Good afternoon. Thank you for joining us on our session this afternoon. We're really excited to share innovations in the interface of EP and multidisciplinary care for what is to come. So we are going to start. We're actually going to switch the order of our talks because Dr. Singh has something else. So he will be speaking to us on innovations, innovative or actually irrelevant for AI and he is coming to us from Harvard Medical School. Thank you. Thank you dear chairpersons and thank you all for whoever here. I appreciate it and it's a pleasure. So this is an awkward topic, right? I mean always innovative or actually irrelevant. But I'm going to start by just with this little saying from Alvin Toffer who many of you know is a medical futurist and is a futurist and a writer who said a lot of cool things about adoption of new technologies and stuff. But I think for the adoption of AI, what I could translate him having said is that the illiterate of the 21st century will not be those that cannot read or write but those who cannot learn, unlearn and then relearn. And I think that really speaks a lot to AI for sure. Now with that said, I think within the term AI, there are many controversies buried, right? It can be almost intelligent, could be always intelligent, could be augmented intelligence, automated intelligence, assistive intelligence or it could be analytical intuition or what people now think is more of the fact that it could be algorithmic insights. So I don't want to get into details with the acronyms but what we're going to talk today about is really the two extreme ends of the spectrum. Talks about actually irrelevant and always innovative and I think where we are and where we should be somewhere in the middle and try to figure out how we need to bridge those gaps. With that said, why this talk? I think this talk is important because if you look at, this is unpublished data which has just been submitted to a couple of journals I think in a salami version but the bottom line out here is that there has been a spate of approvals by the FDA for AI, ML enabled cardiovascular devices. More so in the last two or three years but from 2010 onwards almost 70 to 80 percent of all the AI devices actually have been accepted out there but the important thing out here is you look at that. Most of the devices or the subspecialties, let me say, the ones the subspecialties that actually have most of the devices are the ones that have a digital footprint and those are primarily EP and cardiovascular imaging and within that if you look which are the subcomponents out there, AI ECG seems to be one of the hottest topics for getting FDA approval for a variety of reasons and I think all of us have experience with that on many fronts. Now Hartford and obviously you know recognize that AI is important and we have this whole symposium seminar that you all know about called HRX. Co-enjoined with that we came up with this little statement about where AI can revolutionize healthcare delivery research and education in electrophysiology and I'm not gonna belabor this but every part of EP whether it's arrhythmias, it's sudden death, it's out-of-hospital cardiac arrests, it's heart failure, it's remote monitoring, it's conduction system pacing or CRT or digital twins, there are already artificial intelligence strategies finding their way into each one of these options out there. Now I don't have time to talk about all of them and talk about the innovation and the irrelevance but I'll take a simple example of atrial fibrillation. Do we have a timer up here that I can look at or there's no timer out here? Okay I'll hang on and make sure I'll be within 10 minutes though. There is one up on the screen. Having said that I think it's important to recognize that within the management of atrial fibrillation there are many decisions whether it's stroke risk stratification, whether it's medication management, whether it's screening and diagnosis or whether it's personalized treatment. Whether a patient needs an atrial fibrillation ablation or patient needs a left atrial appendage occlusion device, we use very generalized criteria right now and I think all of us recognize that much of this really needs to be individualized and personalized and that's where AI actually has a role and in many of these areas already is playing a bit of a role out there. But let me address the issue of irrelevant because if it is irrelevant it can lead to patient irreverence and I'll explain you how. Now there are many determinants of relevance of any algorithm and I think the determinants are largely the data. What's the quality of the data? What's the availability of the data? How heterogeneous is the data? Is the data representative of the patient population you're actually studying that algorithm in? We know that there are many ethical issues. There's this whole concept of algorithmic bias now that has become a really important thing besides other forms of biases. And then there's the propensity to over rely on algorithms and we know that if you over rely on algorithms the probability of having false positives and then leading to irreverent actions which I'll talk about is important. And then obviously also there's this whole concept of whatever we have how can we integrate it into the workflow because if we cannot integrate it into the workflow it's irrelevant. And that's where if you have an algorithm which any of these determinants are playing a part it can have a detrimental impact. If you remember simple device diagnostic which is narrow form of AI just to use that as an example can set off an alarm that can cause an increased anxiety, lead to more diagnostic tests. Those more diagnostic tests lead to more hospitalizations and that in turn leads to further expense. All of the things that you are creating an algorithm to actually avoid then get cascaded by those things. Now the other thing is you know is some of the strategies we're using actually imperfect and this is a state-of-the-art that is actually not just public it just got accepted it's from era as well as HRS and ESC and I think it's being published right now or probably is released in the last 24 to 48 hours. But I'm gonna use a simple example from that and that is photoplatysmography right. We all know the PPGs on our watches, the PPGs on our cell phones can help us determine, diagnose disease states as well as atrial fibrillation besides a number of other disease states that I'm not going to go into. But just looking at AFib I mean if that AI science within using the PPG signal has its issues with the quality of the signal, the definition of the ground truth, the appropriate reporting of variability within subjects, the sensitivity and the specificity and all those parameters as listed out here they can all indirectly cause the algorithm that you're using to be imperfect. And if it's imperfect it becomes irrelevant and if it's irrelevant it'll ends up being consequential in a negative way. Now that was just about AFib with a PPG signal but I think all of us know that you can now manage AFib with machine learning strategies using sensor based approaches as well as cloud based algorithms. And what I mean by that is that you can now risk stratify patients from your EHRs using machine learning approaches. You can substratify them into high, moderate and low risk and then subsequently within those subset of patients you can apply sensor based strategies to further categorize whether those patients are predisposed for atrial fibrillation or not. And then we have the AIECG and this is some work we have done where you can look at the first few hours of a patch monitor and predict which patients are going to actually develop atrial fibrillation in the ensuing 14 days. So there are AIECG strategies to pick up AFib you know within 31 days and within five years just based off a static 12 lead ECG and then obviously we have cloud based algorithms. Now the reason I'm saying that is that there can be issues at every level of the data set. The acquisition of data, the quality