UKC

Don't blame me, blame the FT.

Their latest analysis

https://www.ft.com/content/6bd88b7d-3386-4543-b2e9-0d5c6fac846c

Overall it makes very sober reading about the global impact of the disease. 

Not unexpectedly cities have been crushed. London +96%, Bergamo Province +463% 

This is the part where someone comes along and claims all those extra deaths are either DIY accidents or due to mental health problems. 

Those numbers do make very sober reading

That's what I 'like' about this analysis, it's reductionist and shows a clear position.

Talking to my best mate, who's a doc, he thinks the whole thing is potentially just getting started as we really have no robust understanding of whether it can reoccur or not. 

What I find very sobering is that we have ten times as many deaths (261) here as our neighbour, Dumfries and Galloway (26).  Our population is just under 3 times bigger but ten times the deaths.  Spending on health 3% less than the UK average,  over the border is 17% more than the UK average.

Is there something we should be told?

 If hospitals are not overwhelmed in your region, then health spend is unlikely to be a significant factor. Look at your relative population density for a more likely clue.

In the charts (in the article), top right of the first block.. England and Wales + 37%. 

+37% was YTD, +76% was specific week in April vs average of last 5.

A bit of interpretation on the % figures given in the FT article.  I found their presentation rather confusing.

The graphic gives  the percentage increase above baseline excess deaths from 1st January until some recent date.  This is the "area under the peak above baseline" in the "all deaths" data . As the UK continues to lag Italy by about 14 days, the peak for the UK is significantly more truncated than that for Italy, so the numbers compare a measurement at different phases in the different countries infections.  A "Year to date" measure will never be a good comparison as the "start date" is different in each country with respect to the arbitrary year start.  But it's an almost meaningless comparison before the peak is well past.

In another 2 weeks the UK is likely to be at 90% (or more) above normal deaths since Jan 1st.  In both countries, it looks like the peak death rate is about twice the baseline death rate.

Edit: Well, I say we continue to lag Italy by about 14 days; perhaps unsurprisingly given our comparatively weak lockdown vs Italy, our daily infection numbers and death numbers aren't dropping off like Italy did, so we're going perhaps looking at a lot more deaths in the short term as the time racks up.  The long term is anyone's guess.

I was looking at the ONS figures yesterday comparing the number of death last year to last years and the 5 year average. Up to the last 2 reported weeks the fatality rate was about the average, and then there was the expected spike.

Looking at the spike a bit closer there were 14000 more deaths registered than the average for those 2 weeks, but just over 9000 atributed to covid. So either the cause of death is not being reported properly as covid in a 3rd of the cases, or there has been an extra 5000 fatalities in 2 weeks not directly from covid but possibly covid-situation related

Datasets here https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriage...

Mostly covid 19 but NOT confirmed covid 19 cases?

https://www.ukclimbing.com/forums/off_belay/should_we_be_in_lockdown-718370...

Quite possibly. I think that's the assumption in the article anyway.

If the extra deaths change in the same way as covid deaths (same epidemiology) or are disproportionately lung related they're likely to be covid rather than unlikely coincidences.

There's definitely stories of other extra deaths though.

https://m.facebook.com/Channel4News/videos/244222653486948/

What are the infection numbers?   

If there's a lot more infections it could be because you're in a far more popular tourist/second home location and you've got a lot more seeding of cases from the big cities.  Maybe even an effect from people who know they have it choosing to isolate in a second home somewhere nice.

If it's roughly the same number of infections but far more deaths then it has to do with the health service.  Or maybe the demographics of the population if lots of people like to retire to where you are.

Sort of fits in with what I'd been thinking. Which was...

How comes countries with areas with population density issues (for example cities in Pakistan, India and Bangladesh, but also lots of Africa and South America) have such low covid deaths. They are surely under-reporting because of lack of testing.

They are also much younger populations which covid will affect less. 

Equally likely a Lakes local back from Italy went to the pub and infected a load of people there who then proceeded to go back to their own communities spreading it widely.

Similar to that happened in Spain. The biggest outbreak in Spain at the beggining happened in the Basque Country, where there's still higher than average infected and death rates, and the origin of the outbreak was found to have been at a wedding where over 60 of the attendees contracted the virus and then dispersed over a very wide area comprising several provinces.

I don't know where you get the idea that it's people from cities spreading the virus in rural communities via tourism or second home use, but I'd be interested to see what evidence you have if true though.

FWIW and going by the infection rates (admittely innacurate due to the testing pickles but it's all we have), and taking London as an example, it has 16% of the positive cases in the UK while having just under 15% of the population, so not the boiling soup of disease that reading the news might give the impression of
 

True, I'd forgotten about that. There are so many factors that are still not known or properly understood.

I suspect that job prospects for epidemiologists and statisticians will be improved for the next couple of years.

To ClimberEd: thanks for the clarification, year to date wouldn't be my measure , particularly for comparison.. bear in mind now that Italy and Spain now for example will now be coming down by that measure while the UK is going up so they'll be a lot closer at the end of the year.

Apparently some people I know are a 'hot spot' thanks to a ski trip to Northern Italy.  One ended up on a ventilator

Countries with poorer general healthcare may also have fewer COVID deaths because fewer people survive into old age/poor health who would be in the big risk demographics.

