Hospitals are a critical, yet incredibly expensive part of our healthcare system. So why are so many operating on razor-thin margins, with some even facing bankruptcy? The answer lies in a massively complex and broken billing system, leading to billions in hospital revenue leakage every year. This episode explores a groundbreaking solution: using AI for hospital revenue cycle management to ensure hospitals are paid accurately for the care they provide. We're joined by Dr. Michael Gao and Dr. Joshua Geleris, the physician co-founders of SmarterDX, a company tackling this crisis head-on.
In this deep-dive discussion, the HealthTech Remedy team and the founders of SmarterDX unpack the intricate challenges of hospital finances. We explore the historical complexities of the Diagnosis-Related Group (DRG) system and how tens of thousands of ICD-10 codes create a nightmare for manual coders, leading to frequent errors and costly claim denials. This environment is precisely where physician-led innovation can make a difference. We analyze how SmarterDX's platform leverages clinical documentation improvement AI to find missed diagnoses and justify the care provided, a crucial step to reduce hospital revenue leakage. The discussion covers SmarterDX's two core products: Smarter Prebill, which reviews claims before submission, and Smarter Denials, which provides automated claims denial management by generating evidence-backed appeal letters in minutes. Co-founders Dr. Gao and Dr. Geleris share their founding story, which began when they discovered their own names on lists of attending physicians with missed diagnoses at New York Presbyterian. They explain why the combinatorial complexity of medical coding makes it a perfect problem for AI to solve and how their technology delivers a guaranteed 5-to-1 ROI, adding millions in net new revenue for health systems without requiring new staff. This conversation provides a masterclass in applying AI for hospital revenue cycle management to solve a critical, real-world problem in healthcare operations.
Introduction
Dr. Trevor Royce: Hey guys.
Dr. Tim Showalter: Hey, morning.
Dr. Trevor Royce: Guys, it is so hot here right now. I think we need to do our recording location, maybe in Maine.
Dr. Tim Showalter: Oh yeah. Paul, can you hook us up with that?
Dr. Trevor Royce: Guest room, stocked pantry. How are we, how's it looking?
Dr. Paul Gerrard: I've got my recording location here in Maine. Oh, you guys want to come here too.
Dr. Tim Showalter: Yeah. That's what we want. I did, Trevor, I have visited Paul once. During a college visit for my daughter, I visited him and we got a couple of beers at the local breweries. It was good.
Dr. Trevor Royce: Wow. Was that a plane, a train, a boat, and then a car? Or how did you get there?
Dr. Tim Showalter: I guess a car and then a plane and then a rental car. There was Paul just waiting for me.
Dr. Paul Gerrard: Beer in hand.
Dr. Tim Showalter: Yeah, beer in hand. Yeah. Met his wife and everything. He's actually married.
Dr. Paul Gerrard: Poor woman.
Dr. Trevor Royce: Should we do it?
Dr. Tim Showalter: Let's do it.
Dr. Trevor Royce: Paul?
Dr. Paul Gerrard: Well, I think today's episode is going to be a good one. So good it might even make revenue cycle management palatable to people. Might.
Dr. Trevor Royce: Welcome back to HealthTech Remedy, a podcast where three physician leaders unpack health tech innovations. I'm Trevor Royce, radiation oncologist with a background in AI diagnostics and real-world evidence.
Dr. Paul Gerrard: And I'm Paul Gerrard, a former physical medicine and rehabilitation physician who focuses on reimbursement policy and market access for AI-enabled diagnostics.
Dr. Tim Showalter: And I'm Tim Showalter, radiation oncologist and former med device entrepreneur who's now working on integrating AI into clinical workflows.
Dr. Trevor Royce: And today we're spotlighting a quiet revolution in hospital operations. SmarterDX is a physician-led clinical AI company tackling one of healthcare's biggest hidden problems, revenue leakage.
Dr. Tim Showalter: Not everyone knows this, but you couldn't pick a better analyst for SmarterDX than Paul Gerrard. So I think this is mainly the episode where during the discussion phase, we just wind Paul up and let him go. And I'm sure he's just got a lot of perspective and analysis to share about this company.
The High Stakes of Hospital Revenue Cycle Management
Dr. Paul Gerrard: I guess that's the cue for me to let my geek colors fly here and talk about hospital revenue cycle management, probably one of my favorite topics. Few people really appreciate how complicated hospital revenue cycle management is, what goes into it. I think people know hospitals get paid a fixed amount of money on a DRG. But the way those payment rates for the DRGs are set and then how the appropriate DRG is selected is a complicated issue.
