Sustained Atrial Fibrillation or Not: The Vagaries and Inaccuracies of AliveCor/Kardia and Computer Interpretation of ECG Rhythm

The skeptical cardiologist has often sung the praises of the AliveCor Mobile ECG for home and office heart rhythm monitoring (see  here and here.) However, there is a significant rate of failure of the device to accurately identify atrial fibrillation.  I’ve seen numerous cases where the device read afib as  “unclassified” and normal sinus rhythm (usually with PVCs or PACs) called afib both in my office and with my patient’s home monitors.
In such  cases it is easy for me to review my patient’s  recordings and clarify the rhythm for them.
For those individuals who do not have a img_8322cardiologist available to review the recordings, AliveCor offers a service which gives an option of having either a cardiac technician or cardiologist review the tracing. The “cardiac technician assessment” costs $9 and response time is one hour. The “Clinical Analysis and Report by a U.S. Board Certified Cardiologist” costs $19 with 24 hour response time.
Obviously, I have no need for this service but I’ve had several readers provide me with their anecdotal experiences with it and it hasn’t been good.
One reader who has a familial form of hypertrophic cardiomyopathy utilizes his AliveCor device to monitor for PVCs. One day he made the following recording which AliveCor could not classify:
screen-shot-2017-02-14-at-5-43-51-am screen-shot-2017-02-14-at-5-41-41-am
He then requested a technician read which was interpreted as “atrial fibrillation sustained.”
He then had requested the cardiologist reading which came back as Normal Sinus Rhythm.
Finally, he againscreen-shot-2017-02-14-at-5-44-09-am requested the technician
read and got the correct reading this time which is normal sinus rhythm with PACs
When my reader protested to Kardia customer service about this marked inconsistency: three different readings in a 24 hour period, a Kardia  customer service rep responded :

 I was able to review this with our Chief Medical Officer who advised that the recording shows Sinus Rhythm with PACs. The Compumed report seldom provides identification of PACs and PVCs as most cardiologists believe they are not significant findings. The sustained AFib finding was incorrect, so I have refunded the $5 fee you had paid.

Please let us know if you have any other questions.

As I pointed out in my post on palpitations, most PVCS are benign but some are not and patients with palpitation would like to know if they are having PVCS and/or PACs when they feel palpitations.

More importantly, the misdiagnosis of afib when the rhythm is NSR with PACs or PVCs can lead to extreme anxiety.

This tracing clearly (to me) shows regular and similar upward deflections (red arrow, p waves) which are a similar distance from the QRS complexes which follow (QRS complexes). The green arrows point to irregular deflections due to noise which can confuse computer algorithms (and non-cardiologists.) The distance between the QRS complexes is very regular (black arrow, RR interval). Thus, this is clearly normal sinus rhythm (NSR). Later in the recording PVCs (green arrows) are noted occurring every other beat. The distance between the QRS complexes on either side of the PVC is still the same as two RR intervals. This is clearly ventricular bigeminy.

Heres a recording
I made in my office this morning on a patient with cardiomyopathy and a defibrillator.

This is very clearly NSR with PVCs yet AliveCor diagnosed it as “possible atrial fibrillation.”

The AliveCor algorithm  is not alone in  making frequent errors in the diagnosis of atrial fibrillation.

The vast majority of ECGs performed in the US come with an interpretation provided by a computerized algorithm and medical personnel rely on this interpretation until it can be verified or corrected by an overreading cardiologist.

One study demonstrated that computerized ECG interpration (ECG-C) is correct only 54% of the time when dealing with a rhythm other than sinus rhythm
Another study found that 19% of ECG-C misinterpreted normal rhythm as atrial fibrillation. Failure of the physician ordering the ECG to correct the inaccurate interpretation resulted in change in management and initiation of inappropriate treatment, including antiarrhytmic medications and anticoagulation, in 10% of patients. Additional unnecessary diagnostic testing was performed based on the misinterpreted ECGs in 24% of patients.
When lives or peace of mind are at risk you want your ECG interpreted by a cardiologist.
I would like to take this opportunity to personally issue a challenge to IBM’s Watson.
Hey, Watson, I bet $1,000 I can Interpret cardiac rhythm from an ECG with more accuracy than you can!
Are you listening, IBM?
Do you copy, Watson?


9 thoughts on “Sustained Atrial Fibrillation or Not: The Vagaries and Inaccuracies of AliveCor/Kardia and Computer Interpretation of ECG Rhythm”

  1. I have many PVC Beats., Consequently I have many misreadings , excepts when it reads normal it appears to be a normal tracing.

  2. Hi. So where do you stand regarding the accuracy of the Kardia now? I have purchaced one recently, primarily to give me piece of mind and ready insight into what my heart is doing. Should we not believe what the device or the supporting staff tell us? Thanks … I am enjoying your blog.

    • Julie,
      Thanks for your comment. I am a big fan of the Kardia device and continue to recommend it to my patients with atrial fibrillation on a regular basis. If you read what I wrote recently about Kardia Pro which allows me to monitor what my patient’s are recording you can get a feel for how useful it is.
      It is only capable of differentiating afib and sinus rhythm as I’ve pointed out elsewhere so premature beats sometimes are called normal and sometimes unclassified in which case it is useful to have a cardiologist review the tracings.

  3. I purchased the Kardia app at the suggestion of the PA at my cardiologist office after I had a syncopal with SVT at work. It is an interesting but frustrating gadget. I printed several readings that the app calls NSR but shows some elevated hr with what the cardiologist said was probably artifact. He does not seem interested in the ecg. Now I’m wearing the Zio monitor so I’ll be interested in the report it generates. As a former ER tech turned ER nurse & now a NICU nurse I’m very used to seeing ecg rhythms and have shared the Kardia reports with my co-workers who are equally curious since I scared the the crud out of them with my syncopal in the middle of the NICU and everyone agrees – not NSR. Just curious what your thoughts are on it and possible “artifact”. I can’t seem to insert the picture here but if your interested feel free to ask via email.

  4. I live in an area not well served by medical specialists. After a bad bout of pericarditis (misdiagnosed and untreated for two months), my previously healthy heart began experiencing a-fib and pvc’s in abundance. Have been reading your blog for awhile, and noting with interest the reportage on Kardia. A PA recommended it to me when I asked about home devices, and the optional service of having readings done by a tech or cardiologist sealed the deal. I viewed it as an adjunct to my cardiologist who books months out on appointments and is four hours away, certainly not a replacement–but I was hoping for some peace of mind. Two months later I am really disappointed. The computer reading is unreliable and the few readings I have paid for have been indecipherable. What is one to think with a reading of: “REGULAR RHYTHM, UNDETERMINED ORIGIN: A potentially abnormal rhythm whereby the origin of the electrical activity cannot be determined.” and a recommendation to seek review within 24 hours. Seriously?? Is the rhythm regular or not?? Driving 8 hours round trip to find out is not trivial. And Kardia on the whole seems very light on tech support. There were some inconsistencies in the documentation as it pertained to the positioning of the device and I wrote for clarification twice and never received a response. I would happily pay more money for better accuracy and interpretation. Really enjoy your blog BTW.

  5. As usual, I enjoyed your post.
    I am *very* interested to see how you do against Watson! He has quite a rep!
    Costa Rica


Please leave your comments. The skeptical cardiologist loves feedback. He reads all and replies to all that warrant a reply.