UKH

See previous threads for descriptions on plots and methods [1]

For the second week in a row, there's no new estimate of daily infection rates from the ONS.   They're now suggesting there won't be one until January.

• Plot 1 - this is still useful for eyeballing the IFR by comparing the respective peaks but it is not an accurate representation of our situation, as it does not include any estimates of infection rate since cases started rising.
• Plot 3 - IFR estimates are rising for all reasonable lags from infection to death.  I think this is partly a demographic shift in cases and (plot 16 detail 1) and partly because the interpolation of ONS data points I use is anchored to the last data point, where-as the missing data for more recent weeks would have the curve rising up more - meaning more infections and a lower IFR.
• I will drop plots 1-3 until after this week until more ONS data becomes available.
• Plot 5 - this helps understand the exceptionally large "by specimen date" number from yesterday as a Monday's draining of the weekend backlog.  Still, the numbers are rising fast.  I've included a couple of days of recent data that are still provisional and are likely to rise more; the trend line does not fit to those.

[1] https://www.ukhillwalking.com/forums/off_belay/friday_night_covid_plotting__3-...

• Plots 6,7,8 - The rise in cases is now highly established and has clearly translated into hospitalisations.  
• Plot 9 - the new look characteristic times plot.  The exponential rate on the left axis is a bit esoteric for most people but hopefully the "growing/flat/falling" labels give an intuitive understanding.  
  • The right-most 7 days or so of this plot are provisional
  • It looks like hospitalisations haven't reached their final doubling time yet, and deaths are yet to tip over in to growth, but I expect that to be established by next week.
• Plot 9e - the key feature remains from last week's plot - the as-yet unexplained failure of lockdown setting in half way through.

A breakdown of chases and their characteristic times by country.  

• Plot 9x - by request from user Wicamoi -
  • Comparative data for the different nations.  Wicaomi has kindly put together a scoring system for Scotland to compare with lockdown measures for England.  I'm looking at ways of making a plot enhanced by this data.  Haven't got it right yet I'm afraid.  
  • I think the exceptionally bad doubling time of 5 days for Wales around 08-29 may be a glitch in the data.
• There's a general correlation in the characteristic times between the nations except at certain points.  Whilst the cases for Wales are clearly exponential it looks like the doubling time has settled down at around 10 days and isn't continuing to get worse.  
• I hope the doubling time for England levels of shortly at the same level...
• Things are better in Scotland and Norther Ireland but the trend in their doubling times is in the bad direction.

Plot 16

• This shows the case rate per 100,000 people that UTLAs were at in England when the tiering level was decided, and where they are now - which is slightly provisional.
• This plot is getting quite crowded and needs a better annotation layout...
• This has very much continued the trend since the plot first appeared 2 weeks ago.

Plot 16 detail 1

• Demographic breakdown for all London UTLAs to have exceeded their level on November 17th on plot 16.  The left heat map has been de-weekended but is otherwise unfiltered.  The line plots have an SG filter applied as well.  (We're much better at seeing trends in a noisy heat map than on noisy, overlaid line curves). 
• This is both fascinating and really a very unwelcome development.   The high case rates in October/early November were predominantly in ages 25-35.   Now, the cases are spread over a much broader age range going both down and up in years.  Given the apparently exponential relationship between age and seriousness of effects from the virus, this means we're likely see significantly more hospitalisations in London than one would expect from just looking at historic high-level data on cases and hospitalisations.  This is the "demographic shift" I mention in my first post.

Plot 17- A mesoscale analysis plotting the cases / day per English region.  

• There's a scale bar for cases/day.  
• The regions London, East of England and South east are coloured "red", the rest are "blue".  Last week "red" regions were the only ones with recent growth.  This week, all of the "blue" regions except the South West are showing growth as well

Plot 18 - the characteristic times for the "red" and "blue" regions.

• Something changed that broke lockdown
• It changed sooner and/or worse in the "red" regions - the bifurcation of behaviours is really quite clear from these plots
• The most recent 2-3 days on this are quite provisional and may change by next week

Plots of when each region tipped over in to growth of cases (R>1).

Anything dated in December is provisional and is likely to change by next week - this is because the data at single UTLA level is quite noisy, so filtering has to be used over a couple of weeks of data to measure rate's of change well, and data yet to come can influence that measurement.

I've scatter plotted the dates UTLAs tip over in to growth vs distance from Medway - chosen partly for convenience given its location between Essex and Kent. 

• Correlation does not imply causation, but it sure as hell piques curiosity.

Wintertree, you are a legend. I learn more on Friday night than hours of consuming news in the week. Thank you

As you've implied, the failure halfway through lockdown, and R going above 1 in the SE mid November worryingly point towards this new strain. I really hope it's not, but if it is then I don't see how they can possibly justify the current Christmas relaxation, and really things need to be tightened immediately.

At this point, the underlying cause is secondary to the effect.  Cases are increasing and the doubling time is increasing; hospitals are almost full and there’s a demographic shift in a bad direction.  I wouldn’t be surprised if there’s a last minute lockdown in the next week; Wales seems to be running about 10 days ahead of England and I’ve seen no good and evidenced explanation put forwards from leadership for why it’s going so badly wrong.  

Re: the latest set of mutations, it presents itself at just the right time to explain a lot of weirdness.  I’m always skeptical of accepting an obvious explanation to what in reality is a complicated set of interconnected problems, but present itself it does.  Just before the first lockdown I had this to say as one of the reasons I saw loosing control of cases as dangerous beyond the economic and human arguments:

This virus could be a relatively benign precursor for something worse.  The fewer people it infects, the lower the probability of that mutation occurring.

Lets hope I’m wrong.  There’s a middle ground where the mutation has nullified critical parts of the immune response and is re-energising the network of people most likely to transmit the virus but is otherwise not more infectious.

One other contender for failure of lockdown is the approval of the vaccine being widely misunderstood and weakening mass individual commitment to the rules; I don’t understand why that would produce a geographic distribution over time, and the bifurcation in characteristic times happened before this news broke.

https://www.ukhillwalking.com/forums/off_belay/is_it_worth_it-717284?v=1#x9154...

Thanks; I learn plenty putting these together.

With regards my 20:24 and 20:32 posts, all of the Y-axis labels give “Deaths” as the measure when the plots are for cases or hospitalisations.  The actual dataset field used is given in the text at the top of the plots, and the UKC photo captions are correct.

Its too late for me to fix them - hopefully this isn’t too confusing.  I’ve moved to a more unified code for making all the plots this week and it must have a bug in the y-axis legends....  

I'm sure most UKC'ers ignored this trivial mislabelling of axes!

I think we are all very grateful to you for analysing the data so thoroughly for us.

Two things stand out - the spreading out of the virus from the 20-30 year age group to the other age groups, and the mysterious geographical spreading out (of a new strain?) from the Medway area.

Comparing the UK data with other countries, our rate curves look more like those of The Netherlands than anywhere else. Common factor?

I’ve been a bit myopic on the data the last few weeks. I’ll have to spread out again.   The obvious factor was a new mutation via the mink farms in the Netherlands.  The obvious doesn’t make it the true explanation though.

I’ve not done more than superficial reading in to the Dutch variant and the commonality (or otherwise) of that and the new English variant.  It’s worth keeping in mind though that immunity is not binary any more than infection is.  It’s a shame I don’t have more time to spend on this...

As if on cue.. 

https://www.theguardian.com/world/2020/dec/18/boris-johnson-calls-crisis-me...

someone’s ahead of the curve!

as an aside, thanks for pulling this all together so professionally and openly (in terms of your methods). I must admit I was studiously avoiding these threads thinking they were about people plotting how and where to climb in lockdown!!

I clicked last night and was slightly shocked but delighted at what I found.. something about books and covers comes to mind.

;D

Thanks as ever for the work. That case growth rate mid lockdown is pretty frightening now it's confirmed to be real. Looks like the third wave is going to be horrific. If lockdown as we've know it can no longer contain growth we're back to 'stay at home' or quarter of a million plus excess dead in '21 Here' s hoping (but doubting) it's largely behavioural, that at least we have a little control over.

Jk

PM to give a public address at 4 pm.

https://www.bbc.co.uk/news/uk-55376727

Thanks again for all this.

Hertfordshire, our only new Tier 1 today,  has 23% case growth to yesterday (Friday). The government's xmas dinner gift is looking more like a breakfast for the dogs.

I had a look at the raw data for Herefordshire after your other posts and I think it’s largely in the noise right now if it’s growing or falling.  That kind of hinterland means R is about 1, which suggests restrictions shouldn’t be dropped in my view as there’s almost no space in hospitals, large areas are going exponential and it’s not worth the risk.

Now it transpires the government knew about the new variant when they put H. into tier 1, that judgement seems hard to understand - especially as I think it’s going to emerge that the new variant is suspected or evidenced as driving the current Welsh situation across the border.

I hope what is happening is what I predicted a month ago when the four nations decided to allow Christmas meetings i.e. that cancelling Christmas a couple of days before Christmas as an emergency response is a lot easier politically than cancelling it a month in advance.   Only a total maniac would look at the slope on that new infections curve for England and keep going with loosening restrictions for Christmas.

This 'London variant' virus is also a great way to explain/excuse cancelling Christmas.  Whether it is the actual issue or whether its just people's behaviour changing before Christmas explaining it with a scarier variant of the virus is politically convenient.

Well, we shall be out on a family walk for the briefing to watch the planetary conjunction.  I’ll read the news after the dust settles.

Whitty has confirmed to the WHO that this variant can transmit more easily....looks like its shit or bust with the Oxford vaccine. 

Strictly final tonight, hope they don’t cut it too fine with the typical delays they have on a Saturday. 

It's not just Wales being a problem neighbour...  as I said (in plotting #3) the strong growth next door in Forest of Dean is way above being due to noise. Its almost like they felt they needed to award someone a gold star for merit. It smacks of politics and failed scientific input. 

Well at least now we know why we were so wrong about the new tiers being good enough...

Things look bleak now. I can't really see the SE (edit, or Wales) getting out of lockdown (tier 4), or others areas staying out of it for more than a month or so if there is already 5% prevalence of this new variant everywhere (if it has risen from 28-66% in the SE in a few weeks). All hands to the vaccine pumps.

One thing is for sure: we may be totally f*cked with a new easily transmitted variant of Covid but Brexit will carry on.

Most likely long delays at ports will mean lorry drivers in situations where the virus can spread.   Then the sh*t will really hit the fan.   I am so fed up with being governed by complete morons.

I'm still bemused we haven't gone to lockdown given how scary this looks: a third of the population in Tier 4; 70% faster transmission for the new variant; 0.4 on the old variant R rate for similar prevention response; fast growth is already locked in for the next week. I think things will look so bad in the last week of December that lockdown may become inevitable.

Boris just said we will be done with this virus soon: plain irresponsible. I thought a week was a long time in politics. With supply issues we must be looking at early summer as a minimum for anything other than protection for the old and vulnerable. By the end of January we face  dominance of the new variant nationwide with these numbers and current geographic prevalence.

Don't pack your bags and run to beat Tier4!?... I wonder what mobile phone data will show tonight on the motorways out of London.

Yup.  I think it's everywhere given my 20:51 plots last night and their earlier version on the last thread.  

We were watching a jetliner climb out of Newcastle earlier on today on its way to Alicante, followed closely by one on its way to Antalya.  I wouldn't want to be a Brit abroad right now, the only sensible thing for other countries to do is to impose immediate hard quarantine on anyone who has arrived in the last month and close borders with us.  Given the state of much of Europe right now I don't think there's a rat in hell's chance of closing down this lineage though.   

The demographic shift from my plot 16 detail 1 hasn't been discussed yet in the news.

It was all working.  Now the spectre of the Red Queen looms large. 

It's a waiting game now - why is it more transmissive?  What control measures need to change?  What is the efficacy of each vaccine against this new strain?  

More immediately, as Offwidth says, a lot more cases are locked in already.  With the demographic shift in London and the current occupancy levels of hospital, this is looking worse than March I think.

Stay safe everyone.

It’s your plot 18 that persuades me

Yup.  I’m going to make a pair of maps at some point using the turning point and the zero-crossing times from the exponential rates behind that plot, broken down to UTLA level.

Not tonight though.  I think I’m going to open a nice bottle of whisky and sit out and enjoy the night sky.

0.4 on the R number is grim.....the spring lockdown reduced R to about 0.6 so a similar lockdown would only get R to 1...and we had schools shut on the last lockdown.

If schools dont close I don't honestly see how we avert a disaster in the next 2 months.. 

I haven't had to think about that Red Queen scenario since our last exchange about it, months ago. It scared me then, and it scares me now.

I thought I heard Vallance mention 23 changes in the proteins. My basic understanding of virology is poor, but for a change in transmissibility would there not have had to be a change in the spike protein, hence a potential effect on the vaccine?

And would we be able to expect the mechanism of this increased transmissibility to be identifiable in a way that would alter control measures? E.g. can we expect to find out if it's become more robust in aerosolized form, or more viable for contact transmission, or some other such definite mode - or is that sort of identification unlikely?

I'm not a virologist but from what I read, one or more of the 23 mutations are to the spike protein. However, apparently the vaccines target multiple elements of it so it is theoretically unlikely that a single variant would become resistant to them, especially without the vaccines already being in wide circulation and hence giving an obvious advantage to any variant that did have the required mutations to resist them.

