Big Data, AI, and the Journey to Value Driven Care Vivian S.Lee | AAAI-20

Thank you, Kenny and good morning everybody. And this is a very large room for this section. So to the extent that people specially way over there if you want to come on towards the center great, but I understand sometimes I like to sit out on the edge as well. So welcome to this event to the day Health Care is the biggest part of our economy. It's the biggest part of the economy of almost every country in the world about 7 trillion dollars. Beyond being really big is also deeply intimate for everyone of us and all of that

means that it is an area that has a lot of potential for growth for improvement and as it is beginning to change an impact every part of our society is beginning to have very important effects on health and Healthcare and today's discussion is really about that. How that's going to play out and how we can shape it. And we have people who are AI experts who are working on health and we have health experts were thinking about Ai and really a combination of both sets of folks. I will be speaking to you this

morning during the day and we're going to get kicked off with our first speaker. So I'm not going to go too much longer. Let me just very briefly introduce you to our first keynote Vivien Leigh Vivian is a physician. She is the president of Health platforms at verily she is a radiologist by training a member of the National Academy of Medicine has had a long and distinguished career already at her young age and among her other many accomplishments was leading the University of Utah health system really making it

one of the nationally prominent health systems than that was before she went off to verily I'm so what she is going to talk to us about is Big Data aai and the tourney to value-driven healthcare and so without further ado. Thank you for coming. Alright, thank you so much as she she thinks I have guinea for so kindly invited me to speak and thanks all of you for staying around to the end of this meeting. Hopefully today will be very worthwhile for you. I am going to as Ashish alluded to be one of those people who is coming from the

healthcare side and sharing with you some of my perspectives on where I hope the AI Community will help to drive us forward and really enable a true disruption and in healthcare in this country and hopefully for the benefit of others around the world the story that I want to tell you is really about a journey to value-driven care before I start there. Let me just get a sense of the audience. How many of you are actually working in a field that is related to healthcare Life Sciences. Just curious and then how many of you are hoping to get into a field that is related

to healthcare Life Sciences. Okay, so good terrific. Thank you. So I work for a company called vert Life Sciences, it was formerly known as Google life sciences. It is alphabets Healthcare & Life Sciences company and I joined it about 18 months ago. But my career prior to that is related to has been in academic medicine and the story that I wouldn't tell you this morning is really a story that parallels my own career in terms of thinking about the direction of where our Healthcare System needs to go. So this story starts in the summer of 2011 right here

in Manhattan at that time. I was the chief scientific officer at NYU langone Medical Center responsible for the research portfolio there and I was particularly interested in I became very interested in these new fields of population health or implementation science or help other areas of Health Services Research and how those areas could help to improve our Healthcare System. In the summer of 2011 I found myself taking a cab to JFK and boarding a plane for the beginning of the next

phase of this journey, which was a trip to Salt Lake City Utah, where as she said, I was given the privilege of serving as the CEO of the University of Utah health the dean of the medical school and also the Senior Vice-President of the Health Sciences Campus there which meant being responsible for colleges of Pharmacy nursing health and a new dental school and around that time as you may remember the Affordable Care Act was really just taking hold across the country and it was moving up through the the courts and its constitutionality

was being challenged as you might remember I still actually still is but at that time it was going to the Supreme Court and there was a lot of speculation about what would happen to Healthcare in this country. I don't remember the Supreme Court voted just barely to support and uphold the Affordable Care Act with a couple of key exceptions to that. But even through all the discussions within the Healthcare Community, everyone was saying the train is already left the station regardless of the outcome of the Supreme Court case. We in healthcare knew that Healthcare couldn't

continue the way it was that it was really time to move away from fee-for-service to more value-based care and some people said and rode the train is already left the station. We're already on our way. Well as I landed in and got settled in Salt Lake City leading this Healthcare System that was about a three and a half billion dollars system sort of the medium sized Healthcare System. I discovered that actually the train really hadn't left the station and for the most part the train was still in the station mostly across the country and much of the work that needed to be done to improve

