Out of the FHIR Podcast

About Out of the FHIR Podcast

Deep conversations with Healthcare Data & AI experts, implementers, and healthcare innovators. Get practical insights, real-world implementation stories, and the latest developments in healthcare interoperability from industry leaders. evestel.substack.com (https://evestel.substack.com?utm_medium=podcast)

Publisher

Gene Vestel

Category

health & fitness · technology

Language

en

Explicit

No

First episode

23 Apr 2025

Latest episode

15 Oct 2026

Latest episodes

35 episodes in the feed.

1. 

   15 Oct 2026

   Medicaid Data Got Usable. Nobody Noticed.

   Medicaid rarely comes up in interoperability conversations. Mathematica’s Marie Copoulos and Paul Messino said so in the first five minutes, and the rest of the episode explained why that is a mistake. Marie came to Medicaid from Medicare and commercial data. Her point: ten years ago you could not analyze Medicaid across states. Today you can. The national data set is imperfect but usable, and most investors and builders still assume it is a black box. Thanks for reading FHIR IQ playbook! This post is public so feel free to share it. Paul spent eight years at Maryland Medicaid, on a COBOL system built to pay claims. His framing of the core problem: Medicaid grew up on transactional data designed for payment, and now states are being asked to measure outcomes from it. Three things worth your time 1. Wrap the mainframe, don’t replace it. CMS pays states an enhanced federal match to modernize their Medicaid Enterprise Systems and has pushed a modular approach for over a decade. Paul’s read on the results: mixed. Large vendors often sell the same monolith in modules. The approach that is working now is to leave claims processing alone and put a data model, analytics layer, and API or FHIR gateway on top. Benji added that commercial payers and Medicare contractors have hit the same wall: many rewrite attempts end with the core still on COBOL. 2. Health data’s most expensive habit. Marie’s sharpest line was about our own industry. Every organization rebuilds the same claims normalization pipeline from scratch. Her second point was about time horizons. Value-based contracts usually run a year. Chronic illness plays out over a decade and moves people between Medicaid, Medicare, and commercial plans. A ten-year contract would change how you think about attribution, churn, and prediction. 3. Local codes come from deadlines. When a legislature creates a new benefit and the standard code does not fit, a state program with a launch date will create a local code rather than wait on a balloting process. Marie’s caution on the fix: forcing everything into one standard strips out real program differences. The useful work is finding the 80 percent that can be standardized faithfully. On AI and patients We also covered the CMS ACCESS Model, a 10-year Innovation Center model that pays for outcomes from technology-enabled chronic care, starting July 2026. Paul named three gaps he hears from participants: reimbursement, data flow, and matching providers with the right digital tools. On Patient Access APIs, his assessment was blunt. States built them; apps and members have not shown up. Marie added that patients don’t make most care decisions alone. If the clinician has no time to look at app data, the app does not help much. Listen * Spotify: [link] (https://open.spotify.com/episode/1TTzYeZSo1s37fdHafrSd0?si=ZbObEQBBSf6qk2iI0gXVpQ) * Apple Podcasts: [link] * YouTube: [link] Out of the FHIR is independent and bootstrapped. If you work on health data infrastructure and have shipped something worth talking about, [reply to this post or email link]. Guests speak for themselves, not their employers. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit evestel.substack.com/subscribe (https://evestel.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2)

2. 

