International Statistics - UK's record on testing

[DerekC]

The following table shows the now well-known fact that the UK is second only to Belgium in its Covid death rate per head of population. It also shows some interesting facts associated with that, and I am racking my brains to work out what's going on.



The UK also has the highest number of tests per head of population and yet the number of cases per head is only average. The "Tests per case" figure shows the issue starkly. We are carrying out about twice as many tests per case found as comparable countries. And our number of cases found per death is very low - so either our treatment of the cases found is very poor, so more of them die (which I find hard to believe) or - as I would prefer to conclude - our test regime is very ineffective.

The WHO suggests that the average death rate from Covid-19 is about 3%, which would suggest that the UK's total death toll of 45,000 is likely to have been derived from about 1,500,000 cases. If so then we have only found 20% of the cases, which would explain the numbers above.

I can see that we missed a lot of cases when government shut down test and trace from early March to some time in May, but are we still missing them? I looked at the UK government figures for June, which show 36,332 new cases found and 4,537 deaths - a ratio of almost exactly 8. So it hasn't got much better - we are still missing the vast majority of cases.

If this is right then the "Track and Trace" system seems to be totally ineffective. Please somebody tell me I am wrong.

 

[Deerfold]

We are one of the biggest testers. Now. But we took ages to get up to the level of the average European country.

 

[yorkie]

The WHO suggests that the average death rate from Covid-19 is about 3%, ...

Surely they aren't still saying that? Latest estimates are around 0.5 to 0.8%.

I can see that we missed a lot of cases when government shut down test and trace from early March to some time in May, but are we still missing them?

There must be millions of undiagnosed cases early on and there is now no way to be sure if those people had this virus or not; if there are cases being undiagnosed now (which there will be, given many people do not suffer symptoms) it won't be on anywhere near that scale

I looked at the UK government figures for June, which show 36,332 new cases found and 4,537 deaths - a ratio of almost exactly 8. So it hasn't got much better - we are still missing the vast majority of cases.

Given that a huge number of cases will be asymptomatic, it is unavoidable that many cases will never be found, but do bear in mind that people who died in June are unlikely to have caught the virus in June; therefore your calculation is flawed.

Also, what we really want to know isn’t the case fatality rate: it’s the infection fatality rate.

If this is right then the "Track and Trace" system seems to be totally ineffective. Please somebody tell me I am wrong.

It can never be totally effective; how do you define totally ineffective?

 

[jtuk]

He's probably just dividing positive tests by "deaths with covid" ignoring there's going to be millions, if not tens of millions of people that will have had it asymptomatically

 

[Bantamzen]

The following table shows the now well-known fact that the UK is second only to Belgium in its Covid death rate per head of population. It also shows some interesting facts associated with that, and I am racking my brains to work out what's going on.

View attachment 81451

The UK also has the highest number of tests per head of population and yet the number of cases per head is only average. The "Tests per case" figure shows the issue starkly. We are carrying out about twice as many tests per case found as comparable countries. And our number of cases found per death is very low - so either our treatment of the cases found is very poor, so more of them die (which I find hard to believe) or - as I would prefer to conclude - our test regime is very ineffective.

The WHO suggests that the average death rate from Covid-19 is about 3%, which would suggest that the UK's total death toll of 45,000 is likely to have been derived from about 1,500,000 cases. If so then we have only found 20% of the cases, which would explain the numbers above.

I can see that we missed a lot of cases when government shut down test and trace from early March to some time in May, but are we still missing them? I looked at the UK government figures for June, which show 36,332 new cases found and 4,537 deaths - a ratio of almost exactly 8. So it hasn't got much better - we are still missing the vast majority of cases.

If this is right then the "Track and Trace" system seems to be totally ineffective. Please somebody tell me I am wrong.

The single biggest problem with simple data comparisons like this is that different countries are recording deaths in different ways. And so comparing death rates is near impossible without accounting for exactly how each country is doing so.

