Forum menu
Not seen today but yesterday hit 500 admission were lower at around 1.3k rather than 1.5k
So today almost 600 and admissions 1.5k again. Cases and admissions seem to have plateaued though, deaths lagging as expected?
I know that AZ and others are going to be selling their vaccine at cost (at least during the pandemic).
Does seem normal for pfizer to profit from most things.
Could this mean Pfizer have pushed things through too quickly? If they're the only company to not agree to offer it at cost then they have a lot to gain by being first. If they can get ahead of the others and have 4-6 months of being the only player in town and thus charge what they see fit, then drop the price as other options become available what's stopping them cutting a corner or two? I know normally there would be loads of checks and balances to stop this but with everything being pushed through in record time, understandably, right now what's the chance they've missed a step and it hasn't been picked up?
Not wanting to poo-poo their work, especially if it is genuinely a solution we desperately need, but it's the first thing that pops into my head when I read that.
So today almost 600 and admissions 1.5k again. Cases and admissions seem to have plateaued though, deaths lagging as expected?
I can't help but notice that test numbers have levelled out or even slightly dropped though, might help supress the number of confirmed cases slightly. Admission numbers are slightly more encouraging though.
I don't think they will have missed anything, the consequences will be far too high for a mistake like that.
Basically if the vaccine comes out as being dangerous due to missed testing then we will have the current situation for a long time as no one would take any vaccine.
Sadly I just think it is profiteering as normal from them.
Plus I don't think they will be first to market. They have just given an interim readout which being cynical has probably helped them get some more orders. But as I said this is a first world vaccine with it costing £30 a go.
So today almost 600 and admissions 1.5k again. Cases and admissions seem to have plateaued though, deaths lagging as expected?
Deaths by date of death are what really matter - these won't be exceeding 500#.
https://coronavirus.data.gov.uk/details/deaths
A lot of pressure to get a daily bed and death count out, but the 7-day rolling mean is a better measure. Date of death is pretty smooth. I've not seen anything to alarm or reject my previous projections of more than a month ago. Now we have Tiers and lockdowns, things will be heading downwards. North West admissions are already going south.
You wouldn’t be shedding virus… no idea about anti-body testing. TiRed is your man.
If you have been vaccinated, you'll test positive in a lateral flow test for antibodies. If you have a monoclonal antibody as therapy, you may not depending on the lateral flow assay used. The PCR test for virus will not care about whether you've been vaccinated.
#just to bait @thecaptain - actually both projections aren't bad at all given the shifting sands of policy 🙂
I don’t think they will have missed anything, the consequences will be far too high for a mistake like that.
Basically if the vaccine comes out as being dangerous due to missed testing then we will have the current situation for a long time as no one would take any vaccine.
I'm thinking more like the Thalidomide issues, where unknown side-effects may not show up for a few years or more. There's been time for short-term issues to be flagged up but not enough for long-term ones. Absolutely agree that all the pharma companies will be very aware that they have a duty to not cock it up as public opinion on the safety of any drugs has to be very high for it to work.
I’ve not seen anything to alarm or reject my precious projections of more than a month ago.
Agreed. We should be looking at teachers, front of house staff in hospitality, public transport passenger facing staff, and shop floor workers. Anyone who’s job it is to deal with loads of people a day. NHS and care staff first though… of course… they not only deal with loads of people a day, but a lot of those people are the most at risk.
Depending on uptake then a regional approach might also work medically if not politically.
Wipe the virus out in specific areas and make a full lockdown in/out (with the exception of medical who have also been vaccinated)
For me the vaccines are a very different beast to the small molecule drugs.
Vaccines induce the body to do something where drugs have a direct action.
the main concern for a vaccine or anything that affects the immune system is induction of an auto immune response and all that follows from there.
With the numbers being tested any significant incidence would be detected, but yes there will be a lot of hopefully minor adverse events reported.
@stevextc - that's sensible and worked in Australia, helped by the major population centres being widely separated (sorry Albury). Unfortunately we're one of the most densely populated countries in Europe with massive interconnectedness so I can't see that working as well.
