Using Patient Data to Better Understand Disease Progression in NASH
About this Webinar:
Nonalcoholic steatohepatitis (NASH), liver inflammation and damage caused by a buildup of fat in the liver, has historically been under-diagnosed. Untreated, the condition can lead to liver scarring and ultimately cirrhosis. 30 million adult Americans may have the misunderstood disease, with $5 billion currently being spent in annual related healthcare costs. Those costs will increase significantly by 2030 if NASH goes unchecked.
Unlike other liver diseases, NASH development pathways have not been well understood and FDA-approved treatments unavailable. In addition, many healthcare providers have been hesitant to test patients, particularly because a definitive diagnosis requires an invasive liver biopsy.
Recently, growing clinical acceptance of non-invasive biomarkers to measure fibrosis have opened up new pathways to understanding the progression of the disease. New therapies for NASH have now been developed, giving new hope to patients. Yet, for stakeholders like life sciences companies who are focused on fighting NASH, these breakthroughs create a new set of challenges: how to develop a better understanding of the disease’s epidemiology to guide treatment development and delivery.
New analytic approaches that focus on clinical encounter data allow organizations to see the “ground truth” for NASH patients and the healthcare providers who treat them. As life sciences companies shine brighter spotlights on rarer diseases, better data can lead to better clinical and financial outcomes.
In the webinar, we’ll explore how combining large clinical encounter data sets and composite blood test results has helped map the patient landscape for NASH as well as the impact of AI-facilitated insights.
00:00 Aswin Chandrakantan: Alright. Thank you so much everybody for joining today. My name is Aswin. I'm Chief Medical Officer at Komodo Health as Jessica mentioned. A little bit around Komodo Health itself, before we get started. Our mission is focused on reducing the Global Burden of Disease through the most actionable Healthcare Map. Our approach is focused on using the ground truth of a patient's journey through our healthcare system in order to better understand the opportunities to drive a better education and adoption of the right standard of care for patients. Our solutions are focused on building AI software products. We've deployed over now 125 indications, and our clients consist of 17 of the top 20 largest pharmaceutical companies, we also serve about 55 biotechs, a lot of players across the diagnostics and device bases, so a lot of additional applicability across life sciences, and we also cross into other verticals, including payers, providers, and patient advocacy groups.
01:13 AC: I think Jessica has actually provided a good overview of my background, but just very simply, doctor by background, a lot of expertise in healthcare analytics while in management consulting, I got a lot of product facing experience while at Google and then joined Komodo Health about four years ago to lead our product and platform development. In the last few months, I've focused on development and senior partnerships.
01:42 AC: Our agenda today consists of five key sections. First, I want to layout a little bit around the false north in healthcare and some of the issues that relate to data as it pertains to that false north. Next is talking about the visibility into a misunderstood disease, namely NASH in today's conversation. A little bit around our approach to our descriptive and predictive insights on NASH, case studies in other disease areas, and then I'll leave about 10 to 15 minutes open to question and answer from our audience that will be fielded through Jessica to me during the course of the conversation.
02:30 AC: So a predominance of our data in healthcare is open data, and what I mean by that is that you essentially have patients and you have a sampling of patients, a sampling of the visits that they have in the healthcare system, and you also only know a limited set of the encounters that they have. So let's take, for example, a congestive heart failure patient. They go to an inpatient setting where they might have an acute episode, they are in the ICU, they go down to a step-down unit, and then they're discharged to the floors. They're discharged from the hospital, they go to see their cardiologist a few weeks for follow up, their cardiologist then sends them to an echocardiogram. These are three different settings of care within the hospital itself, there were multiple layers and intensities of care. And the challenge with open data is that it, A, may miss the patient entirely if they're outside of the sample of any of these three settings of care. Second, it may miss their visit. So you may see the inpatient setting, but not the cardiologist visit.
