Episode #160. The Hidden Cost Of Siloed Data And How AI Finally Breaks It Open, with Doug Fridsma, CMIO at Health Universe

Summary

---

Risk Never Sleeps Podcast – Episode #160

Title: The Hidden Cost Of Siloed Data And How AI Finally Breaks It Open
Host: Ed Gaudet (with Saul Marquez, Outcomes Rocket)
Guest: Dr. Doug Fridsma, CMIO at Health Universe
Date: December 11, 2025

Episode Overview

In this engaging episode, Ed Gaudet and Saul Marquez sit down with Dr. Doug Fridsma—an accomplished physician, computer scientist, and digital health thought leader—to discuss how siloed healthcare data stifles innovation and the pivotal role of AI-powered platforms in transforming patient safety. Doug shares lessons from his influential tenure in government and the private sector, offering insight into building the technical and cultural bridges necessary for real data interoperability.

---

Key Discussion Points and Insights

1. Doug Fridsma’s Background and Legacy

• Internal medicine doctor and PhD in computer science.

• Chief Science Officer at the Office of the National Coordinator (ONC) during the Obama administration, instrumental in deploying electronic health records (EHRs) across the U.S.

• Former CEO of the American Medical Informatics Association, updated board certifications for clinical informatics.

  "When we started at ONC, 20% of all the doctors had electronic health records. When we ended...80%." (Doug, 02:29)

  “Our hope when I was in the government is that we would have all of this data that would drive innovation. We never really anticipated that we'd have three or four vendors that would really have a lock on the data.” (Doug, 03:25)

2. Evolution of Health Data Infrastructure

• Movement from paper-based to digital health records—unmatched scale and speed over five years.

• The current issue: major vendors “lock” the data, restraining the vibrant ecosystem of apps and tools once envisioned.

  “We hate lazy data. Data that just sits there.” (Doug, 03:54)

3. The Next Layer: How AI Unlocks Data

• The future of healthcare is “agentic and AI-driven”—enabling smarter, more flexible, and efficient systems.

• Doug draws parallels to the 1950s mainframe era: just like computers evolved from mainframes to smartphones, health data must move beyond siloed EHRs to flexible, app-based platforms.

  “We’ve got to move out of these mainframes in the sky...make it easy to deploy these things and actually start to use them in real life.” (Doug, 05:02)

• Health Universe aims to serve as the “operating system for AI in healthcare”, facilitating integration, normalization, and real-time utility of disparate healthcare data.

  “Health Universe in large part is intended to be an operating system...it helps to coordinate agents, does data normalization, makes it really easy...plug[s] into EHRs and HIEs.” (Doug, 05:03)

4. Platform Design and Innovation Enablement

• Health Universe is built on Python and leverages agentic AI, using orchestration to sync data and tools contextually (06:16-06:45).

• Capabilities include scanning and unlocking data even from unstructured sources (PDFs, images) using OCR (Optical Character Recognition).

  “We do things like customize summarizations for cancer. And we can pull out everything from scanned images. We have our own pipeline for OCR because data matters. It’s garbage in, garbage out.” (Doug, 06:49)

• The platform serves both large institutions and entrepreneurial developers: “horizontal,” not “vertical,” so one integration supports many solutions.

• Co-development partnerships help innovators focus on expertise while Health Universe provides infrastructure.

  “So many of these AI solutions are vertically integrated...it becomes overwhelming. So our platform is a horizontal one that allows you to integrate once and deploy many.” (Doug, 08:00)

5. Breaking Open the Silos Through Standards and APIs

• Doug wrote and implemented critical standards (e.g., Consolidated CDA, SMART on FHIR) underpinning modern health information exchanges and interoperability.

  "A lot of those standards I actually developed when I was at ONC...We use that in the HIEs...to pull [data] out." (Doug, 10:51)

• Emphasizes the difference smarter interfaces and APIs make in bypassing vendor “locks.”

  “Well, all the problems associated with...data in healthcare—Doug’s fault.” (Ed, joking, 11:33)

6. Patient-Centered Perspective & History Analogy

• Identifies the lack of patient representation in health AI forums.

• Shares a compelling analogy: early medical use of automobiles was focused on doctor benefit, but ultimately became transformative for patients, enabling them to access care differently.

