UKC

I feel like I should be writing about the weather today, not Covid, it's much more interesting.  This morning we were walking up the south side of Weardale on to the tops in t-shirts in glorious, windless sunshine watching storm clouds try and boil up, then fold over and collapse.  3 hours later we were watching the trees turn white under some of the biggest fairytale snowflakes I've ever seen.  There's a hard frost tonight and the snow was hanging on over 300 meters so there could be a late season winter walk tomorrow.

On the Covid front, last week's update saw a welcome return to decay in cases for all nations.  With the additional data on testing by type or England, we could see that this was partly because the school-related LFD testing ramped down and partly because genuine faster decay set in to PCR cases, with likely contributing factors being the schools shutting for Easter and some unseasonably warm weather, the later having clearly departed us now.

My plots - which try and smooth over some of the noise in the data in a meaningful way, show Wales and Northern Ireland returning to growth in cases, with England and Scotland levelling off.  Looking at the most recent provisional days on the government Dashboard (not used in these plots), I'd say growth in cases is coming to all 4 nations.  The data for England by test type (in a later post) shows that PCR cases are still falling so I presume the provisional growth is driven by LFDs being ramped up in preparation for the return to schools.  By proxy I suppose that could apply to the other 3 home nations...?

Hospital admissions for England look to be rebounding slightly in the last couple of days in plot 7.2e; looking at the raw data it could be that they've stagnated and a bit of noise makes it look to be rebounding.  We'll find out in a few days.  Their stagnation ins't much of a surprise, as PCR cases borderline stagnated during the school term.  Hopefully they'll do some more convincing decay by next week, given the recent decauy in PCR cases.

Plot 9e shows cases halving a lot faster than admissions, but much of this fast halving was driven by the collapse of additional LFD numbers as asymptomatic testing wound down after school term, and we wouldn't expect that to translate in to falling admissions.  

Deaths continue to halve faster than everything else, as the vaccine continues to decouple death in particular from infections.  The curve is getting noisy as the absolute numbers dying become very small.  With a bit of luck it's not going to be long before the first day without any reported deaths; almost unthinkable a few months ago.

The regional plots.  I've censored values from the rate constant plots 18.1 and 18.2 where the absolute numbers they derive from are less than 10/day, as the curves were being driven by noise.  I might start to use some adaptive window for the exponential fitting that increases in size until a minimum number of results are used.

With deaths having become so low at the regional level and with the behaviour of cases curves dominated by many LFDs being turned off, I don't think these plots add much this week.

Plot 22 is close to reaching the origin. Various posters with friends and relatives working in hospitals have commented on these threads and others over the last few months.  It's clear healthcare continues to feel the impact of this, not least in the number of staff suffering from various Long Covid symptoms.  I've seen a few interesting things on the gut microbiome and Long Covid; I wonder if the intense research interest in this is going to result in some breakthroughs for ME/CFS as well?

I have added red circle markers to the end points on plot 22r.  All regions are now close to closing the loop and being bereft of hospitalised Covid patients.  There's still some way to go before occupancy is as low as it was last summer, but we're getting there.

It's notable that deaths are halving consistently faster than hospital metrics; I take this to mean that the 1st dose vaccine doesn't always present infection and acute illness - particularly in older people - but that it does make a radical difference to their survival chances.

No surprises in the demographic plots for England - everything continues to fall.  It's looking pretty conclusive IMO that schools being eventually open leads to a rise in symptomatic (PCR) infections across all ages.  The good news is that we were still holding R<1 for PCR cases before the Easter holidays, and when schools go back, a lot more working age adults will have vaccine granted immunity, which is going to knock a chunk out of R, so hopefully it will still remain <1.  I'm cautious though given the rising case counts in the provisional edge of the government dashboard; I'm hoping that's mainly LFD related.  We can re-visit that next week.  There's no sign of the plateau I thought might be emerging a couple of weeks ago in deaths for younger adults in plot D5 - good.

Punter_pro linked to some plots from John Burn Murdoch on plotting #19 [1] that are something akin to my D5, but for cases not deaths, that show the normalised curves separating from their shared peak, with the most decay for the oldest, strongly hinting at the vaccine also preventing transmission of the virus.  We've seen this in my D3 plots until the LFD data sent them all over the shop.  JBMs are clearer.  Not wanting to be outdone (hard, JBM is the standout best person I've seen for visualisations of Covid data in the press), I've made plot D6 which shows the ratio of cases in early January to cases now.  Other than the dramatic changes for older adults where the vaccine started first, the stand out to me here is the shape in secondary aged children - it's barely got better since January.  A curse from the "Kent" variant perhaps, and perhaps an indication of why schools reopening apparently had such a strong effect.  

