A Guide to Medical Device Reimbursement

October 6, 2025 ░░░░░░

#427 A Guide to Medical Device Reimbursement

This episode tackles the often-overlooked but critical topic of medical device reimbursement. Host Etienne Nichols speaks with Haley King, co-founder and CEO of Paxos Health, about why this process is just as vital as FDA approval for a device's commercial success. They explore the journey a medical device takes, highlighting the distinction between FDA approval and securing reimbursement from payers.

Haley explains the three key pillars of reimbursement: coding, coverage, and payment. She delves into the complexities of CPT codes and the significant difference between a temporary Category 3 code and the gold-standard Category 1. The discussion also covers the immense challenges medical device companies face, including the lengthy timeline—sometimes years—to secure payer coverage, which can be a make-or-break factor for startups. The conversation wraps up with a look at how artificial intelligence is beginning to streamline the cumbersome, manual process of patient access and appeals.

Watch the Video:

Listen now:

Love this episode? Leave a review on iTunes!

Have suggestions or topics you’d like to hear about? Email us at podcast@greenlight.guru.

Key timestamps

1:45 - The initial challenge of making coverage match medical need.
3:30 - The three-part reimbursement pathway: coding, coverage, and payment.
5:50 - Navigating CPT codes and the difference between Category 1 and Category 3.
10:15 - Common challenges for medical device companies seeking payer coverage.
13:20 - The multi-year timeline to achieve Medicare coverage for innovative devices.
15:00 - Advice for regulatory and quality professionals on speeding up reimbursement.
20:10 - How AI is currently being used in patient access and reimbursement.
24:45 - Debating the accuracy of AI and its role in replacing human expertise.

Top takeaways from this episode

Integrate reimbursement strategy early: Unlike FDA approval, which focuses on safety and efficacy, payers also demand evidence of a device's clinical and economic value. Medical device companies, particularly startups, should integrate reimbursement planning into their pitch decks and product development timelines from the outset.
Recognize the two-step process: FDA approval is not a golden ticket to reimbursement. Companies must understand the subsequent and often lengthy process of securing coding, coverage, and payment from payers like CMS and private insurance companies, which can take several years.
Enhance clinical trials for payers: Regulatory and quality professionals can speed up the reimbursement process by designing clinical trials that not only meet FDA requirements but also generate robust data to prove a device's clinical and economic value. This may involve including additional endpoints to justify the cost.
Harness AI for efficiency, not replacement: AI is a powerful tool for automating the tedious parts of reimbursement, such as sifting through patient records and payer policies. However, it should be viewed as a way to enhance, not replace, the work of human experts who can handle complex edge cases and appeals.
Be aware of coding complexities: The distinction between a temporary Category 3 CPT code and a permanent Category 1 code is a major hurdle for innovative devices. Companies must be prepared for a potential multi-year journey to prove their device’s value to the American Medical Association (AMA) and payers.

References:

Paxos Health: Haley King’s company, which helps patients and physicians navigate insurance barriers for medical care.
Current Procedural Terminology (CPT) Codes: The standardized language used by healthcare providers and payers to bill for medical services.
Study on Payer Coverage: A 2023 study referenced in the episode, with authors including Josh Macaur and Dr. Erin Saxton, that found a nominal time of 5.7 years for innovative medical devices to achieve Medicare coverage.
You can also connect with Global Medical Device Podcast host Etienne Nichols on LinkedIn: https://www.linkedin.com/in/etienne-nichols-824a7114/

MedTech 101: The CPT Code Breakdown

Think of a CPT (Current Procedural Terminology) Code as a unique ID number for a specific medical procedure or service. When a doctor performs a treatment, they use this code to communicate with an insurance company (payer) what they did so they can get paid.

Category 1 Codes are like the established, official ID numbers. They're for widely accepted procedures with proven clinical value and are typically covered by insurance.
Category 3 Codes are like temporary ID numbers. They are assigned to newer, innovative procedures that are still being studied. They allow doctors to track the procedure's use, but they don't guarantee that insurance will pay for it. The goal is to collect enough data over time to upgrade to a Category 1 code.

Memorable quotes from this episode

"A lot of times patients are not going to be able to pay out of pocket for expensive medical treatments, and a lot of times providers are not going to be able to write off those treatments on their side. So somebody needs to pay for this. And that's usually the health insurance companies..." — Haley King

"I think that for this sort of a use case [AI], you're always going to want some human in the loop... AI has the potential to be super, super powerful in this space, but I think you're always going to want to have human experts involved." — Haley King

Feedback Call-to-Action

Your feedback helps us create content that matters to you. Have a topic suggestion or a question about this episode? Please reach out to us at podcast@greenlight.guru. We read every email and appreciate your insights!

Transcript

Etienne Nichols: Hey everyone. Welcome back to the Global Medical Advice Podcast. My name is Etienne Nichols. I'm the host for today's episode. Today with me is Haley King, the Co-founder and CEO of Paxos Health, a venture helping patients and physicians fight through insurance barriers to access medically necessary care.

In just their first year, her team secured more than 2.5 million in coverage benefits with a 90% denial overturn rate. You know, that's a. Correct me if I get the numbers wrong here, but first of all, how you doing today, Haley?

