Healthcare Is NOT About Care - Part 1 with Dr Timothy Martens

What if the healthcare system was never built to make you healthier?

In Part 1 of our conversation with Dr. Timothy Martens — congenital heart surgeon at Northwell Health, Director of Data Strategy & Innovation at Cohen Children's Medical Center, PhD in Biomedical Engineering, and General Partner at Picap Fund — we get into the structural truth most clinicians won't say out loud: the current system is optimized for billing, not care.

Dr. Martens breaks down why episodic 15-minute visits can't move the outcome needle, how wearable technology and direct-to-consumer biomarker testing are converging to create a parallel care layer outside the EMR, and why AI-powered startups are outpacing hospitals and universities at actually solving healthcare's hardest problems.

🧠 What you'll learn in Part 1:
→ Why Dr. Martens calls himself a "Technology Fedaykin" — and what the Dune reference reveals about his mission
→ The real reason healthcare workflows are structured around CPT billing codes, not patient outcomes
→ Why pediatric congenital cardiac surgery attracted a data-obsessed biomedical engineer
→ The "great convergence" of EMR, wearable, lab, and genetic data — and which companies are winning it
→ Why predicting heart attacks and Alzheimer's risk is already possible — and what's still missing
→ How the same dopamine algorithms destroying kids on TikTok could be flipped to drive healthy behavior

⏱️ Timestamps:
00:00 — Intro & Dr. Martens' background
04:30 — Technology Fedaykin: the Dune philosophy behind healthcare innovation
06:00 — Healthcare is designed to maximize billing, not outcomes
07:00 — Why pediatric cardiac surgery attracted a data engineer
11:00 — EMRs are built for billing codes, not continuous care
16:00 — Wearables, biomarkers & the great data convergence
22:00 — Can AI actually predict a heart attack? What's still missing

🎙️ IDAAHub Podcast: AI in Finance & Healthcare
Part 2 drops next — subscribe so you don't miss it.

#HealthcareAI #AIinHealthcare #DigitalHealth #HealthTech #MedTech #IDAAHub #PodcastEpisode #HealthcareReform #WearableTech #PredictiveMedicine

Transcript

SPEAKER_00 0:15

Hello everyone. Welcome to Lighthealth Podcast, AI Healthcare. Today I'm extremely excited to introduce Dr. Martens. And Dr. Martens is a surgeon. He has a PhD in biotechnology. He's also an investor. It's really hard to introduce you, Dr. Martens. And you could also tell us what role you love the best out of these three. Can you please introduce yourself?

SPEAKER_02 0:48

Sure, sure. So my name is Tim Martens. I'm a congenital heart surgeon for Northwell Health. I'm the director of transplantation and mechanical circuitory support for the Northwell system. And then I'm also the director of data strategy and innovation for the Children's Heart Center at Cohen Children's Medical Center. I that's my clinical role. I've always had an interest at in the intersection of health care and technology and trying to find ways to improve pretty much every aspect of the care we deliver. Quality, consistency, efficiency, you name it. Because I'm a believer that you can merge technology and humanity and come up with something that's better than uh either on its own. I took a pretty circuitous route in my training, so I have a PhD in biomedical engineering. I was involved with some of the cell therapy efforts to try and stimulate my cardio regeneration when I was in the middle of training, which was fascinating uh benchwork, but had some pretty severe limitations as far as trying to commercialize and scale that really still haven't been solved even today. Um then um uh so there's the whatever half of my brain is involved in surgery, um you know it's very satisfying, but it's there's a uh an untapped side that uh is is involved more in research and innovation and investing. And so um I think the best way to bring technology to market isn't necessarily through always academic or federally funded research channels, sometimes, um especially when the technology is very rapidly evolving. Um the whole startup ecosystem is much more effective at rapid iteration and um development than some of the more traditional pathways. And I think we're seeing that play out today uh with AI, especially, where um, you know, the the regulatory constraints and the legal constraints of trying to innovate within a healthcare system, uh I think um even a very forward-thinking one can still limit uh the pace. Now, obviously healthcare systems in general are very risk-averse, just like the FDA is very risk-averse. Their job is to protect people, not necessarily uh, you know, to innovate at all costs. And so there's there's a balance there that still has to be managed. But if you want to look at what's been built and who's building it um most effectively, uh I think more of it's coming out of the startup ecosystem than is coming out of universities.