  "I think we spend a lot of time with technology to try to make patients better, but I think we're going to end up making better patients." (Doug, 14:48)

---

Notable Quotes & Memorable Moments

• On the changing landscape of EHRs:

  “We’ve got like seven EHRs. And you know what happened after [the mainframe era]? We had personal computers and now everything runs on my iPhone and it’s all a bunch of apps.” (Doug, 04:12)

• On innovation bottlenecks:

  “[Institutions] have to wait two, two and a half years for good ideas to get through the gauntlet...the CIO says, if you can’t give me a seven figure return, we aren’t going to take a look at it...there’s a tremendous amount of bottom up innovation waiting to be released.” (Doug, 05:47)

• On the purpose of Health Universe:

  “We named our platform Health Universe because Health Solar System seemed a little small in terms of our aspirations.” (Doug, 10:31)

• Historical parallel:

  “Between 1905 and 1912...doctors were adopting this new technology...the automobile...to improve and transform healthcare. Actually [it] became a patient’s tool...doctors could have offices and patients could come to them...I think we're going to end up making better patients.” (Doug, 13:00–14:48)

---

Memorable Moments & Banter

• Playful callbacks to classic computer history and standards, e.g., reference to CORBA, Linux (08:53–09:03).

• Doug’s comedic timing:

  “My flight just got canceled.” (Doug, 10:03)

• Closing “lightning round” with personal and humorous responses:

  • Riskiest thing Doug has ever done?
    “Starting a company.” (Doug, 15:36)
  • Advice to his 20-year-old self?
    “Buy Amazon.” (Doug, 15:58)
  • Desert island albums might include: Lady Gaga, EDM, the Beatles. (Doug, 16:50)

---

Timestamps for Important Segments

| Timestamp | Topic / Segment Description | |-------------|----------------------------------------------------------------------| | 01:14 | Doug’s intro: medical/technical background & EHR federal rollout | | 02:29 | National shift to EHR adoption—impact during ONC tenure | | 03:54 | “Lazy data”—current problems with data silos and vendor lock-in | | 05:03 | Future vision: agentic, AI-driven healthcare platforms | | 06:49 | How AI + OCR extracts value from structured and unstructured data | | 08:00 | Platform approach: horizontal vs. vertical integration | | 10:31 | Why “Health Universe” and ambitions for broad impact | | 10:51 | The standards behind breaking silos (CDA, SMART on FHIR) | | 13:00–14:48 | Historical analogy: the car, patient empowerment, better patients | | 15:36–16:50 | Lightning round: risk, advice to young self, desert island albums |

---

How to Learn More or Connect

• Explore Health Universe:
  healthuniverse.com – Free accounts available for app deployment and experimentation

• Contact Doug Fridsma:
  Email: [first initial][last name]@healthuniverse.com

---

Tone & Style

The conversation is fast-paced, technical yet approachable, and peppered with humor and camaraderie. Doug is insightful, candid, and self-deprecating, while Ed and Saul keep the discussion lively and accessible, unafraid to poke fun at jargon and the quirks of health tech industry history.

---

Summary prepared for listeners and readers seeking an actionable, engaging distillation of this Risk Never Sleeps podcast episode.

Transcript

[A]

Foreign.

[B]

Welcome to Risk Never Sleeps, where we meet and get to know the people delivering patient care and protecting patient safety. I'm your host, Ed Gaudet. Welcome to the Aimed Insights podcast sponsored by Outcomes Rocket and Sense in at Risk Never Sleeps. And we're here with Doug Fitzma. Did I get that right or Fitzma? Yeah, that's one of the. Doug's a good man. We've already talked. He's promised this is going to be the best podcast I ever saw. And I'm here with my right hand man, Saul Marquez, Outcomes Rocket.

[C]

Such a pleasure to be here. And Doug, really excited to chat with you today.

[A]

Yeah, I'm looking forward to it.

[B]

And Doug tells us he's got an interesting background. Let's start there. Doug, you gotta wow us now. We've been wowed. Day one was a huge wow, so the bar's high. Day two, you're the first one. First one. Gods are a little angry this morning, so let's make it work.

[A]

I don't want to be the sacrificial virgin. For sure.

[B]

Oh, man. Wow.

[C]

Hello.