Haley King: I'm doing fantastic. Thanks for. Thanks for having me on.

Etienne Nichols: Awesome. I'll just kind of finish this little intro real quick, but Stanford Impact founder fellow Haley built her career at Medtronic and Cooler Heads Care before launching Access to tackle one of healthcare's toughest problems, making coverage match medical need, which I'm excited about.

So, I want to talk about medical reimbursement, how it works, why it matters. However, you want to introduce this topic, but. Yeah, take it away.

Haley King: Sure, yeah. So medical reimbursement and medical device reimbursement is something I feel like a lot of times gets missed until later, later down the road. And I think it's something that is so core and fundamental to a product ending up being successful in the market that I'm glad we're having this conversation.

It's great to talk a little bit more about, about why it matters.

I think the short summary is that a lot of times patients are not going to be able to pay out of pocket for expensive medical treatments, and a lot of times providers are not going to be able to write off those treatments on their side.

So, somebody needs to pay for this. And that's usually the health insurance companies, but the health insurance companies also have to figure out, hey, is this something we should cover? Is this something that is safe and effective from the FDA, but also is this reasonable and necessary?

Is this something that's the best thing for the patient? So, there's this interesting misalignment of incentives in some ways where patients and providers are wanting to get access to these technologies and sometimes payers are saying no.

And medical device companies that earlier on they can be thinking about these dynamics and helping to figure out, hey, how do we overcome some of these misaligned incentives and how do we show and prove out and demonstrate that our product is reasonable and necessary and that it is something payers should be covering.

The earlier you can have those conversations and be smart about how you address them, the better. So really excited to be talking about this topic today.

Etienne Nichols: Yeah. Cool. Well, I'm glad you have all this knowledge because this is something that affects so many people and I think the medical devices, the medical device companies themselves, hopefully they're thinking about it, but I don't know if they all are early enough.

And so, I'm curious if you could maybe talk about the process from. You know, a lot of times there's that big. I was at LSI last week where there's a lot of early-stage companies trying to get funding so that they can get their regulatory pathway through the FDA or whatever.

But when it comes to reimbursement, what does the typical reimbursement journey look like for med tech companies?

Haley King: Yeah, so it's quite variable. There's a lot that it depends on, depending on the type of product and like the, the. Yeah, essentially what pathway they're going to take to getting coverage.

But the general pathway is you usually first have to get FDA or regulatory approval. And so, with FDA, they're typically looking for making sure that the device is, is safe and effective.

But after FDA approval, then there's a question of reimbursement. And I like to split reimbursement into three main categories. There' coding, which essentially you have to get codes in order to basically use as the language to communicate back and forth with payers.

And then there's coverage. So just having a code is not enough for a health insurance company to actually cover a product.

And then there's payment, which is how much are you actually going to get paid out for this product? And so those are all kind of distinct steps in the process.

Which again, everyone talks about FDA approval and when I was at Medtronic, that was the main, you know, the main thing we were all working towards. And of course that's a huge, huge milestone and super important.

But after FDA approval, you then still have to have conversations with payers in order to work on getting codes, getting covered for those codes and figuring out how much you're actually going to get paid.

Etienne Nichols: I've heard a few different things about those CPT codes stretch my memory. Current procedural terminology I think is what it stands for. How do they fit into that pathway? And really, I'm curious about the different.

How do you choose those? Or do you choose those? Are they assigned? You know, I'm kind of, I'm A little bit on the outside. Here are the different categories.

Explain this to me if. Yeah, however, best, whatever makes sense as far as that goes.

Haley King: Yeah, yeah. So, CPT codes, you got it right. Current procedural terminology, those are assigned by the ama. And again, like I mentioned, that's basically a language that's used to communicate what are the different procedures and services that are, that are being requested to be paid out.

So, the first purpose is to just essentially act as a language and, and then within that there are many different types of codes.

So, for a lot of newer, innovative medical treatments, what will happen is you have to go through some steps to actually get to what is usually considered the gold standard, which is a category one code.

And ideally you get a category one code specifically for your product and treatment that enables you to essentially, Category one means that you have reached a standard where payers, or at least Medicare is agreeing that this is something that should be covered and should be paid out.

But what happens along the way is a lot of times if it's a more innovative newer product, particularly if there is not a lot of clinical evidence to support it, they will actually give you a category three code to start, which essentially is a temporary code for tracking the utilization so providers can still bill for it.

But because it's not actually covered yet, that second phase of coverage, there's actually, you know, there's some uncertainty about whether it'll be paid out. So, in a perfect world, without taking into account timelines and how long it takes to get a product to market, in a perfect world, you would set up trials in a way so that you make sure you have enough evidence upfront to support auto going to a category one code.

But depending on how things work out, you may end up in a situation where you have to start at a Category 3 code and, and then you're trying to fight your way to getting to that category one code.

Etienne Nichols: Yeah, I can see that being kind of difficult as far as, I don't know, really getting adoption across all the, the entire industry of healthcare, to use a third degree. How, how, I'm curious if you have any kind of idea how many there are a year or how it, how the utilization works as far as that goes. Any, any idea there?