of the data, the variability of the data and other parameters with the fidelity of the signals for many of these sensor based strategies using cloud based algorithms too. Now we talk about EP, we talk about readmissions and I just want to touch upon that. We talk about heart failure readmissions, we talk about AFib readmissions, we talk about CRT response and you'll realize that when you look at machine learning techniques to actually determine whether the patients are more predisposed to readmissions it really falls apart. And it falls apart because all our baseline parameters are baseline parameters and these are continuous states and that's why you need continuous sensor based approaches and data that is constantly being accrued and being analyzed so that you can better predict which patients are going to develop heart failure. You look at CRT response, there's this elegant paper from Mayo Clinic that actually looked at CRT response using conventional data and they found that it doesn't work. The AUC curve, the curves are really sub optimal at .65 or so, it's almost close to a toss of a coin and why is that? That is because there are missing variables, there is no exact data where the lead is, what its relationship to the activation sequence is, there's no data relation to where the scar is, there are no more I would say individual variables that really determine the response to CRT. And then we we talk about social, economic, and cultural factors which can also impact response that are unmeasured out here. So those are things that require multimodal AI and other strategies that are beyond the scope of this talk. Now if we want to make AI relevant and we want to make it innovative we really need to understand what its position is in our clinical practice. It is centrally located which is interacting with sensor-based strategies with virtual care. A lot of AI based approaches are now virtual. As you're talking to your patient you can make the diagnosis of atrial fibrillation off the face camera looking at facial flow patterns. This is just to give you an example, so virtual care sensor-based approaches and care pathways are important along with the interaction with other newer technology that can interface with this. So it's really important to recognize that nothing of this exists in isolation but in alongside other technologies. So really to I want to give you an example and this is a paper that is again probably just published. It doesn't have a page number on it as yet. It's in the European Heart Journal and I thought I'd use this recent publication to really understand how we have to make irrelevance into innovation. So this paper is largely about using patch monitors, using the first few hours of the patch monitors and predicting which patients still have VT or VF or sudden death within the next 14 days, right, off the first few hours. Now you may think this is a phenomenal idea, right? Works really well. The AUC scores out here a 0.95, really high predictive ability. You say wow that really you know it really makes a good good impact on us and maybe it's really innovative because it's predictor of malignant arrhythmias with just a simple ECG. It is so scalable. It is so cheap. It's so non-invasive and this could enable sudden cardiac death prevention. But the bigger question is that is really irrelevant because we haven't done any predictive calibration as of yet, right? The low, it has such a low pretest probability. I'm killing our own paper that just got accepted for publication but it has such a low pretest probability and it's rare because these conditions are rare. So it really needs to be put into perspective and when you do this there's you need to have a clear action plan. There's no clear action plan on how you're going to use that data, how are you going to notify patients and clinicians and how are you going to take away the anxiety that may actually put this you know in a place that you don't want to use this. And that's where I think we can bridge that gap and we need to bridge that gap between making something irrelevant into innovative by really through pragmatic trials, understanding who actually we need to monitor, how to notify, how to integrate with other covariates and really also do a cost-effective assessment alongside that. So in conclusion I'd like to say that AI is neither irrelevant nor innovative. As of now it's we are somewhere in the middle. Our goal should be to understand the gaps and transform that irrelevance to innovation and it's important we do this if we need to disrupt care pathways because as as Clayton Christensen has said that disruption and here means adopting AI is a process and not an event. Thank you very much. I don't know if there's any questions from the audience because Dr. Singh has to go so if there are please do come up to the to the mic but I've got one for you just because you showed that example of your paper and I think a big thing is about explaining these I guess black box diagnoses or predictions. How do you think we can bring that element of explaining what these algorithms are finding and I guess incorporate a degree of trust in where we're utilising it for clinical decisions. That's the bit I find very difficult. For sure, for sure. So what we did in that paper and you know obviously I didn't explain it is something called saliency mapping where we actually look at our AI algorithm and try to figure out what are the parts of the ECG and the analytical parts that actually contributed towards a sudden cardiac death prediction and there were two really important parts of it. One of them is from a patch monitor you get a heart rate plot and you look at the heart rate plot which looks at the heart rate variability the PACs, PBCs and that visual image digital imprint it was one of the predictors for which patients had a higher propensity because that reflects the autonomic tone and the changes in the heart rate and heart rate variability. The second thing we did was we looked at a saliency mapping to figure out which part of the ECG contributed towards that and that was embedded in the QRS and that makes sense because I think if you have some sort of structure of normality and some activation sequence of normality and that's picked up within the QRS those saliency maps suggest that the explainability is coming from within the ECG and I think for clinicians like us that becomes really important you want an explanation and I think that is buried inside the paper but I think when it becomes something that is going to get into a clinical trial or so obviously I think explaining it to the clinicians involved is going to be really important. Wonderful thank you. Any questions from the audience then at all? Thank you very very much appreciate it sorry to run. Right so we will move on to Dr. Artie Dallal as our next speaker and she'll be talking about remote transmission or remote monitoring in heart rate and transmission of data to actual clinical actions so welcome. Thank you very much I'm really excited to be here and it's always hard following Jack he's a great speaker and it's really lovely to listen to him but thank you to the so the chairs for the invitation. As mentioned, my name is Arthita Lama, Pediatric Electrophysiologist at Vanderbilt. Those previous disclosures are older, but so currently I have nothing active. So I think JAG did a really good job of sort of introducing this, and I'm going to highlight a little bit about AI, but the purpose of today really is to identify arrhythmia innovation technologies that have not reached their true potential. And one such example is remote monitoring. I think we can all clearly agree that there are gaps between data transmission and clinical action. We know that there are benefits to remote monitoring, but there really are some challenges. And what I hope to do is at least just touch on where those challenges are and how we can use current technologies and even optimize where remote monitoring is to really better care for our patients. And so this is the most