Transcript

Basically predictive analytics is using all the inputs. What we call data points that we have gathered passively or actively over the course of time to clear the noise and guide us, our decisions and actions. Now, the issue has happened that especially over the last 10 to 20 years, we have seen an exponential increase in the data that's pushing, I would say, the boundaries of our ability to process that. Thankfully here works and I would say computational sciences, big data. AI computers that is helping us go over the intuitions that would have taken decades and especially with the EMR that's accumulating, I think so that was a big step over the last 10 to 15 years that has allowed us to gather a lot of that information. However, at the same time, I think so computer sciences, data analytics, and a lot of our partners, be it academics or amazing industry vendors, are coming us and showing us the way in how we can do that. And I think so that is the promise that holds and the predictive analytics. We are at a stage one of this. Think so this is the Manhattan Project in healthcare where we can bring in together not only all the experts, we have one of the most immense data resources and the investments to really make sure that we are transforming healthcare. And that is the most promising thing. so we are managing our patient, through full coordination use Methodist coordinated care. Our our goal is to coordinate that care and not drop the ball and not have them come back. So, we've got an inpatient team that is working specifically with all the patients in our ACO. They're doing the discharge planning and they're thinking long term. It's not just stopping the 30 day readmission. I'm responsible for a calendar year, total cost of care, every outcome that happens to that patient. So I'm setting up 30, 60, 90 days of keeping you healthy at home. But the key to that is setting up this discharge deployment and making sure you're seeing somebody within the appropriate time period. Like you said, you know, the top, 20% top quintile of risk, I want to get them in within seven days. I know you mentioned 14, but for that highest risk of mortality or for readmission, I'm getting those in in seven days. The next quintile, those are the 14 days. And then anything below that, that's the ones that I'm going to get 14 to 30 days. there are several thousand million data points in electronic medical record. And because of that, it becomes so complex that, yes, we're entering in a lot of data in that medical record. But then how do you take that information so that as a clinician, whether I'm a nurse, a physician, social worker, how do I take all that information out and how do I understand it? And what has been really interesting is really to take and look at multiple encounters in a medical record. And so as I'm looking at a medical record, not only am I seeing a point in time, but I'm also able to with the work that we're doing is look at multiple encounters on the outpatient side, on the inpatient side, and through that whole transitions of care. Now we have a better look at the complete patient's medical record versus just a point in time. Hello and welcome to Innovation Spotlight Houston Methodist leading medicine podcast. I'm your host, Roberta Levy Schwartz executive vice president and chief innovation officer of Houston Methodist. Innovation is the key to better care and better outcomes, which is why we're fundamentally organized to quickly embrace and adopt new approaches and eager to talk about their impact on our organization and their potential impact on care delivery. More broadly. Join us as our clinical and administrative leaders examine our successful innovations as well as their outstanding problems. We're still trying to solve. So today's episode and today's spotlight is on predictive analytics and the work that we're doing on predictive analytics. I'm going to introduce our guests to you today as they start speaking. So I'm going to start with Doctor Karam Nassar, who is one of our incredible, physicians, clinicians and researchers here at Houston Methodist. And I feel very blessed that he joined us a number of years ago. So he's going to speak to us and start us off today on what is predictive analytics. What promise does it holds? Kind of where is it today and where do we hope it will be in the future? That's, a pretty loaded question to start with. And I'll just try to simplify for most of us to understand, basically predictive analytics is using all the inputs. What we call data points that we have gathered passively or actively over the course of time to clear the noise and guide us, our decisions and actions. Now, as within humanity, we have been doing that for thousands of years. As physicians, we have been doing for hundreds of years. We train and educate ourselves for decades. Which we called experience. And some of us are able to synthesize and process them better than others. And we think those are the experts, innovators and leaders. Many of us, here today. Now, the issue has happened that especially over the last 10 to 20 years, we have seen an exponential increase in the data that's pushing, I would say, the boundaries of our ability to process that. Thank thankfully here works and I would say computational sciences, big data. I computers that is helping us go over the intuitions that would have taken decades for years, and especially with the EMR that's accumulating, I think so that was a big step over the last 10 to 15 years that has allowed us to gather a lot of that information. Now, you know, other industries are doing