Modelling may be in good faith but it’s reasonable to ask why it’s been so wrong especially if policy decisions are going to be influenced by it going forward.
SARS CoV-2 coronavirus / Covid-19 (No tin foil hat silliness please)
- Thread starter Don't Kill Bill
- Start date
What has been so wrong? If it is variant related then modelling would be based on existing variants so will vary when a new variant arises. I'd guess the other factor is that human behaviour becomes more and more important as infectiousness increases. And human behaviour is hard to model on such short timeframes.
I'm not sure what you mean. Wrong in what way? Actions are taken to reduce the impact, so that will affect the outcome.
Are you suggesting that because things didn't turn out as 'predicted' that using models is inherently flawed?
Seasonal? It loves summer here in Australia, I suspect he’s making out it’ll just be a winter thing.
But we’ve been through many months where everyone’s been criticising the lack of restrictions in England and the reality in terms of hospital numbers and deaths have been far below even the most optimistic range of predictions based on the modelling. We can’t turn around now and say that the discrepancy was a result of restrictions that prevented it being even worse - because there hasn’t been any.
Given lack of measures in England for the last few months surely we should be looking to compare the upper end of the modelling predictions that presumably were based around an assumption of minimal public health measures, which is exactly what we have. Surely the lower end of the modelling was based on stringent measures that never happened.
Arguing that modelling designed to take into account changes in public behaviour can’t be expected to accurately take into account changes in public behaviour is also a bit shit
Questioning why it has been so inaccurate seems fair especially given this thread is hardly packed with praise of UK govt for keeping most metrics below the most optimistic of modelling predictions over the last few months.
Last edited:
Are people really still defending modelling at this point? They are attempting to predict the utterly unpredictable, the majority therefore are always going to be way off and have been just that.
35 million booster jabs, starting with the most vulnerable that would have been hospitalised, has to some extent saved the day. I say to some extent as many have still died or now have life-changing long-term effects, I would word that better if I could.
In defence of modelling (I know, I should never have mentioned it in the first place 
).
This is the summary of the modelling that encouraged the UK government to stay open, when the rest of Europe was closing. The models were written when we knew a fair bit about how Omicron affected infection rates, but very little about what happens with death rates amongst older adults, and in particular amongst those who've had boosters.
The models were the basis on which central government went for a Plan B+ that is Plan B plus accelerated booster program, changes in testing rules/advice and a couple of solemn looking press conferences.
Infections and hospitalisations are on track for the lower bounds of Scenario A. We think deaths will come in lower, but we're not sure how much lower - maybe 300/day at peak. The thing the models don't predict are the timing differences between regions and whether regional hospital/care services can cope with their peak.
The models actually were the science advice that suggested stopping football, hospitality, non-essential retail etc weren't actually going to do much other than move some hospitalisations into March, and could potentially cause more deaths overall as boosters started to wane.
I know people view the UK as Plague Island with no rules, but the modelling is part of the reason why they chose a particular gamble on what was worth doing.
On a day to day basis they look at a series of "nowcasts" with daily data on things like mobility patterns, testing, hospital admissions, occupancy where it's available, and predicted data where the daily data is incomplete. A lot of that is public domain, with daily updates on the dashboard, some of the most up to date info appears as a series of .xls files from the NHS.
The short-term (two week) models using that data know the inputs in terms of current government action, people behaviour etc so they have a fair idea where hospitalisations are going - these are the ones that mean government ministers can keep saying "no new restrictions right now" - because they've already decided that if daily admissions stay below 3k we're doing well and we can slam on the brakes if we have to. Different models for different jobs.
).This is the summary of the modelling that encouraged the UK government to stay open, when the rest of Europe was closing. The models were written when we knew a fair bit about how Omicron affected infection rates, but very little about what happens with death rates amongst older adults, and in particular amongst those who've had boosters.
Tweet
— Twitter API (@user) date
The models were the basis on which central government went for a Plan B+ that is Plan B plus accelerated booster program, changes in testing rules/advice and a couple of solemn looking press conferences.
Infections and hospitalisations are on track for the lower bounds of Scenario A. We think deaths will come in lower, but we're not sure how much lower - maybe 300/day at peak. The thing the models don't predict are the timing differences between regions and whether regional hospital/care services can cope with their peak.
The models actually were the science advice that suggested stopping football, hospitality, non-essential retail etc weren't actually going to do much other than move some hospitalisations into March, and could potentially cause more deaths overall as boosters started to wane.
I know people view the UK as Plague Island with no rules, but the modelling is part of the reason why they chose a particular gamble on what was worth doing.