 

[yorkie]

The single biggest problem with simple data comparisons like this is that different countries are recording deaths in different ways. And so comparing death rates is near impossible without accounting for exactly how each country is doing so.

True; we count people as dying with Covid19 even if their positive test was many weeks before, and not the true cause of, death.

Another factor that will skew the death rate is that the chances of dying vary massively by age; among over 80s I believe it is as high as a 10% chance of death, yet for under 40s it is a tiny fraction of a percent.

So if you get more outbreaks in care homes, this pushes the overall death rate up. But that doesn't mean a fit & healthy person is any more likely to die from it.

 

[Bletchleyite]

True; we count people as dying with Covid19 even if their positive test was many weeks before, and not the true cause of, death.

We really could do with sorting that out now we know the mechanisms by which COVID kills in more detail. It's silly including the likes of car crash victims and makes the situation look far worse than it is.

 

[Darandio]

We really could do with sorting that out now we know the mechanisms by which COVID kills in more detail. It's silly including the likes of car crash victims and makes the situation look far worse than it is.

I really cannot see them sorting that now. It allows them to continue the scare tactics that filter down through the media. Anyone thinking that originally could have been accused of being cynical but not now, it's blatantly obvious.

 

[Domh245]

Noticeably, the worldmeters data uses the government figures (PHE & equivalents) and is therefore under-representing the situation. The ONS number is 50,800 to the 17th of July, most of this discrepancy coming from the peak in April when PHE were under reporting and registering a death as COVID didn't require a test to be carried out at all. Given the relatively high (but absolutely small) variance currently between ONS and PHE figures now the situation is reversed (discussed here), this 5000 death gap should close up.

 

[Bantamzen]

True; we count people as dying with Covid19 even if their positive test was many weeks before, and not the true cause of, death.

Another factor that will skew the death rate is that the chances of dying vary massively by age; among over 80s I believe it is as high as a 10% chance of death, yet for under 40s it is a tiny fraction of a percent.

So if you get more outbreaks in care homes, this pushes the overall death rate up. But that doesn't mean a fit & healthy person is any more likely to die from it.

Indeed, its why looking more closely at the data is so important, and not just cherry picking the headline figures.

We really could do with sorting that out now we know the mechanisms by which COVID kills in more detail. It's silly including the likes of car crash victims and makes the situation look far worse than it is.

Its just been mentioned above but whilst the reporting mechanisms can & should be reworked & tightened to ensure that covid mortality rates actually reflect cases where covid is a major factor & not just present. But it will be almost impossible to retrofit to the existing data which will mean the data will always be skewed.

 

[DerekC]

Lots of interesting comments already, but I am not feeling any more confident. Let's address a few of them:

> The fatality rate is probably lower than 3% - agreed - that means there are/were even more undetected cases.
> The number of Uk fatalities is 50,000, not 45,000 - probably right, but doesn't make much difference to the issue
> There are millions of asymptomatic cases out there - the best estimates seem to be that between 10% and 25% of cases are asymptomatic.
> The historic data is inaccurate/skewed - true, but it's all we have to go on and I doubt the UK figures are out by orders of magnitude
> Don't just look at headline numbers - I am not - that's the point of this thread
> Different countries prepare their statistics in different ways and there is no point in comparing them - that's the Johnson argument, but I think you have to do your best to see through the fog. The UK is an outlier in various areas and it seems right to ask why.

@yorkie asked how I was defining "totally ineffective" for the track and trace system. Presumably the it has to work well enough to keep R0 below 1 in a situation where lockdown has been largely removed and we are relying on social distancing and masks. That means getting to a large proportion of the contacts of people who test positive before they reach the onward transmission stage. If we are only picking up 1 in 5 or less of the actual infections and (according to the government's own figures) only getting responses from about half of the contacts, then track and trace isn't going to have much impact on the overall R0. If that's true then it is obviously ineffective (maybe "totally ineffective" was a bit harsh).

