English
Empowering your Emergency Department with AI-driven Clinical Decision Support
Speaker:
Dr. Phillip Levy
Professor of Emergency Medicine, Wayne State University
Fellow, American College of Emergency Physicians
Description:
Artificial Intelligence in healthcare and diagnostics is continuing to make headlines worldwide. From image to symptom recognition, it’s clear that AI is making quite the impact on the healthcare experience for both providers and patients.
In this interview, Dr. Phillip Levy, Professor of Emergency Medicine at Wayne State University and Fellow of the American College of Emergency Physicians discusses how Clinical Decision Support tools powered by AI and machine learning can provide valuable guidance to clinicians in the ER.
| Time Code | Speaker | Transcript |
| 00:00:21:27 - 00:00:48:05 | Narrator | Artificial intelligence and machine learning models leverage the massive computing power that exists today to analyze large quantities of data quickly. These models can identify various patterns in clinical data that may not be readily recognized in a time pressured environment. These models can be leveraged to provide the clinician with additional insights to support a clinician's decision making process, acting as an advisor. |
| 00:00:49:08 - 00:01:14:15 | Narrator | As these models learn over time, they further improve the quality of the insights they provide and deliver ever greater value. The best designed CDS solutions integrate seamlessly with the clinician's existing workflow, do not interrupt the workflow in any way and require no additional clicks. Let's hear thoughts on artificial intelligence and machine learning Clinical Decision Support from Dr. Philip Levy. |
| 00:01:15:05 - 00:01:19:12 | Narrator | Dr. Levy is a professor of emergency medicine at Wayne State University. |
| 00:01:21:11 - 00:01:57:09 | Dr. Philip Levy | I'm Philip Levy, professor of Emergency Medicine at Wayne State University, where I also serve as Associate Vice President for Translational Science. In addition to this, I was Vice Chair of the American Heart Association and the American College of Cardiology Chest Pain guidelines that were recently published. It means my job gets easier because I don't have to assimilate all the different pieces of information in order to make the best decision for the patient that I'm caring for. |
| 00:01:57:25 - 00:02:24:09 | Dr. Philip Levy | Artificial intelligence allows you to pick up hidden signals in the data that would be hard for the human brain to do otherwise. Different points of information. Lab test results. What artificial intelligence allows you to do is take large datasets and try to make the most sense of all the small data points to get the biggest piece of information. |
| 00:02:27:29 - 00:02:47:15 | Dr. Philip Levy | So artificial intelligence and machine learning, clinical decision support has applicability in a lot of different health care settings. But in circumstances where time is of the essence and you have to make quick decisions and you don't have a lot of time to try to sort out all the information you're taking in, such as the emergency department, it can be particularly beneficial. |
| 00:02:47:25 - 00:03:22:04 | Dr. Philip Levy | In that setting, we're often managing multiple patients at the same time, and we have to make decisions on each one of them in a timely manner so that they can get to the next stage of care or get to go home. Machine learning algorithms and clinical decision support that results from that does all the heavy lifting for us and makes it a lot easier to get to the right answer. |
| 00:03:22:08 - 00:03:42:26 | Dr. Philip Levy | Conditions like chest pain and sepsis are particularly amenable to clinical decision support using artificial intelligence and machine learning because they're very hard to predict who has the condition that we're most interested in knowing about. So everybody who comes to the emergency department complaining of chest pain thinks they're having a heart attack, but only about 5% or so actually are. |
| 00:03:43:11 - 00:04:07:28 | Dr. Philip Levy | And the ability to quickly identify that 5% and separate the 95% that are not having that problem is critical. Not just for that individual patients care, but for all the other patients who are in the emergency department at the same time and all the other chest pain patients who we're seeing, because there's a lot of them. We have to figure out who gets the next level of testing, who needs to go to a cardiac intensive care unit, who may need to go to the cardiac cath lab? |
| 00:04:08:14 - 00:04:25:06 | Dr. Philip Levy | The better and quicker we can narrow down that decision making, the better off we are. And the same holds true for sepsis. There's a lot of people who come to the emergency department with a high fever. Maybe they have pneumonia, maybe they have a urinary tract infection, but they're not always septic. It could be sick, but not septic. |
