Daily numbers, graphs, analysis thread 12

[BigChocFrenzy]

Welcome to thread 12 of the daily updates

Resource links:

Slides & data UK govt pressers
UK dashboard sub-national data, local authorities
Beta Uk dashboard deaths, cases, hospitals, tests, partially sub-national
UK stats updated daily by PHE & DHSC
ONS UK statistics for CV related deaths, released weekly each Tuesday
PHE surveillance report infections & deaths released every Thursday with sep. infographic
NHS England stats including breakdown by Hospital Trust
FT Daily updates
HSJ Healthcare updates
Worldometer UK page
Plot FT graphs compare countries deaths, cases / million pop. / log / linear
Covidly.com filter graphs compare countries
Plot COVID Graphs Our World in Data

We welcome factual, data driven, and civil discussions from all contributors 📈📶👍

Zoe app. Numbers are rising slightly every day. Very very frustrating

I've been getting urticaria every since I think I had the virus. It's now 4 month on and I have to take antihistamine everyday or they appear.

Morning all!

apple.news/A2gWbWkB9T5uw8tiVEOTuSQ

Interesting article that throws a lot of the statistics and conclusions drawn re COVID into question.

@itsgettingweird

Morning all!

apple.news/A2gWbWkB9T5uw8tiVEOTuSQ

Interesting article that throws a lot of the statistics and conclusions drawn re COVID into question.

That very interesting article illustrates how difficult it is to completely eliminate the COVID infection even in a supposedly isolated group.

Either one sailor had an unusually long incubation period of > 14 days (a 1% chance) and was infected in the hotel despite quarantine conditions,
or it was surface transmission from supplies put on board

Also, 57 out of 61 becoming infected, 2 hospitalised.
More understandable, as it would have a much smaller accommodation area than the cruise ships and it just takes one superspreader.

However, it is news because it is an outlier

@BigChocFrenzy

What we don't know is if the virus has become milder - permanently or just over summer - and how much medical advances have improved survival rates

vs
How much is down to elderly people being better protected - incl by public SD
How much to better immune systems in summer
How much to the virus not being as powerful in summer heat
(yes, we know it can propogate in summer, but the death rate seems lower)

It depends what happens once we enter late Autumn & winter

Will death rates remain so low ?

If so, then that is indeed a game-changer that affects the measures we need to keep in place

No evidence that genome has changed significantly to make it milder.

However according to BBC yesterday R rate had started to drop before lockdown.

• Could be behaviourial eg SD
• Could be that most vulnerable to infection without T cell immunity had already got infected. Asymptomatic passing to asymptomatic might result in many more 'under the radar' cases.
• Could be Vitamin D

As you say if death rates remain low that is million dollar question.

"However according to BBC yesterday R rate had started to drop before lockdown."

We know habits had changed considerably for at least a couple of weeks before lockdown - common in many countries:

• Recommendations to wash hands - public health officials and later the PM's Happy Birthday
• Lower footfall in restaurants, gyms, leisure & hospitality generally
• High absences at some places of work
• High absences of school staff & students

It is encouraging news, although expected,
because now we know much better which measures are the most effective,

we have a much better chance of squashing new outbreaks before they spread
and avoiding a full 2nd lockdown in winter

oh and re R falling before lockdown:

I should have mentioned that a number of employers moved to WFH in the 2-3 weeks before lockdown,
before it being officially recommended

I've been getting urticaria every since I think I had the virus. It's now 4 month on and I have to take antihistamine everyday or they appear.
Same with me, but not urticaria - chilblains. In 23C heat I've needed to wear thick socks.

R rate dropped before lockdown EASED ie in May, not in March, according to BBC.

www.bbc.co.uk/news/uk-53414363

Changes the whole meaning!

Though there was a small effect in March as well from increased social distancing etc as well I see (sorry)

The pre-lockdown R dynamics above is based on the famous "Imperial chart" but I am not sure it actually holds true if you look at hospital admissions and death stats.

Hospital admissions in England plateaued on March 29 (3,121) and peaked on March 31 (3,483). If R indeed stayed over 3 up until March 24, we would have continued seeing rapid growth of hospital admissions up until early April. This did not happen. Death stats followed hospital admissions peaking on April 8.

My own amateur interpretation of hospital admission trend is that R fell very close to 1 at around March 16 when Boris introduced soft social distancing measures. People did take them seriously (at least, in London) and acted very quickly. Of course, Imperial College analysts are much more qualified but I am struggling to see how they came up with their estimates.

Actually, if I remember correctly, the Imperial pre-lockdown R chart is relatively old and might have been published before PHE started releasing official stats on daily hospital admissions.

UK did not run any meaningful C19 surveillance testing in March - early April so we will not be able to backtest R modelling results with actual transmission rates. However if the models did not produce accurate estimates in May, it is not unreasonable to assume they might have had similar deficiencies in other time periods.

Full lockdown obviously massively reduces the spread of infections

However, the change in behaviour in the preceding couple of weeks also reduced this

• but nowhere near enough to bring R down sufficiently once infections had already spread so widely;

hence lockdown

Lockdown has reduced infections to a low level,
at which there is a reasonable chance that targeted measures

• now we have more knowledge of what works best - can keep ínfections down

As we've seen in the UK and elsewhere, R>1 is not a problem for a few days when there are only a thousand infections or fewer

but could lead to disaster when there are already a million infected people

Yes, agree with this - and this is why I would have expected to see something like:

• R falling to 1 at around March 16 and continuing to fall until March 24
• R dropping to a very low level (significantly under 1 and possibly under 0.5) post lockdown (March 24)

Imperial chart does not appear to reflect this dynamic which seems intuitive based on hospital admission numbers.

