S3 | E11 | Cracking Clinical Coding with AI: How CodaMetrix Is Rewriting the Rules with Hamid Tabatabaie - CEO @ Codametrix

Welcome to the ThinkData podcast brought to you in partnership with DataWorks. If you want to stay up to date with the latest breakthroughs and trends in the world of data and artificial intelligence, And if you're curious about some of the strategies that companies and founders use to launch data and AI products, then you're in the right place. Our aim is to bring together a diverse lineup of fantastic guests.

from the founders through to accomplished leaders and product owners at some of the most fascinating data and AI companies worldwide. They will each offer you their own unique insight into what it takes to launch and scale a great data business. Thanks for tuning in, and I hope you enjoy the episode.

This is the Think Data Podcast in partnership with DataWorks, and today I'm really pleased to welcome Hamid to the show. Hamid is the president and CEO of Codametrics. They're a really interesting Series B AI-powered clinical coding health tech company. Hamid has over 25 years of experience within the healthcare technology space.

And it's really good to get you on today. I know we've obviously had a kickoff call prior to this and you're back and kind of speaks for itself. You've been at Codometrics, what come out too?

kind of eight, nine years now. So obviously, you know the space well. It's a hot space. Obviously, when you put AI up on any kind of solution, it obviously... turbo powers it but i'm really fascinated to dig into your background and also really get under the skin a bit more about codometrics and look healthcare tech huge market or huge kind of

target market but i'm also really interested to kind of get under the skin of kind of what problem it's solving as well so yeah over to you if you wouldn't mind not at all uh

It's been more than 33 years, actually. I'll give you a little bit of background on that. But I started out as an engineer. I went to Lehigh in Pennsylvania. first program in the country for engineering and medicine started at Boston University. And I moved there to become both a doctor and engineer. And shortly after my tenure, I switched to pure engineering because there was so much opportunity to do interesting things at the time. This was in the mid-80s where...

Epic and Meditech and Cerner and HBOC, now called McKesson, and others were all using mini computers and they were pretty much at the mercy of... the operating system and the capabilities of the hardware. So I got assigned to all these accounts and assigned to the health systems that they did business with inside of the... Boston Beltway called 128. So I learned, even in the very early stages of my career, a lot about healthcare IT. Now I've been a CEO for 33 years.

four different startups. And my focus has always been on a category defining capability. One of my earlier... startups was amicus. We were trying to switch from film being used throughout the health system by radiology to digital interpretation of images. as well as letting images be accessible across the enterprise with whatever PC or device they may have had.

That was a lot of fun. That was a growing company. It's gone on and survived after multiple acquisitions. It's still the same kind of product that we built. I'm very proud of that. The second one was life image. And that was when we wanted to do away with patients having to carry CDs and DVDs around. And that very much came from a personal experience.

My mother had Alzheimer's in early stages. We were shopping around for hope. So we went from one health system to another. And everywhere they repeated a CT exam because they didn't have access to the last one.

So that caused me to start LifeImage and is now the largest network for image sharing. Quotometrics, on the other hand, came to be as a result of my relationship with Mass General Brigham. They had developed something internally and asked me to take a look at it for commercialization purposes. I can tell you more about that, but that's my background. Fascinating, yeah.

I don't think we're going to get on to this, but I think, you know, you've seen so much evolve over those years. And obviously, on one hand, you've had that kind of... personal motivation obviously the cd the imagery um you know to set up that company then encoder metrics is quite unique in the sense obviously it was started inside mass general so obviously we'll we're going to revisit that because i think it's important to focus on it but as a ceo founder kind of

Those kind of taking a product to market and actually getting commercial viability and buy-in and then obviously earning revenue off the back of it is one thing a lot of organizations fall short of. They might have raised a serious amount of money, but ultimately... the product market fit is.

so much companies struggle with but well you know can you talk about you know the solution and you know when you had that kind of light bulb moment thought yeah we this is a commercial opportunity here and equally what steps did you then take to commercialize air? It is probably fairly different and distinct from one type of product to another, but there's a lot of common elements. I've always...

to start building a product with the customer, meaning not inside the vacuum of our own research and development capabilities. That doesn't mean you... take everything that the customer says as gospel and decide to create it because customers voice typically talks about today's problem. And by the time you solve today's problem. Tomorrow arrives and they may want things differently. So my first stage always is learning the market and or the people that a product serves very well.

