>47% false positive rate on COVID-19 test?

  • #1
BWV
592
495

Summary:

>47% false positive rate on COVID-19 test?

Main Question or Discussion Point

interesting:

https://www.ncbi.nlm.nih.gov/pubmed/32133832?fbclid=IwAR3KtTRFTp9eb0aA-8ThVONEMGeGhecOnjMJtIblbClp8ufahx71HOFSWdg

Objective: As the prevention and control of COVID-19continues to advance, the active nucleic acid test screening in the close contacts of the patients has been carrying out in many parts of China. However, the false-positive rate of positive results in the screening has not been reported up to now. But to clearify the false-positive rate during screening is important in COVID-19 control and prevention. Methods:Point values and reasonable ranges of the indicators which impact the false-positive rate of positive results were estimated based on the information available to us at present. The false-positive rate of positive results in the active screening was deduced, and univariate and multivariate-probabilistic sensitivity analyses were performed to understand the robustness of the findings. Results: When the infection rate of the close contacts and the sensitivity and specificity of reported results were taken as the point estimates, the positive predictive value of the active screening was only 19.67%, in contrast, the false-positive rate of positive results was 80.33%. The multivariate-probabilistic sensitivity analysis results supported the base-case findings, with a 75% probability for the false-positive rate of positive results over 47%. Conclusions: In the close contacts of COVID-19 patients, nearly half or even more of the 'asymptomatic infected individuals' reported in the active nucleic acid test screening might be false positives.


child that tested positive in LA now believed to be a false positive:

https://losangeles.cbslocal.com/2020/03/16/coronavirus-child-testing-negative-false-positive/
 

Answers and Replies

  • #2
atyy
Science Advisor
13,727
1,852
The Westerdam passenger who initially tested positive after disembarkation later tested negative. Either a quick recovery or a false positive.
https://www.ktoo.org/2020/03/11/city-working-through-what-ifs-as-the-westerdam-and-cruise-ship-season-makes-its-way-toward-juneau/

There are also reports that the problem with some of the first CDC test kits that led to the shortage of tests in the USA were due to false positive problems
https://www.propublica.org/article/the-fda-is-forcing-the-cdc-to-waste-time-double-testing-some-coronavirus-cases

There are some labs that have done a full sequence in addition to the PCR test. If you do a full sequence, you can essentially rule out false positives. This article mentions two labs that do full sequences in addition to the PCR, but my guess is that this is not routinely done, and the labs do the full sequence for other reasons.
https://www.the-scientist.com/news-opinion/how-sars-cov-2-tests-work-and-whats-next-in-covid-19-diagnostics-67210

- "Yvonne Doyle, the medical director and the director of health protection for Public Health England, tells The Scientist in an email that once a sample is received by a laboratory, it takes 24–48 hours to get a result. Commenting on the test’s accuracy, she says all the positive results to date in the United Kingdom, a total of 36 so far, have been confirmed with whole genome sequencing of the virus isolated from patient samples, and “the analytical sensitivity of the tests in use is very high.”"

- "The laboratory at Westmead Hospital also does a complete sequencing of every virus sample to look for possible new strains of SARS-CoV-2 and has shared some of those sequences in the international Global Initiative on Sharing All Influenza Data (GISAID) database for other researchers to study. The staff also cultures the virus and images it using electron microscopy. “That’s not really a diagnostic test, but gives you some confirmation of what you’re seeing in the laboratory,” Dwyer says.

He adds that, so far, there’s no suggestion of false positive findings, because every positive test has been confirmed with whole genome sequencing, viral culture, or electron microscopy. As for false negatives, he adds, it would be hard to know if any infected patients were mistakenly given the all-clear."
 
  • #3
Vanadium 50
Staff Emeritus
Science Advisor
Education Advisor
2019 Award
24,015
6,567
Bayes theorem says that you will get a high false positive rate whenever the actual incidence is very low.
 
  • #4
atyy
Science Advisor
13,727
1,852
BTW, the link to the full article didn't work for me, but if I understand the abstract correctly, the 47% false positive is only for screening asymptomatic close contacts. It is probably related to the calculation @Vanadium 50 did in https://www.physicsforums.com/threads/wuhan-coronavirus-containment-efforts.983707/post-6310535 (post #994)

This may be one of the reasons that (as I understand it) there is presently no recommendation to test asymptomatic people.

The other reason for not testing asymptomatic people is a kind of "false negative", in which a person has low viral load early in the incubation period, so the test is in some sense correctly negative. In another sense, of course, it is false negative since the person has been infected, so it may give people who have been infected a false sense of security such that they don't get tested when the viral load has increased after the incubation period and they have symptoms and would test positive.
 