However, they obviously need to check it.

This bmj has an easily digestible summary which is better than the press.

https://www.bmj.com/content/371/bmj.m4857

Via the website of the organisation who found this variant, I also found this website which provides live tracking of the different variants currently in the UK. It runs very slowly on my tablet and I don't have the expertise to understand what I'm looking at without a deal of research and time, but if someone else is able to have a look and summarise where the variant is currently known to be prevalent around the UK that would be really useful.

https://beta.microreact.org/project/4ebHf85NFurDYcuc8tysBe-cog-uk-2020-12-0...

Your first question - I think the good people at Porton Down and perhaps the LMB aren’t getting much vacation time right now.  

Your second question - I don’t know.  I suspect it’s changed the minimum dose needed to cause infections, so things like mask wearing levels, physical spacing and air turnover rates will need to change accordingly.  I don’t know how quantitative studies of this stuff are.  I assume the virus can be cultured and studies using cultured/perfused organs have come on a lot in the last decade; I can imagine some fascinating research but I’ve no idea what’s going on.

Well we had the cases increasing in the SE during lockdown 2 with schools open. If this is fully the effect of the new mutation rather than lockdown fatigue then this is really grim. 

I wasn't happy with the planned Christmas relaxation, but the reason for its cancellation is not what I wanted to hear.

It also puts paid to all those saying we should allow the virus to circulate and protect the vulnerable. Tolerating the virus circulating has now given it adaptations to make it more of a problem.

My only hope is that the government knew all along they couldn't relax at Christmas and that even tier 3 wasn't enough, they just needed something they could exaggerate as an issue. Hoping that the government is bending the truth isn't really a good place to be.

Thanks. You're right: that BMJ article is great. I had a look at the Microreact site, and the identifier codes they're using don't seem to have a legend, but I can work out that UK5 had been the dominant strain throughout, and that UK1219 has now become much more dominant. Whether that's the same as this VUI – 202012/01 strain is way beyond me.

Yes - it must be panic stations for those monitoring mutations, and I don't envy them.

I take your point about viral load - that would change all our current countermeasures. But I'll hope for a mutation that changes only one...

Our of interest what do you do for a living?

Forcing my optimists hat on, a mutation that made a larger fraction of people asymptomatic would also lead to wider spread given our reliance on symptomatic testing.   This would lead to a higher R and a faster doubling time, with a smaller fraction of infected - but absolute larger numbers - going on to become ill during the growth phase of the pandemic.

The range of interpretations is wide open right now.   

Varied.  Postdocs then lectureship in the physical sciences (not biology) but this year was a kick up the arse that it wasn’t the life for me, so I’m striking out in biotech as co-founder of a new business.  What we do isn’t applicable to vaccine technology or otherwise of use to covid studies so there’s no conflict of interest there with me interpreting the data IMO.  Other than that I want things back to normal, because this is a difficult time to be trying to grow a new business.  I believe that the better people understand what is going on, the sooner normality can return.

There was a blog post early on in the pandemic - I can’t find it now - from a scientist about how scientists could pitch in to help with the pandemic.  The author suggested not jumping in to the modelling as you can’t add anything and hundreds of people are doing likewise but suggested instead trying to explain and help others understand, so I took that to heart.  

As it stands I maintain I could add to the modelling by making the modellers take an honest look at implicit degrees of freedom / free parameters in their models and by training the non-modellers and policy makers in the limitations of modelling and how to use models to interpret the present and the limits to how they do or more pertinently don’t predict the future.  My submission to RAMP went nowhere...

With the new variant and a hard lockdown coming they should rethink the priorities for vaccine use.  It could be used to lower R and protect key workers that need to move about across regions and/or come into contact with many people during a lockdown.  Health workers and HGV drivers for starters.

Maybe they should also start using every vaccine candidate that has passed safety tests and can be manufactured to get faster coverage in the hardest hit regions and not wait for efficacy testing.

Ha, not dissimilar to me then.. funny old world. 
 

just a thought, does the data support producing a ‘plot18’ for deaths? It seems the rate constant turnover point is one of the earliest indicators of change. Could looking at the growth rate of deaths relative to cases give a sense of whether the new strain IFR is similar to the original one? I realise deaths lag so this may not be useful immediately, but next week...?

Sadly, the herd immunity proponents might found out what their crazy idea looks like in practice...

They might.   And if this government does not go to the EU and ask for an extension on Brexit for at least 6 months until the vaccine kicks in and it is summer they should be done for manslaughter.   It is criminally irresponsible to go ahead with no deal Brexit in these circumstances.

I agree that's a good idea, especially if focused on South East, London and East of England with a separate plot for the rest of England so you can see any difference in evolution of death rate between the two areas over the next 2-4 weeks from now. This should remove any effect on death rate of early vaccinations from the analysis, assuming that should be roughly common across the country (although obviously that is also of interest.) 

Let's keep these threads to covid data and discussion and not let them descend into argument about Brexit or Scotland please.

Sort off.  As there’s a broad statistical distribution of time from detection to death, sharp features in the cases plots are quite blurred in deaths - you can think of it as convolution with a broad kernel.  (Although the distribution is always changing and isn’t published despite the NHS presumably having all the longitudinal data, and given how useful this information would be for interpreting the situation).  There’s also - thankfully - orders of magnitude fewer people in deaths than cases which means there’s a lot more random Poisson noise, so when broken down to regional or UTLA level, more filtering will be needed to get a reasonable curve for interpretation, blurring out features more.

The bigger question is where can I get daily regional deaths from?  I recall seeing some data on hospitalisations on a spreadsheet from the NHS but not for deaths.  Any pointers and I can try...

Thanks for all the work you have done in analysing and presenting the data. This is so useful and informative.

Thanks!

I’ve only seen weekly death stats. To be fair hospitalisation might be better anyway. Less lag, less dispersion, better power (higher absolute numbers.. I’ll go data hunting later

This page:

https://coronavirus.data.gov.uk/details/deaths?areaType=utla&areaName=L...
 

has the data on a regional basis broken down to daily cases, deaths and hospitalisations..

but I don’t know whether you can query it in an automated way?

edit: there’s a page about downloads which gives permanent links you can just run periodically or if you’re more dev savvy than me there’s a whole API documentation guide(!)

I presume you've seen the downloads from:

https://www.england.nhs.uk/statistics/statistical-work-areas/covid-19-daily...

and

https://www.england.nhs.uk/statistics/statistical-work-areas/covid-19-hospi...

These basically split the daily hospital data into "by region" and "by hospital". As of today there's "deaths" data up to 18/12/2020 but from 14/12/2020 is "likely to change". For the hospital activity it cuts off slightly earlier (varies for the different tables) but there's no indication that the latest data is incomplete.

One of my quick "have I got it right posts"...

From looking at your plots, it appears to me that in London & the SE & E lockdown "failed" about halfway through lockdown-2 in that in those areas things started going up again when it was expected that lockdown-2 would be sufficient. In the other areas, "failure" didn't occur until we came out of lockdown 2 into the new tiers. The most likely cause of this is the new strain which appears to have originated from the Medway area and so has hit London & SE & E before the other regions.

Any panic fleeing from London is of course going to wonderfully help the spread of the new variant.

Oh f**k here we go again.

You appear to have the right end of the shitty stick.

Thanks both for the links.  The dashboard one won’t be much work for regional level.  I’d seen the hospital data ages ago but lost the link.  I’ll see what I can do for the next update...

You can't appreciate or plan for Covid without considering Brexit.   It's pointless to look at historical Covid data and plan for the future based on it while ignoring a second absolutely massive shock to the country's critical systems which is about to happen on January 1.   It is equally pointless to plan for Brexit without considering the latest variant of Covid.

It looks like we are going to have hundreds or thousands of trucks queued up in Kent.  Those lorry drivers are not going to have proper facilities and if Covid starts to spread among them and large numbers of drivers get sick or forced to isolate the whole Brexit transport train wreck is going to be much worse.

And that's assuming the EU countries don't impose quarantine measures on the UK to stop this new variant of Covid.

Anyone looking at the data dispassionately would say f*ck this Brexit can't happen on January 1.  We can't stop Covid but we can stop Brexit or at least postpone it until the summer months and after a reasonable level of vaccination.  Stopping or postponing Brexit is a necessary public health response in exactly the same way as a lockdown or a change in policy on use of limited vaccine supplies.

That’s my take.  Lockdown worked everywhere then failed in area after area, with that effect spreading out with time from the Thames Estuary.   Where it’s established before lockdown, cases remained exponential throughout as we now also see in wales.  It’s already everywhere but hasn’t grown enough to drive cases in many regions yet but I think it’s coming, and every person leaving London risks accelerating that process.  

That’s as maybe but this thread is about plotting and understanding the data, not discussing political ramifications or policy. Start another thread?

Looking again at your case plots for the uk nations, is it possible this variant originated in South Wales and was seeded to Medway region early on?

otherwise how do you explain the failure in wales happening so hard and fast?

Interesting observation; I haven't spent much time looking at the other nations - I only split them out this week.  I am working on a plot with the various control measures shown per-nation - or will be when I get around to it...   So I don't have an opinion for now.

Wales had an earlier, shorter (19 day) lockdown than England that ended on November 9th. We then went back into relative openness despite cases still being high. You can see the cases taking off again shortly afterwards.

Reports saying the new strain or something similar has been found in Australia. If correct, that’s extraordinary given that Australia has proper quarantine on entry and relatively few cases (now growing). Of course this could be coincidental similar mutations but seems a bit too much of a coincidence. Let’s just hope it’s not vaccine resistant.  

Ah, thanks. I didn’t realise that.. doesn’t preclude the hypothesis but certainly changes the analysis.

NERVTAG report out....they suggest this new varient increases the R value by between 0.39 and 0.93...omg.

Not happy reading.  

One confirmed export to Australia 

Data on reinfection and immunity will take another week, as will data on the demographics of infection with the new variant.  I think plot 16 detail 1 above is a strong hint of where to come.

35928 new cases.....this could get very ugly, very quickly.....I am seriously worried. 

36k cases reported today. That’s on a Sunday, which usually has relatively low numbers. Twice as many as last Sunday. F*cking hell!!!

Could be another case of finding a bunch of positive results under the sofa but I doubt it. Looking at the specimen date data, this week’s data has been consistently on steroids.

Sorry what do you mean by data on reinfection and immunity?

Worrying Tier 2 areas above 100 cases per 100,000 (and red growth on 'On the Rise' on the Guardian map):

Area; cases per 100,000; weekly increase

Adur; 128; 164%

Arun; 139; 100%

Brighton & Hove 141; 112%

Bromsgrove 236; 112%

Cherwell; 201; 121%

Cotswold 103; 132%

Eastbourne; 188; 138%

Huntington 108; 127%

Lewes 156; 210%

Mid Suffolk 123; 146%

S. Cambridge 150; 143%

Uttlesford 227; 125%

Winchester 102; 130%

Wychavon; 146  ;136%

As a comparison in Tier 3 of cities under serious debate:

Leeds 142; 2%

Manchester is 177; 0%

I assume from their wording they mean data on how many people wort the new variant are re-infections of people who had an old variant.

Immunity - the minutes refer to lab tests of immunity - in a lab setting how does vaccine efficacy change with this variant.

Even in T2 areas without clear growth the recent news makes to me a clear cut case everything needs to be T3 minimum right now until the experts understand what the hell we’re dealing with.

I'm sure the government scientists know this. Given the London exodus last night the government have the perfect excuse (and revenge).

I don't think it can come too soon.  The case counts developing in the provisional window of the government UK dashboard are showing a staggering amount of growth - and they're not final yet. 

At risk of being pedantic, I've said it before but it didn't start in Medway. Cases in Kent took off first in Swale (Isle of Sheppey) then Margate (Thanet) and Medway a week or two behind. Medway overtook Swale at some point although recently Swale had been higher again at times.

These areas are similar sizes geographically but because Medway is its own LA, it stands out on the covid infection maps (Swale and Thanet are always reported as part of Kent at utla level.)

Cases actually started increasing fast in Basildon and parts of East London in the very first week of October, a week or two before Swale.  Medway overtook them during lockdown although interestingly like Medway, neither Basildon nor Redbridge (as examples of South Essex and East London respectively) saw a drop in cases at any point through lockdown (Swale did.)

This is all a bit of a ramble, but basically I'm trying to say if you are looking for the source, don't just look at Medway.

In reality because of the above and because of geography, more cases have probably spread to elsewhere from Essex than from Kent, and far more from East London.

Sorry; I didn't help by picking it out of convince for my plot.  I might switch to a location in the Thames estuary, say 51.4595°N, 0.7853°E.  That seems appropriate.

Regardless of any Brexit politics I think you're spot on with this one.

It's basic common sense, deal with big unknowns one at a time rather than all together.

(pedantically, you mean lengthening the transition period since we've already left but I think everyone'll know what you mean)

At the moment this is really looking like the worst case scenario isn't it? I haven't been this concerned since March. Even the best case scenario  of R 0.39 higher we'd still have R about 1 in a full lockdown and we're getting 2000 admissions per day. R 0.93 higher doesn't even bear thinking about - this will be an Exercise Cygnus scenario.

Hopefully once more data comes in the picture will improve. 

