
Posts: 1326
Joined: November 4th, 2016, 12:11 pm
Has thanked: 122 times
Been thanked: 831 times

Re: Coronavirus - Modelling Aspects Only

#377820

Postby funduffer » January 16th, 2021, 3:30 pm

spasmodicus wrote:Interestingly, none of the Nightingale hospitals show any occupancy, except for NHS NIGHTINGALE HOSPITAL NORTH WEST which had a maximum of 47 patients in the summer. It had none as of 6th January.
S


It is therefore a legitimate question to ask why we still have Nightingale hospitals at all? If they cannot be used now, then I suspect they will never be used again.

It seems it is easier to transport patients from one end of the country to the other than to staff and open up a local Nightingale.

Seems strange to me that no use can be found for them, having spent so much creating them back in the spring.

FD

swill453
Lemon Half
Posts: 7962
Joined: November 4th, 2016, 6:11 pm
Has thanked: 984 times
Been thanked: 3643 times

Re: Coronavirus - Modelling Aspects Only

#377872

Postby swill453 » January 16th, 2021, 5:47 pm

funduffer wrote:It is therefore a legitimate question to ask why we still have Nightingale hospitals at all? If they cannot be used now, then I suspect they will never be used again.

The NHS Louisa Jordan, which is the "Scottish equivalent" of a Nightingale hospital, is vaccinating 5000 people today https://www.bbc.co.uk/news/uk-scotland-55688152

Scott.

spasmodicus
Lemon Slice
Posts: 259
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#378049

Postby spasmodicus » January 17th, 2021, 2:28 pm

I wanted to do an update of the infections model that I have been working on and was waiting for the latest update of the ONS infections survey, due on 15th January, but now it seems this is delayed, see
https://www.ons.gov.uk/news/statementsandletters/coronaviruscovid19infectionsurveylaboratorydelays

I wonder what's going on.
Anyway, here is Model 4, the current version. The green and red markers are minimum and maximum 95% certainty estimates from the ONS random infections sampling , for al England. I have assumed that vaccination is taking place at a rate of 100,000/day, with a delay of 20 days for effectiveness to be achieved. Full vaccine effectiveness for the first dose is assumed as 80% and it is also assumed that the effect of vaccination will be reduced because people who are already wholly or partly immune are also being vaccinated. Starting population immunity is 18% and I also had to tweak the effective R value with time to make the curve fit. Current RT value is 1.6, compared with RT of about 1.2 at the model start day. This roughly fits with the idea that the virus has increased infectiousness since the beginning of November. It is difficult to estimate the effect of the everchanging lockdown rules, but this is effectively included in the changing RT.

Image

The predicted peak is 25th February, i.e. in 8 days time. This may seem a little late, as new infections are reportedly going down already in some areas. The last data point is 30th January and this is about as far as a model like this can plausibly be pushed! It will be interesting to see the new ONS estimate, for around January 6th, which will probably require some adjustment to the RT value to make the model fit.
S

johnhemming
Lemon Quarter
Posts: 3858
Joined: November 8th, 2016, 7:13 pm
Has thanked: 9 times
Been thanked: 609 times

Re: Coronavirus - Modelling Aspects Only

#378057

Postby johnhemming » January 17th, 2021, 2:43 pm

funduffer wrote:Seems strange to me that no use can be found for them, having spent so much creating them back in the spring.

There was a lot more uncertainty as to how the disease would progress in the spring. I do not criticise the government for doing this sort of contingency planning.

As it stands putting the patient interest first it is better to go to a hospital that has all the facilities than try to create a field hospital.

dealtn
Lemon Half
Posts: 6072
Joined: November 21st, 2016, 4:26 pm
Has thanked: 441 times
Been thanked: 2324 times

Re: Coronavirus - Modelling Aspects Only

#378068

Postby dealtn » January 17th, 2021, 2:59 pm

spasmodicus wrote: I have assumed that vaccination is taking place at a rate of 100,000/day,


That appears to less than half of the current vaccination rate. Further, more capacity is being added, widening the difference.

tjh290633
Lemon Half
Posts: 8206
Joined: November 4th, 2016, 11:20 am
Has thanked: 913 times
Been thanked: 4096 times

Re: Coronavirus - Modelling Aspects Only

#378085

Postby tjh290633 » January 17th, 2021, 4:13 pm

spasmodicus wrote:The predicted peak is 25th February, i.e. in 8 days time. This may seem a little late, as new infections are reportedly going down already in some areas. The last data point is 30th January and this is about as far as a model like this can plausibly be pushed! It will be interesting to see the new ONS estimate, for around January 6th, which will probably require some adjustment to the RT value to make the model fit.
S

Point of order m'lud. It is 17th January today, so 25th February is 39 days hence.

