A Bayesian approach to the differential diagnosis of wide complex tachycardia

We use sensitivity, specificity, and likelihood ratios (LR) to interpret lab tests all the time. This is known as Bayesian analysis. It is an essential part of the clinical reasoning we do everyday. We  don't ordinarily apply Bayesian analysis to ECG interpretation but the concept is valid. A review published in 2000 outlines the Bayesian approach to the diagnosis of wide complex tachycardia, linked here:

The Bayesian approach improves the electrocardiographic diagnosis of broad complex tachycardia

Like any other uses of the Bayesian approach, one starts with the pretest probability which can be converted into prior odds. We have data on the likelihood ratios of various ECG findings in wide complex tachycardia. To apply a Bayesian analysis one must first determine the pretest probability (or prior odds). The authors posit prior odds of 4 (80% pretest probability) as appropriate for most cases, based on data from a study of unselected patients. They acknowledge that different prior odds may be applied based on clinical judgment and published data in selected patient groups. For example, in an adolescent with WCT a supraventricular mechanism is more likely with a probability of VT of around 40% in published studies, equated to prior odds of 0.67.

Table 2 from the paper, showing various  ECG findings and their likelihood ratios, is shown here.

Pretest odds are multiplied by the LRs of various findings sequentially. The final product is the post test odds. Post test odds of greater than or equal to 1 lean toward  VT. Odds of less than 1 point to a supraventricular mechanism. The further the number is from 1 the greater the strength of the diagnosis. 

When should a Bayesian approach be used as opposed to a conventional approach using algorithms and scoring systems? In WCT with bizarre QRS morphology measurements may be difficult causing problems in application to algorithms. A Bayesian approach may be advantageous in those situations. Moreover, for teaching and learning purposes, Bayesian  analysis provides a good exercise in clinical reasoning as applied to electrocardiography.

VT or not VT?

A fast heart rate with a wide QRS complex (greater than 120 ms) can be scary. Here’s the crucial question to consider when you encounter wide complex tachycardia (WCT):   Is it VT or not VT?  The distinction is clinically important.

If it's not VT it is usually a supraventricular mechanism such as atrial fibrillation, atrial flutter or SVT with aberrant conduction (bundle branches not working properly at tachycardic rates). So it's mainly a binary differential diagnosis between VT and supraventricular wide complex tachycardia (SVWCT). However, a couple of very infrequent exceptions should be noted for which the diagnostic rules to be described herein may not apply:  Some WCTs in Wolff Parkinson White syndrome and suraventricular mechanisms with severe metabolic disturbances, eg TCA overdose. Otherwise for most purposes, in the binary differential diagnosis between VT and SVWCT certain rules are helpful. Through the years many proposed rules and algorithms have been put forth. This has led to some confusion. However, finally they have all been compiled in one place in the review linked below: 

Wide Complex Tachycardia Differentiation: A Reappraisal of the State‐of‐the‐Art

Here is a graphic from the review that highlights some of the key findings, any of which, if present, favor VT. Most of these findings have high specificity but low sensitivity. 

Figure 2 from the review summarizes the various algorithms and other approaches:

Some key points to be gleaned from figures one and two are summarized here:

AV dissociation

This finding has high specificity but low sensitivity. This is when P waves can be seen independently of the QRS complexes. If this is seen it is strong evidence for VT. There are some caveats. First, many cases of VT, (in some series up to half ) exhibit retrograde conduction to the atria (in other words, AV association ). Whan AV dissociation is present, particularly at faster VT rates, it can be difficult to see because of interference from the wide QRS complexes and the ST segments and T waves. 

Morphologic criteria

Does the tachycardia have a typical bundle branch block appearance? If not, it is more likely VT. This test is difficult to apply unless one knows the ins and outs of bundle branch block morphology. (The details of BBB morphology are covered in the archived lecture series).

