Pure data thread #1: Daily numbers, graphs, focused analyses

[BigChocFrenzy]

This is pure data, NOT for the "worried about Corona"

We welcome calm factual, data-driven contributions
Please try to keep discussion focused on these and avoid emotional venting or politics
📈 📉 📊 👍

Resource links

UK:
Uk dashboard R, deaths, cases, hospitals, tests - by postcode, 4 nations, English regions, LAs
Interactive 7-day rolling cases map click on map or by postcode
UK govt pressers Slides & data
SAGE Table Interventions with impacts and R
Imperial UK weekly tables & extrapolations LAs, cases / 100k, table, map, hotspots
School statistics Attendance - Tuesdays
ICNRC Intensive Care National Audit & Research reports
UK testing and NHS England track & trace - Thursdays
ONS Roundup deaths, infections & economic reports
ONS England, Wales & NI Infection surveillance report - Fridays
ONS Datasets for surveillance reports
Our World in Data UK test positivity
R estimates & daily growth UK & English regions - Fridays
Modelling real number of UK infections February in first wave

England:
NHS England Hospital activity
NHS England Daily deaths
PHE COVID Clinical Risk Factors Non-respiratory by region, area, district etc
Cases Tracker England Local Government
PHE surveillance reports Covid, flu, respiratory diseases - Thursdays
CovidMessenger live update by council district in England

Scotland, Wales, NI:
Scot gov Daily data
Scotland TravellingTabby LAs, care homes, hospitals, tests, t&t
PH Wales LAs, tests, ONS deaths
NI Dashboard

COVID-19 Risk Factors
Alama Personal COVID risk assessment
PHE Clinical RFs - summary & social vulnerability indicators
PHE Clinical RFs - respiratory disease
PHE Clinical RFs - non-respiratory - CVD,T1, T2, obesity, flu jab coverage
PHE Non-Clinical RFs - deprivation, demography, economic inactivity, ethnicity
PHE Non-Clinical RFs - Vulnerable Groups (1): care / nursing home, MH, visual disabilities
PHE Non-Clinical RFs - homeless, children in care, ESL

Miscell:
Zoe Uk data
ECDC rolling 14-day incidence EEA & UK
Worldometer UK page
FT DIY graphs compare deaths, cases, raw / million pop
Local Mobility Reports for countries
UK Highstreet Tracker for cities & large towns Footfall, spend index, workers, visitors, economic recovery
NHS Triage Dashboard Pathways - triages of symptoms
NHS Triage Dashboard Progression - # people pillar 1&2, # triages

⏭ Our STUDIES Corner ⏮

Stupid question - I obviously haven't read enough about this - but I presume those 75 symptomatic in the vaccine trial are hopefully among those given the placebo? And then they can assume that the same proportion have been in contact with the virus in the 'really' vaccinated group?

I'm guessing it's more complex actually.....

(Prepares to feel stupid when someone explains it to me)

Can I ask.... on the dashboard - why haven't the daily tests processed figures updated since Thursday? Is it just my browser?

Sorry if asked previously!

@MsWarrensProfession

gidmk.medium.com/most-positive-coronavirus-tests-are-true-positives-60c95fe54fec?

@Augustbreeze

Stupid question - I obviously haven't read enough about this - but I presume those 75 symptomatic in the vaccine trial are hopefully among those given the placebo? And then they can assume that the same proportion have been in contact with the virus in the 'really' vaccinated group?

I'm guessing it's more complex actually.....

(Prepares to feel stupid when someone explains it to me)

Hopefully! But I imagine it may not be as straightforward as that. As I understand it, if the vaccine is more than 50% more effective than the control it may be granted an emergency use licence.

@MsWarrensProfession

I have a request for some evidence please.

On other threads I’m increasingly seeing people claiming that things aren’t nearly as bad as the government are claiming “because false positives”. To my well-informed amateur eyes this is bollocks. False positives would be a problem if we were doing large scale “moonshot” random testing in a population with very low prevalence, but they have virtually no impact on the daily figures we’re currently getting on the news, and in particular no impact on the growth curves.

I do think this needs pushback and challenge wherever we see it, because if it goes unchallenged it looks superficially plausible. I’m struggling to find up to date positivity figures for the UK though, which are key evidence, and because I’m on my phone and quite distracted I’m also struggling to formulate a rebuttal that goes beyond “that’s bollocks, come back when you’ve learned some maths” which would be less than persuasive.

Does anyone have any time on their hands and fancy formatting a well-structured rebuttal please?

Does this help?

www.bbc.co.uk/news/54270373

@MsWarrensProfession

on false positives...They are only really an issue when prevalence is very very low in the population, lower that the failure rate of the test.