TJH

spasmodicus
Lemon Slice
Posts: 259
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#378115

Postby spasmodicus » January 17th, 2021, 5:46 pm

tjh290633 wrote:
spasmodicus wrote:The predicted peak is 25th February, i.e. in 8 days time. This may seem a little late, as new infections are reportedly going down already in some areas. The last data point is 30th January and this is about as far as a model like this can plausibly be pushed! It will be interesting to see the new ONS estimate, for around January 6th, which will probably require some adjustment to the RT value to make the model fit.
S

Point of order m'lud. It is 17th January today, so 25th February is 39 days hence.

TJH




Oops, finger trouble again, your honour. That should be 25th January. For clarification, this is the peak of current infections. New infections peak slightly earlier on 21st January.

Regarding the rate of vaccination, I used a figure which ramped up to 2.9 milion 3 days ago (reported by the Beeb I think) . The susequent rate/day makes only a small difference at this stage, because of the 20 day or so delay as the immunity builds up, the dilution by those vaccinated who are already immune. The effective number vaccinated 3 days ago was only about 730,000. Changing the daily increment since then from 100000 to 200000 makes a small difference to the peak day,which is still 25th January and the peak of new infections moves a little earlier to 20th January.

When the new ONS figure for randomly tested infections comes through, I will adjust the data to match the reported vaccinations on that day and he esimated vaccination rate,which looks as though it will be over 300,000.
S

spasmodicus
Lemon Slice
Posts: 259
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#378323

Postby spasmodicus » January 18th, 2021, 12:42 pm

Frustrated by the lack of new ONS random sampling data, I decided to take a look at the 7 day positivity for England, under
Weekly number of people receiving a PCR test and positivity
https://coronavirus.data.gov.uk/details ... me=England
as at 18/01/21 and comparing that with the ONS data which still only report up to 30th Dec. The ongoing testing is heavily biased towards suspected cases and shows a positivity rate nearly 9 times higher than the random sampling suggests (about 18% against 2% at the beginning of January.


The % positive from ongoing testing regime for England, divided by % of population testing positive according to ONS up to30th Dec. (across 9 data points) gives a ratio of 7.38 stdev 0.54 (best fit obtained by moving the ONS data back by 1 day).
The ongoing daily testing (weekly averaged) data suggest that the peak ofinfections was already reached on about 05/01/21. by scaling these testing data using the last ONS Dec 30th value
2.06% * 18.2/18.1 or 2.071% of the population are infected.
I generated new estimated data points, at 5th (2.07%) and 12th (1.7%) January, being the expected dates for the so far unforthcoming ONS survey results, I used +/-14% for the max and min values, corresponding to roughly 95% confidence interval, as for the ONS data. These points are in effect estimates of estimates. I also increased the vaccination rate to 200,000 a day.


Image

To make the model fit the two new data points, the RT value has to be reduced dramatically from 1.60 on Dec 30th to 1.23. If this really reflects the effect of the post Christmas lockdown then that’s quite encouraging.
It also shows that such simple modelling will give highly variable prognoses, unless the effect of lockdowns can be predicted in advance. The error bars on the last two data points are very wide, so that’s enough predictions for the time being!

S

spasmodicus
Lemon Slice
Posts: 259
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#379364

Postby spasmodicus » January 21st, 2021, 3:07 pm

The ONS still hasn’t published their promised survey of random infections report (University of Oxford, University of Manchester, Public Health England and Wellcome Trust) for 15th January


Meanwhile, Imperial college’s REACT survey, which I had not been paying any attention to, hit the BBC headlines today. From their website,
Covid: Infections 'must be brought down to help NHS’
Scientists tracking the spread of coronavirus in England say infection levels in the community may have risen at the start of the latest lockdown.

Infections in 6-15 January were up by 50% on early December, with one in 63 people infected, Imperial College London's initial findings suggest.

