We need better medicine...

An article in the New York Times ( http://www.nytimes.com/2009/02/01/business/01unbox.html?em ) on disruptive, systemic improvements in medicine got me thinking:

30 years ago published studies in a peer-reviewed journal, (based on nothing more than detailed interview of the patient - no electronic medical records required) found that a simple decision-tree based computerized patient history/complaint program could do as well as the average doctor in diagnosis (half of all doctors are below average...). if we can match the 75th percentile, we'll have improved care for 75% of patients. Automated diagnosis should have been widespread decades ago.

Obviously, a computer-assisted diagnosis would be the very best, getting the best of both worlds. Also, such a system would help with self-care enormously.

The bottom line is that the quantity of medical information outstripped the ability of humans to process it many decades ago - that's why we have specialists. Right now they're doing quite badly: repeated studies find that common ailments don't get the standard of care from 40-80% of the time. We desperately need to automate as much as possible.

I've observed doctors quite a bit professionally and personally over the years in settings that range from primary care to tertiary care in the best teaching hospitals (including Mayo Clinic and the ER on which the television show was based), and I've found that most doctors do most diagnosis with very simple "cookbook" algorithms/decision trees. Any honest physician or participant in the health system will confirm this. This is not at all hard to duplicate (and then out-perform) in software. The doctor detectives you see on television, who diagnosis rare illnesses and track down problems with psychic accuracy....well, they may exist, but most patients don't see them.

The suggestion that software that out-performs most physician diagnosis couldn't have been developed decades ago seems laughably absurd to me. It's obviously physician resistance to competition.

Most physicians (and vets, for that matter), are miserably bad at interviewing, recording and remembering complete, accurate patient histories, primary and specialty physicians alike. That's the very first thing that should be automated, and it forms the basis for good diagnosis.

Also, treatment needs to be computer-assisted. There are many studies that say that common problems, for which standards of care exist, don't get the proper treatment 40-60% of the time.
Automated diagnosis & treatment can't do miracles. But most people get sub-par diagnosis and treatment, and this can be changed. Also, such systems would help patient education, compliance and self-management.

Physician hostility to expert systems prevents them from improving. If such systems were welcomed, widespread and worked with by a very large number of physicians, they would begin to capture the very best approaches.