Hello, everyone. My name is Dr. Lisa Charbonneau, and I'm the chief medical officer for Encompass Health. I'm very excited to be with you here today virtually and to introduce our two speakers on this topic of reducing transfers to acute care after admission to inpatient rehabilitation. Those of us that work in inpatient rehabilitation know that this is a very important metric that we follow. And so with that, I will first introduce Dr. Alberto Esconazi. Dr. Esconazi is a professor and chief medical officer of Moss Rehabilitation and the John Otto Haas Chair of Physical Medicine and Rehabilitation. He also was awarded the Frank Kruzan Award for Lifetime Achievement by the Academy of PM&R in 2019. And I'm pleased to say he's a former attending and mentor of mine. And secondly, we will hear from Dr. Joseph Stillo. Dr. Stillo is vice president of medical services at Encompass Health. He received his medical degree from the University of South Carolina and a PhD in chemistry as well from Lemoyne College. And Dr. Stillo has many years of experience in inpatient rehabilitation as medical director of Encompass Health in Toms River, New Jersey. And so I'm very pleased to have these two subject matter experts to talk to us about this important topic. Thank you, Lisa. Thank you for that introduction. So I'm going to talk a little bit about the idea of decreasing unplanned transfers. And I really call this an ongoing and continued challenge. And I am very hopeful that Dr. Stillo will give us better answers than what I have. So it's my hope that with this full program, you'll really learn new things and new actionable interventions that you can put in place in your facility. So I don't have any conflict of interest to declare. These are the learning objectives, and you have those in the notes, so you can come back at them at any time. So let me start by presenting the problem, which we know that with nearly half a million admissions a year for patients with Medicare fee-for-service is provided in rehabilitation hospitals in units across the country. And there are around 1,100 of those. One of the problems is that we have about a 13% average unplanned transfer out of rehab and to acute care. And this is a significant problem which affects our patients in an adverse manner. And so I think it's worth talking about this important issue. The readmissions and length of stay appear to be an important issue because we know that in acute care, decreasing the rate of hospital readmissions has been targeted as a high priority item in the healthcare reform. And although there appear to be some concerns on the validity of using discharge to acute as a quality metric, it is a rate that is now used to measure Medicare ACOs, Medicaid programs, and IRFs and SNFs. And so it's become really an important item that we need to consider. Government has passed legislation that applies financial penalties for acute care readmissions to patients who are Medicare recipients and now mandates reporting for inpatient rehabilitation facilities and skilled nursing facilities of this transfer rate. So there is great importance in lowering readmissions because acute care hospitals are looking at you and saying, hmm, if you have a high rehospitalization, you're not a good place for me to send my patients to. If you talk to physicians, they may say, I don't want to send my patient to that place because they end up coming back with new or ongoing complications. If you are a person who is seeking rehabilitation care, you may go to IRF Compare and find your data there and say, uh-oh, I have a higher rate than other facilities and so they may just bypass you. And certainly there are things like, you know, US News report that will be using this metric as an important parameter. And so if you want to really shine as a facility that provides excellent care, that acute care hospitals look at you with good eyes and that referring physicians want to refer to you, you really need to look at this parameter. We know that there are some predictors of unplanned transfers. We know that patients who are older in age or that have high complexity or intense medical needs, patients who have cardiovascular disease or respiratory disease or changes in mental status or those that have had a longer and potentially more expensive acute care stay tend to have a higher indices of unplanned transfer. From the point of view of rehab, we know that there is an association between reduced functional status and increased acute care readmissions. And so the more difficult your patient from the point of view of functional acumen, the more likely that you're going to have a patient that gets transferred out. And those with neurological orthopedic transplantation and medical categories tend to fall in those groups. And there are opportunities to better manage functional impairments as well as improve patient outcomes if you can reduce that. We also know that there are some modifiable factors that affect readmission. So we know that when patients come out of acute care prematurely, that may be a reason for a patient going back. There are also clinical concerns that might not have been fully resolved in acute care and that may imply a need for return. Failed handoffs, problems with medication reconciliation, complications following procedures, of course hospital-acquired infections, pressure ulcers and falls are also very key in projecting patients who will return to acute care. When you look at the reported causes of unplanned transfers from IRF to acute care, you see that acute respiratory failure, cardiac decompensation, GI bleeds, possible sepsis are very important. At Moss Rehab, we also see things like changing mental status and occasionally deep vein thrombosis. So I'll give you briefly what we're doing at Moss. We have 192 inpatient beds, which are separated into five inpatient locations. All of them are co-located with acute care hospitals. And we have nine inpatient discrete specialized units within these five facilities. And we have 15 cough accredited programs. This is data for a period of 12 months. Of course, I'm looking back because it's the data that we have available immediately. And you can see that that blue line indicates where we would like to see. That's our target. And sometimes we do very well. We are well below. And then sometimes we exceed it. And that's not a good thing. So we need to be sure that we keep our target in line with our performance. So when you look at this, you need to look at the data. And so we've broken our data into not only the different units, our