After a patient with a life-threatening heart attack waited hours in an emergency department before anyone recognized the crisis, Dr. Maya Yiadom began asking a different question: Are we measuring what matters most? In this conversation with Dr. Matthew Strehlow, she explores the gap between healthcare metrics and patient outcomes, and how data, analytics, and AI can help identify missed opportunities, improve clinical decision-making, and build a safer, more responsive healthcare system.
After a patient with a life-threatening heart attack waited hours in an emergency department before anyone recognized the crisis, Dr. Maya Yiadom began asking a different question: Are we measuring what matters most? In this conversation with Dr. Matthew Strehlow, she explores the gap between healthcare metrics and patient outcomes, and how data, analytics, and AI can help identify missed opportunities, improve clinical decision-making, and build a safer, more responsive healthcare system.
Dr. Matthew Strehlow: Hello, everyone, and welcome back to the Stanford Emergency Medicine podcast. I'm your host, Dr. Matthew Strehlow, vice chair of Emergency Medicine at Stanford. Today, we're talking with Dr. Maya Yiadom about how we decide what's worth measuring in the emergency department and what happens when the metrics we rely on don't actually reflect how care unfolds in real life.
Emergency medicine measures a lot: door-to-doc time, patient experience, time to ECG. We have dashboards and benchmarks everywhere. But those of us that have worked a busy shift know that there are moments that matter deeply that never show up in those dashboards. Counting is not the same as understanding. That's what today's conversation is about. Dr. Maya Yiadom is an associate professor of emergency medicine at Stanford and leads the Stanford Emergency Care Health Services Research Data Coordinating Center.
Her NIH-funded research examines how emergency departments measure care and how clinical data and predictive modeling can be used to build learning systems that improve the care we provide. Dr. Yiadam, Maya, welcome.
Dr. Maya Yiadom: Why thank you, Dr. Strehlow. I appreciate the invitation to be here.
Dr. Matthew Strehlow: Well, it's great to have you here.
You've shared before that an early patient encounter stayed with you and shaped how you think about care delivery. Can you start by telling us that story and the impact it had on you?
Dr. Maya Yiadom: It's so interesting. It was a single patient encounter, as you had said, and, to this day, this patient has no idea how their experience-has changed my career trajectory. But I was a very excited emergency physician coming out of training, where I really had this MacGyver perspective, drop me anywhere and I can do emergency medicine. So I was moonlighting for the Indian Health Services. I had my main role that was in an academic medical center very close to where I grew up. I was also moonlighting in smaller hospitals in the region that were more urban, more rural. And in one of these locations, I was working on a day where I was a new doc. The health system wasn't familiar to me. And, the hospital was overcrowded, and the emergency department was full and backed up with many boarders, so we had a lot of people in the waiting room.
There was a particular woman who was found to be disruptive to the triage staff who were overseeing the waiting room. And, they had come to me in the main ED and said, "There's a patient who needs some Ativan. Can you give me an order for Ativan?" And I said, "What does she need it for?" And they said, "She just needs to calm down."
So, I said, "I don't know the patient," and quite frankly, I didn't know that team. So I said, "I'd like to see her first. Can you just do your best to get her back here?" Two hours later, the patient got back into the room. It was a 60-plus-year-old female, of Hispanic ethnicity who had been telling the staff that she was having difficulty breathing, shortness of breath, and she was very nervous and anxious about it.
The shortness of breath, as you know, is one of the two cardinal symptoms of a heart attack, chest pain being the cardinal symptom, shortness of breath a close second, and the cardinal mis-case is a woman who is non-white, and also not a primary English language speaker who reports shortness of breath.
This staff, they were all practicing for a very long period of time in their careers. It wasn't a lack of experience but there was something heuristically and environmentally and even maybe even compassion exhaustion that coalesced in a moment where this woman was missed.