03:44 AC: Third, in terms of longitudinality, you have patients that are constantly coming in and out of the sample so that means that this particular patient may change their insurance provider or move to a different geography, and the second that happens, you know nothing about what happens in that patient's care journey. And last is missing linkages, which means that you have a specific patient that touches a number of different providers, a number of different doctors, nurses, pharmacists, institutions, and the inability in using open data to accurately represent that patient's journey through healthcare, poses a large problem for us, as we're gonna talk through diseases like NASH, but also more generically across any therapeutic area.
04:40 AC: Komodo Health has focused its last three to five years on building this ground truth of the patient journey. And so people oftentimes ask, "What does it mean to have a ground truth?" So I talked about our Healthcare Map, but to describe it in words of this master truth we built for healthcare, it's the journey of 320 million patients, the diseases they have, the providers and institutions where they seek care, the therapies and interventions they use to treat those conditions and the outcomes that relate to that.
05:14 AC: And so that is where we focused a lot of our effort is creating a master view of nearly every patient in the US, and all of the activities that relate to their interactions with the healthcare system. In the middle, we talk about linkages because you knowing about a patient, but not knowing about the provider that they saw or the institution that they touched, only provides a partial view of the potential actionability and your ability to influence the standards of care. So not only do we have a master view of patients and the entities in the system, but you link them together because we see this as an organic vehicle of driving change in the healthcare system. And so knowing a provider without knowing the de-identified patient, knowing a provider without knowing the institution that they're linked to, all of those create gaps in actionability in healthcare. And lastly, we've created access to this ground truth through our software and our platforms, which we will talk about today.
06:23 AC: So just to hammer home the point around open claims versus actual patient journey data. Let's take, for example, renal cell carcinoma. There is the top patient journey represented in open claims, which is the patient has blood in their urine. You see none of the steps on how the way it end in which the system or the data shows you a "newly diagnosed" RCC patient that has been curiously placed on second line therapy. I just recognized that in every single one of these, there's lost intelligence in terms of understanding the patient, the journey, the providers and interventions that they've received and then also accurately modeling where this patient is in terms of their overall trajectory through the healthcare system.
07:18 AC: When you look at actual patient journey data that focused on where you see every single clinical encounter across every single instance of care, you will see that there's a newly diagnosed patient that has touched every single provider. Namely the urologist, the lab provider, the imaging center, having gone to a hospital potentially for a first line therapy and prevention intervention, and then seeing the reoccurrence of the community medical oncology center where they're actually understood to have had a recurrence and properly placed on the second line of therapy.
07:58 AC: The reason this is important here is because, A, on the right, you'll see that this patient potentially got a therapy that they weren't even eligible for or you're gonna think, "No, this provider's really strange, they skipped the first line of care," and this patient is coming on just burst onto the MAP and immediately, there's gonna be a lack of appreciation about all the interventions that has happened to this patient through the healthcare journey.
08:25 AC: So I also wanna use another example here that demonstrates a little bit around the open claims bias versus the ground truth of patient journey data. So I took a highly represented commercially-focused population around MS for a specific state and a single payer. And what you see on the left is a single commercial payer, all the data of that specific data aggregator only sees about 3000 patients and five visits per year for each one of those MS patients. Whereas on the right, when you get the full picture, you see in this force-directed graph the visualization about how many more linkages, entities, all the care providers and institutions that are involved in this these patient's care and you see that more than double the number of patients and five times the number of encounters.
09:23 AC: And that's the richness of today's market intelligence because there's no longer a world in which, "Hey, I just have an MS patient. It's an MS patient that is at a specific... " They're either having a reoccurrence of the disease, and then also understanding where they are in terms of line of therapy. So there's no label indication out there for like, "Hey, this is just for MS patients in general." Getting down to the specificity and the ground truth of where the patient is is incredibly important to understand and drive actionable interventions into the market.
10:03 AC: I think obvious to this is that the ground truth yields better outcomes. So there are sort of three ways that I look at our approach, and so the first is the attribution. So the correct attribution of the right patient, the right disease, the provider that they saw, the interventions that they've got. Next is describing the disease and being able to look at at-risk populations, modeling disease stage in complex patient cohorts, which is gonna be our focus today. And then the third arc is around predicting future behaviors, patients that might benefit from specific therapy to be able to drive the right stakeholder to the right intervention at the right time in the right setting of care.