On a day to day basis they look at a series of "nowcasts" with daily data on things like mobility patterns, testing, hospital admissions, occupancy where it's available, and predicted data where the daily data is incomplete. A lot of that is public domain, with daily updates on the dashboard, some of the most up to date info appears as a series of .xls files from the NHS.
The short-term (two week) models using that data know the inputs in terms of current government action, people behaviour etc so they have a fair idea where hospitalisations are going - these are the ones that mean government ministers can keep saying "no new restrictions right now" - because they've already decided that if daily admissions stay below 3k we're doing well and we can slam on the brakes if we have to. Different models for different jobs.
You can't take a model seriously that has such wide ranges, where the most extreme options actually significantly overlap each other. It's virtually useless. It was just a way for the government to do what they wanted whilst having a study to point at and say 'look, we're following the science'.
Absolutely, but modelling something so unpredictable isn’t really scientific in the true sense of science.
Imperial have now released a study showing that high levels of T-cells from the common cold can protect against Covid infection, which is probably another reason the original models in particular were so ridiculously overexaggerated. In the years to come we’ll have more and more evidence I’m sure to show the first lockdown experiment in human history was based off completely bonkers predictions.
The imperial study backs up this non-peered reviewed paper also.
In a sense that was the point of the model. They couldn't stop it, they could only squash it. They then had to choose how long they would be willing to squash it for. It effectively knocked out the idea that a "short, hard lockdown of 2-4 weeks" starting just before or just after Christmas, would help.
The answer came back that if you don't squash it for months and wait for an Omicron tuned booster or a new antiviral before opening then it didn't help. In fact some measures, short of a full lockdown, made the overall first quarter death toll worse, by adding more infections in March.
Plus, of course if Omicron got replaced by something more virulent, you would see no longer term (12 month) gain either.
If you are a modeller you build in worst case assumptions, even worse than you think, then even if you are wrong, it’s better than being optimistic and wrong the other way.
Also this wasn’t exactly easy to predict, just modellers looking after their Livelihoods or someone else’s.
Also this wasn’t exactly easy to predict, just modellers looking after their Livelihoods or someone else’s.
This is exactly what is happening in Poland. We have several doctors who are 'outside the system' and 'don't fear speaking the truth'. Anti-vaxx movement considers them all heroes who haven't allowed big pharma to bribe them.
Naturally, they all charge 500% of average consultation price, and are all booked for 3-4 months forward by their devoted fans.
Pandemic has been a great opportunity for mediocre doctors to make a name for themselves here and become valued for 'speaking the truth' that people want to hear.
Russia may now be looking at excess deaths approaching 1m now.
https://www.themoscowtimes.com/2021/12/30/russias-excess-death-toll-hits-930k-a75964
Pro-rata against population that's not so far off IC's 500k deaths in the UK if no action taken estimate. The sad thing is that a lot of Russia's death happened in 2021 and occurred despite vaccines being available. The UK managed to delay things long enough to not to see that number of deaths, because the vaccines took over from the other controls before Delta hit - we've still seen 150k deaths though.
Countries that have managed to delay things further, into Omicron territory, particularly if they've got high protection using vaccines in the over 60s - may come in with much lower overall death tolls. They'll still have to take a leap into the unknown though and some still aren't ready for that.
Ok sure but the modelling presumably factored the boosters in to the equation based on low or high take up
The November/December omicron models looked at boosters based on the planned December rollout. One suggestion was that faster/wider booster rollout was a way of making Plan B controls more effective. Another element was to suggest that a stronger push to reach missed high risk groups (basically the ones not mobile enough to get to the big vaccine centres) was needed.
The government responded by asking the NHS and volunteers to increase the number of booster sites. That led to the weekly total of boosters in the UK jumping from around 2.5m/week (late November, early December) to more than 5m/week in the two weeks before Christmas. They also increased the payment to GPs etc who could vaccinate higher risk people like the housebound and other vulnerable patients who'd been missed. Time from dose 2 to booster was reduced and the eligible age group came down.
The models were supposed to give guidance. That information led to policy changes. That's a model working, not a model screwing up.
The booster rollout jump between November and December in England:
I hope I’m not spreading fake news but if this account is legit it gives a good (and incredibly bleak) insight into how China manages to keep covid numbers surprisingly low.
Tweet
— Twitter API (@user) date
Looks like Jaywick.
Hong Kong is doing the same.
A good friend of mine is a pilot there and he tested positive a few days into the 21 day quarantine on arrival in similar looking shipping containers (Penny Bay, Google it). Sent immediately to hospital for a 2 week isolation and observation, and will have to do another 2 weeks in the quarantine centre after that.
