INFORMATION TECHNOLOGY: Artificial intelligence seen as decision support aid
Artificial intelligence seen as aiding decision support
The decades-old quest to bring artificial intelligence to bear on everyday healthcare decisionmaking is moving inexorably forward, according to health informatics experts. So what’s to be made of recent announcements by Indiana University and IBM Corp. about advances in healthcare AI? The apparent progress is just great, the experts said, but it’s still very early in the game.
“I think there is a new generation of technology coming out in AI, in computational power, that will make a big difference,” said Casey Bennett, an IU computer scientist. “So, I think over the next 10 years, you’ll see a big change.”
“I think we’re starting to move to the next domain within clinical decision support,” said Dr. William Bria, a veteran physician informaticist, president of the Association of Medical Directors of Information Systems and chief medical officer of Dataskill, a San Diego systems integrator where Bria is helping develop clinical decision-support, or CDS, systems that leverage the computing power of IBM’s Watson supercomputer system.
How long before sophisticated CDS systems such as those will be in widespread routine use by clinicians? “Another decade,” Bria said. Of the two announcements, the one with the bigger ballyhoo came from IBM, with contributions also by insurance giant WellPoint and New York’s renowned Memorial Sloan-Kettering Cancer Center, all related to Watson.
Famous for its 2011 victories on the TV game show “Jeopardy!”, Watson has since been crammed with 1.5 million patient medical records, 2 million medical journal articles and more than 600,000 “pieces of medical evidence,” according to an IBM news release touting its latest “breakthrough” developments.
One is a Watson-powered oncology clinical-decision support tool. In creating the tool, Watson received about a year of tutoring from oncologists at Memorial Sloan-Kettering.
But the CDS tool itself is still “in training,” according to a Memorial Sloan-Kettering spokeswoman. It is not yet linked to, much less integrated into, the cancer center’s electronic health-record system, it won’t be tested in a clinical setting until fall and a commercial product is not expected for release until the end of 2013, the spokeswoman said.
Meanwhile, in another Watson application, nurses and technicians at WellPoint have worked with the computer for more than a year, with more immediately serviceable results, according to IBM. They have developed a Watson-leveraged system that, in December 2012, began processing provider requests to WellPoint for approval of common medical procedures. WellPoint plans to have more than 1,600 providers using it by the end of the year, according to a news release.
Bennett is a doctoral student and instructor at IU’s School of Informatics and Computing and a research fellow at Centerstone Research Institute, an arm of Centerstone, a behavioral healthcare provider based in Columbus, Ind., and Nashville. Bennett and colleague Kris Hauser, an assistant professor of computer science at IU, are co-authors of a recently published research report, “Artificial intelligence framework for simulating clinical decisionmaking: A Markov decision process approach,” in the journal Artificial Intelligence in Medicine.
The report tells the story of an AI system they developed that second-guessed the decisions of clinicians in 500 cases involving behavioral-health patients with a variety of other chronic illnesses, such as hypertension, diabetes and cardiovascular disease. Their AI system came up with alternative, evidencebased care paths that outperformed those of the human decisionmakers in terms of cost while boosting patient outcomes 30% to 35%, the authors said.
Since publication, they have received some “angry responses” in “some random, unsolicited e-mail,” Bennett said. Accepting a greater level of reliance on computer support will be “a bit of a culture shift,” for some physicians, he said.
“I think the younger generation is more open to it,” he said.
Hard or easy, a shift to more reliance on clinical support tools is going to come, Bennett said.
“With meaningful use and Obamacare, there is going to be a political shift that’s going to force change,” he said. Without it, he believes, soaring healthcare costs will bankrupt the country.
“So, you won’t be able to quibble about it,” Bennett said. “It will happen whether people like it or not.”
“I think it’s all good news, it’s all positive,” said Dr. Mitchell Feldman, who has worked for more than two decades on a computerized diagnostic support tool, DXplain, first developed at the Laboratory of Computer Science at Massachusetts General Hospital, Boston. “It’s trying to bring more evidence and more data and more science and a little less art to medicine,” Feldman said.
For now, though, “No one I know in informatics is ready to trust their patient care to a computer, but is man plus machine better than man alone? The answer is definitely yes.”
Dr. Mark Kris of Memorial Sloan-Kettering Cancer Center, left, and Manoj Saxena of IBM demonstrate medical applications connected to IBM’s Watson supercomputer.