QRS duration

If the QRS is a greater than 160 milliseconds it favors VT. (If the QRS in V1 is upright, greater than 140 milliseconds favors VT). In general, the wider the QRS the more it favors VT. 

Chest lead concordance 

In the chest leads, (V1-6) if the QRS complexes are all upright (monolithic R waves) or are all pointing down (QS complexes) VT is the diagnosis.

Frontal plane (limb lead) axis

Right upper quadrant axis ( -90  to -180) strongly favors VT. Left upper quadrant and left lower quadrant axis (normal ) less so.

Ventricular activation velocity

If the rate of voltage change of the first 40 milliseconds is less than or equal to that of the last 40 milliseconds of the QRS complex, VT is favored. In addition, if the time from the beginning of an R wave to the nadir of an S wave greater than 60 milliseconds it favors VT. (In the Brugada algorithm that cut off is 100 milliseconds). 

Algorithms

Several algorithms have been proposed and have good test characteristics. In the paper the Brugada algorithm, the Vereckei algorithm, the limb lead algorithm, the Griffith algorithm, and the RWPT algorithm are described.

There are also point scoring systems that perform well. These are discussed in the paper. Another, someone novel method, is the Bayesian approach which will be discussed in a separate post.

You suspect your patient’s chest pain is cardiac. Should you activate the cath lab?

Below is an algorithm that has been discussed on wards several times. It incorporates the new acute coronary syndrome guidelines with a modification based on the OMI (occlusive MI) model.

The new guidelines, which have not adopted the OMI model, are liked here:

2025 ACC/AHA/ACEP/NAEMSP/SCAI Guideline for the Management of Patients With Acute Coronary Syndromes: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines

NSTEMI, the catch-all diagnosis

The diagnostic label NSTEMI (non ST segment elevation MI) tends to be inappropriately applied to many cases of acute troponin elevation in the absence of electrocardiographic ST segment elevation. The result has been that NSTEMI has become a catch all diagnosis, to the point of being devoid of meaning.  Actually, troponin elevations without ST segment elevation have a differential diagnosis including NSTEMI, type 2 MI, takotsubo cardiomyopathy,  PE, and acute nonischemic myocardial injury. This last one is frequently seen in sepsis and other non cardiac critical illnesses. Labeling such instances as NSTEMI is problematic because it can lead to inappropriate and potentially harmful treatments such as systemic anticoagulation.

This problem has been addressed in the literature. The paper linked below from JAMA is somewhat dated but useful.

Increasingly Sensitive Assays for Cardiac Troponins

Figure 1  from the paper is a proposed algorithm.

A good general resource, one with which all internists should be familiar, is the following,

Fourth Universal Definition of Myocardial Infarction (2018)

The JAMA article was published before the most recent addition of the universal definition. It contains references to the third universal definition, now out of date. I should also point out that although not explicitly addressed in the fourth universal definition one should not conflate type 2 MI with NSTEMI, as the latter term implies ACS. Thus terms like “ type 2 NSTEMI”  should be discouraged.

Electrocardiographic patterns of acute coronary occlusion that do not meet STEMI criteria

How can you identify patients who need an immediate trip to the cath lab if they do not meet STEMI criteria? 

We're all familiar with the guideline based metric for door to balloon time of 90 minutes or less for ACS patients who meet criteria for ST segment elevation MI. STEMI is a surrogate for acute epicardial coronary occlusion and therefore the need for immediate reperfusion therapy. In this paradigm, ACS cases that do not meet STEMI criteria are designated as non-ST segment elevation myocardial infarction (NSTEMI). Less well appreciated, however, is that up to 30% of ACS cases not meeting STEMI criteria actually have acute coronary occlusion. Because these occlusions tend to go unrecognized and are not under the CMS “quality” metric of door to balloon time, these are patients with acute coronary occlusions who are often deprived of timely reperfusion. Multiple studies indicate that such patients have a higher mortality even compared to STEMI patients who do get timely reperfusion. Most of these patients can be recognized electrocardiographically as they generally fall into several patterns listed below. However, these can be subtle and require a certain degree of electrocardiographic interpretation skill.