Haley King: Yeah, that's a really great question. I, you know, I don't have the numbers top of my mind at this exact moment, but I do know that roughly every quarter they release new codes that have been categorized as category three.

And we'll also call out are these CPT codes are specifically for A treatment. So sometimes there's already an existing CPT code and you're trying to get your product to be covered in that existing CPT code.

So, there's a little, there's enough sub segments that I'm not sure of like an exact number to give you. But yeah, it's, it's a lot of every time a medical device company commercializes, they're figuring out can we use an existing code or do we need to go out and pave the way for a new code for our specific product.

Etienne Nichols: Okay. And I know there's gotta be hundreds and hundreds of codes at some point. It feels like it's almost insurmountable. I know the different coders out there; I've actually some friends who do this for a living but that's what they do that they just connect what the doctor's saying to the codes.

And I know that's their brain.

It seems like at some point it's almost insurmountable for human to do. What's the future look like? There's.

Haley King: Yeah, I mean I think this is a very good use case for AI and so it's something that you know is typically there are some AI companies that are and software companies that are trying to help providers offices with this.

And so there are companies that essentially can connect to the electronic medical record. So, a lot of times Epic or Athena, they can directly connect into that EMR. And then essentially after like when a new treatment is being requested, you can actually have these tools help you in figuring out what sorts of codes to be using.

It can pull out like hey, what were the diagnoses that we have based on that? Here's some of the different options. So yeah there's. This is a really exciting time in history for, for this, these, this industry because I think there's a lot of manual waiting through different documents to try to find what you need.

And I think it's a great use case for AI.

Etienne Nichols: Yeah, I would think even some of the ones that are lesser known, it's just, you know, there's gotta be some that just fall through the cracks like. Well, we don't have that for that.

You just don't know the code probably. But I can see that being a.

Haley King: For sure for sure. And it's coding errors are one of the biggest like, like when you look at the breakdown of why payers will deny some sort of a treatment, one of the biggest ones is related to either like missing documents or like errors in the submission as far as like what codes were submitted.

So that leads to, you know, there's a lot of downstream effects of that. Right. Like patients might be losing out or delaying their care as a result of some of these coding issues.

Because there's so many different codes out there. It's hard, it's hard for a human brain to get that right every time.

Etienne Nichols: Yeah.

Okay, so I guess if we moved away from the, that immediate, there's just a gap or maybe an irritation for medical device companies, not necessarily a problem you could solve today if you're a medical device company. But what challenges do those companies face?

They're trying to get payer coverage, they're trying to get those CPT codes, they've proven safety and efficacy. What are the things that they're going to face when it comes to getting that payer coverage?

Haley King: Yeah, so at that point there is actually almost a higher bar essentially that you then have to meet outside of safety and efficacy when you're trying to get payer coverage. And so, there's just a big challenge in itself, especially if you only planned for FDA approval and didn't think about, hey, what are we going to need to show to payers in order to go through. And so really, I think when it comes to the payer side of things, you're really looking to show that there's, this is a clinically valuable thing to do.

It starts to involve more of economics as well and looking at what are the other treatments that are out there to figure out, hey, is this actually the best thing for the patient or is there something else that's candidly way, way more affordable, that maybe has better clinical data?

And in that case, it's a little bit more obvious that, hey, actually, like, I don't know if we should cover this, this thing is better on paper and then also is cheaper.

So, we should go with that. Obviously, that's a more straightforward example. You start to get into a lot of gray area where something's more expensive, but also maybe it does have better clinical data.

And so, you're essentially in this push and pull with payers, trying to show them that your technology is reasonable and necessary. And so that's not only safe and effective, there's some clinical value to doing this over other treatments.

Etienne Nichols: Okay, that makes sense. Are there what I'm trying to think the, the how to word this question, I guess. But if I was to give them advice on what they should be doing, all the activities, I mean, it's pretty straightforward.

The FDA, you know what you need to do, there's guidance documents out There. What about when it comes to these payers? And when I say payers, we're talking about, are we talking about CMS, UnitedHealthcare, Medicare? Who all are we talking about too?

Haley King: Yeah. So, who we're talking about is usually the first step is typically Medicare is usually the first, which is administered by CMS. And so typically you're trying to get a code, a code through ama, but then get Medicare to cover your product.

Usually private payers like United, Aetna, etc. They will follow suit after Medicare coverage. So, they kind of look to Medicare to be the guiding decision maker of like, hey, is this something that should be covered?

But then I should say they look to them, meaning it's the first step in the process. And so, they wait to see an approval there.

Once they see an approval there, depending on the payer, it can take quite a long time for them to be convinced that they should cover it as well. So, there's kind of that first step with Medicare and then private payers tend to follow after that.

Etienne Nichols: Do you have any idea about how long it takes? And I'm guessing it probably depends on the therapeutic, I'm sure. But whenever I hear, well, you have to do this and then you have to wait for a while, I'm sure the CEOs or whoever's trying to get their medical device to market, they think, well, how long, how do I put that? My pitch deck, you know, what does that look like?

Haley King: Yeah, again, the more you do work earlier, the more you can try to speed up that process later on. But it's a grind, especially if you weren't thinking proactively about what data you needed to be able to show to this pay.

