UKC

A pretty simple update this week - everything remains in decay everywhere at all ages.

A few comments:

• The upwards trend in the rate for cases in last week’s update (Plot 9e, right hand side) has become a small blip and they remain in pretty rapid decay.  
• The decay on all measures (Plots 6e, 7e, 8e) is slower in absolute terms - i.e. shallower gradients - than last week but that’s the nature of exponential decay and does not indicate a failure of lockdown, it's what we expect to see.  The time it takes for cases to halve is remaining about the same, indicating that lockdown continues to work well as well as before in terms of reducing infections and everything that follows.
• We see the same pattern in plot 9e and reoccurring in posts down thread where the exponential decay of deaths is often faster than cases over the last couple of weeks.  I’m coming round to Si dH’s take that this could be a sign of the vaccination which probably doesn’t prevent infection but does reduces severity of illness in the age range targeted for early vaccination - which is also the age range that was contributing the largest number of deaths.  If so, this is highly encouraging because there are a couple more weeks of vaccinations already done in that range that will start to develop in to immune responses and their effects over the next two weeks of data, and vaccination continues at pace and is working its way down to ages where more transmission is likely to be prevented.
• Including the right hand edge of plot 9 - which is very provisional - it looks like halving times for all of cases, admissions and deaths are still getting shorter suggesting the rate of improvement may yet get better still - certainly it doesn't look like they're bottoming out which is one potential warning sign of eventually returning to growth.

So it really does look like this lockdown is holding its effects very well.

Link to previous thread: https://www.ukclimbing.com/forums/off_belay/friday_night_covid_plotting_11-...

The four nations cases and characteristic times plots.

• Cases in all four nations remain in decay
• The decay in Wales might be slacking off in exponential terms.  See Plot 9x on the right hand side - the Wales curve is rising from a halving time of ~15 days towards one of ~30 days
• The extreme dip on the very far right of all curves is likely related to weekend sampling issues and is in the highly provisional zone.  Last week's plot had the same feature and I think it just keeps moving right etch week so long as cases are falling.  
• It's notable that the halving times for cases in England and Scotland are so different at ~15 days in England and ~30 days in Scotland, over the last couple of weeks.
• As with last week, cases in NI and (to a lesser degree) Scotland look to be levelling off - but last week’s levelling off is gone, replaced by decay.  I think this is an extreme manifestation of structured sampling noise from weekend effects - you can see it in the decaying slope on both plots as a sort of “stepped” profile like a sloping staircase.  I try to repair this with the "deweekending" algorithm but the leading edge is still not great.  So with a waffly, qualitative understanding that the NI plot and to a lesser degree the Scotland plot are bumpy, it doesn't worry me if they keep appearing to level off in these updates.  

Plot 17 - Regional cases

• The distinction between my "red" and "blue" regions from last December is now basically gone so I should probably re-work this plot.  
• The shrinking number of cases are about to break the presentation of this plot as now the region names are too crowded.  I like it when I have to adjust a plot because things are getting better.

Plot 18 - Regional Characteristic Times

• All measures remain in decay in all regions.
• There’s a lot of corrugated wobble in the cases data - this is residual weekend sampling jitter. 
• Squinting, it looks a bit like the case rates are slowly becoming more negative in most regions indicating faster exponential decay.
  • Perhaps Yorkshire and the Humber looks a bit worrying close to heading for growth though.  The provisional dashboard data (not used by my plots) isn’t very encouraging there, either.
• In some regions, deaths appear to be on a more rapid exponential decay than both cases and hospitalisations, as indicted by darker purple colours on the RHS of the bottom left plot than on the ones above, as well as by the line curves. 
  • I think this could again be related to the effect of vaccination.
  • Do we expect to see much more purple on this plot next week?  I increasingly think that we do, if this is the vaccination kicking in.

Plot 22, 22r - Hospital Occupancy

• About 5 weeks after the start of lockdown, hospital occupancy has dropped to a level "only" as high as it was during the first peak in April 2020.
  • This is just bananas
• If you squint at the area of pink data markers to the right of the bottom of the colour bar and at about 14000 patients on the X-axis, there's a small loop in the curve.  This is the brief reduction in occupancy caused by the November lockdown enacted to protect healthcare.  
  • Look at the difference in spacing of the data markers before and after this point to see how rapidly things spiralled out of all belief after that lockdown failed.
  • It's almost unbelievable that healthcare has stood up to this, an awful lot of people must have taken an awful lot on physically and emotionally to hold up under that.  
  • I hope we never see another event like that again.
• It's going to be some time yet before the numbers are down to anything "sustainable".

The UTLA Watch list. 

• Plot 16.1 - UTLAs are ranked horizontally by their case rates when we entered lockdown on Jan 5th
• Plot 16.2 - UTLAs are ranked horizontally by their case rates two weeks before the date of the plot

Plot 16.2 gives insight in to more recent changes in the data.

A reminder of the annotations that not UTLAs showing concerning trends:

• ▲ - Cases have risen more than 1.1x above their minimum since the ranking date, suggesting that cases could be on the rise - or it could be noise.  The minimum level is shown by a horizontal line under the data marker.
• ▲▲ - Cases have risen above their level on the ranking date
• ▼ - Cases have fallen by less than 0.1x their level on the ranking date

Baselining from lockdown (Plot 16.1) we have 5 UTLAs apparently rebounding from their post-lockdown minimum.  For each of these I looked at the provisional data on the government dashboard - which is not used in my analysis but can be interpreted somewhat usefully "by eye".  My thoughts:

• Calderdale - the rise in cases looks significant
• Cornwall and Isles of Scilly - this looks like it could be noise in the data, perhaps.
• Bath and North East Somerset - I'm reserving judgement for a week.
• Darlington - could be noise, could be the start of a rise in cases.  
• Bury - looks like it’s on a gentle rise

Some of the places appearing on Plot 16.2 have also been seen before.  It's notable that they're generally plots where cases were lower at lockdown; this does make me wonder about naturally acquired immunity being concentrated in the groups most at-risk of receiving and transmitting the disease.

There's a map to go with plot 16.2 - this includes all the regions from plot 16.1.    The places around the North West and also Darlington have cropped up before; they're definitely struggling to fall as fast as other regions.

It's worth pointing out that there's nothing special about the criteria I use to select regions here; I picked them initially as they returned a handful of places, so it puts a focus on the UTLAs struggling to consistently drop cases the most.  

Plots D1.c and D1.d - Demographic data (England)

• This week, plot D1 has been become D1.c for cases, and is joined by D1.d for deaths.  More detail on this is given at the end of plotting #11.  It uses a ±14 day window to measure exponential rate constants compared to ±8 for cases, as the much smaller numbers mean more statistical noise, so a wider window is used to compensate for that.  Results are censored (white in the plot) where a very small or zero number of deaths occur in the window as otherwise noise just Jackson Pollock's the plot up with meaning-free splats of colour. 
• There's vertical banding in D1.c especially on the right side from residual weekend effect not fully fixed by my deweekending.  Squint and you might convince yourself there darker purple in the bands at the top, perhaps as a result of the vaccination program... 

Plot D3 - Exponential rate constants vs age 

• Data markers show the most recent values from plot D1.c.  The lines show their change over the last two weeks; red for rising, blue for falling.
  • The biggest falls are at the oldest ages, with fits with the vaccine roll oujt
  • There is a worrying trend to markers moving up the plot - i.e. longer halving time and moving towards the axis crossing into growth - for young adults.  I could believe that there's a degree of lockdown fatigue creeping in behaviourally here.  If the rate of change doesn't accelerate, there's perhaps six weeks before these cross in to growth, but this could foreshadow what's coming as the vaccine rolls down to age 50 in the next two months.    

Plot D4 - demographic extrapolation of cases

• This is not a prediction. It extrapolates the current situation forwards to March 8th, the notional day for reopening English schools.  It's not a prediction as things will change between now and then - vaccination, the weather, variants, behaviour.  It's done as a way of understanding changes to the current situation, by asking "if this goes on as it is...".  Michael Hood offered this alternative explanation which I like:  For each date ("from date" - "to date") on the x-axis, the "point" shows the number of cases there would be on 8/3/21 (y-axis value) using the latest data to determine the exponential decay rates at that date (on the x-axis) and maintaining those rates through to 8/3/21 - so every day the curve will approach one "point" closer to 8/3/21.
• I'm still hopeful that the ongoing vaccination program is going to lower the groove through which cases are falling and bring the final number down on March 8th to less than 1500.  That's starting to feel a bit too optimistic for where we are now though. 

The last plots for tonight.

D1.x - A rate constant looks at demographic deaths vs cases

• A more detailed look  at this plot is here - https://www.ukclimbing.com/forums/off_belay/friday_night_covid_plotting_11-...
• This uses heavy polynomial filtering (21 days, 1ˢᵗ order) to reduce the significant noise and make a nice plot.  This spreads out peak exponential rates moderating the values compared to those seen on other plots.
• The idea with this is that we should see the effect of vaccinations pulling the oldest ages (red curves) further down and left compared to other curves.  I think it's too soon to see much effect on this plot given the heavy filtering used.  I'd hoped to see a bit today, not so much in the end.  Maybe next week.

