While the real biostatisticians get back to me here’s a #tweetorial on why it’s super important to be careful in how doctors, people, anyone interprets #covid test or antibody results. Hope it helps in the hospitals 🏥, doctors offices, ppl deciding if testing makes sense
I declined antibody testing until I was able to enroll in a study where my results would be useful and checked longitudinally. Convinced my husband to decline too. We started with a quick review of sensitivity and specificity which he understood from college/news
Working w this basic understanding I used the example of a pregnancy test that is 99 percent accurate- aka has a specificity of 99 percent. That means that if you took 100 not pregnant women, 99 would have a negative test and one a false positive, right?
Well, kind of, because yes that’s true about the properties of that test in the lab. But most people don’t get negative and positive predictive value, which changes based on incidence of the disease. I avoided numbers and talked conceptually.
Let’s say you have a community where 80 percent of women are pregnant. You give 100 women who don’t think they’re pregnant a test, and one is positive. Is this a false positive? Could be. What about a community where 20 percent of women are pregnant? More likely to be false +
Then we went to my favorite, pre vs post test probability. I didn’t even touch a nomogram (I just torture my residents with those.) instead take a group of 100 women who say they aren’t pregnant and give them tests. One is positive. But she missed her menses and can’t stop puking
Do you think this is still a false positive? Or maybe a real positive because the pretest odds of her being pregnant were higher because of her symptoms. Pre v post test is all that matters! A test changes the probability of a dx, but doesn’t diagnose a patient—doctors do!
So I explain to residents that when a covid pcr comes negative on someone who looks like they definitely have covid, in a community where covid rates are high, that is a different result than the low risk, asymptomatic patient with a negative swab!
And for antibody testing, it’s even harder. We don’t know what the sensitivities and specificity are for each test; we don’t know true incidence; don’t think of pretest probability in each situation outside of research study when interpreting results
Let alone know how antibodies translate into clinical immunity. It’s hard to understand all this as a physician. How will using random, decentralized antibody testing give info that isn’t false reassurance or unnecessary anxiety?
A centralized and organized response is key for public health, and this includes testing and “immunity passports.” If this is left to individuals and individual business it will be an absolute disaster. Period.
#epitwitter #scicomm #medtwitter @DrCSWilliam @Pradipsedation @Covid19Docs feel free to correct and weigh in. This is a dialogue I’m not seeing enough of and as healthcare workers we need to be savvy and smart enough to sort through the noise and help the public understand this!