And then on top of that, at the time that a hospital claim is submitted, no human on the face of this earth really knows who owes whom how much money. And so often there are various settlement processes afterwards. So hospitals are under increasing financial stress. In 2023, about 20% of insurance claims were denied. If you've ever looked at a denial, it sometimes is not obvious exactly why. It's some administrative denial or you get not medically necessary or other. And let me add on top of that, sometimes it's not even an English phrase. Sometimes it's a code that then has to be matched to some vague phrase on a table.
Hospitals are constantly trying to make sure that they are capturing all of the revenue that they are legitimately entitled to. And I remember back before I was a medical director for one of the Medicare administrative contractors, I used to sit with the billers for a hospital. I'd go through medical records and look at what are the diagnoses that are important that need to get captured on the claims. A lot of this was based on just my knowledge of how DRGs got captured, how comorbidities got captured.
Sometimes there are a lot of things that the patient would legitimately have. It would be documented in the medical record, but the coders didn't realize that it had a financial impact on the hospital. I would put on my expert billing hat and I would say, "Oh, make sure we capture that one if we've missed it." And often they hadn't because they would think, "Oh, that's some little trivial thing that I don't need to put that in my top 10 or 25 diagnoses. That doesn't need to go in." And so starting to have ways to, A, find that without having a human spend time go through it is really important. Not every hospital has somebody who knows this stuff inside and out, but every hospital has coders who kind of know it, and they just have to manually go through and abstract things out of the medical record.
And so even if they're not thinking about making sure that they are optimizing and capturing everything to which they are legitimately entitled, it's just a very manual process. And then on top of that, making sure that they are really capturing the things that are costly, that therefore drive higher payment rates is one more step on top of that. And that requires not just manpower, but expertise as well. So it's great to be able to start to automate this and have systems that can do it. It can A, save on manpower and B, potentially obviates the need for some of that expertise, which not to toot my own horn, but I think very few people have just because A, it's tedious and boring and B, it's complicated.
Dr. Tim Showalter: Yeah. Paul, I'll hop in. I think as a radiation oncologist, I feel like our coding is pretty complicated. There's a lot of CPT codes. They all have to be justified. And the challenge is when you deliver costly services, and obviously in the hospital setting, a lot of what's delivered is pretty costly. And you've got a denied claim. It can really chip away at hospital's margins. So my experience of this is mostly on the department level setting or the service line setting. And things get really complicated. The cost of the potential denials to the bottom line is really impactful, and there's a pretty high-stress process and can be really confusing to hop in and try to get into the weeds of it. Seeing that there are over 30,000 data points per chart, and there's over 150,000 ICD-10 codes if you look broadly throughout medicine. So it's really no surprise that errors or omissions occur when you've got humans doing this, even with a lot of training, there's just a massive amount of information to wade through.
Dr. Trevor Royce: Just to zoom out on the problem a little bit. And I think you teed that up really well, Paul. But I feel like this is one of the great paradoxes in U.S. health care. The hospitals are such an expensive part of such an incredibly expensive industry. And yet we always hear that they're at risk and they're not doing well. And I think this begins to get at underlying the complexities of that and why that paradox exists.
Dr. Paul Gerrard: There's been increasing financial pressure on the hospitals ever since the implementation of the DRG system. And if we go back a few decades, hospitals weren't getting paid based on DRGs, but there was this growing healthcare costs and health economists and Congress said, hey, we need to figure out a way to get this under control. So they started to have this idea of let's just pay a fixed amount of money, a fixed amount of money for these buckets of care. And that was the birth of the DRG system. The DRG system has, of course, evolved significantly since the 1970s. But one of the things that we've still seen is healthcare costs have continued to rise, even with the implementation of the DRG system.
And the challenges to hospitals are that the way payment rates are set for DRGs are based on these Medicare cost reports that are periodically updated. Medicare is basically using numerical methods, statistical extrapolation to associate certain diagnoses, certain buckets of care with certain costs. And so nobody is necessarily going and saying, hey, based on my substantive knowledge of the kind of care we deliver, we know X, Y, Z is more expensive than something else. Not in general. There are a couple of exceptions where that happens.
But what that means is you as the doctor who have that substantive knowledge, you may in your brain associate something with high cost. But it doesn't necessarily align with what Medicare associates with high cost based on the way they're crunching the numbers. And of course, they're not crunching the numbers based on the actual medical records. They're crunching the numbers based on claims. So hospitals that are submitting claims or coding things using a different way than what the norm is may be subject to different payments for delivering the same sorts of care. It could result in under or overpayment.
Dr. Tim Showalter: Depends on your perspective.
