[ali k]

The BBC, sky news, ITN, channel 4, the telegraph has only just in the last day or two really pivoted away from parroting the last two weeks of doom stats from government.

Pete - just listing a lot of media outlets and saying they’ve reported Covid numbers doesn’t amount to much of an argument to back up your accusation of ‘scaremongering’.

..is nowhere remotely resembling what I actually said. What I did say was that the media have focussed on the doom predictions not the covid numbers.

I haven’t followed the rest of the thread but don’t get pissy about miscommunication if you type ‘stats from government’ when you actually meant ‘predictions [from scientists]’. They’re completely different things and you know that.

[AJM]

It’s not always “scaremongering “ but it’s often over sensational.

Take today’s 100k threshold headlines. Compare “date reported” with “Specimen date”, the latter giving a much better picture:



Quite alarming…



Quite different.

What about the “by specimen date” picture do you think is less alarming? Genuine question, to be clear.

Given you can’t trust any of the greyed out days yet because they are incomplete, and that the earlier data is broadly identical in both graphs with about a day or two time shift (you’d expect a test with a specimen date of today to be reported tomorrow or the day after, give or take, so one graph should broadly be a time shift of the other, which you can roughly see in the shape), I’m not sure what different message I would take from one versus the other.

Give it another few days for the rest of the cases with specimen dates from 18th onwards to finish processing, and those bars will rise, and you’ll get back to a position where one graph is just a time shift of the other (perhaps with more blurring now because the gap between specimen date and processing date is becoming a bit more volatile (I’ve seen suggestion of larger processing delays in some places currently))

[Oldmanmatt]

It’s not always “scaremongering “ but it’s often over sensational.

Take today’s 100k threshold headlines. Compare “date reported” with “Specimen date”, the latter giving a much better picture:



Quite alarming…



Quite different.

What about the “by specimen date” picture do you think is less alarming? Genuine question, to be clear.

Given you can’t trust any of the greyed out days yet because they are incomplete, and that the earlier data is broadly identical in both graphs with about a day or two time shift (you’d expect a test with a specimen date of today to be reported tomorrow or the day after, give or take, so one graph should broadly be a time shift of the other, which you can roughly see in the shape), I’m not sure what different message I would take from one versus the other.

Give it another few days for the rest of the cases with specimen dates from 18th onwards to finish processing, and those bars will rise, and you’ll get back to a position where one graph is just a time shift of the other (perhaps with more blurring now because the gap between specimen date and processing date is becoming a bit more volatile (I’ve seen suggestion of larger processing delays in some places currently))

Because the error on the more recent days is unlikely huge (edit: bar yesterday, which will still be way off). My guess is that it peaked/plateaued about Friday last week. It’s not the precipitous ramp the first graph would suggest and that messaging/measures are having an impact.
Of course, Xmas itself might blow that out of the water.
And, there might be tens of thousands of tests five days late in being reported, but I suspect that’s what inflated todays figure. Wednesdays are always likely to be abnormally high as the combination of late weekend reporting and increased Monday testing, catch up.

[AJM]

Because the error on the more recent days is unlikely huge (edit: bar yesterday, which will still be way off). My guess is that it peaked/plateaued about Friday last week. It’s not the precipitous ramp the first graph would suggest and that messaging/measures are having an impact.
Of course, Xmas itself might blow that out of the water.
And, there might be tens of thousands of tests five days late in being reported, but I suspect that’s what inflated todays figure. Wednesdays are always likely to be abnormally high as the combination of late weekend reporting and increased Monday testing, catch up.

Time will tell, but I’m not so sure. I don’t know how much geographic respreading they do for processing they tend to do - I assume not much given the desire for speedy turnaround and given the issues with the lab they uncovered a few months back predominantly hit the SW. I could easily see a localised spike in demand around London, delays on postal samples due to Xmas post load, staff absences through covid through the chain and so on pushing that delay out far more than we are used to.

If you think that the end of school term/last Friday before Christmas for many is potentially quite a powerful downwards pressure on infection rates, that shouldn’t be feeding through until after the weekend, given a few days from infection to infectious. Cases peaking on the Friday would imply infections peaked early-mid week, and I can’t obviously see why.

Other slight concern on disengagement with testing as the “you’ll miss Christmas” deadline approaches. That was middle of last week under the old rules, which is a concerning coincidence if the data does turn out as you suggest, because that’s not real.

I think we’d be very lucky, personally. But I guess we get to see the level of restatement on the 18/19 over the next couple of days.

