The future is now: Transforming health care
Exploring AI, analytics and the future of health care
- Welcome. I'm Curt Medeiros, I'm the president of Optum Analytics. I'm very excited, not just to kick off the data and analytics track, the Purple track, but also hopefully, over the course of the next couple days to be able to get to know you all. I know we have relationships in some way shape or form with each of your organizations. So my hope for this track is that we can help you think more broadly. Part of what you saw in the session downstairs is not just how we use data, not just how we turn it in to actionable insights, but then how can we actually work with you all to deploy those insights, to do something about it with actual patients? That's always very inspiring to me and part of what keeps me here at Optum, and keeps me fighting for our mission. And I really want you all to be creative at thinking about "what are the other things that we can be doing that we're not yet doing today?" And so we're gonna talk a bit about what we see as the future in data and analytics. I know Norman talked about it as Optum's superpower. I think maybe that's why some of you came here to this session. We believe that we've done a lot of really good work, but in the health care system in general we're just at the beginning. We're at the tip of the iceberg in what we can do to leverage data and analytics. To drive better patient care, to lower costs, and to improve the satisfaction of all of our customers. With that, we get to the main event. Tushar gonna take over, introduce our panel, and then I will come up at the end just to give a few thoughts. So with that, Tushar take it away.
- Very good, thank you Curt. Good morning, and welcome everyone. My name is Tushar Mehrotra, lead Optum's Analytics services business. I'm thrilled to be here today. We have a fantastic topic we're covering focused on how new and emerging data sets are driving innovation in health care, and specifically impacting stake holders across the spectrum. So from payers, providers, life science organizations, consumers, and patients. We have a fantastic group of panelists here that are gonna touch on a variety of topics as part of this, from consumer data, social determinates of health, qualified entity data, and even more broadly, real world data. So with that I'm gonna ask each of the panelists to introduce themselves, share a little bit about their backgrounds, and then which specific topic they're gonna be covering as part of this panel. Paul would you like to begin?
- Yeah, thanks Tushar, I'm Paul Higday, I lead our innovation and strategy teams within the broader Optum context. Joined Optum about a year, or 18 months ago or so, from the advisory board which has been in health care for a long time. I specifically have a focus on a number of different areas, mostly confined to the question of what's next in health care? And today we'll be talking a lot about what's next for consumer data in health care and how we can lead the health care industry forward by about a decade by combining consumer information with medical information to really drive a deferential patient experience and hopefully get some consumers that are actively disengaging from health care to rejoin their care continuum.
- Hi everyone, Jay Hazelrigs, I lead our provider actuarial services practice at Optum. It's just what it sounds like, actuaries working for providers, health systems, ACOs, we work with these organizations, with their journey on value based care. That's probably the easiest way to think about it. Typically we are answering three questions for our customers. One; "if I'm gonna enter in to a value based care contract can I be successful?" "If I do enter in to it what are the areas of focus that I would need to work on or put interventions in to achieve that success?" And three, many times we're being asked by organizations or customers; "I made a contract, a value based care contract, it's sort of mid-year, what's the score, how am I doing in achieving success?" So, I'm gonna be talking about qualified entity data, and we'll talk about how that data, we're applying that to working with our customers and answering those three questions.
- Jim Dolstad, I'm vice president within Optum advisory services and the payer actuarial area. My focus is really working with our predictive modeling tools with Symmetry and Impact Pro. And then also leveraging social determinates of health with in those tools and also out side of those tools, really to help drive clinical performance, as well as financial outcomes.
- Hi everyone, my name is Eleni Argentinas. and I work on data strategy and data partnerships in our commercial data licensing business. Historically I come from the pharmaceutical industry, I have a legal background. And I'll be talking about how real world data has evolved and added increasing value to life sciences organizations and how it might also add value to the rest of the health care eco system.