Talking about confusion: Does anyone have an idea why Italy has a baseline of 4500 and Spain 8000 and  France and E+W around 9000? It looks like they are missing about half of the regions in Italy in the analysis.

Also, if as there is some speculation, the development of CV in the body is actually assisted by the drugs used to treat heart disease, then that also would mean developed countries with stronger health care systems actually coming off worse (not yet proven though)

Which drug is that out of interest?  Or just general immune suppressant drugs?

We have a higher infection rate than Gateshead per 100,000 but you would not call Cumbria an urban environment - is that a likely clue?

could be - though the local saying has been for a long time that it's not a good place to get sick!  Perhaps Gateshead was over here for Mothering Sunday.  Nevertheless the disparity in spending on health is a scandal

ACE inhibitors maybe https://www.reuters.com/article/us-health-conoravirus-blood-pressure-ins-id...

The low (so far) figures in developing countries may be down to lack of reporting.

But they may also be down to the low age of urban populations. Italy for example has a median age of 47 while the UK is 42 but Nigeria is 18. In Italy many older folk live with children and grandchildren whereas big cities in developing countries like Lagos, Kinshasa, Nairobi are mostly filled with younger people who came to find work leaving the older family members back home in the village.

Since your hospitals appear to still be able to treat people it's probably mostly due to a lot more cases in Cumbria, roughly the same fraction of which are dying north and south of the border after demographic differences are accounted for. Is there already a poor health cluster around the Barrow shipyards? Perhaps coupled with some bad luck, maybe a super-spreader, maybe regarding where the virus got established locally early on. Why? Pre-lockdown tourism, the M6 and bad luck would be my guess but I guess it may be unpicked in years to come.

jk

You don't have enough cognitive dissonance. All those people were going to die soon anyway, we have herd immunity now as they covered up that it started in November,  it was all overhyped as the UK hosptitals were not overwhelmed, so it's not much different to flu, and quinine and bleach are being criminally ignored when they could help. Its now vital to end lockdown to protect the economy or that will kill way more people.

I think the idea that urban living accelerates spread is slightly misreported or misunderstood. It seems more like reliance on mass transit than population density per se. They go together some places but not others and there is an element of chance overlayed on all of this, when it's growing exponentially an early (or late) start make a big difference.

jk

If you like analysis, there's an EU project that monitors mortality rates real-time. They release public data weekly.

https://www.euromomo.eu/graphs-and-maps/#z-scores-by-country

Interestingly shows England as having the highest deviation from average mortality rates anywhere in Europe, the rest of the UK affected much less.

Countries like Denmark, Norway haven't really had any significant excess mortality, and a few countries have actually had a negative deviation - eg Austria.

It'll keep health researchers busy for years to come ....

At least that keeps my mrs busy anyway. Silver linings.

jk

Thanks for that link! So much for our hope of buying a nice farm house here in Hesse on the cheap...

Also, what happened in Estonia last summer? Brief bout of immortality?

CB

Mark Handley's graphs would indicate that the North East has the worst problem outside of London

http://nrg.cs.ucl.ac.uk/mjh/covid19/

So Cumbria doesn't seem to be the hotspot mentioned here every so often.

a) Their population doesn't travel as much and therefore the infection took longer to get going.  They're going to be the final peak - First we had Wuhan, then Italy, then Spain, then USA, next it'll be the third world - I think Brazil, then India.

b) some hot places don't seem to get it as badly

c) Under-reporting for sure - the number of cases is also low. If cases were reported properly you'd have loads of cases from the general population but few deaths because their demographic doesn't have many old people.

When the Channel Isles had 88 cases (27th March) I noted the other nations with similar case numbers on worldometers:-

Cameroon 88, now 1621;

Uzbekistan 88, now 1869;

Palestine 91, now 342;

Montenegro 82, now 321;

Cuba 80, now 1369;

Nigeria 70, now 1273.

Belarus 94 now 10,463;

We have 525 cases now and are well over the initial peak. Belarus have no restrictions and a President that thinks drinking vodka and taking saunas helps. The daily percentage increases in Cameroon and Nigeria haven't dropped unlike those in other countries on the list. Like the USA, they may not have been rising as stratospherically as some European countries did to start with, but they've not got it under control by any means.

Look you , all the stories are made up here, a local delicacy

https://www.newsandstar.co.uk/news/18366549.data-reveals-cumbria-high-covid...

is just one of many reports in the press over the past weeks

plus DEATH RATES - Dumfries and Galloway at the time of posting 26, Cumbria 261 - all just made up

and this seems to be indicative, no need to read numbers - look at the shading and compare neighbours

https://www.bbc.co.uk/news/uk-51768274

comprendi?

That 108% excess would grow very fast if the hospitals had started really struggling. Thankfully it seems we've managed to avoid that for now.

Does anyone know what the plan was and is for staffing the Nightingale hospitals? Spread NHS services thinner and retrain/redeploy or maybe employ veterinary staff? I get and approve of the the idea, building an insurance policy while the situation was still quite unknown and alarming but without the ability to staff them they start to look like reassuring white elephants or expensive propaganda.

jk

It was obvious that Ed had shot himself in the foot from the first post as he was blatantly misusing the data. As wintertree said, the excess graph was time truncated with Italy just past the peak and us 14 days behind. Extrapolate for the UK and the peaks would be about the same (as time is now proving).