[1] https://www.ukclimbing.com/forums/off_belay/friday_night_covid_plotting_20-...

The grab bag of left over plots.

Looking the fall in cases week-on-week gives a view of the halving time that's not obscured by the "weekend effect" in sampling.  This is shown for symptomatic PCR cases ± asymptomatic LFD cases.  The week of rapid decay after the closure of schools for Easter seems to be petering out, but odds are there's sampling effects with under-sampling from the bank holidays and over-sampling after that spreads data points out more than normal.  We'll see what a few more days of data does.  Then schools return and the next wave of reopening happens.  This really is a race against vaccination in terms of keeping R<1, with the adults aged 50+ largely now vaccinated and coming in to their immunity I hope we'll see more moderating effects to counterbalance the reopenings.

The vaccines plot shows a big, recent lull around the bank holiday weekend.  Now, we're back to normal rates - but not as high as the few weeks before Easter.  There's a big switch to second doses.  I've only eyeballed it but it looks to me like we're running more like 10-weeks between doses than 12.  I can't analyse that properly as the data isn't longitudinal but I can make some assumptions to estimate the interval; it's on my monotonically growing TODO list...

The variants plot.  I'd mentioned a few weeks ago that the SA variant is likely largely being detected in foreign arrivals.  Longsufferingropeholder pointed out on plotting #19 that JVT confirmed this in a recent briefing.  The other strain I'm plotting is I think probably coming from community transmission.  We don't know how targeting the sequencing giving this data is though, so we don't know that community cases are rising exponentially like this curve.  Something else discussed on #19 was that this is cumulative (total) variant cases I'm plotting; however if it's rising exponentially then so are new daily cases.  If this variant has been rising exponentially during lockdown, that's bad juju.  If the rise is down to the sampling modality, not so much.  There's another local variant cropping up in larger numbers.  I'd quite like to hear these discussed more in a briefing.

It isn't. The dashboard presents the PCR/LFT split for positive tests for all dates including the incomplete ones. There is not yet any notable increase in LFTs up to 9th April, the change is in the PCRs. I think it is just recovery from the Easter bank holiday lull which itself is also very obvious in the raw data.

https://coronavirus.data.gov.uk/details/cases?areaType=nation&areaName=...

If we make an assumption that the effects of the Easter bank holiday are dealt with in the cases data by today ie 10th April (not certain, but looking at the incomplete dashboard data, probably about right) when do you think the data feeding your exponential curves and outputs will be mature enough (ie, not overly affected by the BH) for us to have a stab at understanding the effects of the opening up on 12/04?  Couple of weeks?

Zooming in, it looks to me like there's a big increase in lateral flow cases on the 7th/8th/9th compared to the week before, although the PCR cases themselves have a bit of rebound going on I think that's the normal day-of-week sampling effects with a bit pushed a day further right by the bank holidays as you say.  

We're in to a situation where the true decay between weeks can be less than the false decay within a week from day-of-week sampling effects, which makes interpretation of the leading edge really guggy and subjective.   The week-on-week plot is the best way to think about this I reckon; looking at the PCR only cases on the RHS:

• Big pink gap on Monday ⇒ bank holiday weekend under-sampling
• Small pink gap on Tuesday ⇒ cases displaced from Monday
• Small pink gap on Wednesday (still lightly provisional) ⇒ decay of PCR cases faltering?

The good thing about cases being low is that cases are low.  The bad thing about cases being low is that it's harder to figure out what's going on other than by tuning in next week.

It's going to be really hard.  If we limit ourselves to looking at the PCR data which is only available separately to LFDs for England, then a couple of week seems reasonable - but schools return which likely has a big effect, and there's a material change to vaccination levels in adults aged 50-60, and there's the unlocking happening on the same day as schools.  I think that without longitudinal studies it's going to be another case of staring at the digital tea leaves and saying "phew, that was close" or "oh dear, oh dear" rather than apportioning blame or salvation.

A lot of this is way beyond me but I do like seeing the graphs look like a ski slope. 

That just means I'm not explaining it well enough.  Feel free to ask questions.  

Just so long as they don't have a ramp for a jump at the end...

Photos from staying local in Weardale today.

I'd be fine with a ramp if there was an end.

Exactly, ramps and jumps I know a little about and with enough down hill momentum even the smallest ramp can give some good altitude. 

I would love to believe that the government have a similar set of numbers to you and are watching on a week to week basis. We have unlocked X and the R rate is going up, let's look at that sector and see what we can do. 

