In this episode, experts from Stony Brook Medicine explore how artificial intelligence (AI) is reshaping healthcare today – and what lies ahead. This conversation covers how AI is being used in clinical settings, how its performance is monitored, and what it means to deploy these tools responsibly.
Also discussed are the skills physicians need to work effectively with AI, healthcare inequities, and where human judgment must remain central. We also look toward the future of AI in healthcare and how patients are beginning to incorporate AI into their own care journeys.
The Experts
Jonathan Buscaglia, MD
- Chief Medical Officer, Stony Brook Medicine
Samita Heslin, MD, MBA, MPH, MA, MS
- Deputy Chief Medical Information Officer, Stony Brook University Hospital
Eric J. Morley, MD, MHA, MS, CPHQ
- Chief Quality Officer, Stony Brook University Hospital
What You’ll Hear in This Episode
- 00:00 Opening and Introductions
- 1:30 How is AI currently being used in healthcare?
- 2:57 How is AI performance monitored and evaluated?
- 5:10 Understanding AI and how to use it in responsibly in a clinical setting
- 7:03 What skills do physicians need more of in order to use AI effectively
- 8:22 AI and inequities in healthcare
- 11:03 Where can AI support but not replace physician judgement?
- 13:20 The future of AI in healthcare
- 16:20 How do patients incorporate AI into their care?
- 20:03 Closing Remarks
Full Podcast Transcript
00:00 Opening and Introductions
Description of Video Studio: News desk with Stony Brook Medicine logo on the front. A big screen is behind seated experts with the Healthcast logo (red uppercase lettering with a microphone at the top of the “L”). Music plays as the announcer introduces the episode.
Announcer
Welcome to Healthcast, where leaders and experts from Stony Brook Medicine come together to discuss a range of topics, from leadership and strategic planning to patient care and the inner workings of a successful healthcare system.
Jonathan Buscaglia, MD
Hello, everyone, and welcome to Healthcast. My name is Dr. Jonathan Buscaglia, chief medical officer here at Stony Brook Medicine, and I’m joined by my two colleagues, Dr. Samita Heslin and Dr. Eric Morley. Dr. Heslin, could you take a moment to introduce yourself to our audience?
Samita Heslin, MD, MBA, MPH, MA, MS
Sure, I’m Samita Heslin. I’m an emergency medicine physician and deputy chief medical information officer.
Jonathan Buscaglia, MD
Terrific. Dr. Morley.
Eric J. Morley, MD, MHA, MS, CPHQ
Hi, I’m Eric Morley. I’m also an emergency medicine physician and chief quality officer here at Stony Brook Medicine.
Jonathan Buscaglia, MD
Great, thank you both for being here today. We’re going to be diving into a hot topic today, and discussing the role of AI in healthcare.
AI, or artificial intelligence, is when computers and machines are designed to perform tasks that normally require human intelligence. AI systems can learn, reason, make decisions and even solve problems. It may seem like AI is new, but it’s actually been around for a long time.
What’s different now is that it’s evolving. That’s why we’re hearing about and seeing so much more of it now. We’re going to talk about the evolution today, specifically as it relates to healthcare, and that brings us to our first question.
1:30 How Is AI Currently Being Used In Healthcare?
Jonathan Buscaglia, MD
Dr. Heslin, how is AI currently being used in healthcare?
Samita Heslin, MD, MBA, MPH, MA, MS
So, artificial intelligence, or AI, are systems that learn from data to make predictions, and we are seeing artificial intelligence more and more embedded into healthcare practice today. There are many different applications of artificial intelligence. One of these is early warning systems, and these are systems that are AI models that can look at patient data and look at it continuously in real time and predict potential clinical changes, and that can help clinicians intervene earlier.
We’re also seeing artificial intelligence being embedded into documentation, particularly ambient AI, and these are systems that can listen to patient notes and listen to patient encounters, and create clinical notes based on these encounters.