One other factor we always pointed out to Ed was: yes we are different from Italy (no shit!) even though we were always closely following their mortality curve, things could change. Some of us warned those changes could be for the worst.

https://www.ukclimbing.com/forums/off_belay/we_are_not_going_to_follow_ital...

Well the worst has arrived as next week we will almost certainly overtake the formal death stats of Italy and have the highest official deaths in Europe. Compared to Italy our new cases and deaths are not declining as fast, past peak (quite expected as they have a harsher lockdown), so the gap will continue to increase at our expense.

https://www.worldometers.info/coronavirus/country/italy/

https://www.worldometers.info/coronavirus/country/uk/

Italy got caught early during their flu epidemic and had a massive excess death spike in the initial outbreak area as hospitals were overwhelmed (Covid mortality rates and secondary factors increasing mortality increase local mortality rates almost an order of magnitude when hospitals are overwhelmed). We luckily didn't have any of that. We also had those two extra weeks to prepare. We, unlike Italy, from phone stats, didn't have a huge and dangerous mass exodus from the infected area just before lockdown. The Italian government had all the best excuses but we in the UK end up as 'the sick man of Europe'. Still today our government ministers deny they made errors (testing, PPE, care homes etc....) and insist they always followed the science (starting by burying Cygnus recommendations,  then when China was hit, focussing more on brexit and eventually Scientists like Fergerson, privately advising the government, forced to publicly speak out against herd immunity and the slow response) and that every country was hit hard (yes but the US aside we were hit the hardest).

Cumbria has 24% of residents over 65 compared with England and Wales average of 18%.  

However Dumfries and Galloway has 25% over 65.  So that doesn't explain it.  

Looking at your bbc map the same pattern isnt there on the other end of the border.  I'm not convinced it is a Scotland England thing.  

A very good comparison graph is available here

https://ourworldindata.org/grapher/daily-covid-deaths-per-million-7-day-ave...

However this is half the story as I think as the OP was talking about with his FT analysis. 

Looking at actual excess mortality shows a peak above the mean mortality rate of over 44 standard deviations in England but only 22 in Italy (and 34 in Spain)

https://www.euromomo.eu/graphs-and-maps

Probably safe to say at the moment that England as a country has been the worst hit in Europe, the rest of the UK less so

deaths per capita are much higher in the UK than in the US, double.

new York was bad the rest of the country has largely been ok but we’re undoubtedly going to see a second wave.

It looks like the UK could soon be the worst hit per-capita of any country with more than a very small population.  I’d be more nervous about the badly named “second wave” in the USA than over here through.  #LiberateCovid

I'm curious at what point we overtake Italy, with adjustment for having a slightly higher population.

Their population is about 9% lower than us, their death rate is perhaps 30% lower, and its falling faster than our death rate.  Too many unknowns to sensibly extrapolate a predicted “crossover date” from but it’s the way we are - currently - headed.

I don't understand why this statistic has gained credibility. Yes, the US has suffered 58,000 deaths and the UK 28,000, (perhaps the highest in Europe) but the US has 5 times the population. Surely the figures we should be comparing (if we need to) are deaths as a proportion of population.

On that basis, the US is far from worst affected. And I've heard, though I can't recall the exact figures, that some smaller European countries (eg Belgium) are worse affected per head of population than the UK or Italy, France and Spain.

Each death is a tragedy for the families affected of course.

Martin

Unfortunately I don't think it will be long before we overtake Italy. Totally guesstimating here from the general shape of the "curves", but I can't see Covid deaths in the UK by the end of 2020 being less than 50,000 and likely to be close to or above 100,000 (and that's without considering any further waves).

I looked back at one of the posts on the site (can't remember the name) where the "hammer & dance" was first(?) posted. The post which asks "when should you [i.e. leaders] act" and concludes NOW because delaying 1 day on the exponential increase most countries had would increase number of cases by 40% which was quite staggering.

So the UK delaying by maybe as much as a week has vastly increased the number of cases and hence deaths in the UK.

I'm also reminded about when Israel first closed it's borders (or at least quarantine/trace controlled them) which was quite early and I saw lots of "that's over-reacting" type comments (not on UKC). The basic thing that our government got wrong was the delay and indecision.

Worst thing that could happen if a country over-reacted and brought in too many/tight restrictions "sorry about that, we over-reacted, nobody died (because of that over-reaction)".

Worst thing that could happen if a country under-reacted - just look at us in the UK.

it's true of so many actions - steer early and steer easily.   Steer late and violently and there are dangerous consequences.

I'd like to know who came up with the insane idea that we'd grow weary of lockup so we should delay it to exactly the right moment.

At the time I likened it to an old moon l;anding game where you balance thrust, fuel and descent rate.  This was not, unfortunately, such a game our great, good and unelected political advisors should have been playing with "their tiny plastic hands". .

The Z score data on the web site linked to above is about the best indication and England is already way worse than Italy (and the other countries of the UK).

https://www.euromomo.eu/graphs-and-maps#z-scores-by-country

The comparison is fairer because it is based on 'excess deaths' so it isn't messed up by countries counting things differently and the Z numbers are normalised for the size of population.