She had a heart attack, and not just any heart attack. She was having the most severe form of a heart attack, a STEMI, ST elevation myocardial infarction, and had been in our emergency department for two and a half hours before we noticed it. So, it was really the first time I had realized that we're all trying to get door to EKG. That's the metric that has been on the books, if you will, for over 40 years as our goal for getting the electrocardiogram to identify the worst case of a STEMI. It's a diagnosis made with that tool alone. And we work really, really hard as docs. We're trained to read the EKGs, but there's a whole operation in the front of the emergency department that really has to get that EKG to bring it to us for us to make the determination.
So, I learned that it really wasn't about the physician being prepared. All the training I'd done, all the hard work in high school to get A's, in med school to be a good doc- just didn't matter. The system failed that patient, and there was nothing that I could do to recover her heart at that point.
I had always been really interested in how systems of care align and are set up to deliver the right care for every patient. So oftentimes, there's this difference between measuring a system and measuring situational occurrences as they will play out for any given patient who has an indication for something that we have ready to take care of their problem.
And being able to look at those two things similarly but differently has been really the center of my career.
Dr. Matthew Strehlow: That's a story we can all relate to, right? And that sense of frustration in the moment where you know that the system-- But I think we all own that failure, or at least that moment where we didn't do the best we could for that patient. And we all carry a similar case. I think a lot of us do. You mentioned that we measured door-to-EKG, and we do these things, and we have a lot of operational dashboards to try to figure out what's going on in the system that is causing these kind of failures, or making it not run as much or as efficiently as we want. What are we failing to measure right now consistently?
Dr. Maya Yiadom: It's an exciting time to be in analytics and in metrics for hospital performance. I think we're often measuring our general process. We're measuring our systems, and for that, you want things like an average. On average, how do we do? So, you take let's say, 100 patients who have had an experience with a time measure, and you take the average of that, which gives you a general sense as to how we do. That lets you know if your numbers are bad, your system is bad, but that means that you're failing a lot of patients. Where I have found there to be a lot of power is in measuring non-compliance for individuals. So yes, the system operates at a certain level, and there's a benchmark, and we meet that benchmark, that measure, X amount of times, or on average or median.
That's just your 50th percentile, so half the patients are above, and half the patients are below. I, as a doctor, and you as a doctor, are more interested in the 50% of patients who are below, but our measures don't tell us their story. So, I've been really interested in how do you measure whatever the metric is, and at an individual patient level, when do we miss the mark for a patient?
And then who are those patients who we missed the mark for, and how are they different than the ones we're able to get it right for? Because there's a story that's there about how our system is not wired to capture the presentations, the needs, the stories, the situations of the patients who fall off the mark. You’re going to have a curve, and that curve is never bell-shaped. Sometimes it is, but it's often skewed. It's going to be skewed in the direction of the system working relatively well, and that number for where the majority of cases are is going to be what you get with your systems measures.
But when you're focusing on the patients, there are patients that are in a really long tail, and those patients are not going to be seen in a measure of the system. Really understanding what's happening in those tails can give you so much information of where there's an opportunity to really do better.
And often there's something systematically happening that's causing that tail to be persistent and can really tighten up that curve. And that's been the principal focus area for a lot of the operational interventions. Stanford's a very nimble place and very innovative; we've been able to really ideate and implement to really be able to see, who's in that tail and how do we address their needs so that we get the same outcome as we do for the majority by really addressing this subgroup need for the minority.
And there are lots of cases of this that we can talk more about.
Dr. Matthew Strehlow: So, you're talking about that tail on the maybe where the system's not working as well as it should and I think we think about that tail a lot, but you're right, our dashboards don't generally show it unless it's all or nothing, right? You either have to 100% compliance or it's look at the average which is obviously a problem. It makes me think of that book, "The Happiness Advantage,” where one of the things that that book talks about is measuring the times that things are great and what was happening around that person or in this case-what's happening in the the other side of that tail where somebody came in, their ECG was done in one minute, they were up in the cath lab in 15. Do we look at those patients, those outliers that are actually on the positive end of the spectrum?
Dr. Maya Yiadom: There, there has been a body of work, and this is going back in time, that really studied high-performing facilities and said, "What are they doing that's different than lower performing facilities, or even moderately performing facilities?"
