Medical records are finally digital, but so much of the valuable data is buried in unstructured form in theclinical narrative that it can be hard to make use of it. It’s a great use case for artificial intelligence.
My guest, Chris Tackaberry is a physician and computer scientist turned entrepreneur whose company, Clinithink develops AI software to enable health care and life sciences customers to gain insight from that unstructured information.
Clinithink provides applications in rare disease, population health, real world evidence, and revenue cycle management.
Host David E. Williams is president of healthcare strategy consulting firm Health Business Group. Produced by Dafna Williams.
Medical records are finally digital, but so much of the valuable data is buried in unstructured form in theclinical narrative that it can be hard to make use of it. It’s a great use case for artificial intelligence.
My guest, Chris Tackaberry is a physician and computer scientist turned entrepreneur whose company, Clinithink develops AI software to enable health care and life sciences customers to gain insight from that unstructured information.
Clinithink provides applications in rare disease, population health, real world evidence, and revenue cycle management.
Host David E. Williams is president of healthcare strategy consulting firm Health Business Group. Produced by Dafna Williams.
0:00:11 - David Williams
Medical records are finally digital, but so much of the valuable data is buried in un-structured form in the clinical narrative that it can be hard to make use of it. So it's a great use case for artificial intelligence. My guest today, Chris Tackaberry, is a physician and computer scientist turned entrepreneur, whose company, Clinithink, develops AI software to enable healthcare and life sciences customers to gain insight from that unstructured information. They have applications in rare disease, population health, real world evidence and revenue cycle management. Hi everyone, I'm David Williams, president of Strategy Consulting from Health Business Group and host of the Health Biz podcast, a weekly show where I interview top healthcare leaders about their lives and careers. If you like this show, please subscribe and leave a review. Chris, welcome to the Health Biz podcast.
0:01:00 - Chris Tackaberry
Thank you very much, david, delighted to be here.
0:01:02 - David Williams
Outstanding. Well, before we get into what you're doing today, I'd love to hear a little bit about your background, your upbringing. You know any childhood influences that have stuck with you throughout your career?
0:01:11 - Chris Tackaberry
Okay, as you can hear from my accent, I'm from the other side of the pond Born and grew up in London, got one younger brother and a much younger half sister. My parents, my dad was a lawyer and my mum was an actress and then a literary agent, so I guess the families kind of roots are in the performing arts or talking out loud kind of thing, and there's not a scientist to be seen for a mile around. So I'm an outlier in that respect.
0:01:51 - David Williams
When you were growing up did they wonder like, did you get like a switched at birth or something Like, did you fit into the family or how did they deal with it?
0:02:03 - Chris Tackaberry
They were very encouraging of anything remotely constructive that either of us wanted to do and I have to give them credit for that. I've now got three grown up children on my own and we all know how challenging that is to let the kid focus on what they want to focus on. And I was fiddling with simple coding in Atari modules on very early games consoles and stuff like that. I was one of those, yeah.
0:02:37 - David Williams
Very nice, they let me do that. That's good, they let you do that. I had an Atari computer too. They had the Atari 800 and the 400 and these cartridges that you'd put in Exactly. And they had like 8K on them. You could expand the operating system, but you had to, like open the hood or it would overheat, as I recall.
0:02:56 - Chris Tackaberry
Already, half the people listening to this are thinking well, these guys are talking about stuff that they built out of stone, right?
0:03:01 - David Williams
It's a 6502 microprocessor, as I recall In any case. So what did you do for education?
0:03:08 - Chris Tackaberry
So I went to medical school, did a medical degree. I collected a science degree on the you know, as part of that kind of package. So I spent a long time in university education. I was fortunate enough to be able to, after my intern year as a doctor, I did a masters in computer science and that was quite unusual. I mean it's not that common today, but then it was crazy unusual. It was just an opportunity. I was fascinated by computers and by their power even then, and by luck, the group that I did my MSc thesis research project with was one of the kind of outstanding, you know Xerox PARC type bunch of geniuses in medical informatics and they just happened to be working at my university. I was super, super lucky to get to work with those guys. They were 20 years ahead of their time then and they probably still are, and so they. I got really quite deep insight from that work into the complexity of medical information as opposed to, you know, other verticals that have been transformed through digital innovation and automation and some of the approaches to managing that complexity, and that was really had a profound influence on me the opportunity to work with that group.