different programs, but also by RIC groups. And so we've really tried to distill this information. And we can tell when our programs are doing well, labeled in green, or they are not doing so well when they are in red. And we compare ourselves to both the region and the nation. So when we look at this, we really try to analyze by program. We look at what are the issues that may be playing a role, things like not having access to internal medicine or hospitalists, being sure that we educate the emergency room departments that we may be using as a consulting service so that patients that go there not necessarily get transferred out. And then, of course, we analyze all discharges with a length of stay of five days or less. And then we do a lot of education and try to tag what we call red flags or potential medical issues. This again shows you just an example of how we do that. This is one of our units. We look at that. And then we look at if this unit that had, in a period of six months, 125 admissions had 10 transfers. And so we look at what were the reasons for those transfers, and we can get into the details of that. We then even look further. We look at the dates of the transfer. We look at the time of the transfer, where there are medical consults or other factors that may have played a role, and we document that. And then we go even more in detail, and we want to be sure that it's, are these transfers occurring during the daytime, during nighttime? Is it on a weekend? We really try to look at it as carefully as we can, because we want to analyze this data and share it with our teams. We do very in-depth chart reviews. We look at dates of the condition, change or deterioration. What was the time? What type of status change was noted? When was the last recorded vital signs and other things? Of course, if there were changes in medications within the prior 24 hours or orders. And we look at labs and imaging as well. We then look at human factors, you know, who was the nurse? Who was the physician? Was a resident involved? What actions were taken? We really want to be sure that we understand this as best as we can. We keep looking in and, you know, the point here is we're really doing a deep dive into this data, because we know that these factors are critical in making an informed decision. We also want to be sure that we understand what the acute care partners are doing and share that information with them. As a result of that, we initiated an initiative to reduce on-plan transfers to acute care. We deployed what we've called a modified early warning system for rehab patients. We deployed that into first the units that were having a higher rate of discharge to acute and then we moved it to across all of our units. We also deployed technological video robot interaction for the smaller satellite units. And we implemented a continued monthly review of all on-plan transfers by physicians, nurses, and senior leadership. This is our MUSE rehab modified early warning system. And again, you have access to this information in some of our facilities. We have a rapid response team, and so we've clearly denoted the difference between the parameters for these two groups. But we're looking at blood pressure, heart rate, respiratory rate, temperature, and level of consciousness or changes in mental status. And we made it, you know, you have to call the following individuals and then there are actions that follow as a result of that. That data gets documented in this form. You essentially have to complete this information then gets reviewed by the team. We have had the opportunity to promote better care transitions by really building our relationships, looking at the common goals. Acute care hospitals want to decrease on-plan transfers. We want to decrease on-plan transfers. And so there is a synergy here that we try to work. And we know that when there are less of these transfers, there's increased patient and staff satisfaction. And for acute care hospitals, reduces the financial penalties and improves quality score for both rehab and the acute care hospital. So through these conversations, we found that the process is very instructive, is constructive, and it really has changed the tone of the relationship. We've been able to not only reduce readmissions, but reduce 30-day acute care readmission rate. So we see this as opportunities for improvement. We have integrated PM&R early into the acute care assessment process. We now have physiatrists in the neurointensive care unit, in the transplant unit, and we have direct doctor-to-doctor conversation with all of the acute care partners. And of course, we are looking at how soon can we move that patient into IRF, because that's what the acute care hospital wishes to do and is good for the patient. We do pay attention to persons who have had organ transplantation. And again, we've created embedded physiatrists into those units to be sure that we can work with them as early as possible. As I mentioned before, we implemented video robots for our smaller units where we don't have a 24-hour physician present, but the physicians can now access from here without any problem. And we have found certainly that there are risks and rewards. The transparency of data has been seen as a risk by many, but we believe that this is potentially important because it is publicly available. Our teams are highly engaged and understand the importance of this work and accept really the challenge that it presents. And although there are some technology concerns, we have been able to deploy our robots. One of the major issues that we've had is our inability to really automate this process. We have had to rely on staff looking at the records and triggering the alarms based on their assessment. We've also looked at O2 saturation at nighttime as a way to increase the sensitivity of our tool. But again, that's a difficult thing to do when you don't have a fully instrumented way to do it. So I hope that this has been useful as an introduction, and I am thankful for the opportunity to this team to allow me to participate. And I hope that if you have questions, you certainly can reach us through our website or my email address, which is in my bio. And with that, I will cede the podium to Dr. Steele. Thank you, Dr. Eskenazi. I'm just trying to get the right slide up here. Okay. Thank you. So obviously, as rehab physicians, we don't have to tell you how important it is for our patients, especially our patient population, to be able to come into our rehab hospitals, meet their goals, and be discharged home