Watson -- IBM's Jeopardy!-playing supercomputer -- handily defeated two of the game show's strongest contestants, but can it beat cancer?
The nation's biggest health insurer aims to find out. After seeing the computer at work during the show's taping last year, WellPoint executives began brainstorming with Big Blue about Watson's enterprise potential. Its ability to answer questions posed in natural language makes it a logical match for data-driven industries. "We're dealing with a lot of unstructured data -- from medical evidence to patient information -- and that's where Watson excels," says WellPoint SVP & CIO Andrew J. Lang.
The first pilot, launched in December, involves eight WellPoint nurses using Watson-based systems to respond to physician procedure requests. Watson reviews treatment plans against WellPoint medical policy and clinical guidelines, which improves efficiency, particularly in complex cases. "Watson, as a tool for our nurses, is more than just about speed," says Dr. Anthony Nguyen, WellPoint senior vice president of care management. "It empowers them with information so they can do their job better."
The second pilot, launched this year with Cedars-Sinai hospital, tests Watson's ability to suggest treatment plans to oncologists. It's hard for physicians to keep pace with medical research and apply it to patient care. But Watson, IBM says, can sift through 200 million pages of data, analyze it, and provide precise responses in about three seconds.
The pilot will start with breast cancer and have Watson parse medical literature, population data, and individual health records to deliver probability-based treatment options for doctors to evaluate.
Just as Watson improved at Jeopardy! over time, it's expected to get better at evaluating and suggesting treatment plans. In one sense, working at WellPoint may be easier than playing trivia: The field of information is more limited. But in another sense, healthcare is infinitely more complex. There are few standard answers that apply in every case; each patient has a unique history to take into account.
"You have to train it to know when to apply different components to a specific case, and that requires a large amount of data," says Lang. At Cedars-Sinai, cancer experts enter hypothetical patient scenarios, evaluate Watson's proposed treatment options, and make suggestions to improve responses. Similarly, WellPoint's nurses and IT are working together to fine-tune Watson and suggest ways to streamline data input.
Creating a unique patient record that delivers an up-to-date version in a usable form is IT's big challenge. Electronic patient records are now pervasive but usually tied to a vendor platform and not user-friendly. Lang says his team needs to be sure the system they build meshes well with current workflows to ensure adoption. They also aim to apply Watson's intelligent investigative skills within the context of an individual patient's health record.
The project is small and focused -- by design -- but there are plans to expand Watson's use. "We're driving through on these six-week sprints, but also making time to do things the right way in terms of architecture and design and ability to scale," Lang says. "We want to push this horizontally across our process workflow and leverage this more holistically."