I was very lucky and then I went off into. I did some public health for a while because those guys are really into, you know, crunching large amounts of healthcare data to help facilitate population health. And then I realized I actually wanted to do acute medicine and sort of get that out of my system. I was just a one-trained anesthesiologist and critical care doctor. I did that. I've bordered in both of those disciplines. I did that for quite a few years and passed the point that people normally jump out of a medical degree. So I got to be a proper doctor and worked on a very acute environment for quite a few years. But the IT bug never left and I was constantly thinking to myself. I was doing rounds in intensive care unit at three o'clock in the morning. There's got to be a better way to get the information that I need to look after the patient in front of me to me so that I can make decisions faster than someone having to make phone calls and stuff like that.
0:05:46 - David Williams
Yeah.
0:05:48 - Chris Tackaberry
That's how we ended up getting into IT.
0:05:51 - David Williams
That sounds good, and so I saw a couple of things on your background. Isoft was that? Sounds like software.
0:05:58 - Chris Tackaberry
Yeah, so I was becoming increasingly interested in how technology applications could be applied to healthcare. And this is a long time ago, like 20 years ago In the sort of the relatively early days of the EMR revolution, and Isoft was an EMR vendor. They were quite small when I joined and they grew 10x. Yeah, over the course of my time with them, fascinating really important schooling in commercial software, enterprise class applications which can make a difference in healthcare organizations and those are very complex organizations and I had the sort of field and shop floor knowledge that created some useful insight from a product development and product management perspective.
0:06:54 - David Williams
And then did you start your own consulting practice after that.
0:06:56 - Chris Tackaberry
I did. Yeah, it turns out that the combination of skills I have so, having practiced medicine for quite a few years and done the academic computer science and then worked in commercial software environment, there's not a load of people around with that combination of skills.
0:07:18 - David Williams
That's like the trifecta, you know.
0:07:22 - Chris Tackaberry
It's very dangerous, you know, because it means you know when developers start talking about technical stuff they don't actually lose me as quickly as they'd like to. But and I have no pretense I do not practice medicine anymore, I don't use the title doctor in front of my name, but I am on the register and I do understand a little bit about what my colleagues on the front line of medicine are still doing. So I think that combination of insights is, you know, is valuable in being able to create value in this ecosystem and what was ultimately what led to our setting up a clinic business.
0:08:00 - David Williams
Got it. So you're obviously going from the consulting can be fairly broad and you're solving problems for specific clients, usually more on a services basis, even if you rely on some technology. And so why? Why kind of think what were the origins of that? When you said that this needs to be its own company?
0:08:16 - Chris Tackaberry
Great question. I think the the consulting work I did was was still to software product companies and I was very fortunate I got some experience, well outside the healthcare domain, doing some work for Skype, which was at the you know the sort of fascinating point in that company's journey and so and a B2C business as opposed to B2B, which is what I would have been familiar with up until that point. And yet the fundamentals of software product management, working at how to add features to an existing code base that created incremental value for their users, all those, all that discipline, you know, helped really bed in some of the understanding about managing that super complex challenge of. Here's a bunch of real world problems and here's a single, you know, line of productivity in terms of creating software. But in the through my days as an EMR vendor, I like to describe myself and some of my colleagues now as recovering EMR vendors and post EMR phase. You know we learned a lot about the challenges facing health systems CIOs.
0:09:32 - David Williams
Those are not easy jobs and we also learned about you know they used to say I grew up in Washington DC, so there's a lot of these TLA three letter acronyms and CIO. I always understood it to being career is over once you get promoted to that role.
0:09:50 - Chris Tackaberry
No, that's an overloaded term to use a computer science concept. That's that variable is overloaded. And obviously you know we're talking about chief information officers or chief data officers or you know there's a few different terms now. But if you're responsible for a big health system with a diverse kind of empire, from a sort of landscape perspective, and thousands of employees and thousands of physicians who, let's be honest, there can be an interesting group to work with, so very complex, complex situations, and we knew from that early work that around the admin side of things so capacity planning and bed management and visit scheduling and obviously billing you know there are many very good products out there that help support the enterprise in those endeavors.
