COVID-19

[wilsontown]

I was taking a look at the interactive map that is available on the governments stats pages. Most areas around York now have fairly low case levels, but there are some areas with horrendous numbers - around here Wetherby, South Cave and Hatfield seem to be particularly struggling for whatever reason. So as overall numbers fall things are likely to get less predictable.

[Trinucleus]

I was taking a look at the interactive map that is available on the governments stats pages. Most areas around York now have fairly low case levels, but there are some areas with horrendous numbers - around here Wetherby, South Cave and Hatfield seem to be particularly struggling for whatever reason. So as overall numbers fall things are likely to get less predictable.

Aren't there prisons round there?

[wilsontown]

I was taking a look at the interactive map that is available on the governments stats pages. Most areas around York now have fairly low case levels, but there are some areas with horrendous numbers - around here Wetherby, South Cave and Hatfield seem to be particularly struggling for whatever reason. So as overall numbers fall things are likely to get less predictable.

Aren't there prisons round there?

Indeed there are, I wasn't aware. HMP Humber (between North and South Cave), HMP Hatfield, and HMP Wetherby. There is also HMP Wealstun at Boston Spa, not far from Wetherby.

[OffTheRock]

Prisons, care homes, schools, food processing plants. You get that here in the SW where cases are low. It definitely gets unreliable once the numbers of cases is so low that they are supressed on the map. Areas that had <3 cases suddenly end up with10-20 a week later and then it drops back again nearly as quickly. Especially during lockdown where those additional cases don't seem to spread much. The areas that are purple or dark blue in a sea of green and white almost always seem to be linked to an outbreak in a specific place here rather than generally high levels in that area.

[Millennie Al]

That would imply that simple exponential models will start to fall over at higher prevalences than for other infections

Any models will fail because people behave differently depending on lots of factors, but especially the disease prevalence. Once the numbers get down to levels where people start to lose their fear of the disease, they'll change their behaviour, which will then cause the spread to vary, with a feedback loop which includes delay liable to cause erratic swings.

There seems to be very many vociferous people who seem to think that any advantage gained must immediately result in more permissive rules. As if we want to leave the virus a fair chance rather than just suppressing it completely. Even Chris Whitty was doing it, as reported at https://www.telegraph.co.uk/global-heal ... ine-cases/

Chris Whitty reiterates that the risk to children "is incredibly low from going to school".

"One of the few good things about Covid is that the risks are so much smaller than they are for adults," he says.

"The residual risk is strongly in favour from a child's point of view of going to school and the data on that are unambiguous."

It's amazing that that can be reported after so long when it has been obvious that the major risk of schools is that students spread the disease from one family to another - not that they get seriously ill themselves.

[jimbob]

That would imply that simple exponential models will start to fall over at higher prevalences than for other infections

Any models will fail because people behave differently depending on lots of factors, but especially the disease prevalence. Once the numbers get down to levels where people start to lose their fear of the disease, they'll change their behaviour, which will then cause the spread to vary, with a feedback loop which includes delay liable to cause erratic swings.

Yes, but Covid seems more heterogeneous than other infectious diseases, so it will fall over at a higher prevalence

[headshot]

It's amazing that that can be reported after so long when it has been obvious that the major risk of schools is that students spread the disease from one family to another - not that they get seriously ill themselves.

Yes, but the parents of school children are young and so if they get the virus it won’t be serious. (Make sure no-one mentions the grandparents...)

[Herainestold]

Has there been any progress on a vaccine for un der 18s? I did a quick google and didn't come up with anything.

[Millennie Al]

Has there been any progress on a vaccine for un der 18s? I did a quick google and didn't come up with anything.

It'll almost certainly be the same vaccines used in adults. It's just that young people are at such lower risk from the disease that it's more important to be cautious in providing treatment - in addition to the usual caution over treatments for children anyway.

See https://www.nihr.ac.uk/news/first-child ... open/26870 where they are recruiting in Southhampton for the age range of 6-11.

[Sciolus]

And for the same reason, it would be much harder to demonstrate efficacy. I would imagine the current focus is to see if the vaccines reduce transmission; if it doesn't there is little societal benefit to giving it to young people, on top of the low individual benefit.

