SCCM Resource Library-AI for Augmenting Nurse Recognition of Patient Decline

Video Transcription

Hi, I'm Cheryl Hiddleston. I'm the director of the Emory EICU program at Emory Healthcare in Atlanta, Georgia. I'm also adjunct faculty with the Department of  Biomedical Informatics at Emory University School of Medicine. I want to talk to a little bit today about AI for augmenting nurse recognition of patient decline.  I probably have an unusual viewpoint on this. I have no disclosures or relationships to report. So I think it's probably pretty common knowledge that development of AI,  machine learning, predictive models, etc, etc, in the field of sepsis has increased dramatically in the last 10 years.  Also, there are a lot of publications out that help to guide people through developing these predictive models and how they report, how they work, how they measure outcomes,  evaluate performance, report missing data, etc, etc. This is 10 keys to prediction models. The one thing that is left out of these 10 keys is the human element,  human factors in developing these tools.  What we find with this is that adoption of these machine-based learning tools is not always consistent. It's not consistent from staff to staff.  It's not consistent from different roles. You can see here that for the initial rollout of the tool, which is use 1.0,  there was a decent provider response in terms of how helpful the tool was. As you go on to 2.0, the later version, and people have used it for a longer period of time,  the provider responses indicating that they felt the tool was helpful decreased pretty dramatically.  Meaning initially, they seem to see value in the tool and that changes over time.  When looking through all of this data, determining that adoption is also slow.  I think part of the problem is that data scientists develop these tools and these models. For continuous predictive analytics monitoring to be useful,  clinicians must engage with the data in a way that makes sense to their clinical workflow in the context of a learning health system.  Nurses and physicians and APPs and respiratory therapists are not data scientists. Well, not all of them. There are a few that are, but for the most part, they're not.  If you look at adoption and acceptance of these models, there are primary nurse concerns that come to the forefront. One is understanding the science behind the algorithm.  They don't just want someone to put a tool in front of them and not help them understand why. The sub-themes were, am I the first one to use it?  Is there a lack of standard evidence? Is it moving towards an alarm or is it just bringing our attention? The next issue becomes trust in the data inputs.  Where are they coming from? What noise is involved? Can I believe the data inputs? The third is integration with the electronic medical record.  And this is important because nurses feel like, I don't want one more alarm or alert that's put in front of me that is outside the medical record  because then I'm looking in various places to try and determine what someone wants me to see with my patient. It's not necessarily critical to them to provide care in the moment,  but in order to understand the alerting system, the analytic model, they would like to know more about it. Also, optimize clinical pathways.  So, create something that is more proactive instead of reactive to proactive and then refine the model so they're not trying to decipher  what is important about whatever the alerting number, tool, whatever comes across.  I don't know. I am by training a critical care nurse. I spent 20 years at the bedside in the ICU and things have changed. I haven't been at the bedside now for 12 years.  And nurses now at the bedside, they may not have a big screen in front of them like the picture with Star Trek, but they are trying to decipher information  from all of these different areas of the EMRs and they're tired, especially now since COVID. They have many different data sources coming at them all the time.  One more alert, one more app to look at, one more task to try and incorporate into their workflow.  There's more stress, more burnout, and less staff, but more work that has to be done. And at the center of all of that is a critically ill patient  that sometimes gets lost up in all of that.  So, in thinking about this, who should be the person that adjudicates all of this data from the disparate systems trying to give information about an acuity score,  a sepsis score, any sort of other warning system, early alert warning system, something to let nurses at the bedside know there's a problem with the patient.  When they're looking at all these various alerting scores, then there can be some confusion about which one do I trust the most,  which one is the most pertinent, which one gives me the most information. And that's where I feel like the team you see on this screen becomes very important.  So, at Emory, we have an EICU program. It's a tele-ICU program. We have a group of nurse experts. They have a minimum of 10 years' bedside experience in ICUs,  and they're all CCRN certified. These nurses are trained to access different clinical decision support tools, analytic tools, and models.  And what they hone in on is what is different. They are trained to recognize changes in data trends and to recognize when a patient is deteriorating.  They're trained to adjudicate the findings and go to the source of truth to determine whether or not it's an actionable event  or if it is some sort of artifact in the system or other, you know, artifact in the system or other issue that pulls their attention away.  They're trained also to communicate these changes with the bedside team. So, instead of having the bedside team look at three or four alerts  from different systems, our EICU team is doing that, and they're adjudicating the results to determine the validity. And also, our staff is not physically in the environment,  which is emotion-driven. I know when I was a nurse at the bedside and a patient was arresting, I focused on what was in front of me.  Our staff is not in that emotion-driven environment, and they can take a broader look at everything that's going on  and make decisions from a more, I don't want to say detached, but it is kind of a detached viewpoint. My nurses, I didn't even really know this existed,  and my nurses coined the role that they felt like they were data detectives. They are trained to look at everything that's happened to the patient,  to evaluate all of the alerts that they're seeing, to adjudicate them and determine whether or not they're valid and what the next step should be.  Should they escalate this to a physician? Should they continue to monitor the patient? Should they speak to the staff about it because it's something that they need to intervene on?  They take all of that into consideration in order to be able to deliver the care that's appropriate for the patient. And then once they've evaluated it, they move beyond that  and they communicate with the appropriate area and the appropriate person to make sure that the patient has the care that they need.  Where the innovation will come in is when AI and human engineering is merged. Just creating tools and pushing them out to clinicians at the bedside  is, in my opinion, not going to be the best way to utilize all of the tools that are being brought to the forefront that will be very useful in patient care.  We all recognize that patient care is changing. We have less people, and the expectation is that they continue to produce quality outcomes.  They continue to meet all of the guidelines that are put down from CMS, JACO, all of the different accreditation bodies.  And there's an expectation that quality outcomes remain very high, and you're still doing this with less people. We are deriving data from different AI models,  directing where that data goes, who sees it, who adjudicates it, and how it is implemented is very important.  We need to find ways for our teams to collaborate to create better care. If we have one team focused on the different AI solutions that are pushed out  and our systems decide they want to implement, and then they work with the bedside team to let them know what they've seen,  it takes part of the burden off the bedside team that is decreasing in size every day. And I think creating that collaboration is going to be very important moving forward.  Thank you for giving me time today. I appreciate it.