Fair enough, I looked at exactly the same data and saw no change in LFT (comparing the total dark blue + yellow bars between the different days in the graph, noting that dark blue turn to yellow as tests are confirmed with PCR.) Either way like you say we'll know what's happening by next week, my bet is it's not a real increase at all, but that by then we'll only really be interested in what has happened from tomorrow onwards anyway

I did a plot adding them rather than eyeballing the dashboard one.  

For the last few days in the dashboard data (Tue/Wed/Thu) and the proceeding week, Tue and Thu are similar but Wed is a bit lower in the most recent week; not much difference though.

Yup, I think this is just what the data looks like now with the day-of-week structure.  Unless cases start rising or falling quickly we're doomed to always be a lot more uncertain on the leading edge.  I still think a fair bit of the jitter is to do with test kit specimens being dated by their receipt in to a lab rather than by date of application combined with the Sunday post in particular, but I've made no progress in definitively proving it; a shame as finding a streamlined workflow for using sample dates would improve the data quality - and therefore understanding of the current situation - taking more latency out of the policy based control loop.  I can also see why the logistics of it could just be too complicated to work at scale.  I thought maybe a QR code on the kits you scan to go and enter the date the sample was taken, with some sensible bounds enforced by the system behind the scenes.  

I think the advisors are very good at explaining things to their audience.  They seem to favour different forms of presentation, but many different forms are equally useable so long as you can fit the explanation to the target and bring across the key messages.

Another week and worldwide stats look worse again although some of the previous major hotspots at last seem to be shifting away from growth (Brazil and most of eastern Europe). Turkey and Argentina are two countries not getting as much coverage for worrying growth as maybe they should.

Yup, it's not looking great beyond our borders, but on the whole it doesn't seem to be getting drastically worse, although I'm not sure how reliably any of the data coming out of Brazil can be by now.  

Toerag posted a link to a site visualisation sequencing results on the last thread - https://covariants.org/per-country

Sequencing results need to be interpreted with a lot of caution for various reasons compared to headline case numbers.  Still, the notable thing about Turkey is that the majority of sequenced cases are the SA variant; it could be that that variant happened to be in the right place a the right time to be amplified by the current wave, or it could be that it's being targeted by sequencing, or it could be bad news.

The sequencing results are something of a rabbit hole; so many different ways to interpret it, so much information not provided to contextualise them. I've been wondering about the ORF1a:S3675- deletion - we've not seen anything like so many changes sequenced from the non-spike proteins.   I wonder what that deletion does?  Tweaking the error rate during reproduction?  Improving efficacy against innate immunity (and therefore some emerging therapeutics?). Or was it just in the right place at the right time?  Understandably the people who are unpicking this stuff probably don't have time to explain it to lay people like me right now...

Canadian ski resort with P1 cluster.

https://www.theguardian.com/world/2021/apr/11/canada-ski-resort-p1-covid-va...

Well, there's a surprise. Who would have thought ski resorts could be a problem? It reads as if Canadian Public Health has thrown the towel in... hopefully not.

Hopefully Canada will get on the Red List quickly...

Although this is inevitable at some point it just seems a bit too early in the global vaccination cycle....

I'm just not sure the government listen to them. Every step so far seems to have been let's let things get too far before act just in case a miracle happens and things sort themselves out. 

As an example, it's only in the last couple of weeks that hands, face, space changed to hands, face, space and fresh air. It's been known for a long time that Air flow is important for reducing the viral load indoors but every other day I go in offices that are buttoned up tight and people look at me daft when I ask to conduct the meeting outdoors. 

Looks like it's down to ~100k per day 1st doses. Capacity seems to be rapidly being taken up by second doses.  This does not bode well for increasing the number of people with some degree of immunity in order to win the race you describe above, and looks like it's going to stay that way for some time given the number of 1st doses given in January and February. It looks like they're vaccinating 50yr olds in England now with Wales and NI into the 40-50 bracket. Given that there's about a million people of each age in the UK yet to be vaccinated, so it's going to take 10 days to move down a year at the current rate or 3 years a month.  I can't see how they're going to his the 'all adults one dose' by their end of July target at the current rate - that's "30 year's worth" or 10 months at the current rate. Will be interesting to see the next week's data.

By getting a little ahead with second doses now, May can be a month of smashing through the next few age groups. It's pretty well explained here.
https://news.sky.com/story/the-second-dose-debt-why-millions-of-people-will...
Unfortunately I've not found such a succinct description on any less cesspooly website. Recommend private browsing and reader view.

[...]

I agree with regards advice on air, amazing it's only just happening.  I've ranted before about the number of busses I see with no windows open, and how the bus firms have had a whole bloody year to modify the windows to be unclosable.  