it very seamlessly. And just this morning coming and always told me which streets to avoid. I think so. The geospatial data few weeks back gave us prompted us much earlier. What action should we take as a system? And across other industries, you're seeing like proteins are being discovered. That would have taken all the labs in the country working for decades. However, what we're seeing in in spite of the fact that we have EMR, most of us at the front end have have not yet seen what I would say to impact in how we practice medicine or even conduct research. And the reason is because most of the data so far has been siloed, fragmented, underutilized, and we just don't have the playbook in making the best use. However, at the same time, I think so computer sciences, data analytics, and a lot of our partners, be it academics or amazing industry vendors, are coming us and showing us the way in how we can do that. And I think so that is the promise that holds and the predictive analytics. We are at a stage one of this. We are lagging behind. But for us, I think so this is the Manhattan Project in healthcare where we can bring in together not only all the experts, we have one of the most immense data resources and the investments to really make sure that we are transforming healthcare. And that is the most promising thing. And in this regard, I can tell you that a lot of we're seeing a lot of activities within our system over the last few years. And that is the most exciting point. So I love to discuss more and some of those going into detail, but just being very excited about the opportunities that are present to all of us today. My predictive analytics. So Brenda Campbell, moving on, who was one of our executives here at Houston Methodist over our, DeBakey heart and vascular Center as she sidestepped and, decided that she was going to, slow down a little bit, she made a decision to keep predictive analytics as a passion project, and continue working with us around this. So, Brenda, as I turn to you, you know, the electronic medical record as Doctor Nassar, as Graham has said, it holds incredible promise. There are bazillion, data points within the electronic medical record. Yet despite the fact that there are a million different data points in that medical record, it's not enough. Like speak about the predictive analytics that we have kind of in that electronic medical record. And why at Houston Methodist, we needed to go well beyond that to pull in even more data to really do these predictive analytics on our current patients. Well, thank you, Roberta, and as you said, there are several thousand million data points in electronic medical record. And because of that, it becomes so complex that, yes, we're entering in a lot of data in that medical record. But then how do you take that information so that as a clinician, whether I'm a nurse, a physician, social worker, how do I take all that information out and how do I understand it? And what has been really interesting is really to take and look at multiple encounters in a medical record. And so as I'm looking at a medical record, not only am I seeing a point in time, but I'm also able to with the work that we're doing is look at multiple encounters on the outpatient side, on the inpatient side, and through that whole transitions of care. Now we have a better look at the complete patient's medical record versus just a point in time. And so if I'm a clinician and I'm trying to find a piece or a piece of information in the medical record, you know, as Doctor Nasir has said, and also as Doctor Menna said, us as clinicians, our time is very, very valuable. And how do we go and look in a medical record and find maybe that one important piece of information that could help us in planning for the clinicians? Plan of care. So now with the predictive analytics and integrating it through the electronic medical record, we're actually creating not just predictive risk models, but we're also working on chart summaries. How do you pull this information, get it easily into the clinicians hands so that they can make the decisions. We're not here to make the decision for the clinician to tell the clinician what to do. We're here to give them that information easily accessible in a way that it's easily to be understood. And then they can make those decisions. And what I think is really important is also help the patient and the family be a big part of that care. With more information available. So just building on this, one of the systems that we kind of chose to bolt on that we chose to say an EMR is not enough is a system that actually held, data for, 20 to 30 years of Medicare data and all payer data. Why is that important? Well, I think it's important for multiple reasons. Yes, about 20, 30, maybe 40% of a hospital's, you know, patients are under the Medicare payer. But when we engaged with an outside predictive analytics company, they had information to all Medicare payers, all patients across the entire country. So now what we're able to do is not just look internally and say, yes, are we doing well or are we not doing well? It gives us the ability to start to look externally. And what also we're able to do is to do some what we call twinning, comparing our patients exact patients, not just saying, well, you know, using a case mix index or using kind of similar type of patient trait, we can actually go out and working with this company, twin our patients and find that exact patient sex, gender, age, socioeconomics, comorbidities and say how are we actually doing compared to the rest of the country? And then it allows