10:56 AC: Now, let's jump into NASH. So I don't wanna go through the ideology of NASH in that I think everybody on who is joining today has a good understanding of the disease, but, A, it's just important to recognize that it's a spectrum disease. It's largely predicated on inflammation and liver cell damage. Second is that there's value in catching it earlier stage before it progresses to cirrhosis, because once you're in cirrhosis, you're in transplant territory. And there's been a lack of really good testing modalities, as well as treatments that has led to an underrepresentation of both the disease burden, as well as the potential opportunities and interventions of care.
11:50 AC: So by the numbers, we estimate about 30 million American adults may have the disease. There's about $35 billion in terms of the global market for new drug treatments and there's over $5 billion spent annually in healthcare costs that are directly related to the disease and probably billions more that are more indirectly related. Given a lot of the comorbidities that are associated with NASH, we see that the cost will potentially rise to $18 billion by 2030. A lot of that is in terms of appropriate treatment of comorbidities, also education around better self-care, so exercise and diet. And so this is going to become a really large burden in the US healthcare system if left unaddressed.
12:50 AC: The trouble with NASH is it's silent, it's underdiagnosed and it's misunderstood, so people can have it for years before symptoms occur. You can oftentimes slow it or reverse it but you need to know you have it before you can apply that intervention. It's more complicated than other liver diseases. The development pathways are not really well-understood. There's oftentimes a range in terms of the number of years between patients moving from having just the inflammatory factors to having... Which is NAFL to NASH and then finally to cirrhosis.
13:29 AC: It's traditionally been diagnosed with invasive liver biopsy and, as you might imagine, there's no therapy on the market. There might be no value in actually doing the invasive liver biopsy, so it's largely been sequestered to academic institutions that are trying to understand it a little bit or studying the disease but it's not been adopted on a larger scale. And lastly, there's now a growing clinical acceptance of non-invasive biomarkers, A the measure of fibrosis but the, B, actually tell you about that the patient is positively diagnosed with the disease, opening the doors for potential interventions and treatments.
14:17 AC: So our approach is fourfold and we're serving a number of life sciences companies in this space. And so really digging down into identifying the right compositive diagnostic tests in order to find the right heuristics to build and model the at-risk population here for NASH. So generating that patient cohort is really important and it's now based on a positive test for fibrosis. And so there's ways in that for us to, A, we need to develop a positively labeled patient based on a very specific patient cohort, then exclude all the non-NASH causes to focus on the addressable NASH population and then build a nationally representative predictive model from these truths that are part of the NASH patients based on patient journey, complete repair, complete data.
15:20 AC: So if you look here, there's actually... We've gotten gone from this very invasive sampling method to be able to diagnose a patient with fibrosis, the damage to the liver, to a really nice set of other tasks ranging from FIB-4 fibro test to other commonly used proxies around platelets and bilirubin. And what's really nice here is that we have used our platform of 320 million patients, of which a 120 million are payer complete, which means that we have an eligibility lot population. Every single clinical encounter for them over a multi-year period, we team that up with blood serum results from one of the largest US reference labs and we use that to score the degree of liver damage based on that test data. And then stage each one of those patients, saying, "Okay well, the lab result only tells us what their score is but outside of the context of how many encounters they've had with the system, what drugs and interventions they're on and also understanding then the potential opportunity to drive education and adoption of a better standard care for those patients, we needed to focus on staging to be able to really thin-slice that patient cohort."
16:52 AC: I think a lot of us would recognize that just highlighting patients that have a specific ICD code is not really interesting or relevant to targeting the at-risk population that actually has interventions that can be used to address their disease, and so this is where, our patient journey complete data, we can look back over a 10-year sample and then look at if this patient have a history of alcohol dependence, various types of hepatitis, malignant neoplasm of the liver, cholangitis, all of these toxic insults to the liver that would potentially exclude these patients and they're likely to be non-NASH causes of some of the inflammatory liver disease they might be seeing in fibrosis, they might be seeing... And we were able to exclude over a quarter million patients that through this exclusion criteria to really focus on the population that mattered.
