OK, so I want to take 20 minutes and set the stage for Teresa, who's going to talk about predictive modeling. And I'm going to skip around a lot, but I just want to talk about the state of the art of end-of-life care and communications in the intensive care unit and talk about where we are and the impact on patients and families. I don't have any disclosures, by the way. So a little bit about obstacles to communication, working with family interactions. There's a lot to do in 20 minutes, so I may go a little fast. And I said in our end-of-life group at the small table, this is a self-selected group of people who are here because you're open to improving the quality of care that you provide in the ICU. So already, you're a self-selected group that understands the need to listen and to learn and to get better at what we do. So already, I realize I don't really have that much to talk to you about. But let's talk a little bit about the obstacles to communication. I mean, this is just a small list, but I think one of the biggest obstacles for learning how we as providers, physicians, nurses, respiratory therapists, social workers, clinical managers, et cetera, is the chaotic and frenetic pace in the ICU. I mean, we know that it's so hard to learn how to go from being speedy and needing to make instant decisions and really carrying a lot of weight for each patient on our shoulders and then suddenly stop and be in a quiet environment with a family who's losing a loved one and help them with complex decision-making. And how to go from being speedy and chaotic to silent and still is a skill that we never learn, right? And so that's a real challenge to communication in the ICU. And we're not really trained how to deal with heightened emotion. We're not trained how to deal with anger. We're not trained how to deal with grief. So some of us learn that naturally. Some of us learn that by caring for patients. But for a lot of us, we're thrust, especially as young providers, into environments where people are angry at us because their loved one's dying or in deep grief of losing a partner of 40 or 50 years. And we really don't have the skill set to know how to do that. And so we have to acknowledge that as a real obstacle. And I always say this, but I just was in India giving a talk on mindfulness and communication skills to the Indian Society of Critical Care Medicine. And I asked them, what I ask people here in the United States and in Europe, how many of you had a formal course in communication or listening skills during your training? And maybe 5% of people raised their hand, and they're all young. And yet, it's what we do all the time. Really? I mean, much more important, I would say, than making the right diagnosis is communicating with each other, communicating with our patients, communicating with families, not just during times of need during end-of-life decision making, but period. And yet, we are never trained in communication skills. And we're certainly not, as doctors, trained in listening. And it's many ways, I would argue, the thing we do the most poorly is learn how to listen. Listen, truly listen, deeply listen, and not interrupt. And we do it to each other, and we do it to our patients and their family members. So these are just a small list of the obstacles to communication. This is a great study from a nursing journal from 2003. And this is just another list of some of the barriers to good communication. And I want to move quickly enough so that we can have question and answers. And I'm not going to belabor this. Does it matter? So these are two studies that I show all the time because I love them. This is a French study from 2001. And basically, it was 43 ICUs in France, almost 1,000 groupings of patients and family members. And they measured anxiety and depression. And it's not surprising that almost 60% to 75% of family members had anxiety and depression during the loss of their loved ones in the ICU. But look at the factors that were, through a logistic regression, independently associated with the development of anxiety and depression in family members. Oops. Absence of regular physician meetings, absence of a proper space to have the discussions, and receiving conflicting information. So these are things that happen all the time. We talked about in my small group about the fact that you never know what your consultant is going to say to the family. After you've spent a week preparing the family for the fact that their loved one has almost no chance of surviving the ICU, I always use neurologists because it's my favorite people to get mad at. But they come in and go, well, there's always a chance. And boom, you've just lost a week of it. And honestly, what I see is physicians going in and not bringing the nurse in. And so if the nurse is the one who spends night at the bedside constantly, and we're there once in a while, and if the nursing staff doesn't know what I said to the family member, that's a basis for conflicting information. So this is a good example of that we are very much responsible for some of the anxiety and depression that family members of our patients who are dying in the ICU feel. But just as important, look how long it lasts. So this is a study. There's a US study around the same time that looked at family members one year after they lose loved ones in the ICU, one year. And so not only is what we say and how we handle ourselves with families responsible for engendering the anxiety depression, but it lasts a long time. So you can see 46% perceived conflict a year after their loved ones died in the ICU. And the most common thing they reported was conflict with the medical staff and communication problems. Again, not rocket science. And we see it. This doesn't come as a surprise to anyone sitting here. I know that. And you see some of the other things that you saw in the previous study, and these two are unrelated, was a better source for family discussions and a preference for having the attending physician be the one that talks to you. And yet we know that often, now we also know that sometimes where we work, we'd prefer that the attending physician not talk to the family given their communication skills. On the other hand, family members don't feel that. They feel they're attached to their attending physician, and they would rather hear it from him or her rather than have an internal resident do it. I'm going to skip over this. So why are patients and families frustrated with our communication skills? I also love these next two studies. This first one is from Jim Tulsky. So Jim, as we know, is a really highly regarded end of life researcher. This is in a general medical clinic. And Jim taped family physician. Oh, no, this is patient physician. This is in an outpatient setting. And he found that physicians spend 75% of the time talking in these patient physician meetings. So I always joke and say, that's a great thing. Go home to your partner and say, I've got a new idea. I think, from now on, I'm going to spend 75% of the time talking, and I want you to talk about 25% of the time and see how that goes over with your spouse. And yet, this is what we do. And this followed. This is Randy Curtis. So Randy, to me, Randy and Elie Azoulay are the two foremost qualitative researchers in end of life in the ICU around. And this is in Randy's shop, which is at the University of Washington in Seattle in Harborview. And he did the exact same thing. He videotaped 51 family conferences. So this is families and providers in a very high level institution. And he actually found the same thing. Clinicians spend 75% of the time talking. And I mean, it's like you can't make this up. And yet, I can't honestly tell you that