And really, used that upper group, if you will, to figure out what does best practice look like, and publish those recommendations, and they became part of guidelines. One of the things that's tricky is often resources are a factor in all of that.
One example is from the work that I do with heart attacks and detection in the emergency department. A best practice is to have a dedicated space, so when you identify a patient needing an electrocardiogram because you're worried about them potentially having a heart attack, you have a space they can go right to.You don't have to wait to move somebody or find a chair or find a room or there's no stretcher, let's go get one from the other pod in the emergency department, you're waiting to bring that over. That's lost time.
So, if you have a dedicated space, that's considered to be the best practice. If you look at the square footage of the EDs that have best practice and the square footage of the low performers, there's a big difference.
The low performers don't have space for a dedicated space, and as a result, that ends up being an operational constraint that leads to them having low metrics. And so really trying to figure out how do you create space when you don't have space ends up being the challenge to get a low performer to become a high performer.
So there is, I think, a lot of opportunity to look at the happy place and, figure out what makes it happen, and sometimes you have to modify exactly what the happy people did to bring happiness in places where maybe it's missing right now, in addition to the other way around.
Dr. Matthew Strehlow: Space is not space is not space.
I'm constantly in this refrain, right? Space outside in a tent is not the same as a nicely designed space where we spent 20 years making the room correct and designed just the way we want.
And I do feel like in emergency medicine, sometimes we're so desperate for space in the setting of all the boarding crisis that we'll take any space, but then the expectation is that we'll just perform at the top level as if we were in our actual space and you know, I don't see the surgeons operating in the parking lot much, right? But maybe, maybe in the future.
So you're talking about how there's these differences across institutions, and one of the things that I have always just been so amazed by is how you connect all these different institutions together to form data collaborating centers and integrated networks of data.
Tell us a little bit about why that's so important.
Dr. Maya Yiadom: If you can only see what you're doing in your shop and your ED and your hospital and your hospital system, you have no way of knowing what's possible or what should be done.
And if you highlight your shortcomings, you're now labeling your health system, your ED, your hospital as having problems, and nobody wants that. It's not really good to do for, medical legal reasons, for marketing reasons, for patient care confidence.
So how do you do research on the problems, the opposite of the happy place, and be really honest about what we're not doing and what we're falling short of in order to find opportunities to get better?
And it really required doing multi-centered research when you're looking at system metrics and gaps in patient population care within individual facilities. Being able to have five, six, seven, 15 hospitals who are perceived already as being high performers, and have good reputations gives the work quality, and if they have problems not specific to any one site, it really, speaks to the whole practice as in having an opportunity for change.
It's much safer. Nobody's labeled as being a low performer. Nobody has data reported that could then be turned around and used to undermine the hospital's confidence or in a medical legal case. And so that idea of multi-centered research allows everybody to be really honest. And what we have done is also created reports for sites, their operations teams, that are confidential that they can use internally for process improvement.
So there's been this dualistic approach. But in short, we have really found the use of data that is de-identified across centers created a safe space to have honest conversations about where the practice needed improvement that didn't really put any one institution or ED at risk.
Dr. Matthew Strehlow: You're co-horting institutions that are already have, a strong reputation. So, it's not just institutions that maybe have more resources pointing at institutions without resources and saying, "Why can't you do what we're doing?" Is that, is that a correct representation of what you're saying?
Dr. Maya Yiadom: It is, and I will say there's another level to it too, because the flip side is that you can look at several high performers, they're going to be more equipped and have more resources and be bigger and have scale and be able to level load better than smaller places that maybe don't have all those characteristics.
We've had to be really careful about crafting a universal problem, but testing it in places that are already perceived to be doing really, really well and showing that they have a problem too, and then magnifying the problem, exponentially for all...
I shouldn't say exponentially, but magnifying the problem to scale for what everybody is likely struggling with and saying that the problem is at least as bad as this, so it's probably worse in some places. And that's been our approach.
Dr. Matthew Strehlow: So when you're getting these networks together and you're-
getting more data and more information into the system, is that information actually building a more rich picture of individual systems and care practices, or are we just getting larger pictures?
