Great, thanks so much. So let me start this talk up. All right. I think it will go forward on its own. There we go. So I'm a cardiac electrophysiologist from Los Angeles. I'm not coming from too far away. We're going to talk today about remote monitoring of heart failure. What are we measuring and why? So I think when we talk about remote monitoring of heart failure, most people will think of the PA pressure monitor, right? And we sort of all thought maybe that might be the Holy Grail. But what's interesting is there have been really three big, biggish trials that have come out so far, and all of them have had their sort of issues that you can see that I've written up there. One was done over ten years ago when GDMT wasn't really up to snuff. Another one was kind of inconclusive. And then the last one was really good and showed a pretty good reduction in heart failure hospitalizations, but it was open-label, so people had a little bit of trouble with that. They all sort of combined show that there's a decrease in heart failure hospitalizations, but we really haven't reached the Holy Grail of decreasing cardiac all-cause or cardiac mortality. And why might that be? Well, we know that pressure overload doesn't necessarily mean volume overload because congestion is really a pressure-volume-overload interplay. And to me, as an EP person, we know that it's not just about getting the numbers, but it's about patient selection, workflow, and reimbursement that also might play into this. So right now, if we're going to go by the guidelines, which I'm very much wanting to make sure everyone knows what the guidelines say, the 2022 heart failure guidelines say that it's not truly maybe not quite so useful at this point, and the value is kind of uncertain. And when they updated that in 2024, they felt that really we needed discussion and a team-based approach of whether or not we should be deploying this to our patients on a daily basis. So Camille and I were on the remote monitoring guidelines, and Derek as well. It's been about 10 years since we updated those. We released those in 2023 at the HRS scientific sessions. And what we know is we have these known benefits that we're all familiar with in this room, right? We decreased our missed call, missed evaluations and scheduled visits and in-clinic visits, and we have better earlier intervention. But I think what we should be moving towards in 2025 is that there are other benefits of remote monitoring like heart failure, AFib, and survival outcomes that we should be looking towards. So again, what do the guidelines tell us? Well, the guidelines tell us that when it comes to heart failure and remote monitoring, it's a 2A indication, which means that it's reasonable to do. So you don't have a Class I indication, but it is absolutely reasonable to be remotely monitoring your heart failure diagnostics. And why isn't it a Class I? Well, it's because our studies currently are also sort of mixed, right? Half of them are sort of non-significant, and the other half seem to be positive. And so we'll look at some of those trials. The first trial was the OptiLink heart failure trial that was released in 2016, and that one basically looked at using some of the heart failure diagnostics and looking at that by remote monitoring and comparing that to just usual care, right? And so if we get these diagnostics by remote monitoring, does it make a difference? And what they found is that there was no difference, right? No difference in cardiovascular mortality or all-cause mortality. But if you really kind of dove into the data, what they found is that only about 30% of the people had intervention based on the data that they received, right? So you can't just get data. You have to act on the data, right? The data means nothing if you don't act on it. So contrast to that, that to the combined in-time ECOS didn't trust trials where they actually did show that remote monitoring did decrease heart failure death and heart failure hospitalization. And why was that? That's because when you looked at that trial specifically in the in-time post-hoc, they made sure not only did the patients have daily connectivity that was ensured, but they made that connectivity change into action to make sure that within a week patients had some change in their therapies or their medical therapy, right? So you must have connectivity and action along with your data. So when they went back to look at OptiLink, they said, well, yes, of course, right? So when they looked to see if you actually look at the patients who had action with the data, they did have a significant decrease in their cardiovascular death and hospitalization. So yes, you must have timely, appropriate, and structured reactions to your remote monitored alerts. Again, data doesn't mean anything if you don't act on it. So then when you look at multi-parametric data and you use multi-parameter monitoring, so we're not just talking about impedance, but we're talking about activity, heart rate, heart rate variability, and you make a heart failure score, what you find is that you can use this to predict heart failure, okay? And when they put this into action in the real world with the Moore CARE trial, they found that indeed you could potentially decrease cardiovascular hospitalizations and heart failure hospitalizations. So then let's move on to our heart logic, right? So now just not multi-parametric, but multi-sensor heart failure monitoring. So when we looked at the multi-sense trial and looked at all of the things that we're all very familiar with in this room, that algorithm really performed quite well, up to the point that you could potentially predict heart failure hospitalizations up to about 29 days or a month before it would happen. So that's very, very important to understand. If you follow these out, you could potentially make an impact on patients' lives. Now, that's the algorithm, right? So what happens when you test it in the real world? If you actually put it into action, that algorithm, with, again, clinical action, education, and reinforced adherence, what you can find is that it's not only successful, but it's safe, right? And you can actually decrease BNPs, okay? Now, we're going to look at the harder outcomes, supposedly in January 2025, but that should have already happened, so we'll probably see that paper coming out pretty soon. So when we talk about heart failure monitoring and heart failure management, you cannot just talk about just heart failure, right? Because heart failure and AFib are inextricably linked. I wish Derek was here because Derek was part of this trial. He looked at data from patients with CIEDs and found that the more you have AFib, the longer duration, the bigger burden you have, people with and without heart failure had more heart failure, right? More heart failure, more morbidity, mortality, even if you didn't have heart