Don't most people who get a heart attack die from their first one, with I assume little cost to the system? Of course the general improvements with so many problems means that more people are living long enough to get cancer....
The cost of cancer treatment is high enough that it probably swamps primary care, but one should still run the numbers. Large populations can have results that are non-intuitive to most, e.g. the increased costs of much preventive medicine due to the large numbers that need to be constantly screened to find a few true positives. If I didn't have a family history of aggressive prostate cancer (great-grandfather, 1 out of 6 uncles) I'd listen to my doctor's discouragement on getting yearly PSA testing (now that I'm 49, and yes, I know the test is rather iffy, how iffy I'll learn about after a baseline is established and/or before someone wants to do a biopsy (ouch)). Then again, I'll be paying the full cost of most of those tests.
WRT specialists and outcomes, raw full population outcomes aren't as useful for the individual considering options. In theory/in general the specialist will be able to go beyond the raw numbers and have a better idea of which option is the best bet.
I myself just can't understand the compliance problem ... although I do see the lack of appreciation of complex/the full system issues; even though I ended up going in the directions of chemistry and computers, my early medicine and biology learning was really beneficial.
I myself find the online guides to be useful (well, the Merck Manual is lots cheaper on-line for free :-), for me the biggest limitation in using them is not knowing what's normal. E.g. that tongue coating under these circumstances is no problem, maybe stop taking your antibiotic a couple of days early sort of thing. I know most people aren't in a position to get enough useful out of them to seriously "bend the cost curve", to draw this back to the main topic.
I'd never consider otitis media to be simple, it just obviously isn't, too much hidden, too much guesswork, especially since outcomes are so often ... not optimal (e.g. I lost some hearing in one ear due to one infection in the '70s, a girlfriend as well in the '80s).
Anyway, thanks for the intelligent discussion.