   9 Sept 2026

   Caregivers Are the Interoperability Layer: Brian K Fung on Building haau3

   I asked Brian K Fung a simple question: are you a one-man show, or do you have a team? It’s hard to say yes confidently, because I’m certainly taking advantage of a lot of AI assisted coding. I am, yes, one person, but AI changes that quite a bit with the amount of leverage it has. Brian is a pharmacist. He spent time at ONC. He now runs haau3, a product for family caregivers. What actually changed in December 2025 Brian started using Claude Code in April or May of this year. His description of the effect is blunt: it is very hard for him to just literally handwrite code anymore. But the more useful part is his account of why it became possible, and he puts a date on it. Something happened in December 2025 that completely changed the paradigm of AI coding. It wasn’t like you ask it something, it returns a response, back and forth, back and forth. They created a loop. The whole agentic looping of some sort of goal. Let’s say, write me a FHIR API. Well, that’s a goal. And then it would just keep looping and looping until it reaches that goal. The consequence he cares about is not speed. It is that the thing runs while he is not there. You can tell it to do that and then walk away, and it will just code for him. He is careful about the limit: it is not always right, which is where technical background or domain expertise helps. That caveat is the whole reason a pharmacist building for caregivers has an advantage over a generalist engineer building the same thing. He now teaches this to people with no coding background at all: dentists, nurses, physicians and pharmacists. The cost curve, stated plainly This is the part I have not seen anyone write down honestly, so here it is. Brian started on pay-as-you-go and burned through twenty dollars doing the most simple things within an hour or two. He moved to the twenty dollar subscription and hit the limits where they stop letting you work. He has settled on the hundred dollar plan. I think that might be too much, but hundred dollars might be the sweet spot. He has gone to two hundred when a deadline demanded it, and thinks that was more than he needed. If you are trying to budget this for a team, that is a more useful data point than any vendor page. The warning underneath the enthusiasm Both of us were sitting there enthusing about how fun this is. I described building at night while my wife watches her show. Brian agreed, and then did something I respect: he immediately turned it over. The spicy part is I am starting to hear some different podcasts out there that talk of other users that use it extensively. And their usage has begun to lead to burnout. And then the part that landed: You wake up and you just want to code. If you’re not coding, you feel like you’re not productive. He is not claiming he has reached that point. He is saying the vocabulary other people use to describe burnout matches the vocabulary he uses to describe having fun, and that is worth noticing early rather than late. Are we all about to get priced out Benji pushed on the economics. AI companies are operating at historic losses. That looks a lot like a familiar pattern: get the users first, worry about unit economics later. Brian analogy is the right one. Airbnb, Lyft and Uber Eats were cheap during a period when money was plentiful and interest rates were low. He does not really look at Airbnbs anymore. Those costs were heavily subsidized. I would not be surprised if they bump the prices of these things so astronomical that it is just not reasonable for us to use. His hedge is to keep his own ability to do the work by hand. He admits this is getting harder, for a reason familiar to anyone shipping fast: the codebases are growing large enough that he cannot always remember how something worked or where he put it. Does AI make standards obsolete This comes up every time now, and Benji had the cleanest rebuttal I have heard. The scare version goes like this: someone demonstrates two conversational AIs exchanging a patient’s clinical history in plain language, no formats, no implementation guide, and people conclude that standards are finished. AI is not here to replace FHIR or other data standards. It’s here to leverage those standards to make things even better, faster, more efficient. His argument is that models get more accurate when there is a deterministic data model underneath them, and that CQL is a good example precisely because it is a simple declarative language with one answer. Brian agreed on the principle and then complicated it, which is why he is a good guest. I’ve grown to believe that you will never get to a zero percent chance of hallucination. We always need to think about how we ground or make it as close to deterministic as possible. He then flagged a recent and contested paper suggesting that general frontier models outperformed grounded clinical tools on some benchmark. He was scrupulous about it: he has not dug into the methods. I am repeating it with the same caveat, because it is the kind of finding that gets stripped of its qualifiers the moment it hits a feed. My own take: we have run this argument before. When agent skills arrived, people declared MCP unnecessary. A year of actual adoption later, nobody seriously argues that. Standards do not die when a