Monday, June 20, 2016

It is easy to be led astray by the computer....

I saw this ECG lying around:
The computer called this "normal" with no other comment.
what do you think?





















It is amazing that the computer called this normal, as there are clearly abnormal QRST's in beats 3, 4, and 5.

What are they?

I looked the case up on the McKesson system because one can highlight the run of abnormal beats in lead II across the bottom (see red box) and then one is able to see these abnormal beats in all 12 leads:
Now what do you think?








One could easily be fooled into thinking these are hyperacute T waves.

But the queen was not fooled by this into thinking it is OMI.  

The PM Cardio app does diagnose WPW.  The app called it Left anterior fascicular block.   It clearly recognized it as not normal.  




 






This is clearly WPW.  Among these beats there is clearly a short PR interval and delta waves.  The QRS is very abnormal due to the pre-excitation.

I looked at the patient presentation and it was unrelated (no tachycardia, no palpitations, etc.).  The ECG findings were not recognized by the emergency physicians.   The patient had been admitted to the hospital and no one had noticed.  No final ECG interpretation had been placed in the record before the patient was discharged, and he was discharged without recognition of the diagnosis of WPW.

On record review, the patient had been seen in the ED in previous years for palpitations and the ECGs were actually normal, with no evidence of WPW.  He had been diagnosed with anxiety (which he may indeed have, but it is common for patients with later-diagnosed SVT to be diagnosed with anxiety or panic attack).  No doubt he had been having runs of tachycardia due to WPW.

I am certain that when the final read was placed by the interpreting physician that it would have been correct.  In this instance, I put that final interpretation into the system, added the diagnosis, and notified the primary care physician.

Learning Point:

You cannot trust the computer interpretation!  You must carefully look at every tracing yourself.  Use the computer's interpretation, but do not rely on it.

I suggest:

1. Read it yourself while hiding the computer interpretation
2. Then read the computer interpretation (it may see things that you did not)
3. Then look again

Dr. Ken Grauer has an excellent post on computer interpretations:
http://ecg-interpretation.blogspot.com/2016/05/ecg-blog-126-computerized-ecg.html

Also, see Ken's insightful comments on this case below:

GREAT case Steve! Thanks for citing my ECG Blog #126, in which I review a practical approach for optimizing benefits of computerized interpretations. I am equally amazed as you in this case that the computer did not pick up on at least some abnormality … but the key for anyone who is less than a true ECG expert lies in your 1st suggestion = HIDE the computerized interpretation BEFORE you look at what the computer said. Had that been done, the WPW that is obvious on this tracing would not have been missed.

Often overlooked is the concept that patients who have an accessory pathway may conduct normally at some times and abnormally at other times. And sometimes, they may split the relative amount of conduction passing over normal and accessory pathways even from beat-to-beat (known as a “Concertina effect). The “good news”, is that finding a Concertina effect suggests a relatively longer refractory period for the accessory pathway — and therefore a relatively lower risk of sudden death (http://casereports.bmj.com/content/2013/bcr-2013-009328.full ).

The interesting thing to me is how the 2nd beat in the long rhythm strip in your example would look relatively “normal” by itself. However, when compared to the 1st beat in the rhythm strip, we clearly see the difference. So there is FUSION between normal conduction (PQRST morphology of the 1st beat) — and purely conducted WPW beats ( = beats #3,4,5). Note how there once again is a different degree of fusion for the 6th beat in the long lead II rhythm strip. So we are alternating between normal and accessory-pathway conduction in this rhythm strip … Note also how differently delta waves appear in different leads. Delta waves are EASY to recognize in leads I and aVL (because they are positive). Delta waves are negative in leads III and aVF — and in lead II to we see a multiphasic almost isoelectric initial component to the delta wave. It is because of some fusion with normal conduction and this near isoelectric delta wave appearance that by themself, it would be difficult to identify WPW from beats that look like beat #2 and beat #6 in the long lead rhythm strip.




9 comments:

  1. Great teaching ECG! I have found that intermittent conduction over an accessory pathway is often missed because it is frequently just a few beats and is usually written off as "aberrant conduction." I have also found that my students are much less likely to recognize a negative delta wave than a positive one. I did my own little study some years back (this is anecdotal, not scientific) regarding negative delta waves and how they can be distinguished from Q waves. The most distinguishing feature that I found is the initial slurring. Yes... I WAS able to find some real Q waves that had a bit of initial slurring, but not like what one typically sees with a negative delta wave (like the ones in THIS tracing). You can also use the old "invert it and turn it around backwards" trick to see that the "Q" wave suddenly DOES look remarkably like a delta wave!

    If one is going to depend on the machine interpretation, he or she had better make certain that someone who really CAN interpret an ECG is nearby.

    ReplyDelete
    Replies
    1. Jerry — I completely agree with your comments regarding negative delta waves. However, as to use of computerized ECG interpretation — my point (which is succinctly summarized in the 3-page pdf that Steve links to above) — is that the computerized report CAN be helpful, even to those interpreters who are not expert in ECG interpretation. The KEY is simply that they COVER UP what the computer initially says, and thoroughly make their own independent systematic interpretation first. THEN they can look at what the computer says. The computer might remind them of something that they DO know and simply forgot to mention. On the other hand — IF the computer says something different that they do not understand — then they need to find someone with more expertise. BUT, just like getting an extra opinion from a colleague (who might not necessarily be expert in ECG interpretation) — being provided with other possibilities often evolves to problem solving that leads to an appropriate answer. BOTTOM LINE (that I think both you and I completely agree on) — No one should ever "depend" on the machine interpretation without independent thought and consideration of findings.