I wonder if there’s a hint (and only a hint) that hospitalisations are not quite tracking cases in the very latest data (allowing for lag)..  so there’s maybe a possibility that this variant is slightly less dangerous? But that’s strangling the life out of the last few days of data and it’s statistically laughable in terms of significance I’m sure.

"Even in T2 areas without clear growth the recent news makes to me a clear cut case everything needs to be T3 minimum right now until the experts understand what the hell we’re dealing with."

I'd say everywhere in England aside from Scilly and The Isle of Wight needs to be Tier 4 minimum now (as Cornwall is also red growth zone) better still Lockdown 1 restrictions..  Next week's growth is built in so those Tier 2 places I list above will be mostly 250 to 400 cases per hundred thousand in a week (and growing), even without a xmas mixing effect. It's seriously bad luck for the UK this is happening now with Brexit and xmas/new year reducing focus on response. Most people wont realise the shit storm that's almost certainly coming until after xmas. 

Spreading more quickly amongst younger people initially? I'm now thinking everyone should have been put in tier 4, it might have reduced the exodus from London 

If the r rate of the new strain really is 0.4 higher than the original one we’ll probably start to see field hospitals prepared in the first few weeks of January.

There would be no way to stop it.

Ah didn’t realise they were looking at that, which makes sense. Thanks. 

This NERVTAG statement is very weird. They have announced a range (95% CI?) of 0.39-0.93 on the variant specific increase in R, but no central estimate, and then the publically released figure (R+0.4) is virtually the lower bound.. seems odd to me. Even with a highly skewed CI (which I wouldn’t expect a priori) the most likely value must be higher than 0.4 surely?

It is weird for sure....I have no idea why 0.4 was quoted and yes the mid range would suggest 0.66....if it is that we are screwed unless we can get this Oxford vaccine rolled out en masse very quickly... It could literally be a race against time. 

is there any context to this figure? I'm assuming it means an increase from about 1 to 1.4? If it was from 0.6 to 1 it would be even more concerning. Any chance it could be a estimate if no measures are in place, 3 to 3.4 doesn't seem so bad.

That’s a good point about +0.4 meaning R=1 even with Lockdown 1 measures and hence maintaining 2k admissions a day. Treatment ha s improved but that’s still a lot of deaths and hospitals beyond capacity soon enough. Suspect it might need to be a harder lockdown... unless as you say with further data it turns out to be less contagious after all. 

Mortuary space would be an issue. Grim.

If this tweet is accurate I don’t understand why the panic button wasn’t pressed 12 days ago.

https://mobile.twitter.com/The_Soup_Dragon/status/1340349639946629120

Shiiiiiitttttt... I’m guessing MK has a pretty wide catchment for swabs and that indicated most are now new variant. Right?

Have you tried to do any modelling with the new strain looking at growth assumptions in tier 3 (based on the new tier 4 areas), the current rates of the new strain in the different areas and different vaccine rollout scenarios (eg x people vaccinated a week)? Would it even be possible? It would be interesting to see if we really are a screwed as we seem to be.

I don’t really have much faith in modelling.  It’s so riddled with unknowns it’s not usefully predictive.

I’ll stick to extrapolating current rates and trends 1-2 weeks in to the future.  Right now my peril sensitive sunglasses won’t let me look at the results.  Some number mashing I did a couple of weeks ago - done independently by another poster - suggests the Pfizer vaccine alone isn’t enough to allow us to drop all controls if used in age decreasing ranked order.  So if this is as bad as we all seem to be assuming, the Pfizer vaccine isn’t a magic fix - although it makes things a lot less bad.  Not that we have any more doses of it to hand right now.

I think the AZ vaccine is critical now - production and distribution of this are more scalable I think.  I hope we’ll see it being deployed by Jan 1st targetted at those most likely to catch and spread the virus , and offering any unused capacity at random to people aged above 45.

I’m clinging to hope this variant isn’t more transmissible per se but has reduced co-immunity with old variants and is re-energising the network of people most likely to catch and spread the virus; after that it’ll just be the same old same old for control measures.  Actual microbiologists I’ve spoken with don’t consider this likely, but all epitopes are not equal.  I’m probably wrong

Being cynical, they were aware of it, but it wasn't definite enough and as parliament was still sitting there was no way they'd be able to get the "Cancel Christmas Bill" through?

That’s my reading of it.  

I hope you ARE wrong on that. Reduced cross immunity with the strains the vaccines are based on is exactly what we don’t want. I realise natural immunity is monochromatic where vaccines aren’t, but still.

I think the vaccine can be updated faster than we can wrap up a more transmissible variant.  If it is just reduced cross immunity the W exponential case growth will slow half way through the high-transmission-risk sub populations then we know exactly what to do and how to do it whilst the vaccines are reworked.  Gods alone know which is actually worse.  I don’t.

Worst case it is both more transmissible and has less cross-immunity.  I think we can agree to hope it isn’t that.

Judging from what Chris Whitty said a few weeks ago the mechanics of changing the vaccine would be relatively straightforward. Producing it in any significant volumes would be much more challenging.

Given the ONS projections around what proportion of the population have already had the virus and can therefore be assumed to have some immunity, it doesn’t really stack up (in my head anyway) that there’s less cross immunity to the new strain (I don’t think the current number of immune people are significant enough to slow the spread of either the original or new strain).

Any reduction in cross immunity would possibly reduce the effectiveness of a vaccine.

Let’s hope that it’s just more transmissible than the original strain.

Depends on what assumptions you make about the distribution of those who have had the virus - random or concentrated in specific sub-populations?  

We’ll find out soon enough.

I may be missing something here, I’m no vaccine expert, but it’s taken best part of 6 months to do phase 1-3 trials of these vaccines. Surely you have to do that again if you change them?

I’m no expert here, I’m just badly rephrasing what I saw on the telly (hence my reference to the clever chap who knows much more than I do).

Having said that is you are using RNA to get the body to build copies of the protein spike, all you need to do is work out what the new protein structure is and make a few tweaks to the RNA sequence and bobs your uncle.

Whether this would need full evaluation from the MHRA? Is anyone’s guess. I suspect it wouldn’t given how urgently we need it.

The flu vaccine changes every year - the method of making it remains the same but it targets different strains. I can't see how this would be any different.

And there are trials every year I believe. That’s why it’s based on the Southern Hemisphere strains from the previous 6 months and why it’s sometimes not that effective (flu mutates MUCH faster than CoVs)

France suspends all travel from the port of Dover.  France > U.K. travel not suspended but I can’t imagine many continental lorry drivers queuing up to get stranded over here.

https://www.bbc.co.uk/news/business-55389505

Flight bans are coming in too.

Eurotunnel are suspending services to France from 22:00 today.

https://www.eurotunnel.com/uk/travelling-with-us/latest/

NERVTAG minutes 18/12/20:

PCR ct values: which suggest a decrease of ct value of around 2 associated with the new variant. 

Is that good news amongst the bad? Edit - sorry, apparently not. 

Does anyone have any ideas why this new strain appears to be replacing the previous strain? I would have assumed the previous strain would continue rumbling on in the background, but it seems not. 

It means that the diagnostic test used in the lighthouse labs reaches its detection threshold on a sample sooner.  I’m filing that under “provisionally bad news” as more viral load in the infected seems the likely explanation - although it could have changed when people become symptomatic vs viral load in a good way I suppose.

It’s like “The Highlander”.  It’s a provisionally good sign if one is out competing the other because it suggests cross immunity, although in lockdown then in the T3 areas it had the selective advantage of R>1 vs R<1 for other variants.

I think it is just because the new variant is growing so fast that despite the old variant 'rumbling on' the numbers of the old variant become insignificant as a proportion of the total.

The lower bound is really really bad. At the upper end of that it's cataclysmic, I don't see how we contain that and feed people. I think the let er rip proponents just got their wish.

Jk

If that is the case we are clearly heading for a strict lockdown, hopefully asap. What did the March lockdown get R down to?

SS Richard Montgomery by any chance?

Jk

0.6 I think. That's the problem.

So in best case scenario we can just about contain it with that level of lockdown.........

Yes. This was mentioned above and is explained nicely on the front page of The Times today. Third one down here: https://www.bbc.co.uk/news/blogs-the-papers-5539029

The key point is at the end and I suspect not everyone gets it. The data are not definitive so we can wait and see till there’s better evidence and totally lose control or we can act now and at worst we wilL have suppressed the virus more than expected, which is no bad thing. 

One saving grace is schools are closed and students are back home. I think Gavin Williamson will have to have a long hard think about arrangements for the new term. Unfortunately I’m not convinced he’s capable of a long hard think...

The more I think about this the more terrifying it appears. Either we (nervtag or on this thread) are seriously overestimating this or the national/international media are yet to pick up on the severity. It’s been in and around London since early October. There’s no way it’s not all over Europe and the US (and a load of other places) by now.
 

The horse has bolted, as we will see globally within a month..

I think they it is being taken seriously internationally, look at the travel bans! The media need to sell papers or subscriptions and perhaps they think the headlines are so big already that spreading gloom will worsen their sales.

I would be very surprised if we aren't all put in lockdown post Christmas. I think the Govt will stop schools returning if necessary too. Although they have been very keen to support schools staying open (which I support) they have also shown they are not averse to a U-turn, which I also support when necessary) and this new variant gives them political cover to do almost anything.

I am currently frustrated because for the first time in a while we no longer have any public access I can find to the most important data for policy making: how far this variant has already spread in what quantity. We can infer from the last couple of days infection data that it is widespread south of the midlands but north of there is still an unknown. What we ideally need is the ability to see %infections that were the new variant by utla over time. This is now more important than the all-variants infection rate data (although they obviously need to be considered together, as well as hospital capacity) for the next couple of weeks, but I don't know how to find it unless it is hidden in that link I posted earlier. We have been spoilt by the covid dashboard in recent months but I now feel in the dark again.

Edit to add: I think someone above asked how many of the PCR positive tests are sequenced in the UK to determine the variant. According to Cog-uk coverage reports it is up to 20 %ish but averages more like 8%. Varies with both time and location.

Been doing some searching, haven't found what I was after but the following two articles from cog UK contain some useful information (probably higher level of interest for virologists but also some more accessible information mixed in):

https://virological.org/t/preliminary-genomic-characterisation-of-an-emerge...

https://bit.ly/34xHQxP

It seemed appropriate.

Impressive - the one point I thought the PM slipped in his press conference was his brief boast about sequencing, but maybe it had some justification.

I’m being speculative and maybe worse here but the EU are about to approve the Pfizer vaccine, we did it 2 weeks ago. Did someone here (at MHRA?) see this coming?

I wasn’t convinced the conference was the right pace for a jingoistic boast over our sequencing capability, judge the moment and all that.  If this strain turns out to be as bad news as it looks, I’d not want to be shouting about our capacity - as it means we knew what was going on at least two weeks before we did anything.  

New Zealand are sequencing 100% of infections but their absolute numbers have been small.  

At what point will they call this a national emergency?

Herefordshire up 72% on last week today. This is way above noise.

The Guardian On the rise UK map has gone from bad to grim in just a few days.

https://www.theguardian.com/world/2020/dec/18/covid-cases-and-deaths-today-...

Tier 1 Cornwall is now up 242% on the week

A selection of new Tier 2 red zones:

Cotswold up 211%

Eden up 197%;

Mid Suffolk 157%

Mid Sussex 142%

Rugby 133%

Test Valley 185%

W Suffolk 149%

No - FDA are ahead of EMA as well.

Yes, not good overlap between rising areas and T3/T4.

I think it’s all going to be academic soon.  No reason not to stall for time.

Odds on the Oxford/AstroZeneca vaccine getting emergency approval today?

AZ approval odds: high

Ps. I saw this quote and can’t agree with this guy more:

Dr Stephen Griffin, Associate Professor in the School of Medicine, University of Leeds, said:

“The new SARS2 variant may well be cause for significant concern, and all the more reason to be vigilant.  However, that it has taken an event of this nature to provoke a reaction from the government says more about the current scenario in the UK and the long term view of how this came to be, rather than the naturally upsetting and disappointing false promises over Christmas that have now been revoked.

“It is clear that failure to make good on the sacrifices made during the first lockdown, suppress infection rates over summer, and replace the failed corporate TTI system has led directly to the resurgence seen since September, the need for a second (and likely a third) lockdown, and ultimately an environment in which SARS-COV-2 is able to thrive and evolve.  Half-baked policies neither restore socio-economic harms, nor suppress infections, and the reactionary, fragmented and confusing implementation of Tiers along with their obscure criteria has led to public favour eroding day by day.  We are left at the mercy of misinformation and fringe viewpoints.

“Nevertheless, if it takes the new variant to finally convince our leaders to implement an effective suppression strategy along with the long-recommended safeguarding on TTI, quarantine and internal travel, then so be it.  We cannot simply stand by and wait for vaccines to rescue the situation, the human cost has already been far too high.”

I agree.

Back in March, Coel posted a thread eloquently arguing for the “let it rip” side.   He wasn’t actually advocating it, “just asking questions”, and misrepresenting some things and using his clear smarts to frame the argument to guide people to the way he felt politically.

https://www.ukhillwalking.com/forums/off_belay/is_it_worth_it-717284?v=1#x9154...

As well as arguing against Coel on the usuals grounds of the immediate consequences of this, I noted:

This virus could be a relatively benign precursor for something worse.  The fewer people it infects, the lower the probability of that mutation occurring.