Of course, lockdown was the only option in March given high C19 prevalence but, as we see now, it is not the only tool enabling us to keep R under 1.

"R falling to 1 at around March 16 and continuing to fall until March 24"

Looking at how infections were rocketing, I'd expect:

R falling from 3-5 to maybe 2 with non-mandatory changes in behaviour
then falling to 1 as a result of lockdown

Targeted measures - with the better knowledge we have now - would be much more effective than the early March changes in behaviour,
especially as those were voluntary and hence only some people changed

Estimates of pre-lockdown infections from Oxford Uni & Imperial:

The Times (behind paywall): 22 days of dither

http://archive.is/61c1j#selection-1329.54-1341.153

"back-dated modelling assessing the historic spread of the disease .....
estimates the number of people infected in the UK was indeed doubling every three days during late February and early March,
just as some of the initial reports from China in late January had suggested they might.

The work, produced jointly by an Imperial College London team led by Samir Bhatt and Oxford University, suggests that
on March 3

• the day the government committee was warned about the dire consequences of a mitigation approach -

there were about 14,000 infections in the UK.

Such was the speed of the spread of the virus that 200,000 people were estimated to be infected by the time the government began to change its mind about its policy on Saturday, March 14.

The last nine days while Johnson wrestled over the decision on when and how to go for lockdown were particularly brutal.
By the time the lockdown was announced on Monday, March 23, such large numbers were doubling over such a short period that infections are estimated to have soared to 1.5 million

According to the data, no other large European country allowed infections to sky-rocket to such a high level before finally deciding to go into lockdown."

When the new measures came in on the evening of Monday, March 23, the infections had almost doubled again since the previous Friday and there were an estimated 1.5 million across the UK, according to Imperial and Oxford’s new data.

This its exactly the part of the modelling I am struggling with. The hospital admissions do not seem to support the hypothesis that infections doubled from March 20 to March 23. If this had been the case, you would have expected the hospital admissions to follow a similar trend and double (or increase significantly) from late March to early April. This did not happen. The numbers stayed flat.

I think the shape of the UK infection curve might have been more complex than the traditional exponential graph in the Times article. Our initial trajectory might have been steeper than that in other countries since we were so late to banning mass gatherings (ideal super spreader events). On the other hand, the cases might have actually plateaued earlier than suggested by the model. While the total amount of cases pre-lockdown might have still been 1.5mm (if not higher), the shape of the build up might have been different.

The fact that the UK deaths took so long to fall, compared to our neighbours,
and also,
unlike most other countries in Europe, England has high excess deaths in all regions, not just a few epicentres

all strongly indicate that infections spread widely before the comparatively late Uk lockdown:

"The hospital admissions do not seem to support the hypothesis that infections doubled from March 20 to March 23. If this had been the case, you would have expected the hospital admissions to follow a similar trend and double (or increase significantly) from late March to early April. "

We can't go by the hospitals admissions, because the policy changed:

Unfortunately, until Easter, public health policy in many areas restricted hospital admission to COVID (suspected) cases,
with more severe criteria for admission than for other ilnesses with respiratory distress

Also these estimates of infection numbers, comparing when different countries locked down:

We can't go by the hospitals admissions, because the policy changed:...

But, as far as I can remember, there was no change between late March and early April. Hospital admissions were restricted but as long as the policy had been applied consistently, the numbers must have represented a relatively constant percentage of total infections. If the latter had indeed doubled, the hospital admissions should have followed a similar trend (even with all the rationing).

Even if we dismiss hospital admissions as unreliable data and look at hospital deaths for England, there was a clear slow down in growth between mid March and the first week of April. Between March 13 and March 26 deaths were doubling every 3 days. Post March 27 the growth started to slow down, became more flattish during the first week of April with the eventual peak on April 8. The trend implies the R coefficient was not consistent and had been reducing from a very high number in early March to a level much closer to 1 post March 16.

Also these estimates of infection numbers, comparing when different countries locked down

This is exactly the simplified exponential growth curve that I am questioning.

I am not debating UK locked down late at 1.5mm cases (in fact, the number might have been higher). I think the trajectory to get there had a more complex shape - a very steep growth curve (steeper than in other countries) followed by a flat ('ish) line. It is hard to model and will not come out of a simple regression but if we had done at least some surveillance testing throughout March and had that data, it would have been very useful for understanding the dynamics of C19 transmission in the UK.

Is anyone able to say where we are now?
It is now some time since VE Day and black lives matter protests so should we be seeing an increase in cases/ hospital admissions/ deaths?

New positive tests and deaths UK all settings

Monday 1 June - 1,418 / 111
Monday 8 June - 1,089 / 55
Monday 15 June - 874 / 38
Monday 22 June - 865 / 15
Monday 29 June - 735 / 25
Monday 6 July - 352 / 16
Monday 13 July - 530 / 11

Tuesday 2 June - 1,345 / 326
Tuesday 9 June - 1,059 / 289
Tuesday 16 June - 994 / 236
Tuesday 23 June - 730 / 171
Tuesday 30 June - 569 / 155
Tuesday 7 July - 534 / 155
Tuesday 14 July - 398 / 138

Wednesday 3 June - 1,232/ 365
Wednesday 10 June - 1,087 / 250
Wednesday 17 June - 912 / 184
Wednesday 24 June - 726 / 154
Wednesday 1 July - 605 / 176
Wednesday 8 July - 634 / 126
Wednesday 15 July - 538 / 85

Thursday 4 June - 1,140 / 177
Thursday 11 June - 918 / 152
Thursday 18 June - 945 / 137
Thursday 25 June - 636 / 149
Thursday 2 July - 556 / 89
Thursday 9 July - 649 / 85
Thursday 16 July - 642 / 66