Hyperacute T waves:  these are upright T waves of high amplitude and increased width such that they have an increased area under the curve. They may be the earliest signs of coronary occlusion and may be present without ST segment elevation

DeWinter T waves: these are hyperacute T waves preceded by J point ST segment depression. This pattern is diagnostic of acute LAD occlusion although not meeting STEMI criteria.

Wellens syndrome:  this is a pattern of anterior precordial T wave inversion that occurs after resolution of chest pain. It indicates a proximal LAD occlusion that has recently reperfused. Such patients have a very high incidence of catastrophic anterior STEMI within the coming two weeks and do not respond to medical management.

Precordial swirl pattern:. this is a pattern of ST elevation in V1 and V2 combined with lateral precordial ST segment depression. The ST segment elevation may be subtle and not meet STEMI criteria. This is indicative of proximal LAD occlusion.

South African flag sign: this is a combination of ST elevation in Leeds 1 and AVL and depression in lead 3. There is concomitant ST segment elevation in V2. The pattern is indicative of acute diagonal branch occlusion.

Aslanger pattern:. this is a variant of inferior MI but the ST segment elevation is in lead 3 only. It is accompanied by ST segment depression in any of V4-6. There is also subtle ST elevation in  V1 greater than V2, not enough to meet STEMI criteria. This is indicative of inferior MI, usually circumflex occlusion but occasionally RCA occlusion. It is seen in patients with multivessel disease and is indicative of a large infant.

Terminal QRS distortion:. this is generally seen in LAD occlusions and is manifested by disappearance of the S wave in the anterior precordial leads. It is often accompanied by hyperacute T waves and occasionally may occur before ST segment elevation is seen and therefore not meet STEMI criteria.

Subtle anterior ST segment elevation not meeting STEMI criteria: this pattern can be confused with either normal variant anterior ST elevation, particularly in males, or benign early repolarization. There are a couple of formulas that involve the QT interval, the R&S wave amplitude and the degree of ST elevation in V3 which have good test characteristics for differentiation between early anterior STEMI and these normal variants.

Posterior infarction: because we generally do not employ posterior leads this is a mirror image of posterior current of injury and is seen as ST segment depression in V1-4. It does not meet STEMI criteria and is often missed.

Sgarbossa criteria in left bundle branch block and right ventricular pacing: this consists of concordant ST segment changes. Remember that in left bundle branch block and RV pacing ST segment changes should normally be discordant from the major portion of the QRS. 

Northern OMI pattern: this pattern consists of ST segment elevation in aVL and aVR  accompanied by by T wave inversion in those same leads, and ST segment depression in other leads. It is often indicative of occlusion at the takeoff of the diagonal branch.  It does not meet STEMI criteria.

Right bundle branch block, with or without left anterior fascicle block, not known to be old:  in the context of suspected acute myocardial ischemia, this finding indicates proximal LAD occlusion and a very large infarct with high mortality and high risk of acute complications including pulmonary edema, cardiogenic shock, septal rupture and complete heart block with asystole.  The downward secondary ST segment forces in the anterior leads (due to the RBBB) may obscure the ST elevation and thus fail to meet STEMI criteria.  Moreover, the RBBB itself may be a distractor from what is really going on.

Inferior infarction with only minimal STE not meeting STEMI criteria with reciprocal aVL depression as the principal abnormality.

These patterns are detailed in the paper linked below:

ECG Patterns of Occlusion Myocardial Infarction: A Narrative Review

Tables and Graphics from the paper are displayed here.