18:00 AC: I think everyone here knows that the ICD code for NASH is 75.81. It is severely underused, which means that you can't predicate a national strategy or engagement planning focused on this specific ICD code alone, and when we looked at the analysis, and I'm gonna get to this analysis shortly, we've found that NASH patients who are on F2 to F3 stage, only less than a fifth of them are being treated by a specialist. And so this is largely a disease that is being treated by PCPs or not being treated at all by PCPs is probably the more less generous way of looking at it. So massive opportunity for education and to drive a better standard of care.
18:55 AC: So as we would have thought about deploying the NASH Model, there's a lot to think about. I think a lot of folks use predictive analytics in different ways and I think that the... At Komodo we've kinda come to a very nice teleological method. To first identify patients that may have a disease, understand their disease stage and progression, to understand what is the exact... Are they in the right part of the journey that merits intervention? Develop potential therapy outcomes and then predict future potential cost of treatment but then also the value of interventions on the market.
19:38 AC: So there's a lot that goes into building predictive models, even descriptive models. What I wanted to highlight here is that there's four different steps that involve training, applying the model, identifying patients and then identifying and managing their HCPs. I'm gonna focus on the last bucket almost exclusively today but not to understate the value, the importance and also the sort of intellectual property that Komodo has developed, around building the number one, two, and three, that underpins then the sort of actionable market intelligence that a life sciences company might need to engage the market in this space.
20:23 AC: To just briefly give you a sense of what some of the previous steps look like, A, lab data crossed with Komodo Health's data, so all of these tests that are focused on fibrosis and staging. We built a predictive model that then crossed that with our 320 million patients. We linked it to a positive set of patients based on our lab data, which was a portion of that 320 million patients and then we apply that model to our entire healthcare map. So the way you do that is, you look at the journey of all the positively labelled patients, you identify other patients and that's specifically in that payer-complete journey that actually might benefit for it, you build a new parameter to optimize and you build a really nice model using different... Either types of regression or random forest models to identify providers and institutions that manage those patients. Then you up sample that model to the full 320 million lives across the Komodo data set. And I'll highlight the value of why you start with payer-completed, which none of the legacy data providers out there do and the value of that approach versus other traditional and, I would say, less valuable or frankly wrong, so false North answers that you might get.
21:55 AC: So we found about... There's about more than 500,000 patients that have a NASH diagnosis but eight million that actually have an increased likelihood of advanced fibrosis over time, and over 900,000 providers, which is almost a good, third of all the entire provider base, so not just doctors but nurses, all care practitioners in the health care system. So really a very broad reach in terms of the number of providers that are actually touching patients that will specifically... That might benefit from intervention.
22:34 AC: First, you can build a model, you can throw it out into the market and you can hope that it gives you better answers but the Komodo differentiated approach is that we first pride ourselves on being scientists. And so we went through a lot of the data that looks at an expert NASH study that was done back in 2018, very manual, highly curated, multi-center, hundreds of different collaborators. And they went into a lot of detail in terms of understanding the NASH population and understanding what the specific stage of NASH was across each one of the patient bases.
23:18 AC: And I think what was incredibly valuable in this exercise is that we built the model and then also looked back at all of these expert reviews and we found that our predictive model performs just as well or what you might represent as a national representative sample and it actually largely mirrors what you might find in an expert study. So that gave us a lot of reassurance that, A, that we were seeing the same level of disease burden versus a multi-center study and that even an expert's grading could potentially be built into a model, that then we can use to generalizably understand the NASH market.
24:08 AC: And so we found about 39 million patients in the sub-cohort that we modeled on, using our payer-complete data set. And so our national reference lab was a couple of... It was about a 100 million lives. So we actually believed that the total NASH disease burden might be as high as 2-3X of some of the numbers that we're seeing here, across different stages of the disease.
