At FutureMed, IBM's doctors explained how Watson could improve quality of care, reduce errors, and help doctors better utilize skills.
FutureMed featured back-to-back sessions from Dr. Paul Grundy, and Watson guru Dr. Marty Kohn. We'll give you a rundown of both of their presentations here:
Grundy, IBM's global director of health care transformation, is a proponent of a primary-care model called the "Patient Centered Medical Home." Grundy explained that he developed the model by looking for the pain and suffering in the current health care delivery system and working to change it in "profoundly different ways."
Earlier, when looking at the U.S. health care system, Grundy came to the conclusion that "we in the United States were dead last and we were dead last at twice the price of any other civilized nation on the face of the earth. And it has just gotten worse since then."
- Erection-Inducing Condoms May Soon Be Available in Europe
- Proteus ONE, a Smaller Proton Therapy System From IBA
- Hair May Contain a Record of Your Travel History
But he noticed that there were places around the U.S. where one was significantly more likely to get the level of care that one could get in, say, Switzerland or Australia. Dubuque, Iowa, is an example of that, he said. Dubuque is one of the least expensive health care markets in the country.
Not surprisingly, Grundy explained that he had received questions from the IBM CFO about the regional cost of care in the U.S., which the company was considering when deciding where to locate jobs. As it turns out, IBM decided to locate a call center in Dubuque -- partly as a result of the low health care costs there.
And when looking at the costs of various health care markets, Grundy noticed "tremendous waste." Depending on which economist one listens to, administrative costs and overhead make up 19 to 30 percent of the health care costs, he said.
When asked what kind of health care they wanted, IBM employees stressed the importance of the relationship with physicians. "We decided that we really fundamentally wanted to change the covenant between the buyer and the provider of care," Grundy said. "And that is happening today."
While the U.S. has some of the best partial care in the world, the country is poor at coordinating care. "We do not know how to play as a team," Grundy said.
As a solution, Grundy put forward the concept of the Patient-Centered Medical Home, which emphasizes collaboration. The system has been backed by Anthem, Wellpoint, and government agencies such as the Department of Defense.
Grundy explained that, in a recent study with IBM employees, the new delivery system model resulted in the following gains in efficiency, which he ascribes to "robust prevention in primary care:"
- 36 percent drop in hospital days
- 32 percent drop in ER use
- 9.6 percent drop in total cost
Another study by CareMore that tried out the Patient-Centered Medical Home model found the following:
- a hospitalization rate 24 percent below average
- hospital stays 38 percent shorter
- an amputation rate among diabetics 60 percent lower than average
- these improved outcomes have come without increased total cost
Next, Kohn, chief medical scientist of care delivery systems at IBM Research, gave the audience at FutureMed an in-depth look at the medical applications of Watson, which famously bested human competition in Jeopardy! about a year ago. "Our goal, with [Paul Grundy]'s leadership, is to understand what is necessary to support the transformation of health care to the patient-centered evidence-based outcome-based system that actually makes people healthy and reduces cost," Kohn said.
Kohn explained that Watson in health care could offer the following benefits:
- improve quality of care
- reduce errors
- engage patients
- improve audit trails
- improve efficiency
- better utilize skills
Watson was developed for Jeopardy! and, on the game show, could only access internal data. In the future, the system could be used to actively look for new data from multiple sources, including the Internet. While Watson now only understands English, in the future it could understand multiple languages, which it can leverage to process even more data.
While on Jeopardy!, Watson was only able to answer questions with a single answer, but the system can provide multiple answers to a question, assigning a confidence level to each.
This capability could be used to improve physicians' ability to practice evidence-based care. It could also be employed to help physicians navigate the much-discussed and ever-increasing sea of data. Just looking at journal articles alone, there are 800,000 articles published each year, Kohn said.
What is unique about Watson, Kohn explained, is that it is interactive and will ask for more information when necessary to improve decision making. If a patient comes into a hospital complaining of feeling dizzy, Watson could suggest follow-up questions that the doctor could ask the patient to improve the chance of making the right diagnosis. "We know that patients use terminology that is very different than the way health care professionals use terminology. So, a patient says 'I am dizzy,' and, as a health care professional, I may think they mean 'vertigo' but they may mean something completely different. There are something like 150 different things a patient could be referring to when they complain of dizziness," he said. "Well, Watson with its paraphrasing ability would understand that."
When EHRs debuted, they were first hailed as a panacea for health care. "But they don't help one prioritize information," Kohn said. Watson could help clinicians sort through EHR data to find out what is the most relevant to improve patient care.
Watson also can help physicians overcome problems with self-reinforcing perception bias, which is normal but a source of errors, Kohn said. An example of this bias can be found when a patient comes in and a physician "systematically listens for things that support an original hypothesis and systematically suppresses information that is contrary to it," he said. And, as a result of this, you come out with flawed or limited diagnoses.
By contrast, Watson first looks for possible responses. Then, it searches for evidence sources to evaluate the likelihood that the possible responses are relevant and to what extent. This could work in health care where Watson could sort through a long list of possible diagnoses and then use the evidence sources to make suggestions.
This post also appears on medGadget, an Atlantic partner site.