failure. So, again, AFib, heart failure linked, okay? And what we also know now is that if we can take care of AFib fast or early in our heart failure patients, that makes a big difference in these primary outcomes of death or hospitalization, right? And even though in the real world it was a little bit more attenuated, if you can push down that AFib burden to below 50%, our patients do better, okay? Heart failure patients do better. And that's why we end up with these guidelines that are class one to tell us that we should be pursuing an aggressive rhythm control strategy in our heart failure with reduced EF patients. And what's interesting is that the AFib guidelines also tell us that we should probably be monitoring these patients with a 2A indication for their recurrence of AFib, right? Especially the people who have recovered their EFs if they've had a tachymedic myopathy from AF, so you can monitor our patients. We've had a couple trials now that have been looking at maybe less invasive ways to look at heart failure monitoring and also AF monitoring, and all those trials are sort of coming out with their results relatively soon. Phase one have finished in most of the trials, and phase two is ongoing. And of course I have to not, I can't not talk about my favorite stuff, which is digital health and AI. There are also many trials coming out now looking at multi-parameter and multi-sensor monitoring, and when the link heart failure did that and linked it with smartphones and continuous sort of machine learning algorithms, what they found is, again, that algorithm worked very, very well to predict acute heart failure rehospitalizations. The BMAD trial came out not too long ago at ACC. They also showed that a wearable for heart failure monitoring of multi-parameters also decreased heart failure hospitalizations significantly. And interestingly, when you looked at quality of life and using, again, multiple different types of monitoring for the patient's quality of life, what they found is that there was no difference, except when you looked at just MOOM that was involved. Because why? Because there was human touch. So we have to remember that we cannot just depend on all devices, right? The humans are never going to get out of this game, because people like to talk to people. So there's going to be multiple other trials that are upcoming for sure, looking at multiple ways of looking at heart failure, both with and without AI, and there's a multitude of different devices coming out very, very soon. But again, the success here with heart failure is to make sure that your patient is not only well-connected, but the data is looked at, and we're working in conjunction with our heart failure colleagues in this endeavor. Thanks a lot. Thank you very much, Janet. If anybody has any questions, you can enter them in the chat. Next, we have Yara Karina Spivak talking to us today about how do I manage alerts in workflow. Hi, everyone. I have no disclosures. Hi, my name is Kate Spivak, and I am a pediatric electrophysiology PA at the Children's Hospital of Philadelphia. And today, we'll talk about something that we're all wrestling with, making sense of the chaos that is remote monitoring heart failure data and creating a workflow to address these notifications. So let's start with a show of hands. How many of you love managing heart failure notifications? All right, great, great. We're all on the same page here. But what we know is that if managed properly, these notifications, all this data can help keep the patients out of hospitals. But there's multiple different devices across multiple different platforms. And even when we get these alerts into our hands in the hospital, our workflows are inconsistent, leading to variability in patient care. So in order to manage these heart failure notifications effectively, we need to understand the manufacturer-specific parameters and have a coordinated and intentional workflow. So first, let's talk about the many different devices that we have in our space that can give us information about the status of heart failure. Some of these devices are implanted, such as our ICDs and CRTs, while others are external, such as the HVABs from Zoll. Some of the devices provide specific heart failure alerts, again, like the ICDs or CRTs, while others, even though they don't provide a specific alert, give us some biometric data, heart rate, activity, that can give us a glimpse into how the patient is doing. And all of these different devices are made by a variety of different manufacturers. And these are just some of them. These are the most common. There's a few more that aren't even mentioned on the slide. And all of these manufacturers have their own platform. So think Medtronic as Caroling, Boston Scientific as Latitude. And all of these alerts are now funneling through the portals, through the EMR, and more recently, more commonly, through third-party vendors. And once we receive these alerts, it's important to understand the parameters that we're actually looking at. So just on this slide, like if you look at the parameters, there's almost 20 things in there that we're looking at to help us understand, monitor heart failure. And in addition to that, different devices use different technologies to do this. Impedance, bank radiofrequency waves, and even taking it a step further, some of these manufacturers come up with their own algorithms that combine certain parameters to predict the risk of heart failure exacerbation. So how do we put all of this together? So the first step is, of course, receiving the transmission, however it comes through. Is it a portal? Is it third-party? Is it coming through the EMR? And then we need to have somebody who can identify this alert and initiate the workflow. And then the clinical team must be able to review the alert and must understand what they're looking at and look through the patient's chart to put the story together. And then the next thing is we need to call the patient and figure out how they're doing. We call them with heart failure screening questions. Are they having more shortness of breath? Do they have edema? Are they taking those medicines? Have they been to the hospital recently? Is there a change in their health? And then a decision-making team member needs to make a call. Is this actionable or non-actionable? And if it's not actionable, we need to document why it's not actionable, and so the rationale and our conversation with the patient. And if it is actionable, the clinical team must develop a plan. Are we changing medicine? Is it a clinic visit? Or are we just sending them straight to the hospital? And in the end, we always must close the loop. The patient and the provider must be aware of the plan and must be happy with the plan. And sometimes we might need to change their heart failure monitoring frequency, and so we need to know how to do that and to monitor the patient more closely. So it's clear that heart failure notifications are important, but we face a lot of barriers. On the manufacturer side, there's no standardized alerts, and there's different sources for all these alerts coming in. And on the clinical side, alert fatigue is real, and staff needs to be educated on what they're looking at. And from the patient side, connectivity is key, as we already mentioned. And I don't know about you, but sometimes it feels like this is the hardest part of it, and we need to educate our patients about why we're monitoring them. And there's a few key things that we should think about. Who are our key stakeholders? Can we collaborate with our heart failure colleagues? Our protocols must be clear, our roles must be defined, and our communication needs to be consistent. But all of this is doable. Know your population. Know your devices. Create a workflow and educate key stakeholders about their roles. And develop a system to monitor your success, so that if something is not working, you can pivot and make things even better. And, of course, we need to treat the patient, not the alert, and together we can help keep our heart failure patients out of hospitals. Thank you. Thank you. So our next speaker is Audrey Nicholson, speaking about managing the burden of alerts. What is the role for AI? All right, good afternoon. Okay, so like you said, my name is Audrey Nicholson, and I will be speaking about the role of AI for managing the burden of alerts. So I got to HRS a little bit early, and I got to do some really cool stuff. One of them was go paragliding. In case you don't know what it involves, you strap yourself to a stranger and then run and jump off a cliff. And hopefully, you know, the air picks you up and you go flying, and it is fantastic, 10 out of 10 would recommend. I'm there on the left side, the guy in the jeans, that's my boss. He's the chief of EP, so it can't be stupid if your boss does it too. All right, so what is AI? Here's a quick overview of artificial intelligence. The term artificial intelligence was first coined in 1956 and involves the idea that computer systems can be created that mimic human intelligence in the human thought process. So the base of AI is the field of data science, which combines elements of statistics, computer science, and domain expertise to analyze and interpret complex data. Data science is used to aid in decision making, prediction, and automation, and is used every day. So artificial intelligence refers to machines that can perform tasks that typically require human intelligence, like recognizing speech, understanding language, identifying patterns, and making decisions. These days, I'm finding that a lot of things are referred to as AI, when in reality, they are just algorithms. Think of it like this. AI uses algorithms, but not all algorithms are AI. If you take nothing else from this talk, take that. So I think this distinction is very, very important to understand, especially in our world of cardiology with complex devices and wearables, etc. Every company is trying to jump on the AI bandwagon and may be advertising something as AI, when in reality, it's just a complex algorithm. Now, on to machine learning. Machine learning is a subset of AI where machines learn from data and then adapt to improve their performance. Machine learning requires clear, labeled data, which is most often done by humans. Now, deep learning is a form of machine learning that uses neural networks with many layers to model and learn complex patterns found in large amounts of data. Deep learning is more like the brain in that it can automatically extract features from raw data, like images, text, or sounds, without being told exactly what to look for. It can learn automatically without a human having to program or tell it what to do. Now, big data encompasses all of this. It's data that comes in high volumes, high velocity, and lots of variety. Think about the constant telemetry monitoring for an entire hospital. In the past, processing this data was limited to the power of humans and our brains. Now, with AI and deep learning, vast amounts of data can be processed quickly to uncover patterns and make predictions that were previously impossible. So, a recent single-center study published in EP EuroPace this past January looked at device transmissions from a heart failure service for over a year. All alerts were acknowledged and triaged by a clinician and then sent to physicians for further action as needed. They received approximately 5,200 transmissions from 315 patients. 52% were unscheduled, or roughly 2,700, and they included all types of heart failure patients and roughly 2,700, and they included all types of alerts such as arrhythmias, lead issues, CRT, pacing issues, et cetera. 9% or 241 of the unscheduled alerts were specifically heart failure alerts. So, from these alerts, 241, only three, or 1.2%, resulted in medication changes. Three. Just think about that. So, humans, we are really good at making complicated things more complicated. This graphic was republished in the 2023 JAC statement on remote monitoring for heart failure management at home. You're not meant to read this. I doubt you can even absorb it, but it illustrates the components and the pathway of heart failure management from patient to transmission to clinician and then back to patient. There's a lot going on here. It gives me a headache when I look at it. It gives me a headache when I look at it. It's not even clear on exactly how heart failure alerts should be handled. I think it does, though, a very good job of communicating just how complicated this task is. So, kind of just like Kate alluded to, first, clinicians need to decipher noise from relevant alerts. Then they need to evaluate the alert in the context of the real-life patient, their history, what meds they're on, were they recently discharged? Is this a known salt offender? Do they drink five big gulps a day? Do this 241 times over the course of a year, you can see why only three patients from that previous study would even have any meaningful medication changes. So, to more effectively manage these patients, we need to identify meaningful trends across metrics like blood pressure, weight, heart rate, thoracic impedance, the list goes on. We need to improve triage and decision-making, and most importantly, we need to free up clinician time to focus on patient care. So, AI to the rescue, right? This is an example of a neural network, one of those deep learning components from the first slide. It can be trained to recognize meaningful trends in a patient's health, not just from device data, but also from their medical history, EMR and home vitals, for example. The goal is for AI to learn what is important, ignore what isn't, and then decide who needs to see which alerts. AI uses pattern recognition to see