In reply to HardenClimber:

Ski Resorts?  It's groundhog day all over again.  It's intensity disappointing how little anywhere appears to learn from other places until they get hit directly.

The second doses are going ahead faster than first doses; eventually the gap between jabs will be smaller than is likely optimal for immunity, and they'll presumably switch back to first doses again.  In the mean time, now is the time I've wished I was older rather than younger since problems getting IDd for beer became a thing of the past.

Thanks as ever. I don’t think we’re into growth just as yet. This week’s data is tricky due to the Easter weekend but on the whole the case numbers are still falling. I should think growth will return after the 12th - the question is how fast it will be. The fact that some schools are still on holiday this week should help to mitigate it a bit. Hopefully the growth will be slow given that the high risk indoor socialising (both at home and down the pub) won’t be permitted yet. It’s set to get warmer again by next weekend, which should encourage people to abide by the rules.

Looking at the dashboard map, there are vast swathes of the country where most MSOAs are white, with occasional light green. This includes some large towns / small cities. The problem areas are mostly in Yorkshire, plus a few other pockets (Leicester is on, again). I know people would hate it but I think there’s merit in considering a tiered unlock, particularly for the next unlock round schedules for the 17th (or else push it back for everyone, depending on how things go). 

An alternative is to saturate persistent trouble areas with messaging, enforcement, sequencing, targeted vaccination and the kitchen sink.  Many carrots rather than a stick, led as far as possible by local community and public health resources, endowed with extra funding and support.  It’s never to late to further address the underlying social inequalities.

Tiering is not I think going to improve public buy in.

Well you could do that but we know that’s not going to happen... Tiering won’t be good for morale and some people might ignore ‘don’t visit your friends / relatives indoors’ but there are som easy wins which are more enforceable, such as no indoors hospitality, cinemas, gyms, etc. It would have to be combined with proper financial support of course.

Above all, they need to sort out financial support for those who are meant to self isolate and can’t WFH. It wouldn’t even cost much now that numbers are low. Say you have 100k people self isolating at any one point. At £500 a week, that’s £200m a month. Less in practice as some could WFH. Hardly anything in the scheme of Covid spend but could pay significant dividends.

Re second doses. 7.5m done to date. That number of first doses were done by 27 Jan, so running at 10.5 weeks currently. They will need to do another c. 12m in the next 4 weeks as first doses were up to 20m by 27 Feb (due to the 15m by 15 Feb push). That’s 3m a week, so they will need to do over 400k a day on average. Doesn’t leave much for first doses, as we’re seeing at the moment.

They do have a week and a half of headroom to play with and I suspect the 12 weeks is more important for politics than medically. Still, it’s clear that we won’t be seeing much by way of first doses for a few weeks yet. Albeit even 100k a day adds up over a month. 

Thanks again for another update, the next two weeks are going to be very interesting...

On the subject of data, a poster on Reddit has put together a visual 3D map for England of case rates by MSOA. The Kent variant explosion at 0.26 is terrifying...

https://www.reddit.com/r/CoronavirusUK/comments/morf1b/a_3d_map_of_case_rat...

Also, the chalk research from last year has now been peer reviewed, I had completely forgotten about it. Chalk has been found to have an immediate 99% efficacy against COVID 19 on surfaces.

https://www.abcwalls.co.uk/chalk-research/

Friction Labs liquid chalk vaccine due for approval in September!

Some of my mum's friends have started to get invitations to book 2nd doses this week. All of them are in their late 60s and had/booked their first dose through their GP rather than using the internet and/or a mass centre. The gap between jabs for them will be 8-9 weeks. I'd be interested to know if there were some areas that were running well behind or if there is a whole group who are closer to the 12 week limit, eg care homes. Or, perhaps, the people with bigger gaps are those who booked second appointments when they had their first? Either way, there are some people having 2nd doses well ahead of the 10.5 week average.  I think a policy of accelerating 2nd doses this month before many of them get close to the 12 week "limit" is a good one to mitigate the impact of further supply bumps.

Where I used to live in Derby they are still vaccinating people in their 40s at the moment, I think they have done quite a few now.

But ~11 weeks after these low first dose days we'll only have to give 100k second doses a day and there are 16 weeks till the end of July. Extremely noddy maths warning but assuming 11 week dose gap, 100k 1st doses a day for the next 10 weeks (including last week in the low first doses) then 2 million a week for the following 6 weeks (on a supply of 2.7m a week but with 700k second doses) then this gives 19 million more first doses before the end of july with 20.5 million remaining un-firstdosed adults at the moment so less than one weeks shortfall.