Plot D5 - rate constants for demographic deaths

• This is a line plot of some of the data from D1.d, re-binned to reduce the noise by having broader age bins.
• The younger age bins look to be bottoming out at a having time of about 15 days, which is similar to where hospitalisation are on plot 9e.
• The oldest age bin, 80+, has not bottomed out and is continuing to see shorter and shorter halving times, meaning that deaths are decreasing faster and faster in exponential terms.
  • This could be an early sign of the vaccine taking effect.  It could also be why deaths are moving to a faster exponential decay than cases which is otherwise counter intuitive.  

The NHS and PHE will have full longitudinal data on all of this and I hope they share some results of it soon with us.  

Indeed. I know you're a bit more optimistic about the government's change in mindset, but I still fear the swivel-eyed loons will be arguing that "it's all over now", just because all the numbers are falling, even though hospitalisation figures are still higher than the first wave peak. I still don't believe Johnson has actually 'got it'; I suspect his reluctance to commit is more down to not knowing what to do, than a new, cautious approach.

We still have a long way to go before we should consider even the gentlest of unlocking.

All the right wing press sunday headlines seem to be putting pressure on an early return to normal.

A comment about plot 22 - hospital occupancy

As well as showing the numbers, this shows what proportion of patients are in ICU, higher "angle" from the origin 0,0 point means higher proportion in ICU.

The latest points are approaching the peak of the first wave so people might think that's bad, but my interpretation is - in the first wave these (high proportion) points were approached from below as Covid threatened to overwhelm. In the current wave these (high proportion) points are being approached from above because we're managing to keep more really seriously ill people alive.

The 'approach from above' is more to do with the delay between hospitalisation and the need for ventilation. So, as case numbers fall, hospitalisations fall some time later, and ventilation cases fall even later, as the disease works its course from early signs to severe illness. On a rising wave, hospitalisations lead ventilations, lead deaths, so the curve rises 'to the right'. As cases fall, ventilations lag behind hospitalisations, so the curve 'falls to the left'. It's a classic lead/lag response, or phase delay, that can be shown as a lissajous figure; if' there's no delay, you get a straight line whether rising or falling (assuming there are a fixed fraction of hospitalisations requiring ventilation).

Or maybe I misunderstood your 'approach from above'...

I actually thought BoJo was reasonably cautious in his brief interview today. Holding out the prospect of a relaxation starting with schools but noting the need for caution.

As to the idea of retail and outdoor pub service starting from April, I think a lot depends on what impact getting the schools back will have. Easter is still 6 weeks away and even if the current decay rates slow down,  the key indicators will be fairly low by then (although cases will still be higher than you’d want for a full unlock - but it won’t be a full unlock). Plus at current vaccination rates pretty much all over 65s and vulnerable people who want to get vaccinated will have had their first jabs around the end of the first week in March, so should have a reasonable level of protection by Easter. So a lot of it will come down to what impact the schools will have on cases. 

Boris has to appear cautious and knows his part of the press will push the line he instinctively wants to support.

I wish I could share you optimism on the vulnerable, best you don''t turn over too many stones....

https://www.theguardian.com/world/2021/feb/13/new-do-not-resuscitate-orders...

F*ck me. Anyone for a bit of eugenics?

It's not what effect the schools will have; it's how the government respond to rising cases as a result. Their previous record isn't good, and I fear they won't have learnt the lesson.

I think I'm correct in surmising that arriving at that point from above is better than arriving from below, but I think I may have missinterpreted why.

Your explanation sounds more robust to me.

Edit: if I've understood correctly, you're saying that in any "laggy" system, a Lissajous plot will form a loop. Presumably the "depth" of the loop gives some kind of feel about the lag.

Damn: thought I was being insightful last night ☹️

I agree with you above.  CP explains it well.  When we passed the current X-axis point with cases rising, ITU levels were half what they are now - because only half the fraction of patients were needing ITU as in the first wave.  Now, the Y-axis value remains higher on the way down than it was on the way up because of lag from general admission to ITU admission.  If there was no lag, this plot would trace out a diagonal line forwards then backwards.  As you add lag or a “phase shift”, the diagonal line opens up in to an ellipse.  Engineering types use this sort of plot to eyeball the relationship between two signals.

Try this live demo here - untick the "Animate φ" box, and set the frequency of both channels to be the same (a and b) then vary the phase with the φ slider.

When you set the two frequencies to be different is when you get the classic "Mad Scientist B-Movie" oscilloscope display.

https://academo.org/demos/lissajous-curves/

He's got to contend with the swivel-eyed loons he has filled his benches with, though:

https://www.ukclimbing.com/forums/the_pub/end_covid_restrictions_by_may_tor...

If you take two sinusoids at the same frequency, and use them in an x-y plot, you can show the effect nicely. With no lag/phase difference, you get a straight, diagonal line. If they are 90 degrees out of phase (e.g. they are sine and cosine), you get a perfect circle; that's how sine and cosine are actually defined, after all...

Oooh that's fun to play with, thanks

In reply to captain paranoia:

Thanks, wintertree's link allows that to be nicely seen

There is an extra factor to what people have already told you, which is that the demographic going on to ventilation is very different from the demographic going to hospital (and also from the demographic dying.) Offwidth linked a indie SAGE presentation 2-3 weeks back that showed some good data on this. So, if there is any change in the demographic of cases, these things won't all change proportionally. Which means that even in a zero lag situation the line wouldn't follow the way up back down exactly.

By the way, is there is a reason you only start plot 22 on the 07/04?  It looks like there is an extra few days of data you could use on ITU occupancy on the dashboard, and an extra couple of weeks on hospitalisations, at the beginning of the pandemic. Having more data at the beginning would aid comparisons between the different waves, because the current graph starts too late to reliably see what the gradient was on the way up first time.

And If he does crumble under the pressure and “save Easter” or “April the Thirst” as the front page of the sun is claiming - it will be too soon have enough data from the effect that re opening the schools has had. 

Especially if they use their R value that, as wintertree points out, appears to have at least an 8 week delay compared to what is actually happening with case numbers. 

Yup, and before we know it cases are through the roof again. 

Thanks for this steer.  I thought I was using the earliest day available for ITUs when I set the plots up but I didn't make notes of the set up process so I can't check; either I got the date wrong or a few more days of data have since appeared - the quantity of data on the dashboard does slowly get improved, or I may have been having a slow day...

Updated plot below - it confirms the suspicion you perhaps had that the loop has a much larger area this time, even accounting for changes in scale, implying more lag between admission and ITU admission - which fits with the improvements made perhaps.  

A good point; I think this will only turn a diagonal line in to an open shape, in the absence of lag, if there is a demographic shift going on in infections during the course of the wave.  Which there has been.

The NHS are apparently planning for a fourth wave in July. Don't blame them.

Now that you can grok a Lissajous figure, here's D1.x for times around the failure of the November lockdown.  Comparing that to the present D1.x shows just how much more effective things are this time around. 

Do you need to update the scaling notation now...?

😬

Yes I’ve seen that and it seems to be an omission in the list of the priority groups (though I suspect the JCVI will have considered this carefully) but the article lacks context - how many people are we talking about here? Hundreds of thousands or millions? It’s an issue but if say 90% of the ‘overall’ vulnerable population is vaccinated in the next few weeks, that would be an excellent outcome (it will never be 100% unfortunately).

It seems even the CRG agree that the more vulnerable population needs to be vaccinated before any significant unlock (many of them are over 50 themselves!). My concern is what if there is a full unlock with 21m unvaccinated ‘non-vulnerable’ under 50s hitting the pubs and so on. The cases will shoot up and with that comes a higher risk of more mutations. 

Well I suppose if the schools don’t have much of any impact, there won’t be an issue in the first place. If they do, I agree that it will depend on what the response will be. The obvious thing would be to delay further relaxation of measures.  

Raan was fairly clear this morning (fairly clear for a politician, that is) that they aren’t going to surrender to the CRG. I do think they’ve learned their lesson. Not that I’m a fan of this government or the Tories in general!

It is incredible that Raab is now being seen as one of the "good guys". Shows just how removed from reality those on the nutty fringe are.

If they all reopen on the 8th, that’s 3 weeks before needing to confirm a decision to reopen (say Mon 29th if looking to reopen a few days after that), which is definitely enough to see a rise in cases and should be just about enough to see a rise in admissions. However if they reopen primaries and secondary exam years first (which is sensible), the effect would be muted. If all the other years go back later, any reopening would need to be pushed back accordingly.

My suggestion would be to move the Easter bank holiday dates back 3 weeks, if they’re keen to have a relaxation before the bank holiday weekend. 

Point well made!