The idea that running with significant prevalence “to save the economy” has been Russian roulette from the start.  

Coel got himself himself banned it seems, but he’s reading UKC via another account I think, so if you’re reading this Coel - go back to those early threads and consider how smart you are, and how you used those smarts to twist the perspective to push your political ideology rather than using them to digest as much science and historical evidence as you could, and so use those smarts to try and understand the problem and build up from there.  For shame.

No - it was his one bad moment in an otherwise very sober presentation.

NZ are in an enviable position, eh!

I keep asking my contact over there in public health to send Ashley over.  Oddly they won’t.

I live in Kent, M20 shut now. Is it too late or too early to panic buy?

Get a shotgun and raid the lorry park.

You must have seen operation stack before. 

Ignore it.

Not with the port shut for two days we haven’t!

I’m very close to the Shepherd & Neame brewery so I should be ok for beer!

Sorry but I ship regularly to Europe, its happened before.That is why they have Operation Stack. 72 hours not unknown if they have reaally bad weather in the Channel.

Anyway most freight to Europe had a cutoff date last week.You are seeing the tailend  so to speak.

Do you think their reasons for shutting the ports will be gone in 72 hours then?

They might be, depending on how understanding of the new strain progresses, and depending on what the Government puts in place e.g. testing on departure.

They don’t normally shut the tunnels in a storm.

We have not stopped in bound traffic.So at the moment its more a pain if you are an exporter, not an importer.Soemnthign people are tending to forget.

Fortunately freight trafiic was more chaotic last week pre Saturdays announcement, which is normal st this time of year as Europe and Uk shutdown for the christmas break.

My own view is that this strain will be just as widespread in Europe anyway. I think Germanys figures shot up over the last few weeks for example.So its li9ke closing the door after the horse has bolted anyway.

Barely carriers any freight traffic.Give it 24 to 48 hours to settle down.

You are the acknowledged expert on the spread.What do you think about my view its too late anyway and will already be in Europe.

Depends what happens at the borders and what that does to the markets, we probably won't know that definitively for a few weeks (and the impact will be blurred/compounded by brexit) but if the new strain can be contained in Britain (extremely unlikely seeing how competitive and established/prevalent it appears to be) we're screwed, that will be the moment the recovery trajectories of Britain and our peers diverge radically, borrowing prices soar, border disruption increases hardship and tensions, social and economic support dries up, all just as we need to enter the longest deepest lockdown yet or make the awful decision with our salvation receding back over the horizon to let it rip, ration covid care and accept hundreds of thousands of our elders dead. On the other hand if it has established itself overseas then everyone else is equally screwed, Covid20 will go on to utterly destroy populations and or economies across the globe through 2021, despite the vaccines this is now an order of magnitude worse than where we were this time last year and a direct consequence of our choice to run hot all year rather than pursue eradication on either of the occasion the opportunity presented.

 I suppose we should be thankful given the trajectory of 2020 that it's apparently only significantly more transmissible.

Fu*k, what a roller coaster of a year

jk

We've got a good track record on this.

A shame it was their minor front page story rather than the headline story - they could have headlined "Act now, before it's too late".

The responsible thing for the government (deliberate little "g") to do now would be to impose a strict 2-4 week lockdown NOW which would i) slow things down & ii) give time to make better assessments and judgements about the situation.

However because our revered and trusted leaders always seem to be behind the curve, I don't expect a lockdown until after Xmas by which time of course things will be much worse. ☹

And 3: cause much reduced economic damage because everything shuts down for Christmas anyway, and loads of people will be taking stacked-up holidays they didn't take earlier in the year.

Much as parents will hate it, school holidays are the perfect time for lockdowns because you get the added benefit of schools being closed without loss of educational time.

So yes, March-style lockdown from now to say mid Jan (i.e. close building sites and non-essential manufacturing as well, and keep the schools shut) would make a lot of sense.

Alot of manufacturing closed on Frriday because of stacked up holidays.So possibility that you maybe right.

I am personnally expecting us all to be back in national lockdown ( Tier 4) anyway in newxt couple of days.Reasonably predictable.

This isn't going to be 48 or 72Hrs. Any testing regime able to deliver sufficient confidence in France is going to take days to set up and will likely involve a hold and retest phase requiring huge amounts of space and logistics, getting and documenting tests to stranded trucks rather than visa versa (also food, medication and adequate sanitation), that means we start properly clearing our backlog right about January 1st, just as we introduce a new, unfinished/undefined border regime.

If we're not rationing food within a week I'll be astonished.

jk

I’m not the expert; the fact I called it 5 days before Boris’ announcement implies something else given how they have access to the sequencing information and I don’t.

I reckon it’s in all major destinations from the U.K. by now and will be spreading below many of their radars - many places are not sequencing as much as us so won’t identify it so often.

If this variant turns out to be as bad as a reasonable worst case scenario, imposing hard geographic borders buys other countries 2-3 weeks time in which to prepare I reckon, but the outcome seems inevitable by now. 

I think I would rather have alot more knowledge of logistics before coming to any rash concepts about what may or may not happen.You also seem to ignore that inbound is not affected.

Alot is containirised anyway ( not everythign comes through Dover- that is mainly roll on roll off- there is a hell of alot elseswhere form other European countries)

But considering there are what a few thousand lorries going through Dover a day it would be pretty easy to set up testing with lateral flow tests.The biggest issue will be who pays for it.

As regards people travellling overseas, well the general advice has been not to for some time, so its not really that much of an issue.( other than being a pain in the backside for those seeing family etc).

In a way the timing could not have been better, the previous weekend would have been more of a logistics headache.

This is not to say there will be issues.but a bit of balance for the doom sayers helps.

Any views on Europe.?I would have thought it was already spreading and too late.

Ignore just realised you said it was too late for Europe anyway.

I can’t see many companies sending lorries over knowing they won’t be able to get back.

Of course inbound is affected! Does mainland Europe have a limitless supply of disposable trucks and drivers?

It's not about logistics, it's politics. Obviously the new strain has escaped (we know that) and is almost certainly already well established in France but given the extraordinary severity of the new situation their government is not going to be rushed into accepting that situation and dropping their guard. If they do then what happens next is forever tied to that choice, not the preceding weeks of free flowing traffic and choices made here about what to share and when. It's going to take significant time to agree and implement a satisfactory and reassuring response, at a time of already high cross-Channel tension and with an already large backlog of trucks owing to the pre-brexit, pre-Christmas rush and the short notice border closure. Then brexit. I don't have to be in logistics to see this is a mess which can't be solved in two or three days.

And staples like grain are carried in bulk. Fresh food isn't though so very soon the shop shelves start looking patchy and the public react creating a positive feedback loop. The container ports are still struggling with out of position containers and PPE piles from wave one. As passenger air travel closes down again long haul air freight prices soar.

The biggest issue is the 50%+ false negative rate. That's not going to reassure anyone. Paying for testing is utterly trivial given the alternative.

I couldn't care less about holidays right now. Food, medicine, essential sanitation chemicals. If we can keep those moving and maintain confidence in our economy we're winning.

I hope you're right about this being a mere 'weekend' of disruption and me being alarmist, I'll be more than happy if you are.

jk

Its reported as having got to Australia so think it is safe to say everywhere else aside from maybe North Korea are buggered.

Seems like gesture politics shutting down now (although I would be sticking all air passengers into two weeks enforced quarantine as standard)

Each important event into a country accelerates growth ahead of the natural exponential curve.  Shutting borders will buy them useful time if this is really as bad as it could be.  Time to reflect on the UK where every train out of London was fully booked within hours of the PM's announcement on Sunday.

Any thoughts on why the Zoe data is less accurate; I thought it had been tracking the PHE data reasonably closely?

Can we infer anything about the new strain from this or is it just user fatigue leading to less users and therefore less accuracy?

If there was greater asymptomatic spread of the new strain as you posited above would that explain the inaccuracy? 

I use Zoe and have been looking at their reports every so often for a few months. I think one source of error must be the different density of users they have in different locations. Intuitively I would expect more affluent and younger-leaning areas would have higher app take up. Early in autumn it was noticeable on the map they publish of positive tests throughout the previous two weeks that there was always a relatively high density in London even when London reported rates were low, implying a high density of app users there. Equally, there are large areas of countryside that now have high rates which Zoe shows very few positive tests in, presumably because they have relatively few app users. They must correct for this effect in their reported data, but maybe they struggle to compensate accurately for it.

The second possible source of difference is that if you report symptoms in Zoe that are not the classic three covid symptoms (eg, sore throat or excess tiredness) you often get offered a test (I have been tested twice through this route.) Their research has shown a wide range of symptoms are associated with positive tests. So, for people using the app, you would expect Zoe to pick up a higher proportion of positive cases than the national testing programme does, including people with other symptoms. Again, they may correct for this effect, but it may be difficult to do accurately (or they may not.)

Thirdly they are obviously constrained by simple statistical uncertainty due to the lower number of people they have as participants. There are approx 4.5 million participants but I bet many of those installed the app and rarely used it.

Not many were going this week anyway. Put it this way I had machines being shipped to Poland and Greece for December. The cut off for shipping on road was the 16th( last Wednesday). If it was not on the road by then they were not going to be delivered this year.

Under Operation Stack its normal for lorries not to go (why on earth would you send them anyway if there is a stopage).

Just parcel shipped via UPS to Canada and Poland this morning. Both still on track to be delivered Wednesday.UPS and Fed_Ex etc have their own " air cargo "routes and planes.

Logistics is always fascinating.

Surprised you have not mentioned Brexit as well.

Keep a cool head. I have seen enough  logistics issues over the years _ stuff stuck in warehouses in Tornados. delays due to volcanoes in Iceland- to learn that the system is pretty resiliant  enough to cope.

Christmas/New Year is always excitingand a fun time..had a consignment  going to aerospace customer once that got stuck in a warehouse during this period due to backlogs.

Rampant Zombies are of course the one unknown.

I'm afraid I wrote ZOE off as potentially compromised when its PI, Tim Spector, re-tweeted something from one of the people using their scientific credentials to actively deny the existence of the pandemic. [1].  I’m not claiming anything is wrong with the data, or anything untoward is happening.  I simply do not trust the judgement of the PI and will invest no more time in it.

This wasn't an ambiguous re-tweet.  He shared something claiming "No pandemic" in a brief piece of text.

At that point I lost all interest in the ZOE data and am not going to spend any time looking in to it or thinking about it.    

The ZOE data and the gov.uk dashboard data on infections vs cases do not correlate well at all in terms of the last two months other than the direction of cases - up or down.

The gov.uk data on cases is translating in to hospitalisations.  This is a directly relevant measurable.  

[1] https://twitter.com/timspector/status/1326815908594151424

You can laugh at me to your heart's content in two weeks. Hopefully.

jk

It's perishable & specialist goods that will be the problem, not sending machines around the world. The supermarkets and other time-sensitive businesses want 'drive on drive off' to the distribution centres to save time; this relies on accompanied trailers, and if the cabs & drivers can't get back to Europe the next set of goods aren't going to come in a day or two later.  The only option for the hauliers and ferry companies is to move to 'unaccompanied trailers' where the trailers get left on the dock, stevedore cabs put them on the boat, then the reverse happens at the far end. It's how the Channel isles get virtually all their freight. The only problem is that it requires dockspace and increases un/loading times.

The problem with the vaccine is the public mentality - once someone's been vaccinated they become invincible in their mind and behave as normal. This allows the percentage of people for whom the vaccine is not effective to become superspreaders, and maybe even people for whom it is effective to become a low R spreader before their immune system can deal with it.*  People need to remember the vaccine isn't for personal protection, it's for herd protection and until sufficient people have been vaccinated (mid summer?) it doesn't really help much.

*How does a vaccine affect someone's response time to infection? An un-vaccinated person gets infected on day 1, then the virus builds up so that by about day 4 it then becomes spreadable until about day 10+ when the body kills it off. What happens in a vaccinated person? Does the body react so fast the levels of day 4 are never reached?  Or is 'day 4' level reached but 'day 10' actually arrives on day 5 so they're only a spreader for 1 day instead of 6?

Yep, definitely for the UK, look at the animation here https://twitter.com/bioinformAnt/status/1340751018913648641

A friend of mine went to the UK last week with a plan to travel to Germany on Friday to see the other person in their long distance relationship. Came down with appendicitis and had surgery on Friday or saturday (successful). Other half travels over to the UK from Germany Sunday or today, I'd suggest there's a fair chance both of them will be stuck for some time!

It's been quite amusing reading this thread after seeing the news last night and this morning - UKC is pretty good with its predictions!

My prediction - UK will do a proper hard lockdown after xmas with a view to elimination.  Anything less won't be enough to keep the nation out of tier 3/4, and the economy / society can't cope with that until the summer.  It was hard enough to 'live with the virus' before this new strain appeared, it'll be even worse now.  There's no point in trying to 'live with the virus' when the restrictions required to do so are so close to those required for elimination and the benefits of elimination are so massive.

It's clearly what's now needed, whether it's still possible is open to question. Perhaps with strategic use of the little vaccine we do have.

jk

World at One just mentioned another press conference today.  More restrictions incoming?  Can't find confirmation on the web...