The STEMI/NSTEMI classification: great tool for coders, horrible tool for clinicians

Several times recently on wards I have mentioned a research study under review presenting data that in patients with acute coronary syndrome, the final chart diagnosis of STEMI versus NSTEMI correlated more closely with whether they made it to the cath lab in time to satisfy the applicable metric (120 minutes door to balloon time) than it did the ECG findings. The paper was finally published and I am presenting it here:

Door-to-Balloon Time Outperforms ST-Segment Elevation in Predicting the STEMI vs. NSTEMI Final Diagnosis

This is an observational study drawn from a large database of patients who underwent coronary angiography in an acute setting. 410 patients found to have acute coronary syndrome met the eligibility criteria for the analysis.  The findings, from the abstract of the paper:

Results: Among 410 angiographed AMI patients (mean age 63 ± 13; 71% male), 165 (40.2%) received an FDx-STEMI and 245 (59.8%) an FDx-NSTEMI. D2B time showed 94% agreement with FDx (160/165 FDx-STEMI treated less than 120 min; 225/245 FDx-NSTEMI treated greater than  120 min), exceeding concordance for STE (82%; p less than 0.001) and TIMI 0-1 flow (75%; p less than 0.001). FDx and STE diverged in 75 patients (18%): 60 rapidly treated STE-negative cases were labelled STEMI, whereas 15 delayed STE-positive cases were labelled NSTEMI. In regression analysis, D2B less than 120 min remained the sole independent predictor of discordance (adjusted OR 6.7, 95% CI 3.5–13.8). Conclusions: In this registry, the cath-lab label “STEMI” showed the strongest correlation with meeting a 120 min benchmark, exceeding correlations for STE or angiographic occlusion. These findings suggest that quality-metric compliance, rather than electrocardiographic or anatomic criteria, predominantly drives final diagnosis.

Here is a graphic that summarizes the findings:

Or, better yet, this cartoon (not from the paper):

The authors elaborate in the discussion section:

Our multivariable analysis asked a specific question: what factors explain the 18% of encounters in which the cath-lab label contradicts the patient’s ECG? Overall, door-to-balloon time less than 120 min was far more concordant with the final cath-lab diagnosis than either guideline ST-segment elevation or angiographic TIMI 0-1 occlusion (94% vs. 82% vs. 74%). This observation answers our prespecified question and underscores that, in routine practice, the label applied in the cath report aligns most closely with the treatment timeline recorded in the electronic chart. This suggests that the designation of STEMI is not always a reflection of the original diagnostic framework, but rather a retrospective label influenced by procedural outcomes and time benchmarks.

The authors believe the findings are generalizable, reflecting trends across the U.S.

They conclude, at the end of the paper:

In this retrospective analysis, time to treatment showed the strongest association. in final diagnosis of STEMI vs. NSTEMI, more important than both STEMI millimeter “criteria” and the presence or absence of total coronary occlusion. This has implications for research and quality improvement.

By the way, the 2025 guidelines for acute coronary syndrome are now published. (They’re fairly new, although they've been online since February of this year). You can access them here:

2025 ACC/AHA/ACEP/NAEMSP/SCAI Guideline for the Management of Patients With Acute Coronary Syndromes

There's a growing and now overwhelming body of literature calling for a replacement of the STEMI/NSTEMI paradigm. Despite that, the new guidelines, unfortunately, stick with the existing classification scheme and do not mention the emerging OMI/NOMI concept. There is passing mention in the body of the guideline of a couple of electrocardiographic patterns that are diagnostic of acute coronary occlusion but do not meet STEMI criteria (DeWinter pattern and hyperacute T waves). Because the classification remains in current guidelines this will unfortunately drive MKSAP and ABIM board exam questions for some time to come. So you need to know them while recognizing their deficiencies. In this respect the guideline is 10 years or more out of date. In multiple other content areas there is a world of good information and it is worth the read. 

Why this inertia? One big reason is that the STEMI/NSTEMI paradigm is so entrenched in the administrative and coding world. In addition, a reasonably intelligent junior high student can be trained to use a ruler or count a little boxes and measure the degree and direction of ST segment displacement. The new paradigm requires actual skill in electricardiography. I think the cartoon below illustrates the problem well: