Coronavirus

[Simon B]

It does depend which bits of "the media" are being referred to though.

Ok, it's not on deaths and it's in the Daily Wail*, but:

How's this for a microcosm of cretinous journalistic stupidity?

"Reporting" on results from Imperial's latest large-scale antibody sampling data: "Smokers were slightly less likely to have antibodies than non-smokers – at 3 per cent compared with 5 per cent"

So that's much less likely then (at an unstated confidence interval but with a sample space of 100k, that's a prima facie statistically significant result).

One of the many depressing things about CV-19 is how it has laid bare the lamentable standards of basic scientific literacy and capacity for logic and reason among swathes of the UK's population, journalists, politicians, etc.


*Yes, yes, I know, I know but people read this stuff in large numbers and with unknown levels of suspicion. You want a picture of what the forces of reason and evidence are up against, you've got to read this #>$%

[Simon B]

The chart shows the percentage of cases in each 5 year age band, together with the cumulative age profile: https://imgur.com/8K9Saw3

Interesting. Assuming that data is not invalidated by undersized samples, weighting of that with the known age-vs-mortality distribution is going to give a big apparent reduction in CFR/IFR from its previous apparent value, all other things being equal. 10x or something? Don't have time to sit down and work it out...

Certainly looks like there's nothing there to contradict the hypothesis that sustained fall in deaths/admissions simultaneous with increase in cases could largely be down to the age distro of prevalence. If that's right, that trend is likely to not last and instead go nasty after a while.

Presumably scientists whose job it is to work this out carefully are on the case right now - we don't hear much about it though...

[Dave2002]

I’ve lost touch with the data and trends, and as noted it’s now harder to get a “true” picture - or at least that’s my impression.

This article suggests figures of between 22000 and 28000 estimated active cases - https://www.bbc.co.uk/news/health-53781140 It’s hard to relate to previous data, where we were often given numbers in the form of new cases per day. The number of new cases per day at peak was about 6000 IIRC, but of course not all cases were detected or reported. Deaths were easier to measure, even if some may have been ascribed doubtfully to Covid-19 or other causes - but at least a fairly accurate count would have been obtained.

So have things got better, or worse, or are they just the same - since June at any rate? Deaths have gone down, that seems clear. Maybe there are many more mild cases being picked up. What is a reasonable length of time to assume the virus is active in a person? 28 days would suggest an incidence rate of around 1000/day. At the peak of the pandemic in the UK the reported daily rate of new cases was about 6000, while near the end of June it was lower than 2000, but still more than 1000. Rates were seemingly indicating a downward trend, but seemed to be levelling off, or indeed starting to rise again at the time when regular data updates were suspended.

[johnb]

Dave,

Those figures are taken from today's Infection Survey report from the ONS, where it states:

During the most recent week of the study, we estimate that 28,300 people in England had the coronavirus 1(COVID-19) (95% credible interval: 19,000 to 40,700). This equates to 0.05% (95% credible interval: 0.03% to 0.07%) of the population in England or around 1 in 1,900 people (95% credible interval: 1 in 2,900 to 1 in 1,300). This is based on statistical modelling of the trend in throat and nose swab results.
...
...
The modelled estimates for the latest six-week period are based on 122,021 swab tests collected over this period. During these weeks, 58 individuals from 58 households tested positive.

So we are dealing with a pretty small sample of people testing positive, hence the wide credible intervals - which the BBC, in common with most media, fails to mention.

The report also states, for England:

Based on exploratory modelling (3 to 9 August), we estimate that there were 0.69 new infections per 10,000 people per day (95% credible interval: 0.42 to 1.08). This equates to 3,800 new infections per day (95% credible interval: 2,300 to 5,900).

Our modelling suggests that the incidence of new cases of the coronavirus (COVID-19) has increased following our lowest estimate, which occurred at the end of June 2020, but there is evidence that this trend appears to have now levelled off.

But once again, as far as I can see, the number of positive samples this based on is not that large, note the "credible interval".

So it's a useful indication but not something to bet your life savings on.

Link to the report: https://www.ons.gov.uk/peoplepopulat...es14august2020

[Eine Alpensinfonie]

So we can sing/play woodwind and brass instruments now, because (as many of us already knew) they're no more likely to spread Covid-19 than speaking.

[johnb]

So we can sing/play woodwind and brass instruments now, because (as many of us already knew) they're no more likely to spread Covid-19 than speaking.

Well brass instruments are glorified water vapour condenser coils.

[johnb]

Following on from my reply to Dave, this might (or might not) be a guide to whether the number of infections are growing or declining. It is based on the information in the latest NHS Track & Trace report:

[johnb]

The topic as a whole reminds me of this ad:

[Rolmill]

So we can sing/play woodwind and brass instruments now, because (as many of us already knew) they're no more likely to spread Covid-19 than speaking.

I haven't seen this reported - is there a source for this please?