If it increases the R by 0.4 forget about elimination......a March style lockdown would only get the R to 1......as per my earlier post it is the Oxford vaccine or bust.

Interesting relative lack of cases of the new variant showing in Wales in that animation. 

I had a plot of the spread up on the 15th. The moment the news broke of the strain everything odd about the data went “click”.

https://www.ukhillwalking.com/forums/the_pub/how_can_a_virus_be_faster-728968?...

I’m kicking myself for not looking at ways of plotting the public sequencing data now.

Suffice to say, in my view if the effect is visible in public data at this level, it’s too late.

Some perishables, but not all. After all Peruvian asparagus follows a different route......and so on. It does illustrate the complexities, but also shows the resiliance.

Unaccompanied is about 80% of freight.

It would also be interesting to see issues into/from other countries to gain some form of overall perspective.My understanding is that globally there are issues all over the place for seafreight because of reduced capacity.

Put it this way I know there are delays into New York seaport because of Covid issues slowing things down.But it just slows things.

Just illustrates how interconnected we are all.

I can't imagine that the increase to R would be a fixed addition on top of any level of behaviour. It must scale in some form. If it gives a 0.4 boost to an R of, say 1.2, maybe it's only a 0.2 boost on an R of 0.6.

Press conference for this arvo now confirmed on bbc news website...strap in everybody.  Announcement of T4 for whole country as of tonight?

Yes, factors are closer to multiplying than adding I think.

We can lockdown quite a bit harder than we did in March, it seems unthinkable even from here but frankly so's the alternative. We do also probably have the opportunity to cut firebreaks into society with a few million vaccine doses, that could be enough to maintain provision of essential goods and services while R dips under one. I think it'll be touch and go whether it's possible and it's going to be dreadfully painful finding out but it's too soon to say if it's impossible.

jk

As far as I can see (though it is not really clear), that increase of 0.4 is from the current level (and current level of restrictions), so it might be better seen as a 40% increase if R=1 at the moment (though clearly R is greater than 1, so percentage may be lower). The spring lockdown lowered R to about 0.6, so, if we have a similar lockdown with similar compliance, a 40% increase gives about R=0.85 which should contain it. Even the worst case 90% would give about R=1.15 which is bad but not horrendous. Maybe I am being optimistic, but with a really hard lockdown as soon as possible in conjunction with vaccination it might still be possible to avoid the apocalypse.

True but an increase of R by 0.4 makes elimination impossible when the prevalance is so high.....this is a race against time between the deployment of the Oxford Vaccine and this causing utter havoc....if that doesn't get approved it doesn't bare thinking about.

I don't agree. High prevalence among a number of other issues we've neglected and accumulated make this very difficult but I don't think it's impossible.

The beaten down optimist in me hopes today's press conference is about emergency approval of the AZ vaccine and a revised delivery strategy. More likely it's the starting gun for panic buying and subsequent rationing.

jk

Elimination would take until the summer anyway - remember how slow cases came down in March, even in countries with a harder lockdown than us.

That said, what we might be able to do is elimination *using the vaccine as well as a lockdown* - I wonder if it might be worth trialling that somewhere like the Isle of Wight - stop the ferry services, vaccinate everyone and see what happens?

I think it'll purely be about the border issue.

It won't be an absolute increase of R by 0.4, it doesn't work like that.  It'll be an increase of 0.4 in specific conditions, which would be increased if things were relaxed and reduced if they are tightened.

I think it is probably best described as "R is 0.4 higher than it otherwise would be in our present conditions".

I would agree with you that it is possible. However I think it would probably take too long. Wuhan and the surrounding province were in lockdown for a couple of months to eliminate the virus. It would take us substantially longer because of higher starting prevalence and a likely inability to lockdown quite as hard as they did (the state here does not have the logistical ability to take food to everyone's houses and people would see it as too great an infringement of liberty anyway to literally be under lock and key.) And finally the virus is now more infectious. So realistically from this point, achieving elimination will take (probably much) longer than vaccinating the vulnerable.

I hope the same.

Realistically now, people are going to continue being infected in significant numbers until the vaccine has been delivered to the vulnerable and the level of restrictions in the meantime become a balance between the deaths saved by ever more strict measures and the deaths or long term mental health impacts caused by inability to access healthcare or see other people, obtain care etc. Once you get into various levels of strict lockdown I think that balance becomes quite tricky, especially once a few million people over 80 have been vaccinated and the IFR drops.

A plot of the case numbers and their doubling times in Essex.  The right 3-4 days of the doubling time plot should be considered provisional.

This is the first time I'm really worried about the damage that this virus might cause at a national level.

If we have to go into lock down for any appreciable length of time it will be catastrophic for our economy.

If we have a lockdown that brings infection rates down to those of early August I guess you can estimate how quickly they'll rise again. I've got a horrible feeling it wouldn't be very long. I suspect public health officials in the UK and abroad have done that calculation and as a result we've cancelled Christmas and the rest of the world is putting us in quarantine.

Obviously we desperately need to vaccinate as many as quickly as possible. One wonders if other European countries could be persuaded that it is in their own self interest to divert some of their supplies of vaccine to the UK. That might have been an easier job if we were still EU members.

Yes, this is the first time I've felt genuinely scared. In March it all felt a bit unreal and there seemed to be a feeling it would all be over in a few months. It now feels very real and it looks like this wave with the new variant could make the Spring seem like child's play.

Yes.

Well it took four months last time and would be longer this time. Without vaccines really kicking in we are f*****.

No chance. They know this is probably coming their way too.

If the economy was on the verge of falling to pieces the £ would have collapsed completely by now. It has  moved but it has not collapsed..The £ against the $ is the indicator and that is trading at a reasonable rate.

The economy of the hospitality sector is shot to pieces but the rest of it will continue to be OK.

The French are basically saying we need to reopen the borders with a protocol. It is just agreeing the protocol that is the issue.

It just looks like the damage to the economy could be twice what it already has been. That is pretty bad.

The March lockdown's impact was distorted by mistakes like clearing wards without testing patients sent to homes. We can only really estimate spring cases from deaths but we know cases were actually boosted by pandemic preparations in the most vulnerable groups. We don't have to keep making the same mistakes and while we face grave new challenges we do have new tools and knowledge. I'm not feeling optimistic but all is not yet lost, I think while the financial system remains relatively stable we can still walk a very hard narrow track out of this mess.

jk

Well no - actually that was a reassuring "Don't worry everyone, we're not going to run out of food and as far as we know the recent mutation will not affect the vaccine."  

So why am I so nervous?

Because some people feel vulnerable. 

Yes. I think the press conference could be summed up as nervous, shit-scared fudge. 

I'm quite nervous too - can I join the club? 

As ever, thanks WT. 

I'm going to spread my bets. Full lockdown on the 30th or the 2nd. Bloody terrifying reading this thread. At least the nights are drawing out. 

I’ll get off the fence. Emergency press conference, Boxing Day, 6pm. Lockdown.

Even Boris doesn't believe his bollox anymore....

Vallance basically saying we'll all be going Tier 4 soon enough.

nah Johnson will still be pissed then. 27th instead.

The way the reporting lag and the "weekend effect"  in cases by specimen date works, assuming actual growth continues on the same exponential as now, the next couple of days of data release won't look too bad (relatively speaking), then the releases on Christmas Eve and Christmas Day are going to look atrocious as the weekend effect clears through the reporting lag.   So I expect the media will be pushing the scary numbers by Christmas Day setting the scene for a press conference on Boxing Day.

The outlook for the health services is grave.  I wouldn't wait for tougher restrictions, and the mass wave of border closures provides 40 nations worth of external judgement for the government to use to re-enforce their messaging.

Well if it’s 170 or so trucks stuck in Dover then it really is not an economic catastrophe. Pain the backside if you are one of the drivers. 

Fair. Although he’d probably just get hand cock to announce it..

I was thinking of several months hard lockdown.

Even in Aus which had a  hard lockdown people were allowed to go to work or wfh. 

The lockdown we had last Spring had a massive economic effect. Why would another be any different? 

A mid-week plotting update.

The last Plot 16 I gave at the start of the thread was for Dec 9th (as correctly given in its caption).  

This was  an oversight on my part as I had enough data to do it to Dec 12th or 13th at the time.  Now I have data good up up to Dec 15th.  So, these plots represent 6 days growth over the one near the top of the thread.

This is from several days before people packed every train - and presumably many cars and buses - leaving London for everywhere else in Britain.

The demographic split from November may be re-appearing in the new exponential phase with those aged 20-40 rising ahead of younger and older ages.

Also a plot 18 update

Any chance of getting Plot 16 into a moving day-by-day graphic, with a nice slider that we can all move backwards and forwards in time (and then go "oh shit this doesn't look good").

P.S. Thanks for link to blogpost

A request. Is it possible to include age bands beyond 40-50 ( eg. in Plot 16 - detail 2)?

Thank you for your amazing work - it's hugely appreciated.

This article is very well researched: https://unherd.com/2020/12/how-dangerous-is-the-covid-mutation/

The author suggests that this mutation is already in many other countries, it's just that the UK has very good DNA sequencing resources and may well have spotted this mutation before anyone else. 

I can do a movie that can be slid though.  It'd be great if I could add movies to these posts...

In reply to largeruk:

I can do more - the problem is it starts to get crowded.  The lazy 5-second way to do it is to make two copies of the plot; I've left 40-50 in both as an overlap between them...  It's easy to mis-interpret though because this is absolute numbers and not normalised to population - folding that in with age bands is a bit much for the time I've got for this.  What I'll do is make a doubling time plot for each age bin as a way of visualising the spread mechanic separate from population numbers.

Nah. The first rule of Boris Covid Club is you add 7-14 days to the date you think sensible. 

That too. It would take longer to bring infections down but presumably once restrictions are lifted they'll rise more quickly. We lose the window between lockdowns.

Thanks - I skimmed it and will read in full later.  An importation event would explain the unusually large number of changes to the genome - different genetic drift happened gradually somewhere else then it was imported...

It did occur to me, without wishing to sound too much like a rabid xenophobe, that the spread pattern did seem to suggest the potential for it to have been imported from the continent, or more specifically France?

It would explain why Portsmouth has also become a hotspot without the surrounding areas of Hampshire and West Sussex having such high incidences?

Plot 16 is mindboggling. London and the Home Counties have really taken off. T3 areas trending up but it's just a matter of time. Even the T1 areas are growing.

33k today. 30k tomorrow, 40k Wednesday, 50k Thursday?

BoJo won't want to spoil Xmas. Lockdown press conference on Boxing Day?

He did already for a third of the population.

Of course the Tier 1 areas are increasing - Tier 1 isn't enough to suppress it, we already knew that.

What I meant was he wouldn't want to spoil Xmas even more!

Yes but if you looked at the T1 areas they seemed to continue to drop or at least hold steady even after L2 ended. Was just a matter of time.

Your March lockdown was about 20% on the strictness scale. If you do a proper lockdown the rate of decline will be as fast as the rate of rise. As seen time and time again, if you're going to lockdown you've got to go hard and go for elimination, anything less is a waste of time unless you're just trying to buy breathing space to fix a temporary problem that, once fixed,  will permanently keep R under 1.  As seen now, to keep R <1 you need tier 3 or 4 until easter as there doesn't seem to be any way to improve T&T or means of vaccinating enough people to have a herd immunity effect before then.

Good summary of quotes from PHE and others.  Also gathers much of the speculation from academics that’ve spoken to various press.

https://edition.cnn.com/2020/12/21/uk/coronavirus-variant-uk-intl/index.htm...

A couple of key points below.  Also some oddly specific speciation with regards a super spreading event and a rave...

• "The increase in cases linked to the new variant first came to light in late November when PHE was investigating why infection rates in Kent [in southeast England] were not falling despite national restrictions. We then discovered a cluster linked to this variant spreading rapidly into London and Essex," PHE said.
• The variant has already spread globally. As well as the UK, the variant has also been detected in Denmark, Belgium, the Netherlands and Australia, according to the WHO.

Yup, and there’s another 5 days of growth before the plot reaches the Tier 4 date.

What happens after that is going to depend on how much spreading via children is a component of this new strain.  With Tier 4 and the end of school term, the growth might rapidly stop.  

Alot of businesses just shut down as there was confusion over what should/should not stay open. Alot of businesses have now become more Covid resiliant.

I agree that if you are in Hospitality or airlines etc its not good. But there are still huge swathes of the economy working very well- ..try getting a builder for example.Fabricators- still a shortage of welders in fabrication etc etc.

And in retail there is a huge shakeout due to move online which was going to happen anyway.

The position varies but its not all bad in the economy.

That I think is a resonable assumption.

But it's not is it, it's far more plus the inability of incoming trucks to return. As usual with Johnson: mouth moving, pointless lies spilling out with poorly prepared waffle.

jk

I was wondering about Germany. I know correlation isnt causation but the "Germany handled the first wave well but for some reason has lost control" does seem to fit the pattern.

Does anyone know how accurate Johnsons splutterings about the UK doing more sequencing than anyone else is.

Thanks for digging out the pages - I'd rather astoundingly missed that I could click on "United Kingdom" and that (now?) gets a menu down to UTLA level...

The API is the way to do it.  I looked at this some time ago and didn't use it as it couldn't do what I wanted - the number by reporting date as reported on a specific past date.  That was when I was doing plots of reporting lag.  It's just what I need now though, thanks.  In particular, I should be able to make CFR plots and doubling time plots for different areas.

