Understanding And Leveraging Healthcare Data

Speaker 1:
Welcome to TD Cowen Insights, a space that brings leading thinkers together to share insights and ideas shaping the world around us. Join us as we converse with the top minds who are influencing our global sectors.
Charles Rhyee:
Hi. My name is Charles Rhyee, TD Cowen's health care technology analyst. Welcome to TD Cowen's FutureHealth Podcast. Today's podcast is part of our monthly series that continues TD Cowen's efforts to bring together thought-leaders, innovators, and investors to discuss how the convergence of health care, technology, consumerism, and policy is changing the way we look at health, health care, and the healthcare system. Over the last decade, data and analytics have become increasingly ingrained in the healthcare system. Given the evolving landscape, we want to provide a deeper look into different types of healthcare data, as well as the users and use cases. This podcast is intended to be the first in a series on healthcare data. Today, we'll provide a general overview of how healthcare data is currently being used, with subsequent podcasts exploring subsets in more depth.
To help discuss the topic, I'm joined by Paul Reuscher. He has a master's in applied mathematics and applied economics, spending time in academia researching health economic outcomes. Started his professional career at IMS Health, now part of IQVIA, first as a product manager, and eventually helping develop what is now IQVIA's anonymized patient data solution. He subsequently led a similar development program for DRG, which is now a part of Clarivate. Paul currently is vice president of Clinical Data Products at Forian, a healthcare data intelligence company. Paul, thanks for joining us today.
Paul Reuscher:
Thank you, Charles, for having me. It's really good to be here.
Charles Rhyee:
Great. Maybe before we begin, maybe you can tell us a little bit about the work that you're doing at Forian, and maybe some of your professional career as well.
Paul Reuscher:
Sure. Yeah, I mean, you gave a pretty good snapshot to it. Started off doing economic stuff and projections, statistical methodology. I found healthcare data, healthcare claims data to be pretty fascinating, interesting. Started doing some academia-level type consulting work in health economics. And then from there, forayed into the full-on data products, data product offerings and data strategy. It's an ever-evolving platform, really. There's constantly new data streams coming through, new data sources, new offerings, new competitors. It's a fascinating marketplace, really. The entire ecosystem's really cool.
Charles Rhyee:
Great. Maybe tell us a little bit about Forian.
Paul Reuscher:
Sure. Yeah. Over the last decade, I've always been involved with building out claims data products. Here, I'm building out claims data products with a little more of my fingerprint on it, for hybrid data between open and closed source data. And then linking in social determinants of health consumer data to the offerings, as well as EMR, as needed, depending on different use cases or whatnot. That's what we've been doing at Forian, trying to build out a best-in-class patient longitudinal capabilities. With claims data being the basis of it, but then trying to layer in ancillary data products to give a best in class.
Charles Rhyee:
Great. Let's jump in here a little bit. A lot of folks who might be listening is familiar with the idea of claims data and prescription data, some of the big buckets of data that are out there in the market. Maybe starting there, what are the main sources where this data's coming from? And then maybe touch on, what is valuable about claims data?
Paul Reuscher:
I mean, it all comes back to ... think revenue cycle management software that's being offered in health networks. Providers utilize the software to submit medical claims. Or pharmacy benefit type software where it's submitting pharmacy claims, which typically come through in a different format, submitting these through software that gets to payers, Medicare, whoever is going to reimburse the claims. That's typically for claims data. It kind of breaks into three different buckets. There's pharmacy claims, which is retail, non-retail, specialty. It's kind of its own thing. There are medical claims that are either open source, so it's typically like a software multi-payer. Just more of an open software that's a clearinghouse processing claims. And then there's closed-source claims data, which is either a single or multi-payer system, but typically has some sort of an eligibility criteria or enrollment file that comes along with the closed claims data. That's claims data in a nutshell, these three different components.
Charles Rhyee:
What are people typically using claims data for? What are they trying to tease out of this information? It's generally made for billing purposes, but obviously there's a lot of analytics put on it.
