www.elsevier.com/locate/jelectrocard

Journal of Electrocard

How I read, and teach others to read, ECGs

John E. Madias, MDTMount Sinai School of Medicine of the New York University, New York, NY, USA

Division of Cardiology, Elmhurst Hospital Center, New York, NY 11373, USA

Received 29 April 2005

I was asked by the editor of the journal to compose a few

lines about my approach to reading electrocardiograms

(ECGs), and how I teach ECG interpretation to trainees.

My approach to reading ECGs and instructing others is one

and the same. I taught myself electrocardiography during my

internship by reading Grant’s book [1]. After finishing that

book, I got into the habit of opening the volume at random

and looking at an ECG with the b legend to figure Q covered. Iwould interpret the tracing, then read the legend, and finally

read the part of the chapter dealing with the particular figure. I

extended this approach to my clinical routine where I would

read the ECG without any information to steer me to a

specific diagnosis, reread it after I permitted myself to know

something about the patient, and proceed to seek ancillary

non-ECG confirmation of my impressions. I should include

here that the last step led often to a lot of disappointments!

This went on while I read Grant’s book bcover to coverQrepeatedly. Subsequently, I found that other authors empha-

sized Grant’s approach in ECG interpretation [2-4]; partic-

ularly, Hurst [4] tirelessly continues to this day to extol its

merits in mastering ECG reading, and studying his writings

[4] will be very rewarding. Although I use the vector

approach for comprehending the P wave, QRS complex,

ST segment, and T-wave magnitude and spatial direction, I

still cannot help using in parallel the pattern recognition

approach, that is, an associative linking of the ECG

appearances with general and/or specific pathophysiologic

entities. In using this hybrid method, the pattern approach is

usually activated first, acting in turn as a stimulus for the

vector reasoning of the entire ECG curve, before it fades

away in the background.

In essence, I am the sole ECG interpreter of ECGs in my

hospital and have a large workload of tracings. I interpret

each ECG looking only at it, and I examine the automated

0022-0736/$ – see front matter D 2006 Elsevier Inc. All rights reserved.

doi:10.1016/j.jelectrocard.2005.07.008

T Division of Cardiology, Elmhurst Hospital Center, Elmhurst, NY

11373, USA. Tel.: +1 718 334 5005; fax: +1 718 334 5990.

E-mail address: madiasj@nychhc.org.

measurements and computer diagnosis afterward. With rare

exceptions, I feel comfortable using the automated ECG

measurements, but I find the diagnoses often inappropriate. I

am convinced that there is room for improving the automated

ECG interpretations by altering the diagnostic criteria

embedded in the interpretive algorithms.

A sine qua non element in my workflow is that I seek non-

ECG confirmation of my diagnosis of a particular ECG by

checking clinical and laboratory data of the patient involved

when ECG measurements or features are impressively

abnormal or discordant. I will seek, for example, the non-

ECG information when there is left ventricular hypertrophy

and right axis deviation, left ventricular hypertrophy with tall

R waves in the right precordial leads, bP-pulmonaleQ (rightatrial abnormality ) in the limb leads and bP-mitraleQ (leftatrial abnormality) in the precordial leads, or right axis

deviation in an elderly individual. This does not take long

because data can be retrieved electronically.

When I instruct trainees (fellows, residents, and medical

students) to read ECGs, I advise them to read a text on

electrocardiography cover to cover while they are under my

supervision. Usually, I refer them to the bWagner-Marriott

bookQ [5] and, occasionally, to a specific chapter of

Comprehensive Electrocardiology [6]. I teach, not while

reading the daily load of ECGs, but on cardiac care unit and

consult rounds and during selected clinic sessions. My style

of teaching also depends on the competence or sophistication

of the student.

Although I emphasize the vector approach, I do not

condemn the pattern approach in reading the ECGs. All

consult or clinic encounters start with the ECG, with its upper

part being concealed. The trainee who is not familiar with the

clinical particulars of a specific case is selected to interpret the

ECG, whereas for one or more occasions per session, I

attempt to bpresent the case Q through the ECG myself,

without knowing anything about the patient. The greatest

educational benefit for the trainees and myself is usually

iology 39 (2006) 110–111

J.E. Madias / Journal of Electrocardiology 39 (2006) 110–111 111

provided by cases where their or my predictions are entirely

wrong! For the actual ECG reading, the trainee reads the ECG

tracings without any clinical information or automated ECG

measurements. A second ECG interpretation follows with the

benefit of all the information previously withheld. The third

ECG interpretation consists of checking other non-ECG

information, that is, clinical data and results of other

laboratory tests (chest x-ray, echocardiograms, computed

tomographic scans, magnetic resonance images, thallium

images, hemodynamic information, angiographic images,

etc), which often confirm our ECG interpretation but

occasionally refute it. These refutations are often the starting

point for contemplation about the particular bproblem ECG.QIn giving assignments to trainees, I have found that it is

conducive to their learning to focus only on a narrow area at a

time, that is, one topic, or even a detail about a topic (thus, eg,

I may suggest that they read about left anterior fascicular

block or the differential diagnosis of left axis deviation).

Throughout the process of interpreting an ECG, even the

most seemingly buninteresting one,Q alone or in the company

of trainees, I often attempt to convince myself and convey an

aura to the group that the ECG at hand is an binterestingtracing,Q an ECG with bconcealed complexity,Q one that may

represent a bcuriosityQ or bpublishable material,Q or one

bsuitable for research purposes.Q This attitude benefits the

trainees by creating enthusiasm while tackling ECG records.

Currently, fellows are more interested to learn to read the

ECGs in a bsystematic wayQwith emphasis on the bdiagnosticcoding,Q a feature that they need to have practice on because itis used in the cardiovascular board examination. In reference

to this, I should mention that a biweekly ECG conference is

scheduled, where a series of ECGs selected at random from a

currently popular book [7] is systematically interpreted by

trainees using the coding sheets (also used during the board

examinations), which comprise 14 categories of diagnostic

statements and 128 different codes. I have found the

exposition in this bguide Q [7] to be redundant (for the

competent reader) with the multiple quizzes, answers,

questions, and codes; nevertheless, I consider these exercises

to be very useful for the trainees. Although the above remarks

on how I (and instruct others to) read ECGs refer to the

12-lead ECG for the purposes of this communication, I have

extended this approach to the interpretation of ambulatory

(Holter) ECGs, patient-activated long-term bevent recorders,Qsignal-averaged ECGs, and ECGs generated during exercise

and pharmacologic stress testing.

I had some theoretical backing or justification of the

approach I have used/taught in ECG interpretation many

years ago when an editorial by Jerry Kassirer (then the editor

of the New England Journal of Medicine) recommended the

application of the biterative hypothesis testingQ in medical

practice [8]. According to this idea, it does not matter where

one starts from tackling diagnostic or therapeutic particulars

of a patient’s case. One can start from a laboratory finding or a

symptom, a physical sign, or even a therapeutic detail from a

patient’s file. What is important is that one starts with very

little or no information and builds up the case by asking as

relevant as possible questions, provides a reason why such

questions were asked, acts upon the sequential bits of

information given, and explain why each new piece of

information brings us closer to the solution of a problem at

hand. The exposition of the editorialist, in a way, provided

some legitimacy to the haphazard (nonplanned; random)

bnonsystematic Q approach I have used in interpreting ECGs.Another source of inspiration for me have been the books

of Sodi-Pallares [2,3]. In the foreword of one of the books

[3], the author discusses a bpolyparametricQ approach in

interpreting the ECG and counters others bwho feel that we

make too heavy demands on our ECG interpretations,Q bwepresume more than what is actually recorded on the tracing,Qand bprobably some cardiologists will feel that we are

abusing the ECG interpretations in this polyparametric ap-

proach.Q He goes on to define deductive electrocardiographyas a bcorrelation of the ECG diagnosis to the clinical,

radiological, and postmortem examinations.Q Nowadays, thearray of diagnostic tests available for ECG/non-ECG

correlations is expanding by leaps and bounds, to the

advantage of our efforts to extract as much diagnostic and

prognostic information from the ECG. Sodi-Pallares was an

advocate of many ECG interpretations in series, with the first

starting without knowledge of the clinical information. Thus,

to learn to read ECGs competently, one needs to read a good

ECG book, struggle with each ECG tracing unsupported by

relevant clinical information, b reinterpret Q the tracing in the

context of such clinical data, and b re–re-interpret Q the ECGin the light of insights provided by the non-ECG laboratory

testing. With this approach, the practitioner soon becomes

adept in going through these stages quickly, efficiently, and

with good results. Have fun!

References

[1] Grant’s clinical electrocardiography. The spatial vector approach. 2nd ed.

New York7 McGraw-Hill, Inc; 1970 [revised by Julian R. Beckwith].

[2] Sodi-Pallares D, Calder RM. New bases of electrocardiography. St.

Louis (Mo)7 CV Mosby; 1956.

[3] Sodi-Pallares D, Medrano GA, Bisteni A, et al. Deductive and

polyparametric electrocardiography. Instituto National de Cardiologia

de Mexico; 1970 [English translation by Macossay CR, and Dunn M,

Mexico DF].

[4] Hurst JW. Methods used to interpret the 12-lead electrocardiogram:

pattern memorization versus the use of vector concepts. Clin Cardiol

2000;24:4.

[5] Wagner GS. Marriott’s practical electrocardiography. 10th ed. Phila-

delphia7 Lippincott Williams & Wilkins; 2001.

[6] Macfarlane PW, Lawrie TDV. Comprehensive electrocardiology.

Theory and practice in health and disease. New York7 Pergamon Press,

Inc; 1989.

[7] O’Keefe Jr JH, Hammill SC, Freed M. The complete guide to ECGs. A

comprehensive study guide to improve ECG interpretation skills.

Birminham (Mich)7 Physicians’ Press; 1997.

[8] Kassirer JP. Teaching clinical medicine by iterative hypothesis testing.

Let’s preach what we practice. N Engl J Med 1983;309:921.