So now its at 1500 trucks, which is still way below the normal number for Operation Stack( where you easily get 10,000 or so).

It's accurate as its backed up by the scientists on twitter. We are world class with sequencing and piss poor on taking the information we get from that  seriously enough. Given the data it's obvious things are about to explode where restricttions are low, which must relate to the spread of the new mutation from London. Most Tier 2 areas closest to London are doubling in a week.

Tier 1 Cornwall up 243% on the week, Isle of Wight up 343%, Herefordshire up 74%.

Assuming people largely abide by the T4 rules over Christmas. If not then even in T4 we have another 2-3 weeks of infection chains within households to add to the pre tier 4 circulation. If it is super infectious that may of course all come at once around Jan' 1st with everyone who was around the dinner table getting sick from just one person rather than longer linear chains as it ripples through locked down households. We've probably also got a big boxing day rate bump coming now outside of the SE from the hasty exodus pre T4.

I really hope there is a viable plan to staff and provision the Nightingales, until now they've looked like an empty gesture. In a month they're going to be filling fast. Whatever that plan is it'll deplete our routine healthcare capability and draw from the same pool of resources we need to grow the vaccination program.

jk

WT, I’m thjking about estimating Delta-R from your exponential rate constants.. it seems to me from your plots, k would have reached -0.075 or so during lockdown in the absence of the variant. In fact it almost did in the north.. 

Once failure occurred, it almost went to +0.075 in the south, with the north catching up. Obviously there are gross assumptions in there but I wonder if you can do something better from your data, then convert delta-k to delta-R and see how that fits with the nervtag estimate of 0.39-0.93?

ignore me if that’s too much work/flawed/not interesting!

It's a good observation.

I've studiously ignored "R" because

1. It's not directly measurable and getting a value includes a bunch of estimates that mean the R value ends up with big uncertainties not in the measurable.  It also looks like a lot of work...
2. R doesn't apply to hospitalisations or deaths, were as the exponential stuff does, so it makes comparison between metrics simpler.

If anyone can point me to a sensible ready-to-use method for converting the exponentials to R, I can try...

It must be near impossible without knowing which infections are which (where the new variant is in what proportion at what time) to separate the behaviour of the variant from the Christmas shopping effect, even T3 had all shops open and most of the SE was T2 anyway for most of period of interest.

I think for now we can only hope we're underestimating the impact of Christmas shopping on transmission which may be dramatically worse in London than elsewhere with its high dependence on public transport. That said given the growth looks very bad across varied parts of the SE we should assume the worst, that nearly all the acceleration is virus not xmas/rule-fatigue.

jk

Agree it’s impossible to disentangle in a single location but Xmas shopping shouldn’t ‘diffuse’ across the country from Kent..

Someone asked - I can't remember who - if I can do the doubling times plot for hospitalisations and deaths as well.

Plots below - this is now using the government dashboard API rather than downloading files.  Hospitalisations use a different region system - I have shown data where the region names match, I don't know how precisely the areas correspond.

There's quite a bit of filtering, especially on the deaths data.  The deaths data exponential measurements are  very jittery in September when the absolute numbers were very low.   I decided to blank the region as it ads nothing to interpretation but is visually chaotic.   Cases have been deweekended.

I've stacked the plots vertically which helps to see the lag between the different measures.

I'd like to label the colour bar with doubling times on the other side of the bar, but I haven't figured that out yet.  

You can see how the blue "valleys" in cases become shallower in each of the subsequent measurements - this is the blurring effect of the statistical distributions where all cases from one day become hospitalisations over many days etc. 

Fair point, there's no strong reason to suspect non-London SE counties like Kent should behave differently from comparable counties nearby except for the variant. London with its public transport/commuting situation might be expected to suffer harder and faster from increased mobility with or without the variant, we saw that in wave one.

jk

It was me.. thanks! Yes, noisy but you can see the dynamics stacking up as expected. Personally I love heatmaps for noisy data. Somehow much easier to perceive trends and separate S from N to the human eye

I saw something from the royal society on this. I’ll try to dig it out when I’m not at work later. Basically you need to know tau (generation time) to convert k (or r in epidemiology apparently) to R. It’s an exponential in r with a generation every tau days. From a couple of references tau appears to be around 6 days for COVID-19 (original strain), with wide error bars. 
 

ps. What’s your detailed methodology for extracting the exponential rate constants on a time variant basis? (Apologies if I’ve missed it somewhere above..)

I have seen several articles along these lines:

https://www.pressandjournal.co.uk/author/kathrynwylie/

Thoughts?

Edit: Fifth article down in list.

I think someone should send him this thread and ask how his counter-explanation (behavioural change) can explain the failure of lockdown spreading out geographically over time from an epicentre somewhere near the epicentre of the spread of the new strain.

I'm all for people questioning the data, but it's very lazy to just wave heads for some alternative explanations rather than producing any sort of testable theory or identifying the data needed to do so.  I also think it's not very responsible to use your scientific credentials in a way that could undermine or delay the response needed under a reasonable worst case scenario.   Certainly not when that reasonable worst case is quite so desperate looking.

Agree. He’s a microbiologist so almost certainly taking an experimentalist perspective (not a data driven one). His understanding of ‘evidence’ seems to be limited to lab studies, which clearly haven’t been done yet. 

Yes, the human visual perception system is quite something, and targeting plots at that is very powerful.  I've worked a bit on retinal imaging and vision and could easily side-track myself that way.

1. Cases data goes through my de-weekending filter
2. All measures go through a 7-point (7-day) 3rd order Savitsky-Golay filter to mitigate the noise.
3. An exponential of the form y = k0*e^(k1*x) is fit to a sliding window of size ±7 days around the filtered measures.  k1 is the exponential constant, the characteristic time is ln(2)/k1.  
4. I then put the extracted time-series of k1 through the same filter as 2.

1. This is done with scipy.signal and scipy.optimize.curve_fit

So the measurements are quite de-localised.  I picked the 7-day window on the exponential to minimise aliasing effects with residual weekly structure in the data.  The SG filters are quite narrow; I picked these to get curves that strike the right balance between noise and locality to interpret.  There is always the risk that the filtering smears the data to the point that it creates the illusion that lockdown failed half way through in my "red" regions, when actually this is an effect of the filtering.  Although I've not shown it, I do a version of the plot with no SG filtering and with a ±3 day window for the exponential fit.  This makes it clear its not the filtering.  With the exponential fit, I don't think there is a simple impulse response I can look at to quantify the blurring of the filter (any hardcore mathematicians want to correct me?). Anyhow, you can flip-book between the two images below in the gallery.  It doesn't cause me any concern.

I should really change the method - the SG filter used to smooth the data is a convolution that represents doing a local polynomial fit in a sliding window and reading the centre value out as the filtered value.  (Actual poly fitting is used for the edges of the dataset).  This polynomial can be analytically derived to get the rate of change, those quantities can be analytically related to the exponential constant.  As well as being more elegant, it's much more computationally efficient as it's just a couple of Fourier domain convolutions rather than a whole bunch of function fitting.   

I think the left hand graph without the filter works very well for the heat maps. It's more informative - you can see the noise but it doesn't take away from the ability to easily see the trend. Whereas I think your filtering is more important for some of the trend graphs that get too messy when the noise is there.

Now that’s what I call...

...

...

transparency! Have a round of applause.

Portsmouth and Havant had higher rates than the rest of Hampshire throughout October and early November by a similar margin to the more recent difference. However unlike Kent, East London and South Essex it dropped very rapidly in lockdown through November. So it's worse picture than Hampshire is much more likely due to population density, occupation types and deprivation levels. It doesn't support the theory of the variant being imported by sea.

Yes, the filtering was picked for the line graphs which just turn to the consistency of a seismograph on a bad day without the stronger filtering.  The heat maps are a lot more usable with less filtering.  Good to get other people's views on this.

I should put some science where my mouth is on this.

How do we hypothesis test that there is a space/time relationship with the failure of lockdown?

1. I take the distance from Medway to each UTLA as my "x" axis data
2. I take the date on which lockdown failed as my "y" axis data
3. I do a linear regression between x and y, and note the "correlation coefficient" or "r-value", describing how well correlated the two values are.
   • This comes to 0.72 for my latest set of data.  It'll probably change a bit over the next few days.
   • A correlation coefficient of 0.72 is pretty compelling, but we should test it against a null hypothesis... 
4. For the null hypothesis, I randomise the y-axis data through resampling (choosing values at random from the original ones).  This preserves the distribution of distances but destroys the correlation with the data of lockdown failure.  I do this resampling many times and measure the "r-value" each time.
5. The plot below shows the histogram of values for the null hypothesis test.  They have a mean of r=0.000 and a standard deviation of σ=0.083.  So, in this null hypothesis, there is very little correlation, r never gets above ±0.23.
6. The r from the actual data is 8.7 standard deviations above the mean of the null hypothesis and 3.1 standard deviations above the maximum absolute value of the null hypothesis.

This shows that the chances of this correlation arising out of randomness are basically zero by robust scientific standards.

Other null hypothesis could be proposed and tested, none come to my mind other than the weather but that doesn't seem plausible.

I agree - your analyses are compelling, and the new strain with increased transmissibility is the most likely explanation for the patterns in exponential rate change. Not proven of course, but compelling enough to require government action. Which is what Pennington should have said. I'm disappointed in his remarks, but perhaps the journalist made the most of his cynical comments - who knows?

Anyway, amidst all the thoroughly justified doom and gloom, I have one small positive observation to make. When the Pfizer/BioNTech vaccine was announced as 90% effective I felt hope leap up in my heart, and although I'm definitely towards the cautious end of the behavioural spectrum around Covid, I'm pretty sure I let my guard down somewhat - and I would guess many others did the same. 

That announcement was on Nov 9th. The start of the striking shift in the slope of the exponential change rate in cases in England looks to be about Nov 16th from your graph. I'm not sure what lag we expect from infection to case detection, less than a week, but even so, the interval is in the region that means that vaccine-hope-behaviour could be a contributing factor to the rising slope. And viruses are not necessarily the only thing that radiates out of London - information, fashion, opinion may still do the same, even these days. 

Anyway, if there was any contribution from such a behavioural change it was likely quite small, but that at least will be starting to disappear shortly.

I'm not sure when we should start to see the effect of school Christmas holidays and tier 4 starting in the South East. From a quick look back at lockdown 2 behaviour in Bristol and Hull*, cases peaked within a couple of days of lockdown starting but it was about 7-10 days after lockdown began when the average rates were clearly starting to come down outside of the noise level. Of course it took an extra 5-7 days to see that in the data. If we expected the same timescales for the South East, then we might expect it to be New Year or a few days later before we see the effect in the data of Tier 4 starting on the 19th. Do you have anything in your analysis to alter this? I was recently pleased to see on the dashboard lots of of evidence that testing lags (specimen-->reporting date) had come down, but in the last week I think they might have gone back up with the increased positive rates and the bar charts that show it in an intuitive way have conspicuously been removed.

Edit to explain the *.

* Bristol and Hull (especially) being examples of areas that rose to very high rates just before lockdown 2 took effect having not been under significant prior restrictions, which I think is most comparable to the current situation.

Only the filtering which is a symmetrical kernel so it doesn't introduce lag (unlike the common 7-day moving average), but it does obviously soften hard edges like any other de-noising.  When the data is closer to the leading edge than the filter size, polynomial interpolation is used instead and this is quite twitchy with respect to the noise, especially for the last few days.

I used to do a plot of the reporting lag over time (how many cases were lagged by 1,2,3,etc days for each day) but this needed me to manually download the dataset every day as the historic data doesn't apparently survive in the dashboard.  It was a lot of hassle and I gave it up in the end to do something more productive with my time...

I thought Tier 4 was a combination of cases rising and a high proportion of the new strain?  This explains why Gosport (adjacent to Portsmouth) is in Tier 4 whilst its neighbours with higher rates remain in T2.

This (from the Unherd article linked above) explains why the UK is able to trace this particular mutation so easily:

It’s worth talking about testing, which is where we got a bit lucky. Testing for whether someone has a virus, using a PCR test, is usually just Y or N: the test finds the viral RNA or it doesn’t. Meanwhile, to track the spread of different variants of that virus, you have to get samples and sequence their genome, which is a longer and more involved process.

But PCR tests work by looking for a few short lengths of the viral RNA, not the whole thing. And by lucky chance, the standard PCR test looks at three different chunks, one of which happened to include the two areas removed by 69-70del mutation.

This doesn’t mean that the PCR tests don’t work. The tests look at three sites “precisely to have belt and braces” when hunting the virus, says Birney. So PCR tests which return positive results on two of the sites, but are negative on the 69-70 site, are probably telling us that the patient has B.1.1.7, not the original SARS-Cov2 strain. This means it’s been much easier to keep track of the virus. That’s why, says Birney, Patrick Vallance was able to give so much detail in his presentation of where it has spread in the UK.  It’s also part of why we can make any sort of estimate about the increase in R. If we were relying on genome sequencing, even though the UK has genuinely world-leading genomic surveillance, we’d have much less info.

(The PCR tests will have to be redesigned, because the whole point is to have redundancy against mutations, and now they’ve lost some of that redundancy. But it’s been useful, nonetheless.)