instead of being discharged back to the acute care, which is probably one of the worst, besides death, probably the worst thing that can happen to them. So Dr. Eskenazi went through a lot of the statistics, and certainly what his hospital in Moss is doing to reduce acute care trends, both in terms of identifying patients at risk, and then, very importantly, the review process and things that they have in place. You're going to be hearing a lot of the same themes during this presentation. What I'd like to do is to, okay, somehow this slide is not changing, okay, is to review with you an algorithm that we derived with Cerner, who is the vendor for our electronic medical record, which I will refer to as ACIT, which is the acronym for our medical record, our EMR, that's Advancing Clinical Excellence Through Information Technology, ACIT, describe to you the way we derived the algorithm, the way we validated it, as well as associated tools that we put into the system, the way it displays to clinicians, and then some clinical decision support strategies that we put in place. So how is, I'd like to describe to you what we call the REACT algorithm, all right, and the REACT really stands for, the algorithm is going to be showing us, as I'll show you in a few slides, showing us clinical information in a way that we as clinicians have not seen it before, organized through an algorithm, through a display in the EMR, and presenting it to us to say, doc, nurse, therapist, et cetera, would you like to react to this particular situation? So let me go ahead and show you the, Dr. Eskimo-Nazi already framed the issue that we have, AAC is Encompass Health Corporation, IRF, something happens which causes the doctor to pull the trigger to say, I need to send that patient to the acute care, but we know that that arrow can certainly spin around, so the patients can not only be readmitted to the acute care hospital, but they can also bounce back to us, that with the so-called ER in return. And so, the three groups, and Dr. Eskenazi showed this in some of his statistics, but the three groups of patients that we like to look at are those with acute care transfers that are unplanned, meaning we didn't know at the time we admitted these patients that they were going to be readmitted back to the acute care or presented to the emergency room. And of the acute care transfers that are unplanned, you divide those into those that are readmitted back to acute care and those who go to the emergency room and return. So, we do our data analysis. We like to look at all of these things. So, our purpose in developing the REACT program and tools and algorithm was that we utilize our clinical information system to identify a patient's risk of acute care transfer at the earliest point possible, because obviously the idea is to prevent the acute care transfer. So, the objectives of developing this predictive model was to partner with Cerner to develop a predictive model on the risk of an acute care transfer within the rehab setting. So, the data that Dr. Eskenazi showed you is all according to his IRF, MOS, and the data that we use to build this algorithm is all according to the IRF hospitals in the Encompass Health System, which is about 138, and I'll tell you more about that in one minute. We wanted to design tools around the algorithm to utilize within ACID to see the next REACT risk and the predictors that are related to the REACT risk. So, not just the REACT risk, but what are the predictors that bubble up to the top on a given patient in real time. We wanted to validate the algorithm via pilot program and then design and implement integrated clinical decision support measures. And there's a lot of overlap already. I can already tell you between what Dr. Eskenazi just showed you and our data analysis and our strategies. So, we wanted to optimize the REACT program by designing complementary tools and workflow strategies. So, here's a history, briefly, as to how we developed the REACT model. So, we developed the REACT model, the predictive model for acute care transfer was developed. And the way we developed it was we looked at over 90,000 encounters from over 80,000 patients, over a two-year history of data from, at that time, HealthSouth hospitals before we changed our name to Encompass Health. There were 30 clinical variables, which was selected out of over 300 based on SNFing criteria, clinical relevance, and practical feasibility. Many of you that have been involved in developing these predictive models know very well what I mean when I say we had to spend a lot of time deciding whether or not the clinical features, the clinical variables associated with a higher degree of the acute care transfer actually made sense within the system and what the EMR can tell us how valid that data point was. We then developed a face-up clinician tools in the electronic medical record. The predictive model and the clinical tools were installed in pilot sites back in September of 2016. We validated the model. We felt it was effectively predicting the risk of acute care transfer. We developed Beacon applications, which is a separate IT platform with our IT colleagues at Encompass, and other supportive tools that were added a little bit later. We went live with what we call a phase one by July 2017. Then all the remaining Encompass sites went live in November 2017, and every time now we bring a new hospital into the Encompass Health family, that hospital is live with the electronic medical record ACID, but also live with this model. What is this program not? The REACT risk does not replace clinical judgment. This is an adjunct tool, and that's the way we look at it. It's an adjunct tool to already existing processes to mitigate acute care transfer. Dr. Eskenazi, you heard him talk about the MUSE, right, the medical early warning system. In Cerner, we have the St. John's sepsis protocol, so we have some warning systems already built in that we've been using that are pretty tried and true over the years. REACT is an adjunct tool to supplement those. But the overarching strategy with this system, that with this system tool, the clinical team can be alerted earlier as to the combination of predictive clinical factors, creating a greater risk for acute care transfer, and ultimately intervene earlier in the hopes of stopping that acute care transfer from occurring. What are those features that bubbled up to the top in the terms of deriving the model? This is one of two slides which is going to show you those 30 or so variables. All right, you can see things that really