Obviously lots of potential issues affecting this both ways with production, new vaccines appearing, plus the bumper 1st jab weeks last month to cover etc etc but it is certainly possible they will achieve it on only a moderate increase on current supply.

Obviously we want as many people to get vaccinated as possible, but in terms of the Government target, it is about offering the vaccine to everyone. I think in practice take-up will fall with age* and as a result the number of doses required to meet the target will be quite a bit less than the theoretical maximum.

* Eg in toerags ukc polls in the pub, there is a clear downward trend in intended take-up rate with age from 90%+ in the higher age groups to ~50-60% in the under 20s last time I looked. Very small numbers, but ukc isn't exactly full of anti vaxxers.

I think that's likely; perhaps the single jab J&J vaccine will improve that.

If take up does fall much with age, it  raises a complex question for me.  As an early 40 something, am I better protected by getting the vaccine sooner as a result of poor take up by others, or am I better protected by getting the vaccine later as a result of better take up by others?  

Given the efficacy of all vaccines at preventing serious illness I'm probably better of in the short term getting done earlier as a result of others being hesitant, but the long term consequences of significant hesitance could come back and get me through variants.  The long term consequences will stretch on far longer than the a few weeks or months difference to when I get vaccinated, so I think even from a selfish perspective I'd rather I didn't raise up the queue due to significant hesitancy.

That sky article I linked makes complete sense of all this stuff.

In reply to wintertree:

In my mind this is the only really credible pro for vaccine passports. Even though they can only ever be a tenable proposition for a few months, they would move all of those "I will get it if needed it for a vaccine passport" votes into the yes column.
Worth implementing? Probably not. Worth bluffing the possibility? Almost certainly.

I think the gap is a bit of a post-code lottery with other factors thrown in too (Pfizer or AZ, central booking or invited by GP, etc) - we (Pfizer) were invited 6.5 weeks after our first jab so had a 7 week gap yet our much younger neighbour who's a dentist (AZ) is about to have the 2nd jab nearly 12 weeks after the 1st.

Go figure.

I presume that eventually someone will be able to crunch the numbers to provide gap v effectiveness analyses because at the moment we seem to know SFA about what the best length gap might actually be.

There is a UK based trial ongoing for this looking at different gaps and also using different permutations of vaccine for first and second doses which may be particularly interesting for the carrier virus based vaccines.   

Stories breaking today about this being tried in China. Not that we'll a) hear or b) trust their findings.

My MIL was participant No. 1 in this trial. I'm not sure if she has had the second dose yet, I guess the trial will have to run for quite a long time.

We chanced our luck at an almost empty vaccination centre at the weekend (we'd be in the next group but need to travel to somewhere a bit corona-riddled - permitted reason)  and they jabbed us.  Officially here they're done with the over 50s but GPs aren't allowed to start sending the over 40s even though there's plenty of capacity.  People in their 40s seem to be being called up elsewhere though despite the current guidelines. 

Good to see you worked out what my polls were for .  I ran them on our local FB 'have your say' group and got the following:-

50+ 20yes, 3 for passport, 5 no, 2 wouldn't say

40-50 56y, 4 pp, 15n

30-40 25y, 4pp, 6 no

20-30 54y, 9pp, 4 no, 2 wouldn't say

u20 7y, 4pp, 4 no, 2 wouldn't say.

There's thousands of people in the group so I'll bump the polls this evening to garner more responses. The 40-50 group was quite an outlier and it'll be interesting to see if that changes. Quite a few anti-vaxxers / conspiracy theorists / anti-government ranters in the group.

Updated week-on-week decay in English PCR cases...

I need to sort the x-axis out to be much simpler labelling; probably have a simple "date" axis and have the data points on the centre of the 7-day period over which they're comparing falls.

The last few days of data saw the decay of PCR cases really falter, the data for last Friday - still slightly provisional - has almost no decay.   Funny how we've gone from very fast decay a week before the return of schools to almost no decay by the Friday twice in a row now; again it also corresponds to the end of an unseasonably warm spell.  What is coincidental and what is causal is beyond this data and my limited addled brain; it'll be interesting to look at the demographic data as that emerges by the end of the week.

It seems not impossible that cases are going to tip over in to growth in the data for later this week with the reopenings and the return of schools from the holidays.  Policy in England is no longer explicitly to keep R<1, but having a return to growth is still something I hope (hoped?) to see avoided.  In favour of keeping us in decay is the vaccination program having largely completed down to 50 years of age and nudging in to the 40s in various parts of England by the sounds of it.  What happens to case rates over the next month is lap of the gods stuff now I think; still I hope that any return to growth will see moderate doubling times and that the ongoing vaccination will continue to push things towards decay before the next scheduled reopening.