what humans might miss. It can detect subtle patterns over time, like a small weight gain, a slight dip in blood pressure, or an increased heart rate. On their own, these changes might not mean much, but together they can signal very early trouble. So, instead of comparing numbers to a generic threshold in a device algorithm, AI looks at what's normal for this specific patient. For example, a two-pound weight gain might not be much of a deal for one patient, but could be a serious, serious event for a frail 85-year-old 110-pound woman. AI can reduce alert fatigue by ranking alerts based on urgency and routing them to the right clinician, whether that's the heart failure provider, the device tech, the nurse navigator. This way, high-risk alerts rise to the top and the noise is filtered out. AI can pull data from multiple sources, a pacemaker, a scale in the patient's bathroom, a blood pressure cuff, the EMR, and combine it all into a single meaningful summary. Instead of siloed alerts all going to different teams, now you have one clear picture that goes to the right person. It seems simple enough, but maybe not so fast. Artificial intelligence has huge potential to help manage these alerts, but like any tool, it's only helpful if we use it in the right way, and that starts with how we train it. AI is only as smart as the data we feed it, and if that data is limited or skewed, AI's decisions will be too. Think garbage in, garbage out. Bias is also a huge concern in healthcare. Imagine training an AI using patients who all live in suburban New England with easy access to top-notch healthcare. Now imagine applying that same AI to someone who lives in rural Alabama. They have different health challenges and far less access to care. That's AI's recommendation. The AI's recommendations might miss the mark for that patient in Alabama, so that's why we need diverse representative data covering different geographical regions, incomes, education levels, and healthcare access. So as we know, heart failure is a very complex condition, so naturally, the data around it is gonna be complex. AI has to learn patterns without just memorizing details. This is what we call overfitting. It's like a student who passes a test just by memorizing the answers and not actually understanding or learning the material. And finally, you might have heard of the black box problem. Data goes in, something happens, AI makes a decision, but we have no idea how it came to that conclusion. If we asked it to show its work, the printout would fill this entire room, but that doesn't work in medicine. We like to understand, so we need AI that can show its work. It's what's called explainable AI. If we want providers and patients to trust AI, we need transparency and not just mystery. So I believe that AI will at some point make up a critical component in the management of heart failure alerts from CIDs. The question is how quickly will this come and will you be ready? These are my references. I'm gonna speak on wearables and follow all these wonderful speakers. I did not go skydiving, I wish I had known. I do plan to scuba dive when we're done with this. I don't know what everyone here is passionate about, but I hope you get to do something exciting in San Diego. But while we're sitting here learning about heart failure, let's see what we can find that's exciting here. So I think I got the fun talk, but there is a lot going on. I don't need to tell anyone what the costs are associated with heart failure and the problem of readmissions. So for the ambulatory solution, non-invasive CHF management, we have all the traditional things that we think about, the scale, the pulse oximeter, the blood pressure cuff, and of course, that human component with nursing and other staff reaching out to the patient. Modern days, there are electronic stethoscopes and sensors and smartphones, remote wearable sensors, although I will say some of the things that we would think of as traditional, there are eyeglasses now that can measure your blood pressure. So that's changing as well. So AI plus wearables is really the next generation of ambulatory heart failure care. There's a couple really good review articles if you're looking to get everything in a quick stop and you don't wanna go back and look up all the slides, these articles are two really good ones from the last couple of years to give you some good information on where things are. So we're gonna talk about what's available now and some of the things that are coming, so we're gonna breeze through some of these slides. I do have some of the trial data, feel free to go back and look through some of this, but I just wanted to make you familiar with what's available there now if your current practice is not utilizing it. So this is the Zoll Heart Failure Management System, it is FDA approved, it tracks multiple things, heart rate, respiratory rate, activity, posture, ECG. This management system is a wearable patch, it uses RF for the early detection of pulmonary fluid level increases, and it has up to 90 day wear time. Some people are utilizing this during GDMT titration to help them know where they're headed with things. There is coding and billing available for this. The next wearable we're going to talk about is the edema guard monitor, and that checks differences in impedance. During heart failure hospitalization, improvement in those pulmonary fluid volumes was predictive of lower readmission rates and demonstrated better correlation than some other clinical measures that if you're working in a hospital or acute care setting, you might be using, or even as an outpatient, the NT ProBNP or weight gain. That's from the Impedance HF Trial, Single Blind Multi-Center Randomized Control Trial of 256 patients. It did show that it reduced hospitalizations for acute heart failure as well as the incidence of cardiovascular and all-cause mortality. This is one I hadn't seen before, I don't know if anyone here has seen this utilized. This is called the COVA monitoring system. It's actually a wearable necklace, and it monitors ECG, it calculates stroke volume, cardiac output, and thoracic fluid impedance using the chest bioimpedance. It's FDA approved. This is the non-invasive cardiac system monitoring that does whole body bioimpedance. It monitors heart rate, stroke volume, and several other parameters. It is FDA approved, and it's statistically bioequivalent to a pulmonary artery catheter determined cardiac output. The REDS system, this is also FDA approved. It's utilized in multiple different settings. It involves two sensors embedded in a wearable vest or a shoulder clip. It takes about 90 seconds to make a measurement, and then it uploads to a web-based patient management system. This showed a 79% reduction in cardiovascular readmissions. It's utilized in home, hospitals, post-acute care, and outpatient clinics, and there's one of the studies there and there was a 91% reduction in readmission using this for heart failure management. The LifeShirt system, it detected precursors of hospitalization for heart failure exacerbation with about 76 to 88% sensitivity. It utilizes multiple sensors as well. And then we'll move to ambulatory monitors, which of course, as Dr. Hahn talked about, atrial fibrillation and the incidence of heart failure, it helps with that. I don't know that we always think of that traditionally when we're talking about ambulatory monitors and how they can help with heart failure. VitalConnect is a patch that uses AI that Audrey just taught us about to assess continuous EKG as well as skin impedance, body temperature, and activity level, and it generates a personalized baseline model alert system for CHF. That's with the LINC-HF trial, multi-center study. It's a disposable patch, and it forecasted with 76 to 88% sensitivity and 85% specificity, again, utilizing that AI for a personalized model. So what's on the horizon? Those are all the things that are sort of out there already that you may or may not have seen or heard. What's coming? Some really interesting things. The Hero is a non-invasive CHF monitor that uses speech sampled by a mobile device to detect fluid accumulation. The patient just talks softly into their phone. It got the FDA breakthrough device designation last year, so it's currently undergoing an international multi-center observational and non-interventional prospective blinded single-arm two-period study, that's a mouthful, to collect patient utterances or just speech that they're gonna retrospectively analyze to determine the sensitivity of the use of this for heart failure monitoring. The AQORI, if I'm saying that correctly, is a non-invasive pressure monitoring system. It's got a unique patented hardware technology and proprietary machine learning system. You place the device on the upper chest for a few minutes to capture the data, which is displayed on the screen. This also has an FDA breakthrough device designation. This is through the CAPTURE-HF trial, which is a 1,600 patient international prospective trial. It's comparing non-invasive hemodynamic measures, which is the device that we're talking about, with those utilized through gold standard right heart catheterization techniques and blinded core lab oversight as part of a prospective registry. The Evron, this is a piezoelectric sensor. It's actually, if you can see on the right there, that middle image, it's actually an under-the-mattress bed pad, and it detects subtle physiologic vibrations and it converts them to an electrical signal. They've got a single center study right now with 30 heart failure patients who are discharged home following an exacerbation, and they collected 640 nights of data, and the study found patterns that could be unique among patients at risk for readmission due to heart failure. And of course, probably most of us may have one of these on our wrist. This is the TrueHF study, has completed enrollment, and they are using a smartwatch to evaluate if data gathered by the smartwatch in combination with AI can predict an acute decompensation. So we come to the question again that we had at the beginning. Is there a role for mobile or wearable technologies in heart failure monitoring? I would definitely say yes, there is. Challenges and opportunities, tremendous amount of data that come in from these. The evidence in some of the studies that I've presented to you today do support the use of wearables. There are lots of challenges with how we utilize that data, which some of our wonderful speakers here today have addressed, and there's additional research needed to see how we apply those. So consideration for allied professionals, most of us are here today. We need to assess the device, look at the literature and the approvals, the price, the cost to the patient, and best practices. What is the benefit to the patient and to your practice? Because it doesn't matter how much data they bring into you, if they're like my patients, they bring in their look, and they have all these things, or they want to hand you their phone during the visit that's already too short and ask you to look through all of that, and you're looking at a scratchy single lead EKG and wondering what to do with it. And then there's also clinical workflow integration. So how do you put that into your chart? How do you say anything other than the patient showed me this on their watch? So I don't think we have great ways to do that yet. There are some. But then also data management, rights, governance, storage, privacy. These are all the concerns and considerations that we have to look at here today. And I did just want to, by an informal raise of hands, how many people have used at least one of these devices today in their practice, or had a patient who's used it? Any of them? Okay, so we have a few. And that's why I found this to be an exciting talk, because there is so much out there on the horizon that you may or may not have seen. And I think that when we come to HRS, maybe even a few years from now, the landscape of this talk might look very, very different. So thank you. Our next speaker is Alana Miller, speaking about evaluating implementing third-party remote monitoring solutions. So we just heard all about how this is really, really hard to do and hopefully my talk will help you think about a possible solution to at least assist with some of this. I have no disclosures. So again, my name's Alana Miller. I'm a nurse practitioner from Penn Presbyterian Medical Center in Philadelphia. I'm going to take the next few minutes to talk about evaluating and implementing third party remote monitoring solutions specifically for heart failure. Effective remote monitoring is very complicated, as we just heard, and it can be extremely challenging. Some organizations choose to do remote monitoring with an in-house team, but many health systems are looking for help. And that's where third party solutions come into play. The 2023 HRS consensus statement regarding management of remote device clinics identifies the use of a third party resource to assist with monitoring and triaging patient data. Third party solutions have been identified as an option for health systems looking to improve workflow efficiency, data management, and ultimately patient outcomes. Third party solutions are specialized platforms and services designed to consolidate patient data from each device manufacturer into one location. These companies also offer services to troubleshoot connectivity and improve administrative workflows by automatically scheduling transmissions and streamlining billing processes. So maybe you're interested in using a third party, but there are so many options. How could you possibly choose a vendor? I'm going to take the next few slides to outline how you can systematically