      Delete
    2. Thanks, Ken. I DO agree that interpreting the ECG first and THEN looking at the machine interpretation is a great learning method, but I know that isn't the way it is happening out in the real world based on my own experience with other physicians and colleagues. Anyone who follows a blog of this advanced level would certainly do their own interpretation first, but there are a lot of healthcare providers who have to look at ECGs in the course of their practice and they simply aren't up to the task of confidently interpreting them themselves due to lack of training. I think being exposed to all the analyses and opinions expressed on this site by people who obviously have very advanced ECG interpretation skills can give the wrong impression about the capabilities of so many of our colleagues who are actually interpreting 12-lead ECGs at what I would call a beginners' level. I would be concerned that the person with less training and confidence might still accept the machine interpretation as definitive when it's quite possible that BOTH interpretations may have overlooked something important. And even when the less experienced reader's interpretation IS correct, they may still defer to the machine interpretation because it mentions something of a more serious nature - even though it is wrong!

      The ECG is one of the most ordered (if not THE most ordered) tests in medicine. I strongly feel that there should be formal, scheduled classes in electrocardiography at the medical school level that should take each student to at least an intermediate level of skill. In addition to teaching them how to better interpret an ECG, they would be in a much better position to recognize something that is not right and requiring a more advanced interpreter.

      Delete
    3. Hi Jerry. Just a comment first on your last paragraph — I taught a formal course at UF College of Medicine to their 125 medical students for 24 straight years. I developed the course and wrote the text book, which I gifted to all 125 students. For most of that time, I was given a full 11 hours to teach over a 2-week span — and it resulted in med students outshining some residents on the medical floor during rotations. Then the curriculum was “updated” — and my course was cut to 4 hours over my last few years. Despite offering to continue teaching the course indefinitely after I retired in 2010 for minimal compensation — the course was dropped entirely. I don’t know if anything ever got reinstated … I guess the “Powers” just didn’t feel the subject was important … My impression was all-too-many medical schools just don’t have any formal ECG teaching. You and I have both seen the result — new residents are often woefully deficient, lack confidence, and have trouble assessing acute cardiac situations for the first bunch of months of their residencies until they finally catch up on the basics of ECG/Arrhythmia interpretation …

      Otherwise, if the message NOT to look at the computerized interpretation (until after you make your own independent interpretation) is emphasized, along with emphasis that the computer DOES make mistakes (!) — then I still believe the computer can (and should) be helpful. If clinicians are taught to routinely use a systematic approach with each and every ECG they interpret — then even when they do not know what everything they see is, they should be able to recognize the vast majority of what they know they do not know. And of course, ECGs need to be overread by competent interpreters so that hopefully major findings don’t get missed.

      THANKS for your comments! This obviously is a passionate subject for me. I DO think you and I approach most issues in this regard quite similarly!

      Delete
  2. GREAT case Steve! Thanks for citing my ECG Blog #126, in which I review a practical approach for optimizing benefits of computerized interpretations. I am equally amazed as you in this case that the computer did not pick up on at least some abnormality … but the key for anyone who is less than a true ECG expert lies in your 1st suggestion = HIDE the computerized interpretation BEFORE you look at what the computer said. Had that been done, the WPW that is obvious on this tracing would not have been missed.

    Often overlooked is the concept that patients who have an accessory pathway may conduct normally at some times and abnormally at other times. And sometimes, they may split the relative amount of conduction passing over normal and accessory pathways even from beat-to-beat (known as a “Concertina effect). The “good news”, is that finding a Concertina effect suggests a relatively longer refractory period for the accessory pathway — and therefore a relatively lower risk of sudden death (http://casereports.bmj.com/content/2013/bcr-2013-009328.full ).

    The interesting thing to me is how the 2nd beat in the long rhythm strip in your example would look relatively “normal” by itself. However, when compared to the 1st beat in the rhythm strip, we clearly see the difference. So there is FUSION between normal conduction (PQRST morphology of the 1st beat) — and purely conducted WPW beats ( = beats #3,4,5). Note how there once again is a different degree of fusion for the 6th beat in the long lead II rhythm strip. So we are alternating between normal and accessory-pathway conduction in this rhythm strip … Note also how differently delta waves appear in different leads. Delta waves are EASY to recognize in leads I and aVL (because they are positive). Delta waves are negative in leads III and aVF — and in lead II to we see a multiphasic almost isoelectric initial component to the delta wave. It is because of some fusion with normal conduction and this near isoelectric delta wave appearance that by themself, it would be difficult to identify WPW from beats that look like beat #2 and beat #6 in the long lead rhythm strip.

    ReplyDelete
  3. To some extent it is very surprising that the computer called this ECG normal given several abnormalities (short PR, wide QRS, deep negative T waves); it is more surprising because it fails just on its “job” which is mainly to calculate intervals, segments and so on.
    Things are much worse when the computer interprets ECGs of patients with pacemakers (it does not spot even the spikes).
    In the present ECG perhaps would it be different if there were 8 anomalous QRS complexes instead of three in the rhythm strip?

    Mario

    ReplyDelete
  4. Hello , my novice thoughts , the PR interval eye at about and not under 120 ms in the leads with clear delta wave activity. It therefore made me think of intermittent abberant conduction.

    ReplyDelete
  5. Best post thank you so much too sharing to us!

    ReplyDelete
  6. Hi... Extremely useful discussion

    ReplyDelete

DEAR READER: I have loved receiving your comments, but I am no longer able to moderate them. Since the vast majority are SPAM, I need to moderate them all. Therefore, comments will rarely be published any more. So Sorry.

Recommended Resources