Hi, I'm Susan Etheridge from Pediatric Electrophysiology, and I am pleased to join with my co-moderator. Arun Sridhar. I'm an adult electrophysiologist from Tacoma, Washington. And you're at a debate session where we're going to look at smart remote monitoring tools significantly enhance the efficiency and accuracy of pacemaker and ICD alert management and are cost effective. So we have two speakers. Our first speaker I'd like to introduce is Dr. Omar Kradeh, and he comes to us from Tulane University. So welcome, Omar. Thank you for the introduction, and thank you, everyone, for being here as part of the session. So I'm just going to wait for my slides to load. So I've been tasked with making one side of the debate with regards to smart monitoring tools in terms of the ability to manage remote monitoring-based alerts and interrogations. And I think it's part, so these are my disclosures. It's partly telling that when I had some relief when I found out which side of the aisle I had to make an argument for, because who doesn't want smarter tools, right? So that it's almost kind of makes it a little bit easier. But what I'm going to try to go through is, I think, first of all, it's good to understand why remote monitoring is so important to go back to the roots of the initial trials and the data that supports the role of remote monitoring. Because if we don't need it, then we don't need tools to enable it, right? But then I think we're going to go through some of the challenges that come at the expense of being able to provide these benefits, and then how we can use modern-day tools to kind of try and get around that and have better, more efficient programs. And so the first slide that I'm going to start off with is some of the initial trials that looked at the impact of remote monitoring and allowing programs and patients to reduce the number of clinic visits. And so these were sort of the trials that kind of showed us that it's possible. And what you can see in a bunch of these early landmark trials, you can see that by randomizing people to less in-clinic routine scheduled visits and doing instead scheduled remote monitoring, you're able to reduce the number of clinic visits that they have to come and see you in. And that's obviously very important, A, because it relieves the work of physicians, but also patients don't have to take time off. There's a lot of cost improvement, both from health care's perspective and also from people being able to continue to work and not have to worry about, you know, taking time off of work and all these things. But in addition to that, the benefit isn't just in reducing the specific clinic visits that people have to do, but also by acting early on some of the alerts, you can reduce some other health care utilization. So this is a figure that you can see from the EVOLVO trial. And what you can see with me is the generalized health care utilization in that study. But what you can also appreciate on the bottom part of this figure is that people who got remote monitoring tended to have far less urgent visits, either to the hospital or to clinic, which, you know, if you agree with me, typically tend to be more expensive types of care. And there was a shift towards more sort of routine-based clinic visits to adjust things. And I think that that's something that we see time and time again in different remote monitoring studies. There's another study, this time looking at, well, what if people are hospitalized? Well, having remote monitoring enabled typically was associated with reduced length of hospitalization. So why could that happen? Maybe because, you know, you can trust that these patients are going to be monitored, they're going to do better after. Maybe they were better optimized. There's different ways by which you can think that that happened. But this is, you know, another way by which health care utilization is somewhat reduced. So you're not only reducing the number of times you seek care, but also when you end up in the hospital, you're ending up staying less. And that translates to significant hospitalization savings just by length of stay alone. Beyond that, I think one of the more important things with remote monitoring is the crux is in being able to act on actionable alerts and events pretty early. So there are certain things that I think most of us will agree as EPs, we want to kind of see and act on pretty, you know, robustly. Things like VT and VF events, people who are hitting end of life, you know, maybe people who are having noise that's triggering shocks, all those things you want to be able to act on pretty early. And what you can see here in two different studies, the time from a device-related alert to intervention is much, much, much faster when you're using remote monitoring as opposed to just routine and device clinics. And ultimately, there's also concomitantly when you're being able to see these things early before they translate to bad events, for example, maybe seeing noise early before people get shocked or seeing non-sustained VTs of escalating burden before it translates to shock. So in one of the studies, we see here clear, you know, evidence for a reduction in both appropriate and inappropriate shocks delivered by these devices related to getting these remote monitoring events. So a lot of benefit, but is there any adverse events from not seeing those patients in clinic more frequently? You can see here across the board, no real, you know, even sign of increase in adverse events when you try to shift the burden of monitoring to remote systems. So a lot of benefit that we've seen with those studies, but when it comes to real-world implementation, there's a lot of challenges that come with that. And this is, I think, an important study that I'm sure is going to be, you know, used maybe against me. But what you can see is that the number of device transmissions that come as part of programs that do routine monitoring is just astronomical. And so you can see here in a figure, roughly, you know, 90,000 patients. And what that results in was 205,000 alert transmissions just from those patients. And what you can see is that the alert burden is not uniform, you know, across device type. And so you can see devices like ICDs and pacemakers tend to, you know, contribute far less than devices such as ILRs to this sort of overwhelming, you know, device transmission burden. And this is important because the difference is a two-fold problem. These, you know, ILR devices tend to cause a lot more transmissions, but far fewer of those transmissions require intervention, and far fewer are actually related to high acuity problems. And so the issue has become that all of us are getting bombarded with an endless stream of alerts and remote transmissions with only a minority having, you know, the really clinical relevance for patients in an actionable manner. There's another study trying to quantify, you know, the same thing. And what you can look at here is, again, a huge number of transmissions received across, you know, a relatively large cohort, but already 57% of those transmissions can be dismissed pretty early on by trained, you know, techs or by AI algorithms. And then only, you know, a small proportion need to be forwarded to physicians. And even after you discard those, quite a significant proportion of these don't have really high acuity alerts. And again, you see the same thing, loop recorders accounting for a very large proportion of these alerts compared to their, you know, the size of the loop recorders within, you know, the framework of devices, but a much lower proportion of actionable or high acuity alerts. And beyond the burden of dealing with this data deluge that we get from these devices, there are other, you know, really important facets of remote device management that are increasingly recognized as being problematic or difficult to perform. And those are the silent work that happens behind the scenes to be able to have a successful remote monitoring, you know, program. And so what that includes, you know, is things like reaching out to your patients and making sure that they remain connected, having to look into the EMR to figure out whether this new AFib event happened in someone who's on blood thinners already or not, being able to upload the reports or do the appropriate billing. So there's a lot of behind the work, behind the, you know, the scenes work that adds to the burden of being able to execute these programs to get the benefit that we do want to get from them. And so the answer, I think, you know, given this baseline that we started with should not be that we abandon remote monitoring because we've already seen the tremendous benefit that that could offer for our patients in terms of reducing healthcare utilization and providing real, you know, good outcome benefit for them and reducing bad events, shocks, all those things. And so I think the key to managing those challenges is to lean on technology and more efficient tools to be able to better manage the burden that comes with these types of remote monitoring tools. So what are some of the tools that are available to us? So smart remote monitoring tools, we have universal management software. So those are softwares typically that are offered by a third-party vendor that tend to talk to all of these different, you know, manufacturers and kind of unify all of the transmissions into a single portal where it can be much more easily handled by the physicians or the group that's managing these devices. You have tools that allow you to integrate into the EMR, artificial intelligence-based tools that are, you know, relatively early on now, but, you know, some signs that this might have an increasing role within this space. And then some tools to help with billing and reporting and patient connectivity. But what I'll start off by saying is that I think the most important smart tool to help and manage those devices is our own approach of what kind of alerts are we listening to, how do we manage these patients, and how do we program the alerts from the get-go to kind of have a better way of managing them. So the best way to have to deal less with non-actionable alerts is to program them off and not have to worry about the added burden of having those alerts that you know are not going to change your management. And you can see here in one study doing such an approach of just deactivating non-actionable alerts reduces this burden by up to 48 percent, even just by over one year. And so one of the biggest challenges for me and probably the impractical side of potentially using sort of the legacy sort of approach to remote monitoring is managing all these different portals, passwords, usernames, different websites. It's just practically not possible within, you know, a limited amount of time and scope. And so in practical sense, the thing that's not quantifiable is this idea that these smart platforms can move you from having to access four or five different portals to just being able to access everything you need in one place and adjust your settings and look at all of your patients and triage all of your patients in just one place. But beyond that, all of these different vendors have offered different tools that can sometimes help you deal with the burden of these devices in different ways. And so we already alluded to the role of AI, and this is, AI can play a role in different ways. One is by summarizing the reports that are generated by these devices. But another potential role in the future and maybe partially now is to oversee the electrogram interpretations, you