recent sort of expert consensus statement on remote monitoring, published just a few years ago, and, oh, thank you, sorry, I got an alert that had five minutes left. And so the key points of that consensus statement was really to say that, one, we do know that remote monitoring is a standard of care, but it really does hinge upon prompt patient enrollment, regular connectivity, and adherence to remote monitoring. If we don't actually have that, then our ability to care for our patients really is compromised. But that then has a downstream effect on making sure that we have adequate staffing models, including both non-clinical and clinical patient-to-staff ratios. So how do we really address this? And I think when we think about how do we obtain that data in a timely manner, it's about understanding your patient population. So not just understanding how many AFibs versus BTs you have, but what is that infrastructure limitation for your patient? Are they in rural Tennessee, or do they live in the middle of New York City? And then once we receive that data, how do you effectively go through that information? And I think there are some identified challenges which I want to highlight today, which include interoperability and just the data overload that comes with it. So we're going to have two audience response questions, so if you haven't logged in, quickly sort of log into the system. And so the first one is, what do you, in your own practice, what do you find is your biggest challenge to remote monitoring? So is it the lack of interoperability, the fact that we have so many different vendors and they all have their own different remote monitoring platforms? What about your own organizational infrastructure, or just the amount of data that's coming through? Okay, so the lack of organizational infrastructure, that's, I think that's, obviously it's 100% there, but it's, I think we all feel that. And then I'm going to ask you one other question. So the key to sort of the data that's coming in, we keep highlighting alert-based monitoring continuous connectivity. How many of you in the audience have also transitioned your care to scheduled monitoring to alert-based monitoring? And I will address the first question later in a second. And so 50-50. Okay. So I think this is really important because we talk about the importance of continuous connectivity and how it impacts and improves patient care. But this is data from 2019, which was released by the International Telecommunications Union. And it really is highlighting the fact that as recent as 2019, if you look across the globe, and again, because I live in the U.S., I'm just going to highlight the U.S., but in rural areas, only 50% in 2019 had access to the Internet. Now it's gotten much better. More recently, the 2024 data suggests that in high-income countries, 93% of the population has access to the Internet, while in low-income countries, only less than 30%. And so if we're talking about continuous connectivity in our patient population, it's limited by the infrastructure that exists where we live. And so understanding your patient population is important because you may not be able to just go to alert-based care. And if that's really what we're advocating for, maybe the change and what we need to sort of implement is change in the infrastructure that exists. So this is, again, a call to action and thinking about the advocacy that we can do. Because if you don't have that continuous connectivity, you will have to rely on that non-continuous sort of scheduled monitoring that exists. I have patients that will drive to like a McDonald's because they don't have access to the Internet where they live in rural Tennessee. And so that goes along with, when we look at infrastructure sort of challenges, people or identified organizational challenges, I think that also deals with how you go through that data and the many different platforms. And so in 2023, the IEEE 11073 standard, which is essentially a standard that looks at the interoperability and communication between medical devices and IT systems, was updated. And this is really the work of members within our society that are trying to understand how we can better support our patients through CIDs. So for example, if you just look at the different vendors, the vendors have different home monitoring systems and how you can connect the frequency of transmissions. Can you do scheduled transmissions or patient-initiated? And how does that come through? And so that really presents a problem for us as providers because the variability within our CID management system really impacts our ability to care for our patients. And so since 2006, as I mentioned, there has been an active decision to sort of streamline this process. And so there was an HRS committee report published in 2024 that looked at advancing these workflows and how do we sort of make this better. And what the authors present in this guideline is, or this committee document, is identifying critical data exchange points between remote transmissions, EHRs, and third-party platforms. And they really wanted to help develop a concept of mandatory versus optional data elements for each device, ICDs, pacemakers, ILRs, and really wanted to keep each vendor accountable to this. Now, the challenges to this is that devices are evolving faster than we can change these standards, right? So the last one before 2023 was in 2012. And so if manufacturers continually add sensors or enhance existing ones, we are unable to keep up. So we have to sort of rely on industry to modify and sort of keep up with what we would like them to do. While some are being proactive, other CID manufacturers really are just adhering to what is currently the standard. And so there has been work, there continues to be work, and this will continue to be an ongoing need as we think about how do we really optimize what remote monitoring can do. And sort of, again, this is to answer more of that, the organizational sort of workflow challenges that exist in about 50% of your practices, or I think that was the 100%. So when we implant devices, obviously they meet criteria, but there is such a volume of data that is coming through. This, again, these charts are from the 2023 consensus statements. And so if you look at how we program these devices, you can leave them in nominal settings, but you are going to get a crazy amount of data that's coming through on each of your patients, especially if you are doing scheduled transmissions on these patients. And so if you think about reprogramming in a way that is for trying to exclude non-actual alerts, that really will ease the burden on your workflow. And so what the consensus statement tries to do is really help identify red critical alerts versus yellow alerts. And you can sort of go through, and I don't have the time to sort of go through this in detail, but this really does identify ICDs versus pacemakers and ILRs, and you can sort of, again, refer to the consensus statement for more details. But how does that really, how does this translate to clinical practice? So this was a group out of Helsinki, and what they did was they sort of said, okay, how can we decrease the amount of non-actionable alerts that come through our practice and how does that impact the amount of volume that we're seeing in our remote monitoring clinic? And they inactivated non-relevant alerts and kept the ones that were active using sort of the trust study as their guide. And so what they found was when they programmed non-relevant, clinically non-relevant alerts off, they decreased the number of alerts by 44%. And how many, we all get the recurrent alerts within a month, right, on the same patient. When they looked at that specifically, that decreased by over 75% and the ones within the recurrent alerts within six months by 65%. So that truly did offload a lot of excess data and workload on the staff. And importantly, the scheduled action alerts did not change, and most importantly, there were no sort of deaths that were related to adjusting alert