Swab tests from 143,000 people indicate 1.58% had the virus during in early January - up from 0.91% in December.

Ministers say the report does not yet reflect the impact of the lockdown.


Listeners on Radio 4 (me, for example) might be forgiven for thinking that the current falling infection rates reported by Test and Trace do not reflect the true infection situation and the lockdown is not really bringing down the infection rate. Shock! Horror! So, back to the model……

The Imperial College REACT survey provides an independent dataset from random testing, which can be compared with the ONS (Oxford University et al.) survey
https://www.imperial.ac.uk/medicine/research-and-impact/groups/react-study/real-time-assessment-of-community-transmission-findings/


As I attended both these august institutions, I have no reason to favour one dataset over the other, so I plotted up the REACT data along with ONS, as used for model 5 previously. There are only 3 points within the date range for the model. When I heard the BBC headline, I thought that my model 5 infections prediction might have been overly optimistic, given that I had to reduce the RT value by quite a lot (from 1.60 to 1.23 in January following the lockdown, equivalent to the headline R value dropping from about 1.2 to 0.9).

Image

The model is pretty consistent with the REACT infections estimates, which tend to come in a bit lower than the ONS’s, with the 95% confidence intervals just about overlapping. The blue curve (model 4) shows roughly what would have happened with no third lockdown. Recall that the last two ONS data points were guesstimated by scaling Test and Trace data and have large confidence intervals, so I was sceptical about their accuracy. The last REACT point confirms that they are more or less in the right ballpark.

I therefore disagree with the BBC’s reporting, which seems to me a bit like scaremongering, but let’s wait and see how the new ONS data come in.

S

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#379512

Postby scotia » January 22nd, 2021, 12:52 am

Another week of (English) data. To recap:- The Blue Points are the deaths by publish date, summed over the preceding week.
As I described in previous posts there has been a strong correlation between the deaths by publish date, and the hospital admissions of 13 days previous. So the Red Points are the hospital admissions, summed over a week, multiplied by 0.265, and moved forward by 13 days. And these are renamed as being the Projected deaths by publish date. The size of the vertical bars are the statistical standard deviations, assuming a Poisson distribution.

Image

It is now clear that the deaths curve has diverged significantly from the projected values based on the previous correlations with admissions. To get anything like a fit over the past two weeks, I reduced the time difference from 13 days to 11, and increased the multiplicative factor from 0.265 to 0.3. The result is displayed below

Image

One possible justification for such a change from the previous parameters could be the recent pressure on the NHS which has resulted in patients being admitted later in their illness, or not all. I have been number crunching Cases (and Cases per Tests) versus Admissions, with a variety of time slips, but I have found no convincing evidence of such a hypothesis. So that leads me to suggest (and I stress suggest) that the more infective variant could possibly be more deadly.

I feared last week that the divergence of deaths to projected deaths was growing, but the latest data suggests that no longer seems to be the case. Hopefully we are now beginning to see vaccinations resulting in a reduction of the the death rate among care home patients, which should reduce the
Deaths to Admissions ratio.

servodude
Lemon Half
Posts: 8271
Joined: November 8th, 2016, 5:56 am
Has thanked: 4434 times
Been thanked: 3564 times

Re: Coronavirus - Modelling Aspects Only

#379515

Postby servodude » January 22nd, 2021, 2:38 am

scotia wrote:One possible justification for such a change from the previous parameters could be the recent pressure on the NHS which has resulted in patients being admitted later in their illness, or not all. I have been number crunching Cases (and Cases per Tests) versus Admissions, with a variety of time slips, but I have found no convincing evidence of such a hypothesis. So that leads me to suggest (and I stress suggest) that the more infective variant could possibly be more deadly.


That would be depressing and I can see how it could be inferred from what you are seeing.
I can also see that there are a variety of other factors that might be in play and a great deal of noise! so I'm not too worried yet.

I wonder how much you would expect to see in the cross correlation of cases and admissions if there was a change in admission criteria? (or how much would the admissions need to change by before it became readily visible?)

If I do a very rough and ready pull of the english data for cases, admissions and deaths (from https://coronavirus.data.gov.uk/)
- a 7 day rolling smooth
- and scaling to bring in to the same range

Image
I think i can see more space between the cases and admissions in late december
- which might suggest that a lower proportion are being admitted?