The other thing that we’re seeing in the world of medical imaging is artificial intelligence systems that can flag abnormalities within medical imaging and help with prioritization of studies.
We’re also seeing artificial intelligence being used in hospital operations, so we’re seeing AI systems are helping with staffing and flow models and patient safety as well. But there are so many opportunities to use AI in medicine. It’s a very exciting time in medicine as well, and we’re just scratching the surface here.
2:57 How Is AI Performance Monitored and Evaluated?
Jonathan Buscaglia, MD
Wow, I’m shocked to know how many different facets of medicine either AI is currently being used, or there’s potential for, which I guess brings me to my next question. Dr. Morley, how is its performance monitored and evaluated? How do we know that it’s doing a good enough job for us?
Eric J. Morley, MD, MHA, MS, CPHQ
So, I think that’s a very difficult question. I think there’s a lot of excitement right now, and certainly the future holds a lot of opportunity to embed this technology into how we take care of patients. But I think it’s critically important to take a step back, and when you are first evaluating whether you want to implement something or how it functions, is like any other diagnostic tool or treatment modality, you must ensure that it’s been truly vetted with good research, like any other modality that we use. So are there large randomized control trials looking at how they function?
Does it perform safely and reliably? And that’s really important. Also, we need to make sure that our teams evaluate these different platforms to make sure that they’re safe from a security standpoint.
So, if it is the A to Z IT team involved to make sure that data is used appropriately and in a way that keeps our organizations or anybody’s organization safe.
Additionally, there’s compliance issues that come in, and so you have to make sure that all in all that the technology is used in a way that’s safe but also keeps the organization safe while taking good care of patients.
And then ongoing when you implement these types of technologies it’s important to make sure that they function as advertised, and that requires the teams to be very engaged with both from the IT perspective and the clinical perspective to make sure that they’re functioning as they were intended to. And as we know there are many platforms, not to name any, where you can ask some very simple questions and get some very incorrect answers. It’s critically important that our frontline practitioners use these tools and provide good feedback to how they’re functioning, and make sure that they’re still maintaining their expertise and evaluating these tools as they’re implemented and used.
5:10 Understanding AI and How To Use It Responsibly In a Clinical Setting
Jonathan Buscaglia, MD
Fascinating. So, I want to go back to something you had mentioned a minute ago, Samita, about AI listening and ambient listening, and it reminds me of something that Eric just brought up, and, sort of safely incorporating these things into a physician’s workflow.
I presume that these AI capabilities recommend a treatment plan for a patient, for example, when using it for a patient encounter. Is that the case? And do physicians have to be wary of what that treatment plan is, and that it’s in line with what the thoughts are in their head, and how they truly believe the patient should be managed? How does a physician handle that?
Samita Heslin, MD, MBA, MPH, MA, MS
Yeah, absolutely. And that’s a great question. So, as artificial intelligence is being more and more embedded into clinical practice. Clinicians really need to develop a new fluency in artificial intelligence, and it really starts with technical understanding of the AI model. So, understanding how these models are built, what are they tested and trained on and what are their limitations. And that brings in clinical judgment, like you mentioned, as well as critical thinking.
So, when an artificial system gives you a recommendation, you’ll have to decide, as a clinician, whether you’re going to accept that recommendation, whether you’re going to question that recommendation and whether you’re going to override that recommendation. And that comes with your clinical judgment. And AI models are only as good as the data that they’re trained on, so there could be a lot of limitations there.
So, things that need to be thought about are: There are certain artificial intelligence models that will perform differently with different patient populations, so clinicians need to keep that in mind and use that as a factor when deciding which models they’re going to use, when, where, how and why.
And then physicians and clinicians really need to be involved in the implementation of the model in the clinical environment, too — selecting which AI model they’re going to use, how they’re going to implement it, how they’re going to integrate it into their work, and also monitoring and governing that model after it’s deployed.