know, in addition as a second barrier to what the devices provide. So this is one study looking at a machine learning-based algorithm from one of the vendors that took loop recorder-based electrograms and looked at them to try and filter out false positives as they were classified by the loop recorder. And what you can see is that it was able to correctly identify 215 of 283 false positives that were, you know, provided by the manufacturer. And so already helps tremendously by limiting the burden of false and inactionable alerts. Now, I would say here and caution that this was looking at primarily some of the first-generation devices, and so it's not clear, you know, as the manufacturers have moved on to using a lot of AI for themselves, whether third-party vendor AI is going to be able to limit that even further. But I think it demonstrates, you know, the potential role in the future, and this is something that you can foresee can extend beyond just loop recorders and AFib, you know, whether it's going to detect noise, you know, better than what the manufacturers do with their alerts, et cetera. Here's another study looking at these devices, taking different remote monitoring alerts for, for example, AFib, and instead of providing an alert for each AFib alert, you know, provided by the manufacturer, it tries to summarize it into clinically meaningful events. So for example here, increasing burden of paroxysmal AFib instead of telling you each time they go into AFib, changes from persistent to paroxysmal, and what you can see in this study is clearly demonstrated a reduction in the burden of alerts from somewhere, you know, 13 and 18.3 on the left-hand column as transmitted by the traditional CID manufacturers to much lower numbers of really more clinically meaningful and actionable alerts. So beyond that, we talked about some of the things that have to happen in the background of managing patients within remote monitoring systems to be able to keep them connected and kind of get things, you know, moving in the right direction, and so some of these smart programs have, have devices that allow us to do that more efficiently, so this is one example. This is a program that allows you to smartly identify patients who have been disconnected and are not transmitting on time. And then what you can see here is that it automatically sends text messages or phone calls to try and re-engage those patients. And that way, firstly, allows you to maintain patient connectivity throughout long periods of time, but also removes that burden from your staff by doing this automatically. What you can see is that this translates over time to increasing patient connectivity rates, even over a short period of time of one year, by using such automatic patient connectivity groups. And what you can see here is if you use that sort of software, it does almost as good or there's no real difference between that and a person doing a live call. And so it really does allow you to take off the burden from your device clinics. So does that translate in real life? Maybe the theory is it's pretty good. We want the goods, but we don't want to deal with all of the bad challenges that come with it. But does that translate to improved outcomes? And this is a very nice study published recently by the group of Varma et al. And they looked at a French health care registry, all comers. And they split the registry into three different groups. So on the left-hand side, you can see a comparison of patients with devices that are managed using remote monitoring as compared to patients who are managed without remote monitoring. You can see an idea of potentially a survival benefit to remote monitoring when used in the real world in large numbers of patients. But more interesting is even amongst that group that already has a survival benefit compared to no remote monitoring, if you break it down into groups that use a unified universal monitoring system as opposed to people who use sort of a conventional legacy remote monitoring, what you can see is, again, further separation even amongst that subgroup to show potential survival benefit on top of just remote monitoring by using this unified remote monitoring. You can see here the effect estimates pretty significant. So of course, observational study. There could be confounders. Programs that are invested into getting the latest and greatest may be more invested into following these patients closely. But it does lend some credence to the idea that being able to have these smart systems can help you manage your patients better and get better outcomes. And what you can see here is in one of their analyses, in addition to that, there was also a 4% reduction in costs, not necessarily just from remote monitoring, but from using this universal sort of smart tool. And so just briefly, smart monitoring tools. Remote monitoring reduces clinic visits without increasing adverse events. There's a potential link to reduction in ICD shocks and hospitalization and improving the reaction time to actionable alerts. And it's generally cost effective, mainly because of reduced health care utilization and hospital stay. The problems that come from it are burdensome and logistically challenging, a deluge of data. But the answer should not be that we abandon remote monitoring, but rather that we use appropriate smart tools to manage this deluge of data more robustly. Thank you. We thought we would take questions at the end. Thank you very much, Omar. It's my pleasure to introduce Nina Isakardze. She's from John Hopkins. She's an LLTP. She's going to talk about the challenges and benefits of a child. She's going to say that smart monitoring tools present more challenges than benefits, compromises patient care, and increases the health care cost burden. Hi, everyone. It