settings. Now this is still relatively new, and they did say that they still need to continue to monitor these patients to make sure there isn't long-term events. But if we are thoughtful in how we're programming these patients, we may have the opportunity to decrease that workload that's coming through our clinics. Now future directions. I think remote monitoring obviously is important, but what about remote programming? And I think all of us find this an exciting sort of maybe opportunity, but thinking about the challenges of cybersecurity, for example. And so this group out of Bordeaux looked at just the feasibility of remote monitoring, I'm sorry, remote programming on patients with CIDs. And I think the COVID epidemic really sort of forced our hand. It made us be a little bit more thoughtful, maybe creative how we can care for these patients. And so what the organization of this sort of study or this feasibility study was, they had a cardiologist with a workstation, there we go, here, and then you had a programmer that was connected to the patient. And the programmer was then live streamed to the cardiologist via video capture. And then the cardiologist could then control the programmer through a microcontroller that was connected to the system. And there was end-to-end encryption, which ensured security on both ends. And so this group, what they did was they had 110 patients, they did 115 remote interrogations, and these weren't patients that really did not need their devices. About 55% of these patients were pacemaker dependent, so truly needed to have those devices. And they were able to successfully reprogram these patients, about 50% of these patients, including modifying ventricular arrhythmia settings. And so as a feasibility study, it is possible. But what about the patient that you're doing this on? How did patients feel about this? They actually found that patients were okay with it. They felt safe. They were willing to accept this if it would reduce their travel time and consultation days. Now, if you sort of look at the very last question, hypothetically, would you be willing to accept remote control of your device without any human assistance, that obviously went down. And it makes sense, because you want to make sure that you have someone that is responsible on the other end. But truly, the takeaway from this is that we need to be able to move to a model which allows us to care for our patients that are the ones that don't actually have access to us. The ones in rural areas, the ones that may not have the ability to do continuous connectivity. And so again, this demonstrated we could do it. Now we just have to make sure that our cybersecurity is able to meet the needs of our patients. And then I will very quickly, because I think I'm over time, talk about artificial intelligence in remote monitoring specifically. And I did a great job of talking about the limitations of AI and where it can go. But when you think about remote monitoring specifically, I think really where the benefit is going to be is decreasing sort of our workflow, or our workload, rather. This cross-sectional analysis that came out looked at de-identified ICM remote monitoring from Octogos Health, which looked at 140 U.S. device clinics over just about two years. And they specifically ended up looking at the Medtronic AcuRhythm AI, and compared that to the Medtronic Reveal, the ILR Reveal. And what they found in this was that the AI did a really good job of identifying non-actional alerts. And what did they define that as? So basically, transmissions that would be dismissed by device technicians and not forwarded to clinicians for review. And they, again, compared non-AI-generated versus AI-generated. And they found the number of non-actual alerts went down significantly. And also, when you think about the hours saved and the cost in these clinics was significant. So again, improving workflow was key here. And then, what about, how else can we use it? When we think about using AI to detect noise in remote monitoring. So this was a multi-center study out of Japan. They used a convolutional neural network with five-fold cross-validation. And they wanted to see, could you accurately detect noise that may allow us to identify early lead failure accurately? And again, there was success in that as well. So AI has a purpose in remote monitoring. And I finally want to put this picture up. Because if you went to the plenary this morning, Dr. Koloff talked about using the home as a healthcare hub. And Jag sort of alluded to it as well. There is a lot of conversations about deep learning and neural networks. And essentially, what's going to happen is with all of this technology that we have, we're going to be using it to sort of assess the quality, the health of our patient. And our remote, our CIDs are only going to be part of that equation. And so making sure that we are part of that conversation and making sure that we are, make sure good information goes into those neural networks or will be key. And so, again, another plug for HRX. If you want to be part of that conversation, I think HRX is a really good conference to be part of. And so, understanding that future directions really to help us bridge this gap is going to be universal CID language, intentional programming, and then possible remote reprogramming. Thank you very much. Thank you, Dr. Dallal. Thank you, Dr. Dallal, for that wonderful presentation. Next, we'll have Dr. Singleton speaking to us on the EHR of tomorrow. And we'll save questions for the end with the next couple of talks. All right. Good afternoon. Try to keep everyone awake for our last talk before happy hours. Excuse me. This is still loading. All right. So, EHR of tomorrow. So, the first question I have for you is how do we relate to the EMR? If you ask the physicians of more advanced vintage than myself, they'll tell you that the golden age of medicine is gone. It used to be that you talked to patients. Now, you spend all your time interacting with the computer. If that's the case, we have a big problem. What we have to ask ourselves is do you work for the EMR, or does the EMR work for you? Because if the EMR works for you, it should be making your life and the patient's life better, not having extraneous boxes for you to click for no purpose. And people say that the reason doctors are burning out is because of the EMR. I'm not sure that's as big of a boogeyman as it's made out to be, but it might be contributory, and if so, this is a really important thing we need to work on to prevent that. And I propose what we need to do is stamp out EMR inefficiency. In this setting, I would define inefficiency as any work that you do that's not value added for patient care. Because the reason you're here is to see patients and to make their lives better, not to click random box. With that in mind, I have five tips I wanted to go through today on how we can achieve that. Tip number one, regarding clinic notes. I had a 44-year-old man who was referred for AFib ablation. Great. That's something we can do. There were a couple wrinkles, though. It was only one episode that was self-terminating while hospitalized with COVID pneumonia with no recurrences, and it was an AFib. So before we consent for the ablation, let's take a step back. We look, and it's maybe a two-to-one, atypical flutter, EAT, hard to say. So we talked about options. We talked about what this means. He doesn't have any lifestyle limitations. We're not going to do the ablation yet. We'll do some follow-up monitors, and we'll see what happens. The question is, when do you write your notes? And I think you have two main options. The first option is between patients. And the problem with that is then you see less patients. And if your practice is anything like ours, you have a six-month wait list of patients waiting to get into EP. And a bigger problem is if you see less patients per day because you have to make time to write the notes between patients, you're going to have an angry admin, and no one wants an angry admin. You