Dr. Maya Yiadom: That's a great question. I will say they're larger pictures. But what I have tried to hone in on universal problems.
So, it's one picture. You have a picture in your backyard, that's your backyard. But if it's your whole neighborhood, now we have a conversation that's maybe ripe for government or a community organization to step in. And it's been that frame where we're not looking at just getting a big data set.
In fact, a lot of our data sets, the particular items we ask for are actually fairly narrow. We just ask for it in lots of places to demonstrate that there is a universal problem that is worthy of community engagement to solve, as opposed to having everybody trying to solve it in their living room and their dining room as they're looking at their backyard.
So it's been more that perspective. I will say there's this tug and pull once you get into this analytics work, where big is considered to be better. Larger is considered to be higher quality.
And I'd say there is a certain level of bigness that actually is non-informative, that it just almost can give you paralysis of just looking at a lot of detail. Where I think there's a lot more strength is actually knowing what are we looking at across all of these places exactly, and what do we need to know across all these places exactly in order to have insight, and getting really granular at what those details are and getting as much information about that across more sites and not getting lost in bigger data sets.
So, you know, a lot of groups have said, "We have huge registries of heart attack patients."
My question was ,"Do you have the emergency department phase of care?"
"We do. We have all the information before they're admitted to the hospital from the emergency department."
Well, that's great. That's three to eight hours of processing and thinking time and testing to get that data. Our work is at the front door. Do you have any data sets that reflect what's happening to these patients when they walk in the front door? Crickets essentially for that.
And so it really required us to build de novo data sets to replicate the way we practice in emergency medicine. We don't know most of our patients. Well, we know some of them. But we don't know most of them when they walk in the front door. They're coming with new acute issues, that are prompting an urgent visit.
So we start thinking about what's happening right when they walk in. Building data sets that are about that moment is where we found a lot of power to really start to think of our data sets built in our practice, and so we have to replicate data sets that build how decisions are made in series.
We have to replicate decisions made in series on data sets. So we might know nothing more than a name and maybe an age and what you're here for and what they say or what a paramedic said or a family member said, and we start medical care there. How do you build algorithms and build digital health tools that take that into account?
That level of precision for what is it that we're trying to influence has been the center of what data we're collecting, and then the next level has been how much, how many times you have to multiply that to really find a signal that's worthy of community and national engagement.
Dr. Matthew Strehlow: I just don't know if the information is in the system. Like that patient you had, if we did a review of that, the information of what really went wrong, I don't think would be represented in their health record. So, are we going to be able to get to a place in the not too distant future where more is represented in the health record and we can feel more confident in that? And subsequently, what are some of the steps to get there?
Dr. Maya Yiadom: It's a tricky question because you sort of alluded to the idea, is more data better? We have more and more and more data than we had before, but we have the same problems, and are we able to leverage that data to fix it?
In this case, let's say we have more data. There is a lot that's in an electronic health record already that we haven't even learned how to leverage well, and there's a lot of work happening at Stanford and across the country and world where people are thinking about how do we not just put data in, but actually have it feed back and help us practice better.
But it isn't everything, as you had shared. There's a lot that you see, and sometimes in emergency medicine smell, in order to figure out what's going on, and, how do you capture that? There is this tangible physical component to how decisions are made in medicine or not made in medicine.
You know, in the case that I was sharing, it's a busy day, crowded, the noise level is high, distraction level is high, staffing is often not sufficient for the demand, so people are operating within that environmental, those environmental factors in the background.
Are there markers for that in the data system? Maybe. But for the most part, when you look at a patient record, you don't see any of that. So I do think that there is an opportunity to add more. People talk about using visual data and capturing images of how sick people are and the facial expressions. Voice quality can be used. So now we're capturing more of the tangible things humans, collect as we make decisions that's not in an EHR with tabular data, for the most part.
However, I do think that we can't capture everything. And, you know, in emergency medicine, people tease us as the doctors that kind of go with their gut, and we do, but it's what you see and you smell, and it's the facial expression, the skin color, and the mood with the family around them, and it's what's not being said that you learn is data in our decision-making.