evaluate different vendors in order to choose the right one for you. Before you start looking at vendors, I would encourage you to identify your key stakeholders. So for heart failure remote monitoring, this would be cardiology, EP, and heart failure providers, nurses, care coordinators, device techs, members of your current remote monitoring team if you have one, representatives from IT, and of course, administration. Next, do a really good self-evaluation. Assess your existing process, how you collect and use remote monitoring data, identify your current pain points, gather your revenue data, evaluate your billing process. Review your current remote monitoring staffing model and any staffing gaps that you have. Talk to your IT department and find out if EHR integration is possible with your electronic health record and if there are requirements for this integration. Talk to your key stakeholders to determine who will be logging on to the remote monitoring portal and how. Will they log on to a web-based platform or will they require a single sign-on solution through the EHR? Many platforms offer the option to integrate some of this wearable data that Kelly was just talking about and in-person device checks, so if this is important for your institution, that would be important to know before you start your vendor search. And are you currently utilizing remote physiologic management and do you want a third party that can do this? And then really do a deep dive into the nitty gritty of your ideal workflows. So if you could have any workflow, what would it be? Are the physicians or the APPs logging onto the portal to bill? Are the providers reading remote transmissions for their own patients or do they have a designated reader? Will the nurse review the data first before sending it to the providers to bill? And very importantly, are you planning to utilize the third party software as a software only solution, meaning your existing team will provide all patient education, troubleshoot connectivity, contact patients for missed transmissions and manual transmission requests, or do you want a full service solution that will take care of all of this for you? After you've captured this information, it's time to start interviewing vendors. A good place to start is with the EHR integration. This can allow you to narrow down your vendor list right away. The vendor should be able to integrate with all of the device manufacturers that you would like, as well as your electronic health record. High quality integration is bidirectional and this includes the ability to seamlessly trigger billing at the time the report is signed by the clinician. Next, evaluate vendors based on their usability and efficiency. So we all talk about how many clicks we make in a day. This is your opportunity to minimize clicks. You'll wanna limit the need to switch back and forth between the vendor portal and the electronic health record. Ensure that urgent alerts can be easily identified and that the vendor can accommodate a secondary reviewer workflow if this is important for you. And check references, not just the names that the vendors give you. Ask your friends and colleagues to see if you can find someone in an organization similar to yours who are using those vendors so you can get real feedback on their experience. Ask your IT colleagues to evaluate the vendor's privacy and security features. Your health system may exclude vendors specifically based on this criteria. Hopefully, based on that, you can narrow your list a little bit. And then I would next look at these key vendor features. So first, vendors should have high quality customer service. Their service team should be easily available to answer questions in real time. If you're looking at software-only options, you wanna consider asking if they allow you to flex up to full service if needed. This allows increased flexibility so you would be covered if you lose staff or volumes increase. Make sure you ask about what support they offer during the go-live period. Will they be onsite for the first few days or few weeks? And what user training do they offer? Be sure to find out how they train their own staff. Are the remote readers IBHRE certified? Are they nurses or techs? What continuous education and training do they provide for their own staff as technology changes? And do they offer the ability to customize? It's important that the vendor's willing to work with your clinic to customize reports, templates, workflows, and collaborate on software development based on your specific needs. Be sure to ask about analytics and data tracking. As well as how they can help you manage recalls. And then, of course, cost. How much is this gonna cost? What's the pricing model? Do they charge per patient, per transmission, per billable transmission? Are there startup costs or fees for the EHR integration build? And then, of course, the heart failure specific features. So do a deep dive into their heart failure specific features. Heart failure specific monitoring is a rapidly growing field. And many of these companies have new features designed for both the heart failure providers and heart failure patients. From a software perspective, if your heart failure providers are gonna be logging into the portal themselves, does the vendor offer a parallel heart failure only dashboard? This is also where you'll wanna find out if they can integrate all of those wearables Kelly talked about, and remote physiologic monitoring. So remote physiologic monitoring requires some of that discrete data, but also time-based billing. So the platform would need to be able to track the time spent by the nurse and the provider reviewing that data and talking to patients. Customizability we've talked about, but it's particularly important for the heart failure metrics. These often need to be customized, not just by provider preference, but also sometimes per patient. And additionally, companies are developing patient-facing apps and automatic messaging for heart failure alerts. These are interesting technologies that could be hugely beneficial to your patient population. So you've selected a company, and now it's time to go live. I listed here a few important steps in that implementation process that you'll wanna address before the actual go-live date. Review your current workflow with the company and your key stakeholders to make sure everyone's on the same page. Companies have nominal alert thresholds, but these can be customized based on your preference. Talk through the roles and responsibilities of your team versus the third party and map out this new remote monitoring process from end to end so there are no unanticipated surprises. Work with the vendor to set up training for your team. The vendor should offer training sessions prior to go live as well as on-site assistance during that initial go-live period. And integration, this