A study that came out, I believe in 2023 in, and I believe, I think it was the authors were, I know Josh Macaur was one of the authors and Dr. Or in Saxton, I think was another author and they looked at innovative new medical devices, and they found that the nominal time to get to Medicare coverage was 5.7 years.

And yeah, yeah, and I'll call out, that's for like, you know, quite innovative, newer, newer products.

But like it's real, right? Like, that's insane that it takes that long after, you know, getting FDA approval.

Etienne Nichols: So, if that's the category three. Sorry to interrupt. Yeah.

Haley King: Oh no, go ahead.

Etienne Nichols: Is that the category three process where you, you, you get that and then your temporary approval?

Haley King: Yeah, usually in those situations where it's quite innovative, they're going to make you go through, make you go through that process or some, some version of that process. But yeah, so, so again, that, like, there's nuances with that study related to like the, what the sample or, you know, what the population was.

But for new innovative products, it takes a really, really long time. And so that, you know, we're talking years, I've heard, sometimes 10 years. And you know, when you're a medical device company, and particularly if you're a newer, more of like a startup that's really reliant on this one product, like, that can kill you. I mean, if you don't have coverage, you know, providers most likely are not going to want to write off the cost. Patients are probably not going to want to pay out of pocket.

And so, if you don't have that figured out in advance or at least have some plan to get there, that's really, really tough. So, there's obviously, you know, I say five years. I know it sounds really scary. Again, if you do the, the upfront work and try to set yourself up for success with the data, it can be much shorter.

But it's definitely not like day after FDA approval. Cool, we're covered. Like, there is a process you're going to have to go through, which usually takes at least a year or two if you're doing everything right.

Etienne Nichols: And I just want to ask one more question on that. I know we have a lot of different pieces of ground to cover, but, okay, so you got a year. Does it just make sense?

You just build that into your funding or are there other ways to get.

Or any strategies? What do you recommend?

Haley King: You know, there are some ways to do it. I'm not an expert on this, so I don't want to, so I don't want to comment too much on it, but my understanding is there are opportunities and ways and like, there's like grants and there's also some like larger, like for example, like an academic medical center that's like really excited about using these new innovative technologies. Like, I think there's ways to try to like, make things work in those earlier days.

And then, yes, from a funding perspective, yes, 100%. You know, founders should be, and CEOs should be looking at, hey, there is going to be a time period where we are not getting paid out and how are we going to, how are we going to cover that and what are we going to do in the meantime?

Etienne Nichols: That makes sense. Okay, well, a lot of our audience works in quality and regulatory. They.

So, reimbursement. I typically, I guess I put that with regulatory as a Maybe, maybe marketing, medical device marketing, why should they be looking at. So, from quality regulatory standpoint, you know, our primary audience, why should they be looking at this and what are some things they can be doing to speed that up?

I mean, aside from just doing the research of whether the CPT codes or how does this apply to me, what's some of the, the actual work that they should be doing?

Haley King: Yeah, it's a great question. And as somebody who was on product development teams for seven years, we really did not talk about reimbursement much. And, and to be fair, I was, I was in a space where it was cardiovascular implants where reimbursement was less of a risk and there was predicate products that were, you know, set as up for success there. But yeah, it's something that is just not usually talked about on product development teams. And I think like as, as I've said, I think it's so core to the success of a product is making sure you get this right.

So, I think it's good for everyone to just be aware of the fact that there is a higher bar for, for getting coverage and getting paid out versus what you have for FDA.

So, from like a regulatory perspective, when they're providing input into, hey, how do we want to set up our clinical trials? And the question should not necessarily be how do we want to set up clinical trials to get FDA approval?

The question should be how do we set this up in a way where we get FDA approval, but how do we also set ourselves up for success from a coverage perspective?

And how do we, how do we have those conversations? How do we pull in reimbursement to make sure that we are like, like a reimbursement consultant for someone on the reimbursement side of things to make sure that when we're thinking about our clinical trials or our go to market strategy, like how do we do that in a way that's going to set us up for success? And then I think like, yeah, on the quality side, it's, it's, it's similar.

I think like you're essentially, you're trying to prove out that your product is safe and effective, but you're also trying to like be thinking, how do I show that this is a clinically valuable solution?

And I think when you're thinking about what your requirements are for the product, even as early as like customer needs like thinking through, hey, how do we show not only safe and effective, that this is something that yeah, is clinically valuable and economically makes sense for people to use I think that's.

It's hard when you're deep in the weeds of designing a product and in the lab just doing test methods all day. And I've been there, and I loved it. But, you know, being able to kind of think a little, a little bit more about the business side of things, I think ultimately will set the product up for better success.

Etienne Nichols: That makes sense. And I guess if, if we're talking about. If I just kind of go back to that clinical data, what are some of the things they can be doing? I mean, I, I can think. I can see you. You mentioned bringing a reimbursement strategist or a reimbursement expert onto the team just to understand that.

So really that's. It may be the same clinical evaluation, just changing a few endpoints or adding a few different things that. Is that kind of what you're suggesting?

Haley King: Yeah, yeah. I think usually that's what it ends up looking like. I mean, basically, the more rigorous you can make your trial, the better. So, like, obviously, you know, if you're able to have an RCT, like have it.