I know how much of the time when I'm in a family meeting, I spend talking. No one's taped me. And for all I know, it could be 70% of the time. But it's not the basis for a good relationship with families. And in fact, Randy showed that, again, not rocket science, the longer families talk, the more they're likely to be satisfied. Well, I mean, but it's nice to see that through qualitative research, we can demonstrate that we talk too much and that when we let families or patients talk, they are more satisfied with hearing and understanding. So this is why it matters. This is why our communication skills and patients and families report struggling with our communication skills because we don't let them talk. And I'm going to skip over this. This is from the same study. And it's worth it. You see the reference here. But what Randy did through qualitative analysis is in listening to these videotaped conferences, he actually and his group identified missed opportunities during these discussions. And these are missed opportunities, you could argue. If I spend 70% of the time talking, it's not surprising to find out that I miss a lot of opportunities to empathize with what families or patients are saying and to respond with empathy to some of them. And you can see here some examples. And they're in the 10% to 20% range. So working with families, I'm not going to take time. I think you all know LaTrette. This is Elie Azoulay. Randy was on this as well. His landmark randomized controlled trial. This randomized controlled trial was about family conferences. It involved giving a brochure. This was back in 2007. So the data were collected in 2004. And it involved giving them two most important things, three, regular family meetings, a brochure that was available to all families, and a script of how to have the family meetings that was based on Randy's value system. And you see here, there are now a lot of mnemonics on how to have a family meeting. Tony Box got a couple from OncoTalk. And I would refer you to them. It's sad that we need a checklist to remind us how to talk to families. But the truth is, we go so quickly. And as the data suggests, we're so busy talking that we don't remember to do a number of these things. Now, hopefully, eventually, we internalize this list and this mnemonic enough so that we automatically cover this. But I do think it's good for younger clinicians, in the same way that we know the value of a checklist for washing your hands, for inserting a central line, for managing sepsis early. I think the value of a checklist during a family meeting is very, very potent. Because it allows you to self-evaluate, OK, what did I leave out? Oh, wow, I forgot to empathize with their heightened emotion. Right, I should have done that. Because I ran away because they were so mad at me, I just shut down. So it's good to have this list of, what do I need to remember that will force me to stand there and just listen to a family? That's why I think these mnemonics are so important. And as you probably know, the reason it was in the New England Journal is there was a statistically significant difference in anxiety, depression, and PTSD before and after this intervention. So this is a landmark study that really showed that when we take the time to communicate better, we make a difference. And I think that's really important. It's, again, should be so self-evident. But it's good to remind ourselves, right? I realize I'm preaching to the choir here, but I have nothing else to do but give this talk. So I'm just going to preach to this choir. That's my job. This is Renee Stapleton. It's sort of along the same lines. I'm going to just show some slides on hints of how to have a good family meeting. And if we had more time, if this were a two-hour interaction, we run these communication programs for caregivers. And one of the most potent things is sharing helpful hints. Because each of us has our own way of facilitating communication. And each of us has something that might evoke in the other, in the rest of us, oh, right. I never tried that. I really should do that. This is Renee Stapleton just assure. This is a common one. And I find myself still saying, let's talk about withdrawing care. And of course, it's not care we're withdrawing. We're withdrawing life-sustaining therapies. But if we communicate care, that's what families hear. And so you're not going to care for my loved one anymore. And so very, very simple things. And how you can never assure a family enough that you're the nurses and you do this all the time. And that to the best of our ability, we're going to make your loved one comfortable. Like you can't say that enough to patients. And remind them what it's going to look like and that it may not look pretty. So these kind of reassurances are extremely important. I said this in our small group. I cannot emphasize, and I know you've all had this experience, where you don't remember to have the pre-meeting with the surgeon, the neurosurgeon, or the neurologist. And this just happened to me about six months ago. One of our neurocritical care docs came into the room. And he's wonderful, right? He trained at MGH, he's really great. And he sat down with an eight and a half, I'm not making this up. And he drew an MRI of the brain. He drew the picture for the family and started pointing out where in the brain. And I was astonished. They had no idea what he was talking about. And he was also not looking up. So he was like drawing it really carefully, drew it from start to finish. Well, this is the area that's involved with motor. And this over here is the one that's involved with speech. And it was complete gobbledygook for this family. And they were looking at this picture. And I could see them look at the picture, look up at him, look at the picture. And they were like beyond deer in a headlight. And he finished, and then he walked out. He was done. So I have reminded myself having that experience to say to him when we walk in, so please don't draw any more pictures. And I love him. I mean, Steve's really great. And I say, well, you've got to use more simple language. And he, the first time I did that, he got a little irritated with me. But I feel like that kind of pre-meeting is essential, because otherwise, you wind up wasting your time. And then you end that meeting, and you have to have another one. And you know that it's true with neurologists. You want to say, tell me what you're going to say to this family when they say, there's no chance that he's going to get better. Because if his response or her response to that is going to be, well, there's always a chance, then you don't want to have that family meeting, honestly. And we do that. We just agree not to have the family meeting. And these are simple things. You all know these. But I always introduce myself. I mean, you have to, right? And yet, I still see people who sit down and assume, well, you know I'm the doctor, and you know what kind of doctor I am. And you're one of eight that's been in that room. So I think introductions and all of this stuff, I'm not going to belabor this. This last one, I feel really strongly about this. I am going to talk about the shared decision making in a few minutes. I feel very strongly that I take the time to find out what a family thinks their loved one would want. And then I take it on myself to help and make a suggestion to them. I feel strongly that part of the shared decision model is me helping by making them a recommendation, not bullying them, but saying, well, you know, this is what I think we should do, given what you've told me about your loved one. I don't feel the value of abandoning a family to make a decision. And that's the way it's described, as we know, in the literature right now. If I say, you know, I think your mother is dying, but it's your decision, so you need to let me know what to do, that's just abandonment. And especially if we have so much uncertainty, and we are the medical professionals, and we don't know, there's no way that they're going to know. So to a certain extent, it's integrating the information together and then making a recommendation. My experience with families is if they don't want my advice, they are happy not to take it. And if I have to know for myself that I'm not bullying anyone, I'm just offering, and I think it makes a huge difference. Oh, so this is another. This is one of Ellie's early studies that I love that he published this. So the average duration of the family meeting was 10 minutes. That's a short family meeting. That's like you walk in the door, tell them who you are, and then leave, right? 