An updated plot.  This estimates the time between the second dose given on the x-axis date and when its corresponding first dose was given.   This is done by matching up the cumulative numbers of first and second doses.   It's a plot version of an estimate I think Misha gave up thread. 

This isn't meaningful for early weeks of the plot as the data is available by-day but just lumped in on the first day of data...  

We're currently running to an estimated ~10.5 week spacing between jabs on average.

Son in law (30, no other factors, in N London) has just been invited to get a jab - AZ - they're trying to decide whether he should or not.

As numbers fall it becomes virtually impossible to avoid wild fluctuations in R which lead to a series of ‘returns to growth’. Essentially it will become useless as a measure as cases tend to zero. 

You could send them this to help them decide:

https://www.newindianexpress.com/nation/2021/apr/03/half-of-covid-patients-...

Not much corroboration but I've heard the same story elsewhere - I can't remember where, possibly a piece on the BBC news. Seems to be a trending story in India for obvious reasons.

Indeed; we're there with deaths and recently hospitalisations on a regional level I think and we're approaching that point with deaths at national level.  There's still some mileage in it for cases however and I think what we're seeing now is real, not noise.

As the noise is most definitely non-Poissonian in terms of having a temporal correlation and in terms of scale, I've failed to put any meaningful quantification on this however.

Important also to remember that these are cases, not infections, and that there's all sorts of stuff going on around those.  I did wonder if people are more likely to go for a test when they're planning to go in to work (the run up to the return to schools) than when they were effectively isolating over holidays or school closures.  

I was wondering on the best way to handle cases dropping more.  I think the week-on-week change gives am much better measurement that exponential fitting to raw or deweekended data.  I was going to try rate constant measurements for 1-week, 2-week and 3-week intervals , with the later retaining more meaning as cases decrease, and using some noise-weighted approach to blend them to give a smooth transition (Kalman filter?).  It all seems a bit excessive though; perhaps just reduce the frequency of these threads to once every two weeks and do a 2-week comparison...

https://www.ndtv.com/world-news/brazil-now-has-more-young-than-old-covid-19...

"Brazil Now Has More Young Than Old Covid Patients In ICUs
The number of people aged 39 or younger in intensive care units with Covid-19 in March rose sharply to more than 11,000, or 52.2 percent of the total, said the Brazilian ICU Project."

Re: monitoring case rates. I would keep things simple. It is easy to conduct an additional check for random noise affecting the English data by comparing different regions; if several regions are doing the same thing then you know it's either something systematic to be understood or a real change.

I thought you had evidence your data showed the doubling time dropping during August last year, when cases were still low - that's sooner than people realised there was a problem from looking at the raw data, so I think it's still worth producing, it might just need more careful interpretation.

I know I'm repeating myself but I really think the trends seen over the last week are more likely to be driven by the Easter weekend than anything else. Case rates dropped a lot at Easter weekend and then rebounded for 2-3 days. Come next week, I think it might be better just to produce this chart completely excluding the data from 02/04-08/04 inclusive. You'll be able to see if there has been a change in doubling time from just before to after that Easter weekend period better like that than by trying to work out the impact of under sampling, I think. It would suit this chart fairly well if explained.

Yeah interesting, possibly a combination of high viral load due lack of effective control measures and a more efficient virus variant?

I've never really understood what it is that drives the young / old disparity in immune response; it seems almost as if the more mature and trained your immune system is, the more efficiently it kills you.

I was coming around to that view, but I've swung back the other way.  New Years Day only displaced cases forwards in time by 1-2 days but we're seeing an effect out to the Friday (+4 days).  Also, the most the weekend depressed sampling by was from a decay fraction of ~0.65x to ~0.55x, but now we're seeing it boosted up to ~0.80x and even 0.90x on the still-provisional Friday.  They're much more give than take.  So I think there's a combination of factors at work, one of which is the bank holiday sampling shift, but there's more going on most likely.

We'll find out in a week - by which time as you observed we'll be back to pondering the leading edge...

Indeed; although I wasn't doing these plots back then so it was retrospective.  It's much easier to spot things going wrong than right when case numbers are low.

For my own interest I've carried on plotting the decay and joining the dots by day of the week, like the proof of concept I linked a week or two ago. I like that view, but like all of them it's only one way of looking at it, and like all of them it's getting really messy.

You should post them - the more ways of looking at it the better IMO.

Also, it'll distract me from sitting here staring at the printers.  In to day 8 of a long parts run and one of them has just started churning out balls of spaff the moment I go away...

I meant to say on that plot - you might limit it to showing the two days that are conventionally the low and high outliers, and then just scatter plotting any days that fall outside the region bounded by those two curves.  That would convey the key points and flag up any gross exceptions.

