Good afternoon. Like Heather said, I'm Maya Dewan. I'm the Division Director of Critical Care Medicine at Cincinnati Children's. And I'm a clinical informaticist and implementation scientist. So I was given the title, Quality Improvement Processes Will Prevent All In-Hospital Cardiac Arrests. But I sort of switched the title because that seemed a little bold for 15 minutes to Quality Improvement Processes Will Prevent All Preventable, Avoidable, Unanticipated Cardiac Arrests. So really thinking about those ones that we have the opportunity to intervene on. In order to get to that first topic, how to prevent all in-hospital cardiac arrests is going to require a lot of change in terms of how we treat diseases like sepsis, ARDS, et cetera. So I think this is a good place to start and something that we could tackle now. So I don't have any relevant conflicts of interest. My work is supported by a K08 from AHRQ. So I have two objectives today. The first objective is to evaluate the impact of a systems-based approach on reducing pediatric cardiac arrest. And the second is to understand how to identify areas for ongoing improvement in cardiac arrest prevention. So like Heather mentioned, CPR is performed in about 1 in 1,300 pediatric hospitalizations. And less than 50% of these patients survive to hospital discharge. And the majority of these events are occurring in ICUs. So taking the time to really focus on prevention within the ICU is where you're going to get the most bang for your buck. And why is prediction important? It's important because we have clear data that show that in-hospital cardiac arrest is usually preceded by early warning signs of deterioration that can be both recognized and treated by in-hospital staff. No longer do we look at in-hospital cardiac arrest as something that's just routine. This is an avoidable harm that we can really make a difference on. And we know that improvement in recognition, situation awareness, and training can impact this. Today I'm going to focus on situation awareness because I feel like that is a key component that can be shifted at any institution right away. So what is situation awareness? So there are three levels to situation awareness. The first is, what do you perceive? What do you gather the information? So a great example is you have a child who's presented with a high heart rate, diarrhea, and the parent's concerned that they just don't have the same activity level that they used to. The second is comprehension, sort of recognizing and understanding what this might mean, that this child with high heart rate and diarrhea is at risk for having significant dehydration and anticipating that this patient could develop hypovolemic shock. The way that I remember situation awareness is I think of it like the stages of medical training. So I think of stage one like our medical students. When they present a patient on rounds, we expect them to be able to just gather all the data and bring it to us. And then we think of level two more like our residents. How do they pull that information together? How do they give us an assessment statement? And we think of stage three like our ICU fellows. What do they anticipate? What is this patient at risk for? What might happen in an hour, two hours, four hours? And so thinking about how we build systems that really support this, similar to how we support our medical trainees. This is important because we know that a failure to recognize or a lack of situation awareness is actually the most common factor related to serious safety events in children's hospitals. And so how do we build these systems? So I'm going to walk you through three steps. I'm going to walk you through improving and standardizing your identification of high-risk patients, making sure that you're sharing a mental model and creating a medication plan. So let's start and talk about improving and standardizing identification. There's not a clear answer for what will work for you because your patients may look different. This is a system that we originally built at CHOP when I was a fellow and have since replicated at Cincinnati that uses passive alerting. So you just hover your icon, the hover to discover of figuring out what could be wrong with this patient. It doesn't interrupt your workflow. It doesn't stop you from putting in an order. It doesn't impact your direct care of the patient. But it says, hey, there's some things here that you might want to look at for this particular patient. These criteria are criteria that were expert consensus criteria and then that we use to validate within a clinical model. But each center is going to be different. The PC4 criteria for cardiac ICUs is obviously a very different set of criteria. If you are a center like we are at Cincinnati with very complex airway procedures, those airway patients are at very high risk of cardiac arrest for us. If you are a center that does unusual transplantations or orthopedic procedures, you're going to have a different risk profile. And so this isn't a slam dunk, right? This isn't a perfect system. But it's one to say, hey, let's standardize the way we look at these patients and let's think about what could be a risk factor for them. And so then we studied this. And we said, hey, if we take it from paper, the electronic health record, how does it do? And it's OK, right? It gives us continuous screening. It allows us to not have to use a checklist to do things every time. But it really impacts our performance. And so thinking of these types of tools, this is not the area under the curve of 0.95 that you want to say, hey, this kid's going to have a cardiac arrest tomorrow. I need to put them on ECMO today. That's not the goal, right? And that's not where we are right now with predictive modeling in the pediatric ICU. Where we are is, who's at risk? What are our most common risk factors? And how are we going to build a team approach around to make them better? We then have recently updated this tool to really include these additional criteria. Because while automated criteria are important, there are certain things that the electronic health record doesn't know, right? So the electronic health record doesn't know who I'm going to intubate this afternoon for a procedure who has an extraordinarily difficult airway. It doesn't know which patients maybe had an out-of-hospital cardiac arrest if I'm not a center that's good at updating my problem list. It doesn't know who the really experienced clinician on the team or the really experienced bedside nurse is worried about. And so adding additional features that are easy and simple, again, creates that environment