To: Scientists, Doctors and anyone writing OpEds or Making Policy Decisions about tests


The number of False Positive tests does *NOT* go up when prevalence of #COVID19 goes down.

The number of False Positives is the SAME regardless of high vs low prevalence.

This idea keeps getting thrown around in OpEds and by decision makers - worried false positives sky rocket once disease drops.

But if you have 10 False Pos per 100,000 tests when incidence is high, you’ll have ~10 False Positives per 100,000 tests when incidence is low.

What changes is not the number of false pos results, but the fraction of all pos that may be false

Of course even if false pos rate is 0.0001 (about that of newer rapid Ag tests), if there is zero Covid, then sure, 100% of positives will be false. But it’s still only 1 in 10,000
So you know at the outset the burden of false pos you may face based on the test.

And if 1 in 1,000 or 1 in 10,000 is too many, then simple solution - confirm w another rapid test! If the first was False Pos, you’ll know in ten minutes.

A small burden compared to the benefits
And just to be clear - i'm talking about community testing for public health in this pandemic

Of course if prevalence of disease is 100%, there would be no false positives. But that's not COVID... Prevalence in population is usually <<1%

• • •

Missing some Tweet in this thread? You can try to force a refresh

Keep Current with Michael Mina

Michael Mina Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!


Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @michaelmina_lab

29 Apr
UK ppl should be concerned that head of @RoyalStatSoc #COVID19 diagnostics grp is inaccurately stating facts about testing-Intentionally conflating false positives w positive predictive value (totally different) to make rapid tests look bad #badscience
In the tweet above, to get false positive rate you need to have the number of true negatives in the denominator. If you fail to do that, you aren't measuring the false positive rate. It's a totally different number. Studies show false positive rates for many tests <0.1%.

In the first tweet I said "UK ppl should be concerned"... but c'mon @RoyalStatSoc - why allow this to continue. He's been stating falsehoods and bending stats or performing flagrantly unacceptable/unpublishable analyses for months now. Sowing confusion. How is this helpful?
Read 5 tweets
23 Apr
Thread: On costs $$ of at-home Rapid Tests

The US government earmarked 10's of billions of dollars for testing. This was a great move!

But why isn't this money being used to make frequent rapid testing more available to the consumer?

Testing is a public health good.

The government should follow the same successful playbook for frequent testing as it has for vaccines: pay for the tests themselves (as it did for the vaccine) or buy down the price at retail (as it does for the administration for the vaccine itself). 

Why is this important?

1. Children still aren't vaccinated
2. Healthcare workers and others who are vaccinated can still contract the virus (and may be able to pass it on)
3. Less than 50% of US adult population is fully vaccinated

Testing is our eyes to see this virus

Read 5 tweets
22 Apr
Following recovery from COVID19, people remain PCR positive for weeks. These ppl do not transmit virus and needn’t isolate **even if only first tested w PCR after recovery

PCR is not specific for the transmission period nor for requiring isolation

This simple fact - that lab based PCR is not specific to what matters most in a pandemic - whether someone needs to isolate - has been entirely missed in this pandemic and unfortunately at the expense of tests that are fast, accessible AND specific to the infectious period

I hope regulatory / public health agencies understand that public health is not the same as medicine, and the tests needed for public health are different and must meet entirely different metrics to be effective.

@K_G_Andersen and I cover this here.

Read 4 tweets
19 Apr
Articles like this continue to miss that rapid tests can be used multiple times in a row and paired together for rapid confirmation - immediately.

Bayes theorem is great, but let’s update our thinking around tests when we invoke it.…
I want to clarify this 👆is a very nice and well written article by @TomChivers - pertaining to conventional one-off testing

Frequent accessible testing is however different & repeated use of one or more tests must be factored in to sens/spec calculations & Bayes probabilities
Since false pos are of concern bc they mean ppl isolate when they do not need to, then w that definition @TomChivers it would be interesting if you follow up w a piece on the VERY high false pos PCR rate bc many/most PCR positives are detected AFTER the isolation window passed
Read 4 tweets
16 Apr
This is why we must offer confirmatory rapid tests along w any rapid test program

If we give ppl 20 of test A, we need to give 1-2 of test B - a confirmatory test to use immediately if a pos shows up on A. Can be Ag or Molecular, at home test.

Many rapid tests have False positive rates of 1 in 1000 or less. This is great. But when using huge numbers to control spread, it’s better to have that number at 1 in 10,000 or lower. A simple test A +B if A is positive algorithm would vastly improve specificity.

If we fail to consider a simple addition of a relatively small number of accessible, rapid confirmatory tests, the population may lose confidence. Like all the testing problems, there are simple solutions. We should run with them. Rapid confirmatory tests is one of those.

Read 4 tweets
9 Apr
Still so much confusion about rapid antigen tests. They get compared to qPCR and bc qPCR is deemed the gold standard, they appear as “low sensitivity”

To be clear, the problem is not the Ag test, it is the v low specificity of qPCR to delineate who needs to be isolated still

qPCR is highly specific as a medical test to appropriately identify SARS-CoV2 RNA....

but qPCR is terribly NOT specific as a public health test to determine who should still be isolating (though looking at Ct values can help)

This is not good for public health.

We never should have made qOCR the gold standard for evaluating public health tests. It was a mistake from the beginning bc as CDC readily says - you stay positive on a qPCR test for weeks after you are done transmitting. The specificity is terrible as a public health tool

Read 7 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!