10 minutes, right? 10 minutes. I wish. 10 minutes, that'd be great, right? We do them all the time. And it's not surprising that the duration of that first meeting, over half of the patients didn't understand diagnosis, prognosis, and intervention. Not surprising in 10 minutes. And sure enough, there was a correlation between the length of the meeting and their ability to understand. So this is a great, just a real great demonstration of why you need to take longer to explain what the prognosis is if you haven't. I think I'm going to skip over this, but these are very, very simple hints from the literature about things for discussing prognosis. The one thing, and I'd love to have it come up, I'm a big fan of I don't believe that anything I say is going to rob patients and families of hope. I think there's a big difference between helplessness and hopelessness. And I'm saying this because I know it's going to engender some discussion, so I want you to ask me later. I believe that we confuse our helplessness and expressing and empathizing with the helplessness of someone facing death, that if we're honest about that, we're going to rob them of hope. Where I think hope is an innate part of the human spirit, and that we can teach, we can communicate joy during the dying process by being honest. I'm going to leave it there and let you sit on that, and then we'll come back to that. Just to remind us, this is a really great study by Doug White at Pittsburgh that showed that there's so much discordance because of our lack of communication skills between what families hear and what we think we're saying. And that leads to the difference, in our opinion, of the level of aggressiveness, the appropriateness of the level of aggressiveness. And you can see here, look at the difference you see between A and C, is the difference in the perception of the physician's prognosis. And on the right, in the dark orange line, is the surrogate's prognostic estimate. So that physician-surrogate discordance, you see that line that goes from A to C. I don't have a pointer. But that's the result of our poor communication skills. And that's very, very common in the literature. This is just a really nice manuscript that demonstrates that even in an institution where they taught communication skills, there's persistence of the discordance between what we think the patient's prognosis is and what the families think the patient's prognosis is. And it's really because of our poor communication skills. So the final thing I want to do to give Teresa some time is talk about shared decision-making briefly. We had a consensus. It was really the only end-of-life consensus conference that we had between the European Society of Intensive Care Medicine and SCCM. And it was back in 2003. And we published it. And these are the key factors in shared decision-making, which we were advocating, was sharing the prognosis, sharing the level of uncertainty with families, and really uncovering what families' preferences are for their loved one. That's the key factors in a shared decision-making. And understanding the balance between what patients want, what the families think patients want, and what we believe is the key to shared decision-making. And this is just to remind you that when you look in the literature and you ask families, who do you want to make the decision, it's always mixed. It's rare that you see a study published. This is just one of them. This is from the US. No, this is from Canada. It's Darren Hyland. Look at this. This is from 2003. When you ask families, who do you want to make the decision, it's almost always a shared decision model. In this case, it's a majority want a shared decision. This is from the Netherlands. Again, the same thing. And this, mind you, this is in Europe, where this decision-making is so parentalistic. And yet, even in the Netherlands, the question was responded that, well, we want to partner together. So shared decision-making, in most of the literature that's been published, is a desire that's expressed by families. Not, I want the doctor to make it, or I want us to make it. And that's including in the US as well. It's almost always a desire for shared decision-making. And I think with that, I'm just going to close with a couple of summary slides. So there is no question, and the data supports this, that good family conferences can improve patient and family outcomes. That unfortunately, we have not taught skills in communication, and we really need to commit. This is a quality summit. And so one of the most important aspects for improving quality in our institution is improving the general communication skills of our providers. Not just about end-of-life care, but just in general. I think it starts with us being able to talk to each other across disciplines. And we know this. If we don't have communication and good teamwork in the intensive care unit, we're not going to have good communication with our families and our patients. This has to start right away. Someone asked, when should you start talking to people about their advanced directives? I always say that we should start early, saying at breakfast table, honey, can you pass the sugar? And by the way, if you get hit by a bus today, and you have a 50% chance of losing all your executive functioning, what would you want? I mean, I'm not really exaggerating. We've had these conversations in my family all the time. I know what my wife would want, and she knows what I would want. And I think the only way to do that is to ask it when you're healthy, not ask it when you're sick, to make sure people know it. Talking about prognosis, and so on, and so forth. Finally, I think shared decision making is the key. So I appreciate you all listening, and I think we're going to save questions till the two of us are done, right? Thanks, everybody. Back on? Did we turn it off? Can you hear me OK? OK, good. All right. So I'm going to get through this kind of quick here. I always put some up there because I do occasionally get travel sponsorship. I did not get any travel sponsorship for this meeting, but I like to disclose that whenever I speak. I'm going to skip through this quickly. I want to stop a minute here because this is really important when you start to think about predictive analytics. So we're kind of shifting gears, right? What in the heck does predictive analytics and predictive modeling have to do with end-of-life discussions? Well, there's this big talk about artificial intelligence and machine learning. And how many people are on Epic or Cerner? How many have tons of best practice alerts and alarms? Yeah, constantly, right? So we have to think about this. And when you really want to build a predictive model, let's say you have access now to a data warehouse and you're going