Paul Reuscher:
Yeah. I mean, all the analytical use cases come back to three pieces. Basically, some sort of a provider intel or targeting, so understanding what physicians or providers are doing, treatment utilizations, medical encounters, prescriptions that are getting processed. Patients, so understanding medical encounters around patients, or understanding their health journeys as they move along in their lives. Medical procedures they have, prescriptions they take. Drug utilization, compliance around that. And then, ultimately, reimbursement. So if a prescription is filled, how much is it reimbursed by which payers? What is the patient out of pocket? What is the patient burden from an economic perspective? Just understanding that. But it's three pieces. It's typically provider, some sort of an intel around providers, some sort of an intel around patients, or ultimately, how are these things being reimbursed?
Charles Rhyee:
When we think about that, who are the main users of this data at this point?
Paul Reuscher:
Pharmaceutical life science companies. Could be pre- or post-launch of a therapy or indication. Medical device companies, same sort of commercial use cases. Provider networks, just to understand referral patterns, leakage, steerage, in-network, out-of-network analysis. The payers are also interested, to some extent, on overall healthcare utilization and cost containment, right?
Charles Rhyee:
Yeah. No, that makes sense. I guess what's interesting though is when we think about this dataset, it seems like there's ... you mentioned earlier, there's a bunch of clearinghouses that you can get this data from, or you can get it directly from big networks, maybe from health plans directly. Is it easy to get access to this data? Are there barriers to entry to trying to set up a business to access and analyze this data?
Paul Reuscher:
The answer is yes and no. Getting ahold of some sort of claims data from somebody out there in the marketplace that's trying to license the data or insights from the data, pretty relatively straightforward. Yeah, it's readily available.
No, on the other side of that, is the fact that they have good answers or good data, or some amalgamation of a lot of data sources together into one single source of truth for analysis. Yeah, the barrier to entry gets higher and it's more expensive. So you typically have to lean on, again, somebody like Forian that's like a data aggregator to help you answer questions, because we've pulled data from many different sources to give you these insights that you're looking for.
Charles Rhyee:
Is quality an issue when looking at these different datasets? I would think that claims data is pretty standard, so that even if you pulled it from different places, it's still a medical claim. Maybe touch on that. How does that look across the industry?
Paul Reuscher:
Normalization is a problem. Normalization of data, standardization of fields across different sources, some data aggregators and providers do it better than others. Looking ahead into the future, this is where generative AI and such seems to be very promising. I'm of the opinion that it's going to be transformative to our industry to help with this normalization problem, cleanliness of data problem, right?
Charles Rhyee:
Yeah. I think when we talk about normalization of data, I think a lot of people tend to think of claims data as fairly clean. It sounds like you're saying that that's not always the case. What are the major parts that are challenges in normalizing something like claims data, which you would assume is pretty standard?
Paul Reuscher:
Well, it's interesting. It could be ... there's a couple different facets. Typically, we look at everything from back to provider, patient, payer. Provider information, so the hospitals, the physicians, the medical offices. You're beholden to how that information is filled into the software. It could be relatively straightforward in how it was captured. The billing address could be the physical address. That somebody rendered a procedure or some medical encounter occurred at said address. Or could be some medical billing office that's part of the health network and that address is not the physical location of whatever the procedure was. So there's a lot of normalization around the place of service for facilities or physicians. That's interesting.
Next, patient. HIPAA compliance is an issue. De-identified data had to be de-identified at some point. There's different approaches and technology used for de-identification, but that doesn't exactly relate to one-to-one, saying, "I de-identified this patient that has one patient ID." There's a lot of work that we do around patient mastering, that we call it. That has us coordinating between different data sources. Looking at things like data of births, age, gender, different other identifiable information that is HIPAA compliant, that passes through on claims or EMR data. We use that to try to master the patients.
And then, ultimately, the last one being payers. So if you have a reimbursement interest, understanding payer and plan name and payer type, that can be kind of a messy thing in and of itself. Again, back to how the data is sourced, that could be some kind of a conundrum you have to work through.
Charles Rhyee:
Yeah, that's interesting. If we think that that's complex, and then we think about moving onto EHR data, which now given the HITECH Act and the mass adoption of EHRs over the last 10, 15 years, that's an area where I think most people look at it and say, "Well, that's even more complex given a lot of the value-added information is in freeform physician notes, voice attachments, imaging files, image file." Maybe talk a little bit about EHR data, the prevalence of its use today. How important do you think EHR data is relative to claims data?
