For Patients With Palpitations and PVCs Kardia's Determinations Can Be Both Advanced and Misleading

Only one consumer device has been FDA-approved to identify PVCs but Alivecor’s Kardia ECG may diagnose normal sinus rhythm even if it identifies PVCs in your ECG

In 2021 I wrote about Alivecor’s Advanced Determinations algorithm which allows the company’s hand-held consumer devices (Kardia 1L, Kardia 6L, and Kardiamobile (aka "Kardiacard" ) to instantly and accurately diagnose two common cardiac rhythm abnormalities: premature ventricular contractions (PVCs) and premature atrial contractions (PACs.)

I’ve been outfitting my patients who suffer from intermittent atrial fibrillation with Kardia devices since 2014 so that they are empowered to know immediately whether the palpitations or racing heart or sense of anxiety they are feeling is due to atrial fibrillation or some other cardiac arrhythmia. 

But I and millions of other Americans suffer from palpitations that are due to premature or extra beats that often confused the personal ECG devices that many have obtained. 

This confusion resulted in many “unclassified” or "inconclusive diagnoses for both Apple Watch and Kardia, the major consumer devices being utilized for personal ECG monitoring.s

It appeared that Alivecor had solved their premature beat problem with Advanced Determinations and that my patients now could be reassured and informed of the precise rhythm abnormality causing their flip-flops or chest rumbling.

I described in detail how this work in my own case of "flip-flops" here.

Since then, dozens of my patients have been using the various Kardia devices to help them better understand their PVCs, PACs, and palpitations.

I had assumed that if Kardia detected a single PVC in the typical 30-second ECG recording it would report out "normal sinus rhythm with PVCs" but recently an astute reader of Skepcard alerted me to a problem:

I just wanted to let you know that, unless people look really hard, Kardia will actually show you if you have any PVCs, where they are located and how many, ON A NSR EKG! I have a 99 BPM EKG and it shows I have one isolated PVCs. So, can you do a post on this? People need to look at their "normal" ekgs also, because they may include PVCs.

Obligingly, I looked into this and found that my discerning reader was correct.

Here is the PDF of his Kardia recording

Note that the Kardia Advanced Determination reads "Normal Sinus Rhythm" with no mention of PVCs.

But, just below the statement on PVC annotation (blue rectangle) it says PVC 1.

The orange rectangle highlights the little black triangle which Kardia uses to mark something it thinks is PVC.

By the way, the lead I and lead II recordings of the beat in question clearly show a knowledgeable ECG reader that this is an artifact, not a PVC. Thus, it is probably good that Kardia didn't make the call off this one beat.

Although Kardia is quite accurate at identifying premature beats in most patients, like every computer-aided ECG (AI or not), it is not as accurate as an expert, experienced human ECG reader.

I asked Dr. David Albert of Alivecor about this and he informs me that:

"the diagnosis of SR with PVCs on Kardia devices is only made if there are more than 3 PVCs in a 30-second recording”

It is not entirely clear to me why a cutoff of >3 PVCs in 30 seconds was chosen before Kardia admits that PVCs are present. Perhaps this was felt to be the best trade-off between sensitivity and specificity.

But for most who purchased the Kardia to see whether their palpitations were due to a cardiac rhythm issue, Kardia is telling them the rhythm is normal when it is not and is therefore, not helpful.

A quick calculation reveals that 4 PVCs in a 30 second recording corresponds to 8 per minute and (1440 minutes/day x 8 PVCs/minute) 11, 520 PVCs in 24 hours. This frequency is greater than the 10,000 per 24 hours that many view as putting an individual at risk of PVC related cardiomyopathy. Perhaps this was the rationale for Alivecor.

But PVCs vary substantially over 24 hours and a 30 second segment will not predict the 24 hour burden so this seems misguided. In addition, only a small  minority of patients with PVCs have these high burdens (and thus increased risk of cardiomyopathy.)

Flippyfloppyingly Yours,

-ACP

N.B. The PVC annotation in the figure above is only visible if you create a PDF of the ECG tracing on your app. You won’t see it if you look at the recording in your health app on your smartphone. This adds another layer of obfuscation to the process.

I’ll delve into the additional information on that PDF in a later post.