[B]

Already off to a good start here with Doug. All right, Doug, tell us a little bit about your background.

[A]

So I'm internal medicine doctor. No, I have a PhD in computer science and I worked for the Obama administration during high tech to deploy all the electronic health records in this country.

[B]

You're the guy. Thank you, Doug. Thank you for your service. Seriously.

[A]

I was the chief science officer there, so all the cool stuff.

[B]

Yeah.

[A]

Took care of that.

[B]

You made it happen.

[A]

Yeah.

[B]

You're the guy we need to thank for EMRs and EHRs and EPIC and Judy Faulkner.

[A]

Oh, yeah. Judy and I go way back. Do you. Oh yeah.

[B]

Oh yeah. She's great, isn't she?

[A]

Yeah.

[B]

Are you a science fiction reader? No, she loves science fiction books.

[A]

Oh, yeah.

[B]

Oh yeah. She's really into.

[A]

Yeah.

[B]

And she can dance. I've seen her at the His Dog Lollapalooza party. Dancing.

[A]

Yeah.

[C]

Oh, yeah.

[B]

With Jonathan Bush from Athena Health.

[A]

Yeah.

[B]

Can you imagine those two dancing?

[A]

Yeah.

[B]

That's crazy, man. That's crazy. That's when I used to drink. I don't drink anymore.

[A]

Yeah.

[B]

Four years, so.

[C]

For the better.

[B]

For the better soul. Yeah. Yeah, yeah. Because I probably.

[C]

You're crazy enough.

[B]

I'd be late right now. I wouldn't. I would have made the 8 o' clock this morning.

[C]

I'd be running this with you.

[B]

That's true. Never missed a meeting. Never in my life. All right, Doug, so what have you seen over those years that really just blew you away about healthcare.

[A]

When we started at ONC, 20% of all the doctors had electronic health records. When we ended, or at least when I left the government five years later, 80%. So we had taken the entire, took 20% basically of the largest GDP in the country and we flipped it in five years from paper to electronic. Now there's lots of work that still needs to be done and that's part of why we've got aimed here. But I think what I've seen is we had the adoption of electronic health records. Then I spent some time at the American Medical Informatics Association. I was the president and CEO there for five years and worked on developing an update to the board certification for clinical informatics and also trying to work on education because once you've got these new tools, it requires you to think differently about how you deliver care. Now, I will say one thing is that our hope when I was in the government is that we would have all of this data that would drive innovation. We'd have all these interesting apps, we'd have all these interesting electronic health records. And we knew that there was going to be consolidation in the marketplace, but we never really anticipated that we'd have three or four vendors that would really have a lock on the data. The thing that my next phase here now is let's find interesting things to do with that data. Can we make it easier to get data used by agents or used by AI? We used to say one thing we hate is lazy data. Data that just sits there. And I think one of the things that aimed is trying to do is how can we take that data and find interesting ways to make it better for patients.

[B]

We don't like lazy data, that's for sure. Not at all. Data needs to get a job.

[C]

Yeah, a big one.

[B]

A big job.

[C]

Especially in healthcare.

[B]

Especially in healthcare. What's left for us to do?

[A]

Doug, I think there's two things. One is we have to start working on that layer that sits above the data. And I think the future is going to be agentic and AI driven. I think that's what's going to be in the future. I feel like we're at that, maybe the 1946, 1950s in mainframe computers when the president of IBM said hey, we only are going to really need seven computers. That's really what the world needs, right? We've got like seven ehrs. And you know what happened after that? We had personal computers and now everything runs on my iPhone and it's all a bunch of apps.

[B]

Ken Olson from Deck said, who's going to need a personal computer on their desk at home? Right. No one will need a PC at home.

[A]

And now I've got a more powerful.

[B]

Computer than ever before. In your hand?

[A]

Yeah, in my hand, yeah, yeah.

[B]

So incredible.

[A]

And I think that's where we need to get. I feel like we've got to move out of these mainframes in the sky, which is all these cloudy hrs and we need to start getting, making it easy to deploy these things and actually start to use them in real life. I mean, Health Universe in large part is intended to be an operating system, if you will. It helps to coordinate agents, it does data normalization, makes it really easy. And then we do all the plumbing, we connect it in with fhir and we connect it into the EHRs and to the HIEs. And that's what we need because we don't need to have big programs that we're going to have to build on our computers. We need to be able to go and pick the app that solves the problem and be able to use it.