know the admin's always saying, churn harder, see more patients. So that's not a very palatable option. Well, what about after clinic? Well, the great thing about that is you can move through your day and then do it all later. The problem is by the end of the day, you've seen 36 patients. They all blend together. You forget a bunch of important items, and your notes aren't as high quality. But the bigger problem is that your kids can't pick you out of a lineup. You've just done a million years of PGY training. You're supposed to be able to be part of the family life, and you can't make it to the 5 and 6 o'clock events because you're still doing your clinic notes. But the even bigger problem with that is that you have an angry spouse. If she has to put the kids to bed alone one more time while you're catching up on notes, I don't even know what's going to happen. So none of these are very good options. So the question is, can we capitalize on AI to have a better option? Can we have happy wife, happy life, happy spouse, happy house? And it turns out we can. So remember that note I showed you from that patient I saw in clinic? That's my note, but I didn't write the note. So what happened is I used the DAX AI software, which is embedded within Epic on my phone. So what I do is I go into the room, I turn it on, and I just have a natural conversation with the patient. Where are you from? What do you do for a living? What brings you in to see me today? We review the data. We talk about the plans. And at the end, I click done. And in about four seconds, that appears in the chart. And before I walk out of the room, I can enter the billing and click done and close the encounter. And I can go straight from one room to the next with my mind cleared, ready for the next patient, not having to be, oh, am I going to forget that? So it's a really nice thing. If you haven't tried it yet, I highly recommend the artificial intelligence note-writing service. It's much better than all the other options I've tried, including offshore scribe and in-person scribe. Tip number two goes with that, and that's the stickies. So the problem with the AI notes is they can take multiple minutes to read. And sometimes you don't have multiple minutes. If you're consenting a patient before the procedure and you need a quick refresher, who is this patient? What are they here for? Or you're about to go into the clinic room. I know this patient. What are they here for? What are we going to do with them? You need all the data in 10 seconds or less. The key to that is something called sticky. So if I'm about to go into a room and see a hypothetical patient missed her test, I click his name on my phone, and what pops up is this. Every patient I've ever seen has a sticky note. And if you look at the sticky note, it tells you who are they. Oh, that's right, the CIA agent with the four kids. I ablated his wife and his dad and his relevant things from his history, persistent AFib, what happened to him, what he's doing now, how he's losing weight, and things like that. All that data is right there. All the important things are right there available on my phone to refresh my mind at any given time. Because that data is so good, I even copy it into my clinic note so other people who then follow up with the patient can see where's the AI-generated garbage and where is the really important stuff that Dr. Singleton actually wrote. And that's copied right here. And if you want to do the sticky and you have Epic, just click on this little yellow box up here and the sticky will pop up. It's a great way to keep your life organized. It pairs nicely with the AI notes as well because instead of spending all your time writing the note, you can put just the important things in the sticky and have that available to you. You can also share it with your care team so they can all know because sometimes you put things in the sticky note that you don't want in the chart, like difficult patient and things like that. Tip number three, BPAs, the good. Hear me out. BPAs can be good. Those pop-ups, the best practice advisories. I had an 85-year-old woman who had abdominal pain, came to the hospital, had a CT, found to have SMA thrombosis. And you know why? Because she had persistent AFib and wasn't on anticoagulation because of fall risk. Not actually fall, fall risk. You know what we think about that. Long story short, she had surgeries, SICU, and she died. We used to see this a lot. We see this a lot less over the last two years because what we did is we built our very own BPA, the AFOACBPA. This is a BPA that fires for patients who have AFib with a CHADS vasc of three or more, with not on anticoagulant or history of left facial appendage occlusion. It only fires to PCPs or cardiology because you don't want this to fire to DERM and GYNOC and people who aren't the appropriate person to either start the blood thinner or refer. It also offers an easy button. It refers to the appendage nurses. In fact, we have some right here who make everything happen and look at the patient, why aren't they on a blood thinner. They can order the CT, the appendage, review with the MD, get the patient in the clinic, and it's a really great option. We've started thousands of patients on blood thinners and closed over 1,000 appendages. If you think about it from a population level, that's a lot of strokes prevented, peripheral emboli prevented, and things like that. You want to be very careful and not be the boy who cried wolf because if you make a BPA that fires too often, people will tune it out. I think a BPA should be surprising every time you fire. It should be like, oh, I haven't seen that in a while. What do I need to do with that? It shouldn't be something people can automatically click out of. We do want to be mindful of how much of people's eyeballs' time we're taking up with BPAs. But if you haven't looked into this yet and you see a need in your population, I recommend looking into building your very own BPA. Now the other side of that coin, BPA the bad. In our practice, we read over 7,000 monitors per year, and the EPs read all the monitors. Unfortunately, most of those patients have BPAs with such important emergency things popping up as this patient has CKD stage 2 and you haven't addressed that in the preceding year. Do you want to do that, doctor? No, I really don't, partly because it's CKD stage 2 and partly because I'm not the patient's doctor. If it's an EP patient, that's one thing, but most of the monitors are ordered by primary care doctors who have a 19-year-old who had palpitations at a party, and they don't have any real cardiovascular pathology. So if we're thinking about who is the patient's doctor, it's not me. I'm really functioning more as a radiologist interpreting a study. If you do the math on all those extraneous clicks, that's over 300 EP doctor life minutes spent per year for zero value added to patient care. And remember, that's what inefficiency is. It's things you do that don't bring value. So it's a big problem, and one of the messages I've taught my kids, which they seem to have taken possibly too well, if you don't ask, the answer is never yes. But if you do ask, sometimes it is yes. That's how I got your mother to marry me, and it worked out well for me. Remember that admin I showed you? He's not really angry. He's Andrew Burke. He is the best admin ever, as you know. And he can solve any problem with one email. So what I said is, Andrew, when the radiologists are reading their chest X-rays, they're in all the charts, are they getting these BPAs? It turns out they're not. And just like BPAs can be turned on with the flick of a switch, they can be turned off with the flick of a switch. So if you're having frustration with the EMR and you hate this thing about it, see if you can find someone who can make it all go away because it might be able to be done. Last tip, laser-focused patient communication. I had a 50-year-old man who had a PVI with radiofrequency energy to the posterior wall. One month later, he went to the