of a shared situation awareness. And because we know in medicine in general, but specifically in pediatric critical care, that the majority of our clinical decision support tools are not used and not studied, we've actually done some usability testing with this. So what you're looking at this graph is a simple usability test called the State Usability Scale, in which we took our respiratory therapists, some nurses, and some providers, and said, hey, I want you to use this tool, and I want you to provide me feedback. The way that this score uses is anything greater than a 70 is considered good, and anything greater than an 80 is considered excellent. Each of these dots on the graph is an individual. And so you'll see the majority of people rated it at least good, and the overall usability was 87.5. There are lots of things within the electronic health record that we can't change. We can't change exactly where it appears on the screen, or exactly how the person interacts with it. But you can study the usability and really make sure that this is something that your team likes, that answers their questions and doesn't interrupt their workflow. And what was the impact of this? So the impact for us was, first, are we predicting more events? And this chart jumps around because one of the things that we're still not great at, even from a shared situation awareness perspective, is neurologic deterioration. And so almost all of these drops that you see are recurrent, either sudden seizures or other neurologic deterioration that was missed. But when we started this work, we were less than a flip of a coin, so we were predicting events 39% of the time. And our goal is to be around 70%. We're not going to be able to predict every event right now. That's going to take additional work that we're doing around machine learning predictive modeling. But 70% seemed good enough for us. So then we said, OK, so we've got a good idea of how we're sort of standardizing the information. Now, how are we going to share that with our teams? How are we going to get our team ready? And how do we share that mental model with everyone who's caring for the patient? So I get the question all the time, do you take care of patients at Cincinnati Children's, or do you just huddle? We do take care of patients. I have some colleagues. You can ask them after the coffee if they take care of patients too. They do. But building redundancy in your system is an important way to make sure things happen, especially when you are at high acuity and high census times. People will miss things. They will be down in MRI. They will be in the ED seeing a patient. And so redundancy is important, right? It's not wasted time. It's important to measure, but you need it. So we have sort of a combination of huddles that we do at Cincinnati. The first is our twice daily safety huddles. We characteristically do this after rounds. I know some centers like to do it first thing in the morning. And we talk about the state of the unit, the patient flow, who are our watchers, or our high risk patients. When a patient is labeled as high risk, we do a bedside huddle, but it's very informal. It's a conversation. It's the fellow with the nurse pulling in the respiratory therapist, talking with the attending on rounds to say, hey, who is this kid? Why are they high risk? And filling out their sign. And then in our old ICU, we had these pods in which there was about five to eight beds around a central workstation. And so we came up with this idea of huddles because we have so many new staff to be able to say, hey, what's the name of the person working next door to me? So when I need help, I don't yell, hey, you. But I actually know someone's name, so they can come and help me. And to share who are the sick patients around. So it's not just the bedside nurse that knows, but when she goes to lunch, and the person who's covering knows that plan as well. We now call our different areas of the unit neighborhoods. So we now refer to these as block parties. So we study our shared situation awareness. And we do this with bedside surveys in which we ask our nurses, our respiratory therapists, and our resident or our APP who's directly caring for the patient, is your patient a watcher? Why are they a watcher, and what's your plan? And people are always like, there's no fellows or attendings on here. I know, right? Because that's not the person who's immediately responding. The person who is immediately responding is responding in those first two to three minutes that are often the life-saving minutes is your bedside care team. And so it's important that they know what the plan is, and so that's what we study. All right, so let's talk a little bit about these mitigation plans. So when we originally started, we worked with our family advisory council as well as our team to come up with what we call action response plans. And I wanted to call them mitigation plans, but our family representative was like, no one knows what that is, Maya. So we went with action response plans, and we built a plan that was both clear to the family and used the appropriate medical language for the team, but also was something that was accessible to families. And we said, OK, what are your risks? What do you need to do? And what are you going to have ready at the bedside? So here's an example of a filled out one. So you have a patient who's a high-risk intubation. You have your clear criteria for when you want to get ready and clear criteria of what you want at the bedside. But there were gaps. So this was a project done by one of our summer undergraduate students who said, OK, but does everybody know? Does the nurse know? Does the respiratory therapist? Does the resident know? At the same amount, right? And you can see that there's a clear gap each time the resident was significantly below the bedside nurse or the RT. And while most people knew, is your patient high-risk or not, when it came down to, what's your actual plan, you can see that it fell off pretty significantly. And this was one of those things where, in medicine, trying to figure out a way to disseminate information in an easy and understandable manner is not something that we're really trained to do. And so we actually partnered with a design collaborative and said, can you help us? Can you help us find out a way that has an improved human factors and an improved visualization to make it easier for people? We went through a design approach using, first, they sort of walked around with us and mapped it out, figured out what was the big concerns, talked about how we can improve the system, and then helped us to