Paul Reuscher:
It's just as important, but the use cases are a little different. This is more patient-centric type insights. But ultimately, back to that normalization problem, we run into some issues with EMR data, where different software has different inputs, different field content. So you could have some of the same fields across different EMR softwares, different answers, different nomenclature within writing out whatever ... some clinical naming convention. I think you could run into issues where you're trying to go through clinical information. You either have to have a clinical mindset, so more like a clinical researcher than a data scientist. That's a gap that you start to run into more with EMR data versus claims data.
With claims data, there's a lot of ... I had somebody that's a really good data analyst. They know how to code and script, and they can kind of Google search their way through clinical knowledge gaps that they have, but they could still probably do a lot of analysis with claims data. When we start looking at EMR data, there's different coding and naming conventions. There could be a lot of permutations around that. So you start to have to need somebody with a little more of a clinical eye to understand the information.
And then, ultimately, when you talk about freeform patient notes or images or audio inputs, we start to get into things that are very sticky in the HIPAA compliance world. When you look at EMR images that are stored for some patient visit, or radiology imagery or x-ray or something that happened, each one of those images typically has to be scrubbed because it had patient name or date of birth or social security number. There was some kind of an identifiable component put into the image. So just working through all that. Same with patient notes. It's different, but more of the same. Could have patients' names or other relevant notes that a medical professional decided to put in the fields, right?
Charles Rhyee:
Yeah. No, certainly. You just mentioned, with EHR data, it's really more patient-centric. But my understanding is even with claims data, more and more payers are asking for clinical documentation to approve reimbursement. Does that kind of lessen the value of EHR data, if the claims data is now being attached with more clinical notation?
Paul Reuscher:
No, because I still think for specific rare disease, specific types of medical visits, EMR data is always going to be ... it's just going to have a little bit more information than what would ever be pertinent to claims data. Again, it's different data for different use cases. But marrying claims data with EMR data obviously leads to this best-in-class, have your cake, eat it too, whatever analogy you want to use. It becomes a lot better when you have the combination of both. And then, again, if you can add consumer data to it. Well, now we're into a whole new facet of patient-centric research.
Charles Rhyee:
Yeah. Maybe before we go into that, touch on some of the most common use cases for EHR data today.
Paul Reuscher:
Mostly in patient journey and understand clinical sequencing. Anything where a lab result or some kind of a test result is pertinent to be able to further come up with a patient distribution sub diagnosis code level, that's where that typically becomes more useful. Claims data is just going to say, "This person has this condition and this is the procedure. This is the drug that was prescribed." Whereas, obviously, EMR data is going to have a little more information around genomics, lab results, what have you.
Charles Rhyee:
Is EHR data being used yet, in terms of ... With biopharma companies, we hear of this potentially creating synthetic control arms for clinical trials or identifying potential patients for trials. It sounds fairly intuitive that that could be a good use of this EMR data. Is that being actually applied today yet?
Paul Reuscher:
In pockets. I don't think it's as ... It's one of those things, it's not as widespread as you would think. But again, synthetic control arms, clinical trial optimization work uses EMR data, claims length EMR data. That's for sure a definite use case for that, but I don't think it's as widespread as you would think it is.
Charles Rhyee:
What might be the reason that's the case? It sounds like it's such a powerful tool, particularly when you think about just the cost of developing a drug. If you could cut one arm of a trial out and do it through existing real-world evidence, what has been the general barriers for that? Is that a regulatory issue, like FDA, or is that the cost or technological barrier?
Paul Reuscher:
I think it's a bit of a technology and data barrier. Again, going back to my ... without reiterating into the cleanliness and normalization of field, that's always going to be this thorn in your side when you're trying to do stuff like that. And then, ultimately, regulation as well. Without opining too much on that, I think there's a ... Where people would like synthetic control arms and clinical trial based on real-world evidence, where the talk track is and where that is from reality, in regulation, I think is two different points.
Charles Rhyee:
Maybe let's talk about this linked claims with EHR. If it's already challenging to normalize EHR data and normalize ... and even claims data itself has some challenges there, I'd imagine linking the two is maybe exponential in terms of really creating a clean dataset to do analytics.
Paul Reuscher:
Well, you would think that, right? Having said that, it's a de-identified patient ID, just tracking patients longitudinally from EMR data to claims data. They can actually be very supplemental as far as the claims data might have better capture of certain fields and information, and the EMR data has less capture. Or it has a mile deep snapshot of these four medical encounters. Having said that, it's missing the other 15 claims for that said patient. So there's a bit of, when you link the two together and use them in combination, they kind of can supplement each other for use cases when you're doing the analysis.
Charles Rhyee:
I guess before we move on, it seems like the demand for this kind of data ... I ask this because more and more companies seem to be talking about, the datasets that they sit on, looking to monetize that data, particularly looking at the life science industry as the end market. Maybe talk a little bit about the market dynamics that you're seeing here as it relates to the demand for data right now.
Paul Reuscher:
It's changed a lot. There are a lot more entrants into the market, where they have some sort of an exhaust data asset that they're pulling together from whatever their proprietary business is. I think that one of the bigger problems you see is that there are a lot of people that have something that's maybe a mile deep and a foot wide for a specific therapeutic area, rare disease, or some snapshot, or some patient registry. I think expectations are bad. It's a long-winded way to get around to it. I just think that the demand is good, but there are a lot of frequent entrant in the market here that have something kind of small, and I think their expectations on what it could be are just ...
Charles Rhyee:
Too high?
Paul Reuscher:
Yeah, they're just too high, really. I see that a lot. But again, buyer beware. There's more and more things to draw your attention. Pharma life sciences, great buyers for data. If there's an industry sector that really knows how to utilize data to learn more about how to be better at their business, it's definitely pharma life science companies. But yeah, I think there's a bit of a buyer beware. Or on the flip side, just bad expectations on just lofty goals for what their exhaust data product could be.
Charles Rhyee:
You talk about buyer beware, but I get the sense though that, from some of the folks we've talked to, pharma is kind of just like an open ... I don't want to say an open checkbook, but they seem to be very open to buying all types of data. With the idea, "Well, if there's some type of insight there, we'll figure it out, but we'd rather have the data than not have it." Is that a fair representation of the biopharma market at the moment?
Paul Reuscher:
I almost snickered when you said the open checkbook comment. I think it depends on who you're talking about as far as who it is in the industry. Look, we do a lot of work with mid, small and emerging companies. The thing that really resonates with us and why we're pretty good at servicing our customers is that, more and more, I speak to people that are getting inundated by a lot of different offerings, but being a mid, small and emerging company, they have to do a lot with a little. So most of the time they're looking for the most bang for the buck. It's not just, "Who could buy the most data for the least amount of money?" It's more like, "How many different use cases can I solve or problems can I solve with something that I can acquire from somebody or license from somebody?"
So in my opinion, yeah, you're not wrong. Sure, there's definitely open checkbook scenarios. They're looking for new and novel. They're always going to be acquiring whatever the new, novel exhaust data is. But I think the vast majority in the space are more mid, small and emerging, and they're trying to be a little scrappier with what they license and how many questions they could solve with what they license.
Charles Rhyee:
I see. I guess we should think of the market maybe in two tiers, right?
Paul Reuscher:
Sure.
Charles Rhyee:
Like the Mercks and Pfizers of the world versus the emerging biotech space. Maybe bifurcate them, big pharma versus emerging, I guess, perhaps.
Paul Reuscher:
I would say there's definitely a top 20, top 25 pharma life science versus mid, small and emerging. But yeah, definitely, there's a bifurcation there.
Charles Rhyee:
Is that the same in medical device as well? The top 10 med device companies, do they behave like the top biopharma companies?
Paul Reuscher:
No, they're a little different. A little less on that open checkbook thing, but it's definitely ... First off, I don't think it's 20 or 25. In pharma, it's easy to say these top 20 companies all are a lot alike, or are more similar. Have seen whatever market caps, or whatever. And then the medical device, I think it's a sharper group. It's like eight companies, maybe a top 10, and then there are a ton of mid-level or small medical device companies. I don't even like to say emerging medical device because ... You want to talk about a miss in our healthcare system, I think if you're trying to be an emerging medical device manufacturer, I think that's a really tough road to go down.
