Daily numbers, graphs, analysis thread 19

[BigChocFrenzy]

Daily numbers, graphs, analysis thread 19

Welcome to thread 19 of the daily updates

Resource links:

Welcome to thread 18 of the daily updates

Resource links:

Uk dashboard deaths, cases, hospitals, tests - 4 nations, English regions & LAs
Imperial UK weekly LAs, cases / 100k, table, map, hotspots
Modelling real number of infections February to date
MSAO Map of English cases
Cases Tracker England Local Government
ONS MSAO Map English deaths
CovidMessenger live update by council district in England
Scot gov Daily data
Scotland TravellingTabby LAs, care homes, hospitals, tests, t&t
PH Wales LAs, tests, ONS deaths
NI Dashboard
Zoe Uk data
UK govt pressers Slides & data
ICNRC Intensive Care National Audit & Research reports
NHS t&t England & UK testing Weekly stats
R estimates UK & English regions
PHE Surveillance report infections & watchlists each Thursday
ONS England infection surveillance report each Friday
Datasets for ONS surveillance reports
ONS Roundup deaths, infections & economic reports
ECDC rolling 14-day incidence EEA & UK
Worldometer UK page
Our World in Data test positivity etc, DIY graphs
FT DIY graphs compare deaths, cases, raw / million pop
Covidly.com world summary & graphs
Alama Personal COVID risk assessment

⏭ Our STUDIES Corner ⏮

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

@Bifflepants

False positives are an issue and need to be taken into account when cases are relatively low. It should be relatively easy to manage through repeat testing of positive cases. However, this would depend on whether the false positives was a random event (in which case the chance of two false positives would be minuscule), or whether false positives are caused by factors that would tend to repeat in particular people. If the latter is the case, assessment of the false positive rate would be much harder to ascertain. It would also potentially skew the asymptomatic rate where infection numbers are very low... A significant proportion of positives would be asymptomatic simply because their positive was false! Surely the ONS factors this into their analysis.

@BigChocFrenzy could you have llook at this please www.independent.co.uk/news/health/coronavirus-nhs-testing-hospitals-shortage-b485589.html

@Bifflepants

It's not just her though, she's an example. A friend I have with a science based PhD and good knowledge of stats is arguing the same on SM, and although I know instinctively it's a dangerous argument when positivity is rising, I just don't know how to counter argument effectively.

Just ignore. It’s not backed up by facts or numbers. Just because you have a PhD it doesn’t mean you always think logically.

I think there was stuff on false positives in more or less yesterday.

The false positives thing is an issue particularly because it appeared from the clip from the interview that Julia Hartley did with Matt Hancock that he thought that it was based on the number that tested positive when it is actually based on the number tested. She then asked David Speigelhalter for clarification. He said if you test 1000 people at random and 1 in a 1000 have the virus (which is approx what the ONS say) the FP rate at 0.8% would give 8 false positives and 1 true positive.
He chose 0.8, it could well be higher, because Hancock said below 1% were FPs but he didn‘t know the actual figure.
Speigelhalter then added that if only people with symptoms were being tested this would obviously affect the ratio with less being FPs.

COVID messenger cheered me up this morning. Cases in my town have fallen past 7 days. Rates have fallen from high of 14.2/100k to 9.6/100k.

Let's hope the rule of 6 can help us see these sorts of figures elsewhere in the country.

Thank you for this thread, I have been an avid follower and find it really helpful.

This is purely anecdotal but I think important because it illustrates the potential enormous underestimation of cases in England. The area I live in has had very few cases; according to little owls helpful email service only one or two a day over a large area.

I teach in a primary school. Child A has been off all this week trying to get a test. 5 other children in the class have been in and out with various mild illnesses (colds, stomach upsets etc). They all sit in the same area of the room as child A for most of the day but switch to other classrooms with other children for some lessons. Two others have been trying to get tests but have not been able. Finally yesterday child A confirmed a positive test result. So for 5 days other pupils and staff could have been spreading the virus. All classes and teachers that child A has been with will isolate next week. But not the classes the other ill children have been in unless they eventually get a positive test.

There could be vast numbers of children with the virus with mild symptoms but because testing is so difficult to get no one will know.

I've just had a look at the figures for New York and their cases appear completely flat at under 1000 per day and tiny death numbers.

What are they doing right?

Is it possible that the circa 20% of New Yorkers with antibodies is making that much of a difference? I know that some if the more radical epidemiologists have said 20% might be enough to approach herd immunity.

Or are they still basically in lockdown?

Bet Rigby @bethrigby
Interesting PM said “we’re now seeing a 2nd wave coming in” which was “inevitable”.

Just two weeks back senior govt figure told me the strategy was to use local lockdowns/test & trace to flatten mini waves in order to prevent need for national action. Seems it hasn’t worked

If of interest, I wrote this Sept 1 setting out - to borrow PM’s words - the ‘turbulence ahead’ as PM tries to manage re-opening schools, the virus & the ec fall-out of lockdown. Seems the headwinds have come harder & faster than perhaps govt anticipated
news.sky.com/story/coronavirus-turbulence-ahead-for-boris-johnson-as-he-balances-covid-19-with-job-losses-12061276
Article from 1st.