- Pretty good, thank you all. So we're gonna start with the topic that you probably have heard about, there's a lot of buzz about it. So what I'd wanna focus on initially is social determinants of health. There's a lot of buzz in the industry, you go to any conference you hear about social determinants of health. What I'd like Jim, for maybe you to speak a little bit about is "what's real and what's not as it relates to social determinants of health, and how are payers and providers using this data today?"
- Sure, thanks Tushar. When we look at social determinants of health it's definitely a buzzword. 17 sessions at AHIP this year had "social determinants of health" in the title. But many of them are talking about it in a different way than Optum does. Most of the industry is talking about pulling information from the member, from survey data, or from using the Z-codes. And we know the z-codes contain about 1% of information from members, and we also know it's very hard and resource intensive to do surveys on the medicated population and some of the Medicare population. We've actually taken it a different step and we are pushing information to clients on 90% of the members because using consumer analytics, social economic data, and community data, we can get information on the individual member, and push it to our clients so they understand member's propensity to engage, their social isolation, their risk of homelessness and several other factors that allow us to drive clinical and financial performance. And it allows payers and providers, because providers are really risk bearing entities in many cases, to be proactive, not reactive. So with no claims we understand whether somebody has a condition or not with pretty good accuracy. We also help them know the member, not just the patient. So we understand what's their primary language going to be, what's their propensity to engage is, and their social economic status. And then we're been able to reverse engineer the process by looking at it and seeing that provider-patient bases vary based upon social determinates of health. So one provider may have people that speak English to a lesser degree than a different provider. There may be a lower propensity to engage, or a higher degree of social isolation. And by looking at the differences in the provider patient base, we can understand how certain providers are in fact really doing a great job, and others can learn from them.
- So building on that, Jim, where do you see the market going for social determinates of health? If you fast forward the next three to five years. how do you see payers, providers using this data, and what do you think is the art of possible as relates to the social determinates of health?
- The sophistication of how our clients are using it is getting deeper each month. It used to be a propensity to engage was used to improve identification and stratification. And social isolation would be used to improve readmission rates. Now we're getting in to much deeper topics of how can we improve coding and documentation at the member level. And at the individual provider level by understanding what the member's needs are, what their likelihoods so that then generates improvements in revenue from CMS in both the Stars area, as well as the coding documentation. We're also able to look at it from the financial experience of understanding why a certain block of business, or certain patient base, may not be performing as well as others. And now we're starting to get some deeper questions integrating it in to Impact pro and some of their more complex things, where there's some RX complex measure which is "what is the complexity of a particular pharmacy regiment that we're asking a patient to undergo?" and now we can start to understand what the member's adherence might be to that because some members are gonna be able to adhere to that regiment more easily than other members. So again I think the complexity's increasing.
- And then where do you see the limitations as it relates to SDOH? What can we not do with this data?
- Some people are pushing the limits, and we have had some of us wanting to predict life influencing events where if something happens like divorce, or a death in the family, then what are all the additional costs that would be associated with that. And so we've steered away from predicting divorce and other things at this point, so we don't wanna go down that path, but I think there's a lot of good traction we're doing in helping the system work better for everybody. And we continue to get challenged by our clients and virtual visits and other areas, and now we're starting to work with some of the digital technologies as well, to figure out "who are the hard to reach patients?" because if the digital technology can reach the hard to reach, then it makes the success much easier because it's easy to reach through the care management and clinicians.
- Perfect, thanks Jim. So let's stay on a similar theme and talk about consumer data more broadly. Paul, maybe can you speak a little bit about how consumer data is being used today to improve health care decision making.