Dr. Paul Gerrard: Yeah, exactly. People can try to game the system. They can say, oh yeah, we can really try to chase after those codes we know are going to get paid more. On the flip side, people may not realize the diagnoses that are associated with a high cost, and therefore they may not be making sure that they are capturing them either very clearly in the medical record. The doctor might be treating them, but the doctor may not note explicitly, hey, the patient's on this drug for X, Y, Z. You need that level of granularity to make sure that the diagnosis is supported by the medical record to justify the ICD-10 code on the claim. So it may be that patients are getting treated for it. The hospital is incurring the cost of treatment, but nobody is explicitly documenting the cost of that diagnosis. And therefore, the claim is not expressing the cost of that diagnosis. And therefore, the hospital is not getting paid. And if you've got a hospital that's already running on thin margins and you're doing high-cost treatment, but not getting high-dollar reimbursement, that could be a real problem.
How SmarterDX's AI Tackles Revenue Leakage: Prebill and Denials
Dr. Tim Showalter: This is a risky and complex area. We should pivot a little bit in terms of thinking about what it is that SmarterDX is doing to try to help out in this area. They have these two flagship solutions. One is Smarter Prebill, and one is Smarter Denials. There's some automation step here. My perspective is it's not only the automation in terms of having some steps that are scalable and can help support the administrators, but also just digesting a ton of documentation. So it's sifting through data really quickly and organizing it and preparing information for submitting claims and for responding to claims appeals.
Dr. Trevor Royce: Smarter Denials was launched in late 2024. [12] And basically automates the creation of evidence-to-back appeal letters. It does that by extracting text from scanned denial faxes, identifies points that might need further clarity or to be rebutted in some way, It crafts tailored responses in minutes, so it automates this process in a thoughtful way.
Dr. Tim Showalter: Yeah, and then, of course, they have the Smarter Pre-bill system as well. And that's really an overall program that provides review of claims before they leave the door initially. So that starts at the very beginning and reviews the entire record and identifies missing diagnoses, I think, to Paul's points, and codes that aren't aligned and provides the opportunity to double-check and make sure that the full value of the care is being captured and the documentation is together before that claim goes out to the payer.
Dr. Trevor Royce: That's a good way to phrase it where you have the Smarter Pre-bill on the front end and then the Smarter Denials on the back end.
Dr. Paul Gerrard: So ideally you're going to be billing accurately initially and then hopefully have reduced problems with denials on the back end. But for those denials you do have, there's now a software program that can help automate that process because as we all know, if you ever get an actual copy of a hospital record, it is insanely long. And the critical information may be in there, but good luck finding it.
Dr. Tim Showalter: Yeah, almost made worse by note bloat. There's so many people are repasting things into their note to have a detailed summary. But when you look at the medical record in its entirety, there's a lot of repetitive information. So I think this is where it's really obviously powerful to have these algorithms going through and analyzing this.
Dr. Trevor Royce: We kind of do all this as humans and physicians anyways. It's just a much more efficient, smarter way to do it. How many times have we been going through the medical record trying to figure out what happened to a patient in the hospital and just reams and reams of documents and papers and medication administrations. And really all you're looking for is that discharge summary that talks about the hospital course in a sea of tens of thousands of pages.
The Physician-Led Team Behind SmarterDX and Its Compelling ROI
Dr. Tim Showalter: And nobody wants to do that. We're happy to have AI do that. Just to focus a little bit on the people behind SmarterDX. So the company was founded by two physicians. So Dr. Michael Gao, who's a CEO, who we have the chance to interview later, and Dr. Joshua Geleris. I know Dr. Gao was Medical Director of Transformation at New York Presbyterian, and then Dr. Joshua Geleris is a Columbia bioinformatics researcher. Thinking of their founding story, obviously, all physicians have some experience in dealing with the complexity of coding, although most of us don't have the detailed knowledge that Paul might have. But I think having a bioinformaticist and a medical director who's really responsible for technology in a healthcare system really sets them up well to execute in this space, because ultimately this is about directly embedding within the hospital level workflows and providing the tools to do a better job in terms of revenue cycle management.
Dr. Paul Gerrard: I suspect that AI like this that is able to really carefully go through the medical record is going to be better than even the best human expert. SmarterDX guarantees five to one return on investment from day one. And most hospitals see $2.5 million in new annual revenue per 10,000 discharges. That sounds like a pretty compelling reason to start to think about adopting this technology.
Dr. Tim Showalter: Some systems have reported large numbers like 12 to 54 million in new revenue based on reports I found on the web. And this is without hiring a new staff member. So it just shows you correcting and identifying some of these missed opportunities based on these automation steps and these algorithms can be helpful. They also describe an implementation timeline as short as eight weeks, which I think is what you would really want to see at the hospital level because everyone's overworked. You want to make sure that if you're going to stand up a program like this, you can get it going quickly.