Have it be randomized and controlled, like that's a great first step versus, for example, you might be deciding, hey, do we want to go the 510(k) route and find ways to be smart about clinical strategy and maybe not have to do a trial?

I think that's obviously in many situations the right thing to do, but I think at the same time, you then need to be flagging. Okay, but wait a second. Even if we don't necessarily have to do this for getting regulatory approval, what does that mean from a reimbursement perspective? Is there enough clinical.

Has there been clinical value shown through predicate devices that we're going to be able to leverage for our submission as well? And so again, it's like this push and pull because it's like you don't necessarily want to slow down getting to approval just to do reimbursement, but at the same time, if you then wait and don't think about it until the very end, then all of a sudden, from an economics perspective, that it's going to get. It's going to be much more difficult to get market adoption.

Etienne Nichols: Yeah, that makes sense.

What are some of the. Have you. Do you. Do you see pretty common mistakes when it comes to this as far as just things that are going to slow you down if you don't do this right, or issues like consistent things?

Haley King: Yeah, yeah. I think the biggest thing is, is going back to maybe like sample size of the trial, how you set up the trial to measure like economics of it. Like if there's ways for you to be like again capturing that clinical value of what you're doing.

So, like just what you're measuring in the first place, how much you're measuring and then just how rigorously you're measuring them. Like the more you can put best practices in place for it to be, you know, randomized, controlled, all those things, like the better you're going to set yourself up to be going to payers and saying look like this is so obvious that this is something that's best for the patient from a safety and efficacy.

But also, this is better for you because it's going to do a better job and therefore you are not going to have this patient have to come back where you're going to have to pay again for some sort of additional treatment.

Like this just makes sense because the data is so much better.

Etienne Nichols: That makes sense.

You mentioned AI a couple times and it's a good use case for different aspects of reimbursement along the way.

How are teams in reimbursement and patient access?

How do they use AI today or do they use AI today?

Haley King: Yeah, so I think it's, we're in the early days which is, is really fun and, and really exciting and that's where you know, Paxos, we're trying to come in and be helpful here.

So typically, when, when a medical device company encounters these, these challenges, there's some upfront work that needs to be done in terms of strategy and, and that's a little less of like where we play and candidly at least today, a little less of where AI plays.

That's where you want real experts who have been through this day in, day out working with medical device companies to be giving you guidance about how should we be thinking about this for trial setup, how should we be thinking about our long-term strategy here?

Once a product is commercialized and you're actually out in the market, you then are having to fight on a case-by-case basis to get patients access to care.

So again there's, you know, sometimes it's up, it's five years for median for that one study. Right. And so, there's, there's this weird period where basically you were trying to help patients get access to these innovative treatments and help doctors.

Because if you're, if, if you know, if I'm a provider and I really want to use this, this new innovative product that I think is best for my patient, but then I learn that I'm going to have to have my office spend seven hours on paperwork per patient in order to like between the appeals and the peer to peer reviews and all that stuff, right?

Like, I'm probably gonna be like, hey, look, I think it's a great product, but I just can't make this work with my, with my practice.

And so, there's basically patient access groups out there that you can either medical device companies will either outsource helping this with this process, or the medical device company might build in house a patient access team to do this.

And what these teams do in either scenario is they help on a case-by-case basis, helping patients try to get access to this care by like writing prior authorizations, writing appeals, et cetera.

So just to give you kind of a little context on like, that's kind of what the, how these teams come into play. And like they're working with providers, offices typically and like trying to help, sometimes they're working with patients directly if it's a more like direct to patient sort of a product.

And they're essentially trying to provide help with like figuring out what do we need to submit, who do we need to submit it to, what should we include, what do we do if there's a denial, things like that.

So that's just a little bit of context on patient access reimbursement teams in general, where AI can start to play a role is that today what typically happens in these situations is you have to comb through a ton of patients clinical records to understand, hey, what is this patient's situation, what are their diagnoses, what have they tried before?

And then if you're doing things very thoroughly, these teams then also go find the specific policy for the specific insurance plan they have. So, let's say, you know, the patient has Aetna, you can go look up for Aetna, there's clinical policies specifically for different treatments they need.

And you can then go see, hey, for this policy, here's what Aetna requires that the patient must meet in order to be covered. And so, then what you can, you can start to see that this is quite a manual process.

You're going to multiple different locations, you're reading tons of documents, and then you're ultimately trying to make that connection between what does Aetna require versus do I have that from the patient's paperwork?

So that is done typically quite manually today. And hopefully as listeners are hearing this, they're realizing this is a great use case for AI because it's quite a manual cumbersome process.

So, one of the ways that AI is starting to be used in this space is it's capable of reading tons of different documents, including payer policies, and starting to make those connections of do we actually have everything we need in order for this to be submitted to the payer?

Is there information missing? Is there something that the patient should try first, like, you know, that might require a specific type of treatment before they will cover this new treatment?

So, AI can start to be used to work through all that messiness of all the clinical documents, plus tracking down payer policies and put things together. There’re many other use cases as well, like voice AI calling, which is super exciting.

But I would say right now the space is just done quite manually because it's always had to be done manually. And so, it's really exciting that we're at a stage where AI can start to be utilized here.