to build a predictive model, these are all the things you have to think of. Do you have a data scientist who knows how to do SQL? Right? Or maybe they use Python. You ever heard of Python? How many people have used Python in the room? Right? These are the type of languages you have to speak in order to extract data out, right? These are in data tables, multiple data tables. So you might need something called a GUID, which is a way that we can take data from one table. We can do MRN when we're doing it within our own health system. But what if I'm using a database that has over 40 intensive care units from across the United States? I have to de-identify that data. And I might have a common key that I use so that when I'm in all these data tables. These are the things you have to understand in order to extract the data. So that's why all of these words are really important. You might be working with, like I did in my research that I did recently, I had over 2 million adult ICU cases that I looked at. And by the time I applied my inclusion exclusion criteria, I had almost a million cases, right? I couldn't do, I couldn't write my code in R, which was a statistical program I used. I couldn't do that on my MacBook Pro. So I had to use a subset of 200,000. And that would take sometimes all night long to run the code. So I had to use a computer at MIT. So these are things you have to think about when you say, I'm going to extract data and develop my own predictive model. And so when I do get to those tables and I find that information, maybe I have my data warehouse or I'm pulling some kind of extract out of my Epic or my Cerner world, this is the kind of data I might find. I might find flow sheet documentation, medication data. But what is that medication data? Is it what was ordered and the time it was ordered? Or is it the time it was administered? These are all things you have to know about these data before you start to develop a model using those data. Do I have diagnosis data? OK, we're going to talk a little bit more about that in a minute. And history and physical and exam data, right? Because that could be what we call unstructured versus structured. So these are things you have to think about. And these are some of the different categories you'd look at. And when you get all of your data, these data that you have together and all these variables together, how are you going to deal with missingness, right? We saw the New Sepsis 3 definitions come out a few years back. And they used, in their first study that they used their big data extract, they had about 30% to 37% that had a Glasgow Coma Scale. And then they had to do imputation. So then you have to think about that. Am I going to do statistical imputation on when I have a missing data element? Which means I might say, I'm going to look at the mean. So if they don't have a Glasgow Coma Scale, I'm going to automatically give them what the mean of the whole population was. And is that actually appropriate for that particular missing data? So you can see this is extremely scientific. And yes, I did cringe some when Judy was talking, because I think too analytical sometimes. But these are the things you have to think about. And I'm going to talk about structured and unstructured data. I like this infograph, because it kind of talks about all of the data that we capture and where it comes from there at the top in the green. And then it dumps into this big condenser that I call big data. And then about 20% goes up here, because the other 80% is unstructured, right? Our physicians that are still using dictation, instead of using structured data elements within a structured note, because it takes too long. And I get that. We want them to see their patients. And we want them to get home in a decent hour and have quality of life with their family. So there's a bunch of unstructured data that I would have to use something called natural language processing to extract that data. How many people have heard of natural language processing? And how many people have it at their institution? They're actually using it. So a couple people. I mean, that's the only way to get at those data, is to invest in natural language processing if you want to use certain variables that can only be found in an unstructured note. So we have to think about all that as we build a predictive model. But some people have been doing it. Look, here's a study here, published in 2010, where they actually validated a predictive model for identifying Medicare beneficiaries who needed end-of-life care planning. So it can be done, right? So I don't want to discount the fact that we could do it. And it might be important to do. But I want you to understand what it takes to get there. You can't just get Epic and build a BPA, in my opinion. Because I think all we're doing is overloading our clinicians with a bunch of information that may not be useful. Now, they did show that they had some improvement in outcomes. There's another study here looking at predictive modeling at end-of-life outcomes using EHR. So they took the data out of the electronic health record. And they were able to develop some fine-grained models to help predict the outcome at end-of-life. So this is important. It's moving in the right direction. And there's another one that I think is really kind of cool, which was the improving palliative care with deep learning. So as we start to get more into this age of a predictive model. So predictive model, you could take a whole bunch of variables and use logistic regression and build a model, right? But when you start to get into artificial intelligence and machine learning, you're talking about a computerized system that actually learns information and is able to then predict things based on what it learns. So this is much more advanced than just the development of a predictive model. So predictive analytics, is it possible today? Yes. We have a lot of data types. The volume is expanding. We have faster, less expensive computers that we can do these predictions on. We have more intuitive software. We have a lot of companies that are doing really great work in AI that we can learn from and start to incorporate. But there are some pitfalls, right? You need data scientists. How many people actually have a data scientist that they work with? OK, we've got a few more. We're starting to get there. But most of our senior leadership does not understand why you would want to invest in data scientists as you develop this. So you have to learn to have those conversations. Interpretation of data are dependent on structure and content of the data. So you really need to understand the source of that data. And this is where I think you really need to have nurses as part of your research team. Because if you're going to be using data that's in the electronic health record and you don't know how nurses document those data elements, you're going to be in trouble as you begin to interpret. So you need a full multidisciplinary team for that. And then implementation is challenging and expensive. This is a good resource for looking at some analytics adoption models. So I wanted to put that there for you. But I want to blow through some of this. This is what I really want to get to at the core, right? We already know about alarm fatigue. How many have had programs at their hospital to reduce alarms, right? Alarm noise. So we know about that. And our patients don't like it. That infogram over on the other side is talking about how patients don't like to be interrupted with these alarms either. So we understand alarm fatigue. So how come we are developing alert fatigue? How come we don't get it that we can't have people answering, so