In terms of excess deaths we are already somewhere between 40k and 50k, that would suggest 60k is already pretty much locked in (as there will be deaths on the way down from the peak) and it could get to 100k if the government screws up when rasing the lock down and there's another spike.

I remember that one, keep correcting the whole way down, start too late and....

SPLAT

Why would Z score be a good indicator, given that it is in units of standard deviations?  (So that, for example, a larger country would have a smaller scatter in data points, so a larger Z score.)  

Surely the more important indicators are excess deaths, as a percentage of the population, or as a percentage of the usual morality rate?

Only problem I can see with the Z-scores (if I've understood it correctly) is that it's showing a scaled version of how many standard deviations you are above/below "normal". If UK has a different SD to other countries (i.e. our "normal" number of deaths fluctuates more or less than other countries) then the comparison may not be ideal.

Ultimately, the simple measure of excess deaths per million of population might be the best measure of how badly a country has been hit (where excess is defined as above/below the average over the last x years, with each country using the same definition).

Edit: Damn, Coel beat me to it.

What, pray tell me is the "morality rate"?

Sorry, couldn't resist 😁

And regardless of which measure to use, it will soon be pretty clear using virtually any measure that we will have the most Covid deaths in Europe - which speaks volumes for the quality of our government's preparedness and response.

Getting your own back I see! 

Assuming bad flu hits large and small countries equally the systematic deviation that year from norm is linear with population. Standard deviation is a measure of those systematic variations year to year so it would be linear with population.

But if flu tends to hit small countries tend in a uniform way, but hits large countries in a patchy way (e.g. North different from London), then patches will average out, so in the large country year-to-year fluctuations will be less, so a Z score will be higher. 

But anyhow, one uses the number of standard deviations if one is interested in the statistical significance.  It is not what one does if one is interested in the magnitude of the effect.

Yes.

My bigger concern is the 3rd or 4th wave in the fall. I think summer living will hopefully keep the second wave less severe but if this carries on and we start going inside and kids in school in the fall, then it could be very bad. 

I'm not a stats person but isn't the 'large countries may have a lower year-to-year standard deviation than small ones' the problem the Z score is trying to address by dividing by the standard deviation?   As I understand it Z score tries to allow a 'fair' comparison between values of two different normally distributed variables.   So it tells you how much worse one country has been affected than another. 

It seems to be widely used for comparing mortality e.g. this EU project which made these graphs comparing between countries and I've seen it in official stats from the CDC in the US as well where they want to compare between states.

Dividing by the SD gives a good comparative measure of how far away from "normal" things are. So you could say that the UK spike in deaths is much more abnormal than (for example) the French spike.

But unless the SDs are similar it doesn't tell you which one is worse in absolute terms.

These z-score stats are probably assuming that the SDs for different countries will be fairly similar, partly because they're dealing with very large populations (from a statistical point of view as well).

I haven't seen anything that demonstrates whether this is the case or not (because I'm not bothered enough about this detail to search) but it's probably out there.

I suspect that comparing us with Belgium would show that our z-score is worse, but they're worse in absolute terms (more covid deaths/million).

Anyway, in the current situation this is basically a "race to the bottom" discussion which whilst maybe interesting (in a morbid fashion), isn't particularly useful.

A score in "number of standard deviations" is telling you how statistically significant the effect is.  It is answering the question "is the increase real?" (as opposed to being a statistical fluctuation). Or, as Michael says, "how different is this from normal fluctuations?".  

This could be very useful for some purposes, such as: "is there a new cause of mortality leading to deaths, or is this just a fairly large fluctuation of normal behaviour?".

But, once you get beyond about 3 or 4 standard deviations (Z score of 3 or 4) those questions are over and done with -- yes it is real and yes it is different. 

What the Z score does not tell you is how big the effect is. Just to illustrate the point, suppose you're testing new drugs:

In one trial of Drug A there was a 70% reduction in death rate.  But there were only 8 participants. The statistical significance is low.  The reduction could be fluke.

In a trial of Drug B there was only a 5% reduction in death rate. But there were 500,000 participants.  The 5% reduction could then be highly statistically significant (ruling out it being just a fluke) even though the effect size is rather small. 

So "statistical significance" and "effect size" are two different questions.

Large countries have a larger standard deviation than small ones! It is just that the standard deviation increases less rapidly with increasing baseline than the baseline itself. A good measure for a comparison between countries would be (deaths-baseline)/yearly average of baseline.

The fact that the z-score does not correct for population size can be seen in the data directly by comparing the z-score of the individual countries with the z-score of all the data together. The z-score of England at the peak is 45. Now add the data of all other countries (most of which are much less affected) and the z-score goes up to 55.

The z-score is pretty good at comparing e.g. the 2018 flu with Corona in the different countries. It is also reasonably good when comparing countries with similar baseline death. (The latter depends not only on population size, btw, the age distribution comes in as well.)

In isolation NY is as bad as anywhere in the world*, obviously it is much worse than anywhere else in the US. However, the very concerning thing is that many of the states are still on an upwards trajectory towards a peak in the first wave. US state borders are similar to Shengen borders in European in practical terms and we saw different timing of infection. I don't know and I don't want to be proved right, but I think in terms of deaths, it is going to get much worse in the US, even in the first wave.

* Edit to add - NY is twice as bad a Belgium and NJ is also significantly worse than Belgium.

