Read on Twitter
I am not sure what’s up with all the angsting about test characteristics of rapid tests. Worried about low specificity and false positives? Repeat the test and only consider repeat positives to be positive.
Worried about sensitivity and false negatives? Test twice and consider negative only if 2 tests are negative.
Tests are just tools to sort something into piles, with each pile enriched in the attribute you’re looking for
The only truly useless test would be one where sensitivity = 1- specificity, since that’s a coin toss. Other tests change the post test likelihood that an individual has, or doesn’t have, the condition.
Sens < 1 - spec?
Useful test, you’re holding it upside down
Sens < 1 - spec?
Useful test, you’re holding it upside down
Does that mean every test should be adopted and used? Of course not.
It’s a matter of the value of the information you get through testing. If the test is exorbitant, or acting on the information acquired causes more harm than good, don’t test.
It’s a matter of the value of the information you get through testing. If the test is exorbitant, or acting on the information acquired causes more harm than good, don’t test.
I don’t think there’s much need for me to argue the case for how imperfect, rapid tests can be used intelligently on twitter. Both @michaelmina_lab and @KashPrime have been talking about that for months now. Both recommended follows.
But folks need to argue this on the merits, not on sens and spec which are changeable via repeat testing.
Epi joke: what’s a 100% sensitive test for cancer? A: “You have cancer”
What’s a 100# specific test for cancer?
A: “You don’t have cancer.”
Epi joke: what’s a 100% sensitive test for cancer? A: “You have cancer”
What’s a 100# specific test for cancer?
A: “You don’t have cancer.”
And if you want to know how to use these tools, ask places that have figured it out, rather than reinvent the wheel.
Nova Scotia, Slovakia, the NBA. They’ll probably tell you.
Nova Scotia, Slovakia, the NBA. They’ll probably tell you.