collaborate as a team. They then took time to diagram it. And what I saw as a physician as a fairly easy process was actually a pretty complicated process, right? So we have patients who come in, sometimes they're admitted to the PICU and they're sick right away. Sometimes they come in and they're not sick and they deteriorate while they're there. Sometimes they get better and then get worse again. And so having someone who was outside of medicine to really walk the process and figure out just how complicated it was was really key for us. We then worked together and built some different boards. This was a fun one that we posted up and would have people write on. People hated it. It was terrible. But we went through it and we prototyped it. Again, something we don't do in medicine, right? We just make something and we stick it out and we're like, this is what we're doing. Do it, right? But this act of multiple prototypes, mostly made from stuff that we found at Home Depot or Office Depot, was what we used to try to move through this. And these are our fancy signs that we now have. And so what we ended up going with, after all that, is signs that aligned with our high-risk criteria from our automated tool and that have clear setups to sort of help in terms of what comes first, what comes second. And we modeled these both with testing with our bedside nurses as well as with our trainees to make sure that we were building them together. We also made sure that we addressed latent safety threats. So for those of you who know me and know I don't have a ton of patients, the same conversation over and over again in M&M would get frustrating. Every single time we'd activate massive transfusion protocol and we would stand there and be like, where's the blood? Blood's not here yet. It's because you have to send someone to get it, right? But it was the same conversation every time. There was a delay in the blood. Well, we didn't send somebody. So when we wrote the signs, we literally wrote out what to do. It's an emergency. You often have young and untrained staff. What do you want them to do? You want them to call the blood bank and say, activate massive transfusion protocol. You want to collect and order the labs. And then this is my favorite, send someone to the blood bank to pick up the blood, right? It's right there for you in the moment of the high anxiety. You don't have to think about it. It's all laid out. And then we don't have to talk about it anymore at M&M. And then when we built our new critical care building, we figured out, where are we going to put these things? Because we used to put them on the door. And you couldn't miss it. And then we had this fancy new building and these fancy new doors. And people were like, you can't put stuff on the door, Maya. And so while we put these to the side, all of a sudden, your eye wasn't drawn there. There's an emergency. You walk in the room, and you've totally missed that this patient is high risk. And so what we did instead is we added these Epic door signs that change color. So within 15 seconds, if a patient is labeled a watcher with an Epic, the sign changes from this sort of tealish color to orange. And right away, there's a visualization as you walk in that room that this patient is different. Our families notice it. Our consultants notice it. And it's right there for you to say, OK, now, let me find the sign, let me see the plan, and let me figure out what's going on. And so what's the impact of this? So this is our impact up through December 2020. And then I have some additional data. But if you look at the top, it's CPR events per patient days. And if you look at the bottom, it's mortality. And what you see in terms of cardiac arrest, you see a 50% decrease. I just want you to just take a moment and let's pretend that I had a drug that I was talking about today, or I had a device that I was selling you in the exhibit hall that decreased your cardiac arrest in the ICU by 50%. What would you pay for that? What would your hospital invest in that? And so when people say to you, this is a lot of work, this is a lot of time, this is a lot of things we have to do, I think thinking about what that impact is and really measuring it is huge. Because we don't have medications that do this, and we don't have devices that do this. And this is where we are now. So as we've moved to our new critical care building, we've maintained that same. And I think my goal really is that zero that we're going for. But that's going to take some technology advances, some vital sign-based prediction, and some additional work that we're going to need to do. Now, I don't pretend like this is easy. This is challenging. And so you measure what's hard about it, and you figure out ways to build systems that support it. So first, we don't spend all of our time in huddles. We time our huddles. We spend less than half a minute per patient. And that's because we have scripted huddles, and so it's really quick to move through the low acuity patients and then spend more time on the high acuity patients. And then we follow something called the number needed to evaluate or the number needed to alert, which is how many patients are watchers for every one cardiac arrest. And we sit around 16. We've decided with our current implementation strategy, less than 20 is probably appropriate. But this is obviously not going to be an appropriate number for me to say to you, put this patient on ECMO. They're going to have a cardiac arrest. You're not going to put 16 patients on ECMO for every one cardiac arrest. So this is why the burden and what we ask people to do is relevant to what this is. As we get this number down, there might be more intensive interventions that we ask our staff to do. But for now, we feel that this is an appropriate amount. So as I sort of wrap up, I feel like there's lots of things that are fancy that we can talk about as we talk about cardiac arrest prevention. But I think something that everybody can do right now today is to think about how they build situation awareness within their own unit, how they standardize their identification, even if that's paper, checklists, three criteria that you use each time, how you share that mental model, and not just with the people in the room, but with the people at the bedside, and how you create a mitigation plan that incorporates the patient, the family, the bedside team, and making sure you're ready for the deterioration that may happen. Thank you, and we'll take questions at the end.

quality improvement processes

in-hospital cardiac arrests

early warning signs

situation awareness

patient identification