[Herainestold]

This disease affects groups differently. It is especially bad for POC and BAME people, the elderly, those with hugh BMIs, and strangely enough, males.
Shouldnt there be different vaccines developed for all these subgroups? And how can schools go back when kids aren't vaccinated?
Maybe just different dose vaccines for groups with different risk profiles? Why is nobody asking these questions?

[sTeamTraen]

So this week's quack cure is Proxalutamide, an anti-androgen. As far as I can tell the entire theoretical justfication for trying this is that the observed higher COVID-19 mortality in men might be caused by high levels of testosterone. I have literally seen the fact that male patients with COVID tend to have alopecia cited in support of this, without taking into account the confounding effect of, er, age. ("But look, kids hardly have any symptoms and they all have great hair!!!1!")

Anyway, it seems to be the usual round of fake sh.t. I have a theory that the assorted quacks, charlatans, and frauds who have been waiting for the right moment to release their miracle snake oil are looking at the success of vaccines and the generally falling case numbers worldwide, and looking to offload and cash in as soon as possible. There's already something that seems to be to be close to a smoking gun here.

[wilsontown]

Well, it's the same virus (notwithstanding variants), so why would you need different vaccines?

[Martin_B]

This disease affects groups differently. It is especially bad for POC and BAME people, the elderly, those with hugh BMIs, and strangely enough, males.
Shouldnt there be different vaccines developed for all these subgroups? And how can schools go back when kids aren't vaccinated?
Maybe just different dose vaccines for groups with different risk profiles? Why is nobody asking these questions?

Well, I'm no expert but my take on your 4 questions is:

1) It's pretty remarkable that any vaccines have been developed in 9-12 months, yet alone up to half a dozen. You want different vaccines for each slightly more vulnerable subgroup? Any idea how long it would take, when there's vaccines which appear to work for everyone?

2) Good question. Much more a political decision than one based on science, I suspect.

3) Working out different dose regimes would require significantly longer phases in the trials. Again, why bother when the vaccines appear to work for everyone and roll-out to as many people as possible is important?

4) Do you think no-one is asking these questions? I suspect a great many drug/medical companies are working on what will happen after the pandemic dies down, such as: "How to respond to different strains?" "Do we need targeted drugs for different vulnerable groups?" or "Do different dosage regimes of existing drugs protect different populations better?" But they've been working flat out to get something to work; you're like someone who's been waiting ages for a bus and then complains when it comes that it's the wrong colour!

[wilsontown]

Anyway, I'm sure you're all dying to hear about my machine learning project, so here goes. Using the deaths by date reported numbers, I've trained a recurrent neural network and predicted the next 10 days. On the graph "Actual" is the real numbers, "Predicted" is what the RNN reckons, and "Forecast" is what the RNN reckons for the next 10 days.

You can see it plausibly falls to very low numbers in the next few days. You could figure that out for yourself by just looking at the data, most likely, but we are all data scientists now...

download.png (44.78 KiB) Viewed 2839 times

[Grumble]

Anyway, I'm sure you're all dying to hear about my machine learning project, so here goes. Using the deaths by date reported numbers, I've trained a recurrent neural network and predicted the next 10 days. On the graph "Actual" is the real numbers, "Predicted" is what the RNN reckons, and "Forecast" is what the RNN reckons for the next 10 days.

You can see it plausibly falls to very low numbers in the next few days. You could figure that out for yourself by just looking at the data, most likely, but we are all data scientists now...

download.png

Let’s hope you’re on the money

[Herainestold]

4) Do you think no-one is asking these questions? I suspect a great many drug/medical companies are working on what will happen after the pandemic dies down, such as: "How to respond to different strains?" "Do we need targeted drugs for different vulnerable groups?" or "Do different dosage regimes of existing drugs protect different populations better?" But they've been working flat out to get something to work; you're like someone who's been waiting ages for a bus and then complains when it comes that it's the wrong colour!

There are just so many strange things about this viral disease and this pandemic and now the vaccines. It seems like some of the vaccines have lower efficacy in older people, I wonder about other subgroups. Maybe we will never have this exact information but it would be useful.

Personally I will take any approved jab that is offered to me, I dont wish to denigrate any of the incredible work that has been done so far.

[Herainestold]

So this week's quack cure is Proxalutamide, an anti-androgen. As far as I can tell the entire theoretical justfication for trying this is that the observed higher COVID-19 mortality in men might be caused by high levels of testosterone. I have literally seen the fact that male patients with COVID tend to have alopecia cited in support of this, without taking into account the confounding effect of, er, age. ("But look, kids hardly have any symptoms and they all have great hair!!!1!")

Anyway, it seems to be the usual round of fake sh.t. I have a theory that the assorted quacks, charlatans, and frauds who have been waiting for the right moment to release their miracle snake oil are looking at the success of vaccines and the generally falling case numbers worldwide, and looking to offload and cash in as soon as possible. There's already something that seems to be to be close to a smoking gun here.

Yeah, I think they just made the whole thing up. I wonder if we have hit peak snake oil yet, or if it is just going to keep getting worse.

[Martin_B]

Yeah, I think they just made the whole thing up. I wonder if we have hit peak snake oil yet, or if it is just going to keep getting worse.

Sadly I don't think we'll ever hit peak snake oil

[Herainestold]

Yeah, I think they just made the whole thing up. I wonder if we have hit peak snake oil yet, or if it is just going to keep getting worse.

Sadly I don't think we'll ever hit peak snake oil

I think you are right. I keep thinking that things have to get better, but it just gets loonier.

[KAJ]

Anyway, I'm sure you're all dying to hear about my machine learning project, so here goes. Using the deaths by date reported numbers, I've trained a recurrent neural network and predicted the next 10 days. On the graph "Actual" is the real numbers, "Predicted" is what the RNN reckons, and "Forecast" is what the RNN reckons for the next 10 days.

You can see it plausibly falls to very low numbers in the next few days. You could figure that out for yourself by just looking at the data, most likely, but we are all data scientists now...

download.png

My classical linear regression looks very similar to your neural network and has an R-sq of 99%. I've put this on linear Y scale, like yours. I'm not brave enough to extrapolate

PubDeaths.png (22.17 KiB) Viewed 2764 times

Code: Select all

At 27/02 fit = 288.8 with halving time = 10.3 days. That time halving in 23.7 days 

Analysis of Variance Table

Response: log(PubDeaths)
              Df Sum Sq Mean Sq F value    Pr(>F)    
poly(date, 2)  2 8.0261  4.0130  874.60 < 2.2e-16 ***
day            6 8.1263  1.3544  295.18 < 2.2e-16 ***
Residuals     33 0.1514  0.0046                      
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Coefficients:
               Estimate Std. Error t value Pr(>|t|)    
(Intercept)     5.91078    0.02778 212.765  < 2e-16 ***
poly(date, 2)1 -3.05065    0.06868 -44.419  < 2e-16 ***
poly(date, 2)2 -0.60342    0.06776  -8.905 2.72e-10 ***
dayMon         -0.11959    0.03912  -3.057  0.00441 ** 
dayTue          1.04954    0.03915  26.806  < 2e-16 ***
dayWed          1.04480    0.03920  26.651  < 2e-16 ***
dayThu          0.71231    0.03927  18.139  < 2e-16 ***
dayFri          0.84127    0.03935  21.377  < 2e-16 ***
daySat          0.74507    0.03946  18.882  < 2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.06774 on 33 degrees of freedom
Multiple R-squared:  0.9907

Deaths by date of death doesn't have the pronounced day-of-week effect, this is on a log Y scale.

DateDeaths.png (10.73 KiB) Viewed 2764 times

Code: Select all

At 25/02 fit = 208.6 with .halving time = 8.1 days. That time halving in 20.4 days 

Coefficients:
                Estimate Std. Error t value Pr(>|t|)    
(Intercept)     6.578925   0.009648  681.86   <2e-16 ***
poly(date, 2)1 -3.553064   0.075979  -46.76   <2e-16 ***
poly(date, 2)2 -1.174903   0.077215  -15.22   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.05827 on 39 degrees of freedom
Multiple R-squared:  0.9827

[sTeamTraen]

Well, it's the same virus (notwithstanding variants), so why would you need different vaccines?