our Healthcare System really still remained to be room. Need to be taken on an so that's what I wanted to start with was just a simple overview of where we are and Healthcare and what some of these initial changes are that we are hoping your community will really help us with so, where are we today? And then most of you are familiar with the rising healthcare costs? I don't know if you've seen the data presented in this way, but over the last 50 to 60 years healthcare costs have increased at a rate that is 50 fold faster than the rate of increase in wages and

another way of saying this is that the rising healthcare costs have impacted everybody in this country often in invisible ways, whether it's through taxes or which support for example Medicare Medicaid and also the employee subsidy Or through your your own out-of-pocket payments through Verizon deductibles co-payments a coinsurance or probably most insidiously through invisibly by really flattening wages through flatten a wage growth because employers are having to take money for wages as well as for

retirement benefits and use it to support employee health costs. So this this is impacting all of us. And as you may know that increasing cost is coming without a corresponding increases and how come so this is one way in which these kind of data are often shown you have life expectancy on the y-axis. You have health care cost per capita on the x-axis. Typically the more you spend in a country on Healthcare. The better your population is with the extreme exception of the US. If you want to read more about this in detail, actually, she's John his colleagues wrote an excellent paper in

Jama a couple of years ago, which I highly recommend And not only is it not seen and outcomes. It's also not seen in the quality of care in the way in which we deliver care. And this is a slide that shows you the defects per million against the percentage of defects kind of the Six Sigma scale and you'll see that while we're Amy for Six Sigma the only place in which were near to achieving that is in anesthesia related fatality. So if you do undergoing a seizure, you're very likely to survive that that's good. The bad news is possible acquired infections adverse drug events even

just the regular following of guidelines, like not giving antibiotics for common cold. We don't do very well on our defects are very very high. And all of this is coming at the expense of a really frustrating our clinicians and our physicians. It may surprise you to hear given that the quality of life you'd expect for Physicians to be reasonably high but study after study showing that about 50% of all positioned are clinically burnt out and that's in large part through in efficiencies in our system are electronic health records on administrative burdens among many other many other

factors. So after I'll just say it in terms of the context of what I'm going to share with you. I did have the privilege of having a year sabbatical after my 6 year tenure at the University of Utah. In order to research more about outstanding organizations across the country and world who are doing really Innovative things in healthcare that actually is culminating in a book that is coming out in May called along fix and one of the things that I was able to do was to ask over a hundred people in health care, whether pharmaceutical companies Hospital people non-for-profit leaders,

even even journalists if you could change one thing about Healthcare in the u.s. What would it be? Did you just change one thing and the vast majority of those said the payment model that the way in which were getting paid and Healthcare the way in which we are expecting our goods and services to be delivered in healthcare is so different from all other Industries in this country. We are paid fee-for-service the more we do to people that were more we get paid the more we value the the service or the good and it's really not based on the outcome or the impact on health.

So let's just talk a little bit about this payment model and then I will move to some examples that do involve some AI so don't worry the current payment model for the most part in this country and it's just for the most part. It's not completely is fee-for-service. I'm an MRI radiologist. So I embody the revenue Center of most Hospital Systems in a fee-for-service world, which means the more MRIs you do the more money the hospital makes that's the old way the fee-for-service model and what we're realizing is that in order to really change the direction of Healthcare in

order to really put the full Market Force behind improvements in health care. We really need to transform to value-based care and in value-based care. We often talk about the quadruple aim, AAA more quadruple aim from Don Berwick, and the ihi that includes better outcomes lower-cost a better clinician experience in a better patient experience and We can start valuing our goods and services and pricing are goods and services including pharmaceutical drugs based on their impact on health. Then we can really start to improve our overall health care system. So let me put this

in little more or less wonky policy terms and just more specifically in an example of a patient. So when I showed up in in Utah in 2011 S. I said we realize that the train was just beginning to leave the station and one of the changes that happened over the next year or two when I was there was an evolution that is taking place across the country in the way in which Medicaid is thinking about how they pay for healthcare. So let me tell you about a patient's that we had I call her Mary that's not her real name and this is not her real photo but she is a real person. So we had a patient and she

was a Medicaid patient in Utah and as the university health system, we took care of many of our Medicaid population of patients and she was a frequent flyer in our emergency room. The previous year in in the form when I'm talking about in 2012. Mary came to our emergency room over 50 times in that one year. She came in she had chronic diseases. She had diabetes. She had some as much had a little bit of depression and she came in more than 50 times because up until 2013 Medicaid was paying us in a fee-for-service model Every Time. Mary came into our emergency room. We got to Bill

Medicaid and we got paid we got paid for every x-ray every prescription every visit to the emergency room. So you can imagine when Mary showed up the 50th time that year we said, here's Mary we're going to make some more money. We wish you didn't come but we weren't very proactive about preventing her from having to come. Well that change pretty dramatically in January of 2013 when the state decided to completely change its payment model and this is a set is happening across the country in many different ways and they changed it to what's called a capitated payment model

and an a capitated payment model with the state decided to do was they would look at all of the patients on Medicaid who were routinely coming to the university they decided in advance. They did a risk adjustment to figure out just how sick they were in predicted roughly how much they thought that patient would need in care. So they estimated marry in the context of the conditions that she had and then they said guess what University of Utah next year for Mary you get $13,000 to keep her healthy and out of the hospital and you have to meet some quality metrics. You

have to keep her healthy. You can't just lock your doors to her, but you would only get a fixed amount of money. So imagine us as a university health system trying to subsidise our researchers trying to subsidise our students trying not to go bankrupt on the clinical side thinking about all of a sudden how we were going to care for people like Mary. So in this value-based care world what we call a value-based care world with this kind of a payment model all of a sudden we started looking very carefully at all of our Medicaid patients to try to anticipate who were the frequent Fliers and

who we really needed to intervene on early. For the first time we matched Mary with a primary care physician. She didn't have one. So we match their up with one and said really stop using our emergency room as your own Clinic. Let's get you into a PCP a primary care provider and we even assigned her a care manager to help remind her about the appointments to help deal with Transportation issues when she had trouble getting to the clinics are getting her prescriptions filled and we just really work very closely with Mariah to try to Keep Her On Track and to keep her healthy and out of the

system. I hope you appreciate what the system with the with this change. Why payment models are so critical for how we think about Healthcare know if you take a step back and think well, how is this relevant to potential any any worth? It might be done in the air Big Data Community. You can think about it at a more macro level witches. We really want to keep the population healthy are there public health prevention strategies that we can use are the ways in which we can work with Community cities and states to address social services are the ways and We can more effectively look at the

date of the we have and predict. Who do we know who we need to intervene on earlier who we need to provide more consistent for example transportation resources food resources and so on to keep them out of our clinics and hospitals. For patients who are are reasonably healthy mostly out of the hospital, but who had some chronic conditions like diabetes or asthma. How do we help them? Stay healthy and keep them out of the hospital's what tools can we develop to provide that support and then if the patient does have to come to the hospital where the most expensive care

does take place. How can we ensure better outcomes safer or consistent care? How do we get closer to Six Sigma care and lower the cost of care? So in that context, I'm going to talk about three examples from verily I just share those these examples because they're the ones I know best but of course across the industry of health and Technology there many many efforts that are undergoing and I should have been with you and as you consider that your own work, I will just emphasize what I say to our own team of Engineers, which is that we consistently evaluate the work based on not

fee-for-service model but on a value-based model, which is for all of our Innovations are we really driving quality and outcomes are we improving the experience of care for patients and Physicians? And are we reducing the cost of care? So I'm going to take you through three examples that I hope they fit into a little bit of this structure. I'm going to start with how do we keep people like Mary out of the hospital out of the clinic and healthy as much as possible and the example I'm going to give you is a joint venture between verily and sanofi a company called on Duo which was established

about two and a half years ago now to look at type 2 diabetes. So type 2 diabetes issues. No is one of the largest public health concerns in the in the world and only rising and prevalence thanks to a generous exportation of American habits. I have to say and I'll tell you a little bit about the story in this could be married. But in this case is another patient. This is a patient whose real name is Steve and he's been very open about his condition. He is a an individual who lives in Marietta, Georgia. He had it had he was pre-diabetic for many years and then one day