But the clinical stuff is predominantly around what, what gets written down. And when I was putting EMR systems in people were still using paper and couldn't type. Luckily for all of us, those people have either learned to type or retired. And so now the visit summaries and the discharge summaries and the radiology reports and all the rest of it, all of those extremely valuable information assets in the healthcare ecosystem, are machine readable, they're computable. And so me and my co-founder, my colleague Dr Pete Johnson, who's actually the genius behind the technology that we've built the products at Quinnique around, had become convinced, independently actually, that there was a disproportionate amount of value, as you put in the intro very succinctly, in the unstructured content and really very few tools to enable organizations to extract that content and create value from it.
0:11:47 - David Williams
Now why is? I've certainly heard this about. The narrative is where all the information is, but there's a lot of effort in these EMRs to actually structure the data, and so why is it that so much falls outside of that? Is it sort of the exception and something that should be ideally all structured, or should we just accept that the narrative is there? Or does the fact that AI exists means we don't have to structure and we can just do whatever we want and then the machine will figure it out later? I try to understand how this unstructured piece fits in.
0:12:14 - Chris Tackaberry
You phrased that question really well and the way to think about it is built from a couple of observations. So the first thing to say is roughly when I left medical school any physician my vocabulary was roughly double what it was when I went in. So this is a big space and it's mostly messy. Medicine is as much an art as a science. That's the first thing. The second thing is the International classification of diseases.
Icd-9 or ICD-10, or ICD-10, there's various flavours of it. Well, that was invented in England in the 19th century by a guy called William Farr, who at the time was a genius, but that thing's getting a bit long in the tooth. But it was never designed to enable the kinds of questions that people want to now ask. Okay, I'll give you an example. There is and I'm not knocking ICD-10, it's required, right, and there's a ton of stuff you can do with that structured data. But there are. Last time I checked, and we are active in this space there's a hundred types of lung cancer, broken down by stage of disease and tissue type and tumor type, and now, of course, all the genetic stuff. There's one code for metastatic lung cancer, one ICD-10 code. So if your billing data is perfect and it might not be perfect. Then you're going to get a 99% false positive hit rate for patients you're looking for in a specific sub-disease or sub-type, if you're going to use structured data. And then the final kind of gotcha is often when I give presentations and talk to people about the work we do and how we can help them solve their problems, they take notes during the meeting and sometimes those notes are written down on paper, even if I'm talking to IT people and I say to them look, imagine you're in a consultation with a patient. You're not going to take notes, you're not going to write stuff down. Of course you are.
So, as you rightly said, what's interesting about some of the tooling that's coming with the large language model revolution is actually going to increase the amount of unstructured data present in the ecosystem, precisely because it does enable the production of that content.
When you're trained to practice medicine, you're trained in a way to think and to describe your thoughts, and those descriptions are the bedrock of the record that accounts for the decisions you made, the care you delivered, and that has two purposes, as we know. One is to inform the next person in the care chain or journey that's looking after you so that they can see easily what you've done. And of course, the other is to recognize that there's a lawyer in the consulting room. So for those two reasons, you have to write down a fairly rich and detailed description of the care, the plan, the observations, the treatments and the decisions. That's written down in a language and that language is very complicated. So it's not either or. But the market serving the analysis of structured data in healthcare is very mature and there's a ton of rich tools out there. And when we start a clinic and to the extent still the case today the tools that serve this unstructured asset are much thinner on the ground.
0:16:04 - David Williams
And in terms of the generation of the unstructured piece or the notes, is AI actually playing a role there too, Because we see that there's like the co-pilot? I interviewed the head of. Microsoft's unit that's doing that, and I'm even thinking on some calls that people do on Zoom and so on. There's a co-pilot that comes along and takes notes and does some analysis. How does that fit in?
0:16:21 - Chris Tackaberry
Well, I think that I mean it's basically accelerating the change. So we know that there's a burnout, a workforce crisis in healthcare delivery on the frontline a clinical, frontline physicians and nurses that's true in most developed health economies and anything that we can do to automate any part of the process that enables those individuals to do more, with kind of less cycles burnt, is good for them and their mental health and their productivity and, of course, for their patients, which is ultimately what motivates us. So tools which enable you to create the narrative that you need to describe the care that is being given, that accelerates that or facilitates, that automates it, can only be a good thing. And there's some. Co-pilot is a great example.