It’s a hypothetical question you don’t need to worry about. Just get it when it’s offered... What I mean is, life’s too short to worry about hypotheticals (I think). Having said that, I tend to agree with your view. 

Anecdotally, someone at work who is no more than 30 was off work with Covid for about 3 weeks and then suffered from brain fog and couldn’t work properly (even though he wanted to) for a few more weeks. Also anecdotally, I’m 40 and my side effects from AZ amounted to a slight temperature (cured with paracetamol overnight) and a slightly sore arm for about 6 days. 

Agree that Easter is messing things up. Today’s reported cases were higher than I expected. Residual reporting / testing delays or is it ‘real’? Hard to say.

I suspect this week’s numbers will be neither here nor there. Next week we’ll start to see the impact of today’s reopening and return to school (noting that some schools are still off this week).

I agree with you except in that we will definitely see an on-paper increase this week because LFT testing of school kids has now restarted in areas where the Easter hols are over. By the end of the week we could start to see if there is a (fast) increase in positives from PCR testing due to relaxations.

Mucking about with testing, then not testing really doesnt help us figure out what is going on. Testing should continue through holidays. I guess it's the difference between the macro need and the micro need; population testing vs individual testing.

45-49s eligible now .. 

http://ukwalls.epizy.com/wtr/plots.htm

I feel your pain

I tried, very briefly, and failed to make this look any good, so gave up.

or did you just mean like this? 
http://ukwalls.epizy.com/wtr/wintertree_special.png

That's the ticket.  I was going to only show points outside the bounding lines, but I think it works as well - or better - with them in, the human eye can rapidly segment them with just two lines drawn.  This way makes makes it much easier to see the envelope of the decay and the modulo-7 structure in x-axis, as well as the radical changes to that structure in recent weeks.

Is that PCR or PCR+LFD data?

I've only sort of skimmed the discussion, is the idea to effectively have 7 data series (one for each day of the week) and show how they behave as a group - hopefully all fairly similarly - rather than doing all the de-weekending stuff. The assumption being that most of the time the same reporting "biases" occur on the same days of each week.

It's just newCasesBySpecimenDate from the csv download. I haven't delved into the api (and wasn't planing to).
Edit: Whole UK, that is

That's what I've done, more out of laziness than any consideration of merit. I find it straightforward to look at that way, but it is very wiggly.

Exactly; that’s what both my take and lojgsufferingrooeholder’s take on it do.  I prefer this as can extract the same measurements (rate constant or characteristic time) not spoilt by noise from the weekly structure without needing deweekending and more filtering.  

Exactly.  Although as the linked plot from LSRH shows, that assumption blows up a bit with the recent steps up, down and up again in LFD testing and cases...

That includes all cases. The old biases on particular days that we were used to are shot. LFT seems to peak on a Sunday but also midweek to a lesser extent, there aren't enough weeks yet and it seems inconsistent.

You can look at PCR only data for England (not whole UK) by downloading "newCasesPCROnlyBySpecimenDate", and that would allow you to do the test you are after with that graph. PCR data is still following the usual trends with the exception that bank holiday figures were low so the Monday was lower than the following couple of days.

Not after any test in particular. It's just how I plot the cases data when I feel like having a look.
There are loads more ways one could split the data out after filtering by nation, but I mostly can't be arsed/don't have time. This is just a quick look I do from time to time, that's all.
Yeah, probably doesn't hold up to scrutiny across the periods where schools went back/ LFDs came in/ schools broke up/etc. etc.

Edit - England only PCR, replete with all the mistakes I've made by banging this together in extreme haste, looks like this: http://ukwalls.epizy.com/wtr/SidH_special.png

Yeah, PCR numbers will be key for comparability but even then Easter messes it up. Although I hope LFTs will have dropped after the holiday period due to less infection overall.

Anecdotally, the bars which were open last night round here (central Brum) weren’t that busy and many have stayed closed ( it enough outdoor space). Did a round robin on my way back from the wall (which wasn’t busy either). Not surprising given it was pretty cold. There were more people around at lunchtime today when it was much warmer but still not super busy. I suspect any rebound will be gradual at first and the case count by specimen date won’t really kick off till the weekend.

Also in many areas of the country cases are so low now that it will take a while for any meaningful uptick. Hopefully. We shall see...

I like Sunday dropping through the floor... if only!

That's the problem with plotting cases by specimen date up to a day ago. That point will come up. That plot needs a whole new set of caveats on top of wintertree's usual ones.

There've been a few discussions over the new variants and community transmission vs border detection.  