Dr. Trevor Royce: From the hospital side, the next logical question is, if you're considering partnering with a health technology startup that uses AI, well, do they integrate with my medical record? And as best I can tell, they integrate with basically every major medical record, Epic, Cerner, Meditech, and so forth. And then of course, do it in a secure way and respect all HIPAA compliance. So those are some key technical details.
Dr. Tim Showalter: A very recent update is that SmarterDX is now part of a broader corporation called Smarter Technologies, and that includes Thoughtful.ai and Access Healthcare. So it looks like they're all put together and backed by New Mountain Capital. So I think it would be interesting to ask Michael about that when we talk to him about what that announcement means for them.
Final Analysis: Why a Physician-Driven AI Solution is a Game-Changer
Dr. Paul Gerrard: One thing that I also wonder about is the potential future here. Right now, it appears that really their focus is on the hospital side, but to me, it seems like this is the kind of thing that also has a significant opportunity on the payer side. Having done things with the Medicare program, even the reviewers on the payer side at Medicare often don't necessarily entirely understand what goes into the claim, how to support medical necessity, how it impacts reimbursement. So, to me, I ask the question, is there an opportunity for this kind of technology to also be used on the payer side for ensuring accuracy of billing and program integrity? The physician-led model helps to build clinical trust. They're clearly tackling a big problem that's tedious, that I think AI is well-suited to address. And it looks like at this point, they're getting good outcomes for their clients. The challenge here is they do have to have technical integrations. That can be complex, but it sounds like they've got a way to do that and get an integration in eight weeks. And anyone who's lived through hospital IT solution changes, eight weeks is breakneck speed for having changes.
Dr. Tim Showalter: And I imagine that's a two-way street. So you have to make sure that the hospital is prepared for their side of that integration as well. What I really like is that they're leveraging technology in a way that is really outcomes driven. So I like seeing that they've got these impressive metrics like a five to one ROI with no new FTEs. I think that's the sort of example, if you're building technology in health care, you've got to be from the very beginning outcome oriented. And then the other theme that I think is really interesting is just simply using AI to reduce the massive burden of information gathering and synthesis. You see that in the analysis of the medical records to help support the coding. And I think that process of digesting the information and then aligning with the appropriate code is something that obviously would take a human user hours to do, and with AI, it can happen quickly. And I think that's exactly what the health system needs to reduce burnout and address a lot of these bottlenecks.
Dr. Trevor Royce: Yeah, I agree, Tim. I think keeping the big picture in mind, this is such a massive issue. And I think the industry is really desperate for a solution. I would love to see the solution be having a more simple system as opposed to a solution to a very complex system. But I think this is probably the best that we're going to get in the modern era where you're coming up with a tool that can better navigate a very complex system, and I do think the market will react to it very positively. They're desperate for a solution.
Dr. Tim Showalter: It's not just an AI play. It's really a physician-driven innovation and it's purpose fit for hospital operations.
Dr. Trevor Royce: All right, up next, our conversation with Dr. Michael Gao, CEO of SmarterDX. You won't want to miss it.
Interview with Dr. Michael Gao and Dr. Joshua Geleris, Co-Founders of SmarterDX
Dr. Tim Showalter: Today, we're diving into the world of clinical AI with two remarkable guests, Dr. Michael Gao and Dr. Joshua Geleris, co-founders of SmarterDX. SmarterDX is helping hospitals tackle their most pressing problems related to revenue integrity and revenue cycle management. And it's a physician-led team with the two co-founders and using AI-powered platforms to improve ROI in this space. Michael and Joshua, welcome. It's great to have you both.
Dr. Michael Gao: Thank you for having us.
The "Aha" Moment: Uncovering the Root of a Billion-Dollar Problem
Dr. Paul Gerrard: We wanted to ask about the why. Every company, they build their business on solving a problem. I guess the question we have is, take us back to that moment when you realized there's a fundamental problem with how hospitals document and bill for care, or at least there was an opportunity to really improve it. What was that moment and how did that translate to the founding of SmarterDX?
Dr. Michael Gao: Josh and I were both co-residents at New York Presbyterian after I was a hospitalist at Cornell and Josh was up at Columbia. For me, I was helping New York Presbyterian lead various AI initiatives. And there's this interesting duality that I saw, one of which was that you had the traditional revenue cycle vendors who weren't always the best at using leading-edge technology or using that to truly transform processes.
But were really good about delivering clear ROI to their customers and being really clear about the business value. And then I saw, to exaggerate a little bit, your 23-year-old out of Y Combinator, whose value proposition was, once I get a thousand hospitals of data, I'll have this amazing product. But in the meantime, there's no clear ROI.