@TiRed, please re-plot your figure of p455 together with the death data that are now available. At a glance it looks like you've massively underestimated reality. I think I hinted at this a few days later on p458...
I wasn't actually going to draw attention to it. But since you insist....
I think 500 deaths/day was the separation, no? (and I mean deaths by day of death rather than the headline figure today). My earlier plots were based only on England data, but since October I've modelled regions and summed to UK. Predicting the nadir and rise looked reasonable, but the precision was poor out beyond four weeks. From November, projection fails due to policy, but I was satisfied with 300 deaths/day by mid-November, rising to approximately 1000 by mid-December in the absence of no policy change. The effect of Tiers will, however, now be pulling down the daily death numbers.
I'm trying to build a model based on ONS survey data and forward projection. This would provide data-driven two-week forward projection.
Not sure what that means @TiRed it looks like a bit of a word salad to me but here is my best attempt to overlay data on your pic:
ok I no longer know how to embed from imgur but the link is here for anyone who wants to check it out.
Feel free to improve my graphic but this prediction was truly awful, no? The tail end where it just pops back in to your dead zone is of course incomplete data.
Or have I misunderstood what that plot was about?
Ah, maybe this will work:
4 weeks ago I said we were on course for about 500 deaths per day by mid-Nov. Admittedly I didn't provide any confidence interval, but that prediction is still looking pretty good to me based on those data up to say 1 Nov. Not clear how much of the subsequent flattening is real vs just incomplete data. There are now clear signs of a slowdown in the epidemic which will feed through into death numbers in due course.
I’m not sure how appropriate Willy waving over a plot of dead people really is. Let’s be careful of the vocabulary and acknowledge the hard work being done avoiding potentially contentious terms.
Murray
@stevextc – that’s sensible and worked in Australia, helped by the major population centres being widely separated (sorry Albury). Unfortunately we’re one of the most densely populated countries in Europe with massive interconnectedness so I can’t see that working as well.
That was without a vaccine though.... I think it would be political...
Ignoring politics then we have the vaccines and resources to start in the M9 corridor and realistically get Scotland Covid free.
Depending how the Irish manage/cope etc. then NI can be done... and Wales .. but just using these "names" shows how politics would interfere.
It might be amusing to see all the "I'm not having a vaccine" people's attitude if they were told "you're not on the list anyway" though. I suspect a lot of switches would occur to "us first"..
@thecaptain - it looks like you're 2 observed data sets are not the same even in late September. If that is correct then you're not comparing apples with apples, more like apples vs quinces.
I think @TiRed's plot was following the trajectory at that time without intervention/mitigation. As we know it's a dynamic system so any prediction 1 month ago won't hold water now due to policy changes Tier 2/3/twelvty etc.
Is it really only around here that schools are struggling thanks to so many teachers testing positive in the last week or so? Utterly confused to see so little about it anywhere.
Nope - our local comp has sent Yr 9 home as they don't have enough teachers to run all years.
They are right on the edge of sending Yr 8 home too.
I’m not sure how appropriate Willy waving over a plot of dead people really is.
No one is willy waving, its called discussion if discussion of important information is surpressed because is hard to hear we are on a slippery slope.
A local secondary school here is on the verge of shutting. Not because they have sick teachers but due to a lack of cleaners to give it a thorough clean every day.
No one is willy waving, its called discussion if discussion of important information is surpressed because is hard to hear we are on a slippery slope.
Well personally I think that accusing people of "word salad" and using expressions like "suppressed" is a little inflammatory, Maybe thats just me and its not my argument have to so I'll back out for the sake of the thread.
Looking back at 455, TiRed's plot is for England and the extrapolation is I assume based on that.
I don't know what thecaptains data is, but an educated guess by comparing to the Gov data here
https://coronavirus.data.gov.uk/details/deaths
on the around 1st Nov point - thecaptain is a bit over 300 and Gov reported 7-day rolling average as ca 320-325 around that period for the UK as a whole
Are you comparing UK to England only?