 

[Bantamzen]

Lots of interesting comments already, but I am not feeling any more confident. Let's address a few of them:

> The fatality rate is probably lower than 3% - agreed - that means there are/were even more undetected cases.
> The number of Uk fatalities is 50,000, not 45,000 - probably right, but doesn't make much difference to the issue
> There are millions of asymptomatic cases out there - the best estimates seem to be that between 10% and 25% of cases are asymptomatic.
> The historic data is inaccurate/skewed - true, but it's all we have to go on and I doubt the UK figures are out by orders of magnitude
> Don't just look at headline numbers - I am not - that's the point of this thread
> Different countries prepare their statistics in different ways and there is no point in comparing them - that's the Johnson argument, but I think you have to do your best to see through the fog. The UK is an outlier in various areas and it seems right to ask why.

@yorkie asked how I was defining "totally ineffective" for the track and trace system. Presumably the it has to work well enough to keep R0 below 1 in a situation where lockdown has been largely removed and we are relying on social distancing and masks. That means getting to a large proportion of the contacts of people who test positive before they reach the onward transmission stage. If we are only picking up 1 in 5 or less of the actual infections and (according to the government's own figures) only getting responses from about half of the contacts, then track and trace isn't going to have much impact on the overall R0. If that's true then it is obviously ineffective (maybe "totally ineffective" was a bit harsh).

Picking up on the differing reporting methods, you really can't peer through the fog without knowing which way to look. For example in some countries only hospital deaths are recorded, whereas in this all deaths in all settings are recorded. And there's the issue in this country where we record all deaths where the patient has tested positive, even if that might have months ago and not necessarily a factor in the causes of death. So really you need to research all the recording methods before even attempting to line the dataset ups, otherwise you leave way too much noise in.

 

[Domh245]

What you have to remember is that now we have the capacity, we can test far more than we could back in April (peak virus). This means that we will return a lot more negatives (as we are testing mild symptoms that wouldn't have been tested previously), and that we are also able to pick up far more asymptomatic cases as a result of either test & trace of just persistent door to door testing (case in point being Leicester - they ramped up testing and the number of cases followed). The flip side is that during April when we abandoned testing and focused only on the most severe cases, we had a high death count and low testing rate, artificially deflating the cases/death number. Had we tested at the sort of capacity we had now, we'd have seen a far higher number of cases back in April, and consequently a far higher number of cases/death.

A low 'tests/case' isn't indicative of better testing, it's indicative of more focused testing, either because of limited testing resources or very high levels of COVID (or both!)

 

[AdamWW]

Picking up on the differing reporting methods, you really can't peer through the fog without knowing which way to look. For example in some countries only hospital deaths are recorded, whereas in this all deaths in all settings are recorded. And there's the issue in this country where we record all deaths where the patient has tested positive, even if that might have months ago and not necessarily a factor in the causes of death. So really you need to research all the recording methods before even attempting to line the dataset ups, otherwise you leave way too much noise in.

That's true of the PHE daily figures, but I didn't think the ONS figures did that.

@yorkie asked how I was defining "totally ineffective" for the track and trace system. Presumably the it has to work well enough to keep R0 below 1 in a situation where lockdown has been largely removed and we are relying on social distancing and masks. That means getting to a large proportion of the contacts of people who test positive before they reach the onward transmission stage. If we are only picking up 1 in 5 or less of the actual infections and (according to the government's own figures) only getting responses from about half of the contacts, then track and trace isn't going to have much impact on the overall R0. If that's true then it is obviously ineffective (maybe "totally ineffective" was a bit harsh).

I suppose that depends if we are aiming at a situation where we rely on track and trace as the only mitigation. That would be ideal, but maybe we'll have to live with local restrictions being imposed too in which case track and trace doesn't have to do all the work.

Also, if immunity lasts long enough, we will see R starting to drop anyway.

And a partially effective vaccine also reduces the work track and trace has to do.

Pedantic note - I think R0 refers to transmission without any mitigation (though clearly it's not a fixed figure beacuse even without restrictions, human interactions vary across the planet and with time), and when we start taking precautions we then call it R.