I think winter might also have a compounding effect on the deaths
- sd

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#379581

Postby scotia » January 22nd, 2021, 11:26 am

servodude wrote:That would be depressing and I can see how it could be inferred from what you are seeing.
I can also see that there are a variety of other factors that might be in play and a great deal of noise! so I'm not too worried yet.

Thanks for the plots - In my number crunching I just had a look at the raw cases figures and decided I could not extract any positive conclusions.
And yet another surprising piece of news this morning which further muddies the water - it is estimated that only (approx) 16% of those with symptoms get a test and self-isolate if positive. The belief is that the remainder can't risk a positive test and the economic cost of self isolating. If this remainder has been growing, then it could lead to later admissions, and a poorer outcome. Just another unknown factor. I'm pleased to see that it looks likely that the government will introduce a one-off payment of £500 to persons who test positive - so hopefully the remainder will start reducing.
And, slightly off topic - some news from Scotland (which I suspect you will be personally interested in) - where we are handling the vaccination program differently from England. The first priority has been Care Home residents and front line heath and care workers. Since taking the vaccine to a number of small geographically separated Care Homes is less "efficient" in numbers vaccinated, compared to a centralised scheme for ambulant oldies, its not surprising that the percentage of (all) persons vaccinated has lagged behind England, however the opposite is true for Care Home residents - and I think that was probably a good idea, although it does mean that we (in our mid seventies) have not yet received any notification of a vaccination.

spasmodicus
Lemon Slice
Posts: 259
Joined: November 6th, 2016, 9:35 am
Has thanked: 65 times
Been thanked: 117 times

Re: Coronavirus - Modelling Aspects Only

#379585

Postby spasmodicus » January 22nd, 2021, 11:36 am

servodude wrote:
scotia wrote:One possible justification for such a change from the previous parameters could be the recent pressure on the NHS which has resulted in patients being admitted later in their illness, or not all. I have been number crunching Cases (and Cases per Tests) versus Admissions, with a variety of time slips, but I have found no convincing evidence of such a hypothesis. So that leads me to suggest (and I stress suggest) that the more infective variant could possibly be more deadly.


That would be depressing and I can see how it could be inferred from what you are seeing.
I can also see that there are a variety of other factors that might be in play and a great deal of noise! so I'm not too worried yet.

I wonder how much you would expect to see in the cross correlation of cases and admissions if there was a change in admission criteria? (or how much would the admissions need to change by before it became readily visible?)

If I do a very rough and ready pull of the english data for cases, admissions and deaths (from https://coronavirus.data.gov.uk/)
- a 7 day rolling smooth
- and scaling to bring in to the same range


I think i can see more space between the cases and admissions in late december
- which might suggest that a lower proportion are being admitted?

I think winter might also have a compounding effect on the deaths
- sd


scotia wrote:It is now clear that the deaths curve has diverged significantly from the projected values based on the previous correlations with admissions. To get anything like a fit over the past two weeks, I reduced the time difference from 13 days to 11, and increased the multiplicative factor from 0.265 to 0.3. The result is displayed below

One possible justification for such a change from the previous parameters could be the recent pressure on the NHS which has resulted in patients being admitted later in their illness, or not all. I have been number crunching Cases (and Cases per Tests) versus Admissions, with a variety of time slips, but I have found no convincing evidence of such a hypothesis. So that leads me to suggest (and I stress suggest) that the more infective variant could possibly be more deadly.

I feared last week that the divergence of deaths to projected deaths was growing, but the latest data suggests that no longer seems to be the case. Hopefully we are now beginning to see vaccinations resulting in a reduction of the the death rate among care home patients, which should reduce the
Deaths to Admissions ratio.



I hope that Scotia is about the vaccinations starting to have an effect, although it's fairly early days yet.

On the subject of correlation and time slip, the peak in the death rate, which I hope will come soon, should enable a better estimate of the time shift. You may then get a better estimate of the Deaths to Admissions ratio.

I saw an article somewhere which said that there is evidence that ICUs at near 100% capacity sadly show a significantly lower survival rate.