I still think TiRed's extrapolation has turned out low but maybe not by as much as thecaptain's plot suggests.
No one is willy waving, its called discussion if discussion of important information is surpressed because is hard to hear we are on a slippery slope.
I'm not sure anyone has a problem with discussion, but sometimes the tone and language needs to be considered. I appreciate that we have two people with - to my untrained eye - really useful skills in handling and trying to explain data and trends, but sometimes the tone can come across as two alpha males seeking dominance.
Edit - which distracts from the message
Missed that key last sentence
Tired’s graph specifically referred to the UK, and the slight difference in data sets is because the past few days are always a little incomplete and retrospectively updated, which of course everyone who uses these data sets fully understands. My data are directly off the gov website, deaths by date of death for the UK.
But of course I’m happy to hear his interpretation of the large and rapid divergence.
I was hearing last night of a London primary school where the new demands on the teachers mean that they have lost their non-contact periods, lost 20 minutes of lunch and have more obligatory meetings. The working day there is now 7 till 6.
@thecaptain let me get this right. You're arguing with @TiRed about whose plot/graph is most accurate?
@thecaptain - You're being a bit passive aggressive about this.
The differences start in Sept not the last few days. For those of us who you deem not in the know ( which of course everyone who uses these data sets fully understands.), why don't you explain what each element of your graph represents?
If you worked for me I'd send your graph back to be updated. It's missing series labels. Graphs are supposed to be helping to communicate something so they are necessary whether you like it or not.
I'll go back to what I said many, many months ago ... All models are wrong but some are incredibly useful. You need to clearly articulate why yours is credible rather than just sticking to the "my graph is better than your graph" approach
I’ve just screenshotted tired’s graph, you can go back and look at his description of it but it clearly says it’s UK death data. I’ve just told you what data I plotted on top. It was tired who brought up the subject of how well his prediction had held up. I disagree with his assessment. Hopefully he’ll be along to explain if I’ve misunderstood something.
I acknowledge that debating death predictions can seem distasteful and apologise to anyone offended. But without credible prediction, there can be no understanding or control. And mistaking bad predictions for good is worst of all because it implies a belief in understanding which is mistaken.
It’s a sorry fact that a lot of the prediction from those who claim expertise has been extremely poor from day one of this epidemic. It is clear to me that understanding the details of epidemiology is not enough and that it requires the involvement of scientists who understand forecasting methods, their design and analysis. I am one such scientist. There are many others but most of them already have full time jobs!
I acknowledge that debating death predictions can seem distasteful and apologise to anyone offended.
This needs to be done far wider and frequently for public money/projects.
I've previously speculated they should publish a "cost per life".
An example is say the Channel Tunnel... contractors should submit a financial bid and a deaths projection and then be penalised for every death above the "accepted" number.
I suspect if this was done the tolerance for deaths on similar projects would drop significantly...
We can do the same on completed projects... such as Grenville
was hearing last night of a London primary school where the new demands on the teachers mean that they have lost their non-contact periods, lost 20 minutes of lunch and have more obligatory meetings
In my secondary we have split arrival and depature times split breaks, split lunches, all these need staffing.
I’ve just screenshotted tired’s graph, you can go back and look at his description of it but it clearly says it’s UK death data.
His graph on p455?
It does in the source but the title says England. Which could be lifted from UK admissions and deaths data, but only TiRed knows.
shes saying new liverpool mass test has shown not much covid in Liverpool?????
You're right that is inconsistent I looked at the bottom graph and missed the top title. Later I believe he also said he summed over the 4 nations. The UK deaths time series seems to fit closely to his graphic (apart from the incomplete days) but of course that is quite close to England-only numbers. If his analysis was actually just England, there's still a large mismatch but it would be a bit smaller than on my pic.
But of course I’m happy to hear his interpretation of the large and rapid divergence.