7:03 What Skills Do Physicians Need More Of In Order To Use AI Effectively?
Jonathan Buscaglia, MD
So that sort of brings me to my next question, which you partially answered, that I had for you, Samita. What kinds of skills do physicians need more of in the future when they’re learning to work with AI?
Samita Heslin, MD, MBA, MPH, MA, MS
Right, that’s another great question. And one of that is when an artificial intelligence model gets deployed into the clinical environment, really understanding what that model does. So, how is it built, working with the companies to understand how that model was created, what kind of patient data it was trained on and what are the optimal use cases for that model. That’ll help clinicians understand the limitations of that model as well.
And one of the other aspects that’s really important as well is how do you explain this model to patients. If you’re using an artificial intelligence tool or model in your clinical practice, explaining that in a way that patients understand as well. And then other skills too, because after the model is deployed, often it’s trained on retrospective data and in other clinical environments. So, how is it performing in your particular clinical environment as well? So, asking questions like, what are the opportunities for this model to improve, and where am I giving feedback, because we want humans in the loop giving feedback about the model, so that we can continue fine tuning those models as well.
8:22 AI and Inequities In Healthcare
Jonathan Buscaglia, MD
Got it. Makes sense. Wow, let’s change gears a little bit and talk about something you mentioned. This question is for Eric. It has to do with inequities in care — something that we’re more cognizant of today than ever before in healthcare.
Can AI help us prevent inequities in care, or do we have to guard against certain AI-driven modalities to prevent biases and other things that we should not allow to creep into the way we manage patients?
Eric J. Morley, MD, MHA, MS, CPHQ
So, it’s a great question, and I think the answer is yes to both of those questions.
Take the second part first: Yes, AI can be used as a tool, when trained properly, to look for inequities. It can certainly be trained to do that, and you could develop a platform that would help do that work by training it appropriately.
But that brings us to the first question: Do we have to guard against this? Absolutely. Actually, this is no different than how we’ve thought about medicine for a long time. I think the best comparison would be how we look at heart disease in women. We know that a lot of the literature maybe didn’t include women, and it’s a very similar example of how that can be amplified if you take an AI model and don’t train it on a diverse population. Those inequities can actually be amplified by AI.
So you need to make sure that you train these models on a very diverse population, looking at many different types of patients and scenarios — not only broad sociodemographic groups. That’s how you ensure that you have a model that can handle that type of (variation).
Jonathan Buscaglia, MD
So that’s interesting. So it sort of goes back to what Samita was saying about how it’s only as good as what you put into it — sort of a ‘garbage in, garbage out’ type of approach.
Eric J. Morley, MD, MHA, MS, CPHQ
Correct. And that’s a limitation, I think, that we see in general, right? The models are only as good as what we teach them to do at this point. There is some ability for them to learn, but that is limited.
Ultimately, they need a very rich source of data to train the model to handle all sorts of scenarios in all different types of patients and really see the breadth of things that we see every day. And we know, as clinicians, that we need the same type of input in order to become good physicians. The AI models need the same type of input.
Jonathan Buscaglia, MD
Yeah, that makes sense. I never thought about it like that, but it makes total sense.
11:03 Where Can AI Support But Not Replace Physician Judgement?
Jonathan Buscaglia, MD
Let’s change gears a little bit. This comes up a lot, I think, in some of my conversations with people outside of healthcare, and that is: Where can AI support — but not replace — our judgment as physicians?
That’s a common question. People often ask, ‘Are physicians going to be obsolete one day because AI will just do everything?’
So I guess my question is: How do we fit into the mix, and what do you think is the best answer to that question?
Samita Heslin, MD, MBA, MPH, MA, MS
So, artificial intelligence can help clinicians in their clinical workflows. That’s a really important part of it, but clinicians possess some traits that are totally irreplaceable in healthcare.
