Been thinking about this in light of some of the MLB players testing positive and then immediately testing negative on a subsequent test.

Anyone happen to have good sensitivity data on the RT-PCR tests in the US?

In the UK it seems they don't (didn't?) know.

https://www.gov.uk/government/publications/gos-impact-of-false-positives-and-negatives-3-june-2020

Running a million tests a day, if the test is 99% sensitive, the US will still produce 10k new detected infections.

This preprint from May https://www.medrxiv.org/content/10.1101/2020.04.26.20080911v2.full.pdf reported a 3.2% false positive rate.

I'm not sure if there are better studies out, or if this is still the same testing protocol that is most widely used. So, I'm not sure if that is still a decent estimate.

Thanks.

That's more than a trivial issue if the US is hypertesting and using absolute number of new detected infections as a benchmark for easing restrictions.

At 98% sensitivity, run 10k+ tests per day in a metro area and you will always get 200+ positive results.

If MLB runs 1000 tests per day, that's 20 positives.

True. The false positive rate can be a problem when your tested population has low prevalence (like frequently tested MLB players and or folks in NYC).

The false negative rate can be quite high depending on when in the infection cycle you're testing folks and the details of the testing procedure I've seen estimates from 3-30%.

At a false negative rate of 20% and a false positive rate of 3.2% you generate more false positives than false negatives when your test positivity fraction is below around 13%. I think the US as a whole has a positive rate of around 8% now. So, if that FPR and FNR are about right, we probably have some inflation of confirmed counts due to false positives.

Of course, models like Youyang Gu's are estimating that we had somewhere around 444K new cases on July 31, and we confirmed something like 67K. So our true numbers are probably still over 5 times the number of confirmed cases.

Keegan, I think you're interested in specificity, not sensitivity.

A 100% specific test will virtually guarantee that any positive is a true positive. Sometimes at the expense of producing false negatives.

A 100% sensitive test will virtually guarantee that any negative is a true negative. Often at the expense of producing lots of false positives.

Right now the estimated sensitivity is close to 80% for PCR. Specificity is estimated to be very close to 100%.

However you have to look at pre-test probability. If you you're testing a population with a lot of COVID, you can assume that any positive is likely a true positive.

Conversely, if you're testing a population with very low prevalence of COVID, it's almost inevitable that labs will error at some point. Whether this is due to contamination, or machine error, it's not uncommon for 100% specific tests to still have false positives. Although there aren't that many 100% specific tests out there.

Data isn't great right now, because we're still collecting data, some machines are better than others, the gold standard to measure different tests against is typically PCR, and we know for a fact that it's not that great for COVID as it's highly dependent on collection method/technique, time since exposure, time since symptom onset, etc.

Yep. Got 'em backwards. Thanks!

I have seen around 6 studies and they are all over the board. Averaging the data available we are running with a general consensus of a 30% false negative rate on the PCR, which is better then the 40% false negative rate on the Abbott rapid antigen test.

And false positive?

(That's what I was really curious about, but I got boneheaded and inverted the terms.)

Surgeon here. We use the PCR test pre op on all patients. The sensitivity and specificity is sufficient with some variability between vendors (ie the company that supplies the kits) and likely the actual labs processing the tests. The stats that would really round out the answer to your question are Positive Predictive Value and Negative Predictive Value. I've not seen these stats nor do I care enough at this point to investigate. I mention this in case someone wants to consult Dr. Google and report back here.

Why don't I care? Testing has kept 100% of our surgeons healthy through the pandemic; a few other non surgeon OR staff have gotten non fatal COVID. We've had two non OR providers die from COVID. Both healthy and endurance athletes; aged 45 and early 30's; non surgeons. These two unfortunate souls probably got COVID from treating COVID patients so a negative test wasn't something that would've helped them.

Put another way, testing seems to have kept most of us healthy as we roam around a workplace that is a virus factory. Testing appears to have helped surgeons avoid the death rate surgeons were experiencing in the early days in China and Italy where they were subjected to inadvertent massive viral exposure (cytokine storm seems to be linked to how much virus you get exposed to at once).

I realize this testimonial doesn't really answer the OP as it relates to testing outside of the application to exposure of healthcare workers. While it isn't black and white science, maybe this experience lends some validity to testing in high risk situations as we await the many retrospective studies on what is occurring right now.

Also, I don't care because I am late for rounds at the hospital...