- Yeah, sure, so we're gonna start with gathering a little data from the room. How many of you feel like health care is more expensive for you personally, than it was say, three years ago? Raise your hands. Okay, everybody who raised their hands this is about the 80% of Americans who pay more today than they did three years ago. How many of you have changed your healthcare decision making process, in other words, chosen to do something different as a result of financial concerns? That's about right, so it's about 40% of the US healthcare population has actively changing it's behavior. So when we thought of this idea of consumer directed healthcare and pushing more of the cost of the consumer to have them understand and make conscious financial choices about health care in the US. We did so on a belief that it would curb unnecessary utilization and basically over emphasis on care when it wasn't needed. And it's done a little bit of that, it's certainly put the breaks on spending in the US health care environment, but the other side of the unintended consequence has been a much more concerning issue which is as significant majority of people who have financial concerns now, are delaying or avoiding care completely for conditions that desperately need care in some way, shape, or form. So we have a gap. You have people who can afford that are not really affected as much by the need to pay, you have people who are kinda on the brink, and you have people that can't afford. And the problem is the disparity discourage people that can't afford care to avoid using care. So what do what mean? That means that the average person that has the onset of orthopedic symptoms, say a pain in your knee joint, delays about five years from the onset of symptoms to actually getting the full treatment they need to recover from the knee pain. During that time they tend to gain weight, they rend to have a higher predilection for diabetes, they tend to become less mobile and their overall health status tends to decline. So yeah, we manage the cost of the orthopedic procedure to time it better for when they can afford it, and we cause an onset of chronic conditions as a result of that behavior. So, consumer data in health care is not about knowing every thing about every body, it's about engaging consumers to get them to use care when they actually need it. So in other words, it's shifting utilization of health care away from the general utilizers, who are generally healthy and in good shape, and in to those people who are most likely to be the most costly and have the lowest quality of life in the near term because they're not engaging in care properly. And the question is, how do you use consumer data to do that? So how many of you have worked outside of health care, say in the retail, consumer package goods, etc. industry? So I came from there a little bit as well. They're about 10 years ahead of us in the use of consumer data. And in general they use it for three things. They use it for marketing, so what we call "consumer acquisition", they use it for engagement, which means trying to keep your brand engaged using their products etc., and they use it for service. So they use it to figure out how to provide you the services that they think you need and that you think you want in the conversation. We do almost none of that in health care today unfortunately. And so the cutting edge is how do you take the same consumer data that's being used in other industries, and roll it in to health care in a way that meaningfully promotes appropriate use of care for consumers?
- That's great, thanks Paul. Just building on that, have you seen any use cases around using this data to predict potential outcomes or conditions? And can you speak a little bit about that.
- Yeah, so, you'd be amazed. A pretty significant portion of the conditions we all suffer from can be predicted without much health information at all. There's a couple that are blatantly obvious. If you smoke, you're much more likely to get lung cancer, that's an easy one. More complex one is what's your predilection to get diabetes based on a combination of your social determinates of health information, your income strata, where you shop for food, when the last time you went to a gym was, a whole bunch of other stuff as well. Which are not truly social determinates, they're more what we call "consumer engagement in the outside world." And what we've been working on is about 45 different condition predictive models that leverage the health information that we know exists in the market, labels consumers with those health statuses, and then tries to predict off of just their consumer information, their likelihood of having a specific condition. Turns out you can do that for about 60 of the major conditions in US health care today, at an accuracy rate of about 10-15 times greater than the incident rate of those diseases in the US population. Which means if I went to a random sample and said "how many of you have diabetes?" I might find 4 or 5, 6% of this room. We can predict somewhere around a 30, 40, 45% rate when we know enough about you from a consumer stand point. So we'll talk later about the privacy implications of that, and what it means and how you use that information appropriately. But in the right hands that information transforms the way payers engage with consumers, by trying to get the people that need screening to come in. It changes the way providers intervene with consumers, which makes sure they get the care when they need it the most, which actually can be volume increasing, whilst simultaneously lowering costs. And it changes the way life science approaches consumers 'cause it switches you from a scatter shot, a mass marketing concept, to a very targeted out reach to consumers that really do need your help and are much more likely to use the product that you manufacture.