us to say, what are some pro quality improvement projects that we can also put into place? Okay, I'm going to come back to that. As we talk some, some exact examples, but I want to introduce Doctor Zuckerman, who actually is one of the leaders in our ACO. So Zach, you also then took this kind of predictive analytic approach, which many acres are doing right, to see kind of what's happening with your patients. But you've you've successfully in an incredible ACO, had great results with you the using the use of predictive analytics. Thank you. Yes. So the I guess the issue that you have in an ACO, in any kind of population health program, you've got a ton of patients and not enough staff to take care of them. So in order to have the results that you need, you're going to work with the highest risk of that population and put as many resources as possible to those that are going to have a negative event, somebody that is going to have a hospitalization, you want to work with them to try to prevent that hospitalization. So that's where predictive analytics is key. Again, to your point about why is in the air our electronic health record not enough? Your patients don't always go to your hospital, to your physicians. You need to know the full story. And so that's why we need claims data that tells you a full picture, kind of a longitudinal story of every, every place that patient has been. What has happened, every hospitalization, emergency room visit, physicians that they're seeing that we may not know about in our year. We've got to collect all of that data and paint a true picture of what's going on with that patient, and it gives you a more robust prediction of what's going to happen. So that's why we've done very well with this outside predictive analytics company, truly looking at everything that's going on with the patient and putting our resources towards the highest risk of our patients. You know, I think this is now getting into kind of we had this big data, you know, use of more than just what we own, right? I mean, we love the fact that the EMR tells us what's happening when they're in the hospital or in the clinic. And if we're lucky, the physician goes back and looks through, you know, the last three hospitalizations and six clinic visits and whatever it is. But that's really hard. I mean, in in one day a patient can have, you know, basically hundreds of notes. And it's exhausting. And I have to look through those, let alone be able to say based on that, what do we think? I mean, so many of our physicians are in the moment, right in the moment and actually thinking at what will happen. Right? There are two ways they're trying to think what's going to happen clinically. If I do something right, what's going to happen after they get home based on the clinical situation that they're in? So going to Brenda first and then of back to, to Crum. You know, we we looked at this, right? We bring in this system that now has, information that we can twin that has data out 30 days, 60 days a year out and says this is based on twinning your population. This is what we think will happen to that person in these 16 different areas. Like, are they likely to come back? Are they likely to die? Are they likely to have a stroke? Are they likely to have these kind of things? You looked across and introduce this concept to all the different service lines. And I there was almost this feeling of, I'm overwhelmed. I don't know what to do with that. That can't be right. Like speak about the initial experience because people come to me and go, how did you introduce this to your organization? Good point. And so we introduced it to the organization, really starting out in our cardiovascular services. And one of the things that we were asking ourselves is why are we not staying? Maybe with our competition in mortality. And I think traditionally hospitals in in most administrators, except for maybe doctor men who works on the, on the ACO side, will look traditionally at what's happening with inside of my hospital walls. And so when we engaged with, predictive analytics, we started, as you said, to say, okay, these are some risk factors, what really is happening once our patients discharged and and what we really found is we were phenomenal on inpatient mortality. I mean, absolutely phenomenal. And that's one of the things that we get constantly I hear that always today. Well we're fine. It it's like okay. But now let's take a look at what happens after your patients leave the facility. And where are they, especially in that first 30 days. And I think health care traditionally has always focused on a 30 day readmission. I talked to very, very few clinicians and even hospitals across the country. Oh, yes. Where we know and we're working on 30 day readmission. My next question is is where are you on 30 day mortality? And that usually is where you get the silence. And so by bringing in the predictive analytics, we're now starting to say, okay, let's manage our patients inpatient. That's great. But let's also manage them Post-discharge. And I love it because many people say, well, discharge planning starts on admission. Well, yes. But what factors are you looking at and how can you continue to look at your patient. So that discharge planning also extends beyond? And where's the right thing to start working with patients and family to prevent readmit issues, but also to prevent mortality where we can now, one of the areas that we're also looking at, and we always people refer to is when do you bring palliative. And again, there are certain risk factors. The medical record is