can be very complicated. You'll likely need to schedule recurrent meetings with IT, administration, and your vendor to facilitate the software build. You don't necessarily need to have integration in place prior to using the third party. This can be done in the post-implementation period if necessary. And lastly, you'll need to notify the patients that you're transitioning to a new remote monitoring process and that they might be contacted by this company. Be sure to provide the patients with the vendor contact information and billing practices if this is going to be changing. Congratulations, you've selected a vendor and you've started using them, but the work is not done. Following implementation, you'll need to continue to evaluate your remote monitoring process, especially during the first three to six months to ensure this transition happens smoothly and to work through any roadblocks. Take a look at your time from alert to clinician response, your patient outcomes, volumes, and overall workload. Plan to provide recurrent training to staff, whether it be regarding new features, technology, workflow changes, or common errors. You'll also want to perform some internal evaluation of competency for both your staff and for the vendor to keep everyone accountable. Schedule regular meetings with the vendor to discuss issues, review feedback, and keep up to date on any new software builds or updates from the vendor side so you aren't surprised if they change something. And again, it all comes back to the money. So are you capturing the billing you were expecting to? Why or why not? And how can you continue to optimize this? In conclusion, I will leave you with the five things that I wish I had known going into this process. First, the implementation phase really can take up to three months. There's a learning curve for everyone involved, especially the vendor and your reader-vendor relationship. It will not be perfect on day one. There's no set-it-and-forget-it solution. Things are constantly changing and evolving, and it's important to maintain open communication and be flexible. Time zones matter. So this was a big one for us. It's important to have a vendor that has readers and service personnel in your time zone in case issues arise or alerts come in first thing in the morning or at the end of the day. Ultimately, customer service is the most important thing. And keep in mind that this is a highly competitive industry, and companies will build features for you if you ask them. So everything is negotiable. And that's it for me. Thank you so much for your time. Thank you. So thank you, all the speakers, for wonderful talks. Any questions for the audience? So one thing I'll add is for Alana, when you're thinking about choosing a company to assist with your management of the data and integration into your EMR, particularly, another thing I would add to your talk would be the ability for the company to keep up with changes in technology. And when they don't, then that is a hindrance, and all of the benefits that you got from the workflow process then become lost. And so making sure you have good communication with the vendor's team and their engineers as technology changes. So does anyone else, or maybe Janet, you have experience or any thoughts related to that? I think 100% as to what you said about making sure that you can keep up with new technology. I think that one thing I'll add just in general about AI is also that it's already here. It's not like it's pie in the sky, and it's not coming, that it's coming soon. It is already here. It's already integrated in a lot of our technology that we use today, like implantable cardiac monitors. Every single vendor out there already is using AI algorithms to make sure our workload is better, to make sure that you are decreasing all your false positives while maintaining your true positives. So absolutely, AI is already being used, and all the algorithms are already in play. So that's the first thing. I think it's gonna continue to hopefully make our workload better too in multiple ways, I hope, because the data inundation is always a difficult thing, right? So keeping those true positives good and keeping the false positives also low. I think what I'd heard is all companies currently, they're like at 99% with the true positives, and then the false positives are exceedingly low. So they're doing quite well. I think adding into it for the heart failure, we see a lot of that on false pauses and AFib for the ILRs, but I'm not sure that technology has advanced as much for all of these various tools out there in order to enhance it from a heart failure diagnostic standpoint. Yeah, I think the heart failure part is difficult, right? Because there's a lot of great technology that's come out, which is like, when I talked about the VMAB, that is the Zoll product. But you have to remember, in the heart failure guidelines, they don't exist yet, right, of how we are to use these devices appropriately. I think they're great ancillary products, but you cannot rely on them in their current state because we just don't have data, right? And that's where the guidelines come in, and we have guidelines for a reason. So I would love to be able to say, hey, just pop on your smartwatch and tell me how much AFib you have because Apple says that you can monitor your AFib burden. Well, you forget, it's not continuous, right? So I tell my patients, absolutely not, not at this current state. It's coming, and it's great that it tells you that you might be in AFib, but you have to come to me and our clinicians to tell you that you're in AFib. Same goes for the heart failure stuff, right? Great products, they're only gonna get better. The algorithms are only gonna get better, but at the current state, they're not in the guidelines. Completely agree. Thank you. Any other questions? I have a question. Yeah. So do you think for the remote monitoring alerts, do you think that we would still put it, or do you think the AI would adjust it? Like greater than 24 hours of AFib, or the heart logic is greater than 16, do you think we would still have to adjust it to the patient, or would the AI possibly be able to adjust it per patient? So, no, the future would be that AI would do all of that. We're not, like Dr. Han said, we're not there yet, and I think for heart failure especially, we're probably years away. I'm not ready to say a decade, but years. So no, it would be all done by AI. Yeah, I'll put a corollary on that. I would say that heart logic's actually very good, and we saw that sort of coming in, even the Medtronic products, right, years ago when I was a fellow. And so at first, we were saying, oh, man, this just doesn't correlate, you know? But again, algorithms are better, and they will continue to get better. So I agree that to a certain extent, AI will make some decisions for you, but that's not a threat, you