ECG Analysis: Where do we go wrong and how can we improve?

Author: Romeo Fairley MD (EM Attending Physician/Disaster Medicine Fellow, UC Irvine) // Edited by: Alex Koyfman MD (EM Attending Physician, UT Southwestern Medical Center / Parkland Memorial Hospital, @EMHighAK) and Stephen Alerhand MD (@SAlerhand)

Electrocardiograms (ECGs) are among the most commonly ordered tests in the Emergency Department (ED). Emergency physicians will read thousands of ECGs over the course of their career and must be experts in their rapid interpretation

A recent analysis showed that ECG analysis and interpretation changed the ED management of a patient in 32% of cases [1]. Unlike many lab tests, there are few well-defined criteria to designate a yes/no answer for ECG interpretation. It is common for cardiologists, electrophysiologists, and emergency physicians to have poor rates of inter-rater reliability (kappa) when viewing the same ECGs. Even the same individual reading the same ECG at a different time will yield inconsistent results [2, 3]. The American College of Cardiology Foundation states it takes 3,500 supervised ECG reads to become an expert [4]. It thus comes as no surprise that mistakes are common in ECG interpretation. Studies rate non-cardiologist physicians an accuracy of 36% to 96% in detecting ECG abnormalities, and only 87% to 100% successful at detecting acute myocardial infarction [5]. The consequences of these misinterpretations can be significant. It is estimated that these errors cause mismanagement of patients in up to 11% of cases, though the severity of these consequences are variable [5].


This article will not be a review of complex or subtle ECG analysis. Instead, this article will evaluate the systematic errors of ECG analysis in an attempt to limit major sources of misinterpretation. In order to analyze the systematic errors involved, it is important to understand the when, why, and how of the ECG process.

A typical scenario is as follows: an ECG is ordered due to nursing staff protocol (the most common reason an ECG is ordered [1]). The ECG is then shown to a physician who has not yet seen the patient. The physician must stop their current task and rapidly analyze the ECG without a clinical understanding of the patient. Generally the question asked is: “Is this a STEMI?” Most modern ECG machines will give a computer analysis of the ECG to assist the physician. Then at some future time, the physician will see the patient and make further treatment decisions.


Each of the steps in the ECG process can represent the source of a systematic error; these will be discussed in detail below. Resources for further clarification and methods of improvement are referenced where applicable.

Emergency Department physicians are interrupted (i.e. handed an ECG) on average every 9 to 14 minutes [6]. There is considerable debate on if humans can truly multi-task or if it is actually task switching [7]. There are many models being developed to teach people how to cope with interruptions and multi-tasking [6, 8]. During multi-tasking (or task switching), people are more likely to use heuristics to simplify cognitive loads [8].

A heuristic is a mental shortcut. It is a strategy the brain uses to limit cognitive options and therefore quickly reach a viable solution. The use of heuristics allows the brain to function at a high rate; however, they also lead to certain biases (systematic errors due to cognitive processes) [9, 10]. There are many strategies available to counteract the negative impact of heuristics (an in-depth analysis is outside the scope of this article). In one study of ECG analysis, simply being aware of the presence and critically thinking about the impact of heuristics can lead to improvements in cognition and a reduction in errors [11].

“Is this a STEMI?” This question is the source of a systematic error. The physician is primed to search for a STEMI. Once the signs of a STEMI are found or not found on the ECG, it is easy to stop searching for abnormal findings. This is called premature closure, and can lead to the omission of other significant ECG abnormalities [12]. To avoid premature closure, continue to read the ECG for all abnormalities and do not stop once signs of a STEMI have been excluded. The use of checklists also has long been known to reduce cognitive errors. Several recent studies have shown a significant improvement in the accuracy of ECG interpretation by using checklists, irrespective of physician expertise. The only downside noted is an increase in time of interpretation [13, 14].

The next issue deals with the clinical scenario of the ECG process. When an ECG is handed to a physician, it frequently comes without any corresponding clinical history on the patient. Clinical history has been shown to improve physicians’ ability to accurately interpret an ECG [15]. Physicians must be careful, however, as incorrect clinical histories have been shown to negatively impact ECG interpretation [16, 17]. These findings are likely due to the availability heuristic, which creates a bias toward findings that are more easily retrieved by the mind [18]. One method to improve on this source of error is to instruct ECG couriers (whether it is a nurse, technician, student, or resident) to give a brief and accurate history when delivering ECGs. Additionally, all ECGs should be re-evaluated once the physician has obtained a clinical history and physical exam from the patient.

Obtaining previous ECGs is equally as important as clinical history. A standard ECG is a 10 second picture of cardiac electrophysiology. Human disease is a dynamic process. It is important to evaluate prior ECGs in comparison to the current ECG. Knowledge of chronic abnormalities can significantly alter management [19]. Furthermore, it is critical to not rely on a single ECG during a patient’s visit. ECGs should be obtained at 15 to 30 minute intervals for the first hour of an at-risk patient’s visit according to the most recent AHA guidelines [20].

Most modern ECG machines will give a full computer generated analysis of the ECG. It can be tempting to use this analysis as a crutch, especially as a learner. Accurate computer analysis has been shown to improve physician ECG interpretation, as well as increase the speed of interpretation [21]. However, computer automated analysis has been shown to be inaccurate 6% to 42% of the time [5]. Moreover, inaccurate computer analysis has been shown to negatively impact a physician’s ability to correctly interpret an ECG [22]. These errors are likely due in part to the congruence heuristic, in which a tendency exists to search for answers that agree with a belief and ignore alternative possibilities [23]. A prudent strategy would use the benefits of computer analysis and mitigate the harms. One suggestion is to interpret the ECