I know, harsh of me, Tsk

Indeed she is a bore

The woman’s name is blatant cultural appropriation. She should be cancelled forthwith.

dearieme November 1, 2020 at 12:16 pm – ““who drink whiskey “: the bastard, not even patriotic.”

Some whiskey is made in Northern Ireland. That could be patriotic.

@bom4

The following isn’t an endorsement of either the government strategy or the advice they’re receiving. But if you’re wondering what the point was of building the extra capacity if it the plan was to lock down before it gets used – they’re locking down now after they’ve been advised that if they don’t do it by the start of November, there won’t be enough capacity (including the Nightingales) to cope with the peak. Yes they’re unused now. But the lag between policy, changing the epidemic curve and finally hospitalisations is so slow that the models (from various advisers) reckon even the Nightingales could get burned through very quickly once the peak rises. It isn’t like the government has insanely built the capacity it has sworn never to use. It’s done plenty of daft things but that isn’t the logical fault here.

A feature of all the models used to project this stuff is that the size of the peak is very sensitive to the timing of starting an intervention. I suspect a more realistic model might be a bit rounder around the edges here since – like your dad – people have behavioural responses when they see things are getting hairy even without being told to do so, but also not all interventions seem any good at actually reducing contact rates. There was a pretty substantial study released on that this week, the bad news is if interventions don’t seem effective but advisers still think something needs to be done, their only recourse is to call for larger interventions. Unfortunately a sharp on-off switch type model is much easier. And quite how you round off the responses is tricky – behavioural feedbacks is something infectious disease modellers have been looking at over the last decade, and there’s no consensus on the right way to adjust for it.

Perfectly fair to say the government should challenge the advice it’s given, ensure assumptions it is based on are clear and out in the open (they’ll never be totally “realistic” but you at least want the model results to be robust to sensitivity analysis) and to try to avoid groupthink. With so much emphasis on lockdown for example, are there other more targeted approaches to control that are being neglected? Or even with lockdown, did people check whether a truly national one was required by modelling more targeted regional or local ones as a comparison? Those are the kinds of things it would be fair to expect a government to do, I reckon – we can’t expect the government to go away and do the analysis by itself. Well not the politicians anyway, from whom the best we can hope for is that they understand – to whatever extent is reasonable from an educated layperson – the results and limitations. As for the part of the government machinery that works on this stuff, their own modelling units produce forecasts no more optimistic than the academics – but then they’re using many of the same techniques so this isn’t surprising.

If the government were to disregard the advice that the Nightingales will be used up if they don’t lock down this week, and proceed with their regional approach or something else altogether, that would be an extraordinary move. In terms of the political risk calculus, if it turned out they had made the wrong call, there would be enormous consequences. Copy what other Western leaders are doing, end up with middle of the pack death rates, and you might get away with it. People will accept you faced hard choices to some extent. Politically a much safer option.

In fact the political risk reward ratio is being swayed even more than usual by the pandemic itself, I think. Just in terms of “what can we afford to get away with”, previous governments have settled on hospitals bursting at the seams over winter, flu patients in beds down corridors, enormous waits at A&E. Governments didn’t actually like this, and knew that voters didn’t like it one bit either, but it wasn’t something that would necessarily lose you an election whereas the cost of doing something to prevent it – a general enlargement of NHS capacity just to cope with a few weeks during a bad winter – was beyond what their tax-and-spending plans would allow for. However, an equivalent failure over Covid after having a year to prepare for it (even though this is hardly any time at all in terms of healthcare capacity building) would be deemed catastrophic and the public likely far less forgiving.

Governments could also cross their fingers and hope for a mild winter flu, and that the NHS would largely cope with the strain. With Covid there’s a risk, or at least their advisors will tell them there is, that a fairly small miscalculation in intervention timing could result in a wave of hospitalisations that isn’t just slightly over hospital capacity, but massively over it – something like the North Italy experience earlier this year. If governments start to rely on something other than finger-crossing at this point, it’s hard to see that as totally irrational even if you don’t agree with the policies they pursue to try to avert it.