| 00:04:25:06 - 00:05:11:29 | Dr. Philip Levy | Sepsis is a very specific thing. So the sooner you get on the therapies to address those problems that are causing the chest pain, like a coronary artery obstruction or sepsis, like an overwhelming infection, the sooner you can get someone on the path to cure. So the current state of clinical decision support solutions is really centered on automated calculations of scores like the heart score or EDACS. Scores that have been developed from relatively limited patient cohorts that leave way too many people in an intermediate risk range. |
| 00:05:12:14 - 00:05:34:22 | Dr. Philip Levy | And the clinical decision support automation is really just supplanting what I would do at the bedside, which is take the components of the scoring calculator. We're not talking complicated mathematics, but the scores themselves are part of the problem with the clinical decision support solutions that exist, and they're just too limited and they put too many people into categories that don't help me as a clinician get to the right answer. |
| 00:05:35:08 - 00:06:14:29 | Dr. Philip Levy | We're definitely in need of new tools that can take in additional pieces of information to refine that diagnostic differential and help me get to the right answer quicker and more efficiently. So clinical decision support can really help us implement the guidelines. Again, in a chaotic environment like the emergency department, if you have automated clinical decision support, not only can it help you understand when to repeat testing, what to do with pieces of information, but it can also help incorporate things like prior testing. |
| 00:06:15:04 - 00:06:58:27 | Dr. Philip Levy | You don't want to have to delve deep into the record, go back a year or so to just start to make sense of information, and if you could pull it directly in, that would help us be much more in line with the guidelines. This is going to be a big question going forward as clinical decision support, especially with AI and machine learning components incorporated, may make clinicians feel like they're being taken out of the equation because everything is just being converted to an algorithm or a computer program. |
| 00:06:58:27 - 00:07:36:07 | Dr. Philip Levy | It's going to be a bit of a challenge, but not insurmountable. It's one of those things where we make progress all the time in medicine, and I think the ability to trust new forms of data and new pieces of information that help us make decisions better is going to be critical. And it's going to be important for the medical community to have an open ear and an open mind when it comes to this type of stuff. |
| 00:07:36:07 - 00:07:55:11 | Dr. Philip Levy | All too often in medicine, patients are used to being told what's going to happen next, and rarely are they asked what would they like to happen next. When you have clinical decision support, you have something more concrete you can hang your hat on. It's not me coming to the bedside saying, I think this because that's what I think. |
| 00:07:55:24 - 00:08:42:11 | Dr. Philip Levy | I can come to the bedside and say I think this and believe it because we have a computerized algorithm that gave me a very precise percentage of what your risk is. I think we'll both be more comfortable with that decision because they'll understand that it's not just me thinking what I think without any other support. It's having an entirety of research background and an effort that has gone in to come up with what are the most important variables to include in a predictive model. Emergency departments that use artificial intelligence and machine learning |
| 00:08:42:12 - 00:09:11:15 | Dr. Philip Levy | clinical decision support solutions are on the cutting edge, and patients should feel rest assured when they go to those facilities that they're going to get care that is guideline based - that is safe and accurate- because it's supported by years of research that has gone into the clinical decision support tools making its derivation. They should know that when a clinician makes a decision, it's not just what's in their brain, the doctor's brain, that the decision is being made by. |
| 00:09:11:25 - 00:10:13:07 | Dr. Philip Levy | It's also with all of that other information that comes in through the clinical decision support tool. From an operational perspective, incorporating artificial intelligence and machine learning clinical decision support enhances throughput. It makes it easier for clinicians to assimilate information and arrive at accurate decisions quickly. And that's critical, especially for chest pain and conditions like that, such as sepsis, which require very early, very accurate decision making because time is of the essence. From a hospital system perspective, artificial intelligence and machine learning clinical decision supports provide a lot of value. |
| 00:10:14:09 - 00:10:44:07 | Dr. Philip Levy | Oftentimes, chest pain patients are treated with a lot of a lot of reservation, a lot of concern to get to the right diagnosis, but also not to miss the diagnosis. And so what does that lead to? It leads to a lot of people who are kept in the emergency department, may be placed in observation status, or admitted to the hospital for additional testing that they may not need. Uncertainty leads to cautious decision making. |
| 00:10:44:07 - 00:11:03:02 | Dr. Philip Levy | Clinical decision support can minimize that uncertainty and help us make more accurate and correct decisions. |