An open question has been "how large can community transmission get before it's picked up by sequencing following on from test and trace?".  

It looks like there's been sustained community transmission in parts of London that once found has produced 80 confirmed and suspected cases so far.  With the timing coincident with a relaxation of restrictions I don't think the genie is going back in the bottle, and it's possible there are more, smaller out breaks elsewhere in the country that haven't yet hit sequencing; I assume the surge testing with sequencing in London is going to pull capacity away from random sampling elsewhere...?

To no great surprise, it appears the strain came in via someone travelling from a nearby country not then on the red list.      

https://www.theguardian.com/world/2021/apr/13/surge-testing-may-not-be-enou...

https://metro.co.uk/2021/04/14/london-surge-testing-in-another-borough-for-...

Something to keep an eye on.  Over recent weeks, the variants data has been updated every Thursday so perhaps we'll see what that looks like tomorrow.

I had the impression from the reports that they had been tracking that outbreak for a while but only now published it because of the decision to start surge testing. Could be wrong though.

PS case rates in those areas are low and since the start of March there have only been approx 1500 cases across the three boroughs. So the SA variant must be quite a significant proportion of those sequenced.

That does make it unlikely it was going on under the radar - although I am assuming sequencing of pillar 2 samples is applied uniformly across the country.  As far as I can tell, scant little information is available publicly to understand how or where sequencing is applied, or how much of it comes form pillar 1 vs pillar 2 samples etc.  

It would be odd to not publish the presence of an outbreak that potentially partially defeats immunity; you'd think publicising it ASAP would encourage more people with symptoms to go for testing, and some people to be more careful.

In general I feel that a bit more openness around the sequencing would be useful.

Definitely agree

It seems the issue was one person who self isolated but their family members did not as they didn’t have to.

Also worth noting the care home outbreak stats. Something like one of about ten care workers who tested positive are reported to have had a vaccine. Presumably the rest did not.

Res ipsa loquitur.

Surge testing in Barnett now too, cases found there announced this afternoon

https://news.sky.com/story/covid-19-surge-testing-to-take-place-in-barnet-a...

One sentence stood out in the article

That outbreak is believed to have then spread to a care home in Lambeth where 23 cases were reported.

Given the current state of vaccination, that seems a strong sign its evading immunity in terms of causing infection....  Hopefully the consequences of that infection were ameliorated, but the article doesn't say - and indeed it may be too soon to say.

Following on from Misha's comment above, from this article

https://www.standard.co.uk/news/london/battle-control-south-african-covid19...

One of the 13 infected staff had a Pfizer dose two or more weeks before their positive test, a fact which is likely to raise fresh questions over how many care workers are getting vaccinated.

The mind boggles.

I've read somewhere that 1 in 10 samples is being sequenced, but which pillar those samples come from I don't know.

Updated variants plot using data from [1]

The nomenclature has changed from "genomically confirmed" and "genomically probable" to "total confirmed" cases - I was plotting the sum of the first two in previous posts.  I've switched to plotting "genomically confirmed" for before the change and "total confirmed" for after as there seems to be reasonable -not perfect - correspondence there.  I suppose the data is always subject to revision but the time based data series isn't published AFAIK (I could be wrong) although PHE sometimes tweet a picture of it.

There's ~60 new cases of the SA variant in this data, representing seven new days of data up to yesterday.   With similar numbers coming from the London based community transmission number given in articles up-thread, it starts to feel like the growth in SA variant cases might not be primarily driven by detection at the border - unless there is more reporting lag than the wording on the web page implies, and this new cluster of cases is yet to land in the plot.

[1] https://www.gov.uk/government/publications/covid-19-variants-genomically-co...

At one stage during the surge testing in February it was said by Hancock that half of all cases were being sequenced, although data I saw on the Cog-UK website for that period (which excluded surge testing cases) suggested more like 25% average. They do sequence all cases from surge testing.

The Cog-UK coverage report for week commencing 22/03 shows that almost all areas are sequencing above 20% with the UK average being 46% (only 4% in Northern Ireland.) That will only have gone up as cases have dropped further in the last three weeks, so it's way more than 10%, almost certainly over half now.

https://www.cogconsortium.uk/wp-content/uploads/2021/04/COG-UK-geo-coverage...

Edit - the cumulative average throughout the whole pandemic to date is running at 9.6%, so your 10% might have come from that?

Edit2 - the report still implies that capacity used for operation Eagle (surge testing) is excluded from that coverage report, which suggests to me that they were already probably over 50% three weeks ago. That report gives the breakdown by pillar for each nation too.

Oh joy... https://www.theguardian.com/world/2021/apr/15/covid-variant-first-detected-...