Regarding cases vs admissions, I have so far failed to find a satisfactory correlation as these data are quite noisy.

regards,
S

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#379669

Postby scotia » January 22nd, 2021, 3:05 pm

scotia wrote:And yet another surprising piece of news this morning which further muddies the water - it is estimated that only (approx) 16% of those with symptoms get a test and self-isolate if positive. The belief is that the remainder can't risk a positive test and the economic cost of self isolating. If this remainder has been growing, then it could lead to later admissions, and a poorer outcome. Just another unknown factor. I'm pleased to see that it looks likely that the government will introduce a one-off payment of £500 to persons who test positive - so hopefully the remainder will start reducing.

At my age I should know better than report mere conjecture from early morning media stories. By late morning It now appears that the £500 was simply suggested in a government report, and "government sources" are indicating that it never got as far as the PM. Ignore!
And my figure of only (approx) 16% of those with symptoms self isolating is now appearing as only 18% of those with symptoms self isolating for the full 10 days.

tjh290633
Lemon Half
Posts: 8206
Joined: November 4th, 2016, 11:20 am
Has thanked: 913 times
Been thanked: 4096 times

Re: Coronavirus - Modelling Aspects Only

#379732

Postby tjh290633 » January 22nd, 2021, 5:17 pm

The PM has just announced that the latest mutation has a higher mortality rate than the original. That would explain the divergence seen in your post above, viewtopic.php?p=379512#p379512

TJH

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#379868

Postby scotia » January 23rd, 2021, 12:51 am

tjh290633 wrote:The PM has just announced that the latest mutation has a higher mortality rate than the original. That would explain the divergence seen in your post above, viewtopic.php?p=379512#p379512

TJH

I feared as much. But I would have preferred to be wrong.
I'll try to carry out a rough calculation of the size of the effect. It looks like the new variant makes up somewhere between 50% and 70% of the current infections. It also appears that my deaths to admissions ratio has now increased from 0.265 to 0.3 for the most recent data, but if I go back a week, the 0.3 looks to be an underestimate. However if I stick with the 0.3, and go for a 50% infection mix, then the appropriate ratio for the new variant is 0.6 - 0.265 = 0.335. And the fractional difference from the original ratio is (0.335 - 0.265)/0.265 - which is 26%. I could bump this up and down with other assumptions, so I stress that it is a ballpark figure.
On looking at the government reports on the BBC site, the chief scientific adviser has stressed that there is a a lot of uncertainty, but a figure of around 30% is being "hinted" at.
Lets hope this all becomes academic as the vaccinations get to work and the death rate falls significantly.

servodude
Lemon Half
Posts: 8271
Joined: November 8th, 2016, 5:56 am
Has thanked: 4434 times
Been thanked: 3564 times

Re: Coronavirus - Modelling Aspects Only

#379870

Postby servodude » January 23rd, 2021, 1:17 am

scotia wrote:
tjh290633 wrote:The PM has just announced that the latest mutation has a higher mortality rate than the original. That would explain the divergence seen in your post above, viewtopic.php?p=379512#p379512

TJH

I feared as much. But I would have preferred to be wrong.
I'll try to carry out a rough calculation of the size of the effect. It looks like the new variant makes up somewhere between 50% and 70% of the current infections. It also appears that my deaths to admissions ratio has now increased from 0.265 to 0.3 for the most recent data, but if I go back a week, the 0.3 looks to be an underestimate. However if I stick with the 0.3, and go for a 50% infection mix, then the appropriate ratio for the new variant is 0.6 - 0.265 = 0.335. And the fractional difference from the original ratio is (0.335 - 0.265)/0.265 - which is 26%. I could bump this up and down with other assumptions, so I stress that it is a ballpark figure.
On looking at the government reports on the BBC site, the chief scientific adviser has stressed that there is a a lot of uncertainty, but a figure of around 30% is being "hinted" at.
Lets hope this all becomes academic as the vaccinations get to work and the death rate falls significantly.


Indeed. I was reading the report at https://www.bbc.com/news/health-55768627
- trying to understand how they've worked it out

It contains this bit:
This difference is found when looking at everyone testing positive for Covid, but analysing only hospital data has found no increase in the death rate.