ED with three days of fevers, chills, and somnolence. Very concerning. The ED did a viral panel and a chest X-ray, and they were normal, so they were going to send the patient home. I happened to be strolling through the ED to see another patient of mine when I saw them. I was obviously super concerned about an AEF. I ordered a CT with oral contrast, and you know what it showed? It was nothing. But what if it wasn't nothing? What if they had an AEF and we didn't know because we weren't looking for it? So I talked to the patient. I said, how did this happen? I talked to you about this. I talked to your wife. It's in your discharge instructions. And the patient said, you know, the anesthetics, they have antigrade amnestic effects. I don't remember. My wife was thinking about groin hemostasis and what she was going to feed me for dinner, and the discharge instructions you gave me was 14 pages of garbage. If there's one little golden nugget in there, I'll never find it with all the other instructions about now that your ablation is over, you can restart your multivitamin and echinacea and all this other garbage. So what we need to do is find a way to get the message to the patient. And I realized that I have a superpower that no one else in the world has, at least for my patients. And there's one document that I write that, and it's not the clinic note, that patients read. They print out. They keep it on their nightside table. It's their background on their iPhone. They tell everyone. And that's the operative note. So the message I want my patients to get, I put right in my operative note. And this is what it put. And this is what I write. And for AEF, it's very explicit. It says what it is. I want you to watch out for it. If you go to the ED within the next 60 days with any of this stuff, I need to know about it, not whatever non-invasive imager is on call who isn't thinking about it. So you could definitely save a life with this. Now, it hasn't come up because we're using PFA, but you can use this for other things as well, especially I have patients who are very overweight. I put pictures of their ice images in their operative note. And I said, look at all this pericardial adiposity and the inflammation and the adipokines. You're going to have more AFib if you don't lose 100 pounds. It's a great way to communicate with patients. If you also have this superpower of writing operative notes, I encourage you to use it as a laser-focused tool. In summary, the golden age of medicine is today. There's never been a better time to be an EP. The EMR works for you, and we need to stamp out EMR inefficiency. Try these tips. Use the AI scribe. Use stickies. Build your own BPA. Ask for what you want in life and in BPAs. And use your op note for laser-focused patient communication. Thank you. Thank you very much. That was very entertaining and very useful, actually. So we'll come to questions at the end, but it's our pleasure to welcome our next speaker, so Dr. Mehdi Rezavi from the Texas Heart Institute, who's going to talk to us. We're going to change gears, actually. We're going to go to robotic procedures. And can it facilitate remote procedures? So looking forward to this talk. Thank you. The bad news is he was not the last speaker. You have one more before cocktail hour here. So here we go. Thank you for watching! I have been tasked with the concept of discussing robotic procedures and whether or not robotic procedures can facilitate remote ablation. We all have talked about remote navigation of catheters, introduction of the robotic systems for cardiovascular procedures. We all know that from the physician's stakeholder perspective, what the benefits are as interventional or invasive cardiologists. We know that the risk of orthopedic surgery, injuries with lead and radiation-related complications are real. And until now, remote in the sense of not being at the patient's bedside during ablation was thought to be actually a reasonable way of mitigating some of these potential complications that can accumulate for the health care provider. The other part of the stakeholder is also that patients can benefit from this. And in a second, I'm going to kind of discuss the benefits that the patients will have. So the remote robotic ablation systems are precise, and they can also be very stable in terms of the ablation. We know from clinical trials that the efficacy of ablation with the remote systems versus without them is roughly equal, although there are some subtle changes or differences. So the first two are from the perspective of, I think, mostly the stakeholders being both the patient and the physician. You want improved catheter stability. You want precision in your ablation. The third one, I think, is mostly for the operators. And I think we have to be cognizant there are certain things that we have to take care of ourselves. And there's nothing wrong with that. I think that a lot of these talks, often we start focusing almost out of a sense of guilt not talking about the benefits to the allied health care professionals, the fellows that are standing next to us, the nursing staff that stands next to us. And so there are benefits for us for remote navigation of ablation. It also helps with the combined stakeholders. So as a patient, you want your operator to be trained well, to understand what they're doing, and to do so with minimal amounts of practice or practice on a clinical situation. You want to have a resident or a fellow whose training process is somewhat faster than the rest. So in that situation, you have shorter learning curves, and you can have more complex cases to do with the remote navigation systems. So there's two types of remote navigation systems. You have the magnetic navigation, and you have the robotic navigation systems. They are, as I said, they have efficacy in terms of clinical success, about 95% similar to catheter-based without the remote navigation. The studies have shown significant reduction in operator radiation exposure and shorter fluoroscopy times. But these days, that doesn't seem to be as much of an issue as we are introducing fluorophore or fluoroless procedures. What is really important is how this can translate to the patients. And so when we're talking remote in one sense, you're talking about sitting there having a coffee, perhaps, behind the glass partition as your fellow's doing something, or if you're kind of moving the joystick around. And then you have remote ablation, where you're trying to get and penetrate a patient population that has not, who's underserved, who probably doesn't have access to all those fancy technologies for ablation. So in that sense, remote ablation, that changes the tone of our discussion a bit. And it really does, we have to look at both the clinical impacts and the public health impacts. And we'll see in a second, it's really not that straightforward. Of course, the clinical impact, real-time precision, again, we talked about precision, stability, control, high-quality surgical outcomes, despite the geographical distances. I know there are some companies that are actually doing this in Africa right now for remote ablation. And the public health impact, obviously, it expands access to expert care in remote, underserved areas. And it really has the potential for addressing healthcare disparities. But it's not that straightforward. There are issues that we have to look into and issues that have to be looked at. First of all, high-speed, low-latency networks are absolutely key. Even if you have high-speed, latency is a problem. Obviously, with any medical procedure, it is just, that is not forgivable. You have to have remarkably reliable and rapid connections. Real-time data imaging, secure communication, kind of the same as above. Another thing that's not talked about often is haptic feedback. The ability to provide haptic feedback is not only, it's a technological kind of a barrier, but it actually, if we can figure that out, it can help with training the physicians, and it can help with the safety of the