Hi guys. First thing I want to do, because I was asked by my partner who runs this program with me at Emory, was to take a picture of the audience so he could see how many people were here on the last day of the conference at one of the last sessions. You guys waved to him. His name is Tim Buckman. You may have heard of him before. Okay, now he'll be happy. So I want to talk a little bit about clinical decision support, AI, all of the above, and multidisciplinary teams that use them. I do just click. Multidisciplinary considerations when implementing CDI. I'm not going to go through all this. So that is me right there. It is Picken's Nose. It's across the Georgia-North Carolina border in North Carolina. And I tend to want to explore and venture out to see what may or may not be there. That's one of my trips. I have no disclosures to make. So learning objectives. I'm going to talk about the rise of clinical decision support. How should these tools be interpreted? Who should be responsible for the adjudication and use? And let everyone know that I think we need to think outside the box. So medical predictive model development over the last 10 years has increased exponentially. There are studies published every single day around the world, different companies developing these tools, looking at these tools, et cetera. And there are even a lot of publications you can find that advise you how to develop the tools, how to get published when you're trying to share your information and what you did. But you don't see the engineering part in any of this advice. It talks about describe how the outcome labels are developed, conduct internal and external validation, talk about the data, measure model discrimination, calibration, et cetera, et cetera, et cetera. But again, no human engineering here. Adoption is the key to utilization. So if you look at different tools and how the nurses and physicians and APPs and pharmacists, all of the bedside multidisciplinary team use those tools, it's important to know whether or not they're accepting them, whether or not they're embracing them, and more importantly, whether or not they're actually using them. You can see in this one that there was a pretty high percentage of providers that thought the tool was useful initially, but when you went back later on, a year later, it dropped significantly. Adoption is slow, especially with nurses, and let's be honest here, the people who primarily use this information, who implement the changes, who take care of the patients, deliver the care, act on whatever model you have that you're using, are the nurses at the bedside. They're there 24-7. They're the ones you need to buy into this. And the nurses feel that they need to understand the science behind the algorithm, trust the data inputs, integrate, they would like for it to integrate with the EMR so they're not looking at multiple displays with numbers and information and acuity scores in multiple different places because they are already doing so much at the bedside. We need to optimize clinical pathways so they can see how the models are proactive and how they're refined and how they will work for them. So Christopher Horvath did a talk Saturday morning and I got him to send me this slide. He talked to CHAT-GPT before he did his presentation Saturday and said, I'm going to do a presentation at the Society of Critical Care Medicine. And CHAT-GPT gave him six different things that he thought were important for him to talk about. But the one I want to bring your attention to is the last one, which is also, it's important to note that AI should not be considered as a replacement of human judgment, but rather as a tool to assist and enhance the capabilities of the medical professionals. And that's really what I am here to talk to you about. Everyone talks about the data, the technology, the inputs, how they gather it, but it's really important to understand, in the end, without human judgment, those tools don't mean anything and they can't be used. So when you have a multidisciplinary team, it's really important to have highly defined roles with the new tool that's introduced, recognize the data trends, recognizing deterioration, adjudicating the findings, communicate the findings with other clinicians, and act on the information, all of that is really important. But I want you to come up on Pickens Nose with me and look at this a different way. Who should be the people that are looking at all of these tools, all the numbers on the screen, and determine what should be done? I have, I run a tele-ICU program. My nurses run 12 to 15 applications all day long as they're working. A lot of them have data feeds, they have clinical decision support, they have different acuity scores, and my nurses are already experts in critical care, so they're trained, they all have their CCRN. But what I train them to do is to recognize deterioration and to understand when it is actionable and when it is not. So they use multiple scores and multiple systems to help them do this, but they look at it, they adjudicate it. So when they call the team at the bedside, the team at the bedside knows it's not just a number flashing up on the board, it's someone who's experienced that recognizes something is really going wrong so they can act on it. One of my nurses about seven years ago said, you know, we're data detectives, that's what we are, we look at everything, we figure it out, we look at the chart, we look at the patient, we need to call ourselves data detectives. So I came up with this picture and it was because of my nurses, nothing else, and I think it defines who they are. So my proposal is that you have a remote team looking at all of these tools so every nurse, every physician, every provider at the bedside is not constantly looking at numbers and tools. I heard in another lecture that it's important for physicians to also be data scientists. Maybe it is, maybe it's not. And I think it's kind of a lot to ask in addition to everything else they do and all of the training they already have. I think it'd be nice to have a course on data science, on clinical decision support, on predictive models so they have some idea of what they're looking at and what they're using. But I think it's more important to have a small team that is located remotely so they're not in the middle of the drama, they don't have 600 other things to do at that moment, and this is their focus and it's what they're trained to do. So my team, I said they all have a minimum of five years in critical care, they all have their CCRN, they have access to many different analytics, CDS tools and models, they're trained to recognize decompensation and changes in data trends, they're trained to adjudicate the findings, they look in the EMR, they go in the room, they look at the patient. They're trained to communicate the changes with the bedside in a respectful manner. And I don't know how many of you have been exposed to a tele-ICU either as a provider at the bedside or working in one, but 99% of what this team does is being able to communicate respectfully and consciously with the staff at the bedside. It's huge because if they don't, the staff at the bedside doesn't listen to them and they don't act appropriately on the information. They are, I just said, they're not physically in the emotion-driven environment so they can take a step back and look at the information that's being given to them and then they can, as I said, adjudicate and make a decision on how to move forward. So I think AI plus human engineering is what will give us innovation and it already does. You can't have one without the other. You have to use them both. We have to recognize that patient care is changing. All of us know that, especially over the last three years. Stepping away from the bedside and doing the most you can is the most important thing we did during COVID because everyone couldn't be at the bedside. We have less people. Everyone knows the nursing shortage has increased exponentially and it's not going to all of a sudden resolve in the next two years and we're going to be flush with nurses. It won't happen. But we're still requiring them to produce the same high-quality outcomes with their patients and they need help doing that. It allows the bedside team to provide the life-saving, hands-on care that they have to, but it helps take some of the burden of this type of assessment adjudication away from them. Deriving and directing data-driven AI models where they should be with someone who has the bandwidth to take them in and assess them and make a determination. And find ways for teams across the continuum, whether it's at the bedside or in a remote location, to collaborate and create better care for the patients. Before COVID happened, we had a lot of pushback from nurses at the bedside and there are several people in this room who've worked with tele-ICU and they understand what I'm talking about. But during the pandemic, we became a lifeline. What's going on in my room? What does my patient look like? What do you see happening? So that collaboration is what creates the best care for the patients and the high-quality outcomes that everyone is looking for. I think that's all I have. That is all I have. Thank you.

clinical decision support

multidisciplinary teams

nurses

AI tools

remote team