are no surprise. Some things perhaps might be a little surprising, but you can see that we started with appetite, both the initial appetite and the documented follow-up appetite, the initial Braden score and the follow-up Braden score, missing therapy, emission motor index. Dr. Eskenazi talked about, he mentioned the functional level of a patient, well, the emission motor index, which is not exactly the emission motor FIM, did bubble up to the top. Vital signs, including O2 saturation, O2 flow, and pain. And before we move on to the second set of variables, I just want to give you an appreciation because I can tell you that as, well, okay, regarding, let me go back. Missing therapy is a very, very important feature. But in the Cerner system, we have six different reasons whereby the therapist can document why the patient missed therapy. And of those six reasons, three of the six bubbled up to the top in terms of being statistically significant as associated with an acute care transfer. Those are missing therapy because of tolerance, missing therapy because the therapy was on hold or something was on hold, missing therapy because of sickness, okay? So moving on to the second set of variables, the second half of the variables, we see things like some labs, including some medications, including these medications, as well as anti-infectives. But if I show you the way these panned out in terms of order of priority, order of statistical significance, this is what we see, that at the top of the list, the very top predictors were first the Braden change followed by the initial Braden score, respiratory rate pulse. But look at missed therapy. Now we know clinically, we say, well, we know that when patients miss therapy, there's probably something going on, which makes us want to take a closer look. But other things such as pain levels, appetite change versus the initial appetite, vital signs then on down from there through those 30 variables, okay? And regarding the Braden score, as a physiatrist, I've always known that the Braden score was something that the nurses were doing, a standard scale that they applied to judge whether or not somebody is at risk for skin breakdown, okay? In the Cerner system, you'll see that these are the six variables which make up the Braden score, sensory perception, moisture, activity, mobility, nutrition, friction, and shear. Now when you look at sensory perception, the way it's scored, I have it broken out into the way these are scored. So in terms of sensory perception, sensory perception is not just, for example, that you're a T12 Asia A paraplegic and you have no sensation below the T12 level. Sensory perception, according to the Braden scoring, is also can the patient tell you consistently whether or not he or she needs to be turned, okay? So we're starting to see that the significance of the Braden score could very well be couched in the fact that it also speaks to not just sensory level in terms of sensation, but also sensory perceptive level, all right? So and that's the way we explain how this Braden score, but in addition, moisture. Is the patient moist? Is the patient becoming incontinent? Activity, mobility, nutrition, when you already saw that appetite and appetite change is included. So maybe this is giving us some deep, more detail buried within the Braden assessment, okay? So it was really fascinating to me as a physiatrist practicing over 25 years that the Braden score would bubble up like that. So we're giving more emphasis to that. So what I want to show you is the way this algorithm works. I fly a lot, and I probably should not have taken this picture of the cockpit as I board the plane. But as the pilot and co-pilot were going through their pre-flight checklist, it just occurred to me that there were gauges above them, in front of them, in between them, behind them, and way up to the right you can't see, way up to the left you can't see, right? So you can imagine that if something up here were maybe a little off, maybe something down here were a little off, and maybe another gauge were a little bit off, that that when the computerized systems within the plane, see those three things might be a little bit off, but the three taken together is when that red flashing light flashes to the pilots, okay? That's the way this algorithm works. And you can see that the way I've categorized the different variables I showed you on those two slides, where they're all interacting with each other, all right, and as they interact with each other, some may increase the risk, some may decrease the risk, but we are then left with an ACT risk, a risk of acute care transfer, okay? That's the way the algorithm works in the system. So what happens? Because I get this question a lot. The algorithm starts deploying the second the patient's admitted, the second our staff starts inputting data into the electronic medical record, risk stratification, react risk stratification ensues, all right? Then we show, we determine which variables identified for that particular risk. We then display those to the clinicians in a face-up manner, all right? And this all happens in real time. You know, the second a nurse enters an abnormal vital sign, it could very well change the react risk level, and that shows up immediately, within seconds. We then can also show not just how the patient's react level is now, but how is it trending? Is the patient's react risk getting worse, improving, or staying the same? And obviously, if an acute care transfer occurs, these are all the things that Dr. Eskenazi already talked about. So what we do with that is we try to identify, using our data analysis techniques and the IT strategies, identifying the reasons for the acute care transfer, and then try to develop clinical decision support measures, all through analytics and feedback, okay? So like I said, there's already an overlap that I see between Dr. Eskenazi's presentation and this. Who has access to react? What you see here is pretty much the entire clinical staff. And I put this up here to show you that react is a common language, is a common language. When somebody says that, talks to another clinician in the hospital and says, somebody has a high react score, or somebody went from high to very, very high risk, we know what that means. It's a common language. So the algorithm runs continuously within the electronic medical record behind the scenes, in the background, in the back end. It generates a risk score and risk stratification for acute care transfer. The risk levels were broken out into low, high, and