This paper used by Sage suggested 400 of 1600 positives were false. Still happy to have your school child home under house arrest for two weeks?

[Treesofwood]

Big problem with false positive us unnecessary self isolation for the person rested. Now it can lead to 300 children also being asked to self isolate for 2 weeks. This paper suggests when numbers are low, which they are at the moment, falso positives are a big problem. The more we test, the more we find them. www.gov.uk/government/publications/gos-impact-of-false-positives-and-negatives-3-june-2020
And the more children miss education and even being allowed to breathe fresh air if other threads on here are something to go by.
Obviously this is not a peer reviewed paper, but it is of enough interest and reliability to be used by Sage. Interested to hear thoughts, including if I have actually totally misinterpreted it.

With a 0.4% false positive rate you’d get 40 positive cases, 100% of which were false positive

I suspect the false positive rate is much lower.

I am basing this on the ONS survey.

189,000 tests carried out - but only 89 positive people.

That suggests a very very low false positive rate.

This was also carried out on a random sample of the population

Yes, the ONS data is reassuring because it implies that the true false positive rate is very low and that 0.5% is a theoretical upper bound.

@chomalungma

False positives are an issue

Did you see the ONS figures?

The modelled estimates for the latest six-week period are based on 183,994 swab tests collected over this period. During these weeks, 89 people from 84 households tested positive

So not many people testing positive - relative to a large number of tests in a random sample of the population

This. And it doesn't really fit with the daily testing figures either. It did get down to about 400-500 cases per day while doing over 150K tests. 0.4% is the highest figure I've seen for a false +ve rate. Mostly it's 0.1% but I think there was a paper the other day that put it around 0.01% although it probably depends on which test. Although even 0.01% is going to give big numbers of false positives in the (unlikely) event we're testing 10 million people a day.

Also my understanding was that while contamination probably is an issue (when isn't it?), part of the issue of false positives is the test being sensitive enough to pick up very low levels of virus in people that probably aren't transmitting it at that time. Which in some part will be either very early or late testing. So it isn't like there isn't actually some benefit to the isolation of cases/contacts in that case.

It's a subset of 'false positives' but doesn't necessarily mean that the virus has never been there in those cases.

Although even 0.01% is going to give big numbers of false positives in the (unlikely) event we're testing 10 million people a day

I think that if we did that, then we'd be using lateral flow like tests which are more likely to give both false positives and false negatives - especially if they are used 'in the field' by less experienced people.

@phlebasconsidered

Do you believe the op started this as a discussion when she had that in the title? Really? Some discussion has been reasoned. The title is not - and neither was the op, which is what I was addressing.

I have no idea why she started the thread, but I agree the wording is inflammatory. As is yours. And misleading, as no one on this thread that I can see has ‘denied’ Covid. So just as the OP is using inflammatory wording, so are you.

I wonder how they calculate a ‘false positive’ vs asymptomatic carrier. Do they do another test? What we can do is recruit people in labs so they can run over the weekend and expand testing capacity

Phlebas I chose the term house arrest partly to catch people's attention. But also because generally people are being berated for suggesting it might be safe to walk alone at 6am with a child who is self-isolating and who needs outside time. They are treating it like it should be house arrest.

@chomalungma

Although even 0.01% is going to give big numbers of false positives in the (unlikely) event we're testing 10 million people a day

I think that if we did that, then we'd be using lateral flow like tests which are more likely to give both false positives and false negatives - especially if they are used 'in the field' by less experienced people.

Exactly. As a different test it would have a different false +ve or -ve rate, which would have to be recalculated. It might have some benefit as a screening test before large gatherings and getting people to get a follow up test, but I'm not sure that would outweigh the risks to confidence in the testing system.

@Treesofwood

I notice you've not commented on the ONS figures.

189,000 random tests carried out
89 positive people

That has got to be a very very very low false positive rate. Much lower than 0.4%

That’s the problem Namenic. In an ideal world you’d line up a million people and test them all using your test and then retest them all using an alternative method with no false negatives at all and then you’d know for certain. In the real world it’s much trickier and that’s why we’re seeing numbers from 0.01% to 2.3% flying around on this thread. The only thing you can know for absolute certain is that it’s no higher than the total proportion of positive tests on a large random sample in a population with low incidence, but as chomalungma points out, not all tests are the same.

Chomalungma I haven't had chance to look properly at that, but yes that is much lower than 0.4%, which is obviously a good thing.

Coronavirus: Tests 'could be picking up dead virus'

I read this a while ago there were suggests where perhaps the test needed a cut off point rather than a straight yes no.

It does worry me though - two schools here already sent year groups home and I have a Y11 child' - and unlike apparently most schools on MN her school really was dire with home learning. There already been massive disruptions to her GCSE learning I'm really not keen for more.

@LearnedResponse

That’s the problem Namenic. In an ideal world you’d line up a million people and test them all using your test and then retest them all using an alternative method with no false negatives at all and then you’d know for certain. In the real world it’s much trickier and that’s why we’re seeing numbers from 0.01% to 2.3% flying around on this thread. The only thing you can know for absolute certain is that it’s no higher than the total proportion of positive tests on a large random sample in a population with low incidence, but as chomalungma points out, not all tests are the same.

I can't find the paper I'm looking for and I didn't bookmark it, so I'll retract the 0.01%.

Looking at the ONS data choma has posted would give 0.05% IF all of the positives are false positives though. Which would at least suggest that using 0.4% leads to an overestimate. I think it's likely that with more testing data available the authors of that paper would perhaps use a lower level now.