[AJM]

My only other thought is that even if Wednesdays are inflated as you suggest, comparing to 7 days beforehand should strip that distortion out, and hardly shows a pretty picture (that’s the datapoint at just under 80k, if I’ve counted back correctly, so something like a 25-30% increase week on week)

Edit - >30%, I think - 106/78 is roughly 4/3 so 133%?

[Oldmanmatt]

My only other thought is that even if Wednesdays are inflated as you suggest, comparing to 7 days beforehand should strip that distortion out, and hardly shows a pretty picture (that’s the datapoint at just under 80k, if I’ve counted back correctly, so something like a 25-30% increase week on week)

Edit - >30%, I think - 106/78 is roughly 4/3 so 133%?

Oh I’m not trying to suggest there isn’t a wave, just that it’s (I really hope) possibly not the wave that’s being painted. If the transmissibility is as high as suspected, then the measures must be having a moderating effect. Less the restrictions,yet, more the public awareness and experience coupled with vaccine protection; at a guess. 

[abarro81]

Because I'm a prick, I just wanted to quickly come back to say I told you so to Pete 

Alex - yes I was careful in wording it 'severity of outcome'. Not 'severity of omicron' or some-such.

As I guessed, they did indeed include severity of outcome in the model, it was inherent severity of omicron that was modelled as being the same:
https://twitter.com/_nickdavies/status/1473941675634483200

So the answer to

I can't comprehend why modellers tasked by government wouldn't include available real-world evidence of omicron in a population that shows it doesn't have the same severity of outcome as Delta.

is that they did include it.

And as a bonus, if I've understood it right, the inherent severity does now look likely to be a bit lower too.

[danm]

Something not being talked about is long covid - so far there is no evidence that omicron will be less likely to lead to it. Death rate gets all the headlines but the long term impacts of a population ravaged with post viral illness are likely to be huge.

[slab_happy]

Because I'm a prick, I just wanted to quickly come back to say I told you so to Pete 

Alex - yes I was careful in wording it 'severity of outcome'. Not 'severity of omicron' or some-such.

As I guessed, they did indeed include severity of outcome in the model, it was inherent severity of omicron that was modelled as being the same:
https://twitter.com/_nickdavies/status/1473941675634483200

So the answer to

I can't comprehend why modellers tasked by government wouldn't include available real-world evidence of omicron in a population that shows it doesn't have the same severity of outcome as Delta.

is that they did include it.

And as a bonus, if I've understood it right, the inherent severity does now look likely to be a bit lower too.

Obligatory caveat that this is very early days and very small numbers: the Imperial study suggests that for unvaxxed people who haven't had Covid before, the risk of hospitalization is circa 11% lower with Omicron than Delta.  Which would indicate lower inherent severity, but not hugely so.
 
We're definitely seeing lower severity of outcome, but 40-50% reduction in overnight-or-longer hospitalizations is not terrific if Omicron keeps spreading at the same speed -- the impact on the health system of halving the infection hospitalization rate gets cancelled out if cases double, whch at current rates takes 2 days.

It'll all look better if cases level off and then start to fall very soon; it looks like this is happening in SA, fingers crossed, but we don't know why:

https://twitter.com/jburnmurdoch/status/1473985139948535815
https://nymag.com/intelligencer/2021/12/the-south-africa-omicron-wave-is-already-peaking-why.html

[Fiend]

Talking of confusing messages in the mainstream state media....

There are also signs that the effect of booster doses is waning.

Two doses of a vaccine were shown to offer limited protection against catching Omicron, which was then restored with a booster dose.

However, the report says this protection drops by between 15% and 25% after 10 weeks. This is still better than having no booster dose and the protection against severe disease or death is likely to be even greater.

(From BBCCCP)

Is this gobbledegook or what?? Their previous article (sorry couldn't find it) specifically referred to a 3rd vaxx booster being highly effective at avoiding Omicron symptoms (it was just "symptoms", not "serious illness").

Throughout this debacle I've got the impression that vaccines offer "some" protection against being infected with Covid-5G, and "more" protection against getting serious illness from it (the latter seems to have been very much proven by high cases vs low hospitalisations throughout a summer / autumn of Delta).

The BBC seems to be mixing up two different measurements in it's Omi reporting: transmission and illness. Which doesn't strike me as very helpful.


Can someone who follows these things just provide some simple figures for me:

% vaxx effectiveness against catching Covid

% vaxx effectiveness against passing on Covid (if this is useful)

% vaxx effectiveness against getting symptoms (if this is useful)

% vaxx effectiveness against getting serious illness 

Feel free to do a second lot of figures for all of the above, swapping "vaxx" for "booster" and "Covid" for "Omicron".


