Dr. Martin Kohn is the Chief Medical Scientist for Health Care Delivery at IBM Research. Here he discusses work underway to apply the IBM Watson supercomputer to meet needs in the healthcare field today.
How Far along is the IBM “Watson” project in medical applications?
Ultimately Watson makes its real contribution in unstructured data analysis. So that’s reasonably far along. Not complete, but reasonably far along, and we’re getting progressively more complicated in the kinds of information being presented to Watson.
For instance, when Watson played “Jeopardy” it had to be taught the context of “Jeopardy.” “Jeopardy” is answers, then you have to come back with the question. So we’re doing the same thing in health care. We’re feeding it information and questions, and teaching it the reasoning it needs for health care.
As far as an actual implementation, we have spent a lot of time in the last year or so going around the country talking to health care organizations to work with them and have them help us identify what kinds of decisions — what challenges in health care organization management — would be valuable for Watson.
One of our early implementations is with WellPoint, the very large healthcare payer organization. One of the implementations that is being developed with them is the process of prior approval for interventions. It’s a very bureaucratic, complicated process that’s time-consuming for all participants. For instance, WellPoint has a panel of nurses that review these requests, and they can either approve them or decide it needs to be sent to a physician panel; so there’s a lot of time involved, a lot of effort on the part of the providers and the payers. There was an article in “Health Affairs” a couple of months ago that estimated for the average physician in the United States, the cost of interacting with payers was $85,000 per year, roughly 14 or 15% of their total practice revenue. So, if Watson is able to help make these decisions — yes this is a request that should be approved, or should go to the physician panel — and make this more efficient, faster. And with Watson being interactive, for example, instead of the physician having to wait a couple of weeks, to get the response that more information is needed; interacting with Watson can be real-time and that may make the whole process more reliable, less expensive, and more efficient.