And a lot of that is not documented, and I don't think ever will be documented short of cameras running 24/7. And we have a whole world of patient privacy, to resolve before I think that becomes part of what get, what gets collected.
So, I think there's a lot more promise with recognizing that there are two streams of data, that an electronic health record system is just…it's snapshots of moments where activity's happening, where human and the computing system touch.
And there's a lot of things that happen when we're away from a computer or away from a monitor or the patient's away from their monitoring that probably never will be captured if we're moving as quickly as we're supposed to do. So how do we recognize that there's some visual data that humans will be able to process and there is some captured data that computing can capture, and we need to leverage both in decision-making and augment each other.
And, you know, a lot of the work that my group has really embarked on is this idea of augmented human performance, where humans are still the lead for the activities. It's why we have doctors and nurses physically in the space with patients at the bedside. We have social workers, case managers, ED techs, all with different roles, all members of the team on the field, if you will.
But isn't it great when we have instant replay to really see, was it a foul? Is it out of bounds? We can go back and make decisions on the fly informed by the technology that's around us. I think that partnering of humans with the data is probably much more representative of what is a realistic way for us to operate in our future than really thinking we're going to capture.
I think some of the data scientist movement is really creating the impression that it's about giving the computers everything they need to help do the job better or do the job instead of humans.
I think humans in the physical environment are good at some things and capture certain things very well, and there are other things we don't need to be thinking about that computing can help us with, and digital health tools as well.
Dr. Matthew Strehlow: I think that's a really good point. I think it might be harder in the systems work than it is at the individual patient level. We often think it's hard to capture all the patient data, but we're now recording our conversations with the patients potentially, right?
And using, ambient dictation or scribes or whatever. And there are people researching, videoing patients. And whether that actually provides information And so with the patient care, the direct patient care of that individual patient, we're getting better at capturing that.
But when it comes to what's going on around that patient in the system, I don't really see anybody doing a great job of capturing that and integrating it, as much as we are trying to capture, individual patient, information. And so. to your systems questions, should you be just videoing the entire emergency department and then that gets fed into something, because that's what the physician and the nurse and the tech are all taking into account.
They're looking around as they're making that decision and seeing everything that's going on, and what time of day it is, and how cold it is outside, and all of these other features.
Well, we can't talk about data without talking about AI, and you mentioned it, right? We're all thinking about how AI is going to change our specialty. You have used a term that I think is different than what we normally hear. We normally hear we're going to put human in the loop for AI. There's AI, decision-making or driven, and then there's human in the loop. And you've said, "Well, let's think about that as AI in the loop," right?
AI in the human loop. And that's much more in line with the way I think of it. How do you think about that idea of having AI in the loop?
Dr. Maya Yiadom: I think the idea of human in loop is really, the framing that, artificially intelligent systems, are doing something, but somebody needs to check to make sure they're doing the right thing, and somebody needs to have the authority to sign off on what they're doing, and I think that's one framing.
Where I think my mindset is, is really in this augmented space where people are doing things already, and if we're really frank with ourselves, we're not doing it as well as we think we are or would like to. A lot of the work that I do is measuring our failure and our shortcomings, so I feel like I have data to show that despite how hard we work, we're not perfect, and nor is there an expectation that we probably could be.
It's almost impossible for us to achieve the goal that's set for each individual patient. So if we can take that gap and use computing to support us, then we're really putting artificial intelligence in our loop to augment our ability to actually do what's required with consistency and reliability and reproducibility.
And that seems like for now, I'm thinking the next 5 to 20 years in medicine, that seems like it's the game. Because you and I, we're going to be working in the emergency department, and there are many moments where I think, "Gosh, if I could be in two places, I could really get both of these things done very well." And I'm leveraging a plan B or somebody's waiting and if we had an ability to take some of the computational or task load that's in my mind off - like do we need to be checking to see if this belly pain patient needs a CBC, BMP, and a CT scan? Is there a way to look at all past abdominal pain patients with their characteristics and their established diagnoses from their records to be able to figure out if they need a CBC, a BMP, a lab test, and a CT scan on this day, and what kind of protocol is likely to be the case? And then when I have a moment to look at their case, I'm just deciding if I agree with the system that I have trust in because I understand what went into developing the algorithm and its performance was really strong.