Eric Topple, who I recommend everyone follows a guy's an absolute genius. I saw a poster of him a couple of days ago talking about the difference that these platforms will make to the physician workload. Because what you're always trying to do, whether it's patient outcomes or physician workload or both, you're trying to solve a paradox. You're trying to get more done with less capacity and, in my humble opinion, the only way you can do that is with digital automation and, ultimately, ai.
0:17:51 - David Williams
Got it All right. So let's talk about some of these end uses. I'm talking a bit about the technology, the workflow, what's driving you to this point? But I see real-world evidence revenue cycle management, which anybody in health care or private equity is always interested in that one population health and rare disease. So why these four things? What do they have in common, I think, between rare disease and RCM? Those seem pretty different applications off of one technology base.
0:18:22 - Chris Tackaberry
They do, don't they so? But actually the thing that all four of those segments have in common, completely in common, is that at some stage in the processes and the health system end processes that they relate to, extensive manual chart review is required to achieve whatever the outcome is. So, although you're quite right, we have solution offerings built on one technology platform. So we've only we've built that core technology platform. Once We've got a pattern for it, we keep getting baked off against all kinds of other people. We keep winning, so we must be doing something right.
That thing is the bedrock and that's solving the same problem in those four different contexts.
And what it's doing is taking thousands, millions and now tens of millions of documents created at the business end of health care, wherever that is in an electronic format, so we're not dealing with handwritten pages anymore and it's converting that content into computable information.
Once you've done that, it's a relatively small kind of tweak of a few levers in a configurable stack to enable value creation in those four different activities where the beneficiary is either a pharmaceutical company like AstraZeneca or a large health system like Northwell, and we're delighted and very proud to be partnering with both of those organizations.
So it's a fair question to ask. That looks a bit of a crazy spread, yeah, but the process that we are automating is the same in each case and it creates value basically for the same reason in each case, and we've just found that those situations are where we have been drawn by our customers, actually in several cases. So, in relation to the revenue cycle piece, revenue cycle is a big, hairy, hugely complicated set of processes and there's a huge primary and secondary vendor community serving those, and so we avoided it like the plague for many years. And then one of our customers said to us here's a point in that process, specifically around clinical documented improvement and denials management, where we, the customer, see your solution creating value, and it's a brave person that just says to that customer yeah, yeah, sometimes might be the right answer that sometimes, but we didn't think that was the right answer.
0:21:17 - David Williams
But in that case let me see if I can understand it. So, with revenue cycle management, usually the hospital or physicians trying to get paid for something that they did, the insurance company may be saying, well, we're not going to pay it if it's a denial, or we're just. We need certain evidence of that. And then maybe there's a clinical issue that they say, well, this is for a cosmetic purpose, or maybe they didn't have this underlying condition, or you didn't do this test, they didn't need certain things for the step therapy, whatever it may be, and that information is going to be in the clinical note and it's not just sitting there to say, here's the information you need, plug it in by its nature, but you can use your tools in order to go through and find what's there and then put it in the format that's actually going to be needed to make the case for the payment.
0:22:00 - Chris Tackaberry
Do you know what? David is even simpler than that. There's a key. This so they're like I say to the RCM process is 30 or 40 steps and multiple different teams. In a large enterprise like Norfolk Health in New York City, that's a big team, okay, and we've got a very healthy respect for where we, where we add value, which again we were we were drawn into by our customer who knows their processes better than we do, and there were two specific points in their end to end process which supports the remunerable cycle activities, where they perceived our ability to create value. One was in clinical documentation improvement, where we're looking to ensure that the physician has fully documented the care delivered such that the health system gets reimbursed correctly and fully, and that in that scenario, you're actually looking for gaps. Yeah, you're looking for gaps, and again, you can automate that search using our technology and deliver a ton of efficiency.
In the Norfolk Health case and this is in the public domain they did a very nice podcast describing it. Actually you could they could extend the coverage with the same number of people to cover additional episodes of care. A lot large volume Got it. And then, on the denial side, you're looking for the keyword is medical necessity, and so there's evidence that should be in the record that can be used to rebut a denial. And again it turns out you can use our software to find that evidence much at much greater scale and much faster and cheaper than using humans to do it. And there's the sorts of skills that in the in a typical denials management team within a health system. You know you need a lot of physician input into that, and if that's the case, then those physicians are on the front line delivering print. Okay, you know. So that that's a big problem and, as far as I can tell, that problem is getting bigger actually.