capability improves. They move. Why pharmacy never got wired up You can call an Uber. You can book an Airbnb. You cannot do the equivalent with your pharmacy, or with your own medication history. Why is pharmacy so far behind? Brian answer is one word. Incentives. Especially in America, we drive on incentives, payments, something like that. And one of the biggest payments in the last decade was the Meaningful Use, Affordable Care Act, HITECH Act, all those kinds of things. Lots of money, but it focused on providers and hospitals. What was left out, pharmacies. He watched the consequences directly. In 2020 he was at ONC reviewing public comments on pharmacy interoperability. All these mom and pop shops, these independent pharmacies saying you can’t enforce these standards because we don’t have the money for it. That is the whole story. Not a technical failure. A funding decision that was made once and then compounded for a decade. He is more optimistic now, because pharmacy is a named category in the CMS Health Tech Ecosystem. Benji drew the parallel to rural health, where a very large transformation fund is currently being divided among the states, and asked the obvious follow-up: is there going to be an equivalent for pharmacy? Nobody knows yet. It is the right question. Two things make the omission worse than it sounds. Pharmacists are the most accessible clinician in the country. As Brian puts it, you do not need an appointment, you just walk in. During the pandemic he estimates pharmacists were responsible for more than half of vaccinations. That physical network exists, it is staffed, and it is largely disconnected from the record. And on my side of it: I spent five years at Express Scripts, and medication reconciliation has been a Stars measure for as long as I can remember. Pharmacists get paid to call patients and do it. The patient never sees any of it. Why do payers have 150 person quality departments where nobody is actually doing anything about quality? They’re just abstracting charts. The fix here is not exotic. Send the discharge summary to the pharmacist. The patient is going to walk in for a refill anyway. Do the reconciliation there, in a conversation, instead of waiting for an annual visit that may not happen. The line I keep coming back to Brian closed with the framing that made the whole episode click, and it reframes caregiving as infrastructure. There are roughly 63 million family caregivers in the United States. What are they actually doing, mechanically? We think that caregivers are essentially serving as the interoperability layer on behalf of their loved ones. They advocate here. They take the appointment there. They carry the history in their heads between two systems that will not talk to each other. They are performing, unpaid and by hand, the exact function this industry has spent twenty years and enormous sums failing to automate. Once you hear it that way, haau3 stops looking like a caregiving app and starts looking like an interoperability product with an unusually clear-eyed view of who the user is. Their roadmap is public at haau3.com/roadmap, and the next big piece is scheduling, including scheduling on behalf of someone else. Which runs straight into the wall we always hit. I raised the USCDI comments where providers worry that publishing appointment availability lets competitors reroute their patients. Brian recognised the pattern immediately, because it is the same fear independent pharmacies have about sharing data with networks: those are your customers, and they might walk. That fear is not irrational. It is also the thing standing between 63 million people and a system that could stop making them do the work by hand. Where I landed Brian is one person with a pharmacy degree, a policy background, and a subscription. That combination produced a working product aimed at a real and enormous problem. The thing I want to sit with is the connection between the two halves of the conversation. The tooling that lets one domain expert build something real is arriving at exactly the moment we are admitting that the biggest interoperability gap in healthcare is being covered, for free, by exhausted family members. Chapters 00:00 Caregiver origin stories and Rosalynn Carter’s four types 02:12 One person, but AI changes the math 04:39 Teaching dentists and nurses to run Claude Code 06:40 Will an overnight run cost thousands in tokens 10:16 An addicting hobby you do instead of watching TV 12:12 The spicy flip side: coding toward burnout 15:53 Cheap Ubers and Airbnbs were subsidized too 21:00 The claim that AI killed standards, rebutted 22:33 Zero percent hallucination is never coming 28:40 Meaningful Use billions that skipped the pharmacies 30:22 Rural health, filing cabinets, and a very large cheque 36:20 The clinician you can see without an appointment 37:30 Every policy choice marginalizes somebody 40:15 Caregivers as the unpaid interoperability layer 42:06 USCDI, scheduling, and the fear of poached patients Brian K Fung is a pharmacist and the founder of haau3. Benji Graham co-hosted. If you are building in this space, or you are one of the 63 million, reply and tell me what you would want scheduling to actually do. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit evestel.substack.com/subscribe (https://evestel.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2)