And why can't I get traction? And came across the clinical documentation improvement space somewhat accidentally when I was trying to work on length of stay variation. Actually, Josh and I were both working together on length of stay and improving length of stay and realized that clinical documentation was super interesting because auto coding was cool. You could change documents and unstructured text to structured codes. But to improve the documentation in the first place, you actually had to deeply understand the underlying actions of the care team. And then improving documentation led to real actualized dollars for the health system. So it was this really cool intersection of getting to take complex data and think about how AI can be used to understand that data and deliver hard ROI to your customers as that intersection of technology and business value. And that was, I think, when Josh and I started talking about it and realized that it was just a really compelling space to go into.
Dr. Paul Gerrard: Were you initially thinking about starting on the back end of appeals in revenue cycle management? Or were you thinking about starting on the front end, helping to, even as far upstream as just helping to get the clinical documentation right to support having the appropriate diagnoses in the claims?
Dr. Josh Geleris: And, more practically, when I first learned about CDI and when Mike introduced me to the whole space and the concept, at that point, he had, at New York Presbyterian, put together a couple queries against the database, pretty rudimentary, but they were able to find really significant opportunity to capture diagnoses that were missed. And for me the aha moment, I think since we're all physicians on this call you'll all relate to this, was when we looked at the list of of misdiagnoses and looked at who the attending was and I saw my own name and I was like oh this is a real problem that isn't just a bunch of bad doctors, like even I miss some stuff here. And when we were able to connect those misdiagnoses to the value that the hospital could earn from actually capturing them correctly, that really lit a fire under us to solve the problem more broadly and to figure out how we could bring a solution to market that would really help providers who are already extremely low margin, in a low margin business, get more resources to take care of more patients.
Dr. Michael Gao: Most intuitively, it's, hey, we can build better algorithms. Let's insert it to improve efficiency and help the team, whether directly through documentation or while the patient is in the hospital, improve the quality of that documentation.
And it turns out that when you sell to health systems and your sales pitch is, "I know you have no idea who we are but we have better algorithms, trust me," it doesn't go very far. I tried, Josh and I tried for a while. It's obvious once you do it, but it took us a little bit of time. What we started to do is we started to say, okay great, well, actually, we'll put this in a counterintuitive place from a workflow perspective, which is you have finished all your documentation and coding. And then we will QA that. Therefore, if we find something, it's not just, you can't just say we front-ran something you would have captured anyway. It really is net new dollars. And we're able to prove out that value proposition. And that's when the product market set took off.
And it was actually kind of counterintuitive to be behind everything and not to increase efficiency. The neat thing is the amount of extra dollars we find for our customers is greater than the entire cost of their clinical documentation improvement and coding teams for the inpatient space. In other words, we find more value than they would achieve through 100% automation of their existing work. So it actually turned out to be kind of fortuitously the right space to insert for that value proposition.
Dr. Trevor Royce: It feels like we're just on the precipice of taking a deep dive into these products that you guys have built. And I think our audience will be very interested to hear that. Before we go down that path, something you said earlier, Josh, really resonated with me. I have a very vivid memory of my first week as an attending after 12 to 14 years of higher education and getting feedback on those first batch of notes and being like, wow, I have no idea how to craft this note. I've never been responsible for the billing side. This is an entirely different language and just wildly complex. Can you give the audience that may not have that direct interaction on the personalized billing side in your own clinics what the kind of problem is for the hospitals and in the very big picture, what you guys are solving for with these really innovative products that we're about to dive into?
Dr. Josh Geleris: Yeah. What's really interesting about healthcare specifically is that basically all payments are determined based on what physicians write in their notes. And before I started Safari DX, I knew nothing about how hospitals got paid or what's important to include in your note or any of that. And most doctors really just don't have a very good understanding of the system that we use for reimbursement, especially on the inpatient side where it's really complex and ultimately determined by a medical coder who goes through your notes and abstracts the set of diagnoses that you've documented for care. And I think it's even more esoteric in part because the system we have is a historical accident.
My dad's a physician. He's a neurologist. And I have vivid memories of going to the hospital with him and sitting there while he had to sign his charts. And in that time, it was stacks of manila folders with these little post-it notes on them for where he needed to sign. And you couldn't physically send that chart to the insurance company in order to figure out how much value you provided to the patients. They needed some way of taking that sometimes 100-page chart and getting it to a fax machine in a reasonable process. And so what they came up with was the system of using ICD codes to catalog the care that was provided and turning that into a system that could be used for reimbursement because you can fax a single sheet of paper with 25 ICD codes to the payer and come up with a system that way. And ultimately, that system has persisted now, even through EMRs and everything.