Sounds like they didn't get their numbers right.

My suspicion is they've over estimated compliance and underestimated covid restriction fatigue.

I genuinely don't know anyone who isn't struggling atm and even those most concerned and most compliant and comfortable are looking at a new wave of restrictions with despair.

Im not entirely sure the government have grasped this level of feeling yet.

Lonely, sadly this is what all teachers feared.

The post code where my own primary is in had only 4 cases last week. I know that two were children.

I'm sure there is some herd immunity in NY. That combined with social distancing and masks might be enough.

I also think this is why the south is not as hard hit right now but they of course may change. Seems odd tho. When we locked down London and south east had huge numbers uncounted as we were only testing hospitals. The north at that point was relatively untouched. Now we are moving and mixing more the north is rising rapidly.

New York just still are not fully open. If the kids are in school at all it’s part time, indoor dinning opens in the main city next week and at 25 percent capacity, it has been open state wide a bit longer but again massively reduced capacity. They only opened their museums about three weeks ago.

Basically it’s very very cautious re-opening compared to Europe and I think that is responsible more than snything.

Testing capacity being overwhelmed also wasn't predicted, despite it starting in Scotland almost as soon as the schools went back there.

If the "circuit breaker" does not in some way restrict school attendance and students moving to university, surely it's not going to "break" much?

Guidance related to "false positives" indicates that low samples should be retested, ideally with a different test type: www.gov.uk/government/publications/sars-cov-2-rna-testing-assurance-of-positive-results-during-periods-of-low-prevalence/assurance-of-sars-cov-2-rna-positive-results-during-periods-of-low-prevalence.
Sage also suggested double-testing with two methods if tests with a higher rate of false positives are to be used: www.gov.uk/government/publications/tfms-consensus-statement-on-mass-testing-27-august-2020.

Some of the debate on Twitter seems to get a little confused:

1. The reason tests are made very sensitive is to catch early stage infectivity, it is inevitable with any testing method that the approach will also pick up late stage low infectivity.
2. late stage low infection rates are someone who has had covid but is no longer infectious, those should not be conflated in figures with true false positives who have not had the virus at all
3. different countries are using different tests and have different testing protocols for RT-PCR; it would be dangerous to assume that false positive rates are identical in each country
4. double-confirmation, particularly using a different type of test, reduces false positive rates drastically, see the sage report linked above; there is retesting with inconclusive swabs
5. What can be done differently? If a method could be totally certain there is no false positive without giving unacceptably high false negatives, that method would be used instead; the tests in use in the UK are the best options right now.

Is it possible that the circa 20% of New Yorkers with antibodies is making that much of a difference? I know that some if the more radical epidemiologists have said 20% might be enough to approach herd immunity

It's probably not 20% in the areas where it spreads the most, infections are not uniform across the population - poor people catch it. As well as that antibody testing is unreliable, and different methods produce different results.

20% in the population is almost certainly not herd immunity though - the Brazilian study we discussed on these threads showed that, the only reason it would be is if there's cross immunity from something else that we don't know about yet and that had spread.

But, any herd immunity reduces R - if 20% of the potential R can't be infected, R is reduced by that 20%, not exactly of course and can be particularly different when combined with other reduction measures.

In most places the social distancing measures different countries have all seem to get R down to somewhere around 1.4, to get it lower you need something more - such as really efficient track and trace in Germany's case. The level of herd immunity to keep the same number of cases bubbling along as roughly the same from that would be around 33% - but could be many more or less if the immunity is not spread evenly.

For me, the levels of immunity does most to explain differences between London and the northern cities.

Regarding herd immunity in London @whenwillthemadnessend; there have been more cases per '00,000 in the areas that initially had high infections e.g. Tower Hamlets, Haringey, Hackney, Newham, Brent, Harrow... and less cases in the boroughs that were hit least badly last time e.g. Sutton, Merton, Richmond, Kingston.
Reasons could be:

1. the same risk factors as before apply, different people are getting infected
2. the virus muddled along at low levels in those communities and came back as people went indoors.

What it can't be reflecting is herd immunity, because that would show the reverse with lots more cases per '00,000 in Sutton and less in Hackney.

London hasn't risen as quickly as the north of England yet; possible reasons include starting from a lower level and more working from home.

Got some more figures from France.

Positivity at 5,4%. 850 new hospitalisations, and 100 to ICU. There’s conflicting numbers for the deaths. LCI are saying it’s 123 in 24 hours, but aren’t saying anything about catch up numbers.

In Île de France (Paris and surrounds) 20% of ICU beds are occupied by Covid cases, that’s in addition to other occupancy. They’re predicting a pinch point on ICU beds in 2-3 weeks (that’s ARS, a bit like PHE).