European agancy monitoring covid 19 says we are one of the few countries in Europe who need to do more

https://www.theguardian.com/politics/2020/may/04/uk-behind-most-european-st...

Euromomo now show England have comfortably the highest level of excess of deaths in Europe (as it was obvious to many where we would end up when Ed started this thread). We are not following Italy we have overtaken Italy and metaphorically lapped them.

https://www.euromomo.eu/graphs-and-maps/

  I don't really understand how these calculations are done. I have three questions on the euromomo figures:

1) The government is obviously aware of them but I have heard Van-Tam twice say that they are awaiting reliable excess mortality statistics. So what do they think is wrong with the euromomo stats?

2) My understanding is that covid (and overall?) deaths in the UK peaked a few weeks ago so how come the excess number was still rising until the latest week?

3) If the baseline (5 year average?) falls that will presumably increase the excess mortality rate (if current deaths remain constant). I am wondering if the fact that the UK is a few weeks behind eg.Italy, and that baseline mortality falls in the spring , increases excess mortality?

The government know it's true and are just not admitting it. Van Tan works for the government. This isn't a paranoid statement, it's pragmatic reality: in his position either you accept the agreed line of presentation to the public (for the greater good) or you resign.

These excess deaths are real directly comparable numbers, they don't depend on test levels or accuracy or on methods or accuracy of registering C19 deaths. They will be subject to age profile differences, areas of highest population densities etc, but the biggest hit nations in Europe aren't so different in that respect (plus another disadvantage of Italy cf the UK is an older population).

Well, I'll have to take your word for <1>. Another example of how deeply crap the media is for not asking.

But what are the answers to to <2> and 3?

For 2 its obviously that a combination of  community and care home C19 deaths and other deaths inadvertently caused by the crisis (eg people not seeking to deal with urgent health needs, or not seen as normal services have been delayed, etc.) have been increasing after the official C19 hospitals deaths peaked.

For 3 the differences to variations in baseline are now trivial compared to excess deaths. The size of the data is real. Other sources show the varying baseline used as a comparator and the range of that variation.

I guess the shaded yellow region called "corrected for delay in registration" is not yet 100% reliable.

The last datapoint is week 17, so two weeks ago.

Yes, but the effect is small compared to the uncertainty in the death toll numbers. The baseline moves about 100 per week and the excess death is >7500 per week?

I believe the EUROMOMO stats for UK are based on our own ONS figures. The latest stet are out today.

https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriage...

There's a 2 week lag to these.

You're looking at it back to front there. Seasonal variations should mean we have fewer total deaths than if this happened in a month baseline was higher.

Question for you now. The UK will be one of the worst hit countries in Europe, if not the worst. Why do you think we have been hit so badly? Government? National habits?

Far, far too early to tell. We don't understand either the disease or its transmission, the impact of the response, or the decision making which led to that response nor the execution process  of that response. And I don't think UKC is a very good forum for ruminating on it.

Incidentally, I think you've misunderstood my Q3. It's not about the number of deaths, it's about the excess mortality rate.I think my confusion was that I thought I'd I'd seen it represented by a percentage number, which would be very distorted if the baseline fell by much.

Edit: I see that excess mortaility is actually usually measured by the Z score ie.Standard deviation. I would have thought (happy to be corrected) that this would cause some of the same problems as if it were measured by percentage?

  Hans seems to have provided the logical answer: that the baseline seasonal variation is small relative to overall deaths so not that important. Thankyou Hans.

I suspect that as France have just found out, it has been circulating in France and England a lot earlier than we thought. So when we stopped track and trace, and went to monitoring hospital admissions we saw a sudden and unexpected increase in cases. Just like Italy did. This increase is due to awareness. France have just found a sample taken early December that has tested positive. 
 

This isn’t a failure of government, this is directly attributable to China keeping it hidden. 
 

This could actually be good news as it would mean that it’s not a deadly as we thought and there could have been large numbers of asymptomatic people wandering about. 

Major World City in London ( if not the World City)equivalent to New York. Diverse multi cutlural residents with huge amounts of international travel etc.Germany for example had 200,00 citizens abroad, we have 1 million.Staggering difference.

Slow to test, but no structure in place.

It will be interesting to find out what PHE were saying back in December and were the govt ignoring or pushing back on the advice.After all we are told that we are alledgedly global leaders with epedemics ( talk about overegging the pudding)

Focus on election and post election Brexit.Clear out in cabinet of some very experienced ministers, Hunt and Hands in particular. Hunt leaving probably unforgiveable.Weak ministers appointed like Patel, Willamson.

I think this is ONS statistics now reflect what was happening in hospitals, care homes and in ordinary homes in mid April. The daily government announcement reflect what was happening in hospitals (& carehomes?) yesterday. 

2020 deaths, baseline figures and the difference (2020 excess deaths) represent deaths registered by doctors.

A fall in baseline does not cause death in 2020.
A fall in baseline does not cause doctors to start mistaking the living for dead in 2020.
A fall in baseline is not the result of doctors mistaking the dead for living before 2020.

Tens of thousands of doctors registering tens of thousands of deaths incorrectly seems unlikely compared excess deaths caused by something known to science and medicine - mostly Covid19.

https://fullfact.org/health/covid-deaths/ - interesting that excess in deaths not mentioning Covid19 are outside hospitals. 

