IBM's Watson Hits Medical School

IBM's Watson is going to medical school to learn what physicians know and how they think.

The famous supercomputer, which already has mastered many aspects of human language, will interact with medical students and experienced clinicians at the Cleveland Clinic Lerner College of Medicine at Case Western Reserve University to gain an understanding of medical concepts and reasoning in a problem-based learning environment.

Eventually, Watson might be able to help teach the medical students, too. "Watson might be an aid to the students as they do their problem-based learning," said James Stoller, MD, chair of the Cleveland Clinic Education Institute, in an interview with InformationWeek Healthcare. When Watson knows medicine, he said, "it could produce a prioritized list of possibilities regarding the cause of a patient's clinical problem and the probabilities assigned to each of those possibilities, as well as the resources that Watson used to assemble that list."

IBM has paired Watson with a number of healthcare organizations for various purposes. For example, Memorial Sloan-Kettering Cancer Center in New York is using Watson to analyze patient records and search the medical literature to help find the best ways to treat cancer patients. Insurer WellPoint and Cedars-Sinai Medical Center in Los Angeles are deploying Watson to crunch data from electronic health records, test results, clinical protocols, and studies, also for the purpose of cancer treatment. And Nuance Communications, an IBM partner, is combining its expertise in natural language processing with Watson's computing power to improve medical diagnosis and treatment, working with Columbia University Medical Center and the University of Maryland School of Medicine.

But Watson has a long way to go before it understands basic medical concepts and how they're interrelated in particular cases, Stoller said. "It needs to learn medicine before it can teach medicine," he said. For example, Watson must learn how to relate all of the medical terms that describe the same concept, such as "emesis" and "vomiting," so that the terms won't be misinterpreted.

[ Practice management software keeps the medical office running smoothly. For a closer look at KLAS' top-ranked systems, see 10 Top Medical Practice Management Software Systems. ]

In addition, he said, Watson must learn "nuances and associations of meanings in the same way that a student or a seasoned clinician does" to be effective in clinical decision support. And although Watson will be able to sift through enormous amounts of information in the medical literature to find answers to clinical questions, it must learn how to evaluate the strength of evidence in individual studies, said Stoller.

When it does that, Watson could help doctors improve their ability to practice evidence-based medicine. "Watson will advance the best practice of medicine based on evidence to the extent that it learns the most rigorous information," said Stoller. "That will be a function of how it gets trained and what literature gets loaded into it. The intent would be to make sure it gets loaded with the best available information and that it's trained the best possible way by seasoned doctors and medical students recreating the way they think and learn."

Someday, Stoller said, Watson might be able to provide reliable clinical decision support to physicians in their daily work. In this scenario, a physician would interview a patient, do a physical examination, order lab tests, and so on, and a synthesis of this information would be provided to the supercomputer through some kind of interface.

"Watson will have access to the same data set that I do and will use that to create a list of diagnostic possibilities in order of the likelihood of them being correct," he said. "If I wanted to take a deeper dive, I could ask Watson how it was weighting the evidence that gave rise to the probabilities. That would give me a check on my own thinking."

If this scenario became a reality, he said, "It could be a paradigm change in how we practice medicine. It could be a powerful, easy-to-use bedside decision support tool with deep understanding that helps make sure that our lists are complete and that there's concordance between the physician's thinking and the decision support thinking."

Although Watson would not replace the clinician's judgment, it could help him or her evaluate the soundness of medical decisions, he said. Physicians would also still have to respect patient preferences in their decision-making, he added.

InformationWeek Healthcare brought together eight top IT execs to discuss BYOD, Meaningful Use, accountable care, and other contentious issues. Also in the new, all-digital CIO Roundtable issue: Why use IT systems to help cut medical costs if physicians ignore the cost of the care they provide? (Free with registration.)