24:41 AC: I think we also found a lot about when you take a population that is positively labeled, you look back at their full clinical encounter and patient journey. We learned a lot about their comorbidities and the other diseases that are likely contributing to multifactorial disease like NASH, that are constantly driving the liver insults. First of all, no surprise that Type 2 diabetes was deeply linked and glucose levels, all of that was highly tied to NASH but just the diagnosis of Type 2 diabetes alone was a predictor and a data feature that came out loud and clear. Hyperlipidemia, hypertension, and chronic kidney disease were also heavy contributors and really gave us an insight into the sorts of patient populations we should be focusing on.
25:39 AC: There's also a lot of specialties that are not necessarily hepatologists that are treating these diseases. And so it's from cardiologists and nephrologists, urologists, and not to mention, of course, the PCPs where we've found a majority of our earlier stage patients. So the idea here being, A, the patients are more complex than we might have ever imagined, and using this patient complete journey data, we were able to understand the phenotype of patient, but then also characterizing that patient in the context of their broader care team.
26:19 AC: The application into the real world treatment here is A, if you're able to drive earlier diagnosis in treatment, you get to better outcomes. And what our belief here is that we are going to drive... We're gonna identify at-risk populations, and their treating providers years before traditional treatment paradigms. And so what we found is that much more than half of the patients who might actually qualify as having a NASH code are even labeled with the NASH code in our data. And so that opens up a lot of potential opportunity to drive education and adoption around it, like standards of care, so creating that marketing intelligence for life sciences teams. And some of our clients had completely remodeled their field force model predicated on this intelligence around where patients are, which of the providers that touched them, so everything from field team sizing, placement as well as experience, is all tied back to the fundamental disease burden of this population.
27:37 AC: Komodo has actually done other previous case studies focused on disease state progression. I wanted to just briefly touch on them today. One is around hereditary amyloidosis and the second is around heart failure, only to show you other engagements and software products and the value of our platform, where there's been... Our product's been used for many, many months if not many years in a rare disease under-diagnosed patient population.
28:12 AC: First, to talk a little bit about heart failure. And I think there's about 10 median years survival years from diagnosis, three plus acute re-occurrences, so a lot of disease burden as well as a lot of constant in and out of the inpatient setting that drives up the cost of care, but also massively saddles the patient with other potential risks every time they have an acute reoccurrence. So we built using our patients journey complete data, we actually build a predictive model that then one of our software products called Pulse. It started providing alerts around providers, that was then taken by the field team of our life sciences companies. And so we looked at this from both the control as well as the alerted provider's perspective, sampling for providers that were going to be called on any way versus the pulse of alerted providers to get a sense of whether or not we were able to steer field teams to see providers that were treating at-risk populations.
29:27 AC: And so what we see here is the results are pretty spectacular. And that you see two times the number of interventions being written for these congestive heart failure patients. And largely it's because, A, the alerts were coming within a couple of days to a couple of weeks of the provider actually seeing that patient in that acute care setting, so there's a timeliness factor to it. And then secondly that, when the provider has the train wreck patient, that's an education moment that matters in the market. And so we saw a spectacular level of adoption versus our control arm of the study using patient journey data.
30:15 AC: It's not just one time, is the conclusion of this slide, which is that our benefits grew over time, and so what you actually see here is that month over month, there was sustained value in the alerts as well as predicting patients that fell into a specific threshold that was the education market. So this is a broader disease area, but only to highlight the value of, A, knowing, having the full patient journey from an inpatient setting to an outpatient setting, being able to model the key predictors of patients that would potentially benefit from therapy and then the providers and institutions where they seek care.
30:57 AC: The last case example that I'm going to provide is about hereditary amyloidosis. This disease has about 10,000 patients in the world, one of the companies that we were working with, they were typically... They were seeing anywhere from finding about 15 patients per year, oftentimes 10 to 12 years of post-symptom onset. And once you find that in this five years of median survival, even when they got the therapy and there was a lot of clinical encounters in the system as this patient finally got diagnosed, and well there's a lot of multi-system involvement here. So running around, seeing the cardiologist, while going and seeing your nephrologists, talking to the neurologist about your peripheral neuropathy, so multi-organ involvement, complicated disease, rare disease. Differential diagnosis number 70 probably for any provider out in the market.