- which implies that if it's no different for the cohort getting treated

-sd

Julian
Lemon Quarter
Posts: 1385
Joined: November 4th, 2016, 9:58 am
Has thanked: 532 times
Been thanked: 676 times

Re: Coronavirus - Modelling Aspects Only

#379902

Postby Julian » January 23rd, 2021, 10:46 am

servodude wrote:
scotia wrote:
tjh290633 wrote:The PM has just announced that the latest mutation has a higher mortality rate than the original. That would explain the divergence seen in your post above, viewtopic.php?p=379512#p379512

TJH

I feared as much. But I would have preferred to be wrong.
I'll try to carry out a rough calculation of the size of the effect. It looks like the new variant makes up somewhere between 50% and 70% of the current infections. It also appears that my deaths to admissions ratio has now increased from 0.265 to 0.3 for the most recent data, but if I go back a week, the 0.3 looks to be an underestimate. However if I stick with the 0.3, and go for a 50% infection mix, then the appropriate ratio for the new variant is 0.6 - 0.265 = 0.335. And the fractional difference from the original ratio is (0.335 - 0.265)/0.265 - which is 26%. I could bump this up and down with other assumptions, so I stress that it is a ballpark figure.
On looking at the government reports on the BBC site, the chief scientific adviser has stressed that there is a a lot of uncertainty, but a figure of around 30% is being "hinted" at.
Lets hope this all becomes academic as the vaccinations get to work and the death rate falls significantly.


Indeed. I was reading the report at https://www.bbc.com/news/health-55768627
- trying to understand how they've worked it out

It contains this bit:
This difference is found when looking at everyone testing positive for Covid, but analysing only hospital data has found no increase in the death rate.

- which implies that if it's no different for the cohort getting treated

-sd

I've been trying to work it out too. I saw an interview on Newsnight last night with Prof John Edmunds (https://en.wikipedia.org/wiki/John_Edmu ... emiologist)), a member of NERVTAG and one of the people behind the analysis. He was explaining a bit about the caveats behind the preliminary findings although he, maybe not unsurprisingly since he was one of the people involved in the analysis, sounded reasonably confident in his conclusions although did accept that they needed to be validated against higher death numbers.

Apparently the issue with the data is the small number of deaths analysed. That is because they started with the very large pillar 2 (testing in the community) database to pick up the strain with which a patient was infected and then tied that to subsequent deaths in hospital(*). The problem was that a large percentage of people dying in hospital have never been tested at pillar 2, they have instead had their first test and diagnosis in hospital so would not be in the "input" data for this analysis. Essentially, if I understood what he was saying correctly, their data relied on the intersection of people first tested in the community who subsequently went on to die in hospital and that intersection is smaller than they would like (speaking purely about the data analysed rather than wanting more people to die obviously!).

(*) It wasn't clear to me if they explicitly only looked at hospital-reported deaths or if JE was meaning that effectively the deaths they were looking at were all (or almost all) hospital deaths because it is extremely rare for someone to test positive in the community and then die at home because at some point in their decline they will end up in hospital.

Now, I could have got the wrong end of the stick about what JE was saying in that Newsnight interview but maybe it's at least a starting point for people like you Servodude trying to understand how they've worked it out to see if other explanations you come across seem to confirm or refute my understanding.

- Julian

scotia
Lemon Quarter
Posts: 3561
Joined: November 4th, 2016, 8:43 pm
Has thanked: 2371 times
Been thanked: 1943 times

Re: Coronavirus - Modelling Aspects Only

#379906

Postby scotia » January 23rd, 2021, 10:58 am

servodude wrote:It contains this bit:
This difference is found when looking at everyone testing positive for Covid, but analysing only hospital data has found no increase in the death rate.

- which implies that if it's no different for the cohort getting treated

-sd

Yes - it appears that there are far too many factors in the quality and availability of the data to allow any certainty as to whether or not there is a real increased death rate with the variant. The strongest I would put my own observations is that I suggest that there could be - and if the effect I am seeing is due to an increased death rate of the variant, it looks like its in agreement with the government's tentative figure of a 30% increase .
But getting back to more encouraging matters, in Scotland more than 91% of Care Home residents have been vaccinated, and that should significantly reduce the deaths, while having a smaller effect on the hospital admissions, so I'm hoping this will also be reflected in the English statistics over the coming weeks. I look forward to seeing the actual deaths on my plots dropping below the historically projected ones!


Return to “Coronavirus Discussions”

Who is online

Users browsing this forum: No registered users and 7 guests