- You hit on this just a little bit but how do you see access to this data fundamentally changing how providers and payers engage with patients going forward, today versus maybe two years from now, three years from now.
- Yeah, I think we're just entering in the area of what we call micro-segmentation. It used to be that when we did drug development for example, we looked for the largest populations we could possibly reach. So we'd add block buster drugs like Lipitor and Plavix, and all those. We're past the age of block busters in some respects and we're much more focused on individualized medicine. Genomix is just the tip of the iceberg there. What if that information was combined with your consumer information? So, not only do I know that you are prone to diabetes from your consumer info, but I know that the way that you access the doctor leads to a decreased likelihood that you'll get access to your insulin, which is a social determinates question. And also that you're likely to be off schedule on the application of that insulin because your family life's put you in a situation where it's very tough to maintain a regular schedule. Those three things combined put you in a situation where you might recommend a completely different regiment of insulin treatment for a diabetic patient than somebody who's a little bit more stable, or can do things in a normal time period, and keep that blood sugar at a relatively appropriate level. So we think there's two parts to personalized medicine. The first is the actual treatment regiment, and the second is the application of the consumer understanding to that treatment regiment to make sure the consumer can actually adhere to it, follow it, and receive the treatment that's appropriate.
- Perfect, thank you. So let's pivot a little bit and talk about how data's used to improve health care operations. And so, Jay, I'd like you to speak a little bit about qualified entity data. And so maybe just start with, what is it? Who has access to it? And how is it used today in a provider setting?
- Sure, sure. I'll start off, these guys get all the sexy data discussions, they left me with the claims administrative data and what we can do it, but I think you'll like this topic. Qualified entity program, qualified entity data, do folks know what this is? A show of hands. I'm counting nine. Maybe 10. Great, so qualified entity data, what is it? This is, first of all, it's traditional Medicare data, claims administrative data, so that's the 40 million beneficiaries. We have what represents as part A, part B, and part D data. Part D data so that's a new one, we'll talk a little bit about that. So that's all of the 40 million beneficiaries, it's all of the data, and we have access to it. Okay? For nine years. That's a big, new opportunity that we've recently been able to gain access to. So, how many organizations out there, so this is through Optum labs has access to it, we're able to use this data for our customers. Across all the United States there's about nine entities that have access to this. This has really been coming online the last 18 months, year, 18 months. And we've been using that to really refine our modeling, helping those questions that I talked about earlier. Helping our customers really understand those questions better, so that's been a game changer for us. I'm a talk a little bit about that, before we had data that we used for our customers, especially when they were in the federal ACO programs, or the Biopsy programs, the Pathways to Success, MSSPs, and we used to try to answer those questions with what's called limited data set, LDS, the limited data set which was 5% sample. And we would work with our customers with that, and what we always had was a little bit of blind spots in that. Now as actuaries we think we're pretty good at estimating those, or projecting and trying to figure out what those, making assumptions and saying, "okay, this is a blind spot but with all the data coming together we think it's this." and some of those, just the level sets, some of those assumptions we're having to go back four or five years, for these Medicare Assurance Savings programs. And there are programs making estimates. With this data set, now we're not making estimates, we're able to say "okay, for these programs, and all these perimeters and terms and stuff, in answering the question "should I enter in this contract, can I be successful?" we're able to harness that set of data and really refine how we answer those questions and provide a better answer.
- Great, from a payer standpoint, Jim, maybe you could talk a little bit about how qualified entity data is being used today.
- Sure, let the two actuaries talk about the data on QE side bit. I did this presentation at the Society of Actuaries about the same percentage of people knew what the QE data was. It is valuable because it is the A, B, and D data that payers can have access to it as well, although a little bit more on a limited basis. But it's still valuable because it allows a payer, or a provider to see the flow of a member from year over year through different programs, so as long as a payer has the member from a commercial plan or self funded plan, through to maybe a Med-sup plan, in to Medicare advantage, all that data can be linked together across year. The data can also be augmented by using the HC risk going methodology we put social determinates of health in to that and scored it as well, so we know those who are highly socially isolated cost 20% more than those that are least socially isolated. So very valuable for the payers. It is able to address the same types of questions.