he came into the emergency room feeling nauseated vomiting. And his blood sugar was in the 400 range, which is way too high and he was diagnosed with no longer being pre-diabetic but with having type 2 diabetes. For about a year, he tried to manage his own blood sugars. He did the finger sticks test his blood sugar try to control his diet and really struggled and so at the end of about a year or his Blue Cross Blue Shield plan his health insurance plan recommended that he sign up for our initial pilot study with on Duo which is a virtual diabetes Clinic. And the idea

behind on Duo is really this concept of co-producing Hell the ideas. How do we use technology to help people like Steve really manage his own health, mostly outside of the walls of the clinics in the hospital. So it's a virtual coaching platform that has a connected sensor and provide some AI driven insights that enable the patient and caregivers as well as the clinicians to really manage their disease and most importantly Drive Behavior change. So the key components of it down here, there's a continuous glucose monitor.

So this is a device that you can put on your armor on your abdomen. There are multiple manufacturers there when I'm showing you happens to be one that we build for Dexcom and this technology enables you now to no longer have to stick your fingers, but it has a Bluetooth chip in it. And so it can transmit your blood sugars 24/7 cuz it's a continuous glue. monitor to the app With the app you have the ability to follow your blood sugars and make an association between your blood sugars in your diet as well as connect with a coach and the goal is to enable you

to keep your blood sugars under control as a diabetic in order to prevent subsequent complications. So some of the features of it. I think that maybe illustrate where were using some of the AI include the ability to automatically identify foods and meals and snacks. So typically for diabetics you're asked to keep a food log. I'll tell you my sister is an endocrinologist in Sanford and she will she tells me all the time that patience really never keep this food log. It's not surprising but with this app, you take pictures of your meals and snacks and we use some code from

Google to automatically recognize over a million different kinds of food. And that's a point of Engagement. I think with a many patients many people just find it fun to go to their favorite ethnic restaurants and take a picture of it. See if it recognizes their their pod tie or whatever. They're new food is but it doesn't able them to make this visual Association so that they can see with their blood sugar tracing which is the middle part of drawing their what each of those food items like this McMuffin or what would have whatever exactly this is and how it's affected their

breakfast blood sugars and with the albums that we have. We can actually drive some personalized insides for each individual patient because everybody reacts to meals and differently that may not surprise you if we all ate a banana in this room, we would have different blood sugar reactions. And so the algorithms that we are developing and refining all the time and Abel us to translate those into some insights for the patients and maybe even some recommendations. So you we notice every Wednesday you love to have that egg muffin thing. Maybe just try taking

the top half of the muffin off next time. I think that'll do a little bit better for your blood sugars. For example. I think the most meaningful part of this kind of Engagement is still the Personal Touch. So I will say that as much as I think AI driven virtualbox and so on have a lot of promise there is nothing that replaces the human touch in my view. And so we do have the ability to chat and video conference with coaches and with Physicians, although increasingly. Some of the interactions are virtue or are with a a bot as opposed to the real person, but we mix

those up and for the clinician I think very importantly again some of the inside that we're able to drive for each patient are shared with the coaches so that they can make some of these recommendations directly to the patients and the the results of our work and others on on the market. So again, I'm I'm not I'm not intending to To say that we're the only ones that can do this but this model of having sensors and a ai-driven algorithms that can drive inside is very very powerful and as a clinician and I will say that I really really welcome

the continued development of these kind of technology. So this is a charge from a recent publication showing that for patients who have elevated blood sugars who have not been able to maintain their blood sugars. That's the farthest a right on the slide. We see significant reductions in hba1c which indicates much better blood sugar control in very short periods of time with this kind of Technology. Patients with diabetes have a number of other complications and just another illustration of a I really very powerful application of AI is in screening for diabetic retinopathy the common

