For more than a year, IBM has partnered separately with WellPoint and Memorial Sloan-Kettering to train Watson in the areas of oncology and utilization management. During this time, clinicians and technology experts spent thousands of hours "teaching" Watson how to process, analyze and interpret the meaning of complex clinical information, using natural language processing.
To date, Watson has ingested more than 600,000 pieces of medical evidence, two million pages of text from 42 medical journals, and clinical trials in oncology research, the announcement said. Starting with 1,500 lung cancer cases, Memorial Sloan-Kettering clinicians and analysts are training Watson to extract and interpret physician notes, lab results and clinical research.
[ Want to know more about the part Watson is playing in healthcare? Read IBM Watson Finally Graduates Medical School. ]
The product of the collaborative effort between IBM and Sloan Kettering, which is being tested at the Maine Center for Cancer Medicine and WESTMED Medical Group, is called Interactive Care Insights for Oncology. The core of this product is a Watson-based "advisor," available through the cloud, that is expected to help oncologists and researchers identify the best treatment options for cancer patients.
When an oncologist sees a patient, Watson will search that patient's electronic medical record for relevant information; scour the literature, evidence-based clinical guidelines and other sources; and recommend that the physician order certain tests or provide particular treatments. In addition, Watson will reveal the sources of its recommendations so the physician can drill down and see where they came from.
If the doctor is aware of other symptoms or facts that aren't in the record, or wants to include patient preferences in his medical decision, he can dictate that information into the application, and Watson will alter its recommendations to fit the new facts. It will also provide a "confidence" level for each recommendation.
The application also includes a button for seeking prior authorization from a health plan. The button links to two other products co-developed by IBM and WellPoint: The WellPoint Interactive Care Guide and the Interactive Care Reviewer.
To develop these applications, WellPoint's utilization management nurses trained Watson to review authorization requests for common procedures, using 25,000 test case scenarios and de-identified data on 1,500 actual cases.
Watson started processing common, medical procedure requests by providers for members in WellPoint affiliated health plans in December. The methodology is now being tested in five medical offices in the Midwest to make sure it fits into physicians' workflow and adds value. The goals are to streamline review processes, speed approvals, reduce waste and help ensure that evidence-based care is provided, the press release said.