Debateable.. We were leaders in how to manage epidemics. Built on Foot and mouth then heavy input following SARS, Ebola etc. Singapore, South Korea learnt from us.Something has gone astray with all the experience and knowledge we built up.

We really need antibody tests to see what the true rates of infection have been and then what the true mortality rates are. A UK virologist was saying yesterday that they think the mortality rate is between 0.1 and 0.9%  - the lower end of the estimate is consistent with influenza. 

I'm not sure what you thought I was thinking but it wasn't any of that!!

Very much the point made by an academic working on global health policy at the London School of Economics:

https://www.theguardian.com/commentisfree/2020/may/01/uk-global-leader-pand...

What has gone wrong has been thinking it didn’t arrive until the first symptomatic person walking into a surgery in February. If it’s been here since late December and cases have been doubling every 3 days we should have had about 17m cases. Not the 3m we think we have. 

And the same could have happened in Wuhan. Maybe it emerged much earlier and rather than covering it up, they just didn’t know. 

You did ask "I am wondering if the fact that the UK is a few weeks behind eg.Italy, and that baseline mortality falls in the spring , increases excess mortality?" and I responded with what behavior from doctors that would require to be true.

The French case was December 27th.

https://www.theguardian.com/world/2020/may/04/french-hospital-discovers-cov...

See my reply to Mike Stretford. I wasn't asking about the absolute numbers but about numbers as encapsulated by percentage , although I now see that excess mortality is usually measured by standard deviation. Hans gave the answer I think.

Your last comment surprises me given the topics you have wanted to discuss on here in the past.

No, I understood.

Yes. And they would have caught it a couple of weeks before that as the first week or so symptoms are fairly mild. Then you either shake it off or things go downhill rapidly. 

AAAggghhhh! Is this one of the 1)"I know what you were really thinking" assertions? 2) you have not yet read my explanation? 3) You have read your explanation but believe your  answer is still the relevant and correct one?

I concur, a combination of factors. I'm not trying to score political points as I don't think a different government would have fared any better. May would have probs taken it more seriously early on but the push back would have been stronger (same if Corbyn had somehow managed to get into number 10). For all Boris is a cavalier character, at least when he did start taking it seriously people knew that had to sit up and take notice. 

We will have to look at this, there's a lot of denial about at the moment, but given the notice we had, this is bad.

I’m guessing that it’s still pretty high. We don’t see loads of 50 year olds in hospital with flu. And it’s the re-transmission rate that’s so high that’s causing the problems. if you get a bad case of the flu, you disappear to bed pretty much straightaway that you get symptoms, you don’t walk around spreading it for a few days first.

Based on what evidence? S Korea, where testing is excellent, with full track and trace, and the hospital system up with the best prepared, indicates 0.24% mortality. I'd be amazed if any European country comes close to this mortality rate, given the common failures in proper care home protection. The best we have in Europe on upper limits from test data (testing in the EU will miss many positives) is 2.4% in Germany.

Flu mortality rates in Europe are normally well below 0.1%.

https://www.who.int/news-room/q-a-detail/q-a-similarities-and-differences-c...

Except it would be an influenza that infects far more people than normal resulting in far more excess deaths than a normal influneza. 

Alternatively it infects a smaller number of people with a higher mortality also resulting in far more excess deaths than a normal influneza. 

Modified 3, other 'answers' are relevant and correct. Mine was more an observation on your question, which I stand by.

What’s the demographic of S Korea though. 
 

In Chris Wittey’s lecture he seems to think countries outside Europe and the US will do much better because their population demographics are much more weighted towards the under 50s. 

In reply 

If Whitty said that it's yet another really stupid thing he has said (like repeatedly saying we were 4 weeks behind Italy)

Countries by median age:

https://en.m.wikipedia.org/wiki/List_of_countries_by_median_age

Japan 2nd oldest, Hong Kong 8th, S Korea 36th, Taiwan 47th, UK 50th.

The issue will the be economy in all this, we might have protected it far better at lower cost.That will haunt us in years to come.

And we have to remember that financial instuitions had pandemic planning as a scenario and had practised for it.They have not been hit other than with a backlog for loans claims, the finance sector ( as a big earner for the UK) has taken it well . Their planning worked.Would have been a real mess without that.

And what about obesity. If the death toll for obese men is not a wake up call then I do not know what is.

If the asymptomatic fraction were significantly higher than the symptomatic fraction as necessary for millions to have shrugged it off then you'd expect to observe (by symptoms alone) relatively little transmission within 2 person households. That isn't the case as far as I'm aware though I do know of one instance where one got quite sick, one noticeably, the third not at all, the rest of my friends who've had it the household got it and got sick.

jk

With this pandemic those who made the strong case for social distancing earlier and  pandemic preparation earlier have done the best and have the least economic damage as a result. It's another reason why its such a big issue that not only did we fail to wake up after the Cygnus excercise, we buried it.

The paradox being that with the financial instituitions and their pandemic planning it was the Bof E and the FCA that told them to do it way before Covid.

It is typical, tell others what to do, but do not embrace it yourself.

41 in South Korea, 41 in U.K., 47 in Italy. It will help a little.

