Welcome to an episode of Heart Rhythms the Lead, an abbreviated journal review podcast. We're at HRX 2024 in Atlanta, Georgia. I have a very special guest with me today, Dr. Hawkins Gay from Northwestern University. Welcome. Thank you. Dr. Gay, you and your colleagues wrote this editorial on managing data overload, AI wearables and apps. Can you give us a summary of your editorial? Sure. So we're in an age now where there are a lot of personal digital health tools that are available to consumers that, you know, are not being prescribed or ordered by physicians that, you know, our patients have access to, you know, anyone can buy an Apple watch or a Samsung watch or even a, like a personal glucose monitor and start recording all of this data that, you know, has, contains a lot of information about their health that's available to them and that then they are often making available to us sometimes significantly more of it than we want or know what to do with. And so the point of our article was to kind of think about ways of making that data more usable so that, you know, we can utilize it in a way that's helpful to patients, but also how can we digest it all and make it, you know, a little less overwhelming. And I think one of the ways that we're not currently good at doing that, but will one day be useful to us is artificial intelligence will help us kind of digest all of that data in a usable way. And so I think the first thing that we discussed was just, you know, how do things exist now and like, what's the problem now? I think the problem with it is that the data is just coming from so many different things. Some patients have an Apple watch. Some patients have a Samsung watch. Some patients have, you know, a Fitbit. Other patients have devices that, you know, are less well studied or less well known. And they're just recording all of this data from those devices and, you know, they, some of them might give you heart rate or some of them might, you know, a lot of them take an electrocardiogram, but, you know, the quality of that differs from device to device. And it hasn't all been studied or approved by the FDA and kind of ways that we're used to, like when you use a medical device, you expect it to have been, you know, studied and approved and kind of apply to a certain standard. And you can't, you know, necessarily feel comfortable like for a consumer, for consumer generated data in the same way you can for medical grade data. And so we're dealing with all of that. And in addition to the variable quality and, you know, the numerous different devices that this data is coming to, the way it's getting to you is just different as well. Some patients just bring it to their appointment and they're like, hey, can you just look at this data that I have on my phone? And other patients are uploading it to the EHR, you know, through a message on, you know, if we use Epic at Northwestern and, you know, there's something called MyChart and they can just upload PDFs of their, you know, blood pressure or their ECGs and they just send it to you and ask you about it on a message. And so it's just coming in variable formats and it's coming from variable, you know, sources and it's of, you know, different quality and it's a benefit and problematic. And you know, there's no great way of storing it and sifting through it. And I think as electrophysiologists, we've faced this problem before with implantable cardiac devices, you know, because you get a lot of information from pacemakers and defibrillators. And what ultimately happened is that in the, you know, industry players started producing dashboards where all of these implantable cardiac devices could send their data to. And you know, you could collect Medtronic data on the dashboard and Biotronic data and Boston Scientific that didn't matter who made the device, the data was collected in one place. And so I think for us to be able to utilize all of this consumer generated data, we're ultimately going to have to be able to develop similar dashboards or modules, whatever you want to call it, where that information is collected and it's stored in a way that can be usable. It can be pulled up by physicians and quickly sifted through and used. And what we hope to be able to do is have AI help us sift through that information. And you know, a lot of people are doing studies and showing the benefit of artificial intelligence. It can use ECGs to make these superhuman predictions that, you know, even, you know, cardiologists and electrophysiologists can't, you know, identify this type of information from an electrocardiogram. And it can also just pour through like tons and tons of, you know, laboratory values or clinical notes and create useful information. But to be able to do that, the information has to be available to the AI. And right now, a big problem in healthcare is a lot of healthcare data is stored in unstructured ways or variably structured ways. It's siloed in unusable formats. And it's just very difficult to get it all in one place where then you can release, you know, an AI model on the data and have it, you know, use it in a useful way. It's just not collected and stored in a user-friendly format, unfortunately. And so, to get to an era where we're able to use all the information that's being sent to us in a useful way and have AI help us use it in a useful way, I think it has to be stored in a more efficient structured format than it is currently, yeah. How do you see the role of healthcare providers evolving as we integrate wearables and AI into clinical practice, especially considering the potential for data overload? So, if you could expand a little bit on that. I think, you know, healthcare providers and healthcare teams are just going to have to become more efficient at using this information. We're going to have to understand it more and be able to respond to it more quickly. And so, AI will help us respond to it more quickly, but we're just going to have to have some understanding of the quality of it. And, you know, so, one of the problems is, like, for me to know what goes into, like, the Apple Watch, for instance, and a Samsung watch, their algorithms have been FDA approved for consumer use for their, like, irregular rhythm monitoring algorithm is FDA approved and the ECG algorithm is FDA approved. But to know what it takes to have that approval, I have to go into that, you know, FDA approval