- Jay, going back to you, how do you see, you hit on this a little bit but, how do you see this being used differently than traditional administrative claims data? And then, what's the path forward as more and more organizations learn about this? What's the path forward for what you can do with QE data?
- Sure, I think the game changer is the comprehensiveness of the data, and able to, I think you mentioned the micro-strategies. How do you start thinking about, from a network, or from a group of physicians, and what are their, from the panel, what they're managing, you can get a much better insight, and deeper insight in to what really are the customer's needs, the patient needs. So when you think of, as we sort of pivot and become more nimble with all this data, right? We're fighting this, we've got all this data, it's gonna add more information, but we also have to be able to use this data quickly and be nimble with it. So I think it's harnessing the data to be nimble, to really feed those micro-strategies, and bringing it together, that's really how I see that sort of changing. Or that's a path for it. As for "how do you start getting involved with this data?" I think first of all, I said we've only been using this data for the past 12 or 18 months where they came online, I think we're still figuring it out to be honest. What can it be used for from the provider's perspective, from the payer's perspective? But I think we welcome engagement on you know, what's the art of the possible, how do we think about this? I really think for network shaping, creating the system-ness that needs to happen, and the nimbleness that's gonna come online.
- And the last question on this topic is; what could a provider do with this data that historically they wouldn't have been able to do that can fundamentally bring near term value for a provider?
- I love saying this little quick example we have, I'm gonna go back to the Medicare Assured Savings program. So, without getting in to details, you have to say "who's the providers in your ACO?" and then you go back and say "okay, how did they deliver care in historical settings?" and "how are they doing it in the performance year?" and you sort of compare that and whoop-di-doo, are you making a financial success or not? That's sort of how it's laid up. One of the issues that these ACOs had is they were getting new providers, the providers are moving around, the physicians, and what they didn't have access to was "what were the tens that these providers used to be in?" and so, inherently what you were doing was creating this risk, you were creating a risk that you didn't have, you had this blind spot, this history of these providers. We didn't know where they were, we didn't what the tens they were in and therefore we had the blind spot in the back, in the historic data. Now, with this data we can easily link it and follow those providers, we talk about following the patients, we can easily follow the providers too, and have comprehensive information. And I know that probably doesn't sound like a lot, but that could easily make or break someone being financially success in an ACO.
- Right, so let's pivot a little bit and talk about real world data more broadly. And so, we've talked a lot about payers, and providers, and consumers, let's talk a little bit about life science organizations as well. And so Eleni, maybe you can talk a little bit about how is really world data being used today? From a pharmaceutical or a life sciences perspective from a commercial stand point, all the way through from to an RND stand point, with precision medicine, clinical trials, what's valuable to a life science organization, and what are you seeing work well?