That’s also really interesting in that you wouldn’t be able to tell from the first graph the disparity in axis crossing times that you can see in this one via the marker point on Jan 8th. A lot of working age population is dropping away into decay by the 8th whilst 85+ is still on the rise.

Is the inference of a wider arc the same here (the earlier version) as on other plots - the fact that 85+ look to have a tighter arc than say 45-49 is indicative of a shorter lag between case detection and death? (Or rather, between case growth and death growth, which feel like they must amount to nearly the same thing)

This bugs me. Why have things been more effective this time compared to November? The non-educational restrictions are a bit more stringent but not massively so. The messaging has been more severe so that may be part of it. The main reasons I can think of are that L2 simply wasn’t long enough and was followed by T2 in London and the SE (just as the Kent strain was taking off); and schools being kept open. We also have more here immunity now in the demographics most susceptible to infection. I doubt vaccination has had much impact on transmission so far. 

Have you got source for that? 

Earlier in this lockdown there were all sorts of mobility metrics doing the rounds to back up news stories about how strict or unstrict we were being about lockdown. They universally showed that this lockdown was less limiting than last March's but significantly more limiting than November's. In November you also had schools open which is the age group in which cases (briefly) rose first with the new variant and, in the SE I'm sure, some sense of complacency because they were all at low reported case rates when lockdown began and had been in Tier 1. So the environment nationally now is very different from Kent, Essex or London in November. The November lockdown worked very well elsewhere.

I will say that we are not out of the woods in this lockdown yet. General case averages are going down really well but they are lagging more in a lot of more deprived areas if you look at the maps at MSOA level; there are still several big cities with large areas over 4-500 cases per 100k weekly, which are primed to shoot up fast.

It will be interesting to see how well your predictions agree with other vaccination impact models.

This paper from the Association of Anaesthetists published on the 11 February is worth a read.

https://associationofanaesthetists-publications.onlinelibrary.wiley.com/doi... 

The paper is roughly in agreement with one published by the COVID-19 Actuaries Response Group on the 11 January.

https://www.covid-arg.com/bulletins/categories/bulletins

Both papers suggest covid deaths will fall quite rapidly in the next few weeks but the real impact on hospital and ICU admissions will not be seen until the end of March and into April when most over 50s have been vaccinated.

Of course the difficulty is picking out these changes due to vaccinations from the changes that are due to the lockdown. Also, as most cases occur in the under 50s, lifting the lockdown too early could reverse many of the gains already made.

I’m not sure that’s long enough. With the incubation time of the virus plus data lag I don’t think three weeks is enough to have an accurate picture of the situation. Don’t get me wrong, I want this over but I think pubs three weeks after schools would be an unholy gamble. 

Reported on BBC News this morning.

"the demographic going on to ventilation is very different from the demographic going to hospital (and also from the demographic dying.) Offwidth linked a indie SAGE presentation 2-3 weeks back that showed some good data on this."

Copied again here... starts at ten and a half minutes in and runs for about three minutes.

https://www.youtube.com/watch?v=gKTHqyFfzFs&feature=youtu.be

Shows the power of the vaccinations though. We should see deaths drop dramatically over the next 4-6 weeks. By the end of March we really should see huge reductions in the number of deaths.

The ministers repeat that this is based on JCVI but other experts have pointed out there is other data showing big variations within risk groups: for ethnic minorities in particular (and very high risks if you combine all these risks with the vulnerable in care or other residential homes). I've already said essential workers who can't always socially distance should really be vaccinated before my group (the 55 to 65s, not vulnerable, at home).

It's great news less people are going to die but we are still hobbled in healthcare until the hospital occupancy drops a good bit below current levels and that's going to take longer. The independent sage youtube link above explains that clearly.

Where, the bit I saw 10-15 mins said pressure should hopefully ease over the coming weeks at her final slide. And this was 15 January. What's the current status a month on?

Our ICU usage is dropping down. We're still dealing with some of the Christmas surge but our ICU capacity is dropping towards 70%, from 60% it's pretty typical occupancy numbers (60-70% is the typical range of ICU occupancy here). We're now closing down the additional field hospitals and extra covid wards created. We're certainly down to regular flu season numbers now in terms of ICU beds being used. 

This week we should see the superbowl impact but that seemed like small gatherings and so far, 1 week on cases are still dropping. We'll see it this week though.

Right now - as per wintertrees graphs, hospital occupancy currently is basically at the level it was in the April peak.

As of the latest data for the UK, hospital and ICU occupancy are both fairly close to the peak that was reached in the first wave last April - see Wintertree's graph above at 11.55am. However, that graph is a few days out of date and things are dropping quickly. Additionally (and for which I am very grateful) we have never had a situation where the whole country hit a peak at once, so although individual areas have been at capacity and sending people elsewhere, that first peak was not actually very close to what can be managed now with some preparation across the nation. Where I live in Liverpool region, we were the last area to be hit really badly by the recent wave but our hospital occupancy has now dropped below where it was in our previous high in November (we didn't get that high last spring).

I think Offwidth is probably worrying about how many new hospitalisations are generated if we release restrictions too early with loads of people in their 50s and early 60s not yet vaccinated. It's pretty clear that occupancy is going to drop well below capacity over the next few weeks, but depending on your expectations for rates of increase in infections across the population, you could still see a hospital problem fairly quickly if you just "let it rip''.

Yeah I hope we don't.

I think we're a good year off full release here. Masks in school will probably be the norm through next winter I think. But we'll slowly increase occupancy rates of bars, restaurants etc. outdoor living crashed covid numbers last year so I don't think there's any reason to expect it to be worse now we've the vaccines out - J&J should get approval in about a week or two.

But I think we are showing we can get schools open, masked and socially distanced, outside as much as possible.

It actually helps having state control now because we can manage it specifically for that region pretty well.

We really need summer travel to get going though which I think will happen. 

Average incubation period is 5 days I think but let’s allow 10 days. 2-3 more before people go to get tested.  Testing results lag is less than a week, so 3 weeks should be sufficient to see if there’s been a significant impact. I don’t disagree that longer would be wiser. Re pubs, the chat seems to be about outdoor service only. 

Today is the first day below the April peak in England (yesterday was a whisker over). Scotland passed their April peak a few days ago. The numbers continue to drop fairly quickly, which is great but inevitably it will taper off eventually. Wales and Northern Ireland are still some way above their April peaks but their numbers are smaller and noisier so harder to analyse.

A point of pedantry:  What I'm showing is an extrapolation, and is not a prediction - it does not take in to account the effect of changes that have occurred but have not manifested in the data used, nor does it take in to account the effect of future changes yet to occur.  It's a way of trying to view changes to the multivariate present (case numbers and exponential rate constants in each age bin - 38 variables) by projecting them forwards to a single, tangible value.

I think that the exponential decay of cases in this lockdown is far more rapid than a lot of people expected; I hope that by the time we get to mid-April, the case rate will be so low - even if new first vaccinations stopped tomorrow - that the impact of vaccination becomes hard to divine; it will manifest not in the numbers, but in the ability to release more restrictions whilst keeping the numbers low.  All this is presuming none of the new variants are catastrophic.

All fair points, my fear with the pubs though is that it will snowball. Like last time, what starts as very clear working practises soon become eroded and turns into a free for all. 

The imposition of DNRs is abhorrent IMO, and a gross abuse of power. 

However, in terms of prioritising the vaccination of the vulnerable, I am not so invested in this argument.  Many different groups are asking for priority for vaccination, and every argument I have seen has merit to it.  Previous decisions mean that we most definitely did not have the luxury of time at the start of 2021 to make the perfect plan.  I am grateful instead that the government made a good plan and are throwing everything at it to make it work.  

The single best way to protect the vulnerable IMO right now is to hammer cases down as fast as possible, and to free up healthcare capacity, and that is happening.  The situation was so desperate at the start of the year that the clear and absolute priority for the vaccine use was to reduce the number of people going in to hospital, to pull healthcare back from the brink; given the demographics of hospitalisation that was a clear decision.  

Every person vaccinated protects every other person out there, sometimes by lowering transmission and sometimes by preserving healthcare capacity (which, indirectly, lowers transmission through hospital based infections).  

Sorry; I am having a slow day today; please take pity and spell it out in simple words...

I look at the schools around here - car parks, playing fields, hard standings, lawns.  Why on earth aren't their heavy duty open sided tents installed?  It's common that a class of 25 has a teacher and two teaching assistants; put half the class outside, halved the density in the indoor rooms, have one staff member in each and the youngest one move between the two.  

Then again I look at schools around here and the number that have had positive pressure  ventilation installed is 0.  Likewise HEPA filtration.

It seems like there's so much low hanging fruit that nobody has even tried to pick in 12 months.

It's worth noting that that staff who make up that "capacity" have been consistently overworked for most of 2021 and can't keep that up indefinitely; I think a significant and sustained drop in the definition of "capacity" is needed to look after the 400,000 or so medical professionals who have been carrying the brunt of this without a break for some time.  