SPEAKER_00 3:46

Yeah. Um, I guess you are completely um not just about surgery or healthcare, you know a lot about technology and uh uh also the startup landscape with AI currently, right? Um I know you mentioned you talk about yourself. I mean, there's a LinkedIn headline which I saw technology fedicin. So um that's um a unconditional commitment to a cost. So what would you describe this cost to be? Um, I I yeah, I I think um um is is uh I I guess you described it before in terms of um mankind and this, uh but I I know what I'm referring to is the technology fed again.

SPEAKER_02 4:39

You you have it up there, so yeah, yeah, so yeah, okay, so um um it's a reference to the the Dune series, um and in the in the books and the movies uh there were um zealots, uh there, you know, whatever word you want to use from intensely loyal to the emperor, uh with this inner circle uh bacon. Um and so I guess I I mean that in the context of exactly what I what I said before, you know, the there are technological solutions to a lot of the challenges um within clinical care delivery. Um and whether it's you know workflow automation or revenue cycle management or patient-facing things, device development, software algorithms, you know, the the landscape is changing. The driver of that change is technology. And um sometimes it means taking apart, dismantling, and reimagining very old ways of delivering care that are um often meant to maximize billing, not necessarily deliver the best, most efficient care. So um because some of these pathways are very, very entrenched, you I mean, it's not like you can say, okay, this is a better way of doing it, let's do it this way. There are so many stakeholders who are maximized for the current bloated, inefficient system, and so trying to change that threatens their livelihood and isn't uh isn't always well received, I guess would be the best way to state it. But you have to push through that resistance sometimes.

SPEAKER_00 6:32

Yeah, yeah. Um it it's very interesting how you took that technology fedic in word from Dune thing. It's uh um I I now we talked about this a little bit. I mean, you talked about this a little bit in terms of uh you're a pediatric um cardiothoracic uh surgeon. So what made you go in that direction? Because a lot of them don't think of that branch as such.

SPEAKER_02 7:02

Um part of like uh I get bored pretty easily. And so um what drew me to cardiac surgery was the kind of real-time minute-to-minute changes, the need to synthesize a lot of data and respond quickly, the being able to see like you can read about anatomy and physiology, you can learn about it in medical school, but like if you want to see it play out in real time, you know, start titrating drips like epinephrine and dobutamine on a patient who's critically ill and watch what happens to their, you know, cardiac index and their whatever their HBR, like it was it was the same anatomy and physiology that's kind of stale in a book on a page and then comes to life when you have to understand it to make decisions that literally are sometimes life and death. Uh so that part of it was fascinating to me. It was a little bit uh akin to like a sound engineer that's trying to record music has a mixing board with 64 sliders, and you know, you can record the same Tam instruments, but the difference between something that sounds great and is balanced and well produced is all in like how you how you manipulate the settings and um I don't know, I'm trying to think of another good analogy. Like sometimes the patients are like cruise ships, like you're not you're not a very nimble airplane that's gonna be doing loop-to-loops and flips and zig and zag. It's more of a combination of what are the signals now, what adjustments am I gonna make, and then you know, over the next minutes to hours, how are things gonna turn around? I mean, there are a few things you can do to drive a blood pressure up to 300 or down to zero, but for the most part, it's um taking a whole bunch of input signals and then figuring out how to optimize the system, um, almost like an engineering problem, I guess. Um, so that part was appealing. The bypass circuit, like having this machine that sits there and it takes over for the heart and lungs and is literally spinning, you know, 10 liters a minute of blood around in circles for hours and and and sustaining life while you you know you work through the technical details of a procedure. Um that part was interesting. And then why pediatrics? Because no two operations are the same, you know. They're you know, it's I had several mentors who were incredibly well trained and experienced and still, you know, on a pretty recurring basis basis would say, like, you know, I've been doing this for 20 or 25 or 30 years and I've never seen this, or you know, maybe I saw this 10 years ago and I have to go into the depths of my treasure drone, but there's a lot of variation that requires um something more than just machines like brutal efficiency. Sometimes there's a lot of judgment um in pediatrics. So that part was appealing. And then obviously it feels good if you can intervene on something that would be otherwise fatal and restore a normal life expectancy and quality.