being interrupted in their workflow to answer another electronic alert? We don't seem to understand that. And so one of the reasons we don't seem to understand that is because we don't honor the five rights of developing an alert or an alarm. We need to have it go to the right person, right? We have to determine who that right person is. If I'm a novice alert nurse and you send me an alert that requires more complex knowledge to decipher or interpret, what am I going to do with that information? I might not even know what it means, right? So maybe I'm not the right person. I also need to interrupt that person in the right time in their workflow. So if it's something that can wait, it can be a BPA and EPIC when it can wait till I open up the chart, right? But if it's a sepsis alert, why is it going off and it doesn't go off until I open the patient's chart? That doesn't make any sense. Now it might be two hours old and I've missed my mark to give my antibiotics or whatever other treatment. So now I need the right time. I need the right person. I need the right information, right? Don't tell me something I already know. Tell me something I need to know. And it needs to come through the right channel, OK? So it needs to come through the right channel. Do I want to interrupt a physician provider on a mobile device? Or how about one I love so much? At my institution, believe it or not, we have Philips monitors. But we will alert a nurse to send them a text message that tells them there has been an alarm going off on the monitor, but they can't see the monitor on that device. They have to drop what they're doing, because they're not using all the technology. I have an assessment by Philips, by the way, coming. So I see them nodding my nurse friends in the back. But they will then get a text message. They have to drop what they're doing on the telemetry for, go to the monitor and see, or go to the patient's room. Oh, that's a silly interruption. So we have to have the right channel and in the right format. This is actually, do I have time to show this or not? We're getting kind of close to time. So I'm going to show this briefly just because I think it's important if it'll play. Nope, it's not going to play. No sound there. I sound guy there. Danger of computerization that we've seen in terms of its impact on patient safety. In the book, I tell a story of a kid at my place who we gave a 39-fold overdose of a common antibiotic to. And the initial error was just a simple glitch of the doctor didn't realize the screen was on milligrams per kilogram rather than milligrams. It happens. But what was really remarkable was an alert fired to the doctor and said that this is an overdose. The doctor clicked out of it. An alert fired to the pharmacist. The pharmacist clicked out of it. And you might, as an outsider, say, oh, how could they be so careless to do that until you realize that in a month at UCSF, at my institution, the doctors get 30,000 alerts. The pharmacists get 160,000 alerts. And those are just the computerized pop-up boxes. In a month, in our 70 intensive care units, the computers that record your heart rate and your blood pressure and your oxygen, there's an audible alert every six months. If you want to advance that, you can go ahead How would you know to be scared about a patient if the alerts are going off? So how many people has this happened, that you've blown through a BPA or alert? We've all done it, right? We've all done it. Because we are desensitized. So now we have all these alerts that we've built. So do we need or want another alert to tell us to have a discussion when we haven't even figured out how to have the discussion yet, right? Don't we need to learn how to have the discussion? Now, if your organization has already perfected having that family meeting and discussion, and you think developing a predictive model or using one is going to help, then I say go for it. But if you haven't even figured out fundamentally how to have that proper discussion, I don't think you need to go there. That's my opinion. And I love this kind of research. I love developing predictive models. We're running out of time, so I'm not going to go here. But this is someone you can look at for the human-centered design. How many people have human factor scientists in their organization? Couple hands. Does everybody know what that means, human factor scientists, what they do? Those are the kind of people who design in the aerospace industry. They use human factor scientists, right? So they look at, how will somebody use the system I'm designing? How will that person interact with whatever it is I design? And we're not doing that enough in health care. So again, if you're not going to employ the right tools, and you're not going to use an approach that we get the right information to the right person at the right time and the right workflow and the right method and in the right manner, then this is probably you don't want to move to predictive analytics yet, right? If I were going to develop a predictive analytics model, I know the analytical side. I'd get somebody like Mitchell to help me figure out what might be the components I'd want to put in that model. So we need that, but we'd have to work as a team to make sure that we build a model that we actually use appropriately. So we can skip through this one if you want. Just pass that one up. So again, really looking at that clinical human factors perspectives are really, really important as we think about how we're going to use AI, machine learning, or any type of predictive models. Now, Epic is, we're an Epic shop. They will tell you they have these new, great, wonderful predictive models that you can build yourself, of course. And I would say there's a lot more to it than just turning on a predictive model for anything, including when we're having end-of-life discussions. So I think we can take some questions now, because I want to make sure we have time for that. And it's already 4.13. And then I'm actually starting the next talk, so I'll just transition in. But if you have questions now for Mitchell or I, this would be the time. So I can give a couple minutes to that. Any questions? They're getting tired, I think. It's the end of the day. Yeah, I always avoid percentages. Same thing with how long after we extubate. My stock answer is, you know, every time I say it's a day, it's a week. Every time I say it's a week, it's an hour. And I say it over and over again, because it's true. And so I never use percentages. I will say, I will do the opposite sometimes, which is, I'll say, even though I'm not 100% certain, I'll say your loved one has no chance of leaving this hospital alive. Or I'll say, I really don't think I'm going to get your loved one off this mechanical ventilator. The most common scenario is the oncologist will say, oh, no, we can use chemotherapy. As long as Levy can get him off the vent and they're no longer critically ill, and then I have to say, there's no chance we're going to be able to do that. Chronic lunger who's got cancer, I mean. And so I'm comfortable, even if in my mind there's a 1% likelihood, when I'm faced with those kind of odds, I will communicate certainty to those folks, but not use percentages. Because I don't know how to do that. I don't know, 5%? I don't even know how to do that. As you said earlier, that you give your perspective, how do you ensure that you're not? Are there phrases that you tend to use to say, we still are going to have a relationship, even if we disagree? So I'm going to say, I think you should pick A. I would recommend A. But if you pick me up, we're still going to do that. Yeah, that's a great question, David. I am very careful