constantly, you know, updating these risks. But we're not telling clinicians what to do, but we're giving them more information to make the best discharge plan, to make the best inpatient plan, but also to make the best discharge plan. And learning from our ACO partners in how they actually look at the whole transitions of care. And now bringing in the predictive analytics and the modeling and the risk. We're actually saying we can't nor should we do everything for every patient. But when we look at these risks and we start to do risk profiling and we start to do risk stratification, how can we bring the best resources, allocate some very, very crucial resources that can be quite expensive? And how can we allocate those to the right patient population so that we can make the biggest impact for quality and patient care, right. So okay so Brenda comes and basically comes to cardiology and cardiovascular surgery and marches in and basically says hey your 30 day results not so good. Right. And and so Doctor Zogby, our, our chair of cardiology and Doctor Lumsden, our cardiovascular surgery turned to Doctor Nassar and basically say, hey, Brenda has this new system with HDI. We've bolted it on. It's doing this predictive analytics here. Go fix the program. Right. And and you are the recipient of this. So what's it like to be on the inside to convince your colleagues that now before they hospitalized, before they do surgery, before they do anything, they need to check yet another system to decide on not just the right clinical button that they've been trained in, but maybe what's the right outcome, because this patient may have a different untoward pathway that they're going to take. I would say it has been a wonderful journey over the last two years and, you know, learned a lot that I never expected. First of all, when it started, I had pretty healthy skepticism that it works. This is how most of our physicians look. You have to show us. What does it mean? One of one of the key elements, of course, as custodian of excellence, I think. So the spirit was, look, we can do better. We definitely know we are doing, as Brenda pointed out, amazingly in house. But why is our 30 day mortality not comparable? Reflecting that the reason is, which you just pointed out, the limitations of the MRI is fantastic, but it does not provide us the whole journey and especially within that time frame, that we want it. So I think so I was truly amazed and humbled by working with our colleagues and understanding that one of the largest data assets that have accumulated almost of millions and millions of all the Medicare patients. Now, what it really told us it gives gave us the journey, what was happening after they were discharged, the readmissions, their mortality. But the most important skepticism was we are a tertiary medical center. We see the highest of the highest risk, I think. So the one thing that really helped us understand that, how would how do we compare to other systems who deal with similar patients? I think so that digital twinning piece was critical for us to understand getting on board. The third piece of the skepticism that was all around our physicians is, yeah, these models have been created with Medicare data, but our patient population is different, I think. So getting over that hump, the best part was really developing and validating that, again, within our own system that clearly showed now and again once predicted, the question is how well it is for us. The best part was we see thousands of patients. We cannot we have limited resources and attention among those who are the ones that who we should pay more attention. And it was the degree of, I would say, ability to distinguish the noise. For example, it helped us identify if you focus on 20% of the patient, 200 of the thousand, actually, you will be able to capture 80% of the mortality. I think. So from a population health perspective, that is one of the best differentiator. Whereas, yes, you may not have to pay attention to the 40% of those were only about 2 to 3% of mortalities happening, I think. So that was a key eye opener where it helped get a lot of buy in from our journalist, skeptic, physician community. However, this was I thought this was hard. This was the easy part, the hardest part. Now how do we implement it? And it's still an ongoing process, but we had a lot of early successes. And in that process we felt and we found out that actually simplifying the information and keeping it to the basics would rather work, rather than giving them 20 predictive analytics, can we come up with a simple mechanism? With the click of a button, you can access the score, and as soon as you access score, it gives you a green, orange, red. My conversations with my colleagues and fellows was check, recognize, discuss. And then of course, if you want to have them see in the next 7 to 14 days, who's responsible, it is going to be of course, a lot of that time is being set to the discharge times. But the challenge with learning that patients do not want to be seen in the next seven days unless and until we, the physicians, have emphasized that. So in essence, that is the more challenging part that's happening. But clearly highlighting that fact that not only you need your own EMR data. And by the way, this scoring system has been further refined by 11 to 15% by using our own EMR data. So a combination of that Woodward. But it's not just the data is we have to get a sense what are the operational needs, what are the barriers, what do our physicians want. Because in the end, I can bring in the best predictive model. If I