We have multiple, awesome, experienced cardiac experts on ST. A general Q…how.much trust should we have in ECG from wearable devices? What about the automatic arrhythmia detection? I have a reason to ask, but the backstory is uninteresting to most. IOW…if a wearable automatically identifies an arrhythmia and records a wonky looking ECG, should that be accepted as a probable medical issue or more likely to be the result of user error…instrument inadequacy… circumstance…something else?
Great question.
I will put my usual disclaimer up…MD here (but not a cardiologist).
Some questions first:
Which wearable are you or the person in question using?
What warning(s) are appearing?
How frequent have the warnings been?
Do they match with any signs or symptoms you/they may have at the time?
If you don’t mind sharing, what is the back story? What sort of cardiac investigations and diagnoses have you had before?
The first question is important as the accuracy can be device specific and knowing what you are using and what sort of warnings are appearing will help somewhat in terms of how to interpret things and also allow the responders to dive into that specific device’s technology and how they decide something is abnormal versus not…
In the operating theatre there are lots of things that can interfere with our ECG traces during cases, movement artifacts, diathermy, loss of contact etc etc-how each wearable filters that sort of stuff out will depend on the actual device etc…
We have multiple, awesome, experienced cardiac experts on ST. A general Q…how.much trust should we have in ECG from wearable devices? What about the automatic arrhythmia detection? I have a reason to ask, but the backstory is uninteresting to most. IOW…if a wearable automatically identifies an arrhythmia and records a wonky looking ECG, should that be accepted as a probable medical issue or more likely to be the result of user error…instrument inadequacy… circumstance…something else?
As always, you should talk with your own MD about what wearable you have and how you are going to use that information. There is considerable risk in using an EKG from a wearable for diagnosis of an acute MI (or even a potentially life-threatening arrhythmia like WPW etc). The patient and the doctor should understand all risks/benefits of using this information.
Known limitations:
Limited information: Because a wearable device is a single-lead ECG, it provides less information about the heart’s electrical activity than a 12-lead ECG. This means that certain heart conditions may not be detectable with a smartwatch ECG. Limited accuracy: Smartwatch ECGs use a single electrode sensor, which may not provide as accurate measurements as the multiple electrodes used in a 12-lead ECG. This means that the results of a smartwatch ECG may not be as reliable as those from a traditional ECG.Limitations in detecting cardiac conditions: Single-lead ECGs typically cannot detect certain cardiac conditions, such as a bundle branch block or a left ventricular hypertrophy. Interference: The signal detected by the single-lead ECG is prone to noise, interference and artifact which can affect the quality of the ECG tracing. Dependence on correct positioning of the device: Smartwatch ECGs require the user to wear the device in the correct position to ensure accurate measurements. Not intended for diagnosis: The smartwatch ECG is not intended to diagnose, cure, mitigate, treat, or prevent any disease or medical condition and should not be used as a substitute for a traditional ECG or medical advice.That said, EKG’s from some devices are really awfully good-like the Apple watch and Kardia device. They can give very useful data about a possible arrhythmia using basically a single lead (ie lead 1 of an EKG). Heart attacks/ Acute MI is diagnosed using multiple leads and that is main reason why a wearable cannot diagnose a heart attack. It is also why ST changes from telemetry monitors in the hospital cannot be used for that either.
There are quite a number of software aps that use AI with automated software analysis. Using these for detection/identification of afib vs sinus rhythm with ectopy is very, very good with those devices, most ~90-99% accurate at identifying afib. Even Garmin/Fitbit/Withings/Samsung watch have all gotten better over time and I think they are all FDA approved now for arrhythmia detection.
We use ZIO monitors as one of our main 24hr/48hr/3-7 day monitors and LINQ for Implantable Loop Monitors (ILR) over longer periods of time (months-years).
Up to date lists these:
â—DeepRhythmAI – The DeepRhythmAI (DRAI) software is used to detect arrhythmias and automatically analyze electrocardiograms (ECG) . DRAI is a cloud-based AI algorithm that analyzes all the heart beats in the processed ECG signal and based on that classifies them as either correct or arrhythmic.
â—Verily Study Watch with irregular pulse monitor – The Study Watch is a miniature wearable device that continuously monitors the ECG and heart rate, among other data .