3. 

   5 Sept 2026

   380 AI Use Cases and One New Request a Day: Inside Highmark’s Responsible AI Function

   Most organizations I talk to are trying to solve an AI adoption problem. Michael Barber has the opposite problem. He is Senior Director of Responsible AI at Highmark Health, and his group reviews every AI use case that goes to production across an enterprise that is a payer and a provider at the same time: six or seven million members in six states, Allegheny Health Network’s 16 hospitals, United Concordia Dental, and about 50,000 employees. Highmark approved its first AI use case in September 2022. As of the day we recorded, 380 are in production. One hundred and four of those, 30 percent of the total, landed in the last six months. We are now seeing a new request for an AI use case every day on average. So it’s one a day now. The role did not exist until the volume forced it AI governance at Highmark used to sit inside the data governance group. It outgrew that. Michael’s position, and the three positions under it, were created about two and a half years ago specifically because the volume of AI requests could no longer be absorbed as a side responsibility. That is the useful signal for anyone benchmarking their own org. The question is not whether you need a dedicated responsible AI function. The question is what request volume forces one, and Highmark’s answer was somewhere on the path to one a day. One design decision worth copying: when AI is involved in a use case, Michael’s group handles the data governance review too, rather than sending teams to two separate committees. Fewer gates, same standard. The wins are unglamorous, and that is the point Ask what AI is doing for a large payer and you will get a lot of answers about clinical decision support. Michael’s list starts somewhere else. Thousands of faxes a day, still arriving in 2026, read and turned into discrete data elements. Systems that pull the relevant medical policy and chart sections in front of a claims reviewer so nobody has to page through a thousand documents to find three that matter. And a hard line underneath all of it: I’ll say right up front we would never use AI as the sole determinant for any adverse determinations. It’s not what we do. It’s illegal. On the provider side, the durable wins are ambient listening and imaging triage. His framing of why: The way we view AI, especially in the clinical setting, is that it doesn’t replace human judgment. It simply allows people to work at the top of their license. He makes the value concrete. If a tool lifts the bottom quartile, usually newer employees still learning the landscape, up to the midpoint, that is real. It lets someone operate as if they had been there a year. Ambient listening is a standardization layer, not just a scribe This was the part I did not expect, and it is the part interoperability people should sit with. Highmark runs one ambient listening system across AHN. Michael’s argument is that its value is not only that clinicians stop typing. It is that hundreds of clinicians now go through one intermediary that behaves the same way every time, instead of each entering notes into the EMR their own way with their own abbreviations. Now we’ve got one thing that does something the same way every time in between hundreds of clinicians and our EMR system. He is realistic about why that matters. Earlier in the conversation he described trying to combine EMR data across 16 cancer infusion clinics: If you’ve seen one, you’ve seen one. Same Epic instance. Subtly different local workflows. Same fields used for different purposes. Anyone who has built quality measures or a provider data warehouse knows exactly what that costs downstream. If the standardization argument holds, ambient AI is a data quality intervention that happens to also save clinicians time. That reframes it from a documentation tool to infrastructure. Five thousand people complaining about the data My own view is that AI governance cannot happen without data governance, and at the same time it forces data governance to happen. Michael gave the best illustration of that I have heard: Five years ago we had my group of 26 data scientists and software engineers complaining about the data. Now we’ve got 5,000 people complaining about the data. Data quality problems used to be a specialist grievance that was easy to defer. Once you hand an LLM to 50,000 employees, bad data becomes everyone’s problem, visibly, immediately, and with executive attention attached. That is the forcing function a decade of data governance decks could not produce. Access is broad on purpose, and education is the actual work Highmark built an internal wrapper called Sidekick that sits on Gemini inside their secure environment and is available to all 50,000 employees. It includes a retrieval layer over their corporate policies, thousands of them, that returns a summary plus hyperlinks to the real policy. Paid Gemini Enterprise and M365 Copilot licenses sit on top of that, allocated at VP discretion. That is more open than most enterprises, which tend to ration licenses to engineers and a few analyst teams. But the access is not the interesting part. The scaffolding is: roughly 100 AI ambassadors embedded in business functions as super users, an AI center of excellence with consultants who help teams find a solution even when the answer is that they do not need AI, and a lot of time in departmental meetings and town halls. Michael’s summary of where that lands him: I’ve met with other companies and frequently the question is, how do we drive adoption? My problem is the opposite of that. How do we get people to calm down just a little bit? He also names the failure mode most enterprises are walking into right now: applying AI to a workflow that should not exist. Speeding up a broken process is not the win. The win is asking what you would do if you did not have the process at all. The vendor position is stricter than the law requires This is the section I would send to anyone negotiating AI vendor contracts. Generative AI broke the old vendor model. A vendor used to own its model and run it in a known environment. Now, as Michael puts it, everything is an API call to somewhere else, and the work is figuring out what happens two or three layers down. Highmark does not allow vendors to