The point that you can forecast what their future needs are within reasonable timelines. If you forecast too far out and you build something too far out, you end up starving on the vine. You probably are a good product, but ahead of your time to the point that you can't finance yourself in that duration. The second thing we do is never use a single customer to...

develop a product with, use at least three or four different ones, because in healthcare particularly, the practices are very different. Every medical school has trained their physicians. in their own vein, in their own culture. And as a result, the workflows reflect that. And automation and creating things that cause automation are very much... paying homage to that workflow. So once you do two, three, four, you get a baseline that you can establish some

commonality that becomes your basis of the platform. And then you create configurability and flexibility in products so it can be used by different workflow requirements. The really hard time is...

where to start in a market. In healthcare, there are probably three distinct segments for the provider side. One is the smaller... end of the market, and that is individual physician organizations or physician groups. That tends to be pretty crowded with lots of regional players that have grown up to be national players. So there are dozens of solutions and differentiation is very difficult.

The second layer is smaller community hospitals. And there, by virtue of who they are, they can't adapt and adopt. very complicated things very quickly. So they need to start with very much packaged offering, something that doesn't ask them to come up with. too many different variations and configurations based basically they want to fit into a mold

And then there's the large health systems and academic medical systems that is completely the reverse of the latter. And that was because they've had their own version of doing things. They have their own types of... research done and attracted particular physicians and particular patient populations. And their flexibility is very key.

But then the difference, as we talked about before, you have to have a baseline to start with. So to me, going to market requires you to decide which segment of the market you want. Yeah. Depending on the product, chances are you cannot succeed in all three at the same time. Sometimes... It's easier to start at the low end because the proximity to the customer is a lot easier. You can get a hold of the owner that is practicing from 10 to 4 o'clock in the afternoon.

But it also is competitive. So you have to decide on where your strengths are and where your superpowers are, whether or not that meets. I've always started from the larger part of the market working my way down. To some extent, the fact that they're difficult is a moat by itself. It's a competitive advantage if you know how to address that. It requires more financing.

and it requires longer sales cycles. Finally, Once you've decided what your market is, who the first three or four early adopters are with whom you work, then you have to sing the gospel. to everyone in that category. My focus has always been very much limiting to the accounts that are of interest to our product for the foreseeable future. even sales and marketing is done to a set of customers. Whether we go to trade shows or use...

lead generation activities, it's always within that window that we've constrained ourselves to be able to best serve our customers. Let me make sure I pause to find out if... I addressed the question you were asking. No, I think it's really concise. I think it's really interesting. Actually,

I like the fact that you flip this on this head and say, actually, it is a moat in itself by targeting the big players, the larger health systems. And I agree, there are those SaaS companies within health tech that will target the physician groups, will target not so much.

low-hanging fruit because they come with their own set of challenges but will be maybe an easier person to get hold of than the larger cells you know involved with the health systems i suppose on the flip side of that though you also have to consider the

the longer sales cycles, the procrastinated kind of procurement processes, the education piece. So how did you overcome that? Because As a startup or as a business that may be relatively unknown in comparison to some of the larger vendors out there, but how did you start to educate both the end customer, but also your sales and go-to-market teams to kind of...

had the right messaging? I think it all starts with you educating your own. and your own team as to who are the likely early adopters, not just because their name is interesting to you, but because the people you get to call on have had some history of early adoption. I've had some history of letting innovators in. So you have to do your homework and find out the person who used to work at this health system is no longer there.

but they have moved to other health system. And let's find out if they are still in the mode of championing. Then the larger health systems is very much about... Show me and prove it to me. Otherwise, it's in the pilot stage and you have not achieved the customer general release point of view. You have to decide which customers are in that mode.