- Sure, before I start answering that question, could I have a show of hands of how many people are in the life sciences, bio-tech, or device industry now in this room? Okay, all right. So maybe we'll think about this as information we can use to ask and how could this also be helpful in the provider-payer space. But, life sciences companies, drug companies, and device companies historically use real world data in two main functions. One was marketing and sales, predominantly for targeting, it was sort of a blunt instrument to go out and understand what physicians they should approach, and what messages they should deliver, how a brand was performing. And the other was predominantly in their health economics and outcomes groups to do two things. One, establish the burden of disease and the cost of that disease, and unmet medical needs in that space. And then differentiate their products by doing comparative effectiveness for surgeon, and comparative cost effectiveness and cost benefit research. Over time, let's say about 10 years ago, many of you are familiar with the Sentinel, the FDAs said maybe we can use real world data to surveil the market place and identify safety issues after clinical trials. Because as we all know, clinical trials are small, they're getting even smaller now with precision therapeutics, so you don't really know how they'll perform in the real world. And they created a network of 18 data partners, some payers, some large provider networks, and began to use that data on an ongoing basis to identify drugs that had post-market safety signals, and then adjust their labeling and take the appropriate regulatory action. But in that last year there has really been a game changing event that many of you may be hearing about, which is the FDAs exploration, and now, utilization of real world data as real world evidence to make label changes. A really good simple example of that is Pfizer's drug IBRANCE, which is approved for breast cancer in women, was just recently approved for breast cancer in men. Previously this would have had to be done through an organic clinical trial which would have taken an extraordinary long amount of time. And extraordinarily expensive. But through the use of an oncology EMR, a claims data provider, and data from their own safety data base, they were able to negotiate that label expansion with the FDA directly. As someone historically from industry I could not overstate how game changing this is for the industry, because it's going to fundamentally change the way drugs are developed and commercialized. In fact, when you think about the continuum of regulatory decisions, once a drug is approved you can then begin to use that real world data to gain understanding of new populations in whom the drug is demonstrating effectiveness, you could package that data up, and at a minimum, go have a more intelligent conversation with the regulators and say "hey, we're seeing effectiveness in a new age group, a group with different renal function" any dimension of population expansion that you can think of. But what's even more impressive than that is that there are even about 10 or so oncolytics that have been approved for their first indication using real world data. You say well "how was that possible?" because they're using the real world as a control arm. Either as a historical control where they're watching those patients before they enter a clinical trial, and then looking at the intervention and how it's effecting their outcomes, or they're looking at, they're creating a propensity score matched synthetic control in the EHR so they can accelerate the execution of that trial, and also lets say, eliminate some of the ethical issues with creating a placebo controlled trial, and some of these really devastating diseases. This is really changing the entire continuum of drug development and drug commercialization. And I'll go maybe a little step further which is, we think a lot about surveillance for safety but I think the leading edge companies are now thinking about real world data as surveillance for effectiveness. So I will monitor pan-therapeutically across all therapeutic areas how my drug is being utilized, and where I see populations that appear to be benefiting. I'm just gonna either use that to immediately grab those patients and enroll them in trials and go forward, or if I think the evidence is compelling enough I'm gonna go have a dialogue with the regulators right away. So it's a very, very exciting opportunity. But what that makes me ask for those of you in the payer-provider industry is "what does surveillance of populations look like there?" And "do we have the same opportunity to surveil populations for how well they're doing on therapies and care choices, and care decisions?" and, if you could get that data quickly enough, "how might one adjust those, and how might that affect outcomes over time?"
- Great, we talked about a lot of different data sets here. Eleni, can you maybe speak a little bit to how data linking is happening in today's world. If you talk about claims data, EMR data, social determinates, QE, what's the value in linking data? Maybe not just from a life sciences stand point but just broadly. And what's the art of possible with what you can do with it?
- Sure, so, one of the first places that we began linking data together was electronic health record data to claims data, and you can now find that, let's call it readily availably in the market. But what is really important about that is that the nuances that you can derive out of that clinical data, that you can then marry to cost data to really understand to differential costs and the differential outcomes across populations. You couldn't really do that before, claims alone are a bit of a blunt instrument. But clinical data has so many nuanced variables, labs, and clinical assessments, even symptoms. I think one of the interesting analysis we did was we actually mined the electronic health records for mentions of migraine and we were able to take that data, when we linked it to our claims data, and show that for every mention of migraine, there was an associated increase cost to take care of that patient. So that first, let's call it a fundamental kind of linking, we've been doing and our clients are very enthusiastic about that data, but we're also being approached by clients to link their data to our data. So patients in a clinical trail, how are they doing long term, for a long term outcomes? Registry data, many of our clients want to enrich what's in a registry with populations that we have available to us. Several of them are beginning to talk to us about linking genomics data and many of them are entering in to the, what shall we call it? The digital medicine area, where they're developing apps and applications to support patient health, and now we can observe those patients in our data and help them report out on outcomes and demonstrate the value of those digital interventions.