Ta!

Edit: Feel free to answer in runes or cave paintings if it's more appropriate.

[abarro81]

What do you find confusing? Seems to make sense to me...

[Fiend]

Two doses of a vaccine were shown to offer limited protection against catching Omicron, which was then restored with a booster dose.

-vs-

Their previous article (sorry couldn't find it) specifically referred to a 3rd vaxx booster being highly effective at avoiding Omicron symptoms (it was just "symptoms", not "serious illness").

(Noting the latter article didn't mention much about transmission reduction, just symptom prevention.

Which is the important factor, and why do they keep switching between talking about two different aspects??

[slab_happy]

Can someone who follows these things just provide some simple figures for me:

*hollow laughter*

Less simple than you think, because you'd have a different set of figures for each vaccine in relation to each variant. And it's now clear that the vaccines wane in effectiveness over time. So, different figures per vaccine per variant per time period since vaccination. And now for every different combo of vaccines and boosters ...

But your understanding is completely correct -- you've got two very different measures, protection against catching Covid and protection against getting severely ill from Covid.

And the latter tends to be MUCH stronger and hold up much better than the former.

Relevant thing to explain: lots of info and studies will refer to "symptoms" or "symptomatic illness" as their measure, because unless you've got a study population where everyone is getting PCR-tested on a regular and frequent basis (which takes a LOT more infrastructure), you're not going to catch completely asymptomatic cases.

So a lot of vaccine studies will go for testing anyone in their population who gets possible Covid-type symptoms (this has happened to me various times in the ENSEMBLE2 trial, even though I'm now in the "control group of people on rival vaccines"  -- getting a headache or a cold requires me to report it in the app and then I have to do multiple PCRs).

"Symptomatic illness" is the quicker-and-dirtier measurement of vaccine effectiveness, basically, so we've generally got a lot more data on that than on the risk of catching Covid full stop (including completely asymptomatic cases).

And people will often use that as interchangeable with "catching Covid", even though you are correct that it's not exactly the same. But it often gets used as a sort of proxy for it.

[abarro81]

I'm still not sure I understand your objection? I suspect it's best to read "avoiding omicron symptoms" as "avoiding symptomatic infection" if that helps and that's what you're driving at? If not I don't understand still...

Bear in mind that most large-scale data will probably (I'm making an assumption here) look at avoiding symptomatic infection as avoiding infection per se is harder to measure unless you have groups you're regularly testing even if asymptomatic and you use them for your study (rather than the general public which typically gets tested when symptomatic)

[Beaten to it and with a clearer explanation]

[slab_happy]

I'm still not sure I understand your objection? I suspect it's best to read "avoiding omicron symptoms" as "avoiding symptomatic infection" if that helps and that's what you're driving at? If not I don't understand still...

Bear in mind that most large-scale data will probably (I'm making an assumption here) look at avoiding symptomatic infection as avoiding infection per se is harder to measure unless you have groups you're regularly testing even if asymptomatic and you use them for your study (rather than the general public which typically gets tested when symptomatic)

[Beaten to it and with a clearer explanation]

Yeah, this is why the SIREN study is a big deal -- it's a cohort of 10,000 healthcare workers who get tested every 2-4 weeks with PCR and antibody tests, starting May 2020. So LOTS of data coming out of that, especially about things like reinfection risk.

But it's a shit-tonne more work to do that than just to test anyone who shows possible symptoms. And Covid vaccine trials have involved much larger numbers of people.

And then real-world data is even "dirtier", because you're looking at "people who had a positive test either because they tested after having symptoms, or tested as a precautionary measure before doing a thing, or because it's a work requirement". Which will pick up some people who might be asymptomatic but not all.

[Fiend]

Cheers jab_happy and barras, that does clear it up.

Maybe I was a bit surprised that symptomatic infection is used as a benchmark as I've seen enough official posters around warning "1 in 3 people get the covid without any symptoms, take care get jabbed get tested wotever", so I assumed "they", whoever they are, were differentiating between symptomatic infections and total infections.

I see now that the quotes in my 2nd, shorter, post are used to mean pretty much the same thing. And yes the distinction between those/that, and "getting serious illness" is clear.




Now is the Grinah guide pdf available to download yet....??

[slab_happy]

(In fact, those of us who had AZ get an upgrade to our protection, because we get in on that funky mix-and-match action.)