â—Zeus system Zio watch – The Zio Watch Service is used in adults with atrial fibrillation (AF) or susceptibility to AF to analyze the ECG and photoplethysmogram, in order to detect AF .
â—Apple Watch atrial fibrillation history feature – The AFib History Feature analyzes pulse rate data from the Apple watch to identify irregular heart rhythms suggestive of AF, and provides an estimate of the time spent in AF . It includes a medical app on the Apple Watch as well as one on the iPhone.
â—Apple IRNF app – The Irregular Rhythm Notification Feature is a medical application that analyzes pulse rate data from the Apple Watch and notifies the wearer if AF is present .
â—AliveCor KardiaMobile 6L and AliveCor QT – KardiaMobile 6L is a device to obtain an ECG, which clinicians can use to manually measure their patients’ QT interval . AliveCor QT Service is cloud-based software that provides clinicians with the heart-rate corrected QT.
â—Implicity IM007 – IM007 records and analyzes ECG data from Insertable Cardiac Monitors (ICM) to help diagnose arrhythmias .
â—LINQ II Insertable Cardiac Monitor, Zelda AI ECG Classification System – The LINQ II ICM is an insertable device that records ECG . It can be activated by the patient or can be automatically activated. It is used in adults with risk for or symptoms of arrhythmias.
â—Zio ECG Utilization Software and AT ECG Monitoring System, Zeus System – The Zio ECG Utilization Software and AT ECG Monitoring System process ECG data from compatible monitoring devices, allowing for detection of arrhythmias (either automatically or via analysis of patient triggered events) . After data collection, a report is generated for clinician review. These devices are used in patients at risk for or with symptoms of an arrhythmia.
â—VX1 – The VX1 is a machine learning software that helps providers identify intracardiac atrial electrograms exhibiting spatiotemporal dispersion during an ablation procedure for AF .
â—RX-1 Rhythm Express Remote Cardiac Monitoring System – The Rhythm Express is a remote cardiac monitoring system . The device connects to standard ECG electrodes to capture two channel ECGs. An embedded algorithm processes the acquired ECG to detect arrhythmias, compress the ECG, and remove most in-band noise without distorting the electrocardiogram. It can function in one of three modes: mobile cardiac telemetry, event recorder, and wireless Holter, and can be worn by patients for up to two weeks.
â—BodyGuardian Remote Monitoring System – The BodyGuardian System detects and monitors cardiac arrhythmias in ambulatory patients . The device makes use of commercially-available, external plug-in devices such as ECG sensors, blood pressure meters, and pulse oximeters.
â—AI-ECG Tracker – The AI-ECG Tracker uses machine learning to assess ambulatory ECG data to identify arrhythmias in adults without pacemakers . The interpretations by the analysis program can be confirmed, edited, modified, or deleted by qualified clinicians.
â—AI-ECG Platform – The AI-ECG Platform uses machine learning to assist clinicians in the interpretation of a 12-lead resting ECG. The interpretation by the analysis program may then be confirmed, edited, or deleted by qualified clinicians .
â—Eko Analysis Software – The Eko Analysis Software assists in the evaluation of patients’ heart sounds and ECG . The software detects suspected murmurs as well as AF and normal sinus rhythm. It can also calculate heart rate, QRS duration, and electromechanical activation time. It does not identify arrhythmias other than AF and does not identify the type of murmur.
â—PhysIQ Heart Rhythm and Respiratory Module – The physIQ Heart Rhythm and Respiration Module is used to calculate heart rate and heart rate variability, to detect AF, and to determine respiration rate by automated analysis of ambulatory ECG data and triaxial accelerometers .
â—Loop System – The Loop System records arterial oxygen saturation, heart rate, and respiratory rate in adult patients in the home environment. This data is subsequently transmitted to a web server for remote review by a clinician .
â—KardiaAI – KardiaAI is a software analysis library that assists in the assessment of ECG rhythms (eg, normal sinus rhythm, AF, bradycardia, and tachycardia) in ambulatory electrocardiograms recorded from a variety of devices including event recorders .
â—RhythmAnalytics – RhythmAnalytics is a software that assesses for cardiac arrhythmias using single-lead ECG data; clinicians should review and confirm the analytic result .
â—Rx-1 Rhythm Express Remote Cardiac Monitoring System – The Rhythm Express remote cardiac monitoring system provides continuous monitoring of ECG, functioning in one of three modes: mobile cardiac telemetry, event recorder, and wireless Holter .
