We are by definition reacting different, and unfortunately statistics are not helping on this case.
Basically, we have:
- too many illnesses (I guess 10% of top common health problems, covering 90% of all problems)
- for each illness, we have N number of symptoms (which unfortunately all or only a subset of this can occur)
- then we have fake symptoms (statistically out of threshold values but we are living fine with them)
this is generating a big number of combinations.
With this much combinations, it is requiring extensive amount of tests.
As everything health is also optimized for the greater good. Which requires optimum cost management.
So basically medicine today, with statistical approach, is happy to skip 1%, if it can fix 99% with 90% less cost.
I think this is probably the easy thing to fix when it comes to searching, as you could pretty much leave out the common diagnoses as they'd likely have been tried - Imma call this the "Dr House, is it Lupus?" filter.
From there, you could have "try this" filter, and see what the results are. I wonder what the overlap is between tests and symptoms, but surely with 10 tests and 10 symptoms, most illnesses could be mostly identified?
One should be able to then filter the disease database based on symptoms. Doctors seem to do this in their head, but they are not comprehensive stores of information. I suppose the problem is in the encoding of symptoms, it's not that simple.