procedures, too. So that's another technology to be looking at. Keeping in mind that these things are not cheap, robots, software, support, training, all of this is expensive. So if we're trying to penetrate areas that are underserved, there has to be some form of understanding that there will be, initially at least, high implementation costs for this. There's currently no standardized training or certification that has to be worked out. There's going to be, obviously, a steep learning curve for the local teams. The gaps are real, and they can, at times, be disabling to the delivery of care. And then, of course, what we all dread is what happens if you have a complication remotely. There are some things that are also not noticed or not mentioned as much. It's really hard to develop a surgeon-patient relationship when you're ablating or doing any kind of a procedure remotely. And I think that has to be taken into consideration. The confidence of the patient, I always tell my patients, this is a conversation I have with every single patient who's having a procedure. I tell them, you have to want me to do this procedure. If you feel that I'm trying to corner you into a procedure or if you feel that you want someone else to do a procedure, we're not doing a procedure. I don't know how I can get that response if I'm doing this remotely. But that's just a problem thrown out there. And, of course, there is the element of patient vulnerability, especially if you're going to an area that is, they may look at you as potentially, you know, we have the expertise. There is an imbalance, right? There's a knowledge imbalance, a power imbalance that comes along with that. Informed consent becomes more of an issue. There's the technical risks that we just kind of talked about, especially latency. That really is a major, major problem for some of these systems, especially, as you know, in EP. You see heart block. When are you going to come off ablation? Within half a second. And then, you know, these things are real problems. And then fallback plans for emergencies. And then finally, you have the legal and jurisdictional issues, responsibility, licensing, reimbursement. We don't know exactly. These have to be clarified, too. So what are the future directions? AI integration. Most of the speakers here have been discussing that. But as they also have noticed and noted, we have to be very careful with this. We have to look at, in addition to that, augmented reality. That I think is very helpful. And I actually think that's a great way to train doctors locally, local doctors. So at the point of care, you can have physicians. I actually think ultimately before we do remote ablation, we'll be able to remotely train people to do this. So that's going to be very important for us. 6G enabling near zero latency operations. That definitely, I think, is good. I don't think until we have 6G, we're going to be able to do this. Haptic feedback, we discussed. And deployment in rural and low resource settings, also. And with that, I believe that's it. Thank you, Dr. Rizalbi, for that wonderful presentation. We've had a really impressive spectrum of topics pushing the envelope, would be sort of the general theme. If there's anyone in the audience who has any questions, feel free to come. utilize alert-based monitoring in your pediatric population, and how do you decide which patients are best for this model? So actually, this is probably for Dr. Lal. Great question. So it is something that we're trying to transition to. Unfortunately, I am part, fortunately, I am part of a larger program, and so we do use one remote monitoring sort of system, and so we are in the process and talking about transitioning, but we just, it's, Vanderbilt has both a very large adult and pediatric population, so it's a challenge. So currently, we are not transitioned to an alert-based system. And the second part of that is how you decide which patients are best for this model. Yeah, so I think it, number one, is if you have patients that have been, and this goes, I think, it just doesn't have to apply just to pediatrics, and I have my adult colleagues also comment, but I think if you have patients that are stable, and you know that, for example, that their care isn't changing, it's, using alert-based monitoring is great. I think, number one, for ILRs across the board, that should, that should always be the case, but for pacemakers, things like battery depletion or lead integrity, that should always be something that you get alerted for, but if you have, so for example, pacing less than or more, you know, like for a certain percentage, for my pacemaker-dependent patients, it's gonna, those alerts don't always, are not always applicable, or if you have, so a lot of times, a lot of these settings are set for adult patients, so having tachy alerts above a certain number is not going to be as applicable for my patients that have baseline sinus tach, because they're a six-month-old with post-op heart block, and their rates are going to be in the 150s, 160s. So I think it's really important to think about, for me, specifically, what, what the indication for that device was placed, if it's a channelopathy versus if they're a congenital patient, and what am I worried about, and so I will base alert, I think you activate the alert that is going to impact your change in therapy for that specific patient. I don't know if that was too generic. I'm happy to be, if there's a follow-up to that, but, so like, I think the adult correlate would be, if you have patients with AFib, and they have chronic AFib, and they're anticoagulated, you're not going to be doing anything else. You may not need to be alerted every time they have an AFib alert, because that's where they sit. Wonderful. Thank you, and as maybe the lone heart failure provider in the room, and Olivia Gilbert, by the way, I should have introduced myself in the beginning, we sort of jumped in because we knew that Dr. Singh had to go, but I'm a heart failure provider at Atrium Health Week Forest Baptist in North Carolina. What would be your message and or request to your heart failure colleagues with remote monitoring in terms of relationship and interoperability? What would sort of your message be to your heart failure colleagues? Yeah, so I think we have, I mean, our devices now, we receive so much information, so I think, and again, I think on my specific institution, we have a very great model, and I think using, making sure that we are collaborating with these patients, right, using the OptiVol settings and things like that to really understand how is this essentially changing their clinical status, so using that as part of the algorithm to understand not just their BNP and things like that, but using their device settings or their information from their device as part of that algorithm, and I think it really is a collaboration, whether it's from, you know, lab work, data, but also within physicians, making sure that we're working together to optimize the information to provide patient care. I mean, I think one of my last slides was looking at all the different type of sensor data that's gonna be coming from our patients, and I'm sure, you know, whether it's from, if they have a CardioMEMS or whatever their device is, right, there's so much coming in from these patients, and this is where I think, and Jag sort of talked about this, but AI is is helpful because it has the ability to, and you didn't specifically ask this, but I'm sort of going down this line, but AI has the ability to help us go through data and figure out what is helpful and useful versus not, but it's the data, we have to be an active participant in putting that information in. I think there was a session yesterday when we were talking about AI as well, which is it, we can put bad data into it, and then it's not going to actually give us, we're not gonna be able to effectively take care of our patients, so I need to make sure that I am actively part of that conversation when it comes to the PEDS patients, because it's gonna be a