Does this work for other types too?. I.e. as someone who had two Pfizer doses, given the choice, should I get a Moderna booster?

Unclear -- Pfizer and Moderna are very similar (not just the same type of vaccine, but almost identical) so they don't have the effect of mixing different mechanisms of action that you do when mixing mRNA vaccines with viral vector vaccines.

The Lancet study got a response that was a bit higher when stacking a full dose of Moderna on top of two Pfizers:

https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(21)02717-3/fulltext

However, that may just be because Moderna went for a relatively higher dose as their "standard" dose than Pfizer did (which is also why it can hit a bit harder with side-effects).

Which is why they're using half-doses of Moderna for boosters, versus full doses of Pfizer.

On the basis of what we currently know, looks like three doses of any mRNA vax performs VERY strongly, so personally I wouldn't delay getting boosted in order to get a particular one.

Some UKHSA data suggesting that if you have a Moderna booster after 2 Pfizer, the protection from the booster may wane more slowly than if you have a third Pfizer:

https://twitter.com/freja_kirsebom/status/1474070643079536649

So that suggests mixing it up is worthwhile if you have the opportunity.

But that's one note in the middle of a picture with some bad news (as well as some very hopeful news re: hospitalizations): data generally suggests pretty rapid waning of protection against symptomatic infection with Omicron from boosters. Not great.

Scientists are hoping that protection against severe illness will remain much higher -- it should, based on the patterns we've see through the pandemic (and the apparent reduced risk of hospitalization), but we don't have the data yet to be certain of what the picture is.

Excellent breakdown of the whole UKHSA Variant Technical Briefing from Meaghan Kall in this thread:

https://twitter.com/kallmemeg/status/1474071983067389958

[AJM]

My only other thought is that even if Wednesdays are inflated as you suggest, comparing to 7 days beforehand should strip that distortion out, and hardly shows a pretty picture (that’s the datapoint at just under 80k, if I’ve counted back correctly, so something like a 25-30% increase week on week)

Edit - >30%, I think - 106/78 is roughly 4/3 so 133%?

Oh I’m not trying to suggest there isn’t a wave, just that it’s (I really hope) possibly not the wave that’s being painted. If the transmissibility is as high as suspected, then the measures must be having a moderating effect. Less the restrictions,yet, more the public awareness and experience coupled with vaccine protection; at a guess.

Just looking at this a few days on, the 18/19 have risen slightly, maybe 5k cases up or something.

The 20th was massively incomplete though since that has put on a good 30k cases to circa 110k.

21st currently lower but based on this experience this is probably very incomplete still.

https://coronavirus.data.gov.uk/details/cases?areaType=overview&areaName=United%20Kingdom

[Sidehaas]

If you want to estimate how many cases are still likely to be added for specimen dates in the previous 5 days,  the bed way is to go to the dashboard page for England (vs UK) and look at the "Daily change in reported cases by specimen date" graph. This data isn't available for whole UK. From it, you can see what proportion of a specimen date's cases tend to get added after 1 day, after 2 days etc. This is fairly consistent and only changes slowly when testing starts to struggle.
Tuesday and Wednesday from this week are each looking likely to be higher than the day before again.

[Offwidth]

IndieSAGE last Friday had a plot of PCR test % meeting the 24 hour return target: that was dropping fast in data from a week and a half ago. This shows we simply can't trust daily PCR data right now. Add to that people who won't get tested so they don't ruin xmas, plus a likely much larger proportion of omicron infected who think they have a cold so don't get a test (latest reports say 1 in 2 of 'colds' in London are probably omicron), plus xmas disruption, plus some reported logistics trouble accessing tests.

ONS data will be much more accurate on population levels of infection but more delayed.

https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/articles/coronaviruscovid19/latestinsights

The key was always the number of new omicron hospital admissions (on top of all the other NHS pressures, esp staff off sick or isolating with omicron), but not just that... it's also important how ill they are (if on average they stay in hospital fewer days, the NHS pressure will be reduced cf a similar number of delta hospitalised)

[petejh]

Because I'm a prick, I just wanted to quickly come back to say I told you so to Pete 

Alex - yes I was careful in wording it 'severity of outcome'. Not 'severity of omicron' or some-such.

As I guessed, they did indeed include severity of outcome in the model, it was inherent severity of omicron that was modelled as being the same:
https://twitter.com/_nickdavies/status/1473941675634483200

So the answer to

I can't comprehend why modellers tasked by government wouldn't include available real-world evidence of omicron in a population that shows it doesn't have the same severity of outcome as Delta.

is that they did include it.