SPEAKER_00 10:04

Yes, I know. It I I guess it's one of the hardest branches probably to get into. And um, because like you said, no two patients are the same. Um so when we talked last time, you mentioned we talked touched briefly on this. Healthcare is not about care um or um the current healthcare assistant. So would you like to elaborate a little bit on that?

SPEAKER_02 10:37

Yeah, so I mean take primary care as an example. Um the concept of episodic, whether it's every three months, six months, twelve months, doesn't really matter. But the the typical workflow is set up to fill the clinic with as many 15, 20, 30 minute visits, whatever billing code you're you're trying to satisfy with the documentation or the time-based billing, but it's all about pack the patients in the clinic, turn them through the mill. Um, and I I I don't think it's rational to expect outcomes to change very much if your touch point is that 15 or 30-minute visit. And I mean, we've all been through this, whether it's on the physician or the patient side, you you know, you get asked some questions, some version of a partially complete or all the way complete, depending on you know, the visit history is taken, and then some recommendations are made, you know, exercise more, lose some weight, eat healthier. Uh, and then you say, okay, yeah, sure, go get your screening XYZ. And then maybe there's a little bit of follow-up to make sure you do that. But for the most part, you're back out on your own, and in that interval between that visit and the next visit, not a lot of touch points, not a lot of interaction, not a lot of guidance. And uh I think now that we have access to wearable data, um, you know, chatbots, text messaging, interactions, the concept of continuous care, what happens in between the visits, all the little decisions that are made on a day-to-day basis that I think are just as important, if not more important, in driving whatever outcome you're trying to achieve, that has been largely ignored by traditional medicine. And so you know, the EMRs are set up for episodic visits with documentation to satisfy a billing code. They're not really set up to maximize care delivery or outcomes for the patient. That's an afterthought. If you can pull the data to show something, great, but they're designed for billing. Um guidelines, you know, are published based on data, but there's no real system in place to try and enforce them. Um and this this isn't like it's this isn't 100% to say the system, the healthcare system, the provider, the hospital is to blame. Like there's also a profound lack of accountability on the patient side for a lot of these things. But if you really, really wanted to optimize outcomes, whether it's you know cancer through screening and looking at all the environmental factors and diet and other things that go into it, or it's heart disease or obesity, doesn't matter. Uh, an approach that was continuous, technology guided, and patient empowering, I think is going to be far more effective. And all of those things sit outside the current revenue capture system. And so there's not a lot of incentive for someone motivated by profit to address those things within the health system. The health system profit gets maximized by billing codes, CPT codes. So this goes back to the earlier point. So now you have startups that are operating in that space that are responding to consumer need or are trying to show that they're gonna be better at moving the outcome needle. Um, and that's a good example, I think, of technology that's gonna drive change needed in healthcare, not because the healthcare system wants it, but because, you know, a second group of people has figured out how to not just make money but also drive outcomes more effectively than the traditional method, if that makes sense.

SPEAKER_00 14:21

Yeah, it does. You're totally right. I mean, traditional medicine has always been like whenever the patient goes to the doctor, that's when they see them. There is no continuous care, and a lot of things happen at home and not when they are episodic and it's paternalistic. Yes.

SPEAKER_02 14:40

Whether you whether you get confrontational and tell the patient you have to do this, or you say, here's my recommendations, and you kind of you know leave it to whether or not it actually gets followed on, but it's it's just not um it's a model that needs to be reimagined.

SPEAKER_00 14:55

Right. And you're you're right, a lot of this is being driven by revenue, and at-home care is not revenue focused, and not many are gonna make that.

SPEAKER_02 15:07

Well, it's hard because the existing codes, we have some for RPM now. We have we're starting to see codes to try and incentivize isn't the right word, but like if you're gonna do the work um to to try and pay for it. Now, early on, there was a lot of funding that went to companies that would diagnose or flag things, um, but did not have an action arm or a lever to change the outcome. And so there's been a pushback, there's been a lot of carnage in the digital health space because some of the early companies were able to raise and build without demonstrating that they could change outcome. And now that the you know, the funding climate has gotten tight, um, you know, there's a lack of liquidity from exits, there's a lot of factors going into it, but now it's much, much harder as a startup to get funding unless you have a direct tie or proof of concept that shows you're gonna move the outcome needle, change, you know, deliver real ROI, not hypothetical ROI based on a bunch of changes that may or may not happen. Um so yeah, yeah.

