This is why I advocate for computer aided diagnosis. It is not useful in the vast, vast majority of the cases but when you hit the wall it is time to look at the remaining few percents.

A set of questions can help to pin point the unlikely case, like here where the link to the database of the 13 patients could have been a hint.

Now, since this is not done at scale means that there are (good or bad) reasons. I am obviously not an MD.