Charles Rhyee:
Right. Well, moving on. We talked about claims data. We've talked about EMR data. Clearly, seems like this is where most of the value currently is, but at the same time, we're starting to see exploration of other new sets of data. Wearables I think is one category you increasingly hear of. Obviously, a lot of people walk around with an Apple Watch. Fitbit kind of started that years ago. Maybe talk about the utility of some of this wearables data. I mean, I think everyone thought measuring steps might be valuable, but more and more functionalities going into some of these watches, or the Oura Ring. Things like that. What kind of utility do you find in this type of data?
Paul Reuscher:
In the future, I think it'll be amazing. Hands down. Today, I think the collected data is interesting, but I don't know how much insight it provides. I think the wearables market is really most interesting from a ... lack of a better term, a PMR, primary market research component for patients, patient registries, clinical trials. Pharma companies are certainly already utilizing that for ... I have people in this clinical trial, we're going to issue them some kind of a wearables device so that they can start tracking things, monitoring things, or answering questions that we could push out to them. There are a couple companies that have hit the market here recently that show a lot of promise, especially around a PMR use case with wearables.
Having said that, one of the largest wearables company had entered the marketplace and was showing a lot of promise to come out with healthcare use cases, using their exhaust data to power things, and just to ultimately link this all together. They made a pretty quick exit from the space. Again, barrier to entry is pretty high. I think there was a lot of money spent. I think it's very hard to enter this market. I hope that other companies aren't discouraged by that because I think it's the future.
Charles Rhyee:
That's interesting. You said you think in the future it'll be great, but today, maybe not so much. What is that next step that needs to be solved for, to go from not that useful today to being a really important part of data being collected?
Paul Reuscher:
Interoperability, user participation. I've done a little work with some wearables folks. The information that they can collect from the wearable is not ... it's typically kind of beholden to however somebody is using it. Or, "I can issue you some kind of ... like a Garmin, but if you forget to put it on every single day ..." I mean, that's obviously a simpleton answer. But you get the gist, that if it's not collecting information, it's not collecting information. I think that's one of the problems.
And then the interoperability of it. Wearables is instant time ... From that perspective, it's instant health data, but how do I link it to your broader healthcare data? How do I glean insight? Is that linked to, whatever, EMR data or claims data, or whatnot? That part's still kind of a gap.
Charles Rhyee:
Is that a gap in terms of technology, or is that more of a gap in just trying to make that a more intuitive connection of how to apply it into the data that we have?
Paul Reuscher:
Both. I think there's a de-identification issue. There's, how I have to store ... One thing has to be HIPAA compliant. The other, I'm not sure. I say I'm not sure because I think it's still to be determined, with wearables data. And then, also, just doing the initiatives, the analysis of how is wearables data ... What use case does that answer when it's collected with claims data or EMR? When I link the two together, what is that collective story? I think that's still a little unknown.
Charles Rhyee:
Yeah, interesting. Another aspect that we hear more and more of is social determinants of health, consumer related data, because a lot of that has an impact on people's health ... ability to either access healthcare or to get treated, et cetera, or to afford healthcare. How has that part of the market been developing? We hear a lot about it from companies, but curious as to what you are actually seeing in the market in terms of using and utilizing it? Maybe first off, where do these sorts of data typically come from?
Paul Reuscher:
Consumer data is more Google type data aggregators, online data aggregators. Think credit reporting agencies or some kind of an entity set up as a exhaust data subsidiary to credit reporting agencies. It's out there and it's pretty readily available. The level of parity on it, that's kind of interesting. I say that because-
Charles Rhyee:
What do you mean by level of parity?
Paul Reuscher:
Well, it's tough to say, but if there's 10 people out on the market offering consumer data, I feel like six or seven of them are selling you the exact same file. There's a bit of parity between the information that you can gather and you get to this ... There's a pretty quick plateau to how good that information is right now. That doesn't mean that that can't be expanded in the future or won't be better. But at the moment you kind of get to this point of parity where it's good, not great. Its linkability via de-identification is good, not great. I'm kind of waiting for the next step on that, where that'll get better.