Anecdotally in my wife’s office. One person has presented symptoms. Two others were tested as they had close contact and were symptomless positive. The other 8 haven’t been tested as not considered close contact for whatever reason. The symptomless positives families showed no symptoms although that included 4 under 30s. 

Thanks. He only showed a few countries profiles to demonstrate why we would be harder hit. Which is why I asked. 
 

As neilh  has noted. Obesity may be a factor but not sure if that has been positively identified as being significant. I thought that we were seeing large miners of obese patients, because we have large numbers of obese people. ie the admissions mirrored the general population. 

(Edit: Thought I was replying to Postmanpat - must have slipped)

This was said in another thread (or maybe it's way up there in this one - I've lost track), but the z-score which divides by standard deviation only shows you which one is statistically "worse", i.e. which country's actual excess deaths are least likely to happen just because of random variations in the number of deaths every week. The corollary being that if the excess deaths are less likely because of randomness, then it's more likely that there is another factor (i.e. Covid) at work.

What it doesn't show you is which country is actually worse affected - because each country's standard deviation might be different. For that you need a measure like deaths/million or % of population.

Just remember that most of the time, any statistics that support an argument are shown, and those that don't remain hidden - until someone else digs them up.

I agree. The z-score might show something is ten times worse than the usual fluctuations for a stable country but for an unstable country with prolonged civil war, unreliable food supply and no functioning health system the same thing might be only twice as bad as the usual fluctuations.

There are multiple ways of representing Covid19 data with different strengths and weaknesses. The z-score is a good way of saying this is significant (or not) for this country. 

That was my understanding. Are there many obese people without diabetes? And are they over represented in the stats. 
 

How could you even separate them? 

I assume that is the bread and butter of disease investigation as comorbidities (often the same ones) exist for other diseases too.

It might be as simple as not all obese people have diabetes and not all people with diabetes are obese but I expect they have cleverer techniques too.

I find the 200000 number for Germany in the skiing season hard to believe. Are you sure this isn't travel outside of the Schengen area? As far as I know, the border crossings aren't even monitored.

I suggest that it is owing to a whole bunch of factors combined.

One factor that I've mentioned before is that, owing to the English language, the UK is by far the most popular destination for Chinese university students in Europe (and second only to the US in the world).

Thus, around 150,000 Chinese students will have gone home over the Christmas break and then flown to the UK in mid-to-late January for the new semester.    A large fraction of those will have gone to London universities.

Nearly all of these will have been young and healthy adults, and so would have been unlikely to have anything more than mild symptoms themselves (as anyone who teaches in universities knows, it's normal, at any one time in mid-winter for 20% of so of student to have cold-like symptoms) but could have started spreading it.

a. 'we' i.e. the Oxford/Cambridge/London establishment think they are world leaders in everything.  When it came down to it and they had to do it for real it turned out they weren't.    There should be absolutely nothing surprising about this.   If you go to the US or Germany or France or China and ask them where the international experts are they're not going to say London.   Leaving aside UK exceptionalism one would expect the countries which actually dealt with SARS would probably be the 'leaders' in combatting Corona virus, and that is what the outcomes show.  China was ready to deal with Wuhan because it prepared after SARS.

b. the Tories wouldn't spend the money to follow through on the recommendations.   As a result they've got a multiple times worse than necessary crisis to deal with and they are having to spend orders of magnitude more money.

No kidding. 

I guess we'll find out pretty soon. I'm not sure what we do with the information if it is a fair bit less deadly than we fear, it's still clearly too dangerous to live with but I suppose the natural herd immunity hope maybe gets put back on the table.

jk

The early UK cases were contact traced after ski trips to Italy and buisness trips to China at the very end of January when students had been back a few weeks. I'm not aware of any student contact tracing. 

I'm not at all sure that we have a full picture of early cases in the UK or Europe.

It might indicate where we are on that peak. Whether it would have slowed naturally soon, whether a harder lockdown now for a couple of weeks will eliminate it or whether there are huge numbers of asymptomatic infections roaming around and lifting restrictions even slightly will mean it takes off again. 

I'm pretty sure we do know it didn't mostly come in with international students and did mostly come from ski trips in N Italy.  Contact tracing for some of that group was too slow so it expanded too quickly. Wuhan was closed so late arrivals from there were unlikely and any late arrival students with the virus only would have added to a growing outbreak.

This idea the virus has been around for a longer time, spreading slower with lower mortality is dangerous conspiracy theory nonsense spread by those who desperately want lockdown to end. Their conclusion is everyone has had it and goverments are idiots and lockdows are not needed. They misuse data to prove their points (ignoring the social distancing in Sweden, get very excited by the French case, push dubious modelling efforts that don't match real data). There is no evidence anywhere of national mortality rates below 0.2%, growth is uniformly fast without social distancing and not population limited (at 0.1% mortality the NY per capita data of 1285 deaths per million and growing is impossible without little herd immunity) and only a few highly contentious papers about high levels of assymptomatic infection (upto the BMJ paper with 75%). European countries will be lucky to get away with 0.5% mortality as it spread to care homes almost everywhere in the EU. Germany mortality is currently lowest at 2.4% on its current high level of testing (the real number will be lower but not that much). In the UK we have R only just below 1 despite our social distancing. If we loosen up now at current high population levels of infection it will be a disaster and hospitals will be overwhelmed 2 weeks later before any subsequent change could be made.