about that, because I also can tell, as we all can, when families really don't want my opinion. So I only offer my opinion when families are asking for it. I don't say, well, here's what I would do, when the family is saying, I believe in miracles, and they're presenting a wall to me. So I definitely wait until I feel invited to offer my opinion. Now, most families and patients will ask, what would you do? What do you think I should do? So that's when I offer it. So I've been doing critical care now quite as long as you, but I'm not sure I'm feeling confident. And I find sometimes it's helpful to avoid getting stuck on, is it 1%? Is it 5%? When you're talking about, am I going to die or not? Am I going to die? Really pivot and shift the conversation and talk about quality of life. So I can't be 100% sure that your loved one is not going to survive their ICU stay, but I'm 100% sure that they're not going to get back to running a corporation or playing golf every Thursday or whatever it was that's important to them. Can you kind of speak to that? It would be great for both of us. As a physician, I think that's a great question. And often, I look to hear from family members, so how independent was your father? And often, the family will say, oh, he would never want to be in a nursing home. And that gives me a great opportunity to say, well, I don't know whether he's going to survive, but I know with 100% certainty, he's going to have a very long recovery that's going to include an unknown long period of time in a skilled nursing facility. And often, families will go, oh, he definitely wouldn't want that. No, I would agree with that. I think from the nursing perspective, those kind of discussions, many times as nurses, they have the 10-minute discussion, and they leave our physician partners sometimes. And we're the ones really getting to the root of what would your loved one want and what would they not want. And so I think we can continue those discussions when they're starting to grapple with that decision on what to do. Because then we take the decision out of their hands. And then it sounds like if your loved one could make this decision, this is the decision they would make. And then it becomes that person's decision, and it takes that burden off. So I think we'll transition. David Murray and I are going to be talking. David's from Emory. And we're going to do a talk as soon as we get the slides up there. I'll start going on that one. We're going to talk about how to engage large health systems in quality improvement and best practices across diverse hospitals. But again, as I mentioned earlier, even within the four walls of one hospital, you can have siloed care and differences in how you provide care. So there's take-home messages for that. So I'm going to talk about two different organizations. I'm currently at UMass, but I came from Sutter Health about five and a half years ago. I was recruited to go out to UMass. And I was a nurse. I'm going to skip to that because I already talked about it. These are our objectives. I'm going to talk personally about two different health system approaches and identify some different approaches and models of care delivery that have been associated with improved clinical and financial outcomes, because that's important, and discuss the importance of culture, structure, and strategy as an integral part of how care is delivered. And so I want to talk a little bit about disruptive innovation. Has anybody heard of disruptive innovation in the room? So we'll talk a little bit about that, because I don't think we do that enough in health care. We're really good at holding on to the way things are being done. And you guys, I'm probably preaching to the choir because you're innovators. You're probably on the forefront of trying to push forward change in your organization. But we see this a lot. But Clayton Christensen, there's a website with him. And this is one of the quotes from him, is that breaking down these old business models is always going to require leaders to follow their instinct. There will always be pervasive reasons not to take a risk. But if you only do what worked in the past, you will wake up one day and find that you've been passed by. And I think we're passed by in health care. Other industries have way passed us up. We latch on to some technologies. Does anybody remember the first cell phones? My brother-in-law ran a towing company, a big towing company in Sacramento. And he had that big white one. And then later, he got that flip one. We thought, well, that's cool. He can talk on the phone wherever he's at. And that was pretty cool. And then I remember at Sutter, they gave me a BlackBerry. We called it the CrackBerry, because everybody got addicted to it. And they could find you anywhere you went. That is disruptive innovation. How many people have a home landline phone now? Any takers? A couple, right? Not very many people have a landline anymore. That is a great example of disruptive innovation. And has anybody seen this video? This is great. So let's play this. I'm going to only play the first couple minutes. And then when I lift my hand up or wave at you, turn it off, because we're not going to do the whole thing. OK, so go ahead and lift up the box. You have four minutes to dial that phone number. That's it? That's it. That's it. With that phone. Wait, so we have to push the button? Kyle, move your butt over. I don't want your butt in the video. My fingers don't work. There you go. It's all the way from here. Yeah. Seven. Oh no, zero's all the way over. How do I restart? How do you do that again? He said, how do you restart it? Seven? Seven. How many in the room have used a rotary phone before? There's seven. OK, see, we're a little bit older generation. But my children, I'd have to show them how to use it, right? They wouldn't know how to do it. Dude, that makes sense. Wait. We can go ahead and move to the next one. But you can watch the whole video. It's available on YouTube. It's quite amazing. Now, how many people have their grandchildren or children have iPads? That's what this video is, right? We'll just watch a little quick minute of how fast he navigates through this iPad. Hi, this is Bridger. Say, hi, Bridger. Hey, Bridger. He is playing with the iPad. And he just turned two. So why don't you show us? First of all, show us, draw us a picture. Nope. Yeah, draw us a picture, Bridger. You guys four. No, draw us a picture first. OK. Sweet. I'm excited. Oh. This? Yeah. Oh, a bright one? Do a new one. Oh, a bright one? Yeah. Why don't you draw a squiggly line? You're so smart, Bridger. You can move to the next slide for me. Draw something. So if we have our kids and grandkids using iPads, how come we use these? How many organizations used to have pagers? Yeah. I've been asked by a group. I'm one of the clinical people, because I'm actually the director of clinical operations and innovation for telehealth. So I sit down with all the hardware guys. And one of the big things they've been asked to do is fix the infrastructure for our unified communication systems. And every time they go to the clinicians, and they say, what do you want? A better paging system. I can't deliver you a better paging system. I'm sorry. The infrastructure for paging is crumbling very quickly. That's kind of like chiseling in stone. It's not going to happen. I can deliver you a better messaging system. So changing the vocabulary. So that's what I've been working with my telecom people and my group in the IS and IT side is, just because they ask you for a better paging system, that's not what you have to give them. You can tell them, I think what you're asking for is a better messaging system. And I