MBE,

That’s all quite reasonable but at this stage that’s not the point. As I pointed out on Twatter and elsewhere:

On slide 3 they’ve pulled the old “scenarios” trick again:

Early SPI-M working analysis These curves represent scenarios from a number of academic modelling groups. THESE ARE SCENARIOS – NOT PREDICTIONS OR FORECASTS
I have a number of problems with this approach:

1. Crowd sourcing is a good idea only if all the sources are decorrelated and we know that the scientific world can be quite incestuous. Did they go out of their way to find a team that is completely sceptical? Why haven’t these models been published so that the input assumptions can be peer reviewed?

2. Just going for scenarios looks like policy based evidence making – find me a scenario that means I can follow the herd without having to justify the inputs

3. The last time we were given models with scenarios showing a huge number of deaths we know they were wrong, but not just wrong by a small amount but by 90%+. By now Sweden should have had 100k deaths and they have around 6,000 deaths

4. Why doesn’t that slide, or any other slide, show the impact of the 4 week lockdown? That’s the whole point of modelling, it allows you to do “what if”planning.

My problem isn’t that we’re having a lockdown, of sorts, its that we aren’t being treated like adults and shown all the working.

I’ll go further and say that we should have been given a target as to when these measures are relaxed based on what is used to implement them.

The problem we’ve got is that the advice from the epidemiologists the government is listening to have a single track mind and suffering group think and all they’re interested in is suppressing the virus and that needs to be challenged, and robustly. As BO4 said in response to my Tweet:

“With all this modelling, the methodology, source code and data should be out there, placed on Github. The people who noticed the housing market was a bubble and about to crash (and made a fortune shorting it) were not the government, nor the institutional banks. Hank Paulson was telling everyone it was all fine days before the banks started failing “

@BIND

Won’t disagree with that but I might temper expectations on some of it as to how much difference it would have made to do things in a better or more open way.

Like I said, I wasn’t trying to address the big points about the quality of either the advice or decision-making. Just the very specific thing that BoM4 raised: why build the Nightingales then ignore them rather than make use of the capacity? Answer – the plan actually was to use them, and the plan lost political support when NHS planners warned so much future demand was now baked in that they expected the Nightingales to hit their capacity before the peak. My more general point was just why the change of direction didn’t surprise me – if the government had taken said “we think models used have a range of uncertainty that makes them of limited value when projecting more than a fortnight into the future, and case numbers are stabilising or falling in many Tier 3 areas, so we will stick with regional tiers for now” then the political gamble this would have involved (and the risk of high-level resignations from SAGE or their other advisors) would have genuinely surprised me. Whereas if they were getting political cover from high-profile advisors who were prepared to say the models are just indicative, other evidence points the other way, cross the capacity bridge when we get closer to it but glad so much has already been invested in it, serious health and non-health effect of lockdowns, etc, they might have risked it. But since that’s very much a minority position in the UK public health community (academic and professional) that was never likely either.

(As an aside, the UK is one of the major players academically in infectious disease modelling, with several research groups of global significance – eg who often work with other countries’ public health authorities to do their modelling for them. I think it’s possible we have put an exaggerated importance on modelling the Covid response as a result, particularly because our civil service and health authorities try to be very “evidence-based” and modelling is one of the very few ways of “producing evidence” – albeit with severe limitations – about potential, largely untested, interventions. Moreover if we had fewer modellers on the advisory committees and most of the modelling input they received was emailed over every few weeks by a bunch of professors in Italy with a track record about as good as the UK ones have had, I can’t help but think it would be greeted with more skepticism.)

Clearly the lack of openness is concerning. I worry that part of the lack of transparency in the past was the government’s deliberate, even admitted, fear strategy. But weirdly the most recent set of revelations suggests the government was trying to hide bad future news that might have been useful scaring us into compliance, so perhaps it was just traditional government dislike of openness.

How much difference would throwing everything open have made? It would have been nice to be treated like adults. But I can’t see it would have altered the policy trajectory in any meaningful way. There are anti-lockdown voices now, including a significant but fairly small minority in public health academia, and they may have been able to bolster their arguments (or undermine those of their opponents) a little better if there was more openness. “Citizen science” types would have had a field day. But I don’t think it would have swayed opinions at the very top-end of the advisory groups.