0:24:09 - David Williams
On the clinical documentation improvement. Is that a retrospective process or do you need to sort of be in the exam room prompting along the way?
0:24:17 - Chris Tackaberry
Great, great question. So the way that particular solution is implemented on our stack, it's it's same day, so it doesn't have to be absolute real time.
0:24:25 - David Williams
Yeah, it's same day. So in other words, while while the case is still fresh and that visit still fresh and the physicians mind they can say what about this?
0:24:34 - Chris Tackaberry
Well, and also and also the the billing cycle. Right. So the conclusion of the episode and the time to get the bill out. You know, the teams are under quite strict performance. Yeah, and you don't have to wait.
0:24:44 - David Williams
You don't have to wait for the denial to come in and then go back.
0:24:47 - Chris Tackaberry
Well, you well, this work is obviously. The reason that that intervention takes place at that point in the process is obviously to try and reduce the likelihood of a denial downstream. So you're getting two bites of the chair in this particular approach.
0:25:00 - David Williams
Let's talk about rare disease. So one of the things about rare disease a couple things. One is it's rare disease are common a lot of people have.
0:25:06 - Chris Tackaberry
Correct, correct.
0:25:07 - David Williams
So that's one, and then the other thing is that they take some a long time to find out they have it whatever it may be they ever do. So I think about this diagnostic odyssey is a term that I've heard, I've seen you use that.
0:25:19 - Chris Tackaberry
What's the?
0:25:20 - David Williams
what are the issues in rare disease and where do you step in?
0:25:23 - Chris Tackaberry
So again we were, we would lead into that space. It wasn't something we went out looking for. We buy another of our customers, alexion, which is a lot of rare disease from suit company that relationships in the public domain and you know, I full credit to them. They baked us off against everything they could find and for whatever reason they, they selected us to work with and obviously we're thrilled to have that partnership. The specific problem they identified and you're absolutely right, 30% of the population in the United States has a rare disease of some kind. Each individual rare disease is fantastically rare. So an average clinician is never, ever going to guess that diagnosis Unfortunately for that position, because that's not a great place to be and, more importantly, for the patient. So that leads to exactly what you've described diagnostic.
Obviously, as the you know, huge advances have taken place in what they call next generation sequencing, so much more powerful sequencing of your DNA, your genes, which is the root of all rare genetic disease. What we've learned is that the genotype on its own, the genetic information on its own, is usually not easily interpreted to be able to pinpoint that is a rare genetic disease. What you need in addition to that genetic information is what they call phenotypes. So that's the clinical characteristics and those are. There's about 12,000 of them that have been sort of documented for about 7,000 rare diseases. That's a list of 12,000 quite obscure clinical things written down in words.
Then you look at the patient in front of you and as a physician you might be able to observe certain characteristics like you know, one finger is slightly shorter than the other or really quite subtle stuff. You might observe those things and write them down, but you'll never make the connection between that and one of these official phenotypes of the rare disease. That's the process that we automate, so it ultimately comes down and we can do in milliseconds what would take the highly trained physician weeks. So the value creation in that space is helping to shorten that diagnostic odyssey by identifying the sort of signal in a patient's record that suggests they might have the characteristics of the rare disease, which can then be used to guide targeted genetic testing and get to a diagnosis and hopefully treatment to treatment, to spainable, which increasingly it is.
0:28:03 - David Williams
I'll ask you about one more, which is real world evidence. Now, that is a broad term and many different data sources that are used. There's a lot of end uses, a lot of different departments in pharma and elsewhere that use them. Where does RWE fit for Clint? I think?