can work on that for you. And so that's what we're doing right now, is kind of going through an overhaul. How do we get rid of these? How do we pry them out of their fingertips and give them technologies that will work better for them and give them better information? And so why do I tell you all this? Because I'm going to talk to you about two health systems that did something very innovative and very disruptive in how they provided care in order to improve. I'm sorry. Another one by Clayton Christensen. A disruptive innovation is a technologically simple innovation in the form of a product, service, or business model that takes root in a tier of the market that is unattractive. Pagers are unattractive, right? We know that. So how many have seen this? How many likes the New England Journal of Medicine Catalyst? If you're not getting this newsletter, get it. It is so cool. I love it. So this came out actually about a year ago, a year ago today. And it's looking at sources of health care innovation. And here are the sectors in health care that need disruptive innovation. You can see hospitals and health systems. This was like, I think, 519 administrative level people were interviewed in health systems. And 65% said we need disruptive innovation in our hospitals. And we need it also in our IT departments, our primary care, and so on. And then when you look at the top ranking organizations, provider organizations, vendors and supplies, and payers and insurers, so we know we need some more disruptive innovation. Well, in 2003, I helped, as a staff nurse, start the first tele-ICU on the West Coast and the second in the nation in a very large hospital system. It was Sutter Health. And so when we did that, it was pretty disruptive. I can tell you it was a top-down decision from our CEO at the time, Van Johnson, one of my favorite people in the world. But by the way, this is how we have it. 430 ICU beds across 500 square miles. And we ended up with two tele-ICUs, one in Sacramento and one in San Francisco. I started as a staff nurse, became the manager of the Sacramento hub, then the director, and then director of both of those hubs. And one of the things we decided as nurses to focus on, in about 2004, we saw this as a problem, was sepsis. And a lot of people weren't looking at sepsis programs. There were some at that time, but not everybody. So it wasn't really mainstream. And we started screening patients from the tele-ICU hub. But what we met when we did this was a lot of opposition and skepticism. People were like, you're not even really a nurse anymore because you're not at the bedside, or you're only there half-time, so you can't possibly know what you're talking about. We had lack of leadership support at the local level. So we still had people who were like, why the heck is this EICU thing here anyway? Just get out of my room. People thought it was a flavor of the Muntz syndrome. Don't worry, this sepsis thing will go away. There was a lot of role confusion. Who's supposed to do what? Should somebody screen at the bedside or not for sepsis? And then if you screened a patient, what should you do with that information? And a lack of accountability on multiple levels. So we were really here. We were in this innovator category for our health system. There were other places across the country doing this. But we were in the innovator category of about 2.5%. By the way, I added a category. I figured since I got my PhD now, I can add on theoretical framework. So I added one. So that's that one there of non-adopters. Because how many people know a non-adopter? So come on, we can't just say they're laggards because some people never adopt. And so this is where we were when we started this process as nurses. And then we said, well, look, we've got to do something about this because we have a huge problem with sepsis. So we created the sepsis surveillance and data collection process from the remote telemedicine intensive care unit. And Windows is checking for a solution to a problem. So we might lose this. So I'll just tell the story. I'll keep going. There went PowerPoint. So we developed the sepsis screening process. And what was key there is we developed a data collection process. So we could report on, that was the next thing we did, is reported on incidence of severe sepsis in each of the units that we cover from an ICU perspective. And I think we were covering 13, 160 beds at that time. We reported on incidence of sepsis. And we reported on compliance to the bundle. We developed this form with partnering with our IS department in an infopath, because that's all we had back then. But it fed into a Cognos database where we could then extract data out. And then go back and say, hey, in this particular ICU, you had 10 patients because we had very small, some of them. Or you had 30 patients who screened positive for severe sepsis this month. And of those patients that screened positive, this is how many who got their antibiotics on time. This is how many got their lactate drawn and what the lactate was. This is how many got blood cultures before antibiotics, and so on. And we began to report those data back. And when we started to do that, let me go back there real quick. So we reported the incidence and the mortality. Oh, I probably skipped that. The only nurses who were highly trained and skilled, they had to have three to five years of critical care experience. Our average years of experience was 15 at the time we were doing this. So these are experienced nurses with a high level of situational awareness who could basically, with a camera, go into a room looking at the medical record and information that they have in front of them and be able to tell what's happening with the patient. Those are the type of nurses we used. And then from that, we ended up creating a system-wide sepsis initiative by 2008. This is what a tele-ICU center kind of looks like. This is actual one. It's not the one I worked at. But you can see that there are all different people working. They're usually staffed with intensivist providers, critical care nurses who, again, are experienced. Some have pharmacists. We have a pharmacist in our EICU at night at UMass, and nurse practitioners and PAs. Some have clinical nurse specialists. So these are the types of people who can work and do work as teams in these units. This was some of the data that we started to publish the results of some small quality improvement studies. At this time, I was a bachelor's in nursing. And I didn't really know how to conduct research. So I relied a lot on some of my partners to help me put together some different studies. And then we published this. And this really helped, because we began to get recognition outside of our health system. Because again, we were at that early adopter, innovator phase, where people were very resistant to sepsis. And we really hadn't gotten good movement within the organization. But by 2008, we now have early adopters. And we have the early majority. So now we're getting some momentum. At that point, one of my colleagues, actually, Mary Ann Daly, who worked on the ABCDEF bundle and some of the work she did, she came on, came from Providence, and started to work with me on sepsis. We ended up getting a grant from the Betty Irene Moore Foundation for the Sac Sierra region. And she took the ball and ran with it and did an amazing job of doing additional focused education at the bedside and putting together more momentum at the bedside. And so when we look at what happened over time in that, we like to use static graphs at Sutter Health. They are a community health organization. They don't have a ton of researchers. We're doing more and more research there now. But