Indeed.  My estimate is that going for "let it rip" without immunising everyone down to 52 years of age or so could overload healthcare - although not for such a sustained period as we have recently experienced.  This is ignoring the bit where having (high prevalence in younger people) mixing with (large, immune population in older people) provides a near-optimum environment for generating variants with a near- perfect selective pressure to elevate an immune evading variant to prominence.  Vaccination at a societal level is very much like taking antibiotics at a personal level; taking a half course is worse than taking a full course.  It's in everyone's best interested to keep cases low and falling until vaccination is complete.

Latest data from Israel showing a 92% drop in severe disease amongst those who have had 2 shots of the Pfizer vaccine...

Our public school system here is still arguing over updating ventilation systems. We updated all ours, plus HEPA filters and portable air filter/sterilizers we have in every room, plus gazebo type tents last summer. All music/singing classes for example take place outside in these tents. 

I agree with your argument (and it's what I'd propose) but the capacity to sort out the most vulnerable remaining can be mainly separate from the main vaccination effort. It's about finding those people who could do with vaccinating ASAP and then prioritising them as soon as possible. The rest of the population go through the standard system as quickly as the system allows, with prioritised slots at the end for anyone in the next group down willing to wait, using up any spares as efficiently as possible (and never binning useful vaccine if possible).

We had over 50 per hundred thousand in hospital with covid at the peak and are only about 30% below that right now. We are on the low side for ICU capacity in the EU. The NHS repurposed many wards and staff for covid hospitalisations and ICU.

NHS staff are exhausted, many are traumatised, many are off sick, quite a few died. We have a backlog of millions of non covid operations and other health interventions delayed by the pandemic. Probably double covid deaths have long covid and will need some help (a minority of those will be rehospitalised), mental health capacity needs resulting from the pandemic are huge. We are more than a few months off normal.

If we open up near fully, as the UK libertarians want, it won't take many weeks to put us on a certain track to overwhelm basic hospital capacity again, unless we are wrong about this variant and it's a good bit less infectious and serious than we think.

I think you'll read I said a year plus..

There's no chance we're months off normal. I suspect mask useage may now be the norm, especially for the winter/flu seasons in schools.

That doesn't mean schools can't start opening. My wife's an MD in a huge regional hospital and yes medical staff are tired, but they are also parents and want kids educated. And they need kids in school so they can work and clear back logs. I keep hearing non-medical people argue for an extended lockdown to allow medical staff a break. I've never heard any of them ask for that. My daughters best friends mum is an covid ICU nurse and is desperate for us to visit. She's desperate for some sort of normality. My wife did an internal medicine residency so works the wards and many of our friends went into critical care fellowships. I've mentioned that idea to them and had a 'god no' response. That doesn't mean I support a full return to lock down which seems to be the position you think I'm taking.

Our psychiatric admissions have tripled for kids recently. I'd say the most difficult thing for my wife has been dealing with our 5 year old at home when schools were closed. As an oncologist covid has caused the deaths of many of her patients directly, or indirectly through lost timely elective procedures so it's been pretty brutal. The long days and no down time, when she's off she's got a 5 year old or twins. No family help. The idea that she can now take days off and kids can be at school is great for her. 

As usual we seem caught between the choice lock down or full relaxation which is a stupid position to be in. But seems symptomatic of current politics. We have to exist in either extreme and the common sense approach has gone. Schools are a great example how we could have spent a year making significant changes but it largely hasn't happened. We can and should look to gradually release restrictions. That doesn't mean normality is months away. I suspect it's next summer if ever. Covid will have changed our lives forever in terms of where we live, how many days we work, office capacity, masks, indoor gatherings etc. Some of these things many return but many won't. We don't need to work 5 days a week in an office. 

But we know large class sizes are wrong from an educational point of view. We spend millions looking at approaches to learning and just won't reduce class sizes. And now we have a public health reason to get class sizes down. Class sizes and timing of the school day are the two easy things that have been staring us in the face for years and we won't make the changes because it is too difficult, when in fact it's pretty easy. It just takes money.

By normal I meant ITU capacity and a start on clearing backlogs, assuming we have the staff to run this. I really can't see life for ordinary people being completely back to normal this year.

Common sense hasn't gone. Our scientists just urge a careful watch on the R number as schools go back in a phased way. For now the cabinet are following this line. If things keep improving, we can keep opening more of society carefully. The vaccine should help significantly reduce risks of hospital overload in a few months (providing we don't just open everything too quickly).

My partner and quite a few friends work in the NHS so I know how brutal things are in the UK and I"m always glad you share your US experiences here.

You have added data for the period before 04/02, as suggested by Si dH, I think?. But the label showing the plot range colours still starts at 04/02...

It doesnt look like R is the right measure to watch, as it appears to have a very long lag (as pointed out elsewhere). Cases have been falling since the start of the year. Only now has the computed R fallen below 1 (hence the minor fanfare this week). Something isnt right.

Maybe the lag isn't as long on a rising curve, but I think there are simpler measures (case numbers, hospitalisations and the associated exponential growth factors) that wintertree has been studiously plotting for weeks, that give an almost immediate, and clearer indication of what is happening. Computing R is too much of a fudge, based on multiple measures, hence always quoted as a range, not always usefully ( e.g. 0.9 to 1.2).

Yeah agree, with the new variants I think we could have various restrictions in place for a while.

Only now on the ONS data, so with a week and a bit lag, and because the previous two reported weeky reports straddled the peak. I agree SAGE have to keep their eye on the other factors informing government decisions but in terms of public explanation, in a partly unfriendly political and press environment, fundamentally R needs to be below 1 everywhere and not growing: it wasn't in September, it wasn't in Kent in late December due to the new strain. That things were so obvious (even without the early warnings from the likes of the fantastic detective work here from wintertree) is what is so criminal about the response of Boris and co and why they cost so many lives in the late implementation of tiers in September and even more so in the late start of this lockdown.

I think that when a politician says they will be monitoring the r rate to see what decisions to take, it would be far too literal an interpretation to think they will be using the r number that is currently published by government. It is just their shorthand way of saying that they will respond to the direction the pandemic is going in the country.  Boris et al don't have time to read and understand all the data; they will be advised by their scientists who are very capable of looking at all the latest data in an intelligent way (and hopefully they will then take intelligent decisions using it, but that's the bit we don't know). I don't personally think the published r number is worth us on here giving it the time of day, it doesn't tell us anything and it won't precipitate policy change either. It's there as a Government messaging tool for people who don't look at anything else.

R numbers from other sources are generally a more genuine calculation looking at rates of change in cases or modelling of same, so the same lags aren't involved.

Thanks again for another update

I was looking at the dashboard this morning for the latest vaccine figures, I couldn't help but notice the huge drop in reported cases over the weekend from 13,000ish to 10,972 which is roughly 28% lower than the previous Sunday, is this a genuine drop?

I was expecting the decay to slow down the lower we get.

https://coronavirus.data.gov.uk/

I agree with that. But if it's that useless, why persist in using it as a message? Part of the problem we have is government messaging.

Maybe they've kept the published R values above 1 all this time (as opposed to the 'real figures' they have access to), in order to try to persuade the public to continue to comply (I.e. the published R values are made up to present a political case). In which case, why publish a sub-1 figure now...? What is the political motive for that?

If those other figures they have access to are clearer and more compelling, why not use them to present the case to the public?

The numbers have been up and down for the last few days. Day-to-day changes have significant noise, over the general downward trend.

But you're right about the fall not seeming to show signs of slowing in a constant exponential fall; near me, it's looking linear, and may even be getting faster, in absolute numbers. Wintertree's exponential rate factor is increasingly negative, which is consistent with that behaviour.

Yes, cases in England seem to be consistently dropping by about 0.7x week-on-week, with no sign of that slacking off.  

At some point soon the cases could be low enough for contact tracing teams to rapidly trace “up” a level then back down other branches of infection, which might hammer that 0.7x down some more.

Because they want to start talking about opening up, so the message now needs to be that measures have worked.

Being cynical, if I was the government, I would want to be one step ahead of the public, as often as possible. I would not want the public to have all the information I had, only enough that they could support my decisions. The Government are not used to sharing all their information with us. It's great that they are doing so in this pandemic but not surprising that they still present artificial data for a less interested audience.

They've been making similar claims for some time without much sign of policy being linked in any clear way to either the frankly bizarre R number that’s published or to what’s actual happening.  Some of the biggest relaxations last summer happened at times of low cases but high R.  Still, I’m hopeful the behind the scenes changes have people a bit better advised this time around.

It’s an interesting take.  The publishing of data on the dashboard is great and really does deserve feedback and encouragement for more of that sort of thing.  

I think decisions are far harder to judge when cases are low because the change in direction is hard to determine from the noise until you look backwards. Still, some of the decisions made last summer were clearly very complacent and not well thought through.  I hope you are right

Exactly. That was the point of my rather leading question... I don't like the politicisation of 'the science'...