complication and early intervention can prevent blindness in type 2 diabetics and type 1 diabetics. And so this was work that was done by colleagues in Google Google Health Google brain and here looking at the retinal images. These are pictures photos taken from the back of the eye using again large bass datasets really deep learning algorithms biscuit deep learning algorithms to identify the abnormalities automatically. This is critical because we do not have enough ophthalmologist in the US much less the world to interpret all of these scans and the

results were very impressive. This is up from a publication in Jama from about a year-and-a-half ago. It's now been deployed in a number of countries include India one of the things that I will say about this is I feel like the potential for a i with the evaluation of Imaging weather is retinal images or my own field Radiology images like MRI mammography chest CT scans and so on is the potential is really Beyond just simply replacing or improving what a human can do and one example of that is in this work. It was also done at between is in a collaboration between Google

health and barely knew which was to look at these retinal images and and try to examine whether there are further insights that we can get from these images Beyond just looking at the health of the eye and so this paper was published in nature by Michael engineering that showed that the retinal image is not only predicted ihealth, but it actually pretty good cardiovascular risk factors cardiovascular mortality and morbidity and even hypertension and if it's not surprising for those of us who are in healthcare because of the back of the eye of you're just looking at the blood

vessels at the back of the eye, so it's a win. Go into your cardiovascular system, but it certainly does make me wonder at having been a radiologist and having read I don't know if thousands of chest x-rays what information is hiding in these films and what insights can we have to not only predict Health, but also to predict a decline and be able to intervene. So that's a challenge. I hope her from that many of you are thinking about add a second example that I want to share with you is in the broader space of clinical Discovery and this is a project called project

Baseline. It was the core project it barely when we first got started about 5 years ago. The question is in conditions like diabetes where I just showed you we know that blood sugar and then hba1c this other ass say they are very good markers for disease and Disease Control and risk of complications. But what about other conditions where we don't have such good biomarkers? So this project Baseline was started as the brainchild of Rob Calif who is a lifelong clinical trial is to was most recently the FDA commissioner and he now works for barely in Google and Anna Conrad our CEO

and about four or five years ago. They got together and said what if we collected every possible bit of data about a human being and Then followed them for say 10 years to see if we could identify other biomarkers other predictors of disease. And so that's what they did. It's literally terabytes of Health Data from thousands of volunteers lose everything from Whole Body sequencing all kinds of homex, including fecal microbiota new sensor data sleep data EHR data claims data having you name it. We are collecting every possible bit of data and the objective of it

is to to identify these new markers, but what it's enabled us to do along the way is to have the capability of integrating data just a wide array of very disparate kinds of data and to develop actually some tools and Technologies for just simply acquiring an organizing them as well as analyzing them and as we were going down this path and we are continuing with the study. This is in close partnership with Stanford and Duke. We realized that the ability to collect these different sources of data from individuals and analyzing.

I'm effectively was actually very valuable in the field of clinical trials and clinical Discovery. And one of the reasons why for example, Pharmaceuticals are so expensive is because of the very high cost of running these drugs through trials and going through FDA approval. So the clinical trials system is a broken system most trials do not recruit all the patients that they need to recruit. They're very expensive. They're very labor-intensive and they're very costly and also they they

don't the people who participate in these trials don't necessarily reflect the populations. So in this life, you'll see the white areas are where all these trials take place and the red areas are where we have the people who are probably in most need of of healthcare and medical advances and there's just not a lot of overlap between the two so by creating through project Baseline a large data registry where we Invite and welcome patients some who provide terabytes of data others who have a lighter participation to be a part of this registry. It enables us to connect organizations that

are running this clinical trials companies with patients who are from all over the country and in the future all over the world and we hope that this will enable faster and more Equitable Equitable drug Discovery and clinical research as well as identifying new biomarkers for us to to develop additional tools like on Duo a very quickly. I'll just finish with one last example. No, this is sort of just for fun. So this is now about machine learning and AI in public health and this is a project which I think has the best name ever. It's called

