COVID-19

[Woodchopper]

Impact of the COVID-19 nonpharmaceutical interventions on influenza and other respiratory viral infections in New Zealand

Stringent nonpharmaceutical interventions (NPIs) such as lockdowns and border closures are not currently recommended for pandemic influenza control. New Zealand used these NPIs to eliminate coronavirus disease 2019 during its first wave. Using multiple surveillance systems, we observed a parallel and unprecedented reduction of influenza and other respiratory viral infections in 2020. This finding supports the use of these NPIs for controlling pandemic influenza and other severe respiratory viral threats.

https://www.nature.com/articles/s41467-021-21157-9

[basementer]

Three cases found in the community, so Auckland locks down by two levels, the rest of the country one level, as of midnight tonight. To be reviewed every 24 hours.
I think most people I know would be OK with acting fast and waiting for the all-clear.

The new cases have been identified as the UK VoC.

[OffTheRock]

Guernsey outbreak eventually got confirmed as being Kent variant. New cases now down to single digits per day with most known contacts of confirmed cases rather than from unknown community spread. 202 active cases currently. None of the cases in care homes have needed hospital care, which might be good news for the vaccine. They might start relaxing the lockdown as early as Thursday but depends on what happens with the unknown community cases and the what additional testing in high risk areas finds.

Good luck in Auckland.

[basementer]

Guernsey outbreak eventually got confirmed as being Kent variant. New cases now down to single digits per day with most known contacts of confirmed cases rather than from unknown community spread. 202 active cases currently. None of the cases in care homes have needed hospital care, which might be good news for the vaccine. They might start relaxing the lockdown as early as Thursday but depends on what happens with the unknown community cases and the what additional testing in high risk areas finds.

Good luck in Auckland.

I'll be staying out of Auckland personally, but thanks. And i wish you well in turn.

[OffTheRock]

Sneaky little f.cker this UK variant. I doubt it's a coincidence Gsy managed to keep Covid out for 8 months and then ended up with more cases in 2-3 weeks than they had in the whole of the first outbreak.

[OffTheRock]

4 new cases in the last 24 hrs. All from known contacts. Unfortunately also the 1st death of this outbreak, which brings the total number of deaths to 14.

[jimbob]

I wonder what subtle analysis of the case numbers having been consistently falling for the past month leads them to suggest that R is less than 1?

More seriously how do they get that it was more than 1 up until now, despite cases halving roughly every two weeks?

It was something that I had been wondering too.

[lpm]

Is the fall in UK cases slowing?

A quick look at the figures hints that it is. Cases were halving every 2 weeks in Jan, perhaps a bit better. But in Feb it's taken around 18 days to halve.

There would be no clear lockdown reason why. Maybe you could spin a narrative of everyone writing off Jan and staying home in bad weather, then being fed up of it in Feb.

Or the mix of variants? The English Variant becoming ever more dominant?

Too soon to be helped by vaccines - Jan doses concentrated among non-superspreaders.

[bob sterman]

Is the fall in UK cases slowing?

It does look that way...and if SAGE continue to be behind the curve with their estimates of R we could have trouble ahead...

https://twitter.com/jamesannan/status/1 ... 2976528395

[lpm]

Ah yes, should have guessed James Annan would already be all over it.

His 3rd chart is always the quick & easy way to see it, that even Johnson might understand. Just the day's figure divided by the daily figure from a week earlier. If there really is a slowing down of the fall, the red line and blue line will creep upwards from the purple.

[sTeamTraen]

Last Saturday evening, David Davis MP went on Twitter to promote a preprint from a hospital in Barcelona that claimed an 80% reduction in ICU admissions and a 60% reduction in deaths simply by administering Vitamin D.

I'm pleased to report that this preprint has now been taken down, and that I may have been able to contribute to that result.

[KAJ]

Is the fall in UK cases slowing?

A quick look at the figures hints that it is. Cases were halving every 2 weeks in Jan, perhaps a bit better. But in Feb it's taken around 18 days to halve.