And the challenge with that is that physicians, I'm sure you all know, don't think in terms of ICD codes. That's just not our mental model of how we take care of patients. And I, prior to SmarterDX, had seen a few dozen ICD codes. I knew what the concept was, but I certainly didn't have a deep understanding or appreciation of what the catalog of codes was. But ultimately, somebody's got to translate your notes into that system of codes. And there's a lot of loss in that translation process. And ultimately, I think that's what we at SmarterDX are trying to solve by using AI that does understand that catalog to interpret the note, but also the clinical actions providers take in terms of ICD codes so that hospitals can accurately catalog what happened to the patient during this day.
Dr. Trevor Royce: Why is AI the right tool for this? What is it about the system that makes this right for our modern world of AI?
Dr. Michael Gao: Broadly, the first is with AI, it's, why don't people do it well? And for the people on this call who may not be familiar, there's about 30,000 ICD-10 codes and tens of thousands more procedure codes.
And as doctors, you treat a patient and you're like, okay, they have some chronic kidney disease, I better adjust their vancomycin dose. And that's the level of specificity that you care about. But of course, it turns out that whether it's diabetic nephropathy or chronic kidney disease stage three or stage two have wildly different impacts on reimbursement and quality, even though it might not matter for the care that you're providing in that moment. The reason why people don't do it well is fundamentally because the specificity on the billing side is somewhat orthogonal to the specificity that you're actually thinking through on the actual delivery of care side. And then your average hospitalized patient has about 15 of these codes. So you just imagine the combinatorial complexity, 15 codes, 30,000 options, each one that presents.
A number of years ago, the challenge with doing this well with AI is that healthcare data is very, very wide. There's lots of variables that are measured per patient, especially when they're in the hospital. The variables are all tied to each other. Your blood pressure and your heart rate are not actually two independent variables. They obviously physiologically relate to each other.
And I remember when we were trying to recruit a friend out of Google who was doing data science there he was like, "Okay so," and he worked in YouTube and he was like, "Alright great, so how many billion visits do you have that you can train off of?" and I was like, "Well, there's not a billion hospitalizations a year in the U.S.," so someone, he's like, "Wait what?" And so for a long time it was very hard to apply big data, data science techniques to healthcare where data is very broad, but actually not all that big. And I think we were, Josh and I were seeing some of the early technologies, like the early use of transformers and the ability for BERT models to be able to parse some of that clinical contextual data and thought that over time, as that gets better, this kind of unlock would be possible. We didn't think so deeply as to start OpenAI, unfortunately, but deeply enough that we thought this would be an interesting problem to go after.
Inside the AI Engine: How Smarter Prebill and Smarter Denials Work
Dr. Tim Showalter: That's the ideal point to hop right into your products. You've launched two main products, Smarter Prebill and Smarter Denials. Walk us through each of those products and how they're designed and what problems they solve for.
Dr. Michael Gao: The baseline world of hospitals is for hospitalized patients, is that doctors see patients, they document. And there's this entire group of people called clinical documentation improvement nurses. And literally, it's nurses who don't provide any patient care. And all they do is open up the documentation, review that, look at the underlying labs, medications, orders, vitals, the actual care delivered, and use the care to identify potential gaps in the documentation process. And that's the entirety of the job.
And existing software that these nurses used were taking the underlying data and trying to grade the probability or predict the probability that there was some gap in the documentation. So effectively it was generating a work list and a prioritization within that work list and that is effectively a software recreation of a manual workflow. Before they had a bunch of patient folders and they would take the thickest one and the messiest one and look at those first or pre-SmarterDX it was let's rank order the work and have you go through the charts.
What we do is we take the underlying clinical data and we directly infer from that underlying clinical data what diagnoses and procedures were done. And then use that to look for, hey, if the patient had a blood pressure of 180 over 100, they got a high blood pressure medication, high blood pressure should be documented or coded somewhere. Obviously, these are usually more complex, but use that to identify gaps. And then we surface that back to those nurses for them to effectively confirm yes or no. And the value is that where there are these gaps, we're changing the reimbursement for that visit and also actually changing how the quality of that visit is measured. So more complex patients have higher expected mortality and complication rates. So you actually change the relative level of mortality that a hospital is perceived to perform at. And then on the denial side, it's literally the opposite.
It is instead of looking at the underlying data and identifying a potential diagnosis, it is I, the insurance company, have rejected a diagnosis. I don't believe that it's actually justified. And we take effectively the same approach and pull in the relevant underlying data and pre-generate an appeal letter from that underlying data. And so, of course, with large language models, converting supporting evidence to paragraphs is relatively trivial. But the know-how here is what data is relevant to make that argument and being narrowly focused, which increases accuracy rates. And that replaces what is an hour or two of a physician advisor or other clinical person literally flipping through the chart and compiling that appeal letter.