And then, unfortunately, I will say this, people like me or my industry are kind of caught in between a rock and a hard place with consumer data because consumer data, when boiled down to social determinants of health, when linked with claims data and EMR data, is super powerful. Talks about patient diversity. This helps with clinical trial recruitment and optimization, physician capture. Just understanding patient distributions better by race or ethnicity or wealth, socioeconomic status. Understanding that when linked with physician information of who is treating said patients, that stuff's very valuable to solving the problems in healthcare. This is a good dataset, getting back to this interoperability problem that healthcare has. This kind of also, once again, helps smooth out these gaps.
But I say we're caught between a rock and a hard place because, on the other side, there's a huge privacy struggle. California Consumer Protection Act passed. Other states have started to catch steam and sign things into law that have protections around it, consumer privacy. Again, on the basis of making sure that some company isn't selling your data so that you can get robo dialed for days on end, yeah, I get it. That's the reason why the law was passed. But it does also provide this barrier to people like me, people in this industry that are trying to take consumer data to link it to claims data, if every six months or every year there's another missing component on the consumer data. It's kind of an uphill battle there. I don't know what's going to happen, but it's an exciting time to be working with consumer data that you link to healthcare data. At the same time, I don't know, there's this dark shadow coming behind it that I don't know if it's going to mess this up or not.
Charles Rhyee:
Yeah. I would imagine with the consumer data, are the main buyers of this more on the payer side versus maybe the life science side, or is it spread out evenly?
Paul Reuscher:
It's both. Everybody has interest in this. Just understanding race, ethnicity, socioeconomic information about patients by disease or prescriptions of interest, or whatever, it provides a lot of insight and it's very helpful. The CMS, they've published a lot of information around social determinants of health and how important that is. Like I said, I think it's burgeoning. It could be great. At the same time, there's a travesty around the corner if we start limiting its collectability or usability of this data.
Charles Rhyee:
Yeah. You've touched on it a number of times, the idea of interoperability, and the importance of it. This is one area obviously a lot of people have focused on. We recently had the final rule on interoperability. I think that's for EHRs. You had the launch of TEFCA, that exchange. Infrastructure, sort of like a voluntary-
Paul Reuscher:
Tick Data?
Charles Rhyee:
Yeah, on the data structure, right?
Paul Reuscher:
Yep.
Charles Rhyee:
Does that not help solve some of the issues that you're talking about, or are these still just first steps to really getting that kind of level of interoperability that you would ideally want?
Paul Reuscher:
Could be a great first step. There are some great people trying to work with Tick Data, and other things like that, that are getting put out via the environment. Having said that, I've made the analogy, joke that it's kind of like payers were forced to create space junk. They shoot it out into outer space. It's there. If you want to go up there and pull it down, you can try to pull it down and make sense of it, but there's no regulation to how it was stored and how it's being put up there. Some commenter might say that that's not true, but from my perspective, there seems to be no rhyme or reason to how the data is pushed out there. It's gazillions of terabytes. There are great people that are trying to work with the data right now to figure out different insights out of it, but it's kind of like two steps forward, one step back. What's out there is space junk. Maybe somebody will make sense of it.
Charles Rhyee:
It's an interesting analogy. I guess, is this a ... Not to ask for more regulation, because I'm not sure that's the answer here, but when you think about the internet, for example, it seems like the industry eventually kind of coalesced around a limited number of standards, HTML, things like that, where common languages ... everything's kind of built on. Why hasn't that happened faster in healthcare?
Paul Reuscher:
That's a great question. That's a legacy problem. For so long, there was not ... I don't want to say disincentive. Nobody was really incentivized. Again, you mentioned the Affordable Care Act and the stimulus package around EMR. Until that came into place-
Charles Rhyee:
The HITECH Act?
Paul Reuscher:
Yeah. Until that came into place, there really wasn't any incentive for people to do anything like that and have some standards. And then now that it's here, again, it was just too disparate. I think there's just too many ... there's not enough incentives for people to be on a universal platform either. Because again, even health networks that acquire more subsidiaries, more child organizations to join their health network seem to have ... I mean, they can't switch each other off of whatever EMR software they're on. So you end up having these health networks that can tell you, "Yeah, we operate 14 different operating softwares." It's like, "Really?" Again, that's a great question, but I think it's a legacy problem.