very high. And this decision we had to make when we were deriving the algorithm to try to maintain some reasonable levels so that not everybody came out as very high, not everybody came out as low. But we tried to look at it, and I'll show you this more in a minute, but the clinician tools that generate from and support the model include the react patient list, the notifications, the nursing test, I'm going to be showing you each one of these, and trending tools and the react retro application. So early on, we had to go in and validate the model. And at that point, we only had about 3,000 total discharges. And at the time, we had made the decision to set the thresholds into low, high, and very high react risk level, ACT risk level. But what I want to show you here is that if you were low level, that's even. But if the patient were in the high level, that the patient had twice as high odds of transfer. If the patient were in the very high level, the patient had a six times higher level than if the patient were in the low level. And one thing I do want to tell you is that low risk does not mean no risk, okay? You'll see this in a minute. Low risk does not mean no risk. When we try to teach these tools, we ask our clinical staff to be cognizant of react issues as they occur, even in the low risk patients, because some things could be meaningful. So obviously, our goal is to reduce the acute care transfer level. And if we look at those of you who are statisticians, the AUC, the area under the curve, or the C-stat for this particular model was 0.7. So it was very, very reasonable in terms of its predictability functionality. And to show you the acute care transfer percent, actually, this is not just acute care transfer. This is not just patients going to the emergency room. These are patients who were actually readmitted back to acute care in the Encompass Health Enterprise back to 2009. So you see that we started getting a little bit of a nice drop over the years with strategies that were already in place. And then it seemed to level out before REACT was implemented. REACT was implemented at the end of 2017. Then the nice thing is that you'll see at least another, about another 1.1% drop below 10.5. All right. So Dr. Espinosa measured targets out. The target that we have is 10% across the company. So currently, for the entire year of 2019, we're at 10%. So what can we do with our data analysis tools, like what we call the REACT RetroApp? And this is just a screenshot showing you our discharges across the entire enterprise, about 365,000 discharges over the course of two years, 18 and 19. And this is the prevalence of how the patients sort out in terms of REACT risk. And this is based upon the mean, the average REACT risk from admission to discharge for all our patients, no matter where they went. The community, acute care, et cetera. And you'll see that the vast majority, 70% of our patients average in the low range, in the green. Right? About 25, about 20% of our patients average in the high range, in the high risk range. And about 7% average in the very high risk. Now the next slide I'm going to show you is going to show you the same pie chart, but the prevalence of the patients who wind up being discharged to acute. And you can see the patients, this red slice, increase by a significant amount. So the next slide is going to show you the patients who actually wind up being discharged to acute. Okay? So a couple of things here. These patients, about a third of our patients, let's just say it's a third, a third, and a third. A third of our patients in the low range wind up being discharged to acute. And I already said that low risk is not no risk, right? And that's what I meant by that. Patients in our high risk wind up being discharged to acute, but you can see that a greater percentage of the patients in the very high range wind up going acute. Let me just go back to the last slide and just show you the greens, the red slice is the very high range. And look what happens to the patients who are discharged to acute. Okay? Another way of representing this is a way that we validated the model that with that same cohort of patients in 2018 and 19, where we averaged a 10.4% acute care discharge to acute rate, discharge to acute rate. If the patient averaged from admission to the time of discharge, wherever the patient went and patient averaged in the low react risk, the patient had a 5.7% incidence of acute discharge to acute. If the patient averaged in high, this number increased by about three times, threefold. But if the patient were in the very high range, the patient increased it by a factor of about six compared to low range. So I showed you the odds ratio initially with our initial derivation validation was an odds ratio of six times greater. And believe it or not, a year after we implemented this model and we looked at the 2018, 2019 data with so many more hospitals in the system, we can see that odds ratio pertained. It persisted. So this is a way of saying, look, the model is doing what we wanted it to do. It's predicting the acute care transfer, in this case, discharge to acute rate of patients based on a risk, an algorithm. So let me show you how these things display to our clinicians. The first I want to show you I mentioned is the react patient list. And I'll show you exactly, and those of you who have Cerner perhaps have access to a react patient, have access to a patient list. But basically this can be accessed by a hyperlink here in the toolbar where it basically can say I want my two south list or my two north list or my brain injury unit list or my spinal cord list, and that list will show up. Then the next thing we can do is sort, and you can't read this, but it says react risk. And this is very high, and then it starts high. So I just clicked on this and it sorted my patients from the highest of the high risk patients all the way down. If I had scrolled down, it'd be the low patient down at the bottom. Okay. So this patient numerically, and I can show you the numbers, but we don't really concentrate on the numbers where we set the thresholds. But this patient who's named fever pain is the highest of the high. Okay. So what? Well, if we go to the next slide, if you hit this little carrot here, this little arrow, it's hard for you to see. It will actually open up and show you the list of features from the 30 features I showed you. It will show you which things are pertinent to this particular patient. Okay. And I can scroll down here and see them. Okay. So this is one way where I have some