I'm not sure if this question was meant seriously or ironically, but here goes anyway:

- The vaccines actually work.
- The companies have developed them, and they seem to work, so they ought to be able to get a return on their R&D dollars even if they weren't first.
- Relatedly, a priori we didn't know if they would work, so at the absolute worst we're dealing with a problem of abundance rather than one of penury, which is very much the side you want to be on.
- There are only so many pharmaceutical manufacturing facilities in the world. I suppose we could commandeer one of Roche's facilities and order them to make the Pfizer vaccine, but it would keep the lawyers busy.
- Diversity of treatments that work is a good thing (it's less likely that a variant will be immune to two vaccines versus one, especially with multiple technologies being used, e.g., mRNA versus not).
- Competition means the price paid by governments everywhere will be lower.

(It was quite an interesting exercise to think up all of these lines of argument, so thanks, whatever the intent of the question!)

[jdc]

I think you've missed the context for that question Steamy (see below quote). Wilsontown's trying to find out why we'd need personalised vaccines for different demographics rather than why we need to have more than one manufacturer developing vaccines against coronavirus.

This disease affects groups differently. It is especially bad for POC and BAME people, the elderly, those with hugh BMIs, and strangely enough, males.
Shouldnt there be different vaccines developed for all these subgroups? And how can schools go back when kids aren't vaccinated?
Maybe just different dose vaccines for groups with different risk profiles? Why is nobody asking these questions?

[Herainestold]

I think you've missed the context for that question Steamy (see below quote). Wilsontown's trying to find out why we'd need personalised vaccines for different demographics rather than why we need to have more than one manufacturer developing vaccines against coronavirus.

This disease affects groups differently. It is especially bad for POC and BAME people, the elderly, those with hugh BMIs, and strangely enough, males.
Shouldnt there be different vaccines developed for all these subgroups? And how can schools go back when kids aren't vaccinated?
Maybe just different dose vaccines for groups with different risk profiles? Why is nobody asking these questions?

My reasoning is like this: Different groups have different susceptibilities and different responses to the virus. Quite large differences in fact.
In treating these different groups you might have to use different strategies and techniques and treatments. So would different vaccines taking into account the differences between groups not make a difference?
Maybe the first time around we didnt know enough and a one size fits all vaccination is the best we could do -and I realize it is amazing that we have these vaccines. Should vaccines not be refined now to reflect these differences, as it seems the pandemic will be around for a while?

[Martin_B]

I think you've missed the context for that question Steamy (see below quote). Wilsontown's trying to find out why we'd need personalised vaccines for different demographics rather than why we need to have more than one manufacturer developing vaccines against coronavirus.

This disease affects groups differently. It is especially bad for POC and BAME people, the elderly, those with hugh BMIs, and strangely enough, males.
Shouldnt there be different vaccines developed for all these subgroups? And how can schools go back when kids aren't vaccinated?
Maybe just different dose vaccines for groups with different risk profiles? Why is nobody asking these questions?

My reasoning is like this: Different groups have different susceptibilities and different responses to the virus. Quite large differences in fact.
In treating these different groups you might have to use different strategies and techniques and treatments. So would different vaccines taking into account the differences between groups not make a difference?
Maybe the first time around we didnt know enough and a one size fits all vaccination is the best we could do -and I realize it is amazing that we have these vaccines. Should vaccines not be refined now to reflect these differences, as it seems the pandemic will be around for a while?

What do you mean: "Maybe the first time around we didnt know enough and a one size fits all vaccination is the best we could do"? This is the first time around. The vaccine designers have been working flat out, and now they're flat-out producing enough vaccine for everyone. The WHO website doesn't reckon that they'll be enough doses produced for one dose per person globally until 2022. Stop complaining about the colour of the bus and be thankful we have a bus.