By the way, these customers change depending on whether you're doing a clinical solution or a revenue cycle solution or some other category of solutions within their. a larger health system. We picked, as I mentioned, four early adopters. The way we promoted it was creating some incentives. Most of those incentives were that they would have a lot to say about what the product does and how it does it.

Interesting. We essentially call it a consortium because these were people who agreed to find mutual interest in the capabilities of a category that didn't really exist before. The second thing we did is we went after specific subspecialties and specialty areas. So we decided that, pardon the expression, but...

eating the elephant in one bite is not the right thing to do for AI. You have to send AI to medical school to learn all these specialties. So let's create a path of which specialties we want to address. or we have the best chance of succeeding, and then find the right champions for that. We chose a path to start with radiology and pathology because 80% of patients...

have some kind of radiology or pathology record in their records. So it made it easier to find those champions. There were a broader set of champions who would be signing up for that. appreciated that 80% of the headache comes from radiology and pathology sites. So finally, how we get to the market is by showing results in our world. Touchless automation means we are responsible for the answers. The answers generate variety of outputs, whether it's for billing or clinical use.

We advertised the answers. We advertised the success with automation rate, accuracy, turnaround time. And we somewhat got lucky in that. At the same time, COVID hit. Yeah, the labor shortage started. So people started needing to automate. It wasn't so much that they were... as impressed with what we had to offer as they were looking for something to solve the problem. And we stood out as the category leader and we took advantage of that wave.

it's interesting touching timing and luck is kind of when you obviously it's such a big event such as covid it's uh with such kind of unfortunate circumstances, opportunities. Certainly if we look at the healthcare space, they still presented itself, but healthcare system has become more efficient, you know, remove those kind of bottlenecks. And obviously with what kind of metrics were originally set up to do, I see why.

timing was kind of fortuitous and you know for both parties i'm interested in touch on the evolution of the product from mass general right at the top of the the intro you mentioned they brought you in to look at something they were working on can you think back to

what the solution was then versus... what it is now can you kind of summarize i know we touched on it briefly in the intro about what codometrics is but i'm interested to talk about that kind of journey from what you started working on back then to kind of the the current version absolutely First of all, as you know, and hopefully your audience knows, Mass General and Brigham are among the very top of the innovators in healthcare.

Their academic medical centers, in this case, Harvard Teaching Hospitals. And they spend an enormous amount of attention and capital on innovating new things. In fact, that's how they draw some of the best. physicians to come work there because there's a very interesting research angle, not limited to biotech, but also health tech. I met the solution for the first time in 2017. What had been developed was a very high-end workflow automation for coding. Okay.

The Mass General Physician Organization has funded the ability for their coders to code better. because the physician themselves were not just interested in the dollars, but also in the value of codes in the clinical sense. And they tried... variety of systems in the market from large legacy vendors, whether it was at the time 3M and Optum and so forth. And those had not met the standards they wanted to see, the quality they wanted to see, and certainly not the speed.

So this new product was developed over a course of 10 years. It did basically everything a coder, you can imagine a coder would want to do. A year before I met them. They had started to experiment with AI for touchless automation. Okay. Because some segment of the cases that need to go out the door. are repetitive enough that you would think early AI would be able to automate. And there was a lot of success there. So when I met them, my proposal for taking it to market was...

Let's emphasize the workflow as the least important component, the automation as the most important component. we were drastically underinvested in AI. Yeah. So we agreed to fund the company internally. as Mass General Physician Organization, so the doctor's money. went into a two and a half year rewriting of the platform to make it commercially viable and make it more AI centric than it was.