is fantastic to be here today. And the title that you just presented made me wonder if someone is after me or set me up with this title. But joke aside, you know, Dr. Currie didn't just presented wealth of data that shows that remote patient monitoring saves lives. So I'm not going to tell you that remote monitoring should go away and we should go to 1960s when we didn't have it. But hopefully I can showcase that there are significant challenges and echo some of the sentiments from your talk earlier as well. So these are my disclosures. So I want to take a step back and just remind everyone our journey with the remote patient monitoring. So about 10 years after the first pacemaker was implanted, we had trans-telephonic transmission of the device data. And this was done with this device, which looks so old now. But the patients were able to put a magnet on their pacemaker and then connect this electrode to it. And then through this cradle, connect the telephone and dial in the device clinic. And this way they were able to transmit only very limited data. This was only like mainly about the battery life because the devices at that time had like shorter device battery. So then in about 1990s, we had one base transmissions. And this was enabling patients to do interrogations instead of coming into the clinic to do this remotely. But they needed to be physically putting the wand on their chest. And then in 2000s, we moved to the completely wireless systems, which we're using now, where it enables us to do this alert-based monitoring in addition to scheduled interrogations. So again, the definitions, I think it's important to realize that remote interrogations is a separate term as a remote monitoring. Remote interrogation refers to those scheduled like every three months transmissions, where remote monitoring refers to the alert, pre-specified alert-based monitoring of the patients. And just the workflow, we are all familiar with this. But we implant a device. There is some nominal alert systems that we can turn on or turn off. And then each company has their own remote monitoring systems. And then we have this universal platforms that ingest the data from device-specific transmission systems and give us the acuity-based alerts and also pre-populates the report, which is then educated by device clinic nurses or device clinic staff and then finally approved by clinicians. So again, I'm not here to tell you that remote monitoring is bad. This is a study in time, a randomized controlled trial, which was randomized about 700 patients. Most of these patients had CRTDs. And you can see that it actually decreased the composite outcome of mortality, hospital admissions, heart failure, worsening of heart failure. So the remote monitoring improves survival. We should keep doing this. These are the challenges and opportunities that we can address. So the first is alert programming. So this was mentioned earlier as well, that we have this laundry list of alerts that the devices have programmed. Some are mandatory that are programmed on all the time, like the battery failure. But then there are some other parameters, lead integrity, arrhythmias, that can be customized and programmed on or off. But mostly, nominally, it's turned on. So this is a large study out of Australia of 25 centers of 26,000 patients, a retrospective study that looked at the transmissions from 2019 to 2019. And what they saw was that there was a significant overlap. So each clinic sets their own alert acuity thresholds and what they consider to be the red alert, which needs immediate attention, or yellow alert, which needs a can, is not urgent. And as you can see, most of those discrepancies between the alert acuity was based on arrhythmia detection. So atrial fibrillation, VT, VF. And we can all resonate with this because AFib in a patient who has a de novo AFib versus a patient who has recurrent AFib means different things in terms of urgency. But how clinics set this up are very different. So because we don't know which one of these alerts are significant, then this creates kind of additional conversations with the clinicians. Then you need to go in the chart and look at the chart and understand what the patient's condition is. And some of those you find that are non-actionable and patients already have existing diagnosis of AFib, for example. So what is an opportunity here, and which is already the work is being done but we're not there yet, is to have systems being able to integrate the multimodal data from the EHR, taking the patient information, the age, and duration of device detected AFib. I know Noah and Altresia were kind of controversial. But still, take the patient characteristics imaging data and then customize the alerts based on that. Like if someone has already existing AFib in the chart, then don't keep alerting and so forth, like short NSVT runs as well. So now we move up to the deluge of data that the devices are alerts that are generated. This slide was shown by Dr. Kuradian as well. So per each pacemaker in ICD, there's two alerts per year. And for loop recorders, it's four alerts per year. So you can imagine if you have a practice with hundreds of devices, you're dealing with a lot of data. And most of the majority of this is yellow alert, which still requires the device clinician to go into the chart and spend time and understand what they mean. So if you're seeing a lot of yellows, and then you will have like kind of alert fatigue that you see this transmission. For example, this is like a real example where this was marked yellow. But because we're seeing so many yellows, the device clinic, this was kind of missed, where they looked, oh, the previous interrogation also had high thresholds. So let's move on. And thankfully, when it came to clinician check, then it was captured and addressed. But these are sometimes things that a lot of data, a lot of alerts can have the opposite effects. So one of the interesting things is that out of this about quarter million transmissions in this study, in this Australian study, where 40% were alerts, but the 60%, so more than half of those transmissions were from the scheduled interrogations. And there is data from another study that shows that the transmission information from those scheduled remote interrogations, only 6.6% of those data is actionable. Or do we actually change something based on those transmissions? So this begs the question, do we need to be doing both, remote interrogation plus alert-based monitoring? Or can we move to just alert-based monitoring? So the importance of this question is highlighted by the fact that there is a PCORI-funded study that is being in process. It's called Raptor CID, which is going to look at this exact question to see if we can minimize the amount of transmissions we're getting in the clinics and have more streamlined workflows. So the next point is cybersecurity. So it's considered that the device data hacking and cybersecurity is very, very low risk currently. But this is the paper from 2008 that showed that with a wireless connectivity, with a WAN-based, there was less risk. But when you have the wireless connectivity, the risk of cyber attacks can be higher. And these are the least of scary things that potentially could be done remotely if you have transmission-based systems, including inducing BF and reprogramming devices and getting the patient data. So this just highlights that we should be vigilant as the technology is evolving, that we continue to have very strong cybersecurity protection. Another aspect is how do we set those pre-specified alerts, and what are the caveats with that? So for example, you have a patient who is monitored, and you have a capture threshold of 2.5 preset, which is a nominal where if you reach the less than 2.5 safety margin, that's when it's going to create the alert. So it's not going to create alerts if you're heading in that direction but have not reached the alert yet. So this is, again, a real-life example of the trends of the capture threshold of this patient's device, where you can see that since the previous interrogation, the capture threshold has been slowly creeping up. And it was marked green in the shared platform that gives us initial report. And it passed through the device clinic staff as normal. And then on the clinician review, it seemed like too high capture threshold, so I wanted to look back at the trends. And clearly, there is uptrend. And then I'm not showing here, but there was a sensing was coming down as well. So this kind of highlights that those binary thresholds that are preset are probably missing some of the important data that can be captured even earlier. And we can further improve the safety of this monitoring systems. So instead of, again, this is where artificial intelligence can really be powerful, where you can look at the trends of, OK, so this amount of change in the capture threshold or impedance and sensing without meeting the criteria for the alerts is already concerning. And they should send alerts to the clinical team. So overall, I hope that it is clear. And we are kind of in agreement with this. I think we agree to agree that remote monitoring saves lives, improves patient outcomes. But there is a lot that we can do to further improve the way we are delivering remote patient monitoring. So this is the end of my talk, and thank you very much. Thank you. All right, thank you both. This paper is open for discussion or for questions. Anybody have questions? I'll start. Predict the future. How are we going to do this? What does 10 years from now look like in our device clinics with our remote modeling? Personally, I think we're probably going to, I mean, just as you alluded to, probably going to move away from routine transmissions, which carry a very high burden of work with very little actionable findings, and probably more towards alert-based monitoring. And then I think as data kind of concentrates with these large third-party systems, and they get this big data deluge that is burdensome for us, that's probably a treasure trove to develop good artificial intelligence that I think is probably, I would imagine, within 10 to 20 years is going to automate a lot of what we do in device clinics now. So that's what I would hope for, is that a lot of this can be much more safely automated with AI within the next 10 to 20 years. And harnessed for patient predictors and outcomes predictors. I think that gathering it seems like drinking from the fire hose now. But will this help us in the future manage our patients more predictively and better? Exactly. I 100% agree. And I think that there is also interoperability issues with the EMR. And obviously, there is this data that sits in those third-party platforms. But then how can we further enrich them with the EHR-based data? And that brings the whole conversation about how can we harness the data that is in EHR and is so rich to further improve the predictability of those algorithms. And how far in the future do you predict this will all be EHR-based? There'll be one system, and it all comes through the same system. That would be ideal. I think 10 years. Hopefully. But there is a regulatory aspect we should also think about. So if we're now talking about augmentative AI, where now the device is taking patient information and is classifying alerts based on that, rather than pre-specified alerts, then it has to go through regulatory approvals as well. So that will take some time. But it seems to me like it should be able to go to the EMR and say, oh, Johnny's never had AFib before. This is kind of a big deal. I'm going to raise this alert. Or Mr. Jones has heart failure. This four-beat run of VT probably does matter. Right. We have a question. Hi, thanks so much. Sanket Druva, VA, San Francisco. A couple of unrelated questions. Thanks so much for the excellent presentations. One is, I know this has been focused on remote monitoring, but I wonder if you might talk a bit about the potential of these in-person, routine in-person visits where we're bringing patients in every year, sometimes every six months, when we're able to get so much great benefit from remote monitoring. And then a totally separate question, I wonder if you might talk about remote reprogramming. We've done some work in VA where we've reprogrammed loop recorders and shown that if you're really on top of things, we're able to reduce the number of alerts. We're able to catch things and make a benefit. And wondering your thoughts about extension to other CIDs. Great. So I'll take the remote monitoring, the reprogramming for the loops. I mean, I think that is so necessary for the loop recorders. And I think there is, to some extent, we should be able to reprogram devices for pacemakers and ICDs in a way of alerts, adjust the alerts. But we do that now in the third party. We can adjust those things in the third party platforms right now. I'm a little worried about being able to remotely reprogram devices for outputs and modes and et cetera, because that is really opening up the cybersecurity problems. So that may be the challenge in contrast to the loop recorders that then we need to tell patients that there is a higher risk. Absolutely. I agree with Nina. I think this was explored in the past. And the cybersecurity issue is huge to allow for reprogramming of any therapeutic devices. So loop recorders are completely different. Like, I think we can modify or reprogram it without any potential for harm. But I think we are already there in terms of reprogramming the alerts through third party systems or the merge and within the device manufacturers themselves. So I think we are already there. I think there's some value also to patients showing up in clinic. I mean, you get to educate them. You get to find out how they feel in other ways. I really discourage people from putting in an ICD for somebody for secondary prevention of their arrhythmia event and then never really thinking again about what their disease is now that they have the ICD. And I would worry that that would be even more progressive or worrisome. Did we answer your first question? That was the answer to the first question about the value. That's a great question. Thank you so much. And what about the patient who had the secondary prevention ICD put in? Does that patient need to come to device clinic if they're consistently and continuously connected? Should that patient be seen in general cardiology, for example, for their heart failure management? For their heart failure management, yes. But if you're dealing with a secondary prevention, you don't know why they had the event. I don't think necessarily general cardiology. I think it's us. We need to see that patient and think about, did I do everything I need to understand what their disease is? Because we know more now than we did 10 years ago. And maybe they had that device in for 10 years. And I just worry so much. People put an ICD in, and they just turn their brains off. I think there is value in titrating the amount of in-clinic device visits immediately post implantation and later on. I think in the first year, as we are learning the patient as well as the device, there is more value in bringing the patient a bit more frequently. But as things stabilize and if we know that it is a stable patient, maybe that's where the AI systems can help us triage the patients and which ones can be monitored safely, remotely. So that could help reduce the burden on the in-person device visits. Yeah, I agree. I mean, I think there's still value. And you're going to have to continue to see the patient to manage the medical issues beyond just the device management. But even if the question is just about device management, I'd say that I think that the role of remote monitoring, at least for me, is to reduce the amount of in-person visits. But it's not to completely replace it. Because you can foresee situations where someone, there can be a breakdown where they lose communication with remote monitoring. And you want to have at least that safety margin of being able to have, at least in our practice, a scheduled visit once a year where you can review things and see them face to face and make sure, OK, you're still connected. We're seeing your stuff. I think we don't want to miss that patient who's not connected, who somehow, his phone number is lost or is not responding. And then bad things can happen. So I think there is value in continuing to see your patients. But you just don't need to do it as much as we used to before. Thanks so much. Thank you. Thank you. We have some questions online. I'm going to start with the first one. If you aren't resetting the statistics, you may miss a change in the trends. And the device only stores so much data. How do we get around that problem? How do we assess the trends if the device is constantly erasing it? Well, so I'm not sure. A lot of these systems, when you go in to look at the data, you can pull up the long-term trends in sort of a trend view. So oftentimes, this data is not lost. It's actually easier to see. I'm not sure if I've. In person, right? You can't do that. But even the one that I was showing, the trends of device capture threshold, that was from the remote transmission. So you can definitely see the trends. If you are looking at the EGMs, if someone had arrhythmia and you're trying to want to look at the exact EGM, and if there's