Thursday, September 15, 2011
Meet Dr. Watson
Michael Laposata is professor of pathology and medicine at Vanderbilt University School of Medicine in Nashville. A number of years ago he coined the terms “pre-pre analytic” and “post-post analytic” to define the interface between the clinician and the laboratory. One the one side, if the clinician orders the wrong test (pre-pre), then everything that follows is for not. On the other side, if the clinical does not understand the result or misinterprets it (post-post), the same holds true.
There are procedures that can be put into place to reduce error at these two critical points. Electronic ordering systems can insert some specific questions that will direct the test orderer in one way or another. And incorporation of advisory notes and cautionary notes and interpretive notes into the report can provide some assistance. But to date neither of these systems is widely used in laboratory practice. We tend to be comfortable staying within our self-imposed boundaries.
But it appears that the next generation of solution makers is just on the horizon, and its name is Watson.
For those who do not watch American game shows, IBM’s most recent super hero is a mega networked computer named Watson. Watson is the son (or maybe the grandson of Deep Blue, the computer that beat the chess genius Garry Kasparov. Watson is today’s best example of how far computers have moved towards faster-than-human artificial intelligence. Basically one loads up Watson with about 200 million pages of information. When asked a question it can process the information of about a million books per second and determine the best connected answer. Watson is likely to provide recommendations for the right test at the right fashion at the right time, and give recommendations for the most appropriate interpretation.
In a head-to-head challenge with the 2 best humans ever to compete in the knowledge game show (Jeopardy), Watson was faster and over 98 percent correct. In a two day competition, Watson “cleaned their human clocks”.
I raise this because the IBM team has decided that the next best target for Watson is to partner up with a medical insurance company to tackle the science and art of medical diagnosis and treatment. The outcome is likely to be predictable. Like the Olympians say “Higher – Faster – Stronger”.
It is the modern linear pattern. In the early chaos, we find leaders, and in leaders we find body of knowledge authorities. Doctors (as an example) became the repository of knowledge. The internet changed all that with the democratization of knowledge so that everyone could believe that they could be an expert. Watson goes the next step.
I suspect that in a world with so much information, nobody can be a better expert than a supercomputer. Watson will take the art out of diagnosis and fill in with lightening speed relational and probability analysis. It will know and integrate what people forget. It will exclude personal bias and myth and “n-of-one” information.
There are factors that make forming diagnoses more challenging than winning on Jeopardy. Good clinicians are influenced by visual and tactile and olfactory clues that will not be available to Watson. And depending on how one asks a question about symptoms, one can get a different response. So I suspect that on a good day and in a one-off situation, the human diagnosticians will come out not too bad.
But on the long haul, we all tend to be burdened and distracted by so many side issues (hunger and heartburn, noises, smells, fatique, taxes, spousal tiffs, parking tickets) that Watson will do better.
Now what we will need is a machine that will improve the pre-examination phase, the next step along the way.
If Watson is here now, can HAL be far behind?
For a course in medical laboratory Quality with a Twist, visit www.POLQM.ca
Registration is now open.