That would save me three minutes of scanning a waiting room patient on a day where we might have 60 in the waiting room and I want to get care started.
There are lots of ways where I think we can cut the computational load and make us as physicians and our nurses and our techs and our case managers, et cetera, far more efficient in doing the stuff that only we can do, which is to sit at a bedside and have a thoughtful conversation with another human being, for us to tell them what we can do for them.
That data gathering, the rooming, moving between rooms, the switching computers, there have to be ways that we can use algorithms and digital health to really close those gaps to help us work more effectively. And so the AI or digital tools in the loop is more around looking at the existing human workflows and rewiring them, leveraging points where we're just not being efficient with the power and strength we bring to the moment.
And boy, I wish I was at my computer less on shift. I wish I sat at the bedside and looked at the whites of my patients' eyes more often. When you do take a moment to sit with a patient for a little bit longer, you can see they feel it, and there's something about being sick that's beyond medicine. It's just making people feel better, even sometimes when you can't fix their problem or when you are fixing their problem, you just need to go because you need to make some phone calls or do some paperwork and tasks that maybe don't help them in their moment.
So I do think there's a way to stay true to the practice of medicine, making people achieve better health states and averting illness, that, we can do better by using tools to help us do work.
Dr. Matthew Strehlow: Sitting by that patient's bedside and, caring is an important part of care. I also think, though, that the time that you spend gives you a richer picture of that patient, and you'll make better medical decisions for that patient and help them make better decisions. So, if we had more time, simply I think we just do better. And if AI and, other systems that you're working on can help us have that time with the patient, I think it'll improve care as well.
Because there's times that we've all sat by the patient and we had one idea, but the longer we sat there the more we realized that maybe we were going down the wrong path, or maybe we needed to try something else in this patient, or there was something unique about this specific case that should change the way that we proceed. How are we going to decide where are the next steps in data gathering?
Dr. Maya Yiadom: You know, it's tricky because I'll always say, "Why are we gathering the data? What are we trying to do here? What's the problem we're trying to solve? And I think, really honing-
Dr. Matthew Strehlow: Hold on. We can't say just everything, right? Like, we can't say I want to solve every problem…
Dr. Maya Yiadom: I know. We can't say everything. I feel like we need to say everything.
Dr. Matthew Strehlow: Because I feel like everything's a bit of a problem. When there's 60 in the waiting room and the patient who's bleeding didn't get the CBC when you see them three hours later, that's a problem. So okay, if we can't say everything, then tell me.
Dr. Maya Yiadom: I mean, I will say I'm super excited in this moment. We're in Silicon Valley. We're at Stanford University. There's a lot of exciting things happening here, and I'd say there's a lot of tech innovation that wants to fix medical problems, and people find medicine to be a tough nut to crack.
I feel like there are lots of nuts to crack. I could bring you a basket of nuts, and, there's a lot of power to leverage, and people are making things for healthcare that really, when you ask them, "Well, what does it do and what would we use it for and where does it fit?" The question is, "Well, how would you use it?"
And I do think there is almost a backwards innovation, where we're coming with the hammer and looking for the nail, so to speak. I do think there's a lot of value of finding the nails that need to be hit. Like, what are the burning problems?
You know, here at Stanford Health Care, part of it, one of the things that I think we have as our mantra is the idea of capacity. We're talking about emergency department boarding and crowding. It's not just the emergency department. It's hospital-based care. We're trying to think of ways to disseminate that to patients. We have introduced virtual care in the physical emergency department because we need to be able to move more patients or meet their needs much more quickly and precisely.
Leveraging a physician more effectively and efficiently and moving between rooms and locations just isn't the way to do that. We've used tech to help us with that. So I would almost say that there are burning problems that we need to solve, and this is maybe being a little bit too broad, a little bit of a buckshot approach with the tech at medicine, that if we can really take the power of the tech that's available to us and really give specific problems as use cases to build around the nuances to how you solve the problem and the realities of healthcare delivery, the physician-patient encounter, the patient's experience with the health system, trust differentials, how willing people are going to be to follow up, whether or not they are resourced to actually implement plans.
