Healthcare has spent a decade trying to solve the interoperability problem to move patient data cleanly across systems, settings and specialties, and has made uneven progress. For health systems deploying AI at scale, that uneven progress is no longer just a care coordination concern. It has become a constraint on what AI can do.
At Rochester, Minn.-based Mayo Clinic, Chief Digital Officer Nari Gopala sees the problem concentrated at three specific moments in every patient journey: when a patient arrives via referral, when a patient leaves at discharge, and when care moves into the home entirely. At all three, clinical context breaks down and fixing the problem has become the precondition for AI to work.
“We have talked about data interoperability for a decade now at least, but we still haven’t truly achieved it,” Gopala said during a “Becker’s Healthcare Podcast” interview.
The first gap is on arrival. Patients referred to Mayo come from different EHR systems, different hospital networks and outside practices. The records they carry are fragmented across those sources and never formatted the same way. The specialist who sees that patient for the first time needs to know what is clinically relevant, not everything that was ever documented.
“When you’re thinking about the context of all of the data and the clinical information that needs to reach the specialist the moment the patient walks in, that’s where we use AI to sort through all of those outside records and create a pre-visit diagnostic summary,” Mr. Gopala said.
At a multispecialty system, the complexity compounds. A patient moving through cardiology, pharmacy and surgery in a single encounter is being seen by clinicians who share the same record but need different information from it. A record designed around documentation was not built to serve that kind of differentiated clinical attention.
“The idea is that when the patient sits in the center, and then you have the cardiologist, the heart failure specialist, the surgeon, the pharmacist, all of them should not be looking at different tabs of the EHR, but instead should be getting a perspective that is exactly relevant to them, based on the reason they’re looking at that patient,” Mr. Gopala said.
What each of those clinicians receives from an AI tool depends entirely on the context it has to work with. Mr. Gopala says that dependency makes context more critical now than at any earlier point in the technology’s history.
“Without that context, what you get out of any of the AI models today will be very different,” he said.
The second gap is at discharge. After Mayo treats a patient, that patient typically returns to a primary care physician in their home community. The physician needs actionable data and a complete picture of the care their patient received at Mayo Clinic.
“When I actually look at the discharge summary itself, a lot of times I’m actually looking at it as a history book,” Gopala said. “It is almost a book that says, here’s what happened historically, and then I don’t know if this patient actually got better or not when they got discharged.”
The direction he is working toward is a discharge process built around two audiences who are currently underserved by it: the patient who needs to understand what was treated and what comes next, and the primary care physician who needs enough context to continue care without triggering another major intervention. AI is powerful enough to bring that information together and Mayo Clinic is developing a tool to bridge that gap.
The third gap is forming in real time. Patients are increasingly receiving care at home rather than at a clinical facility, a shift that is already visible at Mayo in the form of home chemotherapy. The logistics of making that work require capabilities that most health systems, including Mayo, do not currently have on their own.
“We have proven that it can work at a smaller scale,” Mr. Gopala said, but moving home care to any kind of meaningful volume requires outside capability. “We need partners who have the logistical chops, who have the fulfillment chops, who can get even rural care sort of a thing to work for this.”
Mayo has the clinical model; it does not yet have the operational infrastructure to run it at scale, and the system’s executive team is actively seeking partners.
“The hopes of the future is that we will find those right partners that can help scale it,” Mr. Gopala said. “We do believe, as technology evolves, there will be more care done [outside the hospital].”
The post The hard problem AI can’t fix at Mayo Clinic – yet appeared first on Becker's Hospital Review | Healthcare News & Analysis.