I've seen the photos. It's horrible.
Holy shit. 4 weeks isolation! That’s nuts.
Isn't that from the Fyre Festival documentary?
Well, it finally got me. I'm covid positive. Got tested positive on arrival when I got back home from holiday. Symptoms showed up the next day. Had a super sore throat and was feeling a bit feverish second day of symptoms, but am feeling better 3rd day onwards.
I'm just thankful symptoms aren't worse and I only caught it on arrival back home. I'm boostered with PPM. I don't blame the vaccines. I got a feeling that boozing morning, day and night everyday for 2 weeks and then finishing a bottle of Glenfiddich 15 years on the final night finally took its toll on my immune system.
I'm just thankful symptoms aren't worse and I only caught it on arrival back home. I'm boostered with PPM. I don't blame the vaccines. I got a feeling that boozing morning, day and night everyday for 2 weeks and then finishing a bottle of Glenfiddich 15 years on the final night finally took its toll on my immune system.
Is that Rhyl ?
Seriously though, would a week in one of them with a radio and some books be that much of a hardship? I could live with it.
Tweet
— Twitter API (@user) date
Sure you build in worst case and best case scenarios. You don’t set your parameters between shit and fecking shit. These things influence public policy.
And I’m not sure anyone’s saying it’s easy to predict but surely given how wrong they were we need to learn lessons as to why. Else they just become meaningless as politicians, much less the public, will take much note of what they say.
If we’re so relaxed about their inaccuracy then just throw a dart at the board. It’s cheaper. My issue isn’t with the fact they’re wrong as much as people who don’t see an issue with them being so wrong.
You've completely lost me here. You think building ranged predictive models is equivalent to randomly selecting numbers?
My 3 year old daughters tested positive. Incredibly mild symptoms, literally just feeling a bit more tired and grumpy so did a LFT. Outbreak in her school. Sort of weird how kids have all just been sitting ducks, in effect, with regards to getting the virus. Also have sort of felt like I wanted to ride out pandemic without her getting it. But not feasible I guess for anybody unvaccinated.
So far negative myself. Will really be testing out booster efficacy as I was boosted in October and never had covid.
So far negative myself. Will really be testing out booster efficacy as I was boosted in October and never had covid.
For sure. I think that UK anaesthetist would have passed FRCA, probably the toughest postgrad exam for doctors in the UK.
So definitely “intelligent” in one sense and not sure about mediocre but definitely not the sharpest tool in the shed if declined the jabs seeing what he must have.
but definitely I would say irresponsible to perpetuate his views but I agree. Contrarian docs like him are in minority but get an inordinate amount of air time to say their piece while simultaneously having a siege mentality in their supporters with false accusations of censorship
Nice piece of data from France where they've split the hospital and ICU admissions data between Omicron cases and Delta cases.
54% of their new hospitalisations are Omicron infected patients, but only 19% of their ICU admissions. It comes with the usual provisos - it's still early days for Omicron hospital cases, and ICU proportions may change. However it's certainly in line with the UK experience where the total number of patients in ICU is currently falling, despite the massive rise in Omicron cases.
54% of their new hospitalisations are Omicron infected patients, but only 19% of their ICU admissions. It comes with the usual provisos - it's still early days for Omicron hospital cases, and ICU proportions may change. However it's certainly in line with the UK experience where the total number of patients in ICU is currently falling, despite the massive rise in Omicron cases.
Tweet
— Twitter API (@user) date
Tweet
— Twitter API (@user) date
Cool graph.
Was always going to happen. Co's start reducing sick pay if can't work and haven't vaccinated
https://www.bbc.co.uk/news/business-59930206
https://www.bbc.co.uk/news/business-59930206
Omicron the last minute sub...
If only Donny had the same impact.
Queue outrage on Twitter from people “never stepping foot in IKEA again”.
Just for this I’m now going to start going to IKEA. They’ve won a customer today.
Its because the disease course prediction models of epidemiology is a fake science.There are too many unknowns which influence the course of population spread, so you throw in garbage features and get garbage output. It’s mostly for people who read ‘respectable’ media sources who like to say they’re following the science. Invariably when you have 100s of modellers, one will be right and claim victory, much like the 1000 monkeys with typewriters can write Shakespeare quip.
It really is a waste of time.
On a similar note, the first supermarket that bans shoppers who are not wearing masks (unless medical exemption) will get my custom. Sick of people wandering around, picking stuff up/putting it down and no masks.. I know there are some reasons but locally to me, it's half the fecking store.
How else do you expect people to shop?