 

[DerekC]

What you have to remember is that now we have the capacity, we can test far more than we could back in April (peak virus). This means that we will return a lot more negatives (as we are testing mild symptoms that wouldn't have been tested previously), and that we are also able to pick up far more asymptomatic cases as a result of either test & trace of just persistent door to door testing (case in point being Leicester - they ramped up testing and the number of cases followed). The flip side is that during April when we abandoned testing and focused only on the most severe cases, we had a high death count and low testing rate, artificially deflating the cases/death number. Had we tested at the sort of capacity we had now, we'd have seen a far higher number of cases back in April, and consequently a far higher number of cases/death.

A low 'tests/case' isn't indicative of better testing, it's indicative of more focused testing, either because of limited testing resources or very high levels of COVID (or both!)

I agree with your first paragraph, which is why I was looking for an increase in the crude number of cases per death as time goes by, but that doesn't seem to be happening. I know it's hard to compare cases found with deaths because you need to know that you are looking at the same sample of patients - so as @yorkie pointed out, you should probably compare June's figures for cases with July's for deaths. However both cases found and deaths seem to have been running roughly level for some weeks, so the error should be small. Which makes the worry - maybe in the UK's case the high "tests/case" combined with low "cases/death" is the result of the push for high numbers of tests at the expense of any focus on who and where to test. Germany, for example, has a "tests/case" of 36 and "cases/death" of 23, which looks as though they are much better focused.

 

[Bantamzen]

That's true of the PHE daily figures, but I didn't think the ONS figures did that.

Well to a degree they do, taken from their website:

Deaths registered weekly in England and Wales, provisional - Office for National Statistics

Provisional counts of the number of deaths registered in England and Wales, including deaths involving the coronavirus pandemic, by age, sex and region.

www.ons.gov.uk

More quality and methodology information on strengths, limitations, appropriate uses, and how the data were created is available in the Mortality statistics in England and Wales QMI.

To meet user needs, we publish very timely but provisional counts of death registrations in England and Wales in our Deaths registered weekly in England and Wales, provisional dataset. These are presented by sex, age group and regions (within England) as well as for Wales as a whole. To allow time for registration and processing, these figures are published 11 days after the week ends. Because of the rapidly changing situation, in this bulletin we have also given provisional updated totals based on the latest available death registrations, up to 25 July 2020.

Because of the coronavirus (COVID-19) pandemic, our regular weekly deaths release now provides a separate breakdown of the numbers of deaths involving COVID-19: that is, where COVID-19 or suspected COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions. If a death certificate mentions COVID-19 it will not always be the main cause of death but may be a contributory factor. This new bulletin summarises the latest weekly information and will be updated each week during the pandemic.

So by their own admission there will still be cases they record as covid-releated when covid will not have been the primary or even partial cause of death. And as covid is a notifiable disease it is entirely possible that it could make it onto the death certificate when in fact it was not a contributory factor. Indeed I'm sure there have been some cases make the media where relatives of the deceased have complained about this, and for sure I know of one person whose family member died of heart disease yet had covid recorded, & are currently challenging this.

So you see its really very hard to untangle the mess of data just in this country, let alone in all the others then line them all up.

 

[AdamWW]

So by their own admission there will still be cases they record as covid-releated when covid will not have been the primary or even partial cause of death. And as covid is a notifiable disease it is entirely possible that it could make it onto the death certificate when in fact it was not a contributory factor. Indeed I'm sure there have been some cases make the media where relatives of the deceased have complained about this, and for sure I know of one person whose family member died of heart disease yet had covid recorded, & are currently challenging this.

So you see its really very hard to untangle the mess of data just in this country, let alone in all the others then line them all up.

OK but not the run over three months after a positive test variety though, so it's an improvement.

But this is why I'm in the camp that says excess deaths are the best figure to use - this also includes all the indirectly caused deaths, i.e. due to lack of treatment/diagnosis that would also have taken place.

