“We’re here to help make health care better,” said host Lee Huntsman at the February edition of Under the Boughs. “But boy is it complicated. It’s changing before our eyes in ways that are unclear and anxiety-producing.”
The topic of the program on Feb. 23 reflected this. Joining Huntsman on stage were Scott Parris, co-founder and CEO of Synapse Medical Intelligence, and Robbie Cape, co-founder and CEO of 98point6. The two have a deep history in both artificial intelligence and health care that they were able to draw upon to help attendees understand the history and future implications of artificial intelligence—by everyone’s admission a very broad topic—for health care.
Video highlights of the event are available below:
The two opened by talking about what had brought them to the space. Parris described the time he spent in a Buddhist monastery that helped nurture his interest in understanding how the mind works. Cape talked about how his company 98point6 reflects his family roots in entrepreneurship and his personal passion for helping families achieve better health.
Both panelists agreed that AI’s potential in health care is related to two core capabilities:
- The ability to leverage many different technologies that can be brought together to create something that might be called intelligence – including robotics, natural language processing, machine learning and automated reasoning; and
- The ability to rapidly process extremely large amounts of data, even complex data sets, using both deterministic and stochastic models to arrive at decisions.
The Role of AI in Health Care
Huntsman raised the question of whether or not the now well-established tools of AI have created an inflection point for health care. “The McKinsey Global Institute looked at the potential of automation across 12-15 industries and over 2,000 tasks,” said Parris. “They said 51 percent of the total 2,000 tasks across those industries are subject to automation. Within health care, their judgment was that 36 percent of the tasks and activities within health care could be automated.”
“Even if you argue that those numbers are plus or minus a little bit, those are still huge numbers, with huge economics behind it, and they’re not looking out to 2050 […] they are looking at today, when these things can actually be done,” he said.
Parris raised the role of AI to help make decisions against specific fact patterns and patient profiles, to collect better data from patients, and to attach economics to clinical treatments, therapies and outcomes. He also raised the potential impact of a platform for researchers that contained executable knowledge.
Cape described the value of taking the practice of medicine, and the scientific observation underlying that, and expanding the number of accessible data points required to make better decisions from dozens or possibly hundreds to an available pool of data points that numbers in the billions. He also said that the same benefits can now accrue to any plan administrator, payor, self-insured employer, or medical professional. “By virtue of allowing computers to begin thinking and processing information the same way a doctor does, we can dramatically lower cost, make care come faster and deliver it across geographic lines.”
Barriers to Adoption
Parris and Cape differed on whether or not health care is ready to adopt AI. “Health care is different than routing trucks across country or training a computer to assemble a car,” said Cape. “I think it’s because of the stakes. They’re high.”
“The bar we have to achieve around quality in order […] to deploy it commercially is higher than it’s ever been with any technology adoption,” Cape said. He added that for all of the players in the ecosystem, technology clearly has a role to play, to help everyone get more efficient.
Parris added to this thought by saying “If you show up with something that really is obviously and arguably better, you start using it, I think that’s the history of technology adoption.”
“The question is: Does AI have something that can tee up into this arena?” Parris pointed to the lack of clarity behind a “black box” approach to AI, where the computer’s decision-making process and models are not accessible. He suggested that a “clear box” approach where the reasoning behind conclusions is explicit and the research base that it comes from—as well as why it has been applied to an individual patient—can make a real difference in terms of adoption.
Parris said that the ecosystem plays a role in the lack of adoption, pointing to the need for incentive alignment and an economic lens to examine clinical actions as part of a systemic move from fee-for-service to a value-based approach.
Disrupt or Evolve?
To conclude the panel discussion portion of the event, the three turned to the question of whether or not AI will actually disrupt existing business models.
“I’m not a big believer in disruption [in health care], I think it implies that there are winners and there are losers, and I do not believe that that’s the way that it needs to happen, should happen or will happen,” said Cape. “The ecosystem is not working as well as it needs to work,” he said. “All of the layers in the ecosystem have an important role to play. This isn’t like Uber, where you look at it and you say, ‘There is no role for the taxi driver.’”
“The changes that will happen in health care need to be evolutionary, by virtue of the stakes that we are dealing with and the complexity of the ecosystem,” said Cape.
Parris raised the issue of time horizon in his response. “I think expectations at a one or two year level […] are not there,” he said. “There is an opportunity to create substantial value from AI, but I think it will be a co-evolution, and I think it will pick up steam as it goes forward.”
Audience questions revolved around the challenges of adoption and the best uses of AI being deployed now. “I would continue to encourage you all to continually pay attention to the human dimensions of adoption, it’s a really big deal,” said Huntsman. He pointed to the local work happening with the SCOAP project for surgeons in the state of Washington, which he said has been phenomenally successful building the collegial and professional expectations and assistance necessary to support successful adoption in health care.
When asked about some of the best or boldest examples of using AI in health care, the panel pointed to automated EKG interpretation and radiology imaging. Cape said that in both of those instances, clinical decision support is another notable application of AI that is already contributing value.
98point6 is currently recruiting members to participate in their clinical study. For more information visit the Cambia Grove blog.