â—Cardio-TriTest v6.5 – The Cardio-TriTest v6.5 combines the results from a 12-lead ECG, a 4-lead phonocardiograph, and a 4-lead mechanocardiograph into a contiguous presentation that aids in diagnostic interpretation by a clinician .
â—FibriCheck – FibriCheck can be used by patients to self-monitor for AF on an intermittent basis .
â—Rooti Rx ECG Event Recorder, Rooti Link APP Software – Rooti Rx System allows an outpatient to record a single-lead ECG for review by their health care provider .
â—Peerbridge Cor™ System – The Peerbridge Cor System is a wearable monitoring device that captures and transmits symptomatic events and continuous ECG data for 24 hours and up to seven days and provides the data to the clinician for review .
â—CardioLogs ECG Analysis Platform – The CardioLogs ECG Analysis Platform analyzes data that is recorded from compatible devices used for the arrhythmia diagnostics (eg, Holter, event recorder) . It can be electronically interfaced and perform analysis with data transferred from other computer-based ECG systems.
â—Reveal LINQ – A Reveal LINQ insertable cardiac monitor is a small device placed under the skin that continuously monitors heart rhythm for up to three years .
â—FEops HEARTguide – FEops HEARTguide performs computer simulation of transcatheter left atrial appendage occlusion (LAAO) device implantation during procedural planning . It predicts implant frame deformation to guide LAAO device size and placement.
Thanks for responding. Yes, I’m sure that the interpretation depends considerably on context with the details you ask about. I may not have every bit of info but…20 y/o male with previous suspected arrhythmia and syncope. Woke up in the ambulance. Limb tingling and numbness before lights out. Cardiac conduction study discovered one location for ablation and that was performed. Three weeks later, eating dinner sitting down, limb tingling and automatic afib detection by a withings scanwatch health mate 5.16.1 wrist device. Automatically recorded 30 sec of ECG. Pretty wonky looking with mostly regular, identifiable QRS. But T as intense as Q and no real evidence of P. In 30s, about 6 totally malformed signals, mostly having large negative excursions. No syncope, felt better after a minute or so. No additional ECG.
Not me - son who is in Germany on a language fellowship for a couple of weeks.
Thanks for any insight. Electocardiologist who did the procedure reviewed the ECG and says…motion artifact, carry on, no worries. He (son) will be in college soon, 7 hr drive from home so we are getting him an electrocardiology consult close to school. Might be a few weeks, I suspect.
Edited to clarify that son is in college soon, not electrocardiologist (thankfully!) 
Thanks, as always, for your time, experience, insight and willingness to share! Have a look, if you like, to my response to @Amnesia with some additional details of this specific situation.
What a list! I’ll be interested in looking this over, out of curiosity more than anything. But not right now… Late for my morning run!
Thanks again for all your valuable input.
Thanks for responding. Yes, I’m sure that the interpretation depends considerably on context with the details you ask about. I may not have every bit of info but…20 y/o male with previous suspected arrhythmia and syncope. Woke up in the ambulance. Limb tingling and numbness before lights out. Cardiac conduction study discovered one location for ablation and that was performed. Three weeks later, eating dinner sitting down, limb tingling and automatic afib detection by a withings scanwatch health mate 5.16.1 wrist device. Automatically recorded 30 sec of ECG. Pretty wonky looking with mostly regular, identifiable QRS. But T as intense as Q and no real evidence of P. In 30s, about 6 totally malformed signals, mostly having large negative excursions. No syncope, felt better after a minute or so. No additional ECG.
Not me - son who is in Germany on a language fellowship for a couple of weeks.
Thanks for any insight. Electocardiologist who did the procedure reviewed the ECG and says…motion artifact, carry on, no worries. He (son) will be in college soon, 7 hr drive from home so we are getting him an electrocardiology consult close to school. Might be a few weeks, I suspect.
Edited to clarify that son is in college soon, not electrocardiologist (thankfully!)
Good that he is working with an EP doc as he has already had an ablation (?for WPW or AVNRT or something else-).
Many here on ST know that my son has PSVT from AVNRT, which pre-dated wearables and was more challenging to diagnose back then…
Thanks, as always, for your time, experience, insight and willingness to share
Thanks again for all your valuable input.
your are welcome