I'm sure there are subtleties even there but I think it has to be the best method we have of comparing different countries.

 

[Bantamzen]

OK but not the run over three months after a positive test variety though, so it's an improvement.

But this is why I'm in the camp that says excess deaths are the best figure to use - this also includes all the indirectly caused deaths, i.e. due to lack of treatment/diagnosis that would also have taken place.

I'm sure there are subtleties even there but I think it has to be the best method we have of comparing different countries.

The problem, or at least one, is that over the period of the pandemic many other treatments have been scaled back, postponed or even cancelled. And as the months go on, some patients that may previously have had successful treatments will start to contribute to the excess deaths, complicated even further should some of these test positive for covid. So again, headline figures are not enough, you need to dive deep into the data, and stop fixating on one measure or another.

 

[AdamWW]

The problem, or at least one, is that over the period of the pandemic many other treatments have been scaled back, postponed or even cancelled. And as the months go on, some patients that may previously have had successful treatments will start to contribute to the excess deaths, complicated even further should some of these test positive for covid. So again, headline figures are not enough, you need to dive deep into the data, and stop fixating on one measure or another.

Yes, I believe I mentioned that.

If you're looking for a "how well have countries handled this" comparison, then something which includes this effect is probably more useful than something which doesn't.

But we do need to wait for more time to pass - countries that look bad now in excess deaths might look better in the long term.

 

[CaptainHaddock]

I think all this proves is that, whatever your viewpoint on how serious the virus is and how the UK has managed it, you can find statistics to suit your argument!

 

[AdamWW]

I think all this proves is that, whatever your viewpoint on how serious the virus is and how the UK has managed it, you can find statistics to suit your argument!

In a sense that's true, but some statistics are more appropriate to the use than others.

So you can look to the falling death figures and say that the epidemic is fizzling out, or you can look at the stable/maybe increasing daily infection figures and say that it's still here throughout the UK and just waiting for the opportunity to take off again.

But I would argue that the second view is the only one that's actually supported by the figures.

 

[Huntergreed]

In a sense that's true, but some statistics are more appropriate to the use than others.

All statistics are equal, but some statistics are more equal than others.

So you can look to the falling death figures and say that the epidemic is fizzling out, or you can look at the stable/maybe increasing daily infection figures and say that it's still here throughout the UK and just waiting for the opportunity to take off again.

But I would argue that the second view is the only one that's actually supported by the figures.

Indeed, and often the same statistics show a different picture depending on how they are explained, introduced or interpreted by an individual or group of people.

To me, cases mean nothing, as long as hospitals are able to cope and there's no risk of anyone being refused treatment due to an overwhelmed health board, then individual cases in themselves aren't too much of a concern. Where they do become a concern is when you follow a strategy of viral elimination, which Scotland and Wales are doing, and England is not. This is dangerous, as it means one of the strategies is less likely to be effective (elimination probably unless there's a border closure, which in the UK as far as I'm concerned is not viable or possible).

 

[Bantamzen]

In a sense that's true, but some statistics are more appropriate to the use than others.

So you can look to the falling death figures and say that the epidemic is fizzling out, or you can look at the stable/maybe increasing daily infection figures and say that it's still here throughout the UK and just waiting for the opportunity to take off again.

But I would argue that the second view is the only one that's actually supported by the figures.

Or you could look at the increasing amount of testing vs confirmed cases vs covid-related deaths and conclude that the virus is much more widespread, & that the mortality rate is much lower than previously supposed. Which is why one dataset looked at in isolation may paint a very different picture than multiple sets in combination.

 

[AdamWW]

All statistics are equal, but some statistics are more equal than others.


Indeed, and often the same statistics show a different picture depending on how they are explained, introduced or interpreted by an individual or group of people.

