> - "false positive less than 1% of those tested": <10 in 10,000
1% of 10,000 is 100 not 10
Mike:
The level of false positives might be gleaned from where the curve of cases/tests bottoms out at the 0.5% line
Doubtful. There was surely some virus circulating, and this may well have flatlined during the summer, as respiratory viruses are apt to do. I'd be more convinced if you could show, region by region that there some minimum below which the apparent percent +ve could never be reduced.
The tests could have had the sensitivity changed a couple of weeks ago and a new false positive rate could have resulted at the 1.4% line
I doubt that there was any such systematic step, least of all conspiratorially. What is clear is that the system is becoming stressed as demand for tests increases, and this may affect who is tested as well as the quality of testing (Labs appear to be running short of staff). I've been trying to find a plot of ages of test subjects over time, but can't. Also I'm unable to find a breakdown of the machines being used.
Mike:
The level of false positives might be gleaned from where the curve of cases/tests bottoms out at the 0.5% line
Doubtful. There was surely some virus circulating, and this may well have flatlined during the summer, as respiratory viruses are apt to do. I'd be more convinced if you could show, region by region that there some minimum below which the apparent percent +ve could never be reduced.
The tests could have had the sensitivity changed a couple of weeks ago and a new false positive rate could have resulted at the 1.4% line
I doubt that there was any such systematic step, least of all conspiratorially. What is clear is that the system is becoming stressed as demand for tests increases, and this may affect who is tested as well as the quality of testing (Labs appear to be running short of staff). I've been trying to find a plot of ages of test subjects over time, but can't. Also I'm unable to find a breakdown of the machines being used.
Thank you David. I really welcome your challenges on this.
I saw figures for the range of false positives of 0.8% to 4.0%, median 2.3% obtained from two published government reports:
https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/895843/S0519_Impact_of_false_positives_and_negatives.pdf
https://gov.wales/sites/default/files/publications/2020-07/core-principles-for-utilisation-of-rt-pcr-tests-for-detection-of-sars-cov-2.pdf
It is of interest how these might relate to cases/tests in the real world. To investigate false positives in regions across the country could be more accurate, but starting from a coarse overall view keeps perspective and is expedient. That is what we need right now.
First of all, getting a false positive level from the 'flat lining' is not an absolute, of course. I would also expect some real virus present, but we might say that the false positives can never be below this 0.5% flat line. Or could we? If the minimum level of false positives according to the reports is 0.8%, how come it is below even that? Maybe there is some very shoddy testing that is always missing real positives - to the tune of the difference, i.e. 0.3% (I shudder to mention this because scare-mongers will pounce on it). All we can do is seek some effective false positive level that mixes regions, occasional true positives, shoddy testing and the rest - to get something to work with.
Secondly, the shift from a flat line of 0.5% to what appears to be the beginning of a plateau at 1.4%, is a cautious speculation from my side. I did read that the sensitivity of these tests could be adjusted. Perhaps others, such as yourself, could say more on this. The point is that there does seem to be a peculiar ramp-up and flattening of the cases/tests curve. It does invite monitoring and explaining. A change in 'effective' false positives may serve until we know a bit more about what is going on.
Your investigation in more detail would be really welcome here but, I feel, thwarted by lack of available data. But good luck anyway.
As a bit of background, I am a retired aircraft stress engineer who developed methods of analyses backed up by tests. The amount of test data is limited so one sweeps up some unquantified variables by incorporating test factors into theory. It is pragmatic. If the analysis method gets close to a range of test results, without over-predicting strength, it becomes usable. This is my analytical approach to the present situation. My motivation comes from being a practising Buddhist for 30 years. Overcoming the suffering, stress and anxiety of all beings is what I feebly aspire to. The current situation is anathema to me.
There are three parts to the false positive story.
One) The first part is the RT-PCR calibration false positive rate, which is what was posted in a previous thread. From the SAGE document. The 1%/3% rate quoted is the absolute best case scenario with controlled samples. The false positive rate from field samples is a very different story.
If you look at the RT-PCR testing protocol published on the CDC site for SARs CoV 2, especially the sample handling procedures, you will see very many parts of the process where test sample failure, contamination, etc can taint the final results. The field sample test failure rates seem to be of the order of 5% to 10% in high pressure mass testing scenarios.
Then there is the interpretation of the final curve output from the RT-PCR machine. Before the SARs 2 Catastrophe the interpretation curves were biased towards negatives because of the innate problems of the whole RT-PCR process. Its a quick and dirty hack. But in the last six months we have seen wildly different positive rates from field samples from the same kind of rapid RT-PCR machines in different US states with pretty much the same kind of populations. New York State gets up to 50% positive whereas neighboring states get 5% to 10% positives. These differences are almost all down to the changing of the negative result bias to positive result bias in the interpretation phase of the process.
Two) Based on the published literature the false positive rates using RT-PCR test varies wildly depending on what part of the infection cycle the corona-virus is in. Unless the test is done during a very small window in the corona-virus infection cycle the RT-PCR test results are almost all false positives.
This paper gives a good summary of just how inaccurate the tests are.
Three) There there is the mathematics.
If you do large scale (10 thousands +) testing of a population with a 1% infection rate (the general HCOV corona-virus infection rate for adults) and the test is 99% accurate the false positives will be equal to the true positives. With a 95% accurate test (closer to real world results) false positives are five times true positives. With a more probably SARs CoV 2 infection rate of 0.3% the numbers are even worse. Best case false positive are 75%, typical case around 95%.
Unless all initial test positives are immediately retested the mathematics is very simple. The vast majority of all test positives are false positives. The retesting at least gets the false positive down to around 5%. Just on the mathematics.
Thanks very much for this, jmc. You are clearly more informed on this than I am. I will look into this in more detail, including the reference you gave. I have also looked at the useful tool on this page:
https://www.hdruk.ac.uk/projects/false-positives/
My initial approach was to avoid getting stuck in the undergrowth of this, hopefully getting an overview of the wood away from the trees, let alone the undergrowth.
If we can just pin a reasonable estimate as soon as possible on how poor this testing regime is, we can show how poor this government regime is and prevent further harm to the health and well-being of the population.
At the end of the day, your suggested 'more typical' false positive rate of 95% (of positive cases) does not look a mile away from my trivial musings.