There would be no clear lockdown reason why. Maybe you could spin a narrative of everyone writing off Jan and staying home in bad weather, then being fed up of it in Feb.

Or the mix of variants? The English Variant becoming ever more dominant?

Too soon to be helped by vaccines - Jan doses concentrated among non-superspreaders.

Over the past 6 weeks (+ 5 days "incomplete") the decline in cases by specimen date has been almost exactly exponential (with a substantial day-of-week effect). Note negligible quadratic term and R-sq of 99%.

SpecCases.png (15.47 KiB) Viewed 3788 times

Code: Select all

At 17/02 fit = 10056.7 with halving time = 15.7 days. That time halving in 513.9 days 

Analysis of Variance Table

Response: log(SpecCases)
              Df  Sum Sq Mean Sq  F value    Pr(>F)    
poly(date, 2)  2 10.8612  5.4306 1553.632 < 2.2e-16 ***
day            6  0.9988  0.1665   47.625 7.388e-15 ***
Residuals     33  0.1153  0.0035                       
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Call:
lm(formula = as.formula(aform), data = aDF, weights = weights)

Weighted Residuals:
      Min        1Q    Median        3Q       Max 
-0.170401 -0.015843  0.003698  0.022824  0.099785 

Coefficients:
               Estimate Std. Error t value Pr(>|t|)    
(Intercept)     9.87580    0.02456 402.029  < 2e-16 ***
poly(date, 2)1 -4.01363    0.07798 -51.472  < 2e-16 ***
poly(date, 2)2 -0.02420    0.07837  -0.309    0.759    
dayMon          0.45547    0.03414  13.340 7.59e-15 ***
dayTue          0.36679    0.03417  10.735 2.66e-12 ***
dayWed          0.34027    0.03421   9.946 1.85e-11 ***
dayThu          0.30505    0.03427   8.901 2.74e-10 ***
dayFri          0.26685    0.03435   7.769 5.96e-09 ***
daySat          0.05467    0.03414   1.601    0.119    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.05912 on 33 degrees of freedom
Multiple R-squared:  0.9904

[Bird on a Fire]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

[shpalman]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

It's always exponential - the rate at which new infections occur depends on the number of current infections. An exponential decrease just means that each infected person on average infects less than one new person.

Once it drops below 1 person, in any self-contained group, it drops to zero.

[KAJ]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

I'm not equipped to reasonably speculate about mechanisms.
A constant exponential decrease corresponds to a constant R < 1, in the same way that a constant exponential growth corresponds to a constant R > 1.
Exponential decline never reaches zero, but - especially when numbers are small - does give fractional case numbers; i.e. an exponential declining model is inappropriate at small numbers.

[KAJ]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

It's always exponential - the rate at which new infections occur depends on the number of current infections. An exponential decrease just means that each infected person on average infects less than one new person.

Once it drops below 1 person, in any self-contained group, it drops to zero.

IMHO bolded part is only meaningfully true for constant R.

[shpalman]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

It's always exponential - the rate at which new infections occur depends on the number of current infections. An exponential decrease just means that each infected person on average infects less than one new person.

Once it drops below 1 person, in any self-contained group, it drops to zero.

IMHO bolded part is only meaningfully true for constant R.

Well ok then if R isn't constant it won't look like an exponential because the exponential rate will be changing with time, but the underlying mechanism is still "the rate at which new infections occur depends on the number of current infections" - just that the "depends" is changing.

In the case that we are talking about, it does look like an exponential.

[sTeamTraen]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

I think that (like many countries) the UK is in a phase where the measures in place are stopping exponential growth, but there is a constant background level of transmission due to people breaking/bending the rules, as well as unavoidable transmission due to interactions that can't be stopped (e.g., you can never make a factory or supermarket completely safe).

Going to the next level in terms of numbers—in the UK, getting new cases down below 1,000 per day, say, at which point track'n'trace might actually be a meaningful thing to attempt—will require tougher measures, or at least tougher in certain areas. But there seems to be little political appetite for that.

