[EN] - This might be a sensitive choice, but I think that's a cat.
- Aren't you sensitive too, kitty?
- No, it does indeed have round eyes but that's not a cat, you'll have to be more specific.
In statistics, the sensitivity of a test* is its ability to detect what it looks for (eg. a cancer).
Its opposite is the specificity, or the ability to NOT detect what it isn't looking for (eg. no cancer).
*blood tests, clinical signs, MRI, Rx, etc.
Some of those values have been measured on low amounts of patients, in very specific contexts and with patients not representative of what you would usualy find in a hospital or at home.
- Not ok.
Thus, the presence of a very sensitive sign does not require that a pathology exist. Neither does the absence of a very specific sign indicate that the patient hasn't got a pathology.
This is one of the many reasons why sometimes a doctor believes a patient has a pathology and another does not and they end up ordering more tests.
Another reason would be the prior probability or the prevalence.
Assuming I was trying to diagnose cancers in the general population with a very simple test...
My test would be to check if the person has an arm. If he/she has, then the sign is positive.
Among 100 persons, if 50 (50%) had cancer, they would all be identified by my test, because they all (100%) have an arm.
My test has 100% sensitivity.
Among those 100 persons, if all of them lost their arm, my test would always be negative and it would correctly identify 50 patients "without cancer", but it would be wrong for the 50 others (50%).
My test has 50% specificity.
Pewsner, D. (2004). Ruling a diagnosis in or out with “SpPIn” and “SnNOut”: a note of caution. BMJ, 329(7459), 209–213. doi:10.1136/bmj.329.7459.209.