John Bertrand on FDA Approval, CMS Reimbursement, and AI Healthcare Reality

DC EKG

In this episode of DC EKG, Joe Grogan interviews John Bertrand, former leader at Digital Diagnostics, about the real-world challenges of bri

Key takeaways

• Autonomous AI diagnostics face significant regulatory hurdles because existing frameworks are designed for physician-led care, not algorithmic decision-making.
• CMS reimbursement rates for AI tools can be drastically below cost of goods sold, creating a major financial barrier to adoption.

Main topics

• FDA approval for autonomous AI diagnostics
• CMS reimbursement challenges and policy gaps

Notable quotes

"There's more scrutiny on AI than there is even on a clinician."

Conclusion

Bertrand concludes that while autonomous AI diagnostics represent a transformative potential in healthcare, the current

Transcript preview

Speaker 2 (0:00) There's more scrutiny on AI than there is even on a clinician. What you're monitoring for is what's called diagnostic drift. Does the algorithm continue to perform over time consistently? But yet we don't actually post med school really take a look at a random sampling of physicians' diagnostics and say, are they accurate or not? Speaker 1 (0:17) This is DCEKG, exploring the intersection of politics and healthcare by diagnosing Washington and prescribing solutions. We take a deep dive into the legislation, regulation, and innovation impacting policymakers, providers, and patients. Now, America's health policy expert, Joe Grogan. Speaker 3 (0:39) Welcome back to DCEKG. I'm your host, Joe Grogan, joined today by John Bertrand. from Digital Diagnostics, the company behind Luminetics Core, the FDA-cleared AI diagnostic device, the first one. And John has a tremendous amount of experience in healthcare tech from Epic Systems, executive and residents at 8VC, and then his experience at Digital Diagnostics. He understands both the FDA clearance side and the reimbursement side, which gives him a unique perspective in this area. John, I want to talk to you about AI today, but I really want to thank you for joining the show. Thank you. Speaker 2 (1:27) Thanks for having me. Appreciate the introduction. One quick clarification for our listeners. Formerly of Digital Diagnostics, I wrapped up there January beginning of the year, and I'm now at a different business using AI in healthcare. Still happy to be here and excited to talk more about AI, the FDA, and reimbursements future around this exciting technology. Speaker 3 (1:48) All right. Thanks. So your chapter with digital diagnostics is now complete. Speaker 2 (1:52) It is now complete. Yeah. We got out to about 75 health systems across the U.S. Team scaling. We're rolling. And it was just time for a new adventure. And still playing around with AI, working more on the supply chain and fulfillment side here. biotech side, but still keep my finger on the pulse of what's going on in D.C. as everybody does here. They tend to have a big impact as they evaluate policy on how AI is applied in health care. Speaker 3 (2:19) Great. So I just want a little bit on your background. You were at Epic Systems for a number of years. Can you just talk a little bit? People are always interested in how people got where they went. And just talk a little bit maybe how you ended up at Epic. how long you were there, and then how you came to found digital diagnostics. Speaker 2 (2:40) Sure. So I started my career out of undergrad, went to work for Epic way back before EMRs were mandated in every exam room. In fact, there's still tons of paper when I started way back in 2006. Like many people, I was an analyst. I helped deploy software, make sure end users were happy, fix issues post-go-live. quickly moved into a product management role where I was responsible for managing the RevCycle suite of applications. One thing led to another. I ended up doing some enterprise sales there, worked on a variety of international projects, a bunch of other programs in and around there. You know, over 13 years, it's kind of fair to say I did pretty much everything with program. It was a great learning environment. If you wanted to learn healthcare and technology at that point in time, you know, the EMR was the place to be. We were digitizing the patient chart. Everything revolved around that. That was the big investment of time and industry in the industry. And I got to learn a ton. After about 13 years, you know, started to get a little bit itchy, wanted to find a new adventure. And I ended up in my travels running into the founder of Palantir, Joe Lonsdale, who was just starting and scaling 8BC. They had a small fund out that was really doing angel checks. And they're like, you know, it's getting bigger. We want to invest a ton more in health care. We're really excited about AI data. digital health and some alternative like value-based care delivery models and I thought why not and through that kind of happenstance meeting ended up turning into a role. I worked there for several years kind of like an operating partner or exec in residence is what we called it you know really working with the investment team on what the thesis was for where the hockey puck's going helping portfolio companies you know get to the next level from an operating perspective and you know raise money to continue growing their businesses. Around about that two-year mark, the goal was always to jump in and actually operate something. I got really excited about digital diagnostics. So it was actually founded by Dr. Michael Abramoff back in 2010. So by the time I met him in 19, it was already a nine-year-old business. And Speaker 3 (4:42) those Speaker 2 (4:42) first nine years of the business were really focused on getting the algorithm through and cleared by the FDA to diagnose our flagship use case, diabetic retinopathy. But once he got it cleared, all of the like, how do you make this a reality? How do you get into patient workflow? How do you build a business case? How do you grow your case study suite so you can get others excited and willing to take a leap on a newer technology? It's kind of the next phase. And I spent a ton of my career doing that. And when Michael said, you want to jump in and help me scale this thing? I said, sure. You know, fast forward six years later, again, like I mentioned, we're in 75 health systems. We have a reimbursement code. We're partnered with large medical device. vendors around the industry to automate second, third, and quaternary use cases around AI diagnostics driven by biomarker-related detection. But that's kind of been my career, always building new things, kind of moving to where the new thing is and helping turn kind of laboratory ideas into actual hardened products that are used at scale across healthcare. So it's been a fun journey. Right now, as I mentioned, I'm at PartSource. We manage about half