My worry is that we will get to May and have countries (again, not just trhe UK) that say "fine, the over-55s are vaccinated, everyone else has 99.8% survival, let's open up fully". And then you could have 100,000 cases or more per day, which would still f.ck the hospitals, just with younger patients than at present. And with that many cases, you have more scope for nastier variants to appear.

[KAJ]

It's always exponential - the rate at which new infections occur depends on the number of current infections. An exponential decrease just means that each infected person on average infects less than one new person.

Once it drops below 1 person, in any self-contained group, it drops to zero.

IMHO bolded part is only meaningfully true for constant R.

Well ok then if R isn't constant it won't look like an exponential because the exponential rate will be changing with time, but the underlying mechanism is still "the rate at which new infections occur depends on the number of current infections" - just that the "depends" is changing.

In the case that we are talking about, it does look like an exponential.

I think we're debating the meanings of words here, sometimes entertaining but not as often productive.

I'm carefully avoiding opining on "the underlying mechanism" and sticking to fitting curves to data.

There are many growth/decay curves where the rate of change depends (for some meaning of "depends") on the current number. If that dependence is simple proportionality, corresponding to a "first order reaction", then the curve is exponential with a rate constant which is, well, constant.

If the rate constant is allowed to be time varying then I think the exponential growth/decay model can be fitted to any continuous growth/decay curve. But then I don't think it is meaningful to call it exponential.

[shpalman]

If you look at James Annan's models, his R is slowly varying most of the time and then jumps to a different value when the government changes the rules.

If you're fitting a polynomial, which is at most second order, to the logarithm of the data then you're already assuming that the data is exponential with a slowly varying rate. And it kind of is, or else it just wouldn't fit that well.

[shpalman]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

I think that (like many countries) the UK is in a phase where the measures in place are stopping exponential growth, but there is a constant background level of transmission due to people breaking/bending the rules, as well as unavoidable transmission due to interactions that can't be stopped (e.g., you can never make a factory or supermarket completely safe).

Well, that means R will never be zero but it doesn't actually have to be. The issue is more that if people don't feel like there's that much covid about they'll start behaving in a way which will end up putting R somewhere close to 1 in that environment, even if the government doesn't relax the rules officially. It's an unstable system, but it might be somewhat "self limited" by the actions of people behaving as if covid isn't their problem until suddenly it is. (i.e. if a group of "it won't happen to us" people realize that covid got into their group it will only spread in that group in a limited way because the people will react to it in a way which, if they'd behaved like that in the first place, it wouldn't have infected anyone else in the group in the first place.)

Of course you can develop a model to "predict" any answer once you already know what the answer is, but I think we already know that most people don't give covid to anyone else whereas most people who catch it, catch it at the same time from a minority of "super spreaders".

What we seem to have achieved in Italy is a system which maintains new cases somewhere just above 10,000 per day, with a region's rules being toughened if things get worse and loosened once they get better again. There's obviously no will to just keep the tough measures imposed until cases fall to the levels we had at the end of last spring.

Going to the next level in terms of numbers—in the UK, getting new cases down below 1,000 per day, say, at which point track'n'trace might actually be a meaningful thing to attempt—will require tougher measures, or at least tougher in certain areas. But there seems to be little political appetite for that.

People don't seem to want to accept tougher measures for a shorter time in order to make things better overall.

Sometimes I see social media posts, from people who aren't able to earn during lockdown, saying things like "we should not pay that politician until covid is gone" or something. I mean, I'd be find with not being paid until covid was gone if it went along with a real total hard police-state lockdown for a couple of months because I have the savings (and the work from hominess) to cope with that, and I bet the politicians do too, but the pub owner doesn't.

Then there are also restaurant owners (or similar) on the news here saying that they don't want the government handouts they want to be allowed to work NO YOU f.cking DON'T THIS WILL NEVER BE OVER IF WE DO THAT.