docs who will go from patient to patient before they round to look at their react risk. Anybody who uses Cerner, and actually most EMRs, know there's a lot of ways to do the same thing in the system. Okay. So that's the first thing, is that the way it displays in this particular dashboard, I will show you how it displays in a dashboard in a few minutes as well. All right. So we also built a beacon application, an IT functionality tool, which can now show our clinicians what's going on in terms of react risk, and if the patient's worsening or getting better, right? And this is also available to these particular individuals within the system. We didn't give it to everybody, but we gave it to mostly everybody in considered leadership positions, including especially nurse managers, because oftentimes they are the ones to take a look at this tool in the morning. And this tool can be, the in-house beacon application can also be accessed using a button in the toolbar. What does it look like? Okay, so this will take you into a beacon application. Okay, and here, let me walk you through this. And I just actually made this slide yesterday. So this is actually a profile of what it looks like in one of our hospitals. On the left-hand side, we have the patient list. We have the length of stay. Remember, Dr. Eskenazi measured length of stay as being a very important feature. We agree. The nursing station, the attending physician. Then what you have here is the current REACT score. Now, numerically, the way we set the thresholds is anybody over 0.19, we don't have to worry about the numbers, is in the very high range. What you need to know for the here, for this slide, is that the patients in red are in the very high REACT risk. Patients in orange are in the high risk. And then yellow are in low risk. And this list obviously continues down if I were to scroll down. But what's interesting about this is that we say, okay, is this patient getting better or worse? And you can see that we added the 12-hour trend, the trend over 24 hours, and the trend over the prior 48 hours. So if I were using this in the morning, I'd say, okay, we know that this particular patient here, the fifth patient down, is already in the very high risk. But looks what happened in the last 12 hours. This patient was spiking. Red means this patient is spiking pretty high. The slope, basically, of the REACT risk is getting very high, all right? If this were orange, and same thing in this patient who's in the high range, that this particular, this patient is spiking somehow in the last 12 hours. This patient had been spiking a bit, but then leveled off, you can see. So red means the patient's really spiking. Orange means the patient is going up significantly. Yellow means the patient tends to be trending up. Green means the patient's not spiking as much. Then what I can do is I can take this particular patient, click on it, go to the patient view, and the second thing that our clinicians can see right below this is the following. Okay? So now this is showing up in my screen as yellow, but every time you see a yellow bar, this is when a REACT risk score is being generated by the system. This goes all the way back to the time of admission. I said that things start generating from the time of admission all the way through current. I showed you that the patient's risk level was red, meaning the slope was increasing. That risk level was showing you this, that this patient was really increasing in terms of ACT risk, REACT risk. Why? Well, if you scroll down, what you have is all those clinical indicators I showed you, the clinical features are ranged from left to right across this horizontal plane here, all right? And the way this works is our IT guys are incredible. They said, okay, if you start down here, actually, you can't see my pointer. If you start at the bottom, that's the older, but toward the top is the most recent. And what happens is if any particular feature gets worse, it's indicated in pink. If any particular feature improves, it's highlighted in green. And you'll see, for example, in the pain score right smack in the middle that somebody's pain level got worse and that figured into the algorithm, and then it got better. But you can see toward the right, you've got some vital signs which are fluctuating, all right? You've got this patient is fluctuating in terms of pulse going in and out of low-level tachycardia and blood pressure is abnormal. Okay, so this is something that's accessible to our clinicians, not just in the BEACON application, but to our docs as well, right within the EMR. So if we go to the dashboard, you can see that this is the typical dashboard that we see within the CIRNA EMR. And we embedded that with the CLICK technology, people. We embedded that right within the EMR. So in the morning, I can say, hmm, I see this patient react risk went up, and then I can see the same exact graph trendline as I saw if I went outside in the BEACON application. So I think that one take-home message is also that embedding these tools within the EMR so that they're face-up to clinicians, easy to access, will help them, will help the engagement and adoption of using these tools. And again, down here, I can see what's going on with the actual variables. How else do we notify medical staff? Well, those of you, again, using the CIRNA EMR, and in other EMRs, there are message centers. This is a message center. This is a place where you can go and assign new telephone orders. You can review documents which were forwarded to you by your nurse practitioner, medical student residents, et cetera. But what this does is this gives a react message, and here it's telling me that, hey, still low. You've got somebody who went into the high range and the high range here. These notifications, the way we set this up is that patients who go into the high range, patients who go into the very high range, patients who increase from the high to the very high range will generate a message. These are not alerts. These are just simple passive notifications. Alerts stop somebody from proceeding in the system. These are not alerts. These are simply notifications. And we built a level of efficiency in here where if I'm triaging, signing my telephone orders, et cetera, signing other documents, whatever I have to do in my message center, I can simply select this patient and then the patient name here and go right, navigate right to that patient's chart. So with two clicks, I can go right into that patient's chart without having to go into my patient list, scroll down, find the