From Margin to Mission: The Financial and Quality-of-Care Impact
Dr. Paul Gerrard: That's really interesting. And our understanding is that you've delivered millions in net new revenue to hospitals. And as you just said, you're able in some cases to go to a hospital and deliver more net new revenue than the entire cost of the revenue cycle management. Or maybe I'm not quite quoting it correctly, but something akin to that. I think you said greater than $50 million at one health system, $12 million at another. Do you have any sense on the impact of that? How are hospitals using those funds, and are they able to increase the quality of care they're giving or offer new services to patients, or in some extreme cases, maybe even keep their doors open, whereas previously it was tenuous as to whether or not they'd be able to remain solvent?
Dr. Michael Gao: One of our hospitals in Chicago, one of our customers had actually, before we started, gone bankrupt right before COVID hit. And closed their doors. And then an entire community in Chicago lost their acute care center at the time they needed it the most. And I think that the way I would frame this is that your average hospital margin across the country is about 1.8%. And so, from the people listening to this podcast, that's like at the end of your bi-weekly paycheck, you can save 1.8% after paying your basic rents and and food and utilities. And so we have a world in which the people who allocate healthcare dollars, i.e. the payers, actually have much higher margins than the people who are providing care with those dollars. And our mission really is just to tell the accurate story, but we view the secondary effect of evening that playing field between payers and providers as a positive outcome, Paul, specifically for the reasons you mentioned. Even if you're a non-profit, it's no margin, no mission. And when you're living paycheck to paycheck, but as a health system, investments in new services, investments in an extra MRI machine, every decision you make has such a higher ROI bar that it has to clear because you can't accept long-term ROI on your investments.
Dr. Josh Geleris: And just to put some numbers into context here, we find something on the order of 30 to 50 basis points so 0.3 to 0.5 percent more revenue for the health system and so that margin goes from 1.8 to 2.3 in some cases which is really pretty significant increase on a percentage basis of the value that of what they're able to save and what they're able to invest in the future with that. And so healthcare dollars are big and it's hard to move the needle sometimes. But because we are providing a service that for these providers hardly requires any time or expense on their part, all the value that we're providing goes straight to that margin for them and becomes essentially profit that they can use to invest in things like higher quality care, more resources, more programs for the populations that they serve. And I think we're really proud of that.
Dr. Michael Gao: Something that's kind of unexpected is only about half the time are we brought in by the finance side of the hospital. About half the time maybe 40 percent we're brought in by the quality side of the hospital and this is maybe counterintuitive or maybe it's all too intuitive for a bunch of physicians on the call, but the chief quality officer job is really tough because what you get told is you have to improve the quality of care for the hospital and then you're a cost center so you have no budget with which to do so. And this is actually for them a tool that helps them tell a better quality story but then they get to go to the CFO and say, "Hey, this is actually a tool I want to bring in that's a net revenue center for the hospital," and then they can use that as part of their budget to drive other initiatives. The check gets signed because it's hard to get a check signed without a revenue impact, but actually the buyer is often not the CFO, or the champion is often not the CFO.
Beyond Under-Coding: Building Guardrails for Accurate Documentation
Dr. Trevor Royce: Yeah, I really like the way you framed that. I think your phrasing was you want to tell an accurate story or your mission is to tell an accurate story. And I think that makes a lot of sense. A lot of the premise for this is that we are not accurate on the side of under-coding, I guess, for lack of a better term, correct me if there's a better term there, but what about the other side of the pendulum and over-coding? What guardrails are there? How do you right-size this so that it is accurate?
Dr. Josh Geleris: So we definitely do that too. And I think that's an important feature that we, an important service that we provide to our customers, which is we look at the diagnoses in both directions. And we often find things that are coded, but aren't clinically valid. And we present those to the health system for them to review and evaluate and decide whether it should get removed or not. So that is a service that we provide. In terms of how we evaluate whether our findings are in that over-coded world, A, we have a huge data science team that's focused on trying to do what we described, tell the most accurate clinical story possible.
We're also presenting our findings to humans in the health system. And we present the finding not in the abstract, but also with the clinical justification for it. So what the evidence for, as Mike mentioned, high blood pressure is. What were the blood pressure measurements? What were the medications that were given? That way they can shift their role from searching through what is tens of thousands of data points per patient to validating whether the data in front of them is supportive of the diagnosis that we're recommending. That enables them to be much more efficient in the types of review that we're asking them to do. But it also keeps humans in control and ensures that them using their clinical judgment are making the final determination as to whether this is a diagnosis that should be queried to a physician about or not.
And at the end of the day, that human is only one more step in the process. And ultimately, the physicians are the ones that have to make the update in their notes or respond to the query and are ultimately responsible for ensuring that the diagnoses that they're documenting are accurate and do tell the story. But we enable that process by giving them all the evidence that they'll need to answer the question in front of them of whether the diagnosis is present or not.
