As noted in a report on Monday, Nov. 27, artificial intelligence has been the talk of this year’s RSNA Conference, sponsored by the Oak Brook, Ill.-based Radiological Society of North America, and currently being held this week at the vast McCormick Place Convention Center in downtown Chicago. The terms “artificial intelligence” and “machine learning” are everywhere at RSNA 2017, and the buzz around the AI and machine learning concepts appears to be cresting.
In the midst of all of the buzz and all the discussions, presentations, and pitches, Healthcare Informatics Editor-in-Chief Mark Hagland was able to sit down at McCormick Place for a brief conversation with James Whitfill, M.D., to discuss AI and the implications of its leveraging in medical practice, for the future of radiology and of medicine more broadly.
Dr. Whitfill is chief medical officer at Innovation Care Partners in Phoenix (formerly Scottsdale Health Partners), a physician-led clinical integration network that was formed in 2012 and has quickly grown to 700 physicians serving 35,000 patients throughout the greater Scottsdale, Arizona community. SHP consists of 115 primary care physicians and the balance are specialists from a wide variety of specialties. SHP also participates in the Medicare Shared Savings Program (MSSP) Accountable Care Organization (ACO) program. As an organization with a strong focus on transforming healthcare delivery in the Scottsdale area, bridging the gap between patients admitted to local hospitals and their primary care physicians is a critical component of SHP’s Care Management Program.
As Scottsdale Health Partners, the organization was named the number-two winning team in the Healthcare Informatics 2016 Innovator Awards program, for the results it garnered from a data-driven, multidisciplinary initiative that streamlined its care coordination program using health IT solutions, and for remarkable progress in reducing hospital readmission rates and averting new readmissions.
Meanwhile, Dr. Whitfill has continued to demonstrate leadership not only at Innovation Care Partners, but also more broadly in the healthcare industry. An internal medicine physician by medical specialty, he is now chair-elect of the board of SIMM—the Society for Imaging Informatics in Medicine, a Leesburg, Va.-based national association of imaging informatics leaders.
It is in that context that Whitfill was helping to lead some discussions and presentations in the Machine Learning Showcase area in the North Hall at McCormick Place, on Monday. Whitfill spoke with Healthcare Informatics Editor-in-Chief Mark Hagland on Tuesday at RSNA, regarding his current activities, and the buzz at the conference. Below are excerpts from that interview.
Tell me just a bit about the activity taking place in the Machine Learning Showcase area, that you were involved in yesterday?
What we were doing was that the HIMSS SIMM Enterprise Imaging Workshop [co-sponsored by SIIM and by the Chicago-based Healthcare Information and Management Systems Society (HIMSS)], was giving an update on our work, at the Machine Learning Showcase. We set that group up nearly three years ago, and it's exploded with popularity. We've produced a number of white papers that have appeared in the Journal of Digital Imaging. And we continue to pursue that work.
What’s more, our workgroup meets three times every year in person, at SIMM, at HIMSS, and at RSNA. We’re pursuing a number of initiatives, through that work. There are two areas that I think are particularly groundbreaking. The first involves developing an international maturity model to describe and frame progress in enterprise imaging development. In the workgroup, we talk about what radiology had done, what cardiology had done. We've been laying the foundation for that model. With this maturity model, we're also working with the European Society of Radiology and also across the world.
The second thing we’ve been very actively involved in is that we’ve been working with standards on the issues around image exchange from across medical specialties—not only radiology, but also endoscopy, ophthalmology, etc. So we've been developing some case vignettes to support our work. Here’s one example: imagine that your organization is treating a patient with severe burns; you want to send radiology images to the appropriate clinicians, but also the dermatological images of the skin burns, to the burn center when the patient is being transferred. And do you send all images of all types? So this is a very exciting space to be working in collaboratively right now.
More broadly, there’s a lot of excitement—and also some trepidation—around the potential for leveraging artificial intelligence and machine learning to improve the efficiency and effectiveness of radiological practice, as the U.S. healthcare system shifts away from a fee-for-service-based payment system and towards one based on payment for value. How do you see things going forward, in terms of radiologists and radiological practice?
I would have to say that I have both an optimistic view or vision of the future, and a pessimistic view or vision. Let me discuss the optimistic view first. So, if you look at the macroeconomic reality that we’re faced with, in terms of the percentage of GDP [gross domestic product] spent on healthcare, the fact is that there are a number of things that we in the U.S. healthcare industry have been experimenting within, in the context of this march towards value, including experimenting with risk-based contracts. And that work, including the development of accountable care organizations, has had an impact on the cost curve; but it hasn’t had a paradigm shift-level impact yet. The optimistic part of me says that machine learning could provide us with a once-in-a-generation type of breakthrough that could dramatically increase our efficiency in new ways, as physicians, including as radiologists, but also across all the specialties. The reality is that physician offices and hospitals lack the resources to change a lot of what we do; but if machine learning ended up living up to its hype, it might truly provide a breakthrough in improved efficiency and effectiveness that we haven’t seen happen yet.
And your pessimistic view of things?
Yes, now for my pessimistic view—which is that machine learning is predictably following the pattern of the Gartner hype cycle. And people don’t understand what machine learning can do and not do. Eliot Siegel [Eliot Siegel, M.D., who is the chief of imaging at the VA Maryland Healthcare System, vice chair of radiology at the University of Maryland School of Medicine, an adjunct professor of computer science at the University of Maryland-College Park], whom I have tremendous respect for, makes a distinction between narrow artificial intelligence and broad artificial intelligence. He thinks that AI will help us physicians to accomplish some specific things, but doesn’t believe that it will make very broad change possible. And there's an awful lot of hype and fear. So the pessimistic side of me says we're not moving forward fast enough, and still have an almost existential scenario in healthcare, in terms of what needs to happen, and the gap between that and what we’re accomplishing.