That model was just England data, hence the precision of deaths was low due to the limited data (which isn't a bad description of subsequent outcome). Recent analyses are refined to calculate deaths from admissions (a week earlier). That ties the rates of change together in a more mechanistic manner (like your SEIR model). I also describe NHS regions where the prediction of deaths from admissions is more closely aligned. Then sum all regions to get England and sum countries to get UK. Here one sees a more accurate description (Wales is an outlier due to counting suspected admissions). I still feel very uncomfortable of forward prediction beyond four weeks. That's long enough to see evidence of policy changes. So I consider that fit for purpose.
Expect things to turnover at about half of the April peak in a week or two's time. As for post lockdown - expect stable control in Tier 3, maybe Tier 2 and growth again in Tier 1.
BTW, projection from 2020-10-19 was bang on accounting for negative curvature in admissions that became apparent a few days later 🙂 . But in reality, forward projection of an exponential process is always challenging. Calling a zenith and nadir especially so. Total epidemic size - waste of time for an emergent pathogen.
OK, England is definitely a better fit (also to the hindcast)...but also well outside your forecast range, even before you issued your forecast..
On the bright side, what you are doing is a whole lot better than the utter drivel that the REACT group just press-released. Look, I can draw a wiggly line through noisy data! WTF they think that is useful for, I have no idea.
Is there any indication of the efficacy of the Oxford vaccine?
I’m assuming (always dangerous) that the reason we have heard the Pfizer announcement first is due to different regulatory environments, rather than actual timelines.
I have no idea.
You aren't alone there! Piecewise or local exponentiated polynomial models are far better for inference. I told them that the system does not change that fast and any spline mode will hugely influence the regression by number of splines.
I'm pretty satisfied with my methodology - it's local, time bound and should be interpreted with caution. But it is now a counterfactual for evidence of influence of Tiers.
Hahah do those vertical lines represent the confidence intervals? HAH! Holy shit.
But it is now a counterfactual for evidence of influence of Tiers.
Hang on? So Tiers don't work?
Sefton, that's gammonFM, I'd give it a wide berth.
Hahah do those vertical lines represent the confidence intervals? HAH! Holy shit.
Very low numbers of samples at the ends of each sampling - exact Clopper-Pearson limits for 0/10 samples are wide.
Hang on? So Tiers don’t work?
Tier 1 does not control spread
Tier 2 may contain spread
Tier 3 may shrink the epidemic (R<1) - perhaps.
Lockdown shrinks the epidemic (R~0.8)
Is there any indication of the efficacy of the Oxford vaccine?
Oxford vaccine produces an antibody response that looks similar to Pfizer. It did not produce sterile immunity in animal challenge models. It will likely work the same. Duration and sterility of immunity are not known at this point.
I’m thinking more like the Thalidomide issues, where unknown side-effects may not show up for a few years or more. There’s been time for short-term issues to be flagged up but not enough for long-term ones. Absolutely agree that all the pharma companies will be very aware that they have a duty to not cock it up as public opinion on the safety of any drugs has to be very high for it to work.
You don't go dosing pregnant women with random drugs to find out whether they're teratogenic or not, you do mouse studies. Which will have been done on this. Thalidomide was a cock up of epic proportions because they didn't carry out those types of safety studies in animals.
Anyway, everything kills you - oxygen helps to give you cancer, tasty food kills you, every drug has a side effect. It's all about the risk benefit ratio, I'm way more worried about breathing in city air than I am about getting a vaccine. **** I'd probably take the Russian one if I got to breath clean air for the rest of my life.
Tier 1 does not control spread
Tier 2 may contain spread
Tier 3 may shrink the epidemic (R<1) – perhaps.
Lockdown shrinks the epidemic (R~0.8)
Yup, fits pretty much exactly with my semi-professional guess at how the tiers would work in practice.
Thalidomide was a cock up of epic proportions because they didn’t carry out those types of safety studies in animals.
Actually it wasn't - that's post-hoc reasoning. Up to and including thalidomide, drugs were not tested for teratogenicity. After, they all are. Thalidomide is a very effective medicine, but must never be administered to a woman of child bearing potential without appropriate controls.
It was reports of neuropathy toxicity that kept it from approval in the US not teratogenicity.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4737249/