We are seeing the impact of vaccination and natural acquired immunity (guesstimate that 3 - 4 million people have had Covid since early October to early Feb), so the case rate shouldn’t have a ‘tail’ like last spring/summer. This is what gives me optimism.

Some fag packet maths: currently giving 3 mill vaccines a day. Let’s take mid April as a point in time. Anyone vaccinated etc by then will have some immunity kicking in from beginning May. By mid April (8 weeks) we should have given another 24 million jabs. 3 million will have to be second doses for those who got the first dose before mid Jan. So by mid April, 21 + 15 = 36 mill people will have had a jab, of which 3.5 will have had a second jab. A whole bunch (scientific term, but it will be millions) will have some immunity through having caught it. This makes me optimistic for hospitalisations and deaths etc. Factor in better weather and habitual change (more mask wearing etc) then things don’t look too bad at all. (The libertarians know this so pushing for total easing of restrictions). 
 

Natural immunity also seems factor (from testing donated blood from pre pandemic). But they estimate around 80-100 million have had covid here, 3-4 times the no. of confirmed cases.

Yes, the statistical noise is much worse with low cases.  Minimike has asked before how I’d put a certainty on a statement that cases are rising above some threshold level; it’s a non trivial question given the unusual nature of the “sampling noise”.  I rather look at it in simple terms - when people are arguing about if cases are rising or not, they’re definitely not falling.  They should be falling - as otherwise they’re rising.  I certainly think that looking at the data with the best possible set of tools is the way to ask this question.  

I can believe it. 

....which is fine until the South African or Bristol variants get there, in which case it's possible that only ~50% of those previously infected will be immune.

Yeah we've had 28 million cases and 485000 deaths, which is about 1.7% fatality rate. They suspect the true fatality rate is somewhere around 0.6% so it certainly makes sense.

That's the big unknown I think, how much immunity do previous infections (and the vaccines) offer against the new variants, do people get as ill again? worse?

Of interest to you:-

https://twitter.com/theosanderson/status/1356549735033823235

Has last week's concern regarding Rutland/Lincolnshire/Northamptonshire died down now?

Does it appear to be explained by a mixture of the prison outbreak plus some statistical noise?

I haven't looked at the LTLA level where Si dH had noted concerns last week.  Nottinghamshire as a whole is not falling very quickly; about half the rate of decay of the national average perhaps, and North East Lincolnshire seems to have stopped falling.  The Rutland data does look like one big event dominating the numbers briefly - the spike ran out as fast as it stopped.  Kingston upon Hull, just across the water from North East Lincolnshire is falling very slowly.  

So perhaps a combination of some low case rate but concerning UTLAs that don't seem to be moving much, the prison outbreak and a bit of statistical noise.  As ever though, regions have to be towards the more concerning end of the spectrum in terns of their exponential decay rates for the statistical noise to be able to bring them to prominence.

Another reported drop, down to 9,765.

 I just really didn't expect the reports to go from 68,000 to less than 10k in five & a half weeks, we haven't seen cases that low since September! Long may it continue.

Today's case numbers by reporting date for the UK were below 10,000 cases/day for the first time since I think September 27th 2020.  "Real" numbers are a bit higher, this is because cases are currently being reported from the weekend sampling low.   Hopefully it wont't be long before a trendline of cases by specimen date is also below this number.

It's a largely symbolic milestone but it's one I'm very happy to see reached.  

The press might notice and latch on to this as a good news story - god knows people really need some good news with the level of despondency about.  I do hope however that the media carefully qualify the current situation in terms of (a) how fast it could grow from here without restrictions (*) and (b) the situation in hospitals.  

(*) 6974 cases reported on Sep 27th, 11805 cases reported a week later on Oct 4th - driven mostly or completely by the less transmissible old variants.

Edit: Jinx - crossed posts on the same subject with Punter_pro.  Long may it continue.

Doesn't change anything you said. It's mostly areas at low cases with the consequent extra noise. There are some areas around there (eg Rushcliffe) that seem very flat though. 

Rutland specifically was definitely the prison outbreak - cases have dropped back very low now if you look at the leading edge. The seven day average will put Rutland at the bottom of the tables again within a few days.

So it seems, although that's not a bad thing when it's decreasing. Not good though if it's increasing.

Hospitalisations are the most accurate metric but lag infections by a couple of weeks.

That's the thing, back to late September sounds great but we lost control of cases in early September if not before (as in exceeding T&T's capacity to keep any kind of meaningful lid on it). As you say, October saw rapid growth and it took just over a month from late September until L2 was imposed, which was already too late. So great work (to everyone in the UK or at least to those who've more or less followed the rules as opposed to attending illegal gatherings etc) but we need to keep it up. I don't know how low cases need to drop for T&T to be able to get on top of it - I should think low thousands at most. 

Agree Rushcliffe is a problem area in Nottinghamshire but I still have no local news as to why. It's puzzling, with no obvious reasons and being a very affluent area on average. Even so average case numbers are 35 a day....pretty low numbers, despite being flat at 200 per 100,000 (the Nottingham population rate is higher than that but falling, based on ten times the actual daily cases). Bassetlaw is oscillating above the 200 population rate for weeks now but higher than Rushcliffe recently on a population basis.

Returning to your link to the tweet from David Spiegelhalter‘s  on plotting #11 on the subject of hints of vaccine effect in demographic deaths data:

Here's my take on the demographic death data as of today.

• The left plot on each figure shows deaths in various age bins.  Both figures have raw data ('x' marks) and polynomial filtered trendiness.  
• The first figure is the deaths data as actuals on the left hand side, the second plot normalises all curves to the same peak value, and has a log-y scale, at which point different exponential decay rates manifest as different linear gradients.   This is a good one I think to compare the decay on the different age bins.
• The right hand side of each shows my exponential rate measurements made over a ±7 day window ('x' marks) and polynomial filtered trendiness.  

There was an apparent levelling off of deaths in the 50-59 bin about a week ago which has not persisted - visible in the log-y plot.  This would have significantly affected his 0-64 age bin I think, as the dominant part of that bin.    Now it's working its way out of my rate measurements, the differentiation between the decay rates on the far right of the RHS sub plot is much less clear.  Still, the 80+ age bin continues to have the fastest decay.   

I'm not sure if the situation will ever get much clearer than this; the numbers in the younger age bins are (thankfully) really quite small and decreasing, so the statistical noise there is a lot worse.   I still need to have a scientific go at putting uncertainty values on the exponential rate measurements, my gut feeling is that calculating a difference between any two of the rates with these applied is not going to give a lot of significance to the differences.   That would be the case for any reasonable scenario I think though.

A lot of people are staring at digital tea leaves looking for hopefully messages.  It's great that we continue to be in a position to do that.

Continuing the optimistic trend, I've also put an update of the hospital occupancy plot below.  The smoothed trend lines have crossed, showing that both general and intensive care occupancy is now below that of the first wave.  Both measures have a long way to go yet.

I've not seen much (any?) press on this but I've been quite impressed with the reduction in latency by which cases are reported.  Actually measuring this needs a daily download of a file - it's not extractable from the API from what I can tell - and I've binned daily downloads as it took too long.  But it looks like most cases are now reported within 1-2 days rather than over a long, 5-day+ window.  As reporting seems to tally with entrance to T&T, this is really promising.

For T&T to really hit the power it needs to prevent growth in the absence of restrictions, I think it needs to use sequencing of every case, expert analysis and boots on the ground to piece together infectious routes - or to understand where one is a mystery allowing for local lockdown for a few days whilst it’s investigated - and to be able to trace up, across and down meaningfully.  This is probably in to the 100s of cases a day rather than 1000s?  I doubt we’ll get there, but I’m hoping to be proved wrong.

David Spiegelhalter has updated his original plots on deaths.

https://mobile.twitter.com/d_spiegel/status/1360986041948246018

l must say l think yours are clearer and more convincing. Well done.

Thanks.  If a plot is worth doing, it's worth doing well.  I'm a bit ashamed of the one I rushed out on the last thread where I did a line plot of the raw data.  I think doing feint, raw data markers and a trend line goes a long way to conveying both the underlying behaviour and the statistical noise; with plots like that you can get an intuitive feel for significance quite easily, and for the significance of differences between 2 normalised datasets.  None of which excuses me from presenting the exponential rates without uncertainties where-as Spiegelhalter does give CIs...

I think there's more going on than meets the eye here, and that it makes it quite questionable to compare directly between demographic death rates.

I think we're going to see a "long tail" of deaths come to dominate in the decaying phase such as the present - a small fraction of people who take much longer to die than which ever average one measures.   This is insubstantial during a rising phase but comes to dominate a decaying phase, and I think will have a demographic component.  I haven't really got my head around it yet,  but looking at deaths within 28 days and 60 days of a positive PCR test suggests its important.

The 60 days data is available from the API with the key "newDeaths60DaysByDeathDate".  I've seen 60-day data discussed here and there; I guessed at the API key - Googling for it doesn't actually return any results apart from a UKC post by me.  I should probably email them to verify its precise definition...

