Normality may be dangerous

A story in today's Science Times about the variance in quality of MRI scans by Gina Kolata made me think not about MRI per se but rather about the notion of quality in health care. Kolata tells a few anecdotes about colleagues whose MRIs did not lead to a diagnosis of a problem that turned out to be serious. The article doesn't discuss the frequency of these errors but suggests that use of MRI is likely to be variable for several reasons. Different body tissues respond differently to MRI scans, individual MRI machines vary enormously in the age of their technology and what software they use, certain machines and settings are better for certain types of tissue, skills among technicians who run the equipment vary, and finally many radiologists who read scans are specialized for scans of bone as opposed to, say, lung tissue, others are generalists. Their conclusion? Doctors shouldn't rely on scans alone but on good history taking and a thorough examination. All well and good, machines aren't perfect and the people who run them and interpret their results aren't either. No surprises there. But this made me think of Atul Gawande's terrific book Better (my post about it is here), probably because I was talking with a colleague about it yesterday at the lab, and about Gawande's observation that, like every other industry, medicine's performance can be charted on the bell curve, i.e., most performance is average. Gawande writes:

It belies the promise that we make to patients: that they can count on the medical system to give them their very best chance. It also contradicts the belief nearly all of us have that we are doing our job as well as it can be done.

The thing about the normal curve (pictured left) is that it is, well, normal. It defines a phenomenon in nature. By definition most people's performance falls in the middle, the numbers of those who are a little above or below average is smaller, and those who are exceptionally bad or good are very rare. If you improve the average score with a series of trainings, it is likely that the other scores will settle about that central score in the same pattern. And if occasionally someone's performance goes off the charts as exceptionally good or bad because they won a prize that boosted their confidence, or they experienced a personal tragedy, over time their performance will tend to become closer to average again - this is regression to the mean. It is a truism - the average performance is, by definition, average. Anyhoo, my point is that exceptional performance in any industry, even one that cares for our health - one where we want better than average care if we can get it - is abnormal. If it that is your aim you must find a way to constantly throw nature off course.

Kolata's article quotes a Dr. at Massachusetts General Hospital as saying:

"musculoskeletal MRIs are read by someone who does musculoskeletal imaging every day"... It pays to check the credentials of a center's radiologists.

But that got me thinking, is hyperspecialization really the answer? (I don't mean I want an anesthesiologist switching with a plastic surgeon every other thursday, but I am questioning reading only one kind of scan day in and day out). I can think of at least two advantages of specialization 1) the medical and technological knowledge base is forever growing and no one can keep up with every advance; 2) specialists develop expertise because they look at a far greater number of bone scans than they do liver scans and we want those with the most exposures to our type of tissue to be reading our scan - don't we? I'm not sure. There is an improvement in performance for experts as compared to novices but, on the other hand people habituate to what they are used to and over time there is a fall-off in the performance of those who must be vigilant in the same way for a long time. People burn out when they do the same thing over and over. I wonder if there isn't a point where the advantage gained by expertise is lost to burn out. Has this balance been studied?

As humans develop, we generally start a little hypothesis testing machines. We acquire as much knowledge as we possibly can, and then acquire skills too. After the basics - walking, talking, toilet training, tying our shoes, washing and feeding ourselves, reading, etc... the rest is gravy. Our acquisition turns to learning heuristics, that is we learn to take short cuts, saving time and freeing up cognitive resources for... well, just freeing up cognitive resources. In the wild we might have adapted this way so as to always have some cognitive power left for emergencies, today I guess it gives us time for Dexter and Gray's Anatomy or perhaps blogging. One might say we are wired to be lazy.

If excellent performance is not normal and we must throw nature off balance to continually do anything at exceptionally high level, what kind of environment would provide a balance of the right level of expertise with enough newness to keep those who work in healthcare (or any other industry that aspires to exceptional performance) on their toes?