cannot get the trust of either the physician or the operational leaderships, it's just another tool will have no impact on what we are trying to do, which is improving outcomes, operational efficiency in a very cost effective manner. I think all three of you have mentioned it, but let me just, you know, dig into this for the time being. You know what I've seen move here in in many ways with using this level of data. Now is a conversation that's moved from every patient needs a seven day follow up, every patient needs a 14 day follow up. Every patient must come back to us in person. Every patient you know can go to whatever nursing home they want to to a slightly stronger message on. You're fine. You know, you're good at 30 days. We don't need to concentrate on you. I don't need to hire 15 coordinators to get every single one of you in and chase you down to the end level. You need to come in, but you're fine. And you? I can just do a video visit on like and it won't matter like that that focus on who needs and what type of resources. It's like. We talk about personalized ation of medicine, right? Medications. We talk about a personalization of treatment. But this is a personalization of like visit length visit type. You know, where I'm putting coordinators, where I'm kind of enforcing conversations or in some cases where I shouldn't be doing a procedure that I was thinking about doing. So so speak about those resources, because I know that that kind of all three of you have touched on that. But I think it's funny, is that when you speak to one of the easiest ways and I kind of kind of smile about this, but one of the easiest ways to impact 30 day readmission rates is follow up appointments. Now, most providers will say, as you just alluded to, I need more resources to do that. When we look at our own data and when we look at national data using the Medicare data bases, 14 days for a high risk patient and a high risk patient, we have to find in our organization high risk for 30 day mortality and or 30 day readmission. If you do a 14 day or less follow up rate, you significantly impact both readmission and or mortality. If you wait and extend that out towards more of the 30 day, then you don't see the that the again, all of that. And so this 14 days even nationally has been shown to be a valuable initiative. Sounds easy. It isn't. You have pushback from, you know, clinicians offices. I can't schedule that many patients. I don't have that many appointments. You know, the patients won't want to come in. Actually, what we're doing is we're seeing the lower risk patients faster post discharge, and we're seeing the sicker patients usually when they're coming back to the emergency room. And we have to flip this paradigm. And we have to say, just as you said, using predictive analytics, which patient could maybe extend that beyond that magical 14 days, free up some schedules. And then how do we make the biggest impact to that? But not not easy, but we're working on it. And as we start to show what happens when you do see these patients in 14 days and they're starting to see their readmission rates going down, then they become true believers in the initiative. Zach, you're doing this in the ACA, right? You're probably you were probably the primary use case that like latched on to this within a nanosecond had had we had no problem selling it to you. Oh absolutely. I mean why we are so we are managing our patient, through full coordination use Methodist coordinated care. Our our goal is to coordinate that care and not drop the ball and not have them come back. So, we've got an inpatient team that is working specifically with all the patients in our ACO. They're doing the discharge planning and they're thinking long term. It's not just stopping the 30 day readmission. I'm responsible for a calendar year, total cost of care, every outcome that happens to that patient. So I'm setting up 30, 60, 90 days of keeping you healthy at home. But the key to that is setting up this discharge deployment and making sure you're seeing somebody within the appropriate time period. Like you said, you know, the top, 20% top quintile of risk, I want to get them in within seven days. I know you mentioned 14, but for that highest risk of mortality or for readmission, I'm getting those in in seven days. The next quintile, those are the 14 days. And then anything below that, that's the ones that I'm going to get 14 to 30 days. We also have our nursing programs. Our nursing, team is calling those patients within 24 to 48 hours of discharge. They're working with the high risk patients, making sure they have transportation to the follow up visits, making sure they picked up their medication, making sure that if home health was ordered on discharge, did it show up? Did DME show up? I want to make sure everything is done there. They know what medications they should be taking, so they're not coming back for something that is an easy fix. Then they call the next time after the first follow up visit and say, hey, what did the doctor say? Right? I want to make sure you understood everything that was, you know, told to you at that visit so that you're not going to, come again back to the emergency room because there was something that you didn't understand in that explanation of what is going to keep you, again, healthy at home for the next 30, 60, 90 days. But we are we are quick to grab on to something, change. We move fast, and we want to have the best outcomes for these patients. That's just kind of the nimbleness of a smaller ACO managing a larger population of patients as opposed to the, you