Medusa 19 saliva testing

[Brian Freeze]

Work are paying for us to be offered these saliva tests, should I be signing up?

Im not involved in the decision making processes and cant influence it in any way so its just a case of whether I decide to or not and am not science literate enough to work out if its any good.

We've only seen the company blurb and I haven't a clue if 92% specificity/94% accuracy and 96% sensitivity are much cop. The numbers look nice and high but there is an FT article which isnt very complimentary about them.

Sorry for not posting links, I'm on my phone.

Can anyone shed any light please? I trust you lot.

[Zetetic]

Ask them for a confusion matrix reflecting current estimates of prevalence.

[Zetetic]

Something like this, for 1000 people?



Yellow is false positives, red is false negatives.

I'm getting accuracy of 92%, and that's rounding in their favour, so I guess they're estimating really high prevalence. Or I've cocked something up.

All assuming you believe their sensitivity and specificity claims, of course.

[Zetetic]

https://www.ft.com/content/2f6c4175-19e6-403e-ac7a-632d5df8fb5f

Yeah, FT, points out the false positive rate.

I don't know what you'd do with the information that there's a 1 in 20 probability that you have COVID-relevant antibodies.

I don't know what you'd do with the information that you definitely have them, mind you, but that'd be more interesting.

[Zetetic]

In fairness, I have cocked up a bit, because I was focused on COVID-19 prevalence and antigen testing and not antibody prevalence and testing.

If you reckon that a third of people have COVID-19 antibodies, then it looks something like this:



And a positive test is more like a 6 in 7 chance of being real.

[Brian Freeze]

Cheers Zetetic, I will try and educate myself about what it all means, got a six year old to entertain on her birthday today but will try and do it over the weekend.