Royal Society pdf:

https://royalsociety.org/-/media/policy/projects/set-c/set-covid-19-R-estim...

I checked the London and Thames estuary weather records a week or two back for the early mid lockdown period, there was nothing remarkable in the data I found.

I suppose a more systematic check might find a correlation but I'm not even sure what you'd look for weather wise: cold, rain, wind and drizzle?

jk

True, the school and uni holidays might be a saving grace. On the other hand, Xmas day...

True.

Going back to why, don't the shops (and public transport to them) get busier and busier as December goes on - could that be contributory?  Might also partly explain the London-centric effect - hardly anyone is London commuting from the Home Counties now, as the kind of professional work that that involves is being done from home.  But people were going in in droves to shop on Oxford St etc.

Hi, yes, I've read and understood the article. I'm sure the variant is in Portsmouth now. In fact I'm sure it's across the whole of the south of the country in high %s because all these areas are rising fast. What I was explaining is how you can determine from the historical testing data that Portsmouth did not have much of the new variant during lockdown 2, unlike Kent or Essex, and that there is anyway a precedent for it having higher average rates than the surrounding areas, and a logical reason for that. 

From the above it should be obvious that the variant did not arrive in Portsmouth by sea at a similar time to it arriving in Kent/Essex, which is what I thought you were suggesting.

Ps Tier 4 is just all the areas that had become tier 3 a few days earlier. These were themselves influenced by case rates. Obviously there is a high correlation with places the variant is most concentrated, but it will not be 1:1. 

The current guardian on the rise map is an excellent visualisation of increase rates across the country and hence the effective spread of the variant as of a few days ago:

https://www.theguardian.com/world/2020/dec/22/coronavirus-uk-covid-cases-an...

Edit to add, of you study the cases map on the dashboard much, it becomes clear that infection clusters most often start in an apparently isolated location or not immediately adjacent to another cluster, and then grow from there. In this way, although a high infection rate in town A makes it more likely that town B 20 miles down the road will also get an outbreak, it does not require that the villages or towns in between will also suffer an outbreak first. The infection sometimes jumps around and grows new clusters because of the super spreading nature.

Building on the correlation coefficient between distance from epicentre and time of failure of lockdown, here it is plotted as a heat-map for every possible epicentre in a grid around the British Isles.

A gradient descent minimiser was used to find the location with the highest correlation coefficient.  It's out in the Thames Estuary.  I expect it'll move around as the next few days of data refine the estimates of lockdown failure.   I want to do another one based on when the rate of change of cases bottomed out  - this is based on when it became positive (R becomes > 1).

Current epicentre is Epicentre @ 51.5489°N, 1.1780°E

My distance measurements aren't great as some UTLAs are multiple separate polygons and it's gong to take me a while to measure their barycentre correctly.  For now I take the average of all the coordinates in all their polygons.   I doubt it'll make much difference, but if a job is worth doing...

Continuing apologies to residents of Cornwall, I've got to fix your absence...  

Sunspots?   Less flippantly, the weather decorrelates over shorter spatial and temporal scales than this data.

It’s pretty hard to find a rational explanation other than effectively a pandemic within a pandemic.  Given the emergence and explosion of the new strain in the same land area bounding this epicentre, and at the same time...  

Presumably other countries could use this methods as well as I thought the PCR test was standardised worldwide.

Shopping would be a factor but it wouldn't explain why Kent in particular got badly hit. Also I'm not sure the places on the Thames estuary where this took off are particularly affluent (I may be wrong) and I suspect you need a certain level of affluence for a significant proportion of people in an area to catch the train to London to go shopping there.

A German scientist was quoted in the press as saying it remains to be seen if the new strain caused the outbreak or if something else caused the outbreak and that made the new strain more prevalent. Fair point but it's hard to explain what else could drive spread in Kent and then London as well as South Wales but not so much in other areas with similar demographics. I'm not sure it explains the prevalence of this strain as a % of all positive cases either.

On the upside at least something has finally proven very effective against Covid19. Just a bit of a shame it's Covid20.

jk

36804 new cases

691 reported deaths

Numbers in English hospitals up over 1000 on the previous day.

Numbers in London hospitals increased by 10% on the previous day.

Horrible all round

It's interesting to wonder why higher prevalence should cause a significant extension to the decay period if a really hard lock down were imposed? Full-on 'stay at home', 1 solo exercise per day (makes policing possible), random checks at ring road crossings etc. Pass robust law and enforce it with meaningful well communicated penalties. Issue a public health campaign across all channels explaining the need for the measures, draw on trusted brands to maintain credibility, persuade us all to pull together again. Deliver clear simple risk mitigation instructions for essential activities, eg: shop alone, locally and infrequently, walk if you can. Tighten up the safeguards in essential shops back to spring levels, low densities, distanced outdoor queues, blocked slots for the vulnerable (not the common and mad vulnerable + NHS mix!). Focus TTI not on those with symptoms, (any symptoms, assume the worst and isolate completely) but on those in circulation for essential work (properly defined this time so people like me don't easily get a pass), provide frequent screening rather than diagnostic tests with fast turnaround and result notification. It's horrible and expensive and it will kill people but for most it is survivable for a while at least and it would constrain the vast majority of transmission chains to one final link in the locked down home. Other risk environments like hospitals and care facilities are already where where we have focused our vaccination efforts and they are better equipped for infection control than the general public. Done right most post-lockdown infection chains should be very short and the case rate decline could be quite dramatic but with backbench MPs the way they are it could cost this government its stability, I think realistically it it would require an extraordinary sacrifice by the leaders of both big parties to stand side by side (2m apart) and explain all this to the nation.

We won't eradicate it like that in the time available but we could reduce the problem by orders of magnitude, regain some resilience in case things take another turn for the worse before they get better and allow tools like reformed-TTI to have a meaningful impact on transmission which can then be traded for some social and economic freedom as the better weather starts to help too.

jk

An update with this evenings data - less filtering on the heat maps, more on the line plots.  As ever the last 7 days or so are considered provisional on these plots.  The line colours correspond to the colours on the heat map labels.  This is definitely a bit of a specialised plot.

Some observations

• London - currently showing a doubling time of less than 5 days in cases and a trend of that shortening.  That's such an astoundingly short number I went back to the government dashboard to check that.  7-day rolling average cases for the London region were Dec 15: 7495, Dec 9: 3850 - cases went up by 1.94x in that period.  
  • This doubling time is basically where things were in March/April, except hospitals weren't so full back then.
• The South West appears to have jumped over to a "London-like" doubling time curve. 
• Yorkshire and the Humber appears to be bucking the trend; the data for deaths looks a bit suspect to me though as if some might be missing.  

I was looking at it relatively simplistically - if a given set of measures reduces the infection rate at a certain gradient, then the time to drop to a certain level is directly proportional to the starting level.

Theoretically I understand what you are saying, ie that if restrictions and distancing are very strong then the rate should drop very low in one or two infection durations (let's say approximately a month) but in practice we have not seen that happen anywhere with significant starting infection rates to my knowledge?

Really beautiful data presentation in all these graphs and I think your analysis of the data is pretty compelling. However, I still think it's quite possible that the variant spread can be explained in terms of population dynamics with respect to initial seeding location and subsequent behaviour. For me this is mainly based off the (much less sophisticated) fact that in the regions the new variant emerged, case numbers were pretty low compared with other regions of the UK when the variant first appeared (circa late September). I'm not sure what the compliance with lockdown restrictions was like in the SE but perhaps the less severe restrictions during October could have contributed to a lag in the November lockdown cases(?), which only seemed to briefly plateau in late November after lockdown. In addition, we've *yet* to see it dominate in North of England or Scotland where it had been identified by genomic surveillance at the very end of October/early November.

I should emphasise however that I think we should definitely be reacting to the spread *as if* it's actually significantly more transmissable.

From a biological perspective, the mutations are definitely concerning in terms of contributing to transmissibilty, particularly the N501Y mutation given that it's appearing in 3 variants (SA, Welsh variant (b.1.1.70 - I think), and the b.1.1.7 UK variant), that are associated with spread and case increase. In vitro cell culture work certainly suggests that some of the 17 non-synonymous mutations in the b1.1.7 variant could increase infectivity/transmissibility, and maybe virulence but this is much less clear. The Ct values from PCR testing are also a bit of a worry (plus inferred abundance from sequencing reads). Hopefully, the animal work currently underway can clear this up in a more complex systems context. 

In the coming weeks we should hopefully have a better grasp on the transmissibility via studies currently underway looking at the natural transmission of the variant at other locations throughout the UK (i.e. North of England/Scotland), where it was seeded in a more competitive environment. Likewise, it will be interesting to see genomic surveillance data coming out of other countries and determine when the b.1.1.7 variant was first seen there (hopefully they have some historic samples that have yet to be sequenced). Fingers crossed it was a while ago and missed in sampling due to low prevalence.
 

At risk of bringing this thread back to the UKC mean..

B.1.1.7? I mean sure, but what’s it ever done on grit?

Thanks for the kind comments and thoughtful take.

I agree - the rise to prominence of a new strain doesn't need it to be more infectious, especially with a key super-spreader event or two at work.  The exponential mechanic can amplify a small and apparently insignificant difference from either luck or phenotype.

However, what you raise is a separate matter to the observed failure of lockdown and the distance/time relationship between that failure and an epicentre that happens to be close to the epicentre for this new strain.  A lot has changed in even the last 3 or 4 days with the rise in many places seeing a continued rise in the exponential rates for cases into territory where the only behavioural explanation would be a complete and utter abandonment of control measures by most people, as it's so close to what was reasonably inferred about the exponential rate for infections the first time around.

I totally agree; a point being missed by some of the people being dug up to comment on the news...  

Are you taking these to hint at  increased viral load being shed by infected people?

An article from the guy behind that Twitter plot of b117 test result prevalence. Apparently they only went back and identified the variant samples retrospectively, hence why the panic button wasn’t pressed a fortnight ago..

https://www.theguardian.com/commentisfree/2020/dec/22/new-coronavirus-varia...

Interesting, thanks.  It seems the missing primer lets them do a tentative retrospective identification without sequencing which is a spot of luck, and also explains how this genuinely came out of the blue.  

Now I understand why the panic button wasn’t pressed earlier.

I assume there has been sequencing of cases from the hotspot around the Thames Estuary before then as it sounds like PHE teams were working to find out what was going wrong there.  I suppose the dots weren't joined before hand.

I sent Alan McNally a write up of my analysis of the geographic spread of lockdown failure a couple of days after his tweet as I figured it's a useful independent data point to correlate with the genomic spread - and further weight for the argument to risk over-reacting out of caution rather than to wait for the lab data.  I haven't heard back - I suspect I got deleted as a likely nut-job as he'd just had a depressing sparring match with Yeadon on Twitter which saw the unhinged followers of Yeadon descend on his twitter and work email like the scum that they are.

I agree it makes sense with these 3 variables being coincident with one another.

I guess what I reckon as an alternative hypothesis would be (mostly) behavioural change around the time of lockdown easing, both locally with local authorities and focussed on central London, with London acting as a high interaction travel hub. So I'm think you'd see a pretty strong positive correlation between distance from central London and time of lockdown failure. Although this is likely to be similar to your epicentre model given the distances between London - epicentre aren't that big. 

As a side note, I've probably missed it in the thread but how do you define lockdown failure? Is it at the roughly at the point of exponential case growth in a local authority? Is the correlation robust to small changes around the time of lockdown failure?

Yeah it's infuriating. Is the omission a sacrifice to the altar of balance between apparent lockdown vs. anti-lockdown?

Potentially. 

The lower CT values for N and Orf8 genes in that particular PCR assay (when S gene is dropped due to the 69-70del), and the higher read counts from sequencing data could suggest a higher viral load. Given both methods see this sort of effect would suggest it's not just a technical artifact of the PCR. Whether this leads to or is indicative of increased symptomatic presentation or shedding/transmissibility is difficult to know at this point. 

Thanks again for the great thread!

Yeah, he’s probably quite busy too..

On a happier note, another article which is not getting mainstream coverage but is HUGE immunological good news.

https://immunology.sciencemag.org/content/5/54/eabf8891
 

Something behavioural around London leading to cases leaking out and boosting rates elsewhere is probably the closest counter-hypothesis as you say.

They're not but an optimiser puts this one in the Thames estuary.  It'll be interesting to see where one for an alternative definition of lockdown failure falls (see below).  I'm drawing a bit of a blank on how to put a circle of probability on this epi-centre however so I can't fairly say how significant it's separation from London is...

When the time derivative of the daily case count goes positive in a window opening after lockdown started.  

Yes; the time derivative is done with a 21-point (21-day), 3rd order polynomial filter to reduce sensitivity to noise.  I run it through my de-weekending filter first.

I want to add a second definition, of when the time derivative bottomed out at its most negative, which will occur before the return to positive growth.

I regard PCR with the same leery eye of suspicion I have for most things biological - it's hard for one not skilled in the arcane arts to know when it's indicative of the inputs or of the biochemistry going on inside the machine.  Good to get the opinion of others on the context, thanks. 

For sure.  

That is good news.  

That looks way too technical for me - is it basically saying T cell immunity is long lasting? - which would be good

Yes. B cell mediated immunological memory appeared over several months and was undiminished at 8 months (end of study) implying (some degree of) long term immunity.