Re the modelling in particular, you are of course right that asking half a dozen research groups to do the work isn’t the same as getting six independent opinions. But none of the modelling groups could ever be independent – they use similar techniques (actually there were some fundamental differences in the model types that were used, but their assumptions were largely similar and in fact such models usually come to qualitatively surprisingly similar conclusions), the academics often move between research groups, they all publish in and read the same journals, and so on. The issue is less about finding a “more skeptical” research group, or challenging their input assumptions, but the fact you’re relying so much on these models in the first place. Once you feed into any such model, from any university group, the kinds of trends that were showing up in the ONS Covid household survey data, they were all going to scream “you need to lock down ASAP”. It’s just baked in to all the kinds of models in general use that such a rise in the ominous prelude to a nasty peak, in the absence of other measures.

If you were a model-skeptic, you might reasonably say “well I don’t need a model to tell me that – once infections go up like that, hospitalisations will follow fairly soon, so clearly we would implement some measures to limit the effects of that!” But the people doing healthcare capacity planning would generally like some numbers to look at, even if their uncertainty intervals are impractically wide. If you have a model that implements lockdown it will usually tell you this fixes the problem – again, just mechanistically, if you tell the model contact rates go down, it will tell you the peak goes down. It’s how the models are built. But a more complex package of alternative measures may be very challenging to model.

This really harks back to my parenthetical paragraph – if modelling forms a large component of your advice base, you’re inevitably going to receive a lot of advice telling you about the joys of lockdown. This is a more pernicious problem than sticking code on GitHub can solve. It’s why I thought a lot of criticism of Ferguson’s original code was well-intentioned but misplaced. Don’t like his code because it’s sloppily programmed but being used for a safety-critical decision that tens of thousands of lives rest on? Totally fair, and you’d never let nuclear power plant code get away with it. Reckon that if you cleaned it up you’d reach a different conclusion to his team? Implausible – it predicts, qualitatively, exactly what you’d expect such a model to predict, and quantitatively it produced similar results to other types of model produced by different modelling groups. The harsher, and in my opinion more telling, critique was based on track record, especially early in an epidemic when data is more limited.

MBE
So the modellers have been wrong every time so far but this time they are right.

Hmmm. Ferguson’s record on Swine flu, BSE and Foot and Mouth should have deep sixed his career and have it buried with a stake in its heart in an unmarked grave at a crossroads. The only use for him is as an understudy as Banquo’s ghost in a Christmas pantomime.
And Fauci insisted for years that you could catch AIDS around the family breakfast table.

But yeah, this time we should believe them.

Just for a different critique of modelling and placing too much faith in it, or going outside what it can be expected to do, have a glance through All models are wrong, some models are useful.

“The political dynamics of what’s possible in Sweden are very different on account of what “following the science” means there.”

So there’s different ‘science’ in Sweden is there? The virus behaves differently 700 miles further east of the UK?

The UK doesn’t have ‘scientific advisors’. It has a bunch of wannabe politicians masquerading as scientists. Getting on SAGE requires you to be a fully paid up member of the UK Establishment elite, having crawled your way to the top of the greasy pole in academia, with all the political baggage that comes with that, and the incumbents curating who gets to join them. Its a self perpetuating circle jerk. Its advice on covid is no more dispassionate science than its advice on so called ‘climate change’.

One of the reasons Sweden, and for that read mainly Stockholm, had it so bad is their spring half term was a week before their neighbours and Swedes went skiing in Northern Italy about the time Brits did and the virus was widely seeded before they had time to react. (I can’t find that reference now)

They’ve also admitted the big mistake that we made of letting it get in to care homes.

My “2-penn’orth”… The “Nightingales” were built as industrial-scale ventilator units, specifically to cope with patients suffering from a straightforward but potentially serious respiratory illness. It was rapidly discovered that the vast majority of “covid patients” had one or more serious pre-existing conditions, which the Nightingales were simply not designed to cope with, hence people were treated in “normal” hospital facilities.