We did the consortium part that we talked about, and two and a half years later, we took the product to market with radiology. Worked with CU Medicine in Colorado, Yale, Mount Sinai, and so forth. Then we kept adding specialty areas. Then came pathology. Then came GI and surgery. And finally, you start to deal with things that have some inpatient connections, things like inpatient.

bedside consultation by physicians, ED, ambulatory, all of these specialties we think of as a certain roadmap. So we are pretty far down our roadmap at the moment, but... There's always a lot more to do. The product has evolved, I would say, in three different areas. One is its AI has continued to be... more and more advanced. You just can't wait anymore and rely on technology that was...

okay six months ago. This is a much more rapid adoption of technology environment. Second, we have expanded on specialties. amount of the cases that qualify for automation, because some cases are too complicated to automate all at once. And third is automation rates and quality. There's a good saying that one of our co-founders used to use. It says, if quality doesn't matter, we can automate everything.

So quality and automation have to rise together. Those are the three areas that over time, the product has gotten more and more advanced, more and more mature, and certainly a volume that goes through us. having 30 of the largest health systems in the country, millions of cases a day is pretty substantial and we had to keep up with the infrastructure needs. Yeah, it's fascinating. Obviously AI is your...

Touchdown has played obviously a key part of this. And I know people listening, obviously AI is not. ChatGBT a few years ago, it's been around for a while. It's actually been there behind the scenes. for a number of years. But I know from a health system and a provider standpoint, there is still reluctance. There's still this trepidation and nervousness. Maybe it's lack of knowledge or lack of trust around AI.

When you think about your kind of sales and go-to-market efforts, how are you kind of overcoming those, maybe the reluctance or the nervousness when you're talking about AI as a solution? Can you talk about maybe some of the...

Is there anything that kind of stands out where you think this has been a struggle for us and this is kind of ultimately how we've over had to come then? Certainly. Over the last few years, What we have witnessed and the market has shown us is that healthcare providers are much more comfortable to adopt and adapt. to non-clinical use of AI as an early stage experimentation. Interesting. The labor shortage.

costs have speeded that up for adoption to the point that there was a survey done by class. Pretty famous research organization, healthcare. Nearly everybody in the revenue cycle side of things that is highly manual today is looking for AI-based automation. The challenge has always been education of the end user. In our case, revenue cycle. has always been a people-heavy environment, not necessarily people who are well-versed in technology. They're also within the pecking order of

who gets attention from the technical resources, revenue cycle isn't on the top. But cash collection is. So they had historically just thrown more people at the problem. Now we need to educate the market that there are two parts to this AI path. Part one is automating as much as you can. the same things that quarters have more success getting right, but it takes too long. So the long hanging fruit as if it were.

That has become known as autonomous coding. So basically you've replaced how a coder does things. The harder thing that we've always had on the inside developing, and that was the mission from Codermetrics from the beginning. is to take clinical sensitivity into account. So instead of coding well enough to get paid, coding much more contextual. We call that contextual coding. In that sense, Codometrics is the revenue cycle intelligent operating system.

So we take a look at the whole patient and then quote the case accordingly. That has been a difficult component because there are different buyers in... the healthcare system. The revenue cycle, people perhaps rightly so want to focus on the job at hand and that's get the bills out the door. Yeah. As a result. Many people code things again for population health and clinical decision-making and research and so forth, but because they're not the same users.

Cost centers are different. So we've spent an enormous amount of time educating the market as to what is an enterprise coding. Yeah. Why would you not code once and use many? AI lets you do that. It is more involved. It requires more collaboration with the customer. But that has been an interesting challenge and we're overcoming it. But we knew going in that that's an area we had to invest in.

Yeah, it's interesting. You touched on something really fascinating there about the mission from the beginning. I think a lot of founders I speak to, you know, the successful firms, the ones that got their mission, yes, that kind of... It's the path you follow. But as long as your kind of North Star is very clear, then actually ultimately your end goal is going to be achieved far more successfully. And actually with the healthcare provider space and health systems, I think they...