very different model if it's just all adult data going into these neural networks. Thank you very much. Thank you. Can I ask a more general question, maybe for you two, because you talked about remote reprogramming from your side and you've talked about remote procedures. So in five years time, are we going to be in a position to be starting to roll these types of things out, and what needs to happen with the infrastructure, because you talked about cybersecurity, you touched upon it, and you talked about, for example, connectivity, and I don't know the timeline of these things, but I wonder if you guys have an idea of where, and the trajectory, and the speed of achieving those final goals. I certainly think doing remote ablation versus remote programming of the device is probably gonna, it's a little more challenging with ablation. Frankly, I'm a little, I'm not as optimistic right now for five years. I think that the technology is not quite there yet. I think ablation, specifically ablation, you're talking about, you know, doing something where you need split-second decisions sometimes, right, and I just think that that's probably not ready for five years. I do think we're gonna get there, but probably closer to a decade before we really start seeing this, and, you know, with the world, maybe longer than that, but yeah. A question from, sorry, I should come to yourself about the reprogramming. I think the reprogramming is, and I think if you sort of use the caveat that maybe you use indirect remote programming, where you actually have another person there, and it may not be the electrophysiologist guide, I mean, the electrophysiologist is guiding it, but it might be a nurse, or it could be the general practitioner who's there, and they could sort of be that intermediary. So there's still a provider, and so that could happen, I think, in the near future, but it is dependent upon the security of the system, and that I think is going to be independent, or dependent upon where you're practicing, because that study that I highlighted, they had the video uplink, and it was, you could see the screen, and the reprogram happened within, like, a second. So there was speed, but again, and that was, they tested it as far as, like, 5,000 kilometers. So again, it's, if you have a good, and it goes back to what you're saying, if you have a good system that allows immediate connectivity, and that's consistent, right, and from the internet, then I think, then you can do it, but the question becomes security, and there are different standards, like, for example, remote monitoring in the US has very different standards than remote monitoring in Europe. And so I think the constraints upon that, it's going to be based on something that's even beyond our society. And you would presumably need all those things remotely happening, guests during office hours. So there is some, as you mentioned, someone, a provider, a human, that makes sure that the, let's say, it's occurred and fully, or not being sort of, you know, cut halfway through because of connectivity issues. You need someone to check that it's actually happened okay. I think you do, Bill, at least currently. I mean, I think you have to make sure that there is going to be someone responsible in the event it didn't go well, right? And that sort of goes back to ablation. So I think if you use the word indirect, whether it's indirect, maybe ablation, so having a provider there that you have trained remotely that you would still be part of, and I think maybe we are going to get there. But as we're, the technology isn't, it has the potential to be there, but I think the safety issue is what we still all need to feel confident about. Yeah. I would like to add to the doctor's point. You are exactly right. I think training is where this is going to, we'll see the first results with training the physicians locally before actually doing procedures. And it may not even come to that if we can train them. That's wonderful. Great answers and great discussion. Can I have a question for Matthew? Because I think all of us want to know about the Epic scribe that you mentioned. And the specific question is, does the institution need to add it onto their package that they've paid for? Or is it preloaded? And if so, how do we find it? Yeah. So you have to talk with your IT department. It costs money. It costs a monthly license fee. For us, it was an easy sell because prior to that, we were using offshore dictation that I think was something like $12,000 per year per license. And this, I think the price is about a sixth or a seventh of that. So it was an easy sell for us, even before we knew it was so much better at quality. But yeah, there is a cost to it. It's not part of Epic by default. It's a Microsoft product that's embedded within Epic. But it's definitely something worth talking to your institution about it. You probably won't be the first they've heard of it. I think it's spreading like wildfire amongst people who've used it. And there are competitors as well. I mean, you can buy your own AI scribe outside of your institution. I don't know how that works with, you know, cybersecurity and HIPAA and everything like that. But I've seen ads direct to a physician, you know, trying to get your own AI scribe for your clinic notes. And as a heart failure doc with our notoriously long HPIs, it's been life changing, truly life-altering. Yeah, from a clinic perspective, it's been wonderful. If there are people in the audience, unlike Matt, I don't get my notes done on time. But the really nice thing about the DAX Copilot is that because it has that sort of auto-dictated, you at least have...you remember what you talked about. So again, I don't get my notes done on the same day, maybe a couple weeks later, but then I know exactly what I talked about because it's there. So do try to advocate, I think, for your...and again, this is Epic-specific. So if you have a different type of electronic medical record, I'm sure they have an equivalent. And so you should look into it. What's that called? Did you say... DAX. Dragon Ambient Experience. Okay, great. Any other questions from the audience? I think we've exhausted the ones that have come through the app. If not, I think we'd like to thank our speakers for a fantastic session. Looks like there's one question. Oh, beg your pardon. Sorry. Yeah, thank you. I was just curious about...you mentioned robotics. One of the hurdles that needs to happen is haptic feedback. Yeah. How do you imagine that looking or feeling like in a robot remotely? Yeah, it's challenging. I think the haptic feedback, I'm mostly concerned about the speed of this. I really feel that the amount of...you know, the rapidity with which it has to happen, because a haptic feedback, think about it. If you're just falling behind by even two seconds and you're making the wrong move, you're behind...it really can cause major problems. So I don't know the answer to that. I suspect, again, that between training locally and getting six...I don't think also haptic until you have 6G, you'll be able to really, really have that rapid feedback. But I still go back to training doctors locally. I think if we can do that, you know, the old saying, you know, you feed them a fish for a day, but you teach them how to fish, and then they're set for a lifetime. I think the precedent is there. I think they're on like their sixth generation of the DaVinci robot, and I think it gives really good haptic feedback. I've tried it, you know, unrelated to what I do as an EP. But as you point out, the timing matters. If you're six feet away and it's hard wire, that's very different. Very good. I think, therefore, lovely conversation and discussion. So I think we'll close that session and thank our speakers for a brilliant presentation and discussion. So thank you very much.

electrophysiology

artificial intelligence

multidisciplinary care

healthcare innovation

remote monitoring

AI-enabled devices

algorithmic bias

electronic medical records

robotic procedures

augmented reality