Not long after that, the yanks discovered that ventilating patients was killing a lot of them and methods such as CPAP were used instead, with much better outcomes, so the “ventilation factories” became even more superfluous.

I see this morning that Boris is predicting a higher death rate than that in the spring. Personally, I think that it’s bollocks on stilts and we’re not going to have a “second wave” unless there’s a conspiracy to manufacture one. The present slight increase in deaths creating the hysteria in the MSM is from people dying from “other” things (heart disease, stroke, cancer etc.) who just happen to test positive for covid but were not hospitalised for a respiratory condition, and indeed not carried off by same.

If we haven’t had the catastrophic death rates that are being conjectured by Christmas, SAGE should be dismissed with ignominy, ditto Johnson and Hancock.

@BIND

Yeah, I have reckoned for some time that Sweden’s high death rate has rather less to do with their divergence from their neighbours in overall strategic approach than people think, and rather more to do with some very costly errors in its implementation (also a bit of bad luck, but e.g. the seeding counts as a bit of both, since measures could have been taken to limit it). Regardless of root cause, the difference in death rates is still politically tricky, given the general public have limited tolerance for nuance – I don’t think holding fast to the policy would have been sustainable if their advisors were loudly protesting against it.

@BJ

One thing I notice when talking to laypeople is how many say they don’t understand why Covid-positive patients aren’t all being treated at Nightingales in order to free up the main hospitals and keep them safe, but that shows a failure to grasp the vast range of facilities and equipment a modern hospital has and the limitations of a field hospital, particularly when people often have much more going on than just Covid.

Having said that the Nightingales have always been part of the plan for this winter – not so much that they’re superfluous but that they’d rather fill the normal hospitals first. Presumably survival rates are expected to be better if the full range of facilities are available.

Re “second wave”, the definition of “wave” has always been very fuzzy, but the ONS household survey data isn’t a conspiracy and indisputably shows a rise in infection levels across age groups. That trend isn’t being biased by changes in testing strategy, false positive rates or whatever. So regardless of what you want to call it, a second peak in infections in infections is real, and (though there may be quibbles about how to apportion them) a second peak in deaths seems inevitable. What you do about it is a different issue, but I don’t think there’s much mileage in the “we’re already at herd immunuty” or “it’s just a casedemic” arguments right now.

As for SAGE’s heads on plates – the problem with the kind of figures that Johnson is waving around is that they represent a hypothetical scenario where no action is taken. Which means he doesn’t have to quit if they don’t happen – in fact it looks even better for him if they don’t, since they count as “deaths averted” by his strategy. Unfortunately it’s rather hard to prove beyond reasonable doubt, to the jury of either scientific, political or public opinion, that such counterfactual figures are wrong (or indeed right). The suggestion of comparing them to what happens in reality doesn’t do the job if it represents a scenario that was never expected to play out in the first place…

@MBE..

the ONS household survey data isn’t a conspiracy and indisputably shows a rise in infection levels across age groups. That trend isn’t being biased by changes in testing strategy, false positive rates or whatever. So regardless of what you want to call it, a second peak in infections in infections is real,

I’m afraid that I must both agree an disagree with you. There certainly is a reported increase in “infections” (for which read “positive test results”) but this seems to be purely a function of the increase in the number of tests done. The %age of those returning positives has barely changed.

If you test ten times as many people at random, the false positive rate alone is sufficient to give ten times more “infections” – but they aren’t going to follow through into ten time as many deaths – there’s almost no correlation, the system is so haphazard and getting worse as larger and larger numbers of inexpert testers are recruited to help get the numbers up.

There certainly is a reported increase in “infections” (for which read “positive test results”) but this seems to be purely a function of the increase in the number of tests done.

The sudden switch from reporting numbers of deaths attributed to covid-19, to numbers of “cases” of SARS-CoV-2 infections makes the whole business look more and more like propaganda to me.

cf the sudden jump from “global warming” to “climate change” when it became clear to the laymen that the warming had stalled. It’s clear from the published mortality data that covid-19 was over by June. There have been been a handful of deaths attributed to covid-19 each week since, but far outweighed by flu and pneumonia deaths even at the height of summer.