Chest pain and computer ‘normal’ ECG. Wait for troponin? And what is the reference standard for ECG diagnosis? Cardiologist or outcome?

Written by Jesse McLaren

A 50 year old presented to triage with one hour of chest pain, and the following ECG labeled ‘normal’ by the computer (GE Marquette SL) algorithm. 

 

What do you think?

There’s normal sinus rhythm, first degree AV block, early R wave, normal voltages. There’s hyperacute T waves in I/aVL with reciprocal inferior ST depression, diagnostic of high lateral occlusion MI.

Here’s the Queen of Hearts interpretation (OMI with high confidence):

But many small studies have concluded (erroneously) that triage ECGs labeled ‘normal’ by computer algorithm don’t need to be interrupted by a physician.(see reference below)

 

What’s the gold standard for ECG interpretation: is it cardiologist interpretation?

The latest is Langlois-Carbonneau et al. Safety and accuracy of the computer interpretation of normal ECGs at triage. CJEM 2025. They concluded, "Our findings increase confidence in the normal automated GE Marquette 12 SL ECG software interpretation to predict a benign outcome. Our data corroborate that immediate management of a patient with a normal automated triage ECG reading is not modified by real-time ED physician ECG interpretation."

But like many similar studies, the study was small (one year at one centre with no indication of the incidence of acute coronary occlusion), and it used as the gold standard the final cardiologist interpretation of the ECG - not the patient outcome! 

(Smith comment: this is a very stupid outcome measure)

What if we use that methodology for the ECG above?

The final blinded cardiologist interpretation was only non-specific “ST and T wave abnormality”. If we took this as the gold standard, we would conclude that the computer interpretation was safe and accurate – at least accurate enough to not miss “STEMI”, and that physicians should not be interrupted to interpret it, because there would be no change in patient management.

If this advice had been followed, then the patient would have remained in the waiting room, waiting to be seen by a provider. The first troponin was normal (6ng/L, just barely above the level of detection of 2ng/L, and below the upper limit of normal of 26ng/L). But troponin is a rear-view mirror which shows damage that has already occurred, and is often within the normal range within only 2 hours of onset of acute coronary occlusion. So waiting for serial troponin would have further delayed reperfusion.

Smith comment: we showed that the first troponin, even in full-blown STEMI, is negative 25% of the time.

Fortunately, the emergency physician was immediately shown the ECG and immediately identified high lateral occlusion. So they modified patient management in real-time despite the false reassurance of the 'normal' ECG. A 15 lead was done, which was similar and had the same final blinded cardiology interpretation. Despite serial ECGs being "STEMI negative", the cath lab was activated.

What is the gold standard for ECG interpretation: patient outcome!!!

On angiogram there was a 99% first diagonal occlusion with TIMI 1 flow. Repeat troponin was 4,000ng/L and not repeated afterwards, and there was no follow up ECG. Because of rapid cath lab activation, the discharge diagnosis was “STEMI” even though no ECG met STEMI criteria by cardiology interpretation.

Smith comment: we have a paper under consideration that shows that the only independently significant factor in final diagnosis of STEMI vs. Non-STEMI is time to treatment, not presence of STE millimeter criteria, nor presence of occlusion!

But according to Langlois-Carbonneau et al., even a final interpretation of STEMI does not change their conclusion that computer ‘normal’ ECGs are safe and accurate. 

Case 2 in this same paper

They included the following case (different from the one above!):

A 68 year old woman presented to the ED with a history of "atypical" chest pain (Smith: "atypical" is a garbage description of chest pain). Here is her ECG:

What do you think?

Smith: this is diagnostic of acute inferior posterior lateral OMI

Here is the description in the paper:

“The 68 years old female with a final STEMI diagnosis presented to ED with a history of atypical chest pain. The ED physician’s ECG interpretation of her first ECG was “ST elevation in II-aVF and V5-V6,(…)concave, no reciprocal changes.” When troponins came back positive, the emergent catheterization procedure was activated. The complete cardiology consult was done the next day and the initial ECG was analyzed as ‘slight ST elevation in II-aVF-V4-V5-V6.’ The patient’s final diagnosis remained STEMI but this is contentious as the patient did not actually fulfill STEMI criteria on the initial ECG. Moreover, this patient’s ECG was read as “normal” by the blinded attending cardiologist.”

So even in hindsight, these authors concluded that the computer interpretation on the ECG paper is more reliable for the final diagnosis than an actual acutely occluded coronary artery!

Smith comment: this is even more stupid.

Here's the PMCardio AI OMI Model Queen of Hearts interpretation:

This confirms that computer ‘normal’ ECGs are neither safe nor accurate, that physicians can identify subtle OMI in real time (enhanced by AI), and that the gold standard should be patient outcome. 

 

Take home

1.     Computer ‘normal’ ECGs are not safe or accurate. See > 50 cases on Dr. Smith’s EGC Blog. We’ve also published the largest study on this question: Emergency Department Code STEMI patients with initial electrocardiogram labeled ‘normal’ by computer interpretation: a 7-year retrospective review. Using patient outcome (Code STEMI with culprit lesion) we found 4% had had a triage ECG labeled ‘normal’ by the computer – a third of which were identified in real time by the emergency physician. This underestimates computer false ‘normal’ ECGs because we only included code STEMIs, not Non-STEMIs with OMI.
2.     The gold standard should be patient outcome. See our letter to the editor in response: McLaren, Meyers, Smith, Chartier. What is the gold standard for ECG interpretation: computer, cardiologist, or patient outcome. CJEM 2025 March 10
3.      The initial troponin is not reliable in OMI, and can be in the normal range with acute  symptoms

References concluding erroneously that if the conventional algorithm says "normal" that the triage ECG does not need to be reviewed by the physician.