I've put a plot of this data below.  It's notable that clear water appears between the curves after a rapid decrease in the rate of change of deaths (i.e. a growth phase ends - either levelling off or turning to decay).  

This suggests to me that there's along tail to deaths.  Without full spectrum data on deaths vs time since positive test, broken down demographically, I don't think it's possible to understand the full implications of this, but it makes me very cautious about inferring too much in to demographic data in a decaying phase.  It seems likely there's a strong demographic component to factors affecting the long tail.

Sorry, rant alert, this reminded me of something I'm quite upset about...

Coincidentally, the almost total agreement of these curves at around day 135 (early August) when deaths were low (around 10/day) is a highly significant refutation of the false positive PCR bullshit that originated from Carl Heneghan at Oxford University's Centre for Evidence Based Medicine [1], which has since taken on a life of its own in other circles.

• It's a shame a Professor at Oxford University is apparently so inept that they didn't compare the statistics - if their "theory" on false positives had been right, the deaths within 60 days of a positive test would have been 60/28x = 2.14x those within 28 days of a positive test.  But the ratio was much closer to 1.0, indicating that a positive PCR test was highly correlated with death.  It's hard to think of a more compelling statistical argument against false positives.
• The subsequent rise in hospitalisations and deaths, as well as the 28 vs 60 days data, makes it abundantly clear that their "False positive" take is both wrong and irrelevant.  It is beyond a shame and in to a matter of professional competence and integrity that they have not since retracted their blog post.  Since mid-September 2020, there has been unequivocal grounds for Hengehan to issue a follow up post which could, for example, read "Regardless of the accuracy of my Augsust 2nd post it is clear this it no longer has relevance, as PCR detections, hospitalisations and deaths are all rising exponentially, which unambiguously indicates a pandemic growth phase in action"

Edit:  My interpretation is that the long tail of deaths from the 2nd wave is super-imposed on the growth phase from the third wave.  I could be wrong in that and there could be more at work.  

[1] "COVID cases in England aren’t rising: here’s why", Carl Heneghan, Centre of Evidence Based Medicine, Oxford University, August 2nd, 2020.  Link not given to avoid raising the piece up the Google search rankings. 

Great news if this more sensible approach to determining genuine vulnerability happens. The Jo Whiley headlines today should help as well.

https://www.theguardian.com/world/2021/feb/16/two-million-more-people-in-uk...

https://www.theguardian.com/world/2021/feb/16/jo-whiley-offered-covid-vacci...

I would agree. Although we are at the point now where the range 28 - 60 days ago is the beginning of the 3rd wave. It would be interesting to know whether the demographic distribution of the tail differs much from the rest so that as the total number decays over the coming weeks and the tail becomes a larger proportion of the whole we know whether the deaths from cases longer ago should bias the demographic distribution one way or the other.

Hasn't the false positive theory been long ago debunked for anyone with a grain of interest and sense?

Yes but if you're a "prominent" scientist who spouts off, then you also have a responsibility to say "I got that wrong" when you make a public mistake.

Indeed it has. But it lit a fire in the misinformation brigade that’s not gone out yet, and I think the calls of “false positive cases” will again start rising as cases start falling.  A fire that would be easier for people to put out if the “leading academic” who started it all with a gishgallop and a bad graph would publicly admit that they got it wrong.  I expected that to happen at the end of September, and yet here we are.

It would.  I haven’t tried to get the demographic deaths data from the API yet but the wepbage link is using an API key; I’ll see if there’s a marching 60 day key....

yes, but there's loads of people with no interest or sense, or being paid to post about it. You see the same people posting about it again and again even when you've comprehensively torn their argument apart.

An update on the visualisation of exponential rate constants for cases vs deaths, and of the re-binned rates for deaths.  

I think that the data for ages 50-59 is becoming very noisy due to the low numbers; the others now seem to be stabilising and stratifying with the fastest decay at the oldest ages.

On the cases vs deaths plot, the oldest age ranges are standing out as decaying faster on both measures; perhaps a sign that even for 85 and above the vaccine reduces transmission as well as illness - or it could be that it moderates illness (but not transmission) below the threshold criteria for pillar 2 testing.  

Either way, this seems very encouraging - especially as the vaccine is now being rolled out to younger ages.  I think as the vaccine is rolled out to younger people, spotting the demographic effects on case rates will get harder and harder, as you get to ages that interact with other people more, where vaccination of any subset of those interacting people will lower transmission across the whole group, moderating transmission to all.  

Edit: By request, an intermediate set of dates are now marked on D1.x.  These make it clear that the oldest ages were behind the younger ones in time during the conversion to decay on both measures, yet now they're ahead.  This makes it clear the gains in the oldest ages are larger than it seems just from their final positions.

Latest on outdoor risks:

https://www.theguardian.com/world/2021/feb/19/how-the-beach-super-spreader-...

I’m looking at the provisional UK cases data for Monday Feb 15th on the dashboard with a beady eye. It’s too high for the current halving times and it’s still provisional.  Hopefully it’s just the effect of some noise and not a sign of things to come, but there are some persistently stubborn areas whose slower decay could start to show through at the top level soon.

Are these on biggish case numbers? As I said above Rushcliffe doesn't make sense but its only 35 cases a day which as an example could be just a local care home issue (the way agency staff move between homes to cover sickness absence). Lots of people in care homes are under 70 and not yet vaccinated (Jo Whiley's sister sadly has been hospitalised with covid now).

Good to see some media recognition of the abuse of a telephoto lens to  mis convey the situation.

Conversely, as the White House rose garden super spreader event shows, outdoors is no magic panacea and physical contact still has to be avoided between housesholds/bubbles, so the messaging needs to be clear on the limits to the protection conferred by being outdoors.  

It would be nice if the outdoors was generally accepted as then the media could focus their moral panic, er sorry objective reporting, on many other issues including those the article you link hilights.  Including traffic, insane parking blocking main roads, litter, nose, anti social behaviour and all the other aspects that are I think going to hit the countryside even harder than last year when UK travel restrictions (guidance) are eased in the coming months.

I very much started my stance from a precautionary principle over large outdoors crowds; the evidence seems to be stacking up that with social distancing between households/bubbles, it’s fine from a covid perspective.  

The top level number for the UK.  The most recent Monday is a 13.6% drop week on week vs a 30% drop over the week before.  The most recent Monday data is provisional and could still go up a bit more.

We’ll see what Saturday’s update brings.  Both the north west and the Humber estuary have been starting to look like hold outs this week, but I don’t really trust mid week updates as much as Saturday ones.

Good points. I would say its important to remember the Rose Garden White House event was partly indoors and many there were hugging each other on camera (think what they might be like off-camera indoors). It seems to be a US politicians blind spot: even Biden's inauguration had inappropriate physical contact.

On a different subject does anyone had any sensible theories on the latest React information on Primary school aged spread, as the data seems implausible given most kids are stuck at home:

https://www.theguardian.com/world/2021/feb/18/covid-infections-in-england-f...

The article mentions 25% at school. Comparing prevalence for the 75% home schooled vs 25% at school  might be useful for decisions on re-opening schools.

Possible:

1.  Those children at school are key worker children (so high risk of catching off their parents and then spreading at school (lots of health and care workers)).
2.  Whilst those not at school are at home, there will be lots of mingling:  close up and personal street playing (see lots of this) and in homes (childcare bubble workarounds - hear quite a bit if this)

3. Mid teens upwards: massive numbers have thrown the towel in. I think this is a much bigger issue than many realise - the high profile student parties are the very tip of the ice berg. 
4. many young adults have high risk/low paid jobs. 

Uber-pedantic point but is the US month-day date style on your plots deliberate? Trivial I know, but I find it disconcertingly difficult to interpret

Genuinely don’t get what you mean. 

The data for the 15th does make it look like things have flattened slightly doesn't it? I wouldn't expect it to go up significantly any more now, but I would expect the 16th to go up a little more yet, and that will then be on trend with the 15th rather than with the previous rate of decline. Not sure if it could be a short blip caused by some change in lags.

Below is a screenshot from the dashboard for England cases (this graph isn't shown for whole UK.) This gives you a pretty reliable indication of how much further you should expect the incomplete days to go up.

Edit to add, the issue appears to encompass all regions except for London, South West and maybe South East. It's not confined to the North West and Humber.

It is deliberate - it's an abbreviation of big endian Year-Month-Day which itself is a British standard (BS ISO 8601:2004) and definitely not US style (Month-Day-Year).  

It's just that the truncated part I use matches both styles.  I much prefer the big endian approach as my brain works that way - increasing date is in increasing order of the number for big endian dates.  I think the UK is slowly drifting towards Year-Month-Day, in which context the abbreviation makes sense.  I could add the year to axes labels to remove ambiguity, but they become crowder.  Perhaps I could just add it to the first tick-label on the axes and then to the first tick label in each future year (and I'm very much hoping that is limited to 2021).  