Etienne Nichols: Yeah.

You know, every time we talk about AI, there's two questions that seem to always pop up. And that is, well, is it accurate? Is it actually helpful? And then you got the other side is like, if it's actually helpful, is it too helpful and will it replace somebody?

And so, I'm curious what your thoughts are on both of those points.

Haley King: Yeah, so the accuracy is a really important question, particularly in healthcare. Right. Like, like this space has both healthcare, as I'm helping patients get access to care, plus like financial stakes as well. Right. And so, it's like two pretty important things that you want to make sure that you are covering in the right way. At Paxos, we like to say, you know, we love AI, but we are AI skeptics.

I don't know, maybe it's the Class 3 medical device implant engineer in me that's just a little bit more cautious about this stuff. But we want to get it right.

And so, with that, yes, AI can be, can be accurate, but it can also be really inaccurate if you don't set it up in the right way. I think we've all seen when you just go to ChatGPT, and you put something in and it just hallucinates a random paper that doesn't exist and says it quite confidently too.

Right.

And so all that to say, on the accuracy side of things, there are ways to make it more accurate and there's ways to build out tools and put checks in place to make sure that the AI is doing its job and put the guardrails on, but it has to be built in the right way in order for that to happen.

The second point about replacing humans, again, I We love AI, but I think that for this sort of a use case you're always going to want some human in the loop. And that's for a couple different reasons.

I think there are always going to be edge cases and gray areas, particularly with like innovative new medical devices and whether or not patients do meet criteria. Like there's situations where policies don't exist for a brand-new product that maybe has a code but hasn't gotten coverage yet.

So, you're always going to want to have a human involved in some capacity complexity.

So, the way we think about that then is we've built a tool that uses AI to try to eliminate those cumbersome tasks that I don't think anyone really enjoys doing.

Going through hundreds of pages of medical documents, going online, trying to find the right policy, looking at the criteria like that part of it, we automate that. But we set the interface up in a way so that the user has full visibility to everything that has happened, like citations, where they can click and see, hey, like, you know, the AI told me that the patient did not meet these criteria. Let me click right here to see what the criteria was directly from the policy.

Okay, now let me click right here to see the original source document where it highlights exactly where the AI pulled something out and made a conclusion about what the patient's situation was.

So, all that to say, yeah, I don't think like, I think AI has the potential to be super, super powerful in this space, but I think you're always going to want to have human experts involved to be doing those checks because it's such a high stakes thing that you're working on.

Etienne Nichols: Yeah, that makes sense. What about patient health information? Because when it comes to using tools like this, I can see people having an issue with that or HIPAA, HIPAA compliance.

What are your thoughts there?

Haley King: Yeah, so hopefully again listeners, no one go put Phi directly into, into the platforms, directly like just into the chat bot. That is not, it's not HIPAA compliant. It's, it's a, that data is, yeah, it's not a good idea to do that. But what you can do is you can get BAA's business associate agreements signed with these LLMs to work in a HIPAA compliant environment and be able to just basically call different APIs that are HIPAA compliant.

So that's one big part of it is you should not just be going to ChatGPT or any of these LLMs and just putting that information directly in.

But if you negotiate and work with these different LLMs, they do have processes to be able to get situation where you can use it in a HIPAA compliant way.

The other thing I hear a lot about LLMs related to this is like data privacy. Like you know, everything's getting. The model is learning from every, from every single call that it has right about. And again, use that to train, train the model.

You can turn off your settings for that. I do it for my personal one too just in case.

Yeah, like I don't like any sort of personal stuff I'm doing on there either. I do not want to have the model being trained on that.

But then you also there's just settings you can do particularly for like use cases like this where you just make sure that it's not being trained. So, there is a way to do it in a compliant way.

You just need to be very intentional about what you're doing and make sure you’re; you're being smart about how you set it up.

Etienne Nichols: Yeah, that makes sense. You know, it's funny because about two years ago, I guess when AI was really starting to be. Two and a half years ago be a little bit more prevalent as far as.

I just remember some different things that I was playing around with it, making funny videos or whatever. Didn't really take it seriously. I felt like the industry almost got, they kind of got AI fatigue and now the, I think we've gone full circle because now the FDA has their ELSA model and, and, and a lot of people are using it.

I want to kind of emphasize one thing you said about they think oh this, this one big AI model is being trained with everything that happens. And, and so that's why I, I don't, I sometimes I fall out of practice of this, but I like to use the word like AI's plural because not one big AI out there that's out to get us.

It's just all these multiple AIs that are, that are tools.

What are some.

You mentioned the tool you're building these medical device companies; I wonder are there ways they can start using AI's different models internally with their company to better utilize other AI tools like yours or interacting with FDA's Elsa or whatever the case may be.

Any other really high impact use cases for companies that might be listening. Any thoughts?

Haley King: Yeah, yeah. So, there's the whole patient access bucket that we talked about a little bit. One thing I didn't mention as much in that, that's, that is in patient access, but I think applies like to many different spaces including in the medical device world is voice AI Calling.

So, voice AI, like AI is actually at a stage where you can now build in scripts for what an AI should be saying to whoever it is calling, and it can respond to whatever the person on the other end is saying.