SPEAKER_00 16:18

No, you're right. I mean, we'll we'll get to the funding part, but I I would I wanted to understand from uh now that we do have variables and AI is everywhere, there's a lot of data we are collecting from variables, from um EMR. So, how much of all this data is has actually become clinically valuable, you would say?

SPEAKER_02 16:44

Yeah, so there's a couple different facets to unpack there. So there is a tremendous amount of data out there. It's you know, all the continuous pulse ox, heart rate, all that stuff uh is inherently noisy. So there's some difficulty in denoising and and and kind of normalizing and making the data useful. And there are companies that do nothing but that. Um there are still issues with silos, you know. So traditionally the EMRs would hold on to data and not want to release it. Um and there's legislation, TEFCA guidelines coming out to try and enforce interoperability, but they've been very slow and until recently with really not a lot of repercussion for not following them. Um but I I think what's happened as as the EMRs are slowly being forced to open up or to at least share. Um there are intermediaries, the QHINs, that are trying to take these massive JSON files or massive data that, you know, if you're an EMR, you say, okay, here's 15 terabits of your file, have fun with that. Like there are companies that that have the infrastructure to process that and do all the things we talked about, deduplicate and normalize and make you know make it digestible. Um, but independent of that, uh there's what I call it the great convergence. You know, the the more and more wearable companies are out there trying to take the wearable data and show it to the end users. And in the beginning, it was kind of early adopters, biohackers. Like if you're a marathon runner, you would get a garment and then you would look at your heart rate variability and like get a whole bunch of feeds of like you know, your trends of weight and your step counts and your cadence and all of this data. But it was super specific for kind of narrow use case. Like in the I'm an endurance athlete, and these are the things I want to track. So as we've as more and more of that data has become available, now Apple Watch, Aura, Whoop, Fitbit, um, there's more and more interest from not the biohackers and the early adopters, but just the general folks out there. Um and what these companies have to figure out is how to take that massive unclean data and then structure it, present it, show it in a way that's meaningful. Um and that's all happening outside the EMR. So there, I you know, the wearable data is like these continuous feeds, separate from the health traditional episodic data that sits in your EMR. And then you also have um the laboratory data. So in the you know, the old system, if you if you needed your blood checked, you went to the doctor, the doctor gave your prescription, you went to the lab, maybe insurance paid for it, maybe you paid for it out of pocket, and then you got a whole bunch of numbers that didn't mean much to you. You went back to the doctor and said, okay, this is normal, this is a little high, this is whatever. Um, more and more now we're seeing companies that are in the wellness space are going direct to consumers saying, hey, like if you want, if you have a family history of heart disease for 200 bucks, it might not be covered by insurance, but we'll give you a comprehensive biomarker panel that'll tell you what's your fitness and your biologic age and your risk for this, that, and the next thing. So uh, you know, I put them in another bucket, the kind of lab wellness longevity space. Um but if you look at what's going on, the EMRs that are traditionally siloed on the inpatient side, a little bit more fragmented on the outpatient side, don't have access to that data. But the consumers are getting that data either from Lab Core, Quest, from startups focused on different parts of the biomarker market. And then they're also getting the wearable data. What's really useful is if all of that stuff was brought together. And now there are companies doing this too, a bunch of them operating in this space where you know they're saying, okay, we'll take your EMR data, we'll take your wearable data, we'll take your laboratory data, we'll take your genetic data if you did 23andMe or whatever, and we're gonna put that all together and say, okay, here's your data. And um are focused on specific niches there. Some are, you know, the LLMs, even they're super generalist, but you know, open AI, anthropic within days of each other, offered, you know, APIs and MCP connections to your stuff so that you could just import all your own health data and then start asking questions. So all that is part of the same trend, um, and goes back to the first point, which is really the owner and purveyor of the data should be the end user. It should be the patient or the customer, however you want to look at them. Um, and I think the companies that are gonna win and the strategy that's gonna win isn't like that. You if if you're fighting that battle on the data integration side, I think you're in the wrong place because more and more data is available um at cheaper and cheaper cost. Um and just taking that massive chunk and saying, okay, here's all your data, that's not helpful. That's not meaningful. I think what's needed is the next layer that takes that data and says, okay, in the context of you, with your specific background, with your lifestyle, with with the choices you've made so far. We're going to take the numbers. We're going to take all of that. And then we're not going to show you your heart rate, you know, second to second for the last two years. You know, we're going to we're going to combine your heart rate with other information, maybe your genetics, maybe your medication history, maybe your your visits, whatever it is. And we're going to say we put that all together and here's where you are. Here's what's driving it. And here's what you need to do to change. Like your whatever, your BMI is high, your step count is low, like you're not very active, your diet stinks. Here's three things you can do every day to lower your risk of heart disease, to lower your cholesterol, to bring your your hemoglobin A1C down, to make your glucose control tighter, whatever it is, the the technology becomes a tool to empower the patient. And I think that's a much, much more effective approach, provided you can solve the engagement problem than I'm going to see my PCP every 12 months for the next 20 years and then not follow the directions I get.