If you ever see anything with "middle endian" month-day-year formatting, please insult me until I fix it.  

I understand.  Somehow year-month-day is fine for me to process  but month-day causes brain problems!  I agree month-day-year is  just absurd.

Ha! I'm just so good at expressing myself.

From the linked article:

"The team suggested the relatively high prevalence in younger children could be due to a greater proportion of this age group still attending school. A survey found in February that nearly a quarter of primary school pupils were being taught in-person."

Comparing home schooled vs at school would tell you if it is an age thing or an at school thing.

Newly reported cases shading!  Nice; I stoped doing the daily downloads that needs some time ago as it was a largely pointless chore (for me) and the cases data become more interesting than the reporting.

The shading really drives home the commendable reduction in latency in the data (and presumably contact tracing from what I could infer from the reporting of the .xls debacle).  I've not seen much on this beyond Hancock noting the current performance in a briefing - I am very happy to see this improved.

There are lots of small historical additions gong back 21 days or so; they don't look like much but if they happen for each of 21 days, it could be a quite a bit.

We'll see what happens tomorrow evening when the weekend fully resolves including the delayed samples in Monday/Tuesday.  The most recent weekend's data follows a different pattern to previous ones without a clear decrease Saturday-to-Sunday. 

One source of difference between random sampling surveys (e.g. REACT) and symptomatic testing (e.g. Pillar 1 + Pillar 2 > gov dashboard data) is that the random sampling convolves infection with its duration.  I've not seen a demographic study on the duration of infection but that could introduce demographic effects in the REACT results during a rapid growth or decay phase.  

I think with schools the short period between likely reopening and the Easter holidays gives a chance to evaluate the reopening through symptomatic testing and through targeted mass testing, without committing to much increase in spread if things go worse than expected.

• If cases stay on their current decay, this is going to cause much less spread than the opening of schools in the tail end of 2020.  
• If cases cease to decay in some specific geographic regions, that might need locally reevaluating...

Got you. It does point to schools being a ‘place of spread’ (which may tie in with the point I made about key worker children).  Not what many folk (want to) believe. I suppose what we really need to know is: ‘are they catching it at school (ie how many school bubbles have closed with multiple cases)’

Ironically, and this really is an idle musing, but having key worker children in school could be worse for Covid than the other way round. Although vaccinating key workers * could really nail this.

* btw, had my key worker jab last week but yesterday got letter from GP saying ‘book a jab’. Must be my very mild asthma coz I’m still a young lad. Anyway, thinking of selling my extra slot on eBay just like a Tesco delivery slot ;  

Indeed - especially as some schools are coalescing multiple years in to one classroom as their staff have to do both in-person and on-line teaching.  Which goes on to build transmissive bridges between significant groups of key workers.  

I have re-opened all my case study LA areas on my phone from the dash board. What gets me is: how the SE/London etc has continued to see their rates fall to the extent they have overtaken the ‘usual suspects’ (they were twice as bad in January) and continue to fall at a good rate. Looks like the usual suspects areas really do struggle to keep numbers down. Worries me: at some point I fear another regional tiered approach because I have almost no confidence that the ‘usual suspects’ can keep rates very low. 

BTW, I also note the other ‘non usual suspect’ areas levelling a bit, it’s the big population areas that really concern me. 

My expectation is convolution effects would be more significant in adults. Equally the ratio of 25% in school must be less than adults working (as not all essential workers have primary school age kids). 

I agree there are definitely some specific LAs that have particular problems (Blackburn, Bolton, Knowsley, Middlesbrough off the top of my head.) They can get low but often don't. There was a guardian article a few days ago analysing theories for why more deprived areas are worse hit. Apparently part of the problem is lack of test/trace effectiveness as well as underlying risk factors (more front line workers, more large households etc.) It would be interesting to theorise as to whether this should make them tend towards a higher equilibrium level of cases with a given set of measures - they don't seem to have a problem dropping quickly from high rates but they struggle to get very low.

Inequalities and risk factors would be key. Test and trace etc not so sure. I recall Blackburn introduced its own bespoke system which seemed to work. 
Wigan is an interesting area: we don’t have massive areas of deprivation, have relatively low unemployment , very low BME community, yet one of the highest death rates in UK. 
A big taboo subject: how much of it is about behaviours and attitudes. Eg ‘I’m a tough northerner’.   It could have zero impact, but something is happening that isn’t simply about risk and test and trace. 

Not sure about that, my brother is a mid-30s primary teacher in London, he and his wife told me in about November before the latest wave there began that they and most people they knew outside of vulnerable groups were fairly law abiding but had otherwise just given up and accepted they were going to catch it. Pre lockdown 2 he had told me life in London felt pretty normal for several months. That definitely wasn't the case here! I'm sure the above has changed with the area being hit so hard in December though.

I strongly suspect wintertree's update tomorrow will show cases in Scotland tending towards a plateau. Or should I say glacial valley bottom? I think there's still time left for us to mess this up again and start climbing up the other side of the glen.

Finally something on which you are definitely (😀) wrong. For cultures that read left to right, Littlendian is clearly better. At least we can agree that Middleoutian is an abomination.

Another poster was suggesting this a couple of weeks ago.  I didn't agree then; looking at the provisional window I think you might be right now.  Likewise for Northern Ireland as well as the  apparent slow down in the English exponential.  If this is an effect corrected across all the nations, it's reasonable that Scotland would exit decay first as its exponential rate constant has been consistently less negative (slower decay) than other regions.

As always, cases ≠ infections.

It's possible that we're just getting down to only stubborn chains of infection, and that this lockdown is not hard enough to eliminate them - perhaps unsurprising given it seems softer than lockdown 1.0 and the "Kent" variant is more spreadable.  

Some still frames from my UTLA watch plot movie below; there's no "smoking gun" UTLAs behind a slackening.  The idea that there's some base prevalence that's harder to eradicate fits with the weird behaviour in this movie where case rates (per 100k) drop from a wide range of values to a common non-zero floor.  That floor's been bothering me for a while.  There has to be some source of information external to the case rates and the lockdown measures that lead to faster decay in the highest case regions.  The clear obvious candidate for this appears to be a separate sub-population with an intrinsic lower case rate - a buffer we're now running in to.   Speculation on my behalf, for sure.  If it's true, the breaks are coming on.  I make Scotland's case rate about 16 / 100k which puts it at the sort of level of this common "floor".  

Indeed.  I mean what the hell was anyone thinking with middle-endian dates?  It's just madness.

Anyone else noticed the vaccination rate as plateaued or even dropped? Looking ahead, this will limit the groups who get vaccinations, unless it changes or we stick with just one dose.

I know you don't like Zoe but they are reporting similar trends today, which I think is worth commenting on. The text says they are seeing the change due to a flatten/upturn in rates amongst younger working age adults with rates in older groups still falling, but I find that difficult to decipher from their graphs.

https://covid.joinzoe.com/post/rapid-drop-in-cases-slows-down?mc_cid=4832fd...

Possible slight drop - 1.8 mill done since Monday - but good chance it will get to nearer 3 mill by Sunday night. All going to plan on my opinion. 

Thanks.  Zoe is my least favourite source but it is the only symptomatic alternative to P1/P2, and some of my issues relate to it being (qualitatively to me) seemingly slow to pick up on things worsening, so when a negative message comes through that's time to sit up and notice.

You might see hints of it in their plot "Estimated rate of people with Symptomatic COVID across age groups" but I agree it's far from clear.  This was starting to show in the D3 plot I did last Saturday.  We'll see what it looks like tomorrow.

It seems like we're approaching a crunch point to test the claims made in the press recently of data driven policy.

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

And there's NHS staff on Twitter saying they've been told the 2nd Pfizer jab would be more than 12 weeks after the first.

https://twitter.com/JujuliaGrace/status/1362095163544375301

I posted the vaccine delivery schedule the Scottish Government were working to and it isn't any surprise the rate is slowing down and they're having difficulty doing the 2nd Pfizer jab in 12 weeks. 

https://twitter.com/fatweegee/status/1351583237454389249/photo/1

If you are a politician it is easier to say nothing and have an 'unforeseen difficulty'  later than be upfront about the consequences of your strategy. 

Given the React study also highlights basically cases rising in primary children and younger adults, makes Mondays announcement even more interesting. 
I get most of this stuff wrong, but I’m half expecting the tiered approach to kick as things slowing up in many areas but SE/London doing much better. 
 

Could indeed  be supply limitsThe second jab probably isn't really needed at all, so that may be fine.  

https://www.google.com/amp/s/news.sky.com/story/amp/covid-19-pfizer-vaccine...

There was some serious snow disruption for a few days I think.  I find looking at the vaccine data only at the end of the weekly cycle to be a useful way of looking at it.  I noticed that 2nd doses are starting to rise quite a bit - not enough to consume many first doses, yet.