you can see our mortality in 2004 was above 50% for severe sepsis patients. We had hospitals, some of our small rural hospitals, where mortality was 60% to 80%. You did not want to go there if you had severe sepsis. But we were able, through education first, to start to see some downward ticks. And then when we started our screening and data collection process and reporting process, we saw that come down even further. And then after this became a system-wide initiative, we continued to see downward trends until we were down at about 20%, 25%. You can see there by the end of 2010. And then of the work that Mary Ann Daly and others did, this was where the mortality went after 2008. When they took the ball and ran with it, we continued to tick down. Our baseline in 2008 was 35%. And by 2013, we had reached there. And I know there are. Last year, one of my colleagues presented some data on the Sac Sierra region. And they were down in many hospitals to 12.5% mortality. So they've done an amazing job. But it took an army. And part of that army was a very disruptive, innovative group of nurses sitting remotely, starting to screen for sepsis and really shining the light on that we've got a problem. And so then by 2013, we were there. So I want to quickly go through the UMass experience and then turn it over to David. So at UMass, they used a little bit of a different approach. They have over 150 ICU beds that they monitor throughout Central Mass. And they focused on healthy work environments from the get-go. So Dr. Richard Irwin, who is the editor-in-chief of CHESS, and he's our chief of critical care, he was president of CHESS at the time. He co-wrote this article with Kathleen McCauley, who was the president of the AACN. And they talked about the importance of having the right work environment. And that was kind of foundational to how UMass started their program. And within their structure, they built right into their structure that for critical care was defined as caring for critically ill patients, regardless of the patient's location in the system, through a system-wide virtual department. That was their tele-ICU that uses a collaborative, interdisciplinary patient-centered approach. And they really aligned under a very strong culture that was around strong leadership who communicated a powerful vision. They aligned systems and structures under the umbrella of the Critical Care Operations Committee, which would be likened to the quality committees that some of you have spoke to me about, that you have. This quality committee, or critical care operations, was seven different ICUs, including trauma, neurotrauma. We have a transplant service, medical surgical populations. So quite a few different types of players in there. The teams developed the systems and structures and policies of how they provide care. And they wove in the role of the tele-ICU within all of those policies that they developed. And anti-change agents were managed through accountability. I can't tell you how important that is. If you've ever met Dr. Irwin, he's the biggest small man I've ever met in my life. I mean, nobody wants to get hauled in his office, because he will hold you accountable. He does not shy from accountability, so you need that accountability. They published this study first. I actually came to UMass in 2013. This was the first published study they did that showed that they reduced complications, mortality, length of stay. And they had better adherence to their best practice measures. I want to talk more about this one real quick. And I'm going to skip to the data, because we don't have a ton of time. But this is kind of the schema of the study. So in 2004, we went back and looked at financial data and outcomes data. We started a clinical study. That was that first study you saw in 2005, because they didn't implement the tele-ICU until 2006. They had already instituted the Critical Care Operations Committee, and they had standardized multiple care projects going on for best practice. And they studied how did they do in outcomes, mortality, length of stay, and adherence to their own best practice metrics. And they did improve. But when they added the EICU, they improved even more. And so then they decided in 2007 to start using the EICU, where they had the intensivist 24-7, to be like a logistics center. And they managed flow in and out of the seven ICUs. So if you need an ICU bed today, and at that time at UMass, you had to call the tele-ICU intensivist, who then determined, based on the acuity of the patient, and the need, and the availability of beds, which bed you went to, including whether or not you got the last trauma bed. Who can say they did that? I thought that was pretty cool. And we studied that. And what we found out over time, and these are all the different projects we worked on as a group in which we clearly defined the role. What's the role of the intensivist, the nurse practitioner, and the PA in the EICU today? And what's the role of the people at the bedside? And these are all the different interventions that we did over time, the clinical study logistic, and all the different projects that we worked on over time. And I'll quickly show you what happened. So you can see here the case volume in 2004 was pretty flat to six. It bumped up in 2007. And by the time we got that logistics center going in 2008, it jumped up again. So we saw an increase in volume. We had better length of stay. We opened up the beds. And then we filled those beds. And so that translated to improved case revenue. So you can see here the per case costs versus the case revenue and how that on that line just continues to skyrocket up. And what that really translated to is about $50 million to the good after we paid for the tele-ICU installation because that's expensive to install it over a 10-year period. So we recouped all our costs. This is in the paper. You can review that paper to get more detail. But you can see the volume going up at each stage of the intervention and the total annual costs versus the margin and the revenue. And so basically, I'm going to skip this because I want to give some more time. But basically, flow is more efficient when a non-conflicted intensivist is empowered to move cases. But again, that was done because we had a structure in which we gave them a role and everybody bought into that. And so the right structure, the right culture, and the right strategy are very important. I'll skip through this because I want to give David some time to talk. But I love this quote. The telephone did not come into existence from persistent improvements of the postcard. So we got to look at what are we doing, what tools are we using, and how are we doing it if we want to make change happen. I think for the sake of time, I better let David come up. I'm not going to. What time do we have to be done at 5? Yeah. Because I want to have time for questions. So I'm going to let this. You can go ahead and play this while we switch over. But go ahead and play that. Here's to the crazy ones, the misfits, the rebels, the troublemakers, the round pegs in the square holes, the ones who see things differently. They're not fond of rules and they have no respect for the status quo. You can quote them, disagree with them, glorify or vilify them about the only thing you can't do is ignore them because they change things. They push the human race forward. While some may see them as the crazy ones, we see genius. Because the people who are crazy enough to think they can change the world are the ones who do.

predictive modeling

disruptive innovation

healthcare

communication skills

end-of-life care

ICU

data analytics

data scientists

missing data

interpretation