And so, you can track down different information. So that's really interesting in the patient access world because like if you, once you submit a claim you can actually typically you today have to manually keep calling the payer to be able to ask, hey, what's the status of this claim?

And so that's a really interesting area where voice AI can start taking that over and it can navigate phone trees, it can do a lot of things. So, you can imagine, I think that's just like a really fun one that like we're starting to see like is actually working.

I've always been very skeptical of that one but like data is now showing it's working. So, I feel like that has a ton of crazy, crazy use cases, but one that comes to mind outside of patient access related to AI. And I actually don't know the current state of this industry so I might be speaking a little bit out of turn.

But I do think at least from my time at Medtronic, there's a lot of work that goes into like looking at customer needs to customer requirements and then also looking at different standards and different FDA criteria and just things you're going to need to meet every time.

And so, I think there is a lot of use cases similar to what we're doing in patient access where you're kind of looking at what are the requirements that we need to meet and then have we met them or like how do we help this?

Put us put a process in place that's going to meet like minimum sample sizes, minimum quality efficacy in order to like meet these different criteria. So, I think there's some interesting stuff there that could be fun to explore further.

And then just day to day, I mean I use chat GPT for everything at this point, maybe to a fault. I'm not sure if I can write a real email anymore but, but yeah, I think obviously I know different, different companies have different policies in place for how you're allowed to, to use tools. But I do think it has the, the potential to really change at least from my experience at Medtronic.

Just day to day interactions, different documents you had to put together different SOPs, like checking that whatever you created aligns with different SOPs. I just see that the possibilities are endless.

Sorry, that wasn't a great answer. I'm just like, oh, AI is so cool.

Etienne Nichols: Well, yeah, the way I look at it is kind of like if you had an intern with a PhD, you know, what would you, what would you have that person do?

You know, it's just. And, and whatever you would have them do could probably have an AI do.

Yeah, yeah, that's. I mean, everything is probably not far from. Not far from it, I suppose.

What about, I don't know, the ROI on some of these things? Sometimes I feel like, well, it's novel, but it doesn't necessarily affect the bottom line, especially if you have to go back and check everything versus doing it yourself. When you're discussing patient access, if you were to build out a team to do those, all of those manual checks, I could totally see that where that would make sense.

What are you seeing with companies who are utilizing that sort of structure?

Haley King: Yeah. So, I would say there's two main value props that I'm hearing from folks related to using AI and patient access.

One is around operational efficiency. And just like, you know, one person being not having to manually go through all of those documents and instead just review a very slimmed down version of that.

And especially as you build confidence in the tool, maybe even not review, you're just like, okay, cool, we're ready to go.

So, there's some operational efficiency there and just making any one-person way more effective. And honestly, there's a side benefit of that, hopefully enjoying their jobs more. Instead of having to manually do all these things, they can focus on those edge cases, the more interesting gray areas where they're able to operate at the top of their expertise.

So, I think operational efficiency is one fairly obvious one, I guess, about AI and making things more streamlined. I think another one is the quality of the submissions.

And so, you know, I don't want to quite claim, oh, you can like increase overturn rates per se or increase approval rates in the first place for any one case.

But you can see how the logic works out that that could happen. Right. If typically, you know, yeah, sometimes these like patient access folks, because they have to get through so many cases in a day, they're using like templates where they're just, just copying in different things.

And it's a very similar submission to the payer each time with just patient name changed and a few things you can imagine with AI now all of a sudden every.

You can go find that specific payer policy for that specific patient, look at their specific documentation and generate this really customized argument as to how they are, they should be getting this treatment.

Right. And so, I Think there's some interesting things there just around like the quality of what you're actually submitting, which should lead to higher, you know, higher approval rates and ultimately get more patients access to care if they need it.

Etienne Nichols: Yeah, that makes sense. Interesting. Wow. Well, I'm glad you're doing all this. It's going to be really interesting. I can't wait to see how things change.

I don't know if you've read Marty McCary's book. I guess he's. Well, he's now the FDA commissioner. I was surprised when I, when that occurred. I was expecting something more along CMS, but I read his book a few years ago.

What broke the Price We Pay for Healthcare. What Broke American Healthcare and How to Fix It. And a lot of it is around, you know, just healthcare and all of the different ways it takes to get that, that payment.

And so, yeah, that's, that's awesome that people are working on that.

Haley King: Yeah. No, I mean like the system is, is definitely. I don't think anyone will say it is efficient. I think there, there's a lot of opportunities here and, and I think, you know, I hope one day that we don't need to do these arguments back and forth of trying to figure out and wasting time for patients to, to get access to the care they need. And so, my hope is with technology and, and with folks trying to spend time in this space and trying to make this a more efficient process on both sides.

You know, same for payers like I want. I'm hoping they're putting their best foot forward with trying to make this easier to get patients access to care.

Yeah. My hope is, is that in the long run we're able to get to some point where this is less of an issue. But in the meantime, we'll, we'll use technology and get some, get some folks on it to try to try to do a good job.

Etienne Nichols: Yeah, absolutely. And I don't want to put you on the spot necessarily, but I am curious what your, what your view of the future is for healthcare in America. I guess, you know, it's to isolate the conversation or put some boundaries up a little bit.