1. Winter LJ, Dhillon RK, Pannu GK, Terrazza P, Holmes JF, Bing ML. Emergent cardiac outcomes in patients with normal electrocardiograms in the emergency department. Am J Emerg Med. 2022; 51: 384-387.

2. Villarroel NA, Houghton CJ, Mader SC, Poronsky KE, Deutsch AL, Mader TJ. A prospective analysis of time to screen protocol ECGs in adult emergency department triage patients. Am J Emerg Med. 2021; 46: 23-26.

3. Hughes KE, Lewis SM, Katz L, Jones J. Safety of computer interpretation of normal triage electrocardiograms. Acad Emerg Med [Internet] 2017;24(1):120–4. Available from: https://pubmed.ncbi.nlm.nih.gov/27519772/ 

4. Deutsch A, Poroksy K, Westafer L, Visintainer P, Mader T. Validity of computer-interpreted “normal” and “otherwise normal” ECG in emergency department triage patients. West J Emerg Med [Internet] 2024 [cited 2024 Aug 26];25(1):3–8. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10777178/ 

5. Langlois-Carbonneau V, Dufresne F, Labbé È, Hamelin K, Berbiche D, Gosselin S. Safety and accuracy of the computer interpretation of normal ECGs at triage. CJEM [Internet] 2024;26(12):857–64. Available from: https://link.springer.com/article/10.1007/s43678-024-00790-5

===================================

MY Comment, by KEN GRAUER, MD (3/12/2025):

===================================

Today’s case by Dr. McLaren once again raises the question of, “What is a ‘normal’ ECG?” (Please see My Comment at the bottom of the page in the January 15, 2025 post for "My Take" on what constitutes a "normal" ECG). 

Regular readers of Dr. Smith’s ECG Blog are well familiar with the following facts:

• A “normal” ECG is not an ECG that is called “normal” by any computerized ECG interpretation program other than QOH (Queen Of Hearts). 
• A “normal” ECG is not an ECG that is called “normal” by a cardiologist who does not yet accept and believe in the OMI Manifesto (See the July 31, 2020 post — and — McLaren, Meyers & Smith: J Electrocardiol 76:39-44, 2023 — and — Meyers, Weingart, Smith: The OMI Manifesto, 2018).
• As per Dr. McLaren — a normal initial hs-Troponin in no way rules out the possibility of an acute OMI. To quote Dr. Smith, "Using Troponin to diagnose acute coronary occlusion is like relying on a rear-view mirror to navigate a car pile-up: It shows wreckage behind you that has already happened, but can not see the road ahead." As a result — "A normal initial Troponin can give false reassurance, when there's a head-on collision happening in real time!" (from the March 24, 2023 post).
• Up to 1/3 of patients with an "NSTEMI" turn out to have an acute coronary occlusion that could benefit from PCI — and — More than 1/2 of patients with acute coronary occlusion do not manifest ST elevation (Statistics from McLaren, Meyers, Smith et al: JACC Adv 3:101314, 2024).
• BOTTOM Line: Correlating the history with the relative severity of CP associated with each ECG (and with serial Troponins) — is far more accurate for rapid recognition of acute coronary occlusion than waiting until “enough” ST elevation develops to qualify as a STEMI (and all-too-often with acute OMIs — this "required" amount of ST elevation never occurs).

The Computer Can Not be Trusted:

Returning to today's case — I have labeled in Figure-1 the initial ECG from this 50-year old patient who presented to the ED for new-onset CP (Chest Pain) of 1 hour duration.

• Amazingly — the GE Marquette SL computer program labeled the ECG in Figure-1 as "normal".

As per Dr. McLaren — ECG #1 is diagnostic of acute coronary occlusion. In this patient with new CP — failure to at least note the obvious ST-T wave abnormalities in 5/6 of the limb leads serves to immediately discredit any validity that this ECG computer algorithm might have.

• Considering the modest QRS amplitude in each of the limb leads — the amount of ST elevation in high-lateral leads I and aVL is significant. More than this — the T waves in these 2 leads are clearly hyperacute (ie, much "fatter"-at-their-peak and wider-at-their-base than they should be given how small the R waves in these leads are).
• Equally impressive reciprocal ST-T wave changes in each of the inferior leads instantly confirms acute coronary occlusion until proven otherwise (ie, leads II,III,aVF showing downsloping ST depression with widening of the lowest ST segment point and terminal T wave positivity).

Although ST-T waves in the chest leads are not as acute-appearing — there are diffuse abnormalities in these chest leads that made me wonder about posterior OMI or multi-vessel disease:

• The already tall R wave in lead V2 is a common finding with posterior OMI.
• The gently upsloping slight ST elevation that is normally seen in leads V2,V3 is absent.
• ST segments are uncharacteristically flattened in leads V3,V4,V5.
• Especially considering the hyperacute T waves with ST elevation that we saw in lateral limb leads I and aVL — I thought the broadened T wave peak in lead V6 (if not also in lead V5) represented hyperacute lateral chest leads.

To Emphasize: Computerized ECG reports other than QOH (Queen Of Hearts) simply can not be trusted.

• The above said — the fault does not lie with these computer programs. These programs simply have never been programmed with sufficient outcome data to enable accurate ECG assessment in patients with CP. We need to know this!
• Fortunately — the ECG in today's case was promptly shown to the ED physician who correctly identified the acute abnormalities.

Figure-1: I've labeled the initial ECG in today's case. (To improve visualization — I've digitized the original ECG using PMcardio).