and learn what that is. I'm not taught that in medical school or things like that. And so, unfortunately, I think providers are going to have to start educating themselves on the quality of the different, you know, tools that are being used by consumers. And there are tons of them out there, but, you know, some are used more than others. And so, I think you just have to start, like, looking into what's available, what are your patients using, and educating yourself on the different ones so that you have some sort of understanding of the quality behind them and what goes into it. And to the extent that companies want patients to use this for medical purposes, they might, you know, one day we might get to a place where hopefully, you know, the companies that are producing these things are providing educations to physicians that they know consumers are using it, you know, in a medical way. Maybe they'll start, you know, providing education to physicians to help them get up to date so you can, like, learn, you know, how is your algorithm approved, what was it trained on, like, what did you, like, you know, how accurate is it for different things. So, we're going to have to become more efficient in learning the quality behind this data. And then I think the healthcare team just has to be set up in a way that they can use the data. So, one day, like I said, you know, if this is all stored on a singular dashboard or in some module that's available through the EHR in the background, you know, you could hopefully have AI running so that it can alert you to, you know, if a patient uploads an electrocardiogram that shows atrial fibrillation, the fact that that patient's in atrial fibrillation is going to be more important for some patients and less urgent for other patients. And, you know, it would be nice if there's software that's looking through that module and saying, hey, here's a patient that just uploaded an ECG showing they're in atrial fibrillation. It's pretty important for this person because they're 78 years old and they've had two strokes and they're not taking anticoagulation. And less important for this other patient who's never had a stroke, they're 45 years old and have no other comorbidities. So you can get to it a little bit later. So I think having AI help us with our, you know, how we efficiently use that data will be really important one day. Yeah. Interoperability remains a significant issue in aggregating patient data from wearable devices through EMRs. How can healthcare systems and regulators work together to create a seamless data sharing system? Yeah. So that's being worked on. There's something called the Trusted Exchange Framework and Common Agreement and then the Fast Healthcare Interoperability Resources that this is kind of a framework that was produced by the U.S. government to create a system of recommendations and structured format for healthcare data so that it can be shared between providers and what are called health information networks, which includes things like Epic and Cerner. Because we don't want data to be siloed in a lot of different formats where it can't be shared and can't be used. So just as an example, so for people that have, you know, an iPhone and that wear an Apple Watch or that use different applications on their iPhone, a lot of that, you know, Strava, which is just a personal health app where you can track your workouts and your heart rate and things like that. If you have that application on your, you know, iPhone and you do a workout, then it will upload that information onto the Strava app, but you can also access that information through Apple's, you know, health application that is created by Apple. And the way you can do that is because Apple, through Apple Health Kit, has set certain standards where if you structure the data in a certain way, it can be shared on iOS with the health application. So if you create an app that you want to be able to be incorporated into Apple's health application, you have to format your data in a certain way and then it can be shared and utilized by consumers on their iPhone. The same thing with Google Fit. You know, if you have an Android operating system and you have these applications on your Android device, Google Fit has, you know, certain standards where if you structure your data in a certain way, it can be shared between applications on the Google Fit platform. That same sort of, and people that make the applications are complying with that because they want to be a competitive, you know, product that consumers find useful. In the same way, I think we should have those standards exist for how healthcare data is utilized, especially since consumers are generating this data on their personal applications. So if we set these standards up in a way that, you know, companies that are generating these applications know what the standards are and they can structure their software and their applications to store data in that format, then it can be shared among these health information networks in a way that, you know, it can seamlessly go from one application to another or you can, like, generate the data on your watch and share it with, you know, your doctor on Epic because it's all structured and stored in a way that everyone understands and, you know, is following certain formats and complying with. And that is the purpose of, you know, of the FHIR, you know, kind of structure that was developed. So we're working towards interoperability. The problem is, there's a lot of, you know, data that's, you know, been created in the past that's not structured in that way. So we have to go back in time and change all of, you know, kind of the way we've previously been storing data and update it so that it follows these guidelines and, you know, hopefully people that are developing these new applications that consumers find useful will kind of conform to this certain set of standards. The same way they conform to standards that are needed to have their, the data that they generate be used on an iPhone or be used on an Android phone. They can generate it in a way where it can be used on these healthcare platforms that are helpful to us. Thank you, Dr. Gate, for joining us on this very special episode of The Lead. Stay tuned for more episodes from the Heart Rhythm Society. Thanks for having me.

AI wearables

health data

interoperability

data standardization

healthcare systems