[B]

Right, so who do you sell to? Like developer community or do you have a developer kit or how.

[A]

Yeah, yeah, we sell both to large academic medical centers. They look for a platform, they've got a lot of developers, they've got a lot of interest and they have to wait two, two and a half years for good ideas to get through the gauntlet, if you will, so that they can be incorporated into their ecosystem. The CIO says, listen, if you can't give me a seven figure return on this, then we aren't going to take a look at it. And I actually think there's a tremendous amount of bottom up innovation that's just waiting to be released and that's what we intend to do.

[B]

What are you built on?

[A]

It's all built on Python. We use agentic AI. We were one of the first adopters of the Google SDK around A2A.

[B]

Okay.

[A]

And so it's built in from the ground up with those agents and then we do orchestration, we understand the context. So rather than having a chat interface that just gives you the results from literature search, like open evidence, we actually go and find tools. We find tools that are appropriate to the task and then are able to use those within those environments.

[B]

So you're like an MCP server with an API layer, Is that what you're doing?

[A]

We don't use MCP. It's all A2A and we've got an orchestration around that. So that Allows us to parallel. Parallelize a lot of the work. And so we, we do things like customize summarizations for cancer. And we can pull out everything from scanned images. We have our own pipeline for OCR because data matters. It's garbage in, garbage out. So we get all that data in a way that the agents can then use it.

[C]

And you guys got to help me on this. What is ocr?

[A]

OCR is optical character Recognition.

[C]

Got it.

[A]

So that's taking.

[C]

You guys are a lot smarter than I am.

[B]

It's like scans, like a newspaper.

[A]

Super cool.

[C]

So you can see optical for. Everybody lives listening that doesn't understand. I just needed to know this. Yeah.

[B]

OCR is an old term, Saul.

[C]

Okay.

[B]

You just showed your.

[C]

That's probably why I don't know it.

[B]

Saul's actually just graduating high school, folks.

[A]

Character recognition.

[C]

Super cool. Super cool.

[A]

Yeah. Well, the issue there is that we take data however we can get it right. If it's a scanned image, we'll take that. If it's a PDF, we can manage that. If it's structured from FIRE or from other ways to get it out of the ehr, we can do that as well. So it's all about meeting people where they are taking the information that they have and then being able to do something useful with it.

[B]

Love that.

[A]

And so we have a platform that plugs into an EHR and a lot of academic, medical centers and others like that. Because so many of these AI solutions are vertically integrated. Right. They say we've got this vertical solution that will help you with orthopedics or that will help you with diabetes or whatever. And the problem is when you've gotten four or five hundred pitches a year for every vertical integration, it becomes overwhelming. So our platform is a horizontal one that allows you to integrate once and deploy many. And then the other thing, we do a lot of co development work with some of them that want to market and monetize the things that they're working on.

[B]

Really interesting.

[A]

And then we've got innovators and entrepreneurs who want to focus on the technology and their expertise. And then we provide lots of the infrastructure that then can be deployed through our platform.

[B]

You're like a CORBA for AI?

[A]

Yeah.

[B]

You familiar with korba?

[A]

Yeah, I am. I like to think of it more like we're the Linux for AI.

[B]

Oh yeah. You got to keep sticking with that. That os. I love that.

[A]

We're just.

[B]

I'm just trying to blow Saul's mind. Like we. Korba, right?

[A]

Yeah.

[C]

I'm like we're speaking Greek here, people.

[A]

Yeah.

[C]

This is why I'm here, guys. I'm learning AI from the greatest. So talk to me about Korba.

[B]

Oh, my God.

[C]

In a nutshell.

[B]

Common object request broker, right? Architecture, right?

[A]

Yeah.

[C]

Wow. When I was a glossary for this.

[A]

Podcast when I was at onc, we looked at CORBA as an infrastructure and we said, no, it's probably not good.

[B]

Yeah. Well, then servers came out, and app servers pretty much replaced CORBA for integration. But CORBA was used for many years for really distributed, massive distributed computing. Boeing's whole manufacturing system was based on corba. Yeah.

[C]

Fascinating.

[B]

Yeah, probably still is today, which is a little scary.

[C]

I would explain it.

