Skip to main content

IBM's Watson Heads for Clinics, Spurring Debate

 |  By smace@healthleadersmedia.com  
   February 12, 2013

As a young emergency room physician, Martin Kohn, MD, thought he knew most everything. But once in a while, quiet whispers in his ear were the difference between success and a life-threatening medical decision error.

"What saved me and my patients more often than not were three nurses in that emergency department who had been there forever, and were very diplomatic. I could be sitting there struggling because we were doing stupid things like 24-hour shifts," Kohn says. "I'd be struggling with something and one of the nurses would come up and say, Marty, did you think about such and such?"

Now, Kohn is leading the team at IBM that will bring a technology-powered version of that little voice to clinics starting at the end of 2013.

"In some ways I view Watson as that friendly, helpful nurse who by experience knows these things, and just whispers quietly over your shoulder," Kohn says. "Nobody else has to see it."

In case you missed it on the TV game show Jeopardy, Watson is a set of massively parallel probabilistic algorithms able to break apart and parse natural language in different ways, and to suggest possible answers to questions.

In healthcare, Watson is being trained and tested first in the oncology clinics of Memorial Sloan Kettering Cancer Center, located 30 miles away from IBM's development laboratories in Yorktown Heights, New York.

For months, I've nosed around for more details about Watson's role in healthcare, but those hospitals that are working with IBM are under strict nondisclosure agreements, and are press-shy. But last week, Kohn made a lap around Silicon Valley, speaking at the FutureMed conference and then at an event hosted by Triple Ring Technologies, a startup incubator. I caught both talks.

Both audiences were peppered with physicians, such as one from Stanford Medical School, who I chatted with at Triple Ring. He remarked to me that he wasn't overly impressed yet with Watson.

I would expect physicians to be wary of Watson. Reports of its projected diagnostic prowess have fed fevered speculation that artificial intelligence is at long last about to make doctors obsolete, will turn healthcare over to a set of algorithms, and will allow technology to replace a clinician's gut instinct.

But Kohn and IBM insist that's not Watson's role. A second effort with healthcare payer Wellpoint, touted by that company last week, may place Watson in a more centralized decision-making role, allowing or disallowing requested procedures based on Wellpoint's own use of Watson to scour medical literature to support or challenge a doctor's diagnosis or recommended treatment.

Like the famous HAL computer in the movie 2001: A Space Odyssey, Watson has had to learn by starting with basic stuff. When its objective was to beat human Jeopardy champions, Watson marshaled 12 terabytes of stored data, including questions and answers from previous Jeopardy games, to help decide which information was most relevant in forming an answer to each new Jeopardy challenge.

In healthcare, sometimes there is no one right answer, so Watson is being trained to return a list of possible diagnoses based on input ranging from the patient's own story, ordered tests, experiences of similar patients, and peer-reviewed medical literature, which is weighted most heavily.

To date, Watson has ingested more than 600,000+ pieces of medical evidence, two million pages of text from 42 medical journals, and clinical trial data in the area of oncology research.

Like a good journalist, Watson also learns which sources are reliable, favoring those sources' results over time, and discounting those results that come from less reliable sources. The term "artificial intelligence" is a bit of a misnomer, because Watson doesn't generate new ideas itself, but instead relies upon accessing existing information.

What makes the Memorial Sloan-Kettering test fascinating is that the institution already has 15 years of electronic health records on five million encounters, Kohn says. Over time, this kind of data will become some of the most valuable assets of any healthcare system, and for the first time, it will be powering an AI system to help make decisions.

To those who protest that Watson will simply be too much technology for too little gain, it's worth remembering that the amount of medical data available doubles every five years. In 2010, the National Library of Medicine cataloged 700,000 new articles. "I didn't read them all," Kohn quips. "There's information out there that we could use that we can't get to. Businesses that can process that information are two times as likely to succeed."

Kohn is IBM Research's chief medical scientist for care delivery systems. He co-authored IBM's white paper on the patient-centered medical home. But he is also a graduate of MIT, with engineering bachelor's and Master's degrees. He speaks both geek and doc—and both are definitely required for this "grand challenge," as IBM terms it.

If Watson can help providers prevent "avoidable" adverse events, it might prove to be a prudent investment. But for the time being, Watson will only be available to the largest healthcare organizations. Small practices won't be able to afford it, at least not until someone offers Watson in some sort of pre-packaged offering in the cloud. That will take some time.

Remember that natural language free-text analysis is just one piece of the analytics puzzle. Analysis of structured data and images, offers its own separate benefits to healthcare. And improving healthcare is a lot harder than winning a Jeopardy game.

But techniques such as Watson may also crack some of healthcare's harder nuts, such as how to decide on a course of treatment when multiple chronic diseases are present. For instance, should the physician treat congestive heart failure at the risk of making the same patient's asthma worse?

In some cases, only a massive speed-read of all available literature may be able to help suggest answers to that question and others like it.

Meanwhile, the debate about the role of AI in healthcare is just getting started. I expect a vigorous set of put-downs of Watson in the comments on this column. But just as I wouldn't have bet against Wikipedia or Google, I'm not going to bet against this technology's value either. After all, even Wikipedia and Google have assumed a certain role in today's healthcare technology puzzle, even if it's only to advise patients while they wait for their specialist appointments. (And what doctor doesn't use Google in his or her research?)

In the near future, even the medical experts will find themselves more dependent on and appreciative of ever more technology-fueled answers as well.

Scott Mace is the former senior technology editor for HealthLeaders Media. He is now the senior editor, custom content at H3.Group.

Tagged Under:


Get the latest on healthcare leadership in your inbox.