One of these is in diagnostics and treatment. AI models can help find abnormalities. They can also flag for potential patterns within data and continuous patient data. However, one thing they can’t do is what clinicians (do) at bedside: seeing the patient, examining the patient, evaluating the patient and also understanding the patient’s preferences. These are very human traits that artificial intelligence can’t replace.
Also, clinicians are making high-stakes decisions with limited information, so often we don’t have all the test results back, or all the lab results or imaging results back, and we need to make critical decisions for patients. This requires clinical expertise. It requires clinical judgment. It also requires ethics and real-time situational awareness. Artificial intelligence doesn’t have that. It’s something that clinicians really need to be involved in.
And then, with patient care and building trust, clinicians are the ones sitting at bedside with patients, understanding their goals of care, navigating family dynamics and cultural context, and this all requires empathy and presence and building trust, and those are true clinician traits as well.
So, AI can help with clinical workflows, but there are many aspects of healthcare that really need clinicians.
Jonathan Buscaglia, MD
Yeah, I agree with you wholeheartedly. We always talk about the art of medicine and the art of being a good physician, and to me, what that means is it’s not just about the science and what the lab values show or the likelihood of diagnosis X or diagnosis Y, but all the other things that we incorporate into our clinical judgment based on what the patient may have told us even two visits ago.
So I agree with you. That makes a lot of sense.
13:20 The Future of AI In Healthcare
Jonathan Buscaglia, MD
Eric, this is sort of a two-part question. I’m curious to what you think AI and healthcare will look like in the immediate future, like in five years, and what it will look like further down the road, in like 10 years, let’s say,
Eric J. Morley, MD, MHA, MS, CPHQ
So, as we discussed physician judgment, I think we’re a long way from that, and I would echo a lot of what’s been said. We all know that, as we’ve trained as physicians, the information part — getting the test right — is probably the tip of the iceberg when it comes to what it is to be a doctor, right?
Understanding the facts is a small piece, but pulling together how to get a story from a patient, how to get the subtleties of what the patient is telling you about their symptoms and pulling all of that information in — I think AI is a long way from that.
So I think the immediate future is (AI) helping us with a lot of the daily tasks that we need to do so we can actually focus on good patient care and getting better histories, spending more time at the bedside and talking to our colleagues about the patient.
So things like, we already touched on ambient scribing, right? That’s one example that is very useful for offloading some of the time you need to spend at the computer typing. That is a very big offload for the physician.
I think in the near future we’ll going to see also see technology where we can pull together the chart and ask the AI: ‘Tell me the history for this patient. Tell me their story, their cardiac history, all of their visits. Pull together all their ECHO results or EKG results and summarize them for me.’ so that i can, without spending a half hour reading the chart, I can know a lot about that patient’s cardiac history, for example.
So those kinds of tasks — where you offload work from the physician — are coming really soon. I think they’re already evolving and happening, and AI, is already poised with the right training to get us there.
Down the road, I see AI as being sort of a teammate with you on your shift or while you’re working. Where you’re doing all your work as a physician, and in the background the AI is saying, ‘Hey, you just wrote your note. You diagnosed the patient with sepsis, but everything is telling me this patient might have cardiogenic shock instead of septic shock. Did you consider that? Would you consider that?’’ Or: ‘You diagnosed pneumonia on the X-ray that you ordered, but you haven’t ordered antibiotics yet. It’s time to order antibiotics.’
Things like that — where you have sort of a teammate with you, pulling information and helping you with basic facts while you’re working through collecting that data. I see that as a very realistic intermediate five- to 10-year stage of where we’ll likely end up.
And to me, as a physician, that actually sounds quite useful — having that kind of backup helping you move through your shift.
Jonathan Buscaglia, MD
It’s so fascinating to think about what our lives will be like as physicians just in 10 years from now, and to think about some of these things that you mentioned.
16:20 How Do Patients Incorporate AI Into Their Care?