Do you. Being so close to all of those different things. I am curious if you have an opinion on what will happen. Not necessarily what you want to happen, but right now there a lot of people could point at a lot of things that need to be fixed.

And so, we could fix this part or fix this part. We have our tools. It's. It's difficult to fix an entire system. All at once.

So, I don't know if you have thoughts or. That's a little bit of a nebulous question, I guess. Just kind of a feature opinion.

Haley King: Yeah, no, I mean everything you said I agree with and that there is, there's a lot of things that need fixing and you know, I don't have a great answer as to like ultimately how we're, we're going to get there.

I do think there's opportunities for these different segments to be talking back and forth a little bit more. So, one example of this is I get asked a lot like, hey, isn't it just going to be like AI fighting AI with payers?

Like if we're doing stuff to try to help patients get access to care and then, and then payers are just denying using AI, then like what does that mean for the, for the future?

And yes, that's a pretty skeptical view on it. It honestly like warranted given our current health system.

Etienne Nichols: Sure.

Haley King: But I do see a world where we're trying to get to some ground truth about whether this treatment is the right thing for this patient, given what information we have and what evidence is available.

And so, I do think there's opportunities for like collaboration amongst these different segments that are trying to do things of like, hey, like let's get to some ground truth here and let's figure out how to make this a more efficient process for everyone.

Like payers shouldn't be like they're spending a ton of money trying to like go back and forth with patient access teams. So, I guess my answer to your question is I don't know at a full system level, but I will say that like, my hope is that people that are working in these different pockets and trying to make things better within their pockets, I think there's opportunities for us to kind of get on the same side of the table and try to find ways to get to what the ground truth is and try to get patients access to the care that they need.

Etienne Nichols: Yeah, it's funny, it's probably science fiction or something. I guess. I picture this where you wake up, your bed reads all your vital signs, your toilet reads all your diagnostics, et cetera. And then maybe you need follows you throughout your life and recognizes that you need this, that and the other. And so, it makes the recommendation to your doctor, your doctor does the thing, whatever, and that payment happened. There's a whole lot of assumptions in there and. But yeah, it's interesting.

Haley King: I would love that. Right. And it gets to like, it's just if we can all align on like, what is that ground truth of like, hey, like, you know, based on these diagnostics, this is what makes sense for the patient.

Like, there shouldn't even be this back and forth about whether or not we're going to cover it. It's just per the data logically, this is what could happen. And then, yeah, get more people access to what they need.

Etienne Nichols: Yeah, well, we'll have to see, you know, let me know how we can help to make some of this stuff happen. Because it's exciting future. I think so.

Haley King: Yeah. Well, likewise. Yeah, excited to, Excited to see where things get and yeah, excited to have to talk to smart people like you and hear more about what's happening in the industry and how we can all help move healthcare forward together.

Etienne Nichols: Any last piece of advice to those listening, whether it's on AI tools, patient access, reimbursement, anything, any piece of advice. If you were to sum up one thing.

Haley King: Yeah, I would just say be thinking about reimbursement. It's not maybe necessarily the sexiest topic at all times, but it is something that is really core to making medical technology successful and having that innovation and being able to move the industry forward.

And so, yeah, just be thinking about it. It doesn't have to be your day to day, exactly what you do every day, but it's important and I think the more we all are aware of it, the more we can ultimately help move healthcare forward.

Etienne Nichols: Awesome. Well, thanks so much, Haley. Where can people find you? What's. What's the best place to get in touch with you?

Haley King: Sure. Yeah, we. Our website is paxoshealth.com that's P A X O S Health.

You can also just reach me@haleyaxoshealth.com. so yeah, if anyone's interested in hearing more or just wants to nerd out on AI, please feel free to hit me up.

Etienne Nichols: Awesome. Thank you so much. We'll put that all in the show notes so people can find you. Feel free to look at those links so that you can check out what Hayley's doing.

But until then, we'll let you all get back to the rest of your day. Thank you so much for listening. We'll see you all next time. Take care.

Haley King: Thank you.

Etienne Nichols: Thanks for tuning in to the Global Medical Device Podcast. If you found value in today's conversation, please take a moment to rate, review and subscribe on your favorite podcast platform. If you've got thoughts or questions, we'd love to hear from you.

Email us at podcast@greenlight.guru.

Stay connected for more insights into the future of MedTech innovation. And if you're ready to take your product development to the next level. Visit us at www.greenlight.guru. until next time, keep innovating and improving the quality of life.

About the Global Medical Device Podcast:

Untitled (8.5 × 3 in)

The Global Medical Device Podcast powered by Greenlight Guru is where today's brightest minds in the medical device industry go to get their most useful and actionable insider knowledge, direct from some of the world's leading medical device experts and companies.

Like this episode? Subscribe today on iTunes or Spotify.

Etienne Nichols

Etienne Nichols is the Head of Industry Insights & Education at Greenlight Guru. As a Mechanical Engineer and Medical Device Guru, he specializes in simplifying complex ideas, teaching system integration, and connecting industry leaders. While hosting the Global Medical Device Podcast, Etienne has led over 200...