And as a bonus, if I've understood it right, the inherent severity does now look likely to be a bit lower too.

It isn’t clear from your link? You’d need to look at the original LSHTM study, not ‘Imperial report #50’.

It doesn’t tally with what was said by Graham Medley (LSHTM) about the assumptions used in their scenario modelling, if you read his exchange.

Either way it makes no difference to the point, which is that the media and government over the past two weeks stoked panic by pushing a narrative of the high numbers given in the more worse-case scenarios, with little to no regard given to likelihood of these scenarios actually occurring, and there was little to no discussion of the (much more likely) better case scenarios.
Best reflected in the widely reported selective quoting by government and media of Ferguson’s red flag line: ‘we see no evidence for different severity compared to Delta’.

Despite all the evidence that everyone could see, that suggested something different is in fact going on in real life. 🙄

[abarro81]

It isn’t clear from your link? You’d need to look at the original LSHTM study, not ‘Imperial report #50’.

It doesn’t tally with what was said by Graham Medley (LSHTM) about the assumptions used in their scenario modelling, if you read his exchange.

The guy in the twitter thread is from LSHTM, and explains (quite clearly IMO) that they modelled the same inherent severity but vaccines and previous infection make severity of outcome lower in the model outputs[1]. I don't know what you think is unclear or contradictory to the Graham Medley exchange, which appears to be a discussion of inherent severity (caveat - I've only perused his Twitter)?

Either way it makes no difference to the point,

It makes a significant difference to part of your original point, which was ranting about the modellers not including things in the model:

I can't comprehend why modellers tasked by government wouldn't include available real-world evidence of omicron in a population that shows it doesn't have the same severity of outcome as Delta.

but as he explains they already were predicting lower severity of outcome (this was a model output not an input - presumably vaccine efficacy and inherent severity are the main inputs), it was just a question of how much less the severity would be.

But yes, lots of politicians and media both misunderstand modelling and report it with an underlying agenda. I imagine you'll have a certain empathy on at least the former point 

[1] "There was some confusion over what our models, which assumed Omicron had equal "baseline" severity to Delta, meant for severity in practice. This leads to around a 40% reduction in realised severity within each age group, because more Omicron cases are breakthrough/reinfections." - from my link

[petejh]

Why do you think I’d have a certain empathy on at least the former point?

Have you actually read the Medley/Nelson exchange on Twitter?
Or read/listened to any media reporting over the last 2-3 weeks about the omicron wave? To say the narrative was doom-mongering and focussing on the worst case would be an massive understatement.

Headline in Torygraph today: ‘We are not issuing doomsday warnings says top scientist’.
Probably just a random headline, because nobody thinks they were, right?

I don’t think SAGE or other scientists *were* issuing doomsday warning - I think they were just doing their jobs and giving the various possible outcomes.
I 100% think the media and government spin-artists were though.

[abarro81]

Why do you think I’d have a certain empathy on at least the former point?

It was a joke Pete, because - as far as I can tell - you've misunderstood how they did the modelling.

Have you actually read the Medley/Nelson exchange on Twitter?

I read something earlier between those two, I don't know if it was the thread you were referring to but I assume so as it involved him saying something like "we model what we're asked to model" and the stuff about only modelling interesting scenarios. I didn't see anything in there that backed up your statement about their model involving equal severity of outcome to delta. Please post it up. I've shown you why I think you were wrong (one of the modellers explicitly laying it out), you've not shown my why you think you were right, if you still do?

Yeah, media are often lame. Cool. But your objection to the model having the same severity of outcome is also lame, because as far as I can tell it's wrong. There's a reason I didn't respond to the media-focused parts of your  posts - it's not an area where I had something I wanted to opine on. I mostly listen to PM if I listen to news. I don't recall the narrative being much other than "we don't really know what will happen because we don't know quite how well vaccines will stand up to it and how severe it will be, if we do nothing it's rolling a dice/hoping for the best". Which all seems like a pretty sensible conclusion. They've seemed at pains to explain the uncertainties in a way that even idiots could understand. I read the guardian sometimes but have no objection to doing my own interpretation on statements including words like "up to". After all, if I play the lottery I know I could win up to the X million jackpot, not that that's the most probable outcome.

[slab_happy]

Really good thread on why it matters to unpack the difference between severity of outcome and intrinsic severity (and why intrinsic severity still matters even in a country with high vaccination levels):

https://twitter.com/roby_bhatt/status/1475052548331610112