Video Summary

In this video, Cheryl Hiddleston, the director of the Emory EICU program, discusses the challenges of using AI to augment nurse recognition of patient decline. She highlights the importance of considering the human element in developing these tools and emphasizes the need for nurses to understand the science behind the algorithms. Nurses also express concerns about the trustworthiness of data inputs and the integration of AI tools with the electronic medical record. Hiddleston suggests that a collaborative approach between nurse experts and AI teams can help alleviate the burden on bedside nurses and improve patient care.

Asset Subtitle

Quality and Patient Safety, Professional Development and Education, 2022

Asset Caption

This session will explore the potential for artificial intelligence (AI) to augment critical care. Speakers will introduce relevant concepts using terms and examples that are familiar to critical care professionals, describe state-of-the-art applications of AI in critical illness, identify barriers to real-time clinical implementation of these applications, and propose desiderata for ideal algorithms.

Meta Tag

Content Type Presentation

Knowledge Area Quality and Patient Safety

Knowledge Area Professional Development and Education

Knowledge Level Intermediate

Knowledge Level Advanced

Membership Level Select

Tag Evidence Based Medicine

Tag Innovation

Year 2022

Keywords

Cheryl Hiddleston

Emory EICU program

AI

nurse recognition

patient decline

	Society of Critical Care Medicine 500 Midway Drive Mount Prospect, IL 60056 USA Phone: +1 847 827-6888 Fax: +1 847 439-7226 Email: support@sccm.org	Contact Us About SCCM Newsroom Advertising & Sponsorship DONATE	MySCCM LearnICU Patients & Families Surviving Sepsis Campaign Critical Care Societies Collaborative	GET OUR NEWSLETTER
© Society of Critical Care Medicine. All rights reserved. \| Privacy Statement \| Terms & Conditions The Society of Critical Care Medicine, SCCM, and Critical Care Congress are registered trademarks of the Society of Critical Care Medicine.