Feb 13, 2012

Posted in Artificial Intelligence, Assisted Decision Making, Harlan/Siri

The Future of IBM’s Watson

Share Button


Steve Lohr reports from FutureMed, held at the Googleplex:

Dr. Martin Kohn, chief medical scientist at I.B.M. research, sketched out the future path in health care for the technology behind Watson, the computer that last year outwitted the best human players of Jeopardy!, the TV question-answering game. “You’ll not be surprised to learn that the executive leaders of I.B.M. fairly quickly decided that playing Jeopardy! was not a long-term business model,” Dr. Kohn told the audience of a few hundred people.

But the core transferable technology, Dr. Kohn explained, was the artificial intelligence software that made it possible for Watson to read and understand 200 million digital pages, and deliver an answer within three seconds. In health care, Dr. Kohn said, “we are overwhelmed by information. And we’re only as good as what we know.”

So Dr. Watson, it is.

Watson, he added, is not going to make diagnoses, not give a physician a single answer, but make suggestions, recommendations and determine probabilities. The more information Watson is fed, Dr. Kohn said, the more it learns and understands, in its way.

One area where the technology’s learning and recommendation capabilities may be particularly useful is in determining treatment regimens for patients with more than one chronic condition. And such patients account for a large share of the nation’s health care costs.

There are well-defined treatment guidelines, Dr. Kohn said, for individual conditions like heart disease, diabetes, asthma and emphysema. But the guidelines are far less helpful for patients with more than one condition. For example, a beta-blocker drug is good for heart disease, but bad for asthma, Dr. Kohn noted. What are the trade-offs and what are the probabilities?

Watson, Dr. Kohn said, can “really help us learn about these multiply-challenged patients.” In general, he added, Watson can be a powerful tool in moving toward the long-sought goal in health care of making more decisions based on data and a surer grasp of the relevant scientific evidence — so-called evidence-based medicine — instead of experience and intuition.

The Watson technology, Dr. Kohn added, has the potential to be a “profound enabler of the transformation of health care.”

Lohr writes about some of the legal and economic implications:

It’s first big test-run in health care is with Wellpoint, an insurer. I know, I know, there are no “insurers” anymore, only health management organizations. But an insurer makes money by maximizing revenue — premiums — and minimizing expenses — procedures, lab tests, hospitalizations and treatment.

Watson will do what it’s programmed to do. Eliminating unnecessary, and often repetitive, medical tests is a big cost-saving target — and a good one, for patients and for health-care budgets. But the research on this subject has concluded that about half of patients get too much treatment and about half get too little.

The notion that a technology like Watson, if unbiased, is going to reduce health care spending significantly seems misguided. But such technologies can make a useful difference if the right economic incentives are in place. And that, of course, is what the drift toward “accountable care organizations” is about. That is, a system in which doctors, hospitals and insurers are paid for helping people live healthier lives.

If that happens, the nation’s health care bill won’t necessarily fall. People living healthier and presumably longer lives will be consuming health care services for more years. But at least the money would be spent more wisely.

Share Button
Print Friendly