[B]

That was funny. And thank you to our sponsor, Boeing.

[C]

Well.

[A]

Oh, my God, my flight just got canceled.

[C]

Are you kidding me?

[B]

Oh, that was. Look at Doug coming in with a non sequit joke.

[C]

Whoa.

[B]

All right, man, I like this guy. I had no idea we were among greatness today. I just thought he was just another AI AI guy. Just. No, he's like, thank you, Doug, for joining us.

[C]

Like, legacy innovator, maker. And the operating system, like, so that's what you're building.

[B]

AI os.

[C]

Aios.

[B]

That's what you should tm. Aios.

[A]

Well, we named our platform Health Universe because Health Solar system seemed a little small in terms of our aspirations.

[C]

Yeah, that's fair. I like that.

[B]

Really? Health Universe. Yeah.

[C]

So as far as pulling the data out, though, like, a lot of it is in different silos. How do you guys help with that?

[A]

Well, a lot of those standards I actually developed when I was at onc. So there are standards called the Consolidated cda. There's about a billion transactions a month that happened with that particular one. And we use that in the HIE is the Health Information Exchanges and others to pull that out.

[C]

Okay.

[A]

I was the program officer for Smart on Fire when I was still a government program. Yeah. So look at you, Ken Mandel, and damn that whole group. So Fire is another way that you can build that stuff out as well. And there's APIs and other things like that you can have for integration.

[B]

It's the protocol used for the interoperable.

[C]

Well, all right. We're here with the guy that built the maze so he knows how to get the stuff.

[B]

All the problems associated with back in health care. Doug's fault.

[A]

It's Doug's.

[C]

Doug. Doug has all the keys. Is it him? Guys, I know where.

[A]

I know where the bodies are.

[B]

Yeah, I was just going to say that. Doug, Scary Jimmy Hoffa of Healthcare that's right. That's awesome.

[C]

So you're in the thick of it. Why'd you come to Aimed this year?

[A]

Yeah, well, I think this is the future. We've tried this three different times. We tried it with kind of decision support tools that were rule based and then we had neural networks and things like that. And now we've got something different. And this feels different.

[B]

Yeah.

[A]

The adoption of ambient scribe technology, the adoption of the large language models, it's been transformative. We are still going to end up with things that are going to be problematic and that are going to be hard, but this feels different to me. And so that's part of the reason to be here at Aimed. I think this is where a lot of the future is going to be created.

[B]

Absolutely.

[A]

Yeah.

[B]

It's incredible Turnout this year. 600 plus people with just amazing content. I love this type of format too. Same, like low key, focused on the content, focused on the physicians and the nurses and the other clinicians and what matters to them as it relates to AI and technology. It's like this is the group, this is the community that's going to figure out AI and how to apply AI.

[A]

Now, I actually think there's one group that's missing here.

[B]

The patient.

[A]

The patient.

[B]

Oh, look at Doug. And I like, boom, aligned.

[A]

So let me tell you a story. Okay. Between 1905 and 1912, if you go to the Journal of the American Medical association, there are all these articles about how doctors were going to be adopting this new technology that was coming out and how it was going to affect their practices. And they're all about the physician's automobile. And so they talk about how doctors need to know how to work their compute, their automobiles. They have to be mechanics to be able to understand that. They talked about productivity and how it improved patient safety. And you could get in your car and you could get to someone who had an emergency and all those things. But they stopped writing those articles.

[B]

Was that the ambulance?

[A]

Could have been, yeah. But they stopped writing those articles in 1912. So what happened in 1908?

[B]

1908.

[A]

And don't say the Titanic, because that was true.

[C]

But it's in 1908.

[B]

You're showing your age, Doug.

[C]

I know.

[A]

I said I've been in this a long time.

[C]

Damn.

[B]

I think Doug might be a vampire.

[C]

You look good for your age.

[B]

He looks great for his age. My God.

[A]

So Henry ford developed model two.

[C]

Was that 1908?

[B]

Is this like an ambulance?

[A]

Are we.

[B]

Is this the precursor to the ambulance?

[A]

So what happened is what we thought was going to be a physician's tool to improve and transform healthcare actually became a patient's tool to be able to make them. So now doctors could have offices and patients could come to them and radiology and big equipment and all that. Things started to grow from there. And I think that's one of the things I'm really looking to see what happens, because I think we spend a lot of time with technology to try to make patients better, but I think we're going to end up making better patients.