Charles Rhyee:
Okay. Yeah. It's something that I think we got to somehow figure out before we can move on. Maybe just to close out, you mentioned it early on when we were talking about EMR data. Obviously, all the talk these days is around generative AI. It sounds like you think that it could help make EMR data more usable. What is your thoughts on the utility and reliability of generative AI in analyzing healthcare data?
Paul Reuscher:
I still think this is one of these managed expectations things. I think what people think it's going to do versus what it's going to do in the short term are wildly different. I meet with people constantly that have big plans for generative AI to do complete patient journey analysis or really deep complex analysis, or to just create generative algorithms to solve problems of data. But I still come back to, I think it's a really good tool and first step, especially in its pioneering days here, for cleanliness, normalization, and really kind of filling gaps, imputing fields, fixing what's broken and filling in the gaps. From there, I think will open the door to more complicated analysis to come. I will say that. We'll see how this plays out, but I think ... That's the one thing that just always resonates with me is ... I read the same articles that everybody else does. There's a big emphasis on it solving huge problems and I keep thinking, "Think it could solve small problems first really effectively," but doesn't seem much of a push on that.
Charles Rhyee:
Yeah. Maybe as an example, Teladoc Health recently announced that it's working with Microsoft to use the OpenAI and the Nuance DAX platform to automate the clinical documentation part of a telehealth visit. Is that the small problem, fill in the gap where you see this technology fitting in, in the early days?
Paul Reuscher:
Yeah, I do. Yeah. That's what I'm getting at. So hat's off to them. That's a good way to use that capability, to try to solve smaller problems first. As an example, I use ... because we were talking about it the other day, but we just passed through the anniversary for the Normandy D-Day invasion in early June. I just took a second. Something had popped up. I read something about the day. I think what touched me was I saw these 90-something year old veterans that are still alive, that were there that day, and they had ... I want to say they collected 70 of these gentlemen to be in Normandy. It just caught me. I was like, "Huh."
I looked at Bard. I looked at Microsoft. I looked at ... we'll put ChatGPT. I pulled open three different windows. I was frantically asking it, trying to understand how many people stormed the beach, how many died that day, how many are still alive today. Just, again, coming back to actuarial numbers, kind of an interesting thing. I thought, "We'll see what it says." I don't want to say this is a real scripted answer that somebody definitely has ... it's like a definitive answer, but it kind of is. I certainly got three different answers. I won't say which one was right, but I did get three different answers on the questions I was asking about something that was very finite. Long-winded example.
The reason I bring that up is when I start to ask these things to answer complicated questions, unknown answers, looking at healthcare data, how much precision is there? Because I asked it some pretty known answers and got many different answers, right?
Charles Rhyee:
Yeah. Certainly, once you're dealing with healthcare, where the impact of that, if you don't have the right answer, is obviously very profound versus a simple question on just trivia, right?
Paul Reuscher:
Yeah.
Charles Rhyee:
Well, I think that's a great place to end it. Paul, really appreciate you spending the time to give us this great overview. Look forward to chatting with you again in the future. I want to thank everyone for joining in and listening to the podcast today. Look forward to having you join us on future pods. Thanks, everyone.
Speaker 1:
Thanks for joining us. Stay tuned for the next episode of TD Cowen Insights.

This podcast should not be copied, distributed, published or reproduced, in whole or in part. The information contained in this recording was obtained from publicly available sources, has not been independently verified by TD Securities, may not be current, and TD Securities has no obligation to provide any updates or changes. All price references and market forecasts are as of the date of recording. The views and opinions expressed in this podcast are not necessarily those of TD Securities and may differ from the views and opinions of other departments or divisions of TD Securities and its affiliates. TD Securities is not providing any financial, economic, legal, accounting, or tax advice or recommendations in this podcast. The information contained in this podcast does not constitute investment advice or an offer to buy or sell securities or any other product and should not be relied upon to evaluate any potential transaction. Neither TD Securities nor any of its affiliates makes any representation or warranty, express or implied, as to the accuracy or completeness of the statements or any information contained in this podcast and any liability therefore (including in respect of direct, indirect or consequential loss or damage) is expressly disclaimed.