patient, click, and multiple clicks. Obviously, the number of clicks to do anything in our systems when we work with EMRs, we all know that that's going to make a break whether or not clinicians, perhaps even especially physicians are going to use it or not. In addition, when patients go into the very high react range the nurses are tasked to do an extra standard assessment. Okay, this is really important. And what they will do is what we call the standardized assessment. So this is a basic assessment where they'll look at the patient from head to toe, not just look at whatever the parameter is that triggered the patient going into the very high and then decide if the doctor needs to be called. Okay, so that's another functionality, another efficient functionality. Now I want to show you what the dashboards show you. I showed you the list, but I promised you that I would be showing you what the physician dashboard would look like. And there are three columns of what we call components. All right, the middle, this is a vital sign component, patient information component. This goes all the way down if I were to scroll lab component here. We place the react component in the top right in the third column at the top, oops, at the top right. And that's okay, because what I wanted to show you is that now this will, sorry about that, my fault. What this will show you is the patient's current react score the date and time that that score was generated. I showed you those yellow bars on the trend analysis. And then it will line up the actual features as to what's going on with this patient. Okay, here you can see nurse put in 99.3 and immediately the algorithm recognized it as abnormal and put it in here. But if I were to expand this react component to show you if the doctor were to scroll down that there are things that the doctor will be able to see. For example, when did the patient first miss therapy? And now we've actually enhanced this to show you the last time the patient missed therapy. And the same thing with labs. What was the patient's initial neutrophil? And now what is the patient's current neutrophil count, et cetera, okay? So that this is a demonstration of what the dashboard looks like. When acute care transfer does need to occur, we did revamp the OEF, the order entry format for the transfer to acute care order. And this is what it looks like. But what we have here is two required fields. Any of you who use Cerner know that yellow means a required field. It will not let you place this order unless you put in a required field. So we wanna know, hey doc, was it a planned or unplanned transfer? All right, then there's a dropdown list of reason or reasons and notice here we say you can actually multi-select from this. We wanna know, why did you send the patient out? Dr. Espinosa showed you that, that it's important in doing analysis as to why did we send the patient out, okay? And this will just open up to an alphabetical dropdown list where you can pick and choose one or several of these particular clinical features which caused you to send the patient out. And we tried to discourage use of other, but you can imagine it's not that easy. All right. So I mentioned the data analysis application, the other IT functionality called the React Retro application, all right? And we have all our medical directors across all our 138 hospitals have access to the React Retro application because in doing acute care transfer review, they can actually go in there and decide what were their reasons for transfer? This is a snapshot of the entire company. All those patients I showed you before, all the unplanned, the 10.4% of the 365,000 unplanned transfers, all right? So these are patients who went to the emergency room, not just discharged through acute, but went to the emergency room. And you can see on the pie chart that the main reasons are things that are very reminiscent of what Dr. Espinosa showed, right? Mental status changes, respiratory distress, neuro status change, chest pains, suspected stroke, fall, et cetera. Okay, but we wanna know that in your hospital, in my hospital, that we have to be concerned about the following. Okay, because every hospital has a bit of a different profile. Then Dr. Espinosa showed you that the type of patients can cause increased risk. Well, in this application, we can drill down to actually the RIC, all right? And this is sorted through the highest number, numerical number of discharge to acute. But we can see that it's no surprise that perhaps that stroke patients, neuro patients have a higher degree of, and cardiac patients have a higher degree according to the RIC. But we can also drill down and subcategorize the stroke patients into the 10 corresponding CMGs, right? The case-mix groups. And we know that depending upon that, the case-mix group, the risk goes up or down, okay? So just to show you the type of data analysis adjunct tool that we built along with the REACT tools. If you look at this article of PM&R Journal on Potentially Preventable Within-Stay Readmissions Among Medicare Fee-for-Service Beneficiaries Receiving Inpatient Rehab, it's a great article. But what it does is identify preventable within-stay readmissions. This is exactly the ACTs that Dr. Espinosa and I and Dr. Charbonneau are talking about. Because when you look, they say there are five particular categories of reasons why outpatients go acute care. Inadequate management of chronic conditions. And it's a really good framework. I suggest to you, especially those of us who are educators, to remember this, and as we're educating residents, et cetera, to remember that, look, there are reasons why outpatients go out. They can be categorized as inadequate management of chronic conditions, inadequate management of infections, inadequate management of unplanned events, inadequate prophylaxis, inadequate injury prevention. But if we expand out the first, inadequate management of chronic conditions, we see that on this list, the type of ICD-9 codes at the time that were listed included asthma, COPD, and CHF, and obviously diabetic complications and hyperhypotension, blood pressure management issues, okay? So this will kind of help us to decide. I showed you already that the main reasons why outpatients go out from the encompass health enterprise is mental status change and respiratory distress, right? So if we start now thinking, if we wanted to develop clinical decision support tools, where would we start? So what we did was we combined our data with this, and we said, okay, let's develop a tool which help us