In Île de France they have 2500 Covid cases in hospital, and 250 in ICU. They’ll be opening 20 new test centres in the region to cope with the high levels.

The average ICU stay is reducing though, although they didn’t say by how much annoyingly. Slightly worryingly they’ve said it’s now much more targeted to the severest cases, which suggests the numbers would be higher if they were using the old admissions criteria. But much better for getting an ICU bed if you need one!

The ARS estimate that in the next fortnight they may have to start making decisions to cancel operations etc.

And the public health people have decided now is the time to release an infographic telling people they may have Covid if they have both a cough and a fever. That’s it. No other symptoms mentioned, just that combo.

ISAGE said schools were 0.5. So it's not really surprising.
NY have kids in masks at school.
My friend's dc is at school FT but it's private.
I expect they are quarantining from EU.
The r just doesnt allow full school. With no proper measures as handwashing at minimum doesnt stop other viruses spreading at school

and less cases in the boroughs that were hit least badly last time e.g. Sutton, Merton, Richmond, Kingston.

Tests were not uniform last time, tests required significant symptoms - something like 40% of all positives were hospitalised let alone managed at home. I don't believe we can trust the stats that said that those boroughs were hit less in actual infections, just in symptoms sufficient to get a test, and obviously in deaths.

The number of people with the significant risk factors in the rich south west boroughs are so much less than the poorer ones.

In New York hospitality is still take away only I think.

@sirfredfredgeorge - We seem to agree that risk factors that mattered in the case of London were how likely someone is to catch it and that relative poverty appears to have had a disproportionate impact, that view is backed up by research in Sage reports. I haven't seen any Sage reports suggesting that herd immunity is reducing spread in London and I have disagreed by presenting facts of known cases.

There was testing on the same basis everywhere; with symptoms only. Actually the biggest risk factor for covid hopitalisation is age. With the notable exception of Harrow, which had a very high death rate, the London boroughs hit hardest the first time have a lower average age than the average for London. The London boroughs with lower infections are on average older, so lack of testing would not be expected to have a disproportionate impact on those boroughs. If there's evidence why you think it would, please can you explain?

I haven't seen any Sage reports suggesting that herd immunity is reducing spread in London and I have disagreed by presenting facts of known cases.

It looks like SAGE spent some time discussing pre-existing immunity in early September. In summary, they seem to believe it may be a factor but they cannot quantify its impact:

assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/918780/S0733_SPI-M-O_Consensus_Statement_2_September_2020.pdf

Some groups will more central to transmission than others, for example health care workers and those mixing more in the community. Thus, the first 6% percent of the population to become immune will reduce transmission by more than 6%. The strength of this effect will depend on how heterogeneous contact patterns are and how consistent these groups are over time (e.g. whether healthcare workers early in the outbreak and later in the outbreak are almost the same group or not), as well as the characteristics of immunity.

Also, here are the minutes of another SAGE meeting spent on discussing pre-existing immunity.

assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/918732/S0739_Fifty-fifth_SAGE_meeting_on_COVID-19.pdf

The unsurprising conclusion is that there are too many unknowns and nothing can be predicted with certainty.

@Bifflepants
That 91% false positive story that is going the rounds - mostly among those grabbing any excuse to demand abandoning all Covid measures, SD, tests, isoltaion etc -

was started by Julia Hartley-Brewer

• who has an ideological agenda, as always -

asking David Spiegelhalter the consequences of a 0.8% FPR (False Positive Rate)
and then misusing the answer for mass testing & ignoring his later post on testing with symptoms

As we've discussed before,
the false positive % has very different consequences for random mass population testing vs testing of people with symptoms or their family members

The claim that therefore the rise in infections is fake also ignores the 30% or so of false negatives

If you want to rebut the 91% false claims, then you could copy both Spiegelhalter's tweets on the topic:

David Spiegelhalter @dspiegel

Tweet below is for mass testing.

Very different if test 1000 people with symptoms
then from PHE data expect 40-50 covid, detect say 30-40.

So most positive tests are correct. The current rise is real.

David Spiegelhalter @dspiegel
Replying to @JuliaHB1

If you test 1000 people at random, latest ONS figures estimate 1 will have the virus, and let’s assume you find them.

But with an FPR of 0.8%, that’s 8/1000, and so you expect to find 8 false positives.

That’s 9 positive tests, only one of which has the virus

Hope this is ok

@wintertravel1980 - that report says they think it will reduce transmission for healthcare workers, nothing about full population herd immunity. The reverse in fact:
'"Current levels of immunity are unlikely to significantly mitigate the impact of any new epidemic wave resulting from increased contact rates." and
"Although seroprevalence is higher in some regions such as London, behaviour and contact patterns also varies between regions. It is very difficult to unpick the difference between behaviour and immunity using ecological studies comparing seroprevalence and incidence. We cannot conclude that pre-existing levels of immunity will prevent cases to rise in London."