Apparently, the Zoe data is being affected by people reporting vaccine induced side effects which they are looking to fix over the next few days.

https://twitter.com/marksgraham_/status/1362116545011470338

Interesting, thanks!  I’d been wondering about that in terms of P1/P2 but assumed people would test PCR negative as the RNA wouldn’t make it from the muscle to the nose, lacking a viable virus to reproduce and leave cells in.  But Zoe is different.

The data in the week before the provisional and unusually high data comes from the period of disruptive snow.  Unlike other weeks, the UK level cases data for that week does not rise above the 7-day moving average (weekends pull the average below week data).  I wonder if we’re seeing some testing shunted to later dates by snow, creating a false valley?

I will do a residuals plot in the next update and see if that illuminates anything.  

Hard to tell I suspect until a few more days of data come in.

Well yes, they may get lucky but having a plan which only works if you get lucky is irresponsible.

It still works with two, just takes longer for younger people to get it 

On Scotland....

Or, put another way, Scotland had maintained substantially lower case rates throughout, and so had less far to fall. Scotland has also had the example of catastrophe south of the border to contemplate. Our own troubles are bad enough, but with death rates peaking at about half the level in Scotland compared to England, perhaps we got a bit complacent. Likewise NI.

Yes, and the relationship between the two has probably been changing quite a lot recently, and perhaps differently between the regions/countries. Hopefully those advising the governments understand a lot more about the non-stationarities than I do.

Thanks for that! It's like inverse gravity. Or ping pong balls falling till cushioned and caught by an upward pointing fan.

"That floor's been bothering me for a while.  There has to be some source of information external to the case rates and the lockdown measures that lead to faster decay in the highest case regions.  The clear obvious candidate for this appears to be a separate sub-population with an intrinsic lower case rate - a buffer we're now running in to.   Speculation on my behalf, for sure.  If it's true, the breaks are coming on.  I make Scotland's case rate about 16 / 100k which puts it at the sort of level of this common "floor".  "

It will not be simple to separate your "sub-population hypothesis" from a "population behavioural response hypothesis". Certainly either (or most likely both) could result in the effect you demonstrate with the stills from the UTLA movie. The more scared the population is, the more effective the lockdown. 

Edit - to add quotes where the quoting system failed.

A case study  of English exceptionalism at Cheltenham

https://www.theguardian.com/world/2021/feb/20/virus-dispersion-hub-packed-r...

And a plea for better reporting of covid in the workplace:

https://www.theguardian.com/commentisfree/2021/feb/20/why-cases-covid-workp...

Almost makes one nostalgic for the days when ignorance and native stupidity were the most likely reasons for not taking Covid seriously. See also Liverpool v Athletico Madrid.

Thanks for the many thoughtful comments.

I agree - there are different mechanisms that could be behind this.  Way back on plotting #5, minimike and I and others discussed this [1] in terms of a similar effect where the geographically localised exponential rate constant was more positive when case rates were low, in a rising phase - the growth rate always appears to self-moderate.  People responding to the local situation is one plausible explanation - the more ambulances on the roads, the more friends and people in extended networks getting it, the more emails from schools about cases etc., the more people respond behaviourally.  The same effect could well be in play but on the other side of the axis.  If it is this, it suggests that the right advertising and messaging campaigns could make a difference nationwide.  

The amount of phenomenology visible in this data is incredible; I hope that people will be building their doctoral theses and careers around analysing the data once the worst of this pandemic is in the past - the lessons learnt by explaining everything that's going on will be incredibly powerful for writing the future rule book.  

I'd share my rule #1 but I try to keep my input to these threads forwards looking.

[1] https://www.ukclimbing.com/forums/off_belay/friday_night_covid_plotting_5-7...

This is the world's first fully instrumented pandemic and we are bound to learn an awful lot from it and develop a lot of useful technology beyond vaccines.  It's like the space program for control of infectious diseases and quite likely the main effect ten years out will be that many infectious disease are suppressed by the tools and knowledge we develop for Covid. 

I think the clear timelines of the data also present a clear objective basis with which to compare to some other processes.  Some will disagree.

As with your space race analogy, the technological response to this pandemic is spilling over in to related areas; globally biotech is seeing some of the strongest growth.  For all the death and destruction, the end of a lot of other diseases moves closer, sooner, as a result of all this I think.  It’s hastened the time when our problem will become “what will people die from?” - and it’s driving the wedge further in along the wealth lines that will divide access to some of this bounty.  In many ways the pandemic is accelerating many things - good and bad.  

Yesterday there was in article In The Guardian saying that outdoor gatherings did not spread COVID. This was based on the crowded beaches etc. 
 

Typicsl mixed messaging 

I don't think so.  

The transport situation to Cheltenham has to be considered - a lot of public transport (bad transmission risk), a lot of organised groups travelling on minibuses and lift sharing (bad transmission risk), a lot of peripheral activity to the main event with elevated transmission risk.  ,

Further, crowds packed watching a sport are very different to socially distance household bubbles on a beach, and very little was know about the spread back at the time of Cheltenham, except that it was out of control and exponential at the time in the UK, and that the same had happened and was happening elsewhere.

A key point of the article is that the beaches were not crowded, but that trick shots were being set up of all sorts of outdoor activities, using long lenses to misrepresent the reality.

I think there will be a lot of political pressure for that analysis not to take place. I wouldn't be surprised to see the data vanish in the near aftermath, on the grounds of 'privacy'.

Gather whatever data you can now, and keep hold of it...

Another alternative is for a lot of subtle PR campaigning and astroturfing based around a gross distortion of recent history, and a lot of misdirection by comparing what we did not to what we could have done, but to what others did.   The time was, children were asked "If all your friends jumped off a cliff, would you?.  Now it seems we say "it's okay, all their friends jumped off a cliff, too".  Before long we'll probably see attempts to cast the very complex path to failure as an absurdly reduced strawman.  

I don't know if I missed this in the news in December, or if it didn't make the press due to the worsening Covid situation.  It's a report from Parliament's joint committee on national security.  The cross party committee has more tory members than any other allegiance.

Title: Government failed to act on its security plans for a pandemic

https://committees.parliament.uk/committee/111/national-security-strategy-j...

You really are living up to your pseudonym aren't you!  Of course huge lessons and knowledge will be gained from this pandemic. Though, of course, whether governments of the day act fully on those lessons next time around is another matter.

It's not mixed messaging. Cheltenham was way more serious as it was partly indoors and strangers were in physical contact (packed 6 deep at indoor bars etc) when those beaches were more social distanced than they appeared or were portrayed to be (apart from maybe when the large numbers were forced back, much closer, in the Durdle Door helicopter emergency call out). Time and time again the research evidence indicates wholly outdoor crowds were much less of a problem than feared. The infamous football matches and the Rose Garden White House super-spreader event involved mixing partly indoors.

To be fair, Cygnus was buried.

We now live in a world of Designer Truth so far as democracy is concerned

The knowledge will only be gained if we find out where we went wrong, like Grenfell, not investigating properly allows disasters like that to occur.

£1bn spent world wide on research into viruses and vaccines, pre-covid, apparently.

Before this there had not been a pandemic for a century. It was something which governments could conveniently sweep under the carpet without the public noticing. I really don't think that is going to be the case after this.

I am sure we shall.

Well the vaccine research certainly seems to be paying off.

I think that too was partly luck...SARs and MERs had the capacity to become that with mutations.

Is this not a case of "making your own luck"; do enough reseach and related development, and the chances are you will have a lot of the answers sitting there ready to take off the shelf.

Well that didn't work with this latest coronavirus did it? It was possible to see a significant probability of something like it coming (as the big SE Asian economies did) and yet we buried Cygnus and ignored its lessons. The UK and all its fantastic innovative spirit, was the worst hit, and the US not so far behind, due to political incompetence.

I was referring to vaccine development in my post at 12.59.

And referring to Cygnus and so on at 9.49.

The first couple of months of scientific and financial support of vaccine development happened despite the incompetence of our government in dealing with pandemic risks. The hardest work was done when they took over. Having such successful vaccines took most scientists by surprise...more luck. The more we roll the dice the greater we risk a big loss. Add up rolls on Cygnus,  herd immunity, the unforgivable delays, the PPE, sending the infected back to care homes and by 'running hot' (which maximised the chances of worse mutations), as bad as we have done, it could have been even worse but for luck. I'd rather our politicians stopped gambling and took pandemic risk more seriously, as is the norm in those big SE asian economies.

Maybe you have misunderstood me. I am not in way way trying to defend the UK government over their handling of the pandemic. But I do think lessons will be learnt from their mishandling. 

Well you’d hope so...

There is little evidence of real learning on the political front, which is where it matters most. I'd love to be proved wrong and see school returns phased tomorrow. Yesterday on the news Robert Jenrick said they made the right decisions, at the right times. Has a single minister said lockdown is not a trade-off between public health and the economy (ie the truth is a failure to respond quickly to rising covid levels leads to more economic damage)? Science has improved and a few scientists (who should have known better) were woken out of blundering methodological muddles with masks and modelling.