0:28:19 - Chris Tackaberry
Again, it's a similar principle where, in which you're quite right. Real world evidence is the phrase that our colleagues in the pharma industry use to describe data that's already been transacted, usually as part of the delivery of care. So there's a kind of breadcrumb trail, if you like, of data. Now it's not organized in the way that a orthodox prospective clinical trial will be, where I recruit patients into a trial and then actively collect structured data about them as I perform whatever new intervention I'm testing is, but that's very expensive and it's very difficult to scale, and so over the last few years pharma starts to get very interested in this, the utilization of data that's already been transacted. Now that never comes perfectly organized in the way that you need it, so you have to sort of flex and compromise, and so it turns out that for some of this kind of work, the signal needed to answer the question is not in structured data, it's only in the unstructured data, and that's where we come in, and we've been very fortunate to partner with AstraZeneca, our colleagues at Premier, one of our partners in the US, to undertake some very large scale analyses of tens of millions of documents in relation to LUNCAT. So, for example, looking for evidence in radiology reports which are gain-after-y text of a particular finding that can be a precursor to LUNCAT. So help to try and anticipate that and intervene when the disease is curable, as opposed to when it becomes untraceable later on. So it's that kind of approach which now that is real world evidence work that, in that case, was a project funded by AstraZeneca.
It's also population health, if you think about it, because trying to find more subtle evidence of the early stage of a disease and intervene then and I've got trained in public health a long time ago in my humble opinion it's the only way we're gonna solve the conundrum of healthcare getting exorbitantly more expensive. All of these interventions cost me more and more money. So we have to do this. What the population health people call left shift Left shift in anything from diabetes to cancer is going to increasingly require AI techniques operating on real world data to take advantage of what we already know about you, what we've already collected but which we can be so much smarter in anticipating to say, aha, we can now predict that you've got quite a high likelihood of ex-outcome. But because we know about that now and you can only do that using these AI techniques, we can intervene early, which is cheaper for the health economy. However, that's funded and, of course, a better outcome for the youth.
0:31:22 - David Williams
Chris, my last question is turning away a little bit from work and it's about if you have any good books that you've read lately, anything you would recommend, and then, conversely, anything you would recommend to avoid. You wasted your time on it.
0:31:38 - Chris Tackaberry
Wow, that's a great question. So, books to recommend? I'm currently reading a book called the Expectation Effect, which I certainly do recommend. It's talking about the extent to which the mind can control even biochemistry. So I'm a very keen cyclist. That's what I spend my spare time doing.
So I'm interested in performance on a bicycle and, as I get older, how I can try and keep getting faster, and there's some very interesting work being done now about how the brain creates a kind of future state model based on its previous experience and that has profoundly influences how you perform in certain situations, whether they're at work or in some hobby or something. I'm kind of halfway through it. It was recommended to me by a very trusted source and there's a lot of science behind this, and the phrase I hadn't read before, which I heard before until I read this book, is psychobiology. So that's. We of course know about the placebo effect and the no-cebo effect, which is the opposite. This is next level thinking. This is about how your brain decides what you're gonna be capable of physically, say in a performance environment in sport, and of course, the opportunity is to try and intervene in that process to extract even more out of yourself.
0:33:14 - David Williams
What's your space? How much do you find that to be on the range of sort of rock solid to speculative? Is that more speculative or is it does it seem? I mean, I'd like to believe what you're describing there.
0:33:28 - Chris Tackaberry
Some of the book sites a lot of research that I mean in coming from a public health epitopio beyond your background, unless and in your study is two million people, you sort of dismiss it so that. So I had to kind of get over that. It looks plausible is what I'll say.
0:33:51 - David Williams
Okay, good. Well, that sounds like a good one and you don't need to recommend something to avoid. But if that helps to provide more time for reading what you do wanna read. But if you do have something you can let me know. Most of us don't.
0:34:04 - Chris Tackaberry
I don't have anything. I don't have anything that comes immediately. Avoid things that don't enrich your life or somebody else's, that's my advice Sounds good.
0:34:16 - David Williams
Well, Chris Tackaberry, CEO of Clinithink, thank you for joining me today on the Health Biz podcast.
0:34:21 - Chris Tackaberry
Delighted David. Thank you for having me.
0:34:25 - David Williams
You've been listening to the Health Biz podcast with me, david Williams, president of Health Business Group. I conduct in-depth interviews with leaders in healthcare, business and policy. If you like what you hear, go ahead and subscribe on your favorite service. While you're at it, go ahead and subscribe on your second and third favorite services as well. There's more good stuff to come and you won't wanna miss an episode. If your organization is seeking strategy, consulting services and healthcare, check out our website, healthbusinessgroup.com.