32:03 AC: We were able to identify these patients. We moved it from being like, "Hey, we're finding it 10-12 years post-symptom onset to finding it three to five years post-symptom onset. And so what you see here is all the different disease areas, the codes of everything rating from cardiomyopathy to kidney disease to edema, peripheral neuropathy, being billed by a constellation of providers in the system that finally led to this moment of, "Okay. Let me actually run the test for hereditary amyloidosis." But what's clear from this rainbow chart is that years before this moment, you could actually have the patient tested here because you have all the heuristics that would tell you that the patient would likely benefit from testing for hereditary amyloidosis.
32:58 AC: And so in a real world verification of this model around finding these patients, we had... Normally, there's 0.03% of providers who are ordering this test. We drove that as a 30x testing rate on that base of providers, and that's focused on 3,000 providers that were seeing patients or group of patients that would benefit from testing. So we go from a world in which we're spending years leading up to a test or even decades in the case of hereditary amyloidosis, testing them years earlier, so that disease progression is slowed. The inefficiency leading up to testing is reduced, better outcomes for patients, better care standards for patients, earlier intervention, earlier diagnosis, earlier treatment, and overall savings to the entire system.
33:58 AC: In summary, I come from the background at Google, and everybody thinks of Google as being the best AI company in the word. And the truth is Google has the best AI because they have the best data. And I think the same is to be said for Komodo Health, where if you have patient-journey-complete-data, eligibility log populations for hundreds of millions of patients, and you're able to then fully understand and build data features around that complete data, it's going to get you a better model that's gonna lead to better prediction and better pull-through and actionability to reduce the disease burden for patients, and also the adoption for the right standards of care in the market.
34:42 AC: The second piece as I highlighted here is around when you don't know what's happened yesterday, you really have no business predicting what to do differently tomorrow. And so a lot of the legacy data providers out there, are taking their open-sample data-sets, they try to build models around it, those models are essentially training for all the gaps in the data. And then they come out and they try to use that model, and that model just simply useless because it has so many holes in it that you really need to focus in a world in which label indications are narrower and narrower. You need to focus on the right patient at the right time in the right setting of care to drive the right intervention for that patient.
35:32 AC: And the last piece as I highlighted one of our case examples, and I think that NASH market is due to explode, and I think building and modeling around Komodo's expertise of predicting diseases, predicting patients, predicting providers and institutions where they seek care, it creates massive value for patients, and it reduces the total disease burden, the total cost, and it gets patients to the right standard of care earlier. And given a disease like NASH where there is massive disease burden, but it's under-tested, there is nothing but both descriptive as well as predictive models which should be honed in order to address a disease that very few providers think they actually are seeing in their day-to-day.
36:31 AC: Last slide, but it's important to say that Komodo Health, we've built this ground truth, we have a number of solutions that are built on this ground truth. Aperture is our AI-driven support for market intelligence and planning. Pulse which I alluded to in this particular case example with some of our other disease areas, provides real-time alerts where you can actually look into the providers, that's all patients that might benefit from testing or from intervention. And we've built a really nice clean interface, and the fact that we see encounters within hours or days of the clinical encounter itself, build... Gives you that actionable moment in the market.
37:19 AC: And lastly for some of our most pressing problems that may not be perfectly fitted in to software, we have an advanced analytics platform and a team that can help you take it to last mile in terms of serving your brand, medical affairs, or clinical development needs. With that, and a lot of words there, I thank you for your... Thank you for listening, and open it up to audience questions.
37:49 Jessica Cerka: Aswin, it's Jessica. Based on a few of the questions that we're getting, I was wondering if you could describe how Komodo works with lab providers and how can we leverage Komodo's technology with the lab results combined together.