know, hundreds and hundreds of thousands of patients that you see at a hospital level. So talk about the difference between, I mean, our EMR, right? Epic has has a good module for the ACA, right? Healthy planet. I mean, it's like everyone can say that's that's fantastic. We layered on, we layered on this predictive analytics kind of HDR program as we did it. And we layer that on. How did that change your world? Huge night and day difference. So, the issue with so analytics versus predictive analytics analytics is just looking at the past. How did you do right. What are your outcomes. Predictive analytics is saying how are you going to do in the future based on all the information from the past? Well, the original modules that you have of any kind of risk tool is only as good as the information that's being fed into it. And so in our EHR, it was looking at only a few things. It's looking at the problem list, which may not always be updated. Right. And so it's only as good as an accurate problem is there. It's also the original state of it was only pulling from defined fields in the air. So if it was something that was free texted, it's not captured in that risk score. The third thing was it was only looking retrospectively. Now when we got this new program and we're looking at the history of claims and everything that has happened to this patient before, they showed up to this visit. And actually before this, group, we were working with another team of predictive analytics that was looking at, claims data and a little bit of lab data that was fed into it. So it was pretty good. The issue was it had nothing to do with this hospitalization. And so until this group came in, and I love your example about real time traffic, you know, the ways app it tells you. I'm going to look at how the traffic has always been along this route. But what they also have is there's an accident that has currently happened and so now we have the combination of claims data from the past, the EHR data of what is currently happening in this hospitalization to feed a better predictive, analysis, a better prediction of what's going to happen with this patient. And now we have a good idea of what's the mortality actually going to be based on everyone's Medicare claims from the past, this person's Medicare clients from the past. And now what has happened in the past seven days in this hospitalization, that's been the biggest step forward for us. And really refining a prediction of what's going to happen with these patients. And that's when we know, okay, that someone I truly need to focus on and our nurses are using their nursing judgment and are, you know, getting into the clinicians offices based on that, we're identifying, oh, we are really getting the right patients this time. In the past, it was I don't know why this person's high risk or why wasn't this person high risk. They're, you know, they're a mess. I really have a lot to do with them. They should have been brought to me. I feel like now we're getting to the point that when we say they're high risk, I'm getting agreement from nursing. I'm getting agreement from the administration and analytics agreement from the physicians. Everyone understands this is truly a high risk patient. And I've seen it based on what has happened from bringing them. Yeah, I have seen, you know, kind of going around. I've seen us move from a place people often ask me, how do you get this done? And it really has been a journey from skepticism to belief to utilization, right? As I look at this. But, you know, now when I look at a physician and basically say maybe that nursing home that had a 30% readmission rate for that particular heart failure population, we got two choices. We can reeducate them, right, send over a whole team and basically talk about what they're doing wrong or not doing wrong. Or we can say this patient population doesn't belong at this particular nursing home, right? It's we were able to get it down to that level of looking at and in a very easy, easy format that was, that doctors could really understand in a way that I feel like I never was able to convince them in the past. But but this was like, here, look at this. It's very obvious. It's versus national norms versus local norms. It's versus, you know, everything. So I talk about that skepticism to optimism, to utilization, the kind of movement that our physicians took. And, you know, other people say, how do I do this? Like, how do I move along that journey? Like, what advice kind of how did we do it? And what would you tell them? Yeah, you know, I would say the main challenge and then and again, of course, there is so much promise. And the issue of acceptance and adoption is as a physician community and even the rest of the healthcare workforce, we have never been trained in this mindset. Right. So and then again, so as a result, we have not we didn't have the visibility of what could be possible. So I just need to embed myself in medical school. So we are tied to I would say in most of I would say the operational and the general physician leaders have little visibility on the computational sciences and the predictive analytics. It's basically challenging, at least as a physician, not to rely purely on my intuition. And that can always be a big challenge in saying, I know my patients are at risk. And I'm saying, well, actually you're 70% right and you're missing 25% and 25% that you thought were high risk actually are anomalies. So again, that's that has been a major shift that is truly needed. But clearly what these type of tools is helping us, I would say separate the