Friday, August 11, 2017

Beware Automated Interpretations of Atrial Fibrillation!

See this ECG:

There is an irregularly irregular rhythm.
The Automated interpretation was "Atrial Fibrillation."
What is it?

Look at the lead II rhythm strip across the bottom.  There are clearly sinus P-waves for the first 6 beats, although they speed up.

This change of rate of the sinus node is called "sinus arrhythmia" and is related to vagal tone from inspiration (which increases vagal tone and slows down the rate, but this takes several seconds and this gets out of phase, which means that by the time it is slowing down, the patient is actually expiring).

Then beats 7 and 8 appear and do not show P-waves in lead II.  Are they junctional?  No!  Look above in V1-V3, and you clearly see an atrial beat but of a different morphology (coming from another focus in the atrium, and thus not a sinus beat).  This is occurring because the vagal tone is slowing the sinus node so much that a different part of the atrium "escapes," taking over the pacemaker function.

Beats 9 and 10 also appear to be preceded by subtle atrial activity, but of yet another morphology and thus from yet another focus in the atrium.

So there appear to be at least 3 atrial pacemakers here (3 foci).

When the rate is tachycardic (greater than 100) and there are at least 3 foci, then it is called multifocal atrial tachycardia (MAT), which is usually associated with COPD.  For more on MAT, see this lecture on Narrow Complex Tachycardias from minutes:seconds 23:44 to 26:50.

Since the rate is normal, this is called a Wandering Atrial Pacemaker.  It is benign.

2 reasons for an irregularly irregular rhythm in a narrow complex*

1. Multifocal atrial tachycardia
2. Atrial fibrillation
* Sinus arrhythmia appears to be irregularly irregular during the 10 seconds of a 12-lead ECG, but it has a regular pattern to it over more time (speeding up, slowing down, speeding up, slowing down).

Automated interpretations in atrial fibrillation

We compared the Veritas automated interpretation [a widely used algorithm on Mortara machines which is a conventional (if, then; instructional) algorithm] and a new deep neural network algorithm (Cardiologs).  We used an expert reference standard, and found that the Veritas had a very large number of false positive reads, more than Cardiologs.(1)

--> A 2004 study of 2298 ECGs from 1085 patients which had a computerized interpretation of AF found that in 442 (19%) of these ECGs, from 382 patients (35%), the interpretation was incorrect, and that, in 92 of these 382 patients, the physician had failed to correct it.  These errors resulted in unnecessary anti-arrhythmic and anticoagulant therapy in 39 patients and unnecessary diagnostic testing in 90 patients, and an incorrect final diagnosis of  paroxysmal AF in 43 patients.(2)

1.   Smith SW et al. Improved Interpretation of Atrial Dysrhythmias by a New Neural Network Electrocardiogram Interpretation Algorithm.  SAEM.  Abstract 670.  Academic Emergency Medicine 2017; 24(S1):S235. 


2.   Bogun F, Anh D, Kalahasty G, et al. Misdiagnosis of atrial fibrillation and its clinical consequences. Am J Med 2004;117:636-42.

Learning Point

It is easy to gloss over automated reads without scrutinizing them carefully.  Especially when the read is "Atrial Fibrillation", you must look carefully.  These misreads have adverse consequences!!



  1. GREAT clinical example by Dr. Smith! I’d add the following points: i) There are some things that computerized interpretations are good, or even excellent with. However, there are other things for which computerized interpretations are less good (if not poor) for. In my experience — ANY rhythm other than sinus on a computerized report needs to be carefully checked by the treating clinician; ii) If the clinician sees that the computerized interpretation of a rhythm is erroneous — then it is his/her responsibility to CROSS OUT the computerized interpretation, and to write in his/her corrected rhythm interpretation. iii) One of the most underused concepts in rhythm interpretation — is appreciation of the information that simultaneously-recorded leads can provide. As per Dr. Smith, despite loss of the P wave for the 7th and 8th beats in this tracing — a quick glance at simultaneously recorded leads V1,V2 PROVES that there IS still ongoing atrial activity. iv) I’m not convinced that there is another atrial focus for the last 2 beats on this tracing — as the marked underlying baseline artifact undulations make me uncertain if P wave morphology really changes for these last 2 beats — and since the lead switch is after beat #8 — looking at simultaneously-recorded leads V4,5,6 does not really help. v) In addition to the difference in heart rate — another essential difference for making the distinction between MAT vs WAP (Wandering Atrial Pacemaker) — is that there is beat-to-beat variation in both P wave morphology AND the PR interval with MAT — whereas with WAP, there is gradual variation in which you’ll see similar shapes (and similar PR intervals) for several beats in a row — and then the P wave shape will change. vi) Technically, one needs to see at least 3 different P wave morphologies in order to ensure a diagnosis of “WAP”. As per point v), I’m not convinced we yet have this. This emphasizes the clinical reality that most of the time, a significantly longer period of monitoring is needed to make a verified diagnosis of wandering atrial pacemaker. vii) In my experience, the diagnosis of MAT vs sinus rhythm with either wandering pacemaker and/or multiple PACs are all points on a SPECTRUM. Many patients “fail to read the textbook” — and it is common to see rhythm strips that don’t completely fit all criteria for any of these 3 diagnoses … In such cases, clinical setting may help in distinguishing (ie, a longstanding history of pulmonary disease favors MAT). viii) Although true that most cases of WAP are benign — as with almost everything else in ECG interpretation, Clinical Correlation is the KEY. So, the above rhythm might not necessarily be benign if it was recorded from an older adults with a history of weakness and blackouts — in which case the initial sinus bradycardia and arrhythmia could be among the rhythms associated with sick sinus syndrome. Bottom Line — GREAT illustrative case of many important concepts in arrhythmia interpretation. Thanks to Dr. Smith for presenting this!


Recommended Resources