38:06 AC: Great. So there's two arcs to that question. First of all, of the 320 million patients that we see... We see the tests, the fact that the tests were ordered for all of those 320 million patients. So that's the piece that I think I wanna start off by saying, which is, we see all of those encounters that involve labs being ordered. Of the 320 million, about 120 million, we have partnered with some of the largest reference labs in the US, so we see for that population, we see the lab results themselves. And so it's very simple in that if we have to do a specific study we can bring in that... The reference labs, we can bring in the lab results, we can tie it to our healthcare map, we can build models around it, or we can actually identify patients at a specific point in their journey. So all of these are out-of-the-box capabilities at Komodo Health and are all integrated into our platform.
39:09 JC: Okay. Great. Second question, can you identify changes in the patient journey before and after the launch of a product?
39:23 AC: Yes. We, about three quarters of the brands that we serve are launch brands, and absolutely we use it in order to see, okay, well, is this specific institution or provider-base, are they starting to test? So let's just say that a specific disease requires you to be tested for EBV-positive, we'll immediately see the institutions where they're not even testing for EBV and therefore precludes them from getting that therapy, and then we can also see the realtime launch view of the adoption of the therapy itself. And so a big part of Komodo Health's healthcare map is, A, understanding the journey before, but then also responding to market changes as they happen in the context of launch brand.
40:17 JC: Okay. This is a question about what I believe is a typo on Slide 24, the bar chart on the right did not add up. The percentages didn't add to 100, but I believe the 67% was supposed to be actually 63%, according to our notes here.
40:42 AC: Yes. That's probably correct. I apologize if I... Yeah. One of the pieces also there is there's probably a number of places where there is a little bit of rounding in the calculations, so that may have incorrectly represented on a couple of percentage points.
41:00 JC: Yeah. Yeah. It was 63% according to our notes. Another question is, how does Komodo identify providers that are active treaters or are actively involved in diagnosis? Does Komodo leverage AI and machine learning to profile physicians?
41:16 AC: Yeah. So I think one of the things that I focused on a lot today is the fact that we have more clinical encounters per patient than any of the legacy data providers. And A, within that you see more encounters of care but B, when you actually take into the encounter itself, we have more detail, for example, around service line. So let's just say that you have a specific provider and then during an inpatient setting, you won't simply just see, "Oh. This is internal medicine, their doctor who discharged them," you're actually gonna be able to drill into which is the provider that actually got the diagnosis, wrote it for them, versus who are the folks or the providers that actually prescribed this specific therapy. So all of that is embedded within our healthcare map. And so when I talk about death, it's not just more encounters per patient, but I think the audience member is very astute in asking question of, "Well, what do you actually by death?" And it's not just... It's also about the representation of the number of encounters and the care team members within the context of the specific care event.
42:32 JC: Right. Thanks. Next question is access to data is fundamental to robust analysis, can you comment on how Komodo Health is differentiated from competitors like Optum who conduct similar analysis?
42:46 AC: Yeah. Optum, for example, is a single payer, whereas Komodo has partnered with over 165 payers contractually, so most of the large regionals and nationals. And so one of the things at Komodo we pride ourselves on is that we build software products and intelligence platforms on top of that. We don't settle on a lot of PLDs so what's really nice there is that these payers love working with us because we help them with their scoring, and then they... We get data [43:20] ____ usage rates that we can use in our products, and that allows us to have this eligibility locked population for hundreds of millions of patients that come to... [43:33] ____ or Optum, great. You have a couple of payers here and there, but it's nothing to the robustness of 320 million patients and 120 million payer-complete lives. And we also work contractually with a number of the large EMR data linked assets. We have some partnerships that are both announced as well as unannounced and so a lot of this EWRRW... Sorry, real world evidence and [44:04] ____ AGOR work are... Komodo is essentially powering a number of those products on the market.
44:10 JC: Okay. That's all the time we have for questions. So thank you to everyone who joined, we will have a recording that we can distribute afterwards, and if you have any questions you can send them to firstname.lastname@example.org.
44:37 AC: Thank you so much for your time.
44:40 JC: Thank you.