The ONS think the prevalence in the English population is about 1/150, or 0.66%.

So a completely random testing of 100,000 people would produce around 667 real positives.

Assuming PCR produces a "false positive" rate of 0.1% of tests, testing 100,000 people would produce 100 "false positives".

So a total of 767 positives would be returned, of which 13% might be false.

However, we are not testing randomly. We are testing people with symptoms, or who have very good reason to believe they are infected. So the number of real positives in that bunch will be much higher than the population as a whole. We processed some 1.9m tests in the week ending 22 Oct; at 0.1%, that is 1,900 false positives; positives were actually 116,644 over the same period.

So 1.6% of results might be false positives.

@Augustbreeze

From the way I understood it, they will wait for 75 events to occur, but obviously they won't know who has had the vaccine.
So if an event is defined as a positive test plus being symptomatic, after 75 people meet that definition they will unblind the data on those 75 people only and then count what percentage of those 75 people had been given the vaccine. Because it would be an interim evaluation, the efficacy percentage for an EUA would need to be higher due to the smaller sample size.

Basically yes Augustbreeze. Crudely speaking, they unblind the trial, and then discover how many cases came from the vaccinated group vs placebo.
If 100% were from placebo then we declare a national holiday and it’s Nobel Prizes and knighthoods all round. If it’s 50/50 or worse then the study is abandoned and everyone moves on to other vaccines with different mechanisms. If it’s 70% from placebo group vs 30% from vaccinated group or similar, i.e. it looks like it’s better than nothing, then the experts get modelling to work out whether the results are sufficiently significant and it’s good enough to be worth rolling out. Scenario 3 is the most likely.

Thanks for suggestions all, I’ll bookmark that More Or Less page to have up my sleeves. I do think it’s quite a dangerous bit of nonsense because it sounds so well informed.

Thanks for the explanations all, it's very interesting.

Re some people getting the vaccine and some getting placebo - that isn't for people who are going to get deliberately infected with covid in the challenge tests is it? It doesn't sound ethical to deliberately infect someone who had the placebo.

Don't think the challenge tests have been given approval yet anyway?

Question from a lurker:

I've been keeping an eye on France for a while. They've declared 52000 cases today; a week ago their 7-day loving average was 21000 (as ours is now) and two weeks ago theirs was 16000 (as ours was a week ago). Any reason to think we won't be long at 50k cases in a week or so???

*moving moving average not loving average. Sigh.

I mean we have different preventative measures in France, they are all about masks and we are all about banning it restricting family gatherings (there is no law against having your parents round your house in France for example but they have to wear masks outside).

Have to say I think we will be where France are in a week yes no evidence either country has found a particular effective set of measures to manage the second wave of the pandemic but technically on a policy front we are managing it slightly differently.

This BMJ article has a covid test calculator that allows you to change parameters re test sensitivity and specificity depending also on likelihood of person having Covid as a percentage e.g symptomatic and in contact with someone positive, symptomatic with Covid symptoms and asymptomatic, not sure how you define these as a percentage, but still visually interesting to play around with as it shows false negatives and false positives when you change the input data.

www.bmj.com/content/369/bmj.m1808

I also started watching this video the other day about PCR testing options and around 16 minutes in there is a discussion about sensitivity and specificity in relation to different reagents used with table showing this at about 16:50 so this might also impact on tests if we have reagent shortages for the PCR testing machines/processes we use in the UK.

update on positivity rates at LA level - top tiers attached, most recent 7 day figures plus that for each of the two preceding weeks. District council areas to follow.

district councils in county order

@SmileDaily - they never have updated test figures at weekends, no idea why, maybe the de-duplication bit takes time so it's one less thing for labs to do at a weekend.

On the "not updating" point, how on earth can Spain go into a state of emergency in a Sunday, but it's still too much trouble to do weekend stats? Don't the Spanish get annoyed?

@Qasd - I think science is more on the UK's side comparing indoor get-togethers with outdoor face masks.
Compliance appears to be increasingly relevant and seemingly untrackable in either country. It irritates me each time I see the "only 10% isolate" stat, when the report included things like a required trip to the doctor / hospital or for a test as breaking isolation. I know that may have only made another say 5% difference, but the raw data wasn't there to confirm and it was just wrong.

What happened in Bromley???

I mean, thanks for the stats @boys3. Any idea how Bromley is 73% positivity?

@MRex

What happened in Bromley???

Surely that must be a typo for 7.3?

Any chance Bromley figure is a typo? Actually 7.3%??

Surely it has to be ?

Yep a fairly spectacular typo. I blame AnyFucker - disconcerted by those place marking .s. As they say though post in haste, repent at leisure. Who needs data quality on a data thread though