the supply chain for biomedical medical imaging related critical assets so it's very similar thematic across my entire career find workflows digitize them and seek to automate them Speaker 3 (6:01) that's really cool okay tell me about let's just refer to the audience members who don't know what does the device do Speaker 2 (6:09) So for digital diagnostics, what we're talking about here on this call, we took an off-the-shelf fundus camera. So if you go to an eye doctor and they image the back of your retina, they flash you, and then they take a digital image. We really just interjected ourselves into that workflow and said, use the same hardware everyone's been using for 100 years, but let's connect that to a secure cloud-based platform and push the digital image into it. And from there, we had a neural network that dissected all the digital data or biomarker data captured digitally. on the retinal image and then use that to calculate out whether disease was present or not. So really how that manifests, you know, it's like a long-winded technical explanation of what we did. Really, we automated a piece of specialty hardware that existed only in an eye care office. And by automating it in the manner that we did, we were able to move it into primary care because by removing the need for the specialty physician to be there for the diagnostic process, you can now give the device to a low-skilled high school educated operator. And in 10 minutes or less, actually have them producing a diagnostic output similar to what you'd see from your physician. Same workflow. Image goes somewhere. Diagnostic report comes out. But the unique nature of how we did it meant that your point of care diagnosis was available immediately. And all the downstream subsequent treatment steps can start right then and there. And it also unburdens the specialist from needing to be involved in the diagnostic loop. right you still have the the primary care physician present on the education and managing the downstream next steps for the patient but you really don't need to have a specialist on site or even asynchronously available to provide that diagnostic output Speaker 3 (7:48) So it's awesome, right? I mean, that's a tremendous advance. Is it fair to say you're the first one to get FDA approval? Was the first one to get CMS reimbursement? I can't remember. Speaker 2 (7:56) No, we were both the first ones to get FDA approval for a diagnostic algorithm without a physician in the loop. And we were also the first AI without a physician in the loop to get CMS reimbursement. And I got to tell you, I really underestimated how difficult and challenging that would be. You don't realize how many places in our health care industry's regulatory framework we say the phrase the physician. So when you show up and there is no physician for the first time ever, you start to trip over all these places where it's, oh, the physician needs to be present for the diagnosis or the physician's work effort feeds into the reimbursement calculation. Well, how do you work within the RVU framework of CMS reimbursement calculation if there is no physician time? And you have to kind of noodle through and create new frameworks every time you're running up against one of those kind of like firsts. So there are fun problems and challenges to solve, but they are a lot bigger than you realize when you get started. Again, just because our whole mental model on delivering health care up to this point has been all about a clinician driving care decisions, driving the care pathway and being involved and in the loop. Speaker 3 (9:02) Right. So let's press on this a little bit. I want to start with a threshold question, which is there's a perception, I think, amongst a lot of people, which is if you invent something awesome that's going to improve patient care, it should be able to be adopted. Like you would want in an ideal world physicians to use the greatest scalpel that's been invented, the newest imaging device, the newest. Speaker 3 (9:31) latex glove that that's better than the previous one in your experience being there on the vanguard of a tremendous innovation like this what's the reaction to like how are people's perceptions upside down or is that accurate or or how would you react to that Speaker 2 (9:50) Yeah, you often think like, hey, I made a better mousetrap. Everybody should be using it, or at least like a technologist tends to think that way. The reality of trying to get something that's that new and pioneering out into the marketplace is that everybody's questions center around a couple of different buckets. One is, well, how is this going to fit into my workflow? No matter how great it is, if you can't provide a concrete, succinct example. of how you're going to integrate into what they're already doing today moving through their care areas you're kind of dead in the water if it's an innovation that sits off by itself or isn't integrated into the broader workflow and ecosystem i mean we even needed to do things like integrate into the emr schedule to see if a patient with diabetes is showing up because that's the beginning of integrating into the workflows flagging i have a patient with diabetes and now they need this exam to take place without that integration it just becomes really a fancy paperweight pretty quickly. So there's that component you need only me to answer practically. How do I get this into my workflow? How does it actually, how does it actually manifest day to day from patient to patient? And then unfortunately, or maybe just the reality of how humans work is the next question is, how am I going to get paid for this? And that's why we ended up pursuing a category one CPT code, literally from my walking in the door into the business was, even if we're providing clinical care, That's in an improved way, higher sensitivity, better specificity. The problem still remains of today, when I test someone for this with a human, I'm getting paid X dollars. And even if this is clinically better, I need to understand how I'm not robbing Peter to pay Paul here and essentially making my financial situation worse. The reality for most healthcare providers is they're running on razor thin margins. There's massive meta reimbursement headwinds for everybody. We've got the site of care shift happening where we're trying to triage patients to areas to provide care that are lower cost and therefore lower revenue for health systems. So you've got to be careful you're not taking something away from them on the financial aspects of things when delivering the new technology. What's interesting is if you look at the adoption curve of any product that is an AI product the FDA has approved in the last five years, the only ones that are seeing that ramp up. would be like HeartFlow, Viz.ai, and us. Do you want to know what the clear, consistent trade across all of