Last edited:
  • #5
BWV
592
495
i cant get the full link to work either and if I could, it is in Chinese
 
  • #6
1,113
562
i cant get the full link to work either and if I could, it is in Chinese
What, you can't read and understand Chinese? Well, ok. me neither. 😌
 
  • Like
Reactions: BWV
  • #7
13
9
  • #8
727
168
Don't forget that you have to factor in the prevalence of the disease in the population, if known, otherwise you get a very pessimistic answer.

Cheers
 
  • #9
Tom.G
Science Advisor
3,053
1,805
As another non-medical person, what I get from the Abstract is that:
Of those exposed to an an infected person, somewhere between 47% and 80% picked up enough of the virus to be detected but did not get sick.

Sounds encouraging to me!
 
  • #10
atyy
Science Advisor
13,727
1,852
I do not have a hard science background. Truth , I pretend I know stuff.

However,

If I understand this correctly - If you were around a sick person - in 47% of the cases - if tested - it would say you have the virus when you do not.

https://pubmed.ncbi.nlm.nih.gov/32133832/?fbclid=IwAR3LHLplRtmPZED3jt_Hjf2ZKysmDcXMU5LgVgNPcqNAtb9EzcQBKMtCIoI

This is mind blowing. If true - everyone should know about this.
Yes, that is what I understood too. @Vanadium 50 has pointed out that this may simply be the well understood phenomenon that if a condition is rare, then the test must be very specific to avoid a useless or misleading results.
 
  • #11
atyy
Science Advisor
13,727
1,852
Veneto seems to be using tests for screening. Is it unknowingly picking out lots of false positives? Is there a danger to false positives? The next time they get symptoms, they may think it is not Covid-19, since they had it already and should have antibodies against it. So if they then assume it is just a flu they may let their guard down.
https://www.theguardian.com/world/2020/mar/18/scientists-say-mass-tests-in-italian-town-have-halted-covid-19
https://www.theguardian.com/commentisfree/2020/mar/20/eradicated-coronavirus-mass-testing-covid-19-italy-vo
https://www.voanews.com/science-health/coronavirus-outbreak/italys-veneto-region-test-everyone-coronavirus

A company in Iceland is doing something similar, but 0.86% of random samples seems very high (about the same as Wuhan!), so it is either massively flawed or the samples are not representative of the general population (maybe they only tested the symptomatic subset)
https://cleantechnica.com/2020/03/21/iceland-is-doing-science-50-of-people-with-covid-19-not-showing-symptoms-50-have-very-moderate-cold-symptoms/

For comparison, I expect that if one tests symptomatics in a place that is not an epicenter, the positive tests should be about 1 in 200 to 1 in 1000, going by the example of the fever clinics in Guangdong.
https://www.who.int/docs/default-source/coronaviruse/who-china-joint-mission-on-covid-19-final-report.pdf
"Within the fever clinicsin Guangdong, the percentageof samples that tested positive for the COVID-19 virus has decreased over time froma peak of 0.47%positive on 30 January to 0.02% on 16 February. Overall in Guangdong, 0.14% of approximately 320,000 fever clinic screenings were positive for COVID-19."

Incidentally, while South Korea does test a lot, from what I've read they do not test the general population. At the drive-through centers, one fills in a form first, and one is tested only if one is eligible (presumably only if one has symptoms or is at risk perhaps as a close contact), but I'm not sure exactly what criteria they use to determine eligibility.
 
Last edited:
  • #12
358
108
So, given that false-positives would be included in the infection statistics, did the effective mortality rate just double ?
 
  • #13
FactChecker
Science Advisor
Gold Member
5,380
1,950
So, given that false-positives would be included in the infection statistics, did the effective mortality rate just double ?
With a high rate of false positives, I doubt that they would rely on only one such test.
 
  • #14
358
108
With a high rate of false positives, I doubt that they would rely on only one such test.
I just read about the f.p's today, but they've been giving stats for quite a while.
 
  • #15
FactChecker
Science Advisor
Gold Member
5,380
1,950
I just read about the f.p's today, but they've been giving stats for quite a while.
I mean that for an individual case that gives a positive result, I think they would give that person a repeat test or two. The resulting false-positive rate of multiple tests may be fine.
 

Related Threads on >47% false positive rate on COVID-19 test?

  • Last Post
2
Replies
47
Views
4K
Replies
18
Views
356
Replies
7
Views
178
Replies
86
Views
2K
  • Last Post
Replies
9
Views
176
Replies
4
Views
2K
  • Last Post
Replies
2
Views
2K
Replies
3
Views
3K
  • Last Post
2
Replies
30
Views
5K
  • Last Post
Replies
2
Views
2K
Top