Isn't that why we had a lockdown because there wasn't any science for "would have slowed naturally soon"? 

 "couple of weeks will eliminate" the decline in deaths is very slow (900 per day 3 weeks ago, 700 per day now)* so this looks very unrealistic and improbable even for a harsher lockdown unacceptable in a democracy

"asymptomatic infections roaming around" - this is difficult to gauge from published numbers - increased testing, fairly static number of new cases, fewer people in hospital - worse than at lockdown but slowly declining?  

*Slide 8 of Downing Street Daily Briefing for the numbers

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/...

Every time I see slide 8 I think the orange curve might have the same shape shifted 1 week earlier, squashed 5-10 times lower if the lockdown was 1 week earlier. Since growth much faster than decline it would allow a shorter overall lockdown. That was an opportunity missed costing hundred of billions of pounds and tens of thousands of lives. 

Reports are emerging of non-peer reviewed paper suggesting that a mutation that happened in March allowed the virus to become more transmissible. Certainly that would make sense if it’s been here since December. Why didn’t it take off in late January? 

Quite simply as its unlikely the version in that French test is the new virus or the Wuhan form as both are too infectious and dangerous (unless we were very very lucky)

On a log graph of UK cases you would expect the curve to get steeper if this were the case. In fact the curve gets less steep in march including before the lockdown.

Certainly that would make sense if it’s been here since December. Why didn’t it take off in late January? 

I don't know, but it seems in every country it takes off at different times, possibly at a random time determined by a single failure in contact tracing which is never going to be perfect.

Had another look at the Euromomo graphs, am I getting this correct that the UK's z-score is approx 40 standard deviations above the mean!!!

40 standard deviations, that's like wow. In excel my "norm.dist" function gave 4.6054E-308 for a 1-sided tail 37.5 standard deviations from the mean, at 37.6 it just went f**k it and put 0.

I suspect 40 SDs above the mean isn't due to random fluctuations.

Well yes, but is anyone disagreeing?  The counter-claim, that covid19 does not exist and the extra deaths are simply a statistical fluctuation, would be a somewhat eccentric and minority opinion, would it not? 

Sorry, should have put the sarcasm flag on. It's just that I can't remember ever having seen statistics that had something that far away.

A real thought though, in the same way that statistics are not very useful with small populations or sample sizes, is there any real difference between 20 SDs above mean and 40 SDs above mean; does it say anything meaningful?

They're both so obviously not normal distributions that I wonder whether it is still valid to use them to support being able to say that 40 is worse than 20. Rather they're both saying, these are so not normal that you should use some other basis to compare.

Can we have a statistician to comment please.

20 or 40 SD is meaningful, it means something very unusual, so unusual the distribution is almost certainly not Gaussian.

Hence probability is closer to once every decade, century or millennium* than once every 10**308 years.

*based recorded history of plagues and pandemics

Can't have expressed myself clearly, I realise that both 20 & 40 SD are meaningful in that they're both saying, not Gaussian. What I meant was can you meaningfully compare 20 SD to 40 SD or are they both so extreme that comparison using that measure is not statistically valid and you have to look elsewhere?

Agreed. The only use of such a statistic is in answering the question "is it real?" (as opposed to being a statistical fluctuation). Once that is answered with a "yes" (anything above 4 or 5 SDs or so) it's not a useful statistic for anything else.

Using units of SD rather than percentage, proportion or absolute numbers fine for comparing size but yes, not enough historical data so unknown distribution  and can't compare probabilities.

Long ago I was calculating probability  of crack propagation in optical fibres. The theory had log of log so if an uncertain parameter was 1 out the result changed by 10**(10**1) - that was meaningless! 

There seems to be a lot of unknowns about rates of spread, death rates etc. We have a somewhat unique situation going on here in Saskatchewan which might be worth watching and perhaps modeling for some better info.

Our Northern communities are quite remote and isolated with only one way in and one way out. A couple of weeks ago Saskatchewan was down to 1 or zero cases per day in a population of about 1.2 million with only 4 deaths in total.

About 3 weeks ago an infected oil worker from one of the camps in Alberta went home for Easter. Home is a remote northern community in Saskatchewan of about 2,800 people called La Loche. They now have 138 active cases and add about 20 - 30 new cases per day while the balance of Saskatchewan adds about 1 per day. This is a population that likely did not practice social distancing. They have had 2 Covid related deaths.

They will likely add a much higher rate of active cases per day because there is now a team of health care workers who are up there testing every body.

The data coming out of La Loche might become very useful.

https://globalnews.ca/news/6912579/covid-19-outbreak-saskatchewan-dr-theres...

Interestingly, France is now reporting probable cases dating as far back as November - based on diagnostic images and swabs tested retrospectively.

Is it possible that the virus did not, in fact, originate in Wuhan?  How do we know it did?  Did we just jump to that conclusion because that was where the first "cluster" of cases was reported?  

I'm not suggesting it did not originate there - just wondering why we aren't starting to ask questions about the origins now we know that there may have been much earlier infections in Europe - and probably elsewhere....

And Sweden, according to its chief epidemiologist,, and it was reported that the US military was keeping its eyes on coronavirus in November in case US troops in SE Asia were under threat.  It has also been reported that samples from deaths in November will be checked in the US to see if it was circulating then.