Thread by @apsmunro

🧵
A couple of things people I think people have misunderstood about the school DCT study - lets clarify them!

First, test sensitivity of "only 53%"

Does this mean the test is rubbish?

Well actually, it's irrelevant!

1/

There is a BIG difference between a study of test performance, and real world test implementation

Good tests can be harmful in the real world, bad tests can end up being useful

For this study, we don't care about performance...

2/

This study measured secondary cases compared to quarantine - the outcome we actually care about is people getting infected

There is no noticeable difference here. It could be both are equally terrible! the important thing is there is not a big difference

3/

Next, concerns about the non-inferiority level are overblown, because they ignore that this is relative, not absolute

Rates of secondary cases were <2%. A 50% increase from e.g. 1.5 to 2.2% is negligible in real terms, and would be well worth not dismissing bubbles

4/

People draw attention to relative risks like this to obscure the true impact of the intervention

Besides, the result was well within this margin - the upper CI was 1.22, so a max 22% increase (e.g. 1.5% to 1.83%, or 2% to 2.44%) - tiny real increase

5/

This shouldn't be impacted by community transmission either, as this is rates per index case. More spread elsewhere will just mean more index cases.

There you have it - don't let statistical wizardry or misdirection pull the wool over your eyes!

6/