All those details matter when it gets to the point of well, is it going to change anything? And if we're really looking to change things for the better, we have to figure out, well, what are we trying to change for the better? What does better look like? And then once you figure those things out, going back and finding the data that represents that ends up being the key, and so maybe it can be everything, but it's all the problems. And then the data around all the problems, as opposed to all the data we possibly could collect. So maybe I'm hedging and trying to keep it on the everything level.
Dr. Matthew Strehlow: No, it's good. It reminds me a little bit of the physician in the loop in a different way.
So when I talk to some of our colleagues, that have engineering backgrounds or data science backgrounds, they all say the same thing, which is the hardest thing to do is to get a physician to tell them the problem.
They're willing to code anything, or have an AI bot code for them nowadays, but they don't know what they don't know. I think that is something where we can keep that physician in the loop, and maybe that helps bring up the problems that we need to focus on first and foremost.
Dr. Maya Yiadom: Well, I have to say first, you should have your friends call me up, because I have a basket of nuts. But I will say, I hear this time and time again from founders, from folks in the tech world, digital health, whether it's the computational side, it's the algorithm development side, it is really the software development side that getting a physician is hard.
And I would say you need to talk to lots of physicians, because we do have variation in our experience, and even understanding of what they're trying to get at. We know what we do really well, but even the problem at the department, the systems, the unit level, the community, that varies a little bit.
And if they're trying to build things that are for everyone, they need to talk to more than a handful of people, or even just one. I think we're hard to get ahold of. I don't think, clinicians, broadly speaking, recognize the power that we have in informing the ability for all the innovation to actually hit home and solve problems.
So my soapbox is that I think we need to lift our heads up from the problems and speak more to the community about what they are and be more available to people who have the power to move the needle.
Dr. Matthew Strehlow: I want to come back to the concept we talked about, about building networks of information and data. Because as I think about that, I've also heard that AI algorithms aren't going to be able to be transported from one institution or one system into another and be functional. So what are your thoughts on that, given that you are a builder of networks of data?
Dr. Maya Yiadom: So, there's the people level, conversation level, data level, and then there's if you build things to fix problems at home, how do you then get it to everybody else's house?
It's not easy for various reasons. One is, an algorithm needs to live somewhere, and you can build an algorithm or an AI tool at your institution within your own firewalls, but the moment you go through your firewall into another institution's firewall, you have issues with is it tailored to that use case in that other site. Their patient population's likely different. Their practice flow may be slightly different. Have you accounted for all of that?
You're also going through firewalls. You're creating weak points in firewalls for cybersecurity. Are you prepared for all of that, both in terms of how you buttress the connections through the holes that you've now made, or you protect those holes that you've created, but also are you handling the data in a safe way, and do you, are you covered in case there's an issue? Now there's a liability issue that is new by going to another facility.
And then the last part is if you have data flowing back and forth, who has the responsibility for that data flow? Is it me at Stanford University who's maybe sharing something with another institution? Is that even permitted?
What we're finding is it's more comfortable to move out of the university and then plug into universities separately, and it raises a lot of regulatory issues that are not at top of the list when we have these conversations about how to innovate in healthcare.
We have some archaic regulatory restrictions that didn't account for all of this work. We have a lot of patient protection regulations that are really important for highlighting the importance of patient information, but it's really limiting our ability to innovate, despite the fact that many of us who are innovating know how to handle patient information very well. Being able to create new legislation that allows and enables this kind of work is hard.
Things get really complicated when you're trying to do things that go everywhere, from the computing and software. Luckily, there have been a lot of new tools that provide, protections and create software platforms that handle this, but it's very different than anything that you or I learned training to be emergency physicians.
So, it does require engaging different kinds of expertise to make that level of impact. It's not easy, but it's very doable. And we do have partner colleagues out there who have the knowledge for how to do this.