debug and you'll see why in a second. It's it was the idea of Of one of our leaders in in verily who has a passion for this and this is really trying to address the problem of mosquito borne illnesses. So this is now public health talking about in particular the mosquito that carries yellow fever and Dengue Chikungunya zika. And that is this mosquito called 80s Egypt eye and are this individual and barely who decide to take on this project really wanted to help eradicate

these diseases and focused on these mosquitoes not let me tell you a little bit about mosquitoes. You probably didn't to come to this session expecting to learn about mosquitoes, but I'm going to tell you a little bit by mosquitoes so 80s Egypt time. When they are infected with a bacteria called wolbachia, the sterile males essentially breed with the females and the eggs never hatched. Now. What vodka is a bacteria imagine. None of you probably heard about that wall back in a bit while back he is prevalent in insects and animals about in about 60% of

all animals and insects carry wolbachia. It's not humans don't and for whatever reason 80s Egypt I don't but it's very very common. And so this was our strategy and we didn't invent this I'm just going to say but this is the strategy we decided to adopt in order to try to tackle these these infectious diseases now the challenges that you we need to so what other fact is that the only female mosquitoes bite males dote so our goal was to release as many males with Mobic as possible. About releasing any females because we didn't

really want to increase the burden of of biting mosquitoes in any community. So the challenge here was to enable these mosquitoes to mate and then to separate them. To separate the males from the females. So here is our larval rearing robotic facility in south San Francisco. And here's the AIP switch is to differentiate emails from the females. They have different bodies because I'm a radiologist. I'm really good at spotting those differences, but I highlighted them for you for that was that's a joke. Okay. Anyway, so we

develop the hours of to do that and we've been piloting it in Fresno California for a couple of years. Now, we've gone through two seasons and here's an example of the results. So in red it's the it's the neighborhoods around where we released and then in blue, it's the areas where releases how people do ask me. Did you tell anybody you were doing us? Yes the folks in Fresno California. They know this was actually has been reviewed by many governmental agencies and we are in partnership with Folks in Australia and Singapore in another countries around the world.

So that's an example of where I think there's a lot of potential in ways maybe that are unexpected too. So just a wrap up. I think the potential for what the work that you do and impacting care especially in driving value is absolutely enormous. I know people like to say never waste a crisis and Healthcare. We're in a crisis now, and so there's an enormous opportunity for Change and in that change. I really hope it'll be an AI driven change. So thank you very much for your attention. Happy to take questions right happy to take question.

Tell 5 to 10 minutes of questions. He thinks for the great talk just thinking back to that sort of embarrassing scatter plot of sort of specs Healthcare spending versus Healthcare quality in different countries. So I guess that the start of hypothesis is the main thing that differentiates the US from his other countries is the fee-for-service payment model, but I'm wondering it seems like sort of value-based care might also be a pretty new payment model as far as sort of different countries go and I'm just wondering isn't it sort of like an

uncontrolled experiment? Like why not do something closer to what other countries do in terms of payment? Yeah. It's not the only thing that's difference of course of their of the biggest differences are we don't have Universal coverage. We have private insurance model. It's difficult to say, you know, we would be wonderful if we could wave a magic wand and we could all of a sudden say we're going to just completely blow up our Healthcare System and become the some combination of northern European Canadian New Zealand, you know,

that would be wonderful. I think what we're really talking about now is given where we are now with 18% of the US economy firmly embedded in the current model what changes can we really push for that are likely to be effective enough to drive the improvements that we need to see and that's where changing the payment model is. Probably the most effective lever that we have. Okay. Thank you for a really I'm over here. I feel really interesting talk to question. So I'll try to go fast with respect

to the value based model. That's really compelling at the hospital organizational level. But I'm curious if you drive it down to the level of the pcps themselves. How do you what's the incentive for a PCP to take on a really complex patient whose outcomes are likely to be really bad and it's kind of related to that you you talked about patience like that really needing a community-wide healthcare and yet the tool you showed for diabetes with her an individual. So are you at verily

working on to that actually take into account that patients with complex conditions actually have many many caregivers in positions. That that's a really fantastic question and a fact one of the concerns that people have and we all share is that when we move to the kinds of payment models that were talking about that we might put more complicated patients at risk and people might not want to care for them. So we're that plays out is really in this process of risk adjustment, which is frankly if I were thinking about one area where the AI Big Data Community could really help us