@LadyOfTheImprovisedBath

Coronavirus: Tests 'could be picking up dead virus'

I read this a while ago there were suggests where perhaps the test needed a cut off point rather than a straight yes no.

It does worry me though - two schools here already sent year groups home and I have a Y11 child' - and unlike apparently most schools on MN her school really was dire with home learning. There already been massive disruptions to her GCSE learning I'm really not keen for more.

That's largely what I mean about some of the false positives we are talking about being a specific subset of false positives, not just contamination from somewhere.

But these would be people with a recent COVID infection who are probably not still contagious although they might have been ;last week. How much of an issue this is probably depends on how much asymptomatic testing is going on. This is probably solved by retesting asymptomatic +ve cases, but I doubt we have the capacity for that.

Tbh I would be more worried about false negative.

Given dh shielded and my kids and I severely restricted both contact with him and others for months.

Also still don't get the figures quoted especially as it doesn't tally at all with ONS figures.

I confess, the maths is doing my head in slightly. But as I understand, the ONS 89 positives in 180,000 tests works out at about 50 in 100,000. But if 100,000 of an entirely covid free population were tested, with the "best case" 0.01% false positives, then 10 people (I think) would show up as positive. So as a best case out of the 50, only 40 have covid and 10 don't? That seems like a high proportion to me (20%), and multiplies up quickly if all the contacts of those 10, (and their households if in Glasgow, for example), have to self isolate?

I know the testing system is under strain, but surely a second test priority for all positive cases should be important to avoid that?

I know the testing system is under strain, but surely a second test priority for all positive cases should be important to avoid that

I think there is a Gold standard test to see if someone is positive - you have to take a sample and culture the virus.

My speciality was Clinical Biochemistry with some experience of RT-PCR. Virology was not something I got too involved in - and it was also a long long time ago.

(Part of me wish I stayed - but another career beckoned)

I don’t understand the argument that we cant retest positives because of capacity issues.

I would challenge that.

If 2% of all tests are coming back as positive, then to retest to weed out the false positives, we only have to increase testing by 2%. We don’t have to test everyone twice - just the positives.

I get that there are capacity issues just now, but given that we have to sort that out anyway, increasing by a further 2% above and beyond that doesn’t seem too much of an ask.

Or am I missing something?

Thanks chomalungma ...so theoretically there could be a project where swab tests are run alongside these "culture" tests, to get an idea of true specificity/ sensitivity of the swab tests? I'm sure this was done at some point when the swab tests were developed so they must be acceptably accurate....and they may be getting better over time? But from what I can work out, the problems seem much more serious when rates are relatively low in the population than earlier on in the summer when hardly anyone could get tested unless they were in hospital...but rates were higher. 10 or 20 false positives in 100000 is "worse" if there are only 30 positive tests in 100000 in total, than if there are 3000 positive tests in total. Though it's just as bad for the 10 or 20 having to self isolate...

I did a basic first year uni microbiology course, before specialising in something as far from microbiology as you can get. But I agree, virology got pretty weird pretty quickly. And that was 25 years ago.

thelaststarfighter yes, that's how I understand it too. I'm not sure why it's been written off as impossible so quickly! It seems pretty vital to me. Morally, if nothing else. If we as a society are asking people to put their lives on hold to protect others from a virus they, or someone in their household, have (even if it's not making you feel ill enough to need to do that anyway) I'd say it's absolutely vital to be sure that they DO actually have the bloody thing in the first place!

@TheLastStarfighter

I don’t understand the argument that we cant retest positives because of capacity issues.

I would challenge that.

If 2% of all tests are coming back as positive, then to retest to weed out the false positives, we only have to increase testing by 2%. We don’t have to test everyone twice - just the positives.

I get that there are capacity issues just now, but given that we have to sort that out anyway, increasing by a further 2% above and beyond that doesn’t seem too much of an ask.

Or am I missing something?

I see your point. Mind you there has also been a false negative rate (estimated at between 2-15% when I looked on ONS just now So if you end up with 1 false positive and 1 false negative, do you test that person again?

Just to be clear I’m not saying we shouldn’t have capacity to test, or that we shouldn’t make more capacity. And I’m definitely not writing it off as impossible. It isn’t.

It’s just that right now, with the testing system being close to falling over allegedly due to lab capacity issues, people not being able to get tests, results taking more than 24hrs and cases rising I’m not sure that retesting 3000+ positive cases a day just to find maybe 50 false +ves is the best use of lab time.

Making that capacity should be a priority.

If 2% of all tests are coming back as positive, then to retest to weed out the false positives, we only have to increase testing by 2%. We don’t have to test everyone twice - just the positives.

I wrongly suggested earlier that it was impractical but, yes, you are absolutely right. With low levels of positive tests, I don’t see why this is a problem. If we start getting into double figure percentages we may have problems though... but then, we would be in an altogether more concerning phase of the pandemic.

@chomalungma Am I right in thinking they have a relatively good rate of false positives as they run the test three times? Do they use the same CT level also? I know we confirm a positive at different levels, Taiwan for example have a lower threshold, so less likely to get false positives than us.

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 This seems to imply that labs choose their own threshold? Why would that be? Also recommending repeat test and no isolating until it is carried out. Is, this happening? And at what CT level?

Am I right in thinking they have a relatively good rate of false positives as they run the test three times? Do they use the same CT level also

Honest answer - I have no idea. It's moved on a lot since I worked in the labs (although I have applied for Lighthouse Lab work) and the technology has obviously vastly improved.