In a practical sense of the differential diagnosis is a symptom based
clinical algorithmic approach usually rather than on a laboratory based
approach. The laboratory algorithmic problem is one based on negative and
positive predictive values per clinical condition or disease and it is
really impossible to have a 100% sensitive or specific laboratory test.
Some of the diagnostic criteria include laboratory parameters but it is very
hard to work backwards from one laboratory parameter. Diagnostic workups can
be very expensive and unfruitful when one out of normal range result is
detected serendipity.
After saying that I use it professionally at work because that's all I see
and not the patient. If I see a hemorrhagic effusion and know that 60% of
such effusions are malignant then I will look extra carefully for malignant
cells or look at the BNP consistent with transudates. CBC results are the
same way in checking the smear for clinically relevent morphology. Too many
examples to mention but I really haven't seen a single good book or a single
online source for that. Keep in mind that we have to look into abberant
results not in-keeping with clincial expectations so we look at artifact and
clinically relevent parameters so before we even get to the differential
diagnosis we must be assured that the result are valid in the first place
and only the laboratory can do that in terms of specimen integrity and
methodology.
We have made a diagnosis before the doctor has and have informed them based
on unique results.

David is pretty bright so I wouldn't go there.

Uh, thanks. But for your suggestion it would never have occurred to me
try using a search engine to search for something. (Hint: neither of the
above searches finds the site I had seen previously.)