Study objective: To determine the extent of interobserver agreement in the ECG diagnosis of ventricular tachycardia (VT) by using a four-step algorithm and three observers. Methods: Simulated emergency department setting from records of an urban university teaching hospital. All ECGs taken in the ED during a 2-year period that showed a QRS duration of more than 120 msec and a heart rate faster than 110 beats per minute were reviewed. ECGs were categorized as demonstrating sinus rhythm (SR), irregular broad-complex tachycardia (I-BCT), or regular broad-complex tachycardia (BCT). Copies of the BCT ECGs and short clinical histories were given to each of three emergency physicians, who used a published, four-step algorithm (the Brugada algorithm) to categorize the BCT ECGs as indicating VT, indicating supraventricular tachycardia with aberrancy (SVT-A), or indeterminate. Interobserver agreement was assessed with the κ-statistic. Results: The records contained 178 ECGs, 88 of which were SR, 63 I-BCT, and 27 BCT. The 27 BCT ECGs were selected for review. The emergency physicians disagreed with each other 22% of the time in differentiating VT from SVT-A (κ=.58). Conclusion: Application of the algorithm to actual clinical practice in the ED would probably result in the misdiagnosis of a substantial minority of patients having BCT, with potentially serious adverse consequences.