Records Travel, But They Don’t Translate: The Interoperability Shift Nobody Noticed

For two decades, interoperability was the slide nobody believed. Every conference, the same data silos, the same promise that this time it would be different. So when our guest told us he can now pull back 85 to 90% of a patient’s actual record from almost anywhere in the country, the right reaction was skepticism. Then he showed his work.

The shift nobody noticed

Mika Newton, CEO of xCures, traces it to a quiet regulatory turn. Driven by the 21st Century Cures Act moving from “you may exchange data” to “you must,” provider participation went from roughly 30% in 2022 to 85–90% by 2025. The pipes — FHIR, document exchange, the national networks — finally carry near-complete records. As Leon put it on the episode, the capabilities of EHR data aggregation changed fundamentally in two years, and almost no one noticed.

Transport was the easy part

Here’s Newton’s core line: “Records travel, but they don’t translate.” A complete record arrives as a bundle of prose, scanned PDFs, and duplicated notes. His own example: he moved care from Kaiser to John Muir, brought his records, and they were scanned in as images. Now a national query returns his history twice — once as structured data, once as pictures embedded in another system’s output. “It just keeps stacking up.” The first job on connecting to a network is to normalize, deduplicate, and convert to FHIR. The rest is meaning.

Being honest about the 80%

xCures reports about 90% document retrieval and about 80% raw parsing accuracy. We pushed on that 20%. Newton’s reframe was the most useful idea in the hour: stop measuring how much of the record you parsed, and start measuring whether you can answer a specific clinical question, consistently. On those defined “checklists,” the bar is 95%-plus, with published results — and a hard rule that you must always keep a path back to the source document. His warning about the alternative stuck with us: “If you say trust but verify, and you go back and verify every single time, you haven’t moved the problem down the street.”

The patient becomes the access point

The deeper shift is about who the customer is. With TEFCA’s Individual Access Services and identity verified through something like CLEAR, a patient can now pull their own complete record and point an AI agent at it. “For the first time,” Newton said, “I see patients having a voice in healthcare, because they are the mediators of their data.” That reopens the question the field keeps dodging — who captures the value — and Newton’s answer is refreshingly blunt: not the patient, in dollars. What patients should want is the service, and clarity about who’s in the room.

Moving data used to be the hard part. It isn’t anymore. Making it mean something is.

Listen to the full conversation: https://practicalaiinhealthcare.com/episodes/#newton

You Might Also Enjoy

• S1E43 with Isaac “Zak” Kohane, MD, PhD — Who controls healthcare data in the AI age, and how SMART on FHIR quietly freed it: the standards backdrop to Newton’s “records finally move” story.
• S1E27 with Charlie Harp — Healthcare data quality and the PIQI framework: the deep version of Newton’s “transport solved, meaning didn’t” problem.
• S1E41 with Hugo Campos — Patient-directed AI and building OpenKP for personal data access: the patient-as-access-point shift, told from the patient’s side.