chef from the wheat and give us a clear indication from a patient's perspective. If my mom's hospitalized, I really want to know which nursing hospital will have the best outcomes. And I think so. It also meets our requirement from a patient's perspective. So clearly, as far as I would say from an ACO where this is the mission vision and I think so extensive appropriate resources allocated, this is a great tool. The challenge on the our our site, on the fee for service is can we find the appropriate resources to build the infrastructure. Are there true incentives to move in that direction? And apart from the incentive of doing the right thing, what else can be done to move? Adding to that is still, an evolving process that would need more input. I would say on my side of the fence and prevention, we see a huge value proposition on so many things, and the EMR is truly helping us because now we have an ability not to look at single labs and medications and outcomes, but a very multimodal approach, putting all of them together. For example, just the simplest intervention in prevention, like start and utilization for patients with cardiovascular disease, extremely suboptimal. But now with these analytics, we have an ability to do a deep dive to not look at the patient physician system and societal levels. I'm actually that's guiding us where the resources thing. And we got a big NIH grant just to study this and implement it. One of the biggest challenges that's happening is in the diabetes world, how can we prevent people moving from prediabetes to diabetes? It's almost kind of the same analogy. We have too many admissions. Who are the ones who are going to die, are readmitted. We have therapies which are expensive. We have limited attention and time. And where do we allocated, prior risk scores? Who are absolutely inaccurate? The reason was we weren't looking at the right thing because now taking social and environmental factor, taking your geocode spatial and plugging it with the data actually is telling us that 15 of the top 30 predictors were based on that, and it significantly improved our prediction. We are presenting that at American College of Cardiology. Now as part of that, we are building a whole Texas consortium where the models that we are going to be developing will be tested all across the Texas hospitals to see if we can identify those in the diabetes early. And so for physicians, I'm saying this is the most exciting time. Get on to the bandwagon. Most of the institutions, especially here, are having projects that in one way or other, like the ACL, the, the Medicare readmissions, the prevention, the imaging that has some of predictive analytics, digital solutions. I going on get yourself a little comfortable with the basics of data pipelines. What is an AI modeling? You don't need to be a data scientist, but getting some insights will be helpful. The biggest missing piece today is we don't have domain expertise guiding individuals who are creating the tools to see whether they are applicable or not, and partner with data scientists and engineers. As part of a new initiative, the Digital Health Institute. We're partnering with rice for these specific things. The key thing is this is going to be part and parcel of our care, if not in 5 in 10 years. Let's get used to it. It's challenging what we used to think and how we used to think and only physicians, operation leaders who would get fully embedded into this would be the ones leading. So it's an amazing opportunity. And I would actually ask, if you have a problem and you think there is a digital solution, come to us. There are so many of resources that we want. We will provide help, but more importantly, we want to hear from you. What are the challenges that we have not anticipated? So couldn't be a more exciting time for physicians to be here. So okay, Zach, he threw down the gantlet, right? He threw it down and basically said, okay, now he ate this data. This data is great. It's helping me a lot, you know? But if you had a dream, like, what else could he build for you that would even, you know, and I know I throw you, you know, a curveball here, but, like, what? What would it be that would add on here? And is it more specificity? Is it more, you know, is it down to the level of the specialization of. They just need follow up. A follow up in is it you know, give me the best place for them to go. Is it. You know what. What would it that you'd like to see to help even better manage this population? Right. I think, next frontier is going to be disease specific progression. You mentioned diabetes, prediabetes, diabetes. We talk about, CKD, a lot. We have, end stage renal disease or SCD, is, one of the most expensive populations are spending upwards of a $100,000 a year or more. In their later years in life with the, the ACO. And what we see is that if we can keep them in CKD three B, we can prevent that progression. We can keep them healthier. But the key is identifying, again, which of those patients are most likely to progress so that you can like you said, some of the, resources are very expensive for keeping people from progressing. So we identify which patients are most likely to regress. Send your resources there. I think we've got the same thing in heart failure. You got the same thing in many other diseases. And so it's identifying which chronic diseases to focus on, which patients within that subset to focus on and see which conditions are most likely to bring those patients in, what is most likely to bring kind of down their overall health. Let's focus on that and make sure we're