- Great, and then two final questions for you. How do you see real world data being used as real world evidence of part one of the question? And then, specifically, how do you see this data being used from a precision medicine stand point, either from a life sciences stand point or from a provider stand point?
- So I think we touched a little bit on the real world data as real world evidence, and the concept of on-going surveillance over time, which I think might become the mantra of the entire industry is this ability to monitor populations. But if we can reach a tipping point in the genomics world, it's going to not only supercharge the ability to, let's say develop new drugs that are highly, highly targeted, but also make extremely targeted point of care decisions on which therapy is really gonna be appropriate for which patient. And then when you think about endogenous to exogenous data, imagine we have a place where we can link all of this data together from your fundamental biology, all the way out to your behaviors in the real world, as some of my colleagues elaborated on. That I think is the holy grail that we see as the potential for real world data that we're chasing.
- That's great. And then the last question Paul, maybe I'll just go in to you. We talked a little bit about predicting conditions, maybe you can tie it to what Eleni is talking about bringing together different data sets and using that to sharpen our ability to predict outcomes.
- Yeah, so three quick things. The first of all, data mastering, is a non-trivial problem. So data mastering is the concept of how do you link a whole bunch of external data sets together through some common set of identifiers. Best in class performance is around 82-82% of fully definitive matching. You miss some beyond that for lots of reasons, not the least of which is there's a lot of John Smith's in the world, a lot of people move, some people pass away etc. So you're never gonna get perfect, that's problem number one. However, when you can get that data linked you can view it in two different ways. You can use the clinical information as we talked about as labels that allow you then to model the consumer information, so in other words you say, here's my patient population with diabetes, and you point some really advanced AI and ML models at that labeled population to try to find the underlying predictive analytics for the actual consumer traits. And when I say underlying I mean not "you eat a lot of sugar" that's a relatively straight forward thing. But the much more hidden variables that are usually the only things we know about a consumer. So we see a lot of that. Then there's the other side, which is, can you use consumer labels, normally we talk about social determinates of health, socio-demographic labels etc. To then map to a whole set of clinical variables. In other words, you use the clinical variables to map back to consumer traits. And that one is where we're just starting to scratch the surface. So let me give you a very simple example and then we'll move on. Imagine if you could say "here is a patient that we know has disengaged from care for the last nine months and their symptoms have gotten worse all the way through it" and you could predict, based on the condition, the likelihood of a patient to disengage. And it turns out patient's view of health care systems change radically based on a acuity and severity of health care. So one of the most important factors in consumer engagement is the condition you have, because you're much more likely to engage the sicker you are, you are much less likely to engage the much longer term your condition is. And so you can do some really advanced things where just knowing a little bit about the consumer gives you a whole predictive model of how that consumer is gonna react in different pathways of care. And believe me, it's not consistent. You may not ever go to an ED when you're sick, you'll like "shlulff it off" but when your doctor tells you you could have cancer you're the one using the most health care in the country. And so it's not a simple A B comparison, it's a very differentiated view, micro-segmented, etc. I'm gonna wrap up with one last question that I'll open up to any body on the panel here. If you were to imagine the world a couple of years down the road what data set maybe is not being used in health care today that you think could fundamentally drive further innovation in terms of care, quality, outcomes? Open it up to any one on the panel.
- That's a tough one, gosh, what do we not use? We're trying it.
- I'll throw one out that's provocative just because we haven't mentioned it but, weather data. I mean in some of my prior exposures to the use of big data weather was being used to predict pandemics and the spread of disease and viruses, as well as try to identify which patients may not be able to access care. So that you could enable more mobile technologies to get out and reach them. And it's probably another good predictor of whether patients will engage if they are geographically located in an area of severe weather. Those things in combination may make them more or less likely to be able to get the care that they need.
- And I'll give you one; physician effectiveness. It turns out physicians can not always treat the same condition the same way with a different patient population. We're probably just at the tip of the iceberg of developing physician panels that are highly customized and tailored to an individual set of consumers that they treat. Some physicians wanna spend 25 minutes with a patient, really get to know 'em, engage, others are pretty comfortable with the five minute engagement. And we don't line consumers up with that at all, so we don't really have good longitudinal history on the different ways physicians and clinicians engage with patients and whether just simple engagement metrics make a massive impact in health care. So you can see that starting to come around, we're doing a little bit of work on patient panel alignment. But we're really at the tip of the iceberg on that one.
- Great. Well thank you, I'm just gonna hand it over to Curt for a few minutes just to do a little bit of a wrap up. We heard a lot about a lot of different data sets, how they're bing used today, and he's gonna bring it all together and talk about what it means from an Optum perspective.
- How did they do?
- Pretty good? So I'll be quick, 'cause I really wanna get to the conversation part of this but, talk a little bit about where we see things going and I think you heard snippets from the group. Everyone's at a different point in their data and analytics journey. Put together this very simple view of where even we were, as Optum, five years ago. You heard a lot of great stories earlier in the morning. Most of those stories weren't possible five years ago, even in our own organization. And we're helping a lot of you to try to get to a similar place. First, even within our own organizations data is all over the place, we don't have similar definitions for how we're approaching analytics. People are trying to all do different things, get different answers. It makes it very, very difficult to make it actionable. Getting together, centralizing, being able to come up with common approaches is one of the first steps. Second, once we have that in place, we can start to really act at scale. And again, I think you saw some of those examples earlier today, where data and analytics underlay the ability to then act at scale, right down to the individual patient. Really, really important, and again, some of you might be in this phase of your journey. Where we see things going, and this is where it might get a little uncomfortable, okay? 'Cause we're gonna have to think carefully about how we approach this together, is creating an ecosystem where we're actually sharing the data. Consented. All appropriately used, but sharing the data and having real time analytics across our organizations. It's not gonna be good enough, right, just to have our own acts together. But how do we actually act as a health care system? And share data, and drive analytics, common approaches, the ability to act in a coordinated fashion, across all of the different entities, all of the different groups really trying to do great work for the patients. So this is what we're pushing for, and part of it is "how can we help you on your journey based on where you are today?" And I think what you heard from the panel is we're moving in all different directions at the same time, trying to figure out how do we bring this together and create this future world? So it's very exciting, and again, we think we're at the very tip of the iceberg in what we can bring with you all to the health care system. Second, you heard different perspectives, right? And part of what we're trying to do is how do we actually bring this together? Each group actually brings a unique set of experiences, each group has invested differently, and so there's really an opportunity to learn. I know most groups will have been engaging in clinical trials, in some form of patient outreach, and in some sort of consumer engagement. But being able to learn from that broadly, across all of the different players that are tryna do this is something that we believe we bring to the table, and again, at scale. So it's a really exciting time, and again, we have to push our boundaries. How can we actually learn from the others that are trying to do the same thing?
Making good on the promise of analytics
As an industry we're on the cusp of being able to offer simple, smart and seamless health care experiences by harnessing advanced analytics and a wealth of data like social determinants of health. Learn how innovators are making a difference now and what you should know for the future.
Curt Medeiros, President, Optum Analytics
Elenee Argentinis, VP of Advanced Data and Analytics, Optum Life Sciences
James Dolstad, VP and Practice Leader, Optum Advisory Services
Jay Hazelrigs, VP and Practice Leader, Optum Advisory Services
Paul Higday, SVP of Strategy and Innovation, Optum
Tushar Mehrotra, SVP of Analytics, Optum Advisory Services