know, that's definitely to me not a threat to our jobs. I think Paul Friedman said once in a talk that I really liked, that he said that AI really is the light to help you see better, right? So it's not gonna replace you, it's not gonna make all your decisions for you, because they can only be so good. And that brings in a whole nother talk that we could talk for another hour about, about ethics and legality, right? And where do we stand with that? When we as healthcare practitioners decide how much do we allow AI to do, and if our patients are gonna trust it, and again, what is the sort of legal ramifications of that, if you trust AI and they make a mistake, what happens, right? So yes, to a certain extent, it'll make decisions that we as humans probably have to make the final decision. As of yet, AI is not licensed to practice medicine. I do think it's really interesting, though, if you look into what the AI algorithms, like she said, when they bring it outside of the box, if you look at how your loop recorder, the different companies, how their AI algorithms work, to actually get to the point where it's giving you that, well, this is the data, it's fascinating. They don't all work the same. Yeah, and some of the Kelly stuff that you had in your slides, some specific companies, I'm not gonna call them out, that are using AI, just like I said, they're actually just using algorithms right now, but they're kind of advertising as AI, and so that's what everyone has to be careful about, is that double check, learn what AI is, learn how it's used, so that way you can better evaluate these products and also their results and what you're seeing. I think it made a difference in what vendors we were choosing for certain things when I actually sat down and learned what was going on behind the scenes, that at the surface, you just think a product is doing something that you're using to give you data, but it could vary, and I think it's also telling that so many of us, it was really interesting for me to do the talk today, because I learned a lot about what's out there, but I think there's a reason why many of us are not using any of those things, even though multiple of those wearable items are FDA approved, they're just not in use, because what do you do then with that data, and what are the medical legal implications of those items? Yeah, and I think the other thing that people don't always know in that process, is that in the FDA 510K database, is that you just have to have a predicate product, right? Meaning that you have to have a pre-existing product that your new product is similar to, and you have to have testing validation, you know, and external validation data sets, and if it looks similar or up to par to the predicate device, it passes sometimes in the FDA, but what you need is hard data to make sure that it works in our patients, and it hopefully improves outcomes, and that's what is really, that changes the guidelines. And I found that when I was doing my research, some of the trials, as I was searching for some of those different products, they had a preliminary, very tiny study, and then the next thing you find is that FDA approval, because it's similar to something else, and there's really not any more data on it. So Dr. Hahn, I have a question for you. Do you think, heart failure related, that AI might be a situation where it is driving the guidelines versus the guidelines driving AI, because of the data, the amount of data that will be collected? So I'll ask that question again. You think if AI is going to drive the guidelines? Right, because it just allows for so much more data to be collected and analyzed. Well, I think it depends on your AI model, number one. It depends on the data that you're looking at, number two. I think, you know, we will always sort of, again, go back to that adage that I always say that Paul Friedman says, is that it helps you to look with a better light and better eyes at what you're trying to look for. So I think humans will probably always shape the guidelines, because data's only as good as what you make of it, right? You can get as much data as you want. AI can tell you all the things, but it's your human that's sitting in front of you, and whether or not that AI algorithm or output applies to that patient. And we have a question from the audience, the online audience. What are the panel's feelings regarding remote monitoring where patients are able to input daily vital signs, or I guess whatever input on a daily basis, example is cadence, and the role these play in heart failure management? I think the parameters need to be taken together. I think it's always good to have more information, but we need to figure out, you know, how it all works together. So it's a hard question to answer, like if you just have, you know, like blood pressure monitoring. But I think if we put it all together, it might make sense to monitor. I think that's where the multi-sensor data, okay, I think is helpful, too, if they're putting in, there is some value in having the blood pressure and the weight and the things that people are putting in from home or giving to you, but as I think everyone here agrees, it's that human touch. I don't think we can get away from that in contenting the patient and making those medical decisions. We also need to make sure that we're teaching these patients how to do this correctly, you know, that they're using a good scale. Do they know how to take their blood pressure? Do they have a good monitor to do that? Yeah, I think it's a great aspect of care in all of these new wearables and these devices that are coming out are fantastic. But the problem is right now, AI hasn't caught up with those so that's just a deluge of information that we as clinicians are gonna have to either deal with or just be okay ignoring. And so I think it's great, but it's a scary, I see a tidal wave in the distance of data that's kind of scary. For the specific systems that they described on the question, I think it's looking at trends, right? Because it's looking at the trends of a patient. If they're every day putting their weight in, putting their blood pressure in, then you're following that over time. And when you see variance in it, then it tells you I need to contact the patient and look for whatever may be going on with them. So I think that that can be useful in that standpoint. And recognizing whether your patient overall is stable or not stable. Using weight alone as a marker for heart failure exacerbation you know, is a poor surrogate. And so I wouldn't use that or any one value, but I think using it overall as a trend for your patient, it can be useful. So, all right. Thank you very much.

remote monitoring

heart failure

cardiac electrophysiologist

PA pressure monitors

AI-driven wearables

clinical workflow

patient engagement

2024 heart failure guidelines

human oversight