isn't thinking about risk adjustment. So that's really looking at a patient and figuring out for example in Mary's case. Maybe we should have gotten paid $18,000 for her and then somebody a little healthier maybe 6200 there really healthy there is risk adjustment to risk that kind of adjustment right now in government models as well as some private insurer models. How effective is it how good they are think I could always be better because the date of often for example. Gaston medical claims. So just how much did

you use the healthcare system before they don't necessarily have that much information about your access and you know your access to food to exercise to you know, whether you're actually receiving good health care in an environment that supports that so so I think maybe that's maybe the answer to how we avoid that and then how are pcbs thinking about it will a 1 model that's actually moving around the country. Is it a Medicare Advantage model? So for patients over 65 the positions that are caring for them increasingly. There is a version of

Medicare Advantage where Medicare so the government is paying physician groups a fixed amount. That's a risk adjusted for caring for those patients. And actually the payment so far is quite generous even more about it. Just how may be overly generous? Sometimes they are but those are really intended to enable PCP. To take the time to really invest in complicated elderly patients to try to keep them healthy and outside of the hospital. So so there is hope tools for 10 people. The tools were like if you're working on that, so I think there are

the tools that I tried to show you were really more about how we really Empower patients. I think there are tools that are very important for the clinicians for the community for the people who are paying and we are looking at all of those. Thai II photo talk I see the Uncle for a little bit more about how your team feel this app. How does sensor and a I come to this together and how does clinical researchers recognize and realize this one is a working not only for patients and that is also help for the prevention work like the high so people can you can use this one track of

what they eat how many sugar leaves to use for prevention? Thank you. Okay, so so I think I started with a disclaimer that I'm a physician and I work very closely with Rai with our Engineers, but I really probably can't tell you specifically I probably wouldn't even be allowed if I could how those algorithms work. So but just in terms of the general way in which we use them examples that I tried to show included the algorithms to recognize the meals and snacks that's completely in a i driven algorithm ways in which we identify at

patterns between a person's meals or the other ways in which they're eating and exercising and how their blood sugars are changing. We actually are using an AI in order to find those associations and also using it. Well actually using it Britney wherever we have enough data in order to derive some more insights. Okay close to ring of time into adopting on Duo for type 1 diabetes. There's a lot of different disciplines here today, and I was wondering what are some of the other things that Healthcare leaders mentioned besides

the changing value model. Was wondering about that look like I was wondering what identification and I think you know the picture of something you might recognize the food, but you won't know how much sugar is in an type 1 diabetes. It's when we started we really needed a larger populations and we started with type 2, but there's absolutely no reason why this can't work for type 1 it's just not that we haven't gone for the update process yet the meals and snacks. You're right. It doesn't tell you how much

the person consumed. It's really a recognition of the food. I think it's more in my mind. It's more of a point of Engagement for the patient even then actually being used to guide their diets, but we can predict for example the carb the effect of the carb load of the meal if you ate the whole thing. This is why Carb load would be and millions of people using this app for other purposes. So they're not just using it in on Duo. So there's pretty why recognition for that and for frog for millions of different meals, you know, it's not just from our patients who are using it and then what did other

people say other than payments GoPayment model was about 2/3 of the answers and other answers include really needed to focus a lot on social determinants of health because we realize now as we try to keep people out of the hospital that really it's a lot more about what you eat or their you smoke whether you exercise what you inhale or drink that affects your overall health than it is what happens once you get into the healthcare system. So people are worried a lot about that. I really want to invest a lot more in Social Services in this country than in healthcare

cuz spending on Healthcare is too late people are also very worried about Equity or widening gap between the well and healthy. I'll tell you when I get Is talk in front of my colleagues at verily meaning we have great health benefits, but we don't recognize that about 10% of Americans still have no insurance and a rising percentage of people can't afford their co-pays and are underinsured so that that widen Gap. I think those are to do a couple of other areas of thanks. Thank you very much.