a new answer, AFib, or it's like atrial high rate episode, and you want to look at the EGM to truly see that that's AFib, I mean, more than 100%. You need to bring that patient in and then look at the exact EGMs. Or if it's a patient who had an appropriate shock or things like that, but otherwise. Yeah, and also in the third-party systems, now we have everything stored to the eternity of the patient's life. I think as the computing is increasing, so is the cost of the memory is extremely low, exponentially decreasing over the last 20 years. So the storage is probably not going to be a problem anymore in the near future, actually, within the next couple of years. A question from the audience. Does either speaker know current and projected guidelines for physician billing for remote monitor event adjudication? Are those activities billed for at your institutions? We don't bill, but there is CPT codes for remote patient monitoring that can be used. But there is lack of education and actual implementation in practice. I don't know how you do it. Yeah, I mean, I think the institution obviously bills for the remote monitoring. No, but if you are on top of that, you have to call the patient. No, that's not separate. And you're just thinking of your time, not all those people that got that remote, not the technician's time or the people who we pay. It's their time. I think in our system, we bill the patient quarterly when we do get the remotes and everything else in between. If there are alerts, those are just uncompensated. I do have a question for Nina. So not thinking about the future, thinking about the present, are we there yet? Do you think we have a good enough AI system to take the data from these pacemakers and ICDs and integrate that with EMR and come to a conclusion intelligent enough that a physician can rely on? So I think there's two aspects, a great question. So one aspect is those binary alert systems. And I know device companies are already working on those algorithms where you can predict the lead failure. Instead of waiting for it to reach that 2.5 threshold for it to generate the alert, you can kind of predict the terms. I think a more complex AI thing where we're kind of a little further away is that personalization to patients in terms of arrhythmia alerts. And arrhythmia alerts are a significant portion of the AI alerts. So I think we're there close to that aspect of the binary alerts for the lead management. But for the arrhythmias and patient personalization, we're probably a little further work to do. Yeah, my worry is that we have so many. The majority are the arrhythmia alerts. The number of lead issue alerts are a minority compared to the number of arrhythmia alerts. And if you are relying on an AI system, it better be great, right? The risk of missing something that we don't oversee that AI has done is extremely high right now. And the natural language processing has improved quite a bit in the last few years. But still, integration with the EMR has been an extreme challenge with many of these remote monitoring devices, remote monitoring solutions. Yeah, I think that's true. But the natural counter argument is that already you have to rely on these alerts. I mean, so whenever a rhythm is classified as AFib on a loop recorder, you've relied on an AI algorithm that classified it as AFib and classified a lot. Not currently, right? I think right now it's all traditional algorithms. They're not AI-based algorithms. No, so a lot of the manufacturers have moved towards AI. But the idea is you've also classified other rhythms as normal rhythm. And so you've kind of trusted that a lot of the stuff that hasn't shown up for you to review was properly and accurately not selected to be stored. And so we already, even without thinking about it, are relying on these algorithms and technology to figure out what are the things that we have to review. And I think it's a similar step forward that if instead of this thousands of alerts, you just have to deal with 10, and you're relying that it was trustworthy, that it's only showing you stuff you need to see, I think it's not as big of a leap in terms of trust since we're already doing that to a certain extent. Right. So for the rhythm classification, including wearables like Apple Watch or Cartier mobile, et cetera, they have those rhythm. We are pretty far with the EKG interpretation, like electrocardiogram interpretation with those proprietary algorithms. But here, what the issue is, how do we classify alerts? Do I need to give this red alert and push it to you? Or it can be when you routinely check on it, that's when you look at it. And that's why it's going to be FDA, probably regulatory oversights, and when we have augmentative AI tools, whether they are actually accurate. Because they perform differently, also, those augmentative AI tools. If they are using, for example, academic institution where someone may have more protected time or less clinical, they are not reading like 100 echoes and nuclears, you need to make sure that this AI model is suitable for everyone. And it's not kind of only, doesn't relate too much on human input. So that could risk the error. I think we're all in agreement that all of this needs to be very well validated and very well regulated. But ultimately, I think the idea is we're increasingly using these tools and relying on them, and I don't really foresee that that's going to, I think the trend is going to be for us to have to rely on them more and more. Any other questions from the audience? Well, thank you. I'll thank to our speakers. Excellent discussion. Thank you very much. Thank you.

smart remote monitoring

pacemaker alert management

ICD alert management

AI integration

alert fatigue

cybersecurity risks

healthcare utilization

remote reprogramming