[B]

You're onto something. We're going to connect again. I'm going to have you on the Risk Never Sleeps podcast and we're going to geek out. We're not going to let Saul hold us back here. I was just kidding, Saul. I love this guy. It's called Korba. Saul is so amazing.

[C]

I've enjoyed this conversation.

[B]

Saul, you just got smarter. Like 10 points.

[C]

I think you are the average of your five closest peers.

[B]

That's so true. Yeah.

[C]

When you're the smartest person in the room, you gotta get out.

[B]

You gotta. You gotta find another room.

[C]

That's why you're here, right, Doug? You came here because there's a lot of other smart people.

[A]

Absolutely. Absolutely.

[B]

Awesome. Well, Doug, it's been a pleasure having you on the show. We do. We have a lightning round. We have an option. Okay. All right. Lightning round. Riskiest thing you've ever done. This is a rated PG show too, so gotta keep it 13. So you can go a little crazy.

[A]

I don't. Starting a company.

[B]

Yeah, starting a company. Yeah, it's risky. Never jumped out of a plane or swimming sharks.

[A]

Well, yeah.

[B]

Oh, yeah, that's.

[A]

Yeah, that was. That's not as risky as it starting a company.

[B]

I agree with getting married. That's pretty risky.

[A]

Yeah, that's pretty risky too. Yeah.

[C]

Doug, here's the second one. What would you say to your 20 year old self?

[B]

Yeah, there we go.

[A]

Buy Amazon.

[B]

Oh, I did.

[A]

My 20 year self was not listening to me.

[B]

You should have talked to me.

[C]

Good one.

[B]

Wait, you were much older when they were out and available on the market. You weren't 20. Come on, go back in time, let's go.

[A]

No, it would have been like I was 20.

[B]

Buy IBM would have been like, by Microsoft. By Microsoft.

[C]

Yeah.

[A]

I would have told myself, buy Amazon. And I'd be like, what is Amazon? And then I would have missed it.

[B]

Yeah, but you weren't 20 then. Like, come on, dude, go back in time. You're 20 years old.

[A]

Yeah.

[B]

What are you doing? You're 20 years old. You see yourself standing there.

[A]

When I was 20, I was lonely.

[B]

Doug, in the corner.

[A]

I was in medical school and my residency, and I had no life.

[B]

So get a life.

[A]

Get a life.

[B]

Get a life, Doug.

[A]

Yeah.

[B]

Damn it. All right, Desert Island 5 records. What would you bring with you?

[A]

Gosh, probably a whole range of things. Get Lady Gaga, some edm, maybe get. Who are you, Beatles. Okay, back to reality, folks.

[B]

I'm telling you, this guy does not listen to Lady Gaga. There's no way in hell.

[A]

All right.

[B]

We'Re going out with Doug tonight, I think.

[C]

Let's party. Yesterday's dinner was fun.

[B]

Tonight with Doug is just going to be epic. In fact.

[A]

Yeah, I've got three speakeasies already lined up.

[C]

Let's roll.

[B]

All right. Let's roll, man. Let's go with Doug.

[C]

He's got the map.

[A]

Yeah.

[C]

All right.

[B]

You want to. You want to take us out here?

[C]

Yeah. Listen, Doug, it's been a pleasure to have you on Best way that people could learn more about Healthcare Universe, Health Univers Universe. Thank you. And. And how to get in touch with you and learn more about you.

[A]

Well, I think it's healthuniverse.com. yeah, it's free. You can set up an account. It's super easy to just deploy an app and just play around with it. It's very easy to do that, and that's probably the best way. And it's just my first do. Last name at Health Universe.

[B]

Awesome.

[A]

And that's the way to get in touch with me.

[C]

All right, we're going to have all that in the show notes.

[B]

Well, Doug, pleasure. Thank you, sir.

[A]

Thanks for having us. It's.

[B]

Thanks for listening to Risk Never Sleeps. For the show notes, resources and more information and how to transform the protection of patient safety. Visit [e-mail] that's C-E N S I N-E-T.com. i'm your host, Ed Gaudet. And until next time, stay vigilant because Risk never sleeps.