[Simon B]

Terminological inexactitude is a worsening and befuddling difficulty.

In particular, continuing imprecision in the use of "cases" versus "infections" is not helping. The above cited BBC article is a case (pun not intended) in point since it's solely concerned with attempts to estimate the infection rate.

Given the absolutely critical conceptual and numerical distinction between IFR and CFR it would help if journalists consistently used these terms in their correct technical sense rather than the vernacular for the time being. This isn't just pedantry for the sake of it - there still seem to be numerous folk out there (not here) conflating CFR with IFR to make the improbable claim that CV-19 has a 4-5% mortality rate.

At the peak of the pandemic in the UK the reported daily rate of new cases was about 6000, while near the end of June it was lower than 2000, but still more than 1000.

Isn't this an indirect way to re-demonstrate that only a small proportion of cases were detected or reported at the peak? The peak CFR was ~20% where now it looks more like ~2%. Meanwhile, a plausible estimate for the overall IFR is <= 1%, so there would have been a peak of ~100,000 infections/day at some point.

The ONS data cited gives 58/122k +ve swabs over 42 days. Extrapolating that to a population of 68M gives 758/day, assuming nothing has changed over the last 42 days. The oddity with that is that actual confirmed cases exceed that number by some margin in recent weeks.

If we assume instead that the rate has doubled over the last 6 weeks that gives 505/day at the start and 1011/day now. In a remarkable and meaningless coincidence this tallies closely with the actual confirmed cases. To that extent, something still has to be a bit fishy as there's no way we are finding and testing 100% of infections and declaring them as cases.

Clearly there's something much more complex in the way the ONS extrapolates an estimate of overall prevalence from their 58 positives from 122,000 samples over 6 weeks as they get 2,300 to 5,900/day at the 95% confidence interval from a mean of 758/day in their sample. If I had time to read the report it might be clear how/why...

[Simon B]

I haven't seen this reported - is there a source for this please?

There's this for brass and singing:

https://www.bbe.org.uk/news/04082020-1628/brass-bands-england-releases-laboratory-research-results

Doesn't include woodwind though. WW seems a much different/bigger issue as air is blown through or over them to produce sound unlike brass/singing.

[Rolmill]

There's this for brass and singing:

https://www.bbe.org.uk/news/04082020-1628/brass-bands-england-releases-laboratory-research-results

Doesn't include woodwind though. WW seems a much different/bigger issue as air is blown through or over them to produce sound unlike brass/singing.

Thanks Simon. It's singing that mainly interests me, it will be interesting to see if the research specifically into singing (e.g. by Declan Costello) produces similar results to this.

[johnb]

I'm not sure whether anyone will find it of interest but I've updated the Excel file that has a scatter chart of local authority case rates vs change in numbers - where you can scroll through the dates using a scroll bar. It goes up to 8th August (I'll bring it a bit more up to date soon).

I've moved the controls to the side of the chart (after having tried the file on a laptop).

It will only work properly in Excel - and the macros need to be allowed.

https://app.box.com/s/7ugl5c8iecft9o1cy0uf2x7nnyuqq222

[Dave2002]

Thanks to John and Simon for reminding us of some of the technical details. Although the BBC link may have had some terminological inexactitude, and seemed to present numbers without indicating the range limits, one interesting thing it did provide was a postcode/health area link to estimted infection probabilities. Thus in my area this is estimated at around 1 per 100,000 - so I'm not very likely to meet anyone who has the disease, and even if I take no precautions (which is illegal if I go out ...) the probablilty of my catching it, or even worse (for me ...) dying of it, are pretty low. It seems that the average in most of England is now abou 6 in 100,000 but there are still some spikes in hotspots which seem to emerge each week.

The recent chart which John provided showing cases broken down into areas was helpful, and I hope we can continue to have that kind of presentation, either done by our local number crunchers (thank you again, John) or provided by other credible official sources.

I don't know where the current hot spots are this week, though some newspapers are always looking for a sensational story - but that doesn't mean that they do have reliable knowledge and information. Places like Dumfries are reported to be having a spike, while today Oldham seems on the verge of another lockdown.

[Simon B]

... estimted infection probabilities. Thus in my area this is estimated at around 1 per 100,000 - so I'm not very likely to meet anyone who has the disease, and even if I take no precautions (which is illegal if I go out ...) the probablilty of my catching it, or even worse (for me ...) dying of it, are pretty low...

At risk of being tediously repetitive, that data panel on the BBC article is giving case rates, not estimated infection rates.

My recent local area case rate is given as 4/100,000. Meanwhile, the ONS estimate of active infection rate in a (much wider) area including my area is ~1/2000 or 50/100,000. So, about 10x more. This tallies with the rule of thumb I've been using - earlier in the outbreak I'd have gone with 20x or more.

So, in my non-expert reading, you're about 10x as likely to encounter a person with an active infection as the 1/100,000 figure given.