But Lukov’s true medical condition wasn’t quite so grim. The tumor did have a driver—a third mutation few oncologists test for in this type of case. It’s called KRAS. Researchers have known about KRAS for a long time, but only recently have they realized that it can be the driver mutation in metastatic lung cancer—and that, in those cases, it responds to the same drugs that turn it off in other tumors. A doctor familiar with both Lukov’s specific medical history and the very latest research might know to make the connection—to add one more biomarker test, for KRAS, and then to find a clinical trial testing the efficacy of KRAS treatments on lung cancer. But the national treatment guidelines for lung cancer don’t recommend such action, and few physicians, however conscientious, would think to do these things.
Did Lukov ultimately get the right treatment? Did his oncologist make the connection between KRAS and his condition, and order the test? He might have, if Lukov were a real patient and the oncologist were a real doctor. They’re not. They are fictional composites developed by researchers at the Memorial Sloan-Kettering Cancer Center in New York, in order to help train—and demonstrate the skills of—IBM’s Watson supercomputer. Yes, this is the same Watson that famously went on Jeopardy and beat two previous human champions. But IBM didn’t build Watson to win game shows. The company is developing Watson to help professionals with complex decision making, like the kind that occurs in oncologists’ offices—and to point out clinical nuances that health professionals might miss on their own.
Information technology that helps doctors and patients make decisions has been around for a long time. Crude online tools like WebMD get millions of visitors a day. But Watson is a different beast. According to IBM, it can digest information and make recommendations much more quickly, and more intelligently, than perhaps any machine before it—processing up to 60 million pages of text per second, even when that text is in the form of plain old prose, or what scientists call “natural language.”
That’s no small thing, because something like 80 percent of all information is “unstructured.” In medicine, it consists of physician notes dictated into medical records, long-winded sentences published in academic journals, and raw numbers stored online by public-health departments. At least in theory, Watson can make sense of it all. It can sit in on patient examinations, silently listening. And over time, it can learn. Just as Watson got better at Jeopardy the longer it played, so it gets better at figuring out medical problems and ways of treating them the more it interacts with real cases. Watson even has the ability to convey doubt. When it makes diagnoses and recommends treatments, it usually issues a series of possibilities, each with its own level of confidence attached.
Medicine has never before had a tool quite like this. And at an unofficial coming-out party in Las Vegas last year, during the annual meeting of the Healthcare Information and Management Systems Society, more than 1,000 professionals packed a large hotel conference hall, and an overflow room nearby, to hear a presentation by Marty Kohn, an emergency-room physician and a clinical leader of the IBM team training Watson for health care. Standing before a video screen that dwarfed his large frame, Kohn described in his husky voice how Watson could be a game changer—not just in highly specialized fields like oncology but also in primary care, given that all doctors can make mistakes that lead to costly, sometimes dangerous, treatment errors.
Drawing on his own clinical experience and on academic studies, Kohn explained that about one-third of these errors appear to be products of misdiagnosis, one cause of which is “anchoring bias”: human beings’ tendency to rely too heavily on a single piece of information. This happens all the time in doctors’ offices, clinics, and emergency rooms. A physician hears about two or three symptoms, seizes on a diagnosis consistent with those, and subconsciously discounts evidence that points to something else. Or a physician hits upon the right diagnosis, but fails to realize that it’s incomplete, and ends up treating just one condition when the patient is, in fact, suffering from several. Tools like Watson are less prone to those failings. As such, Kohn believes, they may eventually become as ubiquitous in doctors’ offices as the stethoscope.
“Watson fills in for some human limitations,” Kohn told me in an interview. “Studies show that humans are good at taking a relatively limited list of possibilities and using that list, but are far less adept at using huge volumes of information. That’s where Watson shines: taking a huge list of information and winnowing it down.”
by Jonathan Cohn, The Atlantic | Read more:
Illustration: Bart Cooke