I've just found that teaching them the nitty-gritty of what we actually do moment to moment, like teaching a software integration team about emergency department arrival, the first 30 seconds of an ED visit has taken over 17 meetings. But those 17 meetings are critical for them to have everything they need to take us to the next level.
So, I do think there's a lot of opportunity in that scaling of AI intervention that really allows for generalizability, but it's a whole different ballgame that gets into legal, cybersecurity, software platforming, institutional comfort, and also, who owns liability for where this lives and sits.
That is a whole other level of conversation that I don't know that we're quite ready to talk about. But I do look forward to the energy it takes to do that kind of work coming down in the next couple years. Sooner, if we can get there.
Dr. Matthew Strehlow: Just around the corner. So, the next couple years, are we going to be able to build these rich pictures and leverage AI tools and put them in our loop, to really move the needle on care? What-- In five years- what do you think?
Dr. Maya Yiadom: I think so. I think it'll happen in five years. We have elementary school kids who use LLMs as part of school projects. That all happened in the past five years. I do think it really requires advocates who understand the data perspective, the informatics of how hospital data systems are built, how we use the data, how it's entered, what we're doing when the data's generated from the clinical side, that can really do that whole journey to lead those conversations towards solutions that get past all those barriers.
But it's an interdisciplinary conversation, and getting things done when you have seven people around a table who speak completely different technical languages is not easy. But I do see that happening in the next five years.
I do think it's going to just be a two or three-year window of nothingness, and then it's going to click and then exponentially, really grow. So I look forward to looking back and saying, "Matt, remember when we had that podcast and this, we talked about it being five years," and it felt like it was very aspirational, and looking back and saying, "Gosh, that was such a long time ago, and we've really made so many strides in medicine and, with innovation on the tech side."
Dr. Matthew Strehlow: Well, I think with your compassion from that story you told, your obvious passion from listening to you, and your expertise, I feel more confident that if you're guiding us, that we'll have the success that you're, you're outlining.
Because we're struggling at times to, to have that positive view of where things are heading in healthcare. But it's great to hear that you feel that we can get there and that we will improve care, and I'm on board. Sign me up.
We like to end the podcast or the show with a little bit of a lighter question. I know we had to strongly recruit you from different places in the country, and you've lived in the Northeast and the South, and now you're out here in California for a few years. So, tell me, what do you miss about living in the South?
Dr. Maya Yiadom: Oh, gosh. I will say the thing that struck me the most when I lived in the South was eye contact. You know, when you're on the subway in New York City or you're walking on the sidewalk in Boston or you bump into somebody, you just don't make eye contact if you don't know them. If you do, you quickly look away.
And living in the South, people make eye contact and they maintain it, and there's a stream of kindness that comes from them. And they don't want anything from you. They'll make a few comments and you'll pass. And that recognition of like the human in your space, I just... It was so novel to me coming from the Northeast. First I was like, "What's going on? What do they want?" You know…holding my bag tighter and looking with the side eye. And then I just realized it's just this pleasantry of just acknowledging the humans in your space, and, I think I've brought that with me. It's here in California too -
Dr. Matthew Strehlow: When we look up from our phones, you, we make eye contact. Is that what you're saying?
Dr. Maya Yiadom: That’s right. From all of our devices. But that was the thing that I really loved about the South, and I do miss it because it is something that gives you a warm and fuzzy feeling throughout the day. And, you know, having spent a lot of time in Philadelphia, New York, and Boston, you know warm and fuzzy is not what you look for.
Dr. Matthew Strehlow: I do think that sense of community is really important, and I appreciate you sharing that.
That's a wrap for today's episode. Huge thanks to Dr. Maya Yiadom for sharing her insight and her experience. Clearly, we've only scratched the surface of this important topic. For those listening, be sure to check out the show notes and links to Dr. Maya Yiadom's latest work and studies. If you liked today's episode, don't forget to subscribe, leave a review, and share it with a colleague.
Thanks for tuning in. We'll see you next time. Until then, keep taking care of anyone with anything at any time.
Dr. Maya Yiadom: Thanks, Matt.
Dr. Matthew Strehlow: Thanks, Maya.