My worry is that we will get to May and have countries (again, not just trhe UK) that say "fine, the over-55s are vaccinated, everyone else has 99.8% survival, let's open up fully". And then you could have 100,000 cases or more per day, which would still f.ck the hospitals, just with younger patients than at present. And with that many cases, you have more scope for nastier variants to appear.

Yes, I agree with this, it's kind of why I think we should anyway aim for zero covid and not consider an acceptable number of deaths to be acceptable. With 100,000 cases per day it's going to be way harder to stop it being 200,000 cases per day the following week. And then 1 million cases per day the following month. I think people will lock themselves down out of terror at some point, but with the lag between infections and symptoms and then the lag between mild symptoms and severe symptoms, and the lag between severe symptoms and death, there'll already be a whole load of dead persons walking by then.

[shpalman]

Now I'm increasingly hearing that the Yellow restrictions just don't work and the whole of Italy should be at least Orange. But for the moment quite a lot of Italy will remain Yellow, including Lombardy, even if there are a four councils in the region which are separate self-contained Red zones due to localized outbreaks. Some regions will move from Yellow to Orange this weekend.

[KAJ]

If you look at James Annan's models, his R is slowly varying most of the time and then jumps to a different value when the government changes the rules.

If you're fitting a polynomial, which is at most second order, to the logarithm of the data then you're already assuming that the data is exponential with a slowly varying rate. And it kind of is, or else it just wouldn't fit that well.

I think we'll have to agree to differ on terminology. When I read "exponential with a varying rate" I think "not exponential", for "exponential with slowly varying rate" I think "approximately exponential".

As I've said before, I'm using regression for purely descriptive purposes, seeking the simplest description which adequately (!) captures the variation. I plot on log-linear scales because I expect a straight line on those axes (i.e. an exponential model) to capture much of the variation, so the description can be a simple constant doubling/halving time.

I include a quadratic term really to see if that description is inadequate. If simple curvature substantially improves the fit then the exponential model doesn't fit well and a simple constant doubling/halving time is an inadequate description. I've played with higher order terms but don't really regard them as useful in this context.

I posted in response to lpm's question:

Is the fall in UK cases slowing?

A quick look at the figures hints that it is. Cases were halving every 2 weeks in Jan, perhaps a bit better. But in Feb it's taken around 18 days to halve.

... and pointed out that the last six weeks had "been almost exactly exponential .... Note negligible quadratic term and R-sq of 99%", intending (but maybe not succeeding) to say that the halving rate was effectively constant. I didn't show the fit without the quadratic term, but the size of that coefficient indicates it wouldn't have been substantially different. Using yesterdays data the comparison is here, note the similarity in RSS:

Code: Select all

Analysis of Variance Table

Model 1: log(SpecCases) ~ poly(date, 2) + day
Model 2: log(SpecCases) ~ date + day
  Res.Df     RSS Df  Sum of Sq      F Pr(>F)
1     33 0.10304                            
2     34 0.10679 -1 -0.0037446 1.1992 0.2814

[shpalman]

If you look at James Annan's models, his R is slowly varying most of the time and then jumps to a different value when the government changes the rules.

If you're fitting a polynomial, which is at most second order, to the logarithm of the data then you're already assuming that the data is exponential with a slowly varying rate. And it kind of is, or else it just wouldn't fit that well.

I think we'll have to agree to differ on terminology. When I read "exponential with a varying rate" I think "not exponential", for "exponential with slowly varying rate" I think "approximately exponential".

When I write "exponential with slowly varying rate" I intend "can be described as exponential over a limited time span".

"approximately exponential" could also mean "over a long term such that fluctuations away from the ideal behaviour smooth out" but in this case would mean "getting increasingly inaccurate the further away we get" (from the "limited time span" I mentioned). As such it can't really be extrapolated very far.

[jimbob]

So what would an exponential decrease suggest, mechanistically? That there's some asymptotic limit below which transmission won't fall for other reasons? In which case, getting below that limit would require a further development.

I'd say it suggests that although it's a difficult term to determine, R is a reasonable parameter to explain what's happening.


Deaths declined almost perfectly exponentially from the first peak to August