SPEAKER_00 22:57

Right. Yeah yeah. No, there's a lot that can be done with data for sure. And with AI it's making all that possible right so what what what about um now people want to predict diseases also. So all this together you can probably predict I don't know whether it's possible to even predict uh when you when you can have a heart disease or heart heart attack for that matter.

SPEAKER_01 23:25

Yep.

SPEAKER_00 23:25

Can can it be done?

SPEAKER_02 23:27

Are we there yet or still so I mean in some sense when people write you know the framing and heart study they you know you do your multivariate regression and you say here are the risk factors. So if you have these risk factors you're 12 times more likely to develop coronary disease to have a heart attack to have orally mortality like that's prediction.

SPEAKER_01 23:46

Right.

SPEAKER_02 23:47

But it's it's a little bit static. It's you know it was old study data based on a relatively homogeneous population. I think the power is is that same level of statistical inference and um you know analysis can be done in near real time and it can be much more sensitive selective whatever the right word is much more you can develop a cohort or a population that's much closer to who you actually are. So the you know the when you when you set these studies up you say okay we're going to look at people between the age of 30 and 50 that you know whatever you set up your characteristics for your intake um but what would but what would be much more interesting is I'm 50. So a 50 year old white male living on Long Island with you know the your these genetics has this risk like the we we have the ability to do that the data is there. It's a question of you know do you have the resources to burn the compute to run the numbers? And if you do um again it's not just the data. It's okay what does the data tell you and then this is the hard part once you have it how do you find a lever find something you can change to change the to change the outcome so if you can if you stay on this course you're very likely to have a heart attack before you're 60 you know eat eat fewer than three pounds of bacon every weekend. I love bacon but um like finding a way to identify things that can be changed and then finding a way to engage the patient to actually change them. Like there's a lot of extra steps in there that go just beyond the data.

SPEAKER_00 25:42

Yeah. You're right. I mean it's also the what action you can take based on that inference right I mean if if you can't tell you're going to have heart heart attack and you can change your diet that's a possibility but if there is something like neuro neurogenerative diseases like Alzheimer's or something I don't know what what you can do even if you know you're going to yeah for some yeah 100% for some things that we don't have a treatment.

SPEAKER_02 26:13

So yeah it's it's very dependent on the use case but we're we're also seeing a lot of technology you know in Parkinson's and even with Alzheimer's there's there are certain studies out there that suggest if you read a certain amount of day if you remain cognitively active you you know your chance of developing it is is less. And so knowing that and again going back to like your personalized data your genetic background your family history your biomarkers if you knew that someone was at an increased risk but you also knew that you know instead of doom scrolling on LinkedIn or staring at Netflix for two hours if you instead did a crossword puzzle or Sudoku or engaged in something much more dynamic that's something that you know the the the the PCP visitor, the neurologist visit every three, six, 12 months isn't going to address but if your health companion your phone your app Alexa on your speaker said hey you've been on the TV for 45 minutes you know you usually go to bed at 10 o'clock it's 845 it would be much you know if you want to slow your progression of Alzheimer's if you don't want to end up like grandma jones uh you know turn off the TV pick up a newspaper go walk the dog whatever the hell whatever it is like make a recommendation drive behavior change. And we I mean you've seen the the bad side of of um optimizing algorithms for attention and eyeballs the Facebook lawsuits like we've proven at least with our kids that we can turn them into you know screen zombies through you know repetitive dopamine hits and and drive bad behavior I don't think it's unreasonable to think that similar mechanisms could be used to drive good behavior.

SPEAKER_00 28:06

Right, right. No you're right I mean we're we're all I mean especially these reels and all everybody's getting addicted to you just scrolling and