Jonathan Buscaglia, MD
My last question really has to do with, switching gears from us, the perspective of us as the physician, but now the perspective of a patient. I’d like to get each of your thoughts on how AI is used from a patient’s perspective. What are some of the things that patients do which incorporate AI into their care?
Samita Heslin, MD, MBA, MPH, MA, MS
Yeah, when we think about artificial intelligence from the patient’s perspective, it’s really about how the technology meets the patient in their everyday life.
So, let’s talk first about wearable devices like smartwatches. These are devices that are constantly tracking patients’ heart rate, telling us if there’s irregular rhythms, but also tracking other things like sleep cycles, so they are producing continuous clinical data. Artificial intelligence models can incorporate that in real time, notify patients if there’s any abnormalities, so patients can actually seek help earlier.
The other thing is that along the same way, there is remote patient monitoring, and these are really helpful for patients with chronic conditions like hypertension or diabetes or heart failure. These models can keep track of certain health factors, and then let healthcare teams know if there’s any issues or any clinical changes as well.
We’re also seeing artificial intelligence being more and more embedded into virtual health assistance, so these are software that patients can use to help with scheduling their appointments, as well as managing their medications, and then also contacting their healthcare teams,
And then one great example of artificial intelligence is how it’s being used for patient education, and there’s a lot of complex medical information out there, and then what some of these artificial intelligence systems can do is take that complex information and integrate it into personable, personalized education for patients,
So there’s a lot of opportunities for artificial intelligence in healthcare for patients from their perspective, and it can really make healthcare a little bit more continuous and accessible and personable for patients.
Jonathan Buscaglia, MD
Anything else, from your perspective, that a patient might engage in AI?
Eric J. Morley, MD, MHA, MS, CPHQ
Very, very thorough answer, of course, but I would like to touch on that patient education part, you have the patient portal now, which has a tremendous amount of information at the patient’s fingertips, right? But it can be incredibly confusing, and we all know that.
And to have that power to go through and say, ‘I just had six appointments with six different doctors, tell me what I need to know, what’s my next step,’ in a very simple way, and to be able to maybe get an answer where — and that’s you’re starting to see some systems work with technology like that, where you can go and say, you know, what’s my diagnosis, what does that mean, who do I need to see, and start to answer some of those questions. What should I ask my physician the next time I see them, what test should I be asking for, what treatment should I be asking for, right? And that’s probably coming very soon.
Jonathan Buscaglia, MD
This, it’s a great example, you know. As a gastroenterologist, I see some of my colleagues in and around the field referring patients to the very equivalent of the patient portal to get results from, you know, pathology from colon polyps.
And, you know, I still write sort of old-fashioned letters. I mean, I type them in Cerner, but I still send them out for this exact same reason, because I don’t think that a patient necessarily knows the implications of a tubulovillous adenoma versus hyperplastic polyp versus an inflammatory polyp. Nobody’s explaining that to them in the portal, and I have to do it myself in some fashion, whether it be an old-fashioned letter or phone call or the next visit, you know. And so this is exactly an example of what you’d be talking about. Makes a lot of sense.
20:03 Closing Remarks
Jonathan Buscaglia, MD
Well, that’s all the time we have. I want to thank our experts for a wonderful conversation, and thank you to all our listeners and our viewers. If you found this information interesting or helpful, don’t forget to like and subscribe for more content just like this. Thanks again, everybody.
Stony Brook Medicine is Long Island’s premier academic medical center. We transform lives through scientific discovery, education and care, and we bring together innovative research, advanced education and extraordinary healthcare expertise to set the standard for how healthy communities thrive. For more information, visit stonybrookmedicine.edu or follow us on social media.
*DISCLAIMER: The information provided in this podcast is for educational and informational purposes only and is not intended as a substitute for professional medical advice, diagnosis or treatment. If you think you may have a medical emergency, call your doctor or emergency services immediately.
