Accelerating the integration of digital healthcare tools in a responsible, clinically meaningful way is the primary focus of the American College of Cardiology’s innovation agenda, said John Rumsfeld, MD, PhD, the college’s chief innovation officer.
To accomplish this, Rumsfeld said stakeholders in the cardiology industry have offered to collaborate with technology companies more than ever before in the hopes of robustly but quickly validating new technologies that can improve the cost and effectiveness of care. Rumsfeld coauthored the ACC’s 2017 Roadmap for Innovation, which was published in November and generated interest from tech leaders who are eager to join forces with cardiologists.
“[T]o make this work, rather than being passive participants in this technology transformation like we were with the implementation of the electronic health records … we want to help lead the digital transformation of healthcare,” said Rumsfeld, also a professor of medicine-cardiology at the University of Colorado. “We know that there is risk involved in this in that not all technologies will be successful. But I think that unless we guide these technologies to the right clinical settings and how to use them optimally in the right clinical settings … we’re still going to have the same healthcare system delivery problems we’ve had for the last 50 years.”
In an extensive interview with Cardiovascular Business, Rumsfeld divided digital healthcare into three phases: virtual care, remote monitoring and artificial intelligence (AI)-driven care. He doesn’t consider the phases incremental steps but rather “parallel bowling lanes,” each with their own specific challenges which can be tackled concurrently.
Rumsfeld said he first heard of these phases from Kamal Jethwani, the senior director for connected health innovation at Partners Healthcare, and finds them useful in thinking about different ways technology can transform healthcare.
1. Virtual care
This phase is “ready for primetime,” Rumsfeld said, but only if payment models align.
Virtual care takes what physicians have already been doing during in-person visits and allows them to serve patients in another location via virtual visits. This can improve access for patients in rural areas and increase convenience for both providers and patients.
Researchers have already demonstrated telehealth can be useful in managing type 1 diabetes, and a recent Tampa Bay Times article highlighted how adding walk-in care kiosks to grocery stores enables patients to quickly see a physician for simple health questions.
Rumsfeld said he’s aware of payers reimbursing health systems for virtual visits to the same extent as they do for in-person care, but this remains uncommon.
“Right now, in many settings in the U.S., productivity is judged by how many in-person visits you are conducting,” Rumsfeld said, “so until that paradigm shifts that those interactions with patients could also be virtual and count toward your productivity, we’re not going to get very far.”
Rumsfeld said there are certain examinations that are better to complete in the clinic, but added telehealth is an effective solution for chronic disease management and follow-up care.
2. Remote monitoring
Like virtual health, the technology already exists for this phase and payment models are a hurdle to wider implementation. But there are additional barriers to overcome, too.
For one, wearable technology and biosensors can contribute to data overload for clinicians, Rumsfeld said.
“It doesn’t particularly help doctors and nurses to have 24-hour feedback on patients’ physiologic parameters unless we know what to do with them to change their care and to improve their outcomes. A lot of these new technologies want to feed back all this information, but we don’t have a way to receive it or turn it into (actionable) information.”
Another challenge is integrating all that data into a provider’s regular workflow. For example, if the information filtered to a separate website outside of the electronic health record, it would be a logistical headache to identify and incorporate relevant pieces of information into clinical decision-making.
And finally, even though the possibilities of remote monitoring are exciting, its value hasn’t been demonstrated yet.
“Most of the studies of remote monitoring to date haven’t been positive,” Rumsfeld said. “In other words, even when we are armed with extra data from home, for example, on our patients, it hasn’t necessarily led to care decisions or management decisions that have translated into better outcomes for the patients or to more cost-effective care. But that doesn’t mean we’re giving up on the concept.”
3. AI-driven care
Even though the technology also exists for AI-driven care, it is the furthest away from widespread adoption in clinical practice, Rumsfeld said.
It has the same challenges as the other two phases—aligning payment incentives, deriving meaningful information and showing evidence to support its use—but adds another layer of complexity by requiring interpretation of AI.
“An inherent thing about artificial intelligence is that the computer is evaluating all possible associations and interactions and doing this pattern recognition that’s beyond the comprehension of the human brain and because of that, it can be difficult to interpret the results for us,” Rumsfeld said. “If it’s outside of healthcare and you’re doing something like recommending a new movie or a new book or another product, I think we’re willing to take the advice of the artificial intelligence.
“In healthcare, the stakes are much higher. We rightfully require an understanding of why artificial intelligence is recognizing an image a certain way or predicting a certain outcome for a patient or recommending a specific therapy. We’re going to have a clinical skepticism about that unless we can understand how it’s coming to that recommendation to a certain degree.”
Another important point, Rumsfeld noted, is algorithmic outputs are only as good as their underlying data.
Because images are relatively easy for a physician to interpret and there is a wealth of good data in that area, Rumsfeld believes AI applications for cardiac imaging are likely to be among the first success stories in this phase.
“If you have an artificial intelligence-driven interpretation of an image, it’s still very easy for us as clinicians to compare it to the actual image and understand how close the AI is,” he said.
On the other hand, AI for risk prediction or recommending therapies is tricky because the underlying data is worse and the mechanisms for how the algorithms come up with those results are less easily understood.
“I think we need to have a high bar of evidence and be very wary of unintended consequences before we adopt them in clinical practice, and at this point that still seems far away,” Rumsfeld said.
Different burdens of proof
Even though Rumsfeld preached caution with the implementation of some AI-driven platforms, he said the goal of the ACC’s partnership with the technology industry, academic research centers and other stakeholders is to streamline development and testing of digital healthcare solutions.
For remote monitoring especially, he said some research is necessary to prove the practical utility of products, but wearable devices and biosensors shouldn’t be held to the same rigorous testing standard as medications or medical devices that may be implanted in the body.
“Rapid-cycle testing is a major focus” of the ACC’s innovation agenda, he said, adding that technology companies need their medical partners at least as much as the medical industry needs them.
“They are otherwise a little bit isolated to show the clinical utility of these tools and they also are often under quite a bit of time pressure, because with a certain amount of capital from venture capital or seed funding they only have a certain amount of time to develop a product and show it can work,” Rumsfeld said. “Traditional academic research cycles are probably going to be too long for how much time they have, especially for startup companies, so they really are extremely interested in creating relationships where we can more rapidly and efficiently evaluate these things in clinical practice.”
Rumsfeld predicted the deep-pocketed companies such as Google, Facebook, Amazon and Apple would be the most likely to disrupt the healthcare industry as much for their human resources as for their capital. When they acquire smaller companies, he pointed out, it’s as much to accumulate the brainpower of those employees as for the products they assume.
Regardless of how it happens, Rumsfeld said the ACC and some other early-moving professional medical societies are committed to driving the digital healthcare transformation—not being passengers.
“It’s easy to see in the healthcare environment that something has to change,” he said. “The continued rise of cost of care, the continued recognition of unexplained variation of care and waste in the system has not gotten better despite the availability of technology.
“We think part of that reason is there hasn’t been this overt bridge … between the clinical enterprise and the technology enterprise. If we can cross that bridge, we do think care could be more efficient and transform how care is delivered in a way that’s good for patients and good for clinicians in the health system.”