And yeah, the workflow automation, the removing of bottlenecks, the kind of helping free up staff time is where Codometrics is really kind of focused on, which I really like. And I know you've been in the space for a while now, but... If you look at kind of AI as a, I suppose, as an evolution of kind of where it was kind of 10, 15 years ago to now and what this means for code and metrics, how do you sum that up?

It's becoming commoditized pretty fast. Yeah. It isn't yet. It is not affordable at a commodity level. No. It is discreetable. distributable at a commodity level. Anyone can log on and get lots of AI capabilities. And it's becoming better understood by the technical staff. It used to be that AI engineers were the lonely people who were in a corner somewhere, and now they sit in the center of their own. The whole concept of data-driven informatics.

is very different than workflow-driven informatics. And unless you are data-driven, AI won't be terribly helpful. What I've seen change drastically is the evolution of the thinking of becoming more data-centric. It is by no means pervasive, but... Healthcare moves pretty slowly. It's a $5 trillion market and it doesn't move very fast. So you need to be very selective of where you apply AI.

Otherwise, you end up in the hype cycle. And as a leader of the company, you will end up having more frustration than you would if you don't pick and choose the early adoption. adoption or early adoptable solutions while you're looking at your North Star. So we want to do this code once and use many. But we hit the wall of, hey, I'm revenue cycle. Don't confuse me with the facts. Get things right. So we design things for the future. But then.

sell it and provide it and support it as adoptable components. It doesn't lose its merit for us. We become the category leader. It's easier to buy things from us. We can solve tomorrow's problems. But at first, your question was about AI. AI hasn't changed as much. the adoption curve in healthcare yet. Yeah. On the administrative side, it's about to.

We're about to get there within the next year or two. I remember when I did one of my first companies on the turning films into imaging. First year. Or two, it took maybe a percent of the market adopted. By year five, maybe 5%. But by year seven or eight, 90%. All of a sudden, you couldn't do without it.

You've got like two worlds colliding here, haven't you? Because you've got this race at which AI is progressing and the pace at which health systems and providers typically embrace new technology. And it's almost shaking them up to think that actually, you know... they're having to move quicker to, you know, you know,

move with the times and what can we expect to see from you guys then because i know you know your your business is evolving rapidly as you just touched on there but over the next kind of with this month or years what can we as kind of in the u.s can people

start to see from codometrics coming forward? I think the reality is that healthcare encounters start with care and in code. Yeah. And the reality is codes are the backend and the language of healthcare. Yeah. If you get the codes right, not only you get paid, but also you can...

do prior authorization faster. You can do denials management better. You can get research done much more quickly. What you can expect from us... is this idea of being the revenue cycle intelligent operating system is essentially showing the market how you have to graduate from autonomous coding to this contextual coding world and what the requirements are. We will start that in the next month or two at an upcoming HFMA.

conference and speed that up over the course of the next year. You can see new modules from us. You can see new specialties from us. And perhaps you can see new services from us. A lot of our customers struggle with keeping up with the volume, even of the segment that doesn't get automated. And it stands for the reason to take a look at. offering that strategically. We don't want to become a coding company, but we do want to solve our customers' problems.

Fascinating. Yeah, I think it's no surprise that you've had the successes that you have from a growth standpoint, from an adoption standpoint. And it's unique where it was conceived and actually how you've kind of commercialized it, but also the problems you're solving and obviously with AI.

such a kind of a huge part in that. Now, it's been great to have you on. Maybe in parting words, suffice it to say that what Uber has done to transportation is that... They can deliver pizza. They can pick you up. They can pick up your dry cleaning. It has become an operating system for getting someone to do something for you. Similarly, Harvey is doing that for legal.

Sprinter is doing that for care at home. We want to do that end-to-end for all of the coding inside a health system with a lot of accuracy. I will show you what that means on the operating system side. Fantastic. We've got no doubt you'll do that. Yeah, thanks so much for coming on this morning. I really enjoyed that. Thanks again. I appreciate all the time you've given me and it's been a fun discussion. Thank you. Thanks, David.