train on its data. It does not allow Google or Microsoft access to its data. And it goes further than HIPAA requires: Even de-identified, we don’t allow vendors to improve their general product with our information. It has business value. His reasoning is commercial, not legal. Once data is de-identified under safe harbor or expert determination, HIPAA is out of the picture and vendors often assume they can do what they like. Michael’s position is that the data still has value, and if a vendor wants to use it to improve a product they sell to competitors, Highmark should get something back. Anything externally facing gets red teaming and prompt injection testing before it ships. Fifty states writing fifty AI laws The most pointed part of the conversation. Michael testified before two Pennsylvania house committees last spring while they were drafting the state’s AI act. They asked for a list of every AI use in the enterprise. I said, well, what are you going to do with that? At the time we had 330 use cases. I’m going to give you a list of 330 things, what are you going to do with it? What is the problem you’re trying to solve? He told them his list would look like everybody else’s, because everyone is doing the same things for the same reason: they work. Every hospital system in the state is using ambient listening. His objection is not to oversight. It is that roughly 90 percent of what state AI acts cover is already federal law, and the genuinely new part is reporting. Maryland and New York are asking for enterprise-wide inventories, including what data systems were trained on. That is administrative cost added to an industry that is supposed to be reducing cost, in exchange for a document nobody reads. Pennsylvania removed the reporting requirement from its draft bill and replaced it with something closer to what he proposed: confirm that the entity has a governance process, rather than collect a list. He called that a big win, and the bill has not reached the House floor yet. He also lived through the version where this gets expensive. Under the original Colorado AI Act, Highmark had to carve Colorado members out of a program making positive outreach to people being discharged from hospitals. Colorado residents did not get the benefit. Colorado has since retreated from that position and those members are back in. The line I keep thinking about Benji asked whether legislators simply do not understand the technology. Michael’s answer: If you’re looking for votes, it’s a nice thing to say. We’re going to protect you against the big bad AI. I said, I wouldn’t want to go to a doctor who is anti-AI. Would you? I pushed back on some of this in the episode, because I think there is a real category of harm worth regulating. California’s law focused on adverse outcomes, including what chatbots tell teenagers in crisis, seems to me like the right shape: regulate the outcome, monitor the harm, do not demand an inventory. Michael’s answer on that was practical. Every externally facing use case at Highmark gets tested for exactly those scenarios. What happens if someone says they are having chest pain. What happens if an employee asks Sidekick something in a bad moment. His observation is that the frontier labs have already built a lot of that in. The last time he tested it, Sidekick not only declined to give the wrong kind of advice, it surfaced the employee assistance services actually available to Highmark employees. Nobody programmed that. Where I landed Two things I am taking from this conversation. First, the governance story and the data quality story are the same story. Highmark did not fix its data and then deploy AI. It deployed AI broadly, which made the data problems visible to 5,000 people instead of 26, which is now driving the data work. That is backwards from how most of us have been taught to sequence it, and it appears to be working. Second, the patient is still the last to see the benefit. I said this to Michael and I will keep saying it. I have spent a long time in this industry and the person whose data it is still retypes their history into a 40 page form. Michael’s answer, that consumer expectations set the bar and healthcare has to at least stay in the game, is the right instinct. The gas pump screen knows where you are going on vacation. The scheduling system still does not know you had that colonoscopy. That gap is the work. Chapters 00:00 Welcome, and how we met in Pittsburgh 01:04 The Responsible AI job that did not exist 02:42 Where AI is actually paying off at a payer 04:14 Never the sole determinant, and the fax problem 05:29 The FHIR guy pulls a thread on structured data 08:47 Working at the top of your license 12:11 Ambient listening as a standardization layer 15:36 From 26 people complaining about data to 5,000 17:20 Sidekick, and giving an LLM to 50,000 employees 20:55 380 use cases, and one new request every day 27:03 What does the patient actually get out of this 32:47 Guardrails, vendor contracts, and your data 38:31 Fifty states writing fifty AI laws 42:16 “I wouldn’t want to go to a doctor who is anti-AI” Michael Barber is Senior Director of Responsible AI at Highmark Health, where he leads the group that reviews and approves every AI use case that reaches production across the enterprise. Benji Graham co-hosted. If you are working on AI governance inside a payer or health system and want to compare notes, reply to this email or find me on LinkedIn. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit evestel.substack.com/subscribe (https://evestel.substack.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2)

Who has been a guest on Out of the FHIR Podcast

Names identified in recent episode analyses. Showing up to five guests.

• Cody Ebberson
• Darren Devitt
• Grahame Grieve

Reach and audience

Public platform figures. Ratings count people who left a rating, not total listeners.

YouTube views

1,712

Score snapshot 15 Sept 2025

Apple Podcasts (US)

5.0 / 5

1 ratings

Podcast Authority Score: 29 / 100

A composite of feed quality, social presence, YouTube performance and engagement. Read the methodology.

Quality

32

Social presence

0

YouTube

68

Engagement

0

Contact Out of the FHIR Podcast

Guest appearances

Books guests

Based on episode analysis; this does not confirm that the show is currently accepting guests.