Saying that perhaps 25 percent of all healthcare errors are errors of diagnosis, Kohn noted how getting the diagnosis right can prevent all kinds of unnecessary complications and spending. “Of course, if you’ve made the wrong diagnosis, picking the right course of treatment becomes a challenge,” Kohn said.
And after the diagnosis, Watson can prevent treatment errors by, say, scanning EMR data for patient allergies to recommend against a drug that might cause a harmful interaction, then suggest alternative therapies. Kohn presented the case of a 29-year-old pregnant woman who was diagnosed with Lyme disease. A common treatment is the antibiotic doxycyline, but Kohn noted that it’s contraindicated during pregnancy.
Watson, according to Kohn, draws preliminary conclusions according to presenting symptoms, then scans multiple sources of information to present recommendations. Watson does look at the notoriously incomplete and inaccurate Wikipedia, Kohn said, mostly because that user-edited site covers so many topics, but then verifies information from other sources.
Watson then displays reasons why it believes the diagnosis may be correct so the doctor can make an informed decision. “It won’t let you ignore all the possible diagnoses,” Kohn said. But it won’t actually make the final call. “Watson is going to be in a supportive role rather than actually making decisions.” Kohn added.
What the supercomputer does is process vast amounts of data in a short amount of time., something that even the sharpest human mind could never do. And that’s what clinical decision support is supposed to be all about.