This sounds like a great idea, unfortunately I can offer little to no help.

However, I was also thinking about your earlier post where you plotted distance from Medway. Wouldn't a similar method be sufficient to test the London transport hub effect? Maybe choose an arbitrary point of central London, or the centroid of all London borough areas and map distance to all local authority centroids from there to correlate with lockdown failure?

I just wonder which method might capture the strength of the connectivity between London and nearby regions more accurately? 

I think that the Cog consortium did check more historical samples as it was becoming clearer that b1.1.7 case numbers were increasing. But, they had identified it fairly early and were following it because of the mutations it harboured. 

Prof Andrew Rambaut an investigator in the Cog consortium suggested that the prior probability of the variant causing the spread wasn't insignificant especially because they were following the virus at low prevalence due to the mutations. In contrast to a lower prior probability for the variant itself causing the spread when the mutations were retrospectively identified in an 'overrepresented' variant........

I effectively did what you suggest in my 14:59 post today - the correlation coefficient of time of lockdown failure vs distance from every point on the map.  The effect is marginally weaker for London.  The highest point is identified with a marker.  It’s in the sea but that’s because there happens to be sea between Essex and Kent I think.

You basically need a model of mass human movement.  Google have it, a combination of the train operating companies ticket sales and APNR data will have it, I imagine GCHQ have it.  Then you plug that in to a basic pandemic model and it’ll give you a more finessed structure.

a more readable editorial on the paper I linked above on B cell immunity.

https://medicalxpress.com/news/2020-12-covid-immunity-months-reveals.html

A hypothesis for this location is that this point is roughly between the commuter / stockbroker belts of Kent and Essex - where a lot of city boys have been having nights out entertaining clients in the run-up to Christmas.

I don't live there any more to witness if that is what has happened this year - but on past experience the trains after about December 1st were even more awash with drunks than usual - and to call them "drunks" was a understatement. All going home from Fenchurch Street, Liverpool Street and London Bridge to repeat the performance the following day. 

Papers this morning full of rumours about Tier 4 being expanded on Boxing Day.  Minimike take a bow.  Why does this info get leaked rather than announced FFS?  Why wasn't this called earlier?  Sorry don't mean to distract from the serious analysin' going on by the boffins here : )

On that note next Friday night is Christmas right?  I hope you folks, particularly you wintertree, are not tempted to spend too much time plotting and running data at the expense of family time to bring us the next edition.  

Looking at wintertree’s correlation strength plot with variable epicentres it doesn’t look like there would be a significant discrimination between London and the Thames estuary..

On that basis the ‘alternate’ hypothesis that this is just driven by behaviour seem plausible, with people moving out of London driving the spread of the new variant to the rest of the uk.. except:

1) if the new variant were simply equally infectious as the old we would not see it rapidly out compete the old one and dominate in the PCR testing results.

2) it would not affect the overall growth rate, which would be purely behaviourally driven. If that were the case you would expect the whole country to react together (albeit with a *bigger* effect in London) so the lockdown failure times should be basically synchronous.

in other words the temporal correlation of growth onset tells us something is propagating across the country from the south east. Behaviour drives that but it requires something novel in terms of infectiousness to actually spread. If it was spreading an equivalently infectious variant we would see it in the PCR (but not dominating) but not in overall infection rates, as the old version is already present everywhere. 

I worry that Tier 4 everywhere isn't going to be enough.  It hinges on how much of the increased transmission was driven by more transmission in secondary school pupils and how much by the virus just being more transmissive across the ages.  If it's the former, the end of term is going to start winding us back from the emerging crisis.  If it's the later...

A/B testing of policy by observing the public reaction?  It's how DC ran his campaign against the regional parliament back in the day here in the North East, and it's how a lot of the Brexit campaign was guided.  It is however no way to make policy in a crisis situation.  Even if they're not A/B testing any more, the process of leaking policy and observing the reaction seems to have been institutionalised.

I've been pondering that.  By the time the little ones are in bed, I'm going to need some downtime and half an hour of putting the plots out is just that, but I don't really think anyone else should see them on Christmas Day with the way things are going.  No need to spoil other people's day.

But wintertree..

”All I want for Christmas.. is you(r plots)!”

Thanks Wintertree for your robust analysis of potential epicentres for the recent spread. Given that we know lockdown failed earliest (or rather never worked) in Kent, south Essex and East London), I think it was fairly clear this would be the result, but I like the approach you have taken.

For what it's worth though I'm sure the London transport effect is very significant in the enhanced spread, just not the root cause. It's not even reasonably practicable to get to Kent from most of the country without going through the London area - either round the M25 or via changing trains at a London station. I've never had any reason to visit Essex but suspect that the situation is practically not much different unless you live in East Anglia or Cambridgeshire. I suspect that 90% of the transmission paths from someone in Medway to someone in Milton Keynes pass through someone in London.

To be honest I didn't really agree with your first definition of lockdown failure date (I'm assuming it's still the same). From looking at the results, it doesn't seem to distinguish sufficiently between areas that have seen the strong upturn that I think you are trying to spot to indicate a link to something causing a significant change in virus behaviour, and areas that simply flattened, rose very slowly after lockdown because of coming out into a lower set of restrictions as planned, or went up and down a bit, of which there are many. I would suggest an alternative measure - the first date on which a ULTA first saw infections rise for 3-5 consecutive days, after doing your deweekending effect - and see how your results look. Noise effects will prevent most flat or very gradually rising LAs from being caught by this. An alternative if you don't like this would be to use the date at which the rate of change of IFR, averaged over a few days, first reached a given value, or the date at which the time derivative reached a certain minimum positive value (maybe something equating to a 30% week on week rise), rather than when it first hit zero. Of alternatively, apply a higher order polynomial and then use the very last date the time derivative was zero, not the first date after lockdown. All of these would give slightly different results. I'm sure you'll have reasons that some of these ideas wouldn't work well but one of them might be worth trying. You could even use a few different measures to create a ranking and then compare that ranking vs distance from the epicenter.

I think you could then get a more representative date for when various LAs really started rising consistently (and surprisingly) than what you currently have.  The worth of the above depends whether you plan on using this data for anything else, or not - I don't think it would change your conclusions about the epicenter, which align completely with a qualitative analysis of the dashboard map through time.

Edit to make clearer and add a couple of things.

By the way, in case you hadn't spotted it, I noticed yesterday that the dashboard has started showing hospital admissions, number of cases in hospital and number on mechanical ventilation down to individual trust level.

I’d suggest analysing the rate constant  plots with something like a wavelet matching approach, as used in LIGO etc.  We know what the signal we are looking for is. If we chose a crude function like -sinc(t) as a moving matching function, it would identify the point at which the rate constant turned as the peak of the cross correlation function and the strength of the signal as the height of said correlation peak. This could then be used to discriminate against weak correlation caused by noise or other weak fluctuations in the data.

Thoughts?

Yes, the approach above would be robust to this if done properly

I'm not sure there is a "good" universal definition.  

This is effectively the case now - when I started doing this, there were regions that were only just returning to growth, and the filtering isn't so good at mitigating noise at the leading edge of the data.  As that date moves in to the past, the filtering is effectively going what you suggest.  I'm not a fan of more categorical measures such as "5 days of +ve growth" as they're not very holistic and raise flags when there's an aliasing effect (weekends).

I'm going to add one based on the turning point (most negative d/dt) as well as the current one (d/dt=0).  I think I'll add one based on d/dt passing a threshold as well, as you suggest.   Each of these can be used to find their own epicentre.  

The other possibility is to weight the contribution of different UTLAs to the epicentre based on either their population or their case numbers to reduce the contribution from areas most prone to noise.

I know you were concerned about the method with Knowsley - and presumably areas like it - a week ago.  I've put a quick plot cases in Knowsley for today - the turning point in cases (0 crossing in d/dt) is pretty ambiguous as the plateau is so low; this is where your suggestion of a d/dt threshold will be more precise; although it also puts us back in the leading edge of the data where we don't have enough data to filter the noise out well.

I'm trying to put the data in front of the people making decisions, because I've not seen it anywhere in what they're looking and it feels like an important independent correlate with the variant that goes a long way to parking the claims of "but there's no (microbiological) evidence it's more transmissive".  So - the key for the methods is to try and be simple, robust and widely explainable which minimikie I'm afraid I'm not going to adopt wavelets - although I agree they could get a more precise measurement, I want something simple.  

Yup - minikmike put me on to these.  It's very helpful.  I point blank didn't believe my most recent regional doubling times plot and went back to the new regional level data in the dashboard to check it "on paper".  Things are bad.

Re: wavelets, fair enough! It’s a fair chunk of work. Could you throw me the plot 18 regional data and I’ll have a crack? If it works nicely I’ll send you the code..

deal?

Deal.  It’ll be a few hours before I’m back on a keyboard.  Off to enjoy the festive rain!

*nervously raises his hand to ask what is probably a dim question*

This thread is great, although incredibly information dense so I'm still trogging through the meat of it and therefore apologise if this has been covered already:

Did the decision to put London and parts of the SE in Tier 2, rather than Tier 3, have a significant bearing on what we are now seeing? 

Ta, Jim.

My 2 cents - cases started growing in lockdown itself in much of the SE corner, so it would seem reasonable to assume that neither Tier 2 nor 3 (or 4, since that is effectively a return to lockdown? Albeit this time the schools are largely closed for Christmas) would have been enough to stop an exponential spread of the virus.

Having said that, it also feels likely that it’s rate of spread would have been slower in Tier 3, which means that we might have been able to work out what was happening and move to Tier 4 with a lower accumulated burden of cases built in.

Which I think is a qualified yes - certainly some bearing on the severity of where we are now but not in itself probably important enough a decision to flip between “problem” and “not problem”.

Enjoy. I just managed 2 hrs with my boys in the semi sleet. Bone chilling

Having looked at the tier 4 legislation - it's basically just been added to the existing secondary legislation that defined the 3 (new) tiers - tier 4 is pretty similar to lockdown 2.

It's exactly the same other than churches I believe.  This isn't a coincidence

If you look at when it started kicking off at the end of summer (i.e. when kids went back to school) tier 4 + schools closed (or on a half-attendance rota basis) could well be (potentially better than) enough.

Against the September version of covid, sure. But if what we're seeing now is Covid 2.0 with better spreading ability then the September rules don't matter any more.....

They do, because you need to use those figures to extrapolate to what will deal with "COVID 20".

I suspect some 'the virus isn't here I don't have to take care' mentality occurred for sure - people in the southeast hadn't had firsthand experience of people being hospitalised etc. like people in the north had, thus had grown a 'it's not here' mindset and thus dropped their guard. As we know, people directly affected by the virus or scared by media will go further than the restrictions, and people unaffected or not scared will ignore them, thus tier 3 in the south would have resulted in tier 2 levels of behaviour and when they were put into tier 2 compared to elsewhere they'll have behaved like it was tier 1.  All of which will have assisted the spread of the new variant.  Jersey have had a similar meteoric rise in cases (~4% per day) seeded by a couple of superspreader parties, almost certainly due to a 'its not here' mentality.

Tier 3 was enough to bring rates down slowly and Tier 4/lockdown quickly, with schools open, in areas without this new variant, in October, November.

Tier 4/lockdown with schools open was only enough to very briefly flatten but not reduce rates during November in a small number of key areas. Averaged across London and the South East it dropped for a couple of weeks then turned upwards. You could infer this happened as the new variant spread further but there could also be other factors at play.

My best guess would be that Tier 4 without schools closed will struggle to hold the new variant down in some locations. It might be enough in areas with less deprivation, lower popular density or other factors, but we don't know.

Tier 4 with schools staying shut or only key workers traveling to work would again be more likely to work, but we don't yet know.

We should start to see the effect of Tier 4 and Christmas hols on the London rates in the first few days of January.

So we now have a  new new variant that is even worse. Hmmm.

Weren't they full in London then though? You are right to be worried though, because the current virulence means things are likely to get worse faster than the government can/will react and it'll be lockdown 3.

See my 'it's not here' mentality post.

So...

Nearly 40,000 new cases today.

Hampshire (ex. New Forest), Sussex, Oxfordshire, Suffolk, Norfolk and Cambridgeshire in Tier 4.

Bristol and most of Somerset/Gloucestershire back up to Tier 3 along with Warrington and all of Cheshire (not Liverpool, which concerns me.)

Everywhere except the Scilly Isles up to Tier 2.

Have I missed any?

I'm going to try and see how obviously this is related to current infection rates.

Quite. Quarantine away South African holiday makers...

Jenny Harries showed a graph today that showed London hospital occupancy is now virtually at the April peak.

Edit if Blunderbuss is right then I must have remembered this wrongly, a lot to take in.

On Monday there were nearly 21,000 people in hospital - so now 1-2 days away from being above the April peak. 

Previously the case rates per upper tier LA were dominated by Wales but within a week large areas of the SE of England are now at similar levels. Some Welsh health boards are approaching 50% of their beds being occupied with Covid patients.

Oh dear.  That's with two weekend days in the window whose cases are being reported today. That's really bad.

It isn't......it's 3559 compared to 5201 at the peak in April

Anyone know what our surge capacity is?

I'm seem to remember before Lockdown 2.0 they said 26k before urgent care was put at risk....this seems quite low compared to the number of beds.....90k in England alone.