progressing at preventing the progression of those different diseases. So I know you've done a lot of work kind of in palliative care. Yes. You know, kind of that end of life. Yeah. You've worked with, coda, you've worked with HDR, you've worked with a lot of that. What is that risk of mortality and kind of managing kind of at that end, what have you found? People often look for what are what are success for kind of initiatives that can be taken on in that that kind of period with that population? I think the the key that we have found in switching to this predictive analytics company is I again, identifying the right patients that I mentioned before. You know, we thought we had a good way to identify patients. We reach out and find out they weren't necessarily the right patients. It's a very awkward conversation if you're reaching out to patients about end of life discussion, and you had the wrong analytics telling you their end of life and they're not even close, or you're reaching out and they've already passed away. Unfortunately, if things aren't up to date, so identifying the right population of patients and then for us with both palliative care or getting one step before that and just starting that end of life discussion with patients before, they're unconscious in the ICU, things like that. It's really talking to them about the different options they have at end of life, explaining that there is a difference between palliative care and hospice care, and that if we are talking about either of these two levels of care, we're not giving up on you or providing the right resources for the level that you need at this point in time. And so, you know, with the you mentioned a, coda to a company that provides education to patients. And so there's different tools that we may use, but having patients be able to go through education about what end of life care may look like, you know, CPR or, you know, going through your, your chances of survival. What will your life be like after coming back from CPR? You know, will you have the same quality of life having that education given to a patient while they're in still a a state of clarity of mind, and they can make their decisions? It really changes what their wishes are, allows them to document that and put it into the system, and it changes the way that we assist them and help them with what care they need for the rest of their, you know, health care journey. So we've seen a lot of benefits in identifying the right patient, putting the appropriate resources and educational tools in place so that they can make their decisions and we can help to respect their decisions. So going around what's kind of one thing that you're the most excited about in kind of the next year of our revolution of, of predictive analytics from, the first is if you look at the predictive analytics, the basic philosophy has been historically identify risk resource allocation, I think. So the most exciting piece is moving beyond that and more prescriptive, especially with the LMS and especially in the realm of, value based care. Now we are seeing technology that's coming, automates chat bots that will actually have those conversation of the follow ups and guide you through the management. We are actually working in creating one to help educate the patients on starting utilizations, weight loss medications, the drug, diabetes medication. So I think so there is going to be some revolutions that's going to come in there. And the second thing is sensor ability to identify patients who are deteriorating at home much earlier, so that apart from the fact that you have significant resources allocated to follow them up, you can get those signals much earlier so that evolution is upon us beyond predictive analytics to more prescriptive, analytics that will go end. All right. Zach, I think we're actually on the cusp of this. We're leaning into this. It's present very, very near future. But I think natural language processing, reading through everything in the chart and using that not only for the predictive analytics, but also for chart summarization and kind of feeding it to the physician, feeding it to the nurses that are making those follow up calls. Because as you mentioned before, doctor kind series, you know, there's so much knowledge should be taken in by the physician. You know, they're reading through so much of the chart. Then we kind of cut it down to, here's a score, here's a red, yellow, green. I think we have to get somewhere back in the middle. It's a it's a green or it's a red because you give that chart summarization and then they can say, your red your high risk. And here are the five key factors. Even though we know there's 175 key factors behind that really populated that. And Brenna and I would go in, I would add to that, but I would also go into the quality aspects here. We are identifying how we could potentially prevent some complications in the hospital being one of them. Stroke being one of them falls is another initiative. Also through the eye with looking for certain patient populations that might meet certain criteria. How do you take that? I how do you get that summary? And then take your quality teams out there. Give them those patients to focus on, develop your programs, develop processes, and really improve quality. Thanks, Brenda. That's going to do it for this episode of Innovation Spotlight. Houston Methodist leading medicine podcast. If you enjoyed today's discussion, please share, like and subscribe wherever you get your podcasts. Thanks for listening. And until the next episode. Keep thinking outside the box.