To me, cases mean nothing, as long as hospitals are able to cope and there's no risk of anyone being refused treatment due to an overwhelmed health board, then individual cases in themselves aren't too much of a concern. Where they do become a concern is when you follow a strategy of viral elimination, which Scotland and Wales are doing, and England is not. This is dangerous, as it means one of the strategies is less likely to be effective (elimination probably unless there's a border closure, which in the UK as far as I'm concerned is not viable or possible).

Can't agree.

If you're pursuing a policy keeping the hospitals coping (which seems to be what England is doing) I don't think you can afford to base it on looking at hospital deaths, because deaths lag infections and once infections start to grow they can do so rapidly and by the time you see it in the death figures there's a good risk you'll have to take drastic action to get infections down again.

 

[AdamWW]

Or you could look at the increasing amount of testing vs confirmed cases vs covid-related deaths and conclude that the virus is much more widespread, & that the mortality rate is much lower than previously supposed. Which is why one dataset looked at in isolation may paint a very different picture than multiple sets in combination.

Random sampling gives us good idea of how widespread it is, unaffected by changes in overall numbers of tests carried out and how those tests are selected.

It won't tell you if mortality rates are going down, I agree.

And yes I agree looking at just one dataset doesn't give a full picture.

 

[DerekC]

But we do need to wait for more time to pass - countries that look bad now in excess deaths might look better in the long term.

Again, that's the Johnson argument - it's too early to do "lessons learned". The problem is that we need to learn lessons now so that we can reduce the impact going forward.

To me, cases mean nothing, as long as hospitals are able to cope and there's no risk of anyone being refused treatment due to an overwhelmed health board, then individual cases in themselves aren't too much of a concern.

Are you distinguishing between infections and cases? If not, that's an argument for a reactive approach, where we just find cases as they self-report or turn up at hospital. If testing is worth doing it must be because either it reduces the number of infections (and therefore cases) through track and trace or it highlights areas where the infection is spreading so that preventive action like local lockdowns can be taken.

 

[AdamWW]

Again, that's the Johnson argument - it's too early to do "lessons learned". The problem is that we need to learn lessons now so that we can reduce the impact going forward.

I don't think it means it's too early to look at what might have been done wrong. But it's too early to come to a final conclusion.

Are you distinguishing between infections and cases? If not, that's an argument for a reactive approach, where we just find cases as they self-report or turn up at hospital. If testing is worth doing it must be because either it reduces the number of infections (and therefore cases) through track and trace or it highlights areas where the infection is spreading so that preventive action like local lockdowns can be taken.

Or indeed it highlights that infections are spreading too much across the country and national preventative action has to be taken.

 

[talldave]

I don't think it means it's too early to look at what might have been done wrong. But it's too early to come to a final conclusion.



Or indeed it highlights that infections are spreading too much across the country and national preventative action has to be taken.

Only if those infections have a detrimental effect on society. Everyone seems to be assuming that infection = hospitalisation = death. That's not the case and hopefully that's because the vulnerabilities are known and protective action can be taken to minimise infection of the vulnerable.

 

[AdamWW]

Only if those infections have a detrimental effect on society. Everyone seems to be assuming that infection = hospitalisation = death. That's not the case and hopefully that's because the vulnerabilities are known and protective action can be taken to minimise infection of the vulnerable.

I think most people here are aware that not everyone infected has symptoms, and not everyone with symptoms needs hospitalisation or dies.

If were were confident that we could prevent the vast majority from becoming seriously ill if infected, then obviously the attitude to letting infections build up again would change.

 

[yorkie]

....But this is why I'm in the camp that says excess deaths are the best figure to use - this also includes all the indirectly caused deaths, i.e. due to lack of treatment/diagnosis that would also have taken place...

and you would include deaths caused directly and indirectly by lockdown, along with deaths due to people unable and/or unwilling to seek treatment for other illnesses as a result of either policies, or patient decisions, to reduce the spread of the virus, right?

If were were confident that we could prevent the vast majority from becoming seriously ill if infected, then obviously the attitude to letting infections build up again would change.

The vast majority do not get seriously ill if infected; we already know this!