What's Next? The Vision for a Broader Smarter Technologies Platform
Dr. Tim Showalter: That's great. That makes a lot of sense. It sounds like you guys have worked out an impressive amount of solutions and I could see that you're solving for a lot of problems at the health system level. I'm also really curious about what's next for SmarterDX. I did read and actually congratulate you about the investment from New Mountain Capital and the creation of Smarter Technologies. I'm curious, how does this impact your mission and what's next for you guys with SmarterDX?
Dr. Michael Gao: We've been really excited about working with somebody who can help us go from point solutions or a set of point solutions to a real platform partner for health systems. And there's two broad trends with AI. There's digest more and more complex data and it'll generate insights from that data. And then there's agentic technology and how can we get a much, much more sophisticated version of what RPA tried to promise a number of years ago? And what's been really interesting about the construct is that Access Healthcare is on the surface a company that does outsourcing. It turns out to do outsourcing well because you're taking really low-cost labor.
You need to homogenize processes across customers and then set up really tight guardrails on what good looks like and really clear knowledge bases on how to do the process well. You can think of a call center as an example and like decision trees that get shown. You can only run a good call center if that's really clearly laid out. And so it's actually an incredible substrate for where to deploy healthcare agents because you've pre-built that substrate. So I think this is a really interesting fusion of these two big trends within AI. And for us, it's about using the clinical AI that, the ability to generate data from insights from clinical data to go into a number of other verticals and provide more and more values for our hospital partners.
From Clinician to Founder: Career Advice for Physicians in Health Tech
Dr. Paul Gerrard: I guess as we get close to closing out here, we wanted to ask about career advice. You guys are doctors. You've clearly done something creative and different with your careers. We have listeners out there who are also in the medical field, might be thinking about trying to make a difference, get into health innovation. What advice would you give them?
Dr. Josh Geleris: For me, I accidentally stumbled into this more than anything. I have a background in software coding from before medical school and was always looking for ways to use that to improve whatever I was at when I was in med school, when I was in residency, as an attending, and do interesting stuff.
And for me, I put myself in a position to do something like SmarterDX because I really spent a lot of time exploring different areas within the health system and trying to make a difference using technology that was the thing that I wanted to do. And I think in terms of advice, it's hard to predict what opportunities are going to come up, but I found that by knowing what I really wanted to do and putting myself in position to capitalize on that, I ended up with a lot of great opportunities. And so I don't think there's this concept of, oh, I'm a full-time physician. I'm just going to jump ship and go do something totally different. I imagine that does work for some people. But in my experience the people who are most successful at making a career change if that's what you want to do are ones that have been working at the thing that they want to do for a long time and are ready when the opportunity does appear. And so what I would say is figure out what direction you want to do. Being a doctor is amazing because you can do a lot of different things, but you also can use your time as a physician on the side to get really good at the thing in the area that you want to explore.
Dr. Michael Gao: If you don't mind me rendering a bit of a spicy take here, many physicians we talk to that have wanted to do other things generally say, "Great, I'm a physician, so I can give general clinical advice. And then, I don't know, I'm pretty organized, I can help with some project management." And I'm always like, "Do you have any idea the level of project manager you can get for like $300,000 a year? You can get a chief operating officer at that level. Trust me, you are not good at project management."
Unfortunately, then, just being able to give clinical advice doesn't differentiate you from hardly any other physician. So I do think that developing adjacent skills is super critical. Josh and I both know how to write code. I spent some time on the operating side of the hospital as well. And without those skills, we could not have made SmarterDX successful. It was a number of years of bootstrapping, putting together demos, working with data without any funding before things started to work. And so you don't need to be the world's best software coder. As Josh was alluding to, my stuff at NYP when I was working on it was just a small pile of Jupyter notebooks and some SQL queries. But I think it is where the more adjacent skills you develop, the more you can identify insights that are at the cross of multiple disciplines and bring a really unique set of skills and insights to whatever you want to do.
Dr. Josh Geleris: One of the things that we have noticed is physicians work really hard, residency is hard. We all struggled through that. And I think that's an advantage that we have is we know how to do that really well. We know what really hard work looks like. And so putting in the time and working hard in areas, I think physicians do have an advantage and can move into these things, but it does take that. It does take time. It does take hard work.
Dr. Trevor Royce: Well, it's been terrific to have you guys here with us today. I think that'll do it for today's episode. I hope everyone enjoyed the conversation. If you did, subscribe and check out the past episodes of Health Tech Remedy. Looking forward to seeing everybody next time. Mike, Josh, really appreciate you having on the show today and best of luck with everything.
Dr. Michael Gao: Thank you so much.
Dr. Josh Geleris: Thanks so much.