survey patients with congestive heart failure, CHF, right? So we developed a power plan. A power plan is a combination, a set of orders, which would facilitate a medical staff user to survey patients as to whether or not they are getting worse, perhaps if their CHF is exacerbating or not, okay? We did the same thing with COPD, all right? In the next slide, we'll show you the COPD exacerbation surveillance power plan. And these check boxes means that any particular doctor, any particular site can check or uncheck depending upon customized. These are all customizable and be able to save as a favorite, all right? And the second slide of this, we're currently in the process of instituting an alter mental status workup power plan, because you saw the mental status changes were right up there along with our story distress, okay? So what are some of the strategies that we have instituted in our system? And I'm happy to tell you that many of them, you already have heard from Dr. Espinosa, it's great. We ask that the multidisciplinary team huddle review every day, at least, looking at the react patient list that I showed you, looking at the in-house application. And to show you actually some pictures I took during a hospital's morning huddle, you can see that they're projecting the screen in front of everybody in that huddle, but you can see that you can recognize the in-house application. They're actually looking to say, okay, what does our profile look like today? Who are the ones who might be spiking in the last 12 to 24 hours? And then you can see that they actually drill down to look at the trending profile. So these are tools that we encourage our hospitals and staffs to use in a real-time basis. So besides integrating those into clinical huddles, we try to have focused communication and the admission watch, looking at those patients who might've been admitted in the last one to three days, all right, and try to look at those for our high-risk groups and cardiac. In addition, we talked about the formalized additional assessments for very high-risk patients. Okay, now there are probably more that we can do in addition to asking the nurses to do a head-to-toe assessment on patients who spike into the very high risk. But again, as we develop clinical decision support tools and integrate them with Cerner in something that Cerner is developing and rolling out now, something called SmartZone, where hopefully it can look like, hey, doc, your patient seems to have gone into the very high risk. And guess what? Your patient has CHF. So you may want to link to the CHF power plan where your patient looks like he or she is declining in mental status. So you may want to go into the ultramental status power plan or get a speech language workup, start some initial workup in terms of pulse ox, things that Dr. Eskenazi mentioned. So as I'm finishing up, I'll show you the last slide with regard to best practices. So the idea is to develop best practices based upon data. All right, Dr. Eskenazi did a great job talking to you about the way acute care transfer reviews occur. We encourage our clinicians to our hospitals to have physician-led ACT reviews and basically ask the question, we know that at the time the patient went to acute care, we know that based upon our capabilities in the hospital, the patient was better off in acute care for a workup. Okay, we know that. But the question we ask them to ask is, could we have seen it coming? Could we have prevented this acute care transfer use of the emergency room for all the reasons that we know in the earth world? Okay, including do we have the respiratory management protocols that we need? Do we need expanded lab service? Perhaps you couldn't get that stat lab because we don't have a contract for that. Perhaps we couldn't get telemetry to monitor a patient because we didn't have the contract. Do we need to expand our specialty in consulting services? Could we have handled the clinical issue in-house? If not, why not? And that includes both in terms of assessing the patient, working up the patient, then treating the patient. And lastly, I will mention something that I felt very, very strongly over the years in my practice, that the process for effective communication with the emergency room, if a transfer does occur. So I would call the emergency room doctor and say, look, this is Dr. Still, I'm sending this patient over. No, the patient's not coming from a nursing home, patient's coming from a HealthSouth or Encompass. And here's the reason why I'm sending the patient over. I need you to do this and this. I'm in touch with the PCP who knows that I'm sending the patient there in case the patient needs to be readmitted tonight. He can hand off to his partner PCP. And I'm in touch with the family as well. And I told them I would be calling you. I may even be stopping by later on. But here's why I'm sending the patient over. I am not looking for you to readmit this patient back to acute care. I'd like you to assess this, this, and that on the patient. Give me a call. We will decide the best disposition for this particular patient. That alone, and I tell our doctors across our enterprise, that alone, I feel in my heart, can help to reduce the acute care transfers, the patients who are actually readmitted back to the hospital. Just that one phone call, physician to physician phone call. So with that, I will say that acute care transfer reduction is a very, very important metric, as Dr. Resk, when I was beautifully explained, that helps us distinguish IRFs from other levels of a post-acute care. It can be leveraged to develop algorithms to assist our clinicians in identifying those who are at risk of ACT and when that occurs. And this can occur on a real-time basis. So combining an algorithm with other smart technology can facilitate the development of best practices according to the individual hospital needs or even the enterprise profile of acute care transfers, as I showed you, with the development of clinical decision support tools. Okay, so thank you. And I wanna thank Dr. Charbonneau and Dr. Eskenazi for their contribution to this very, very important issue. And now I'm gonna give it back to Dr. Charbonneau. Thank you so much. And I hope that you all feel able to contact the speakers. If you have any questions on this topic, their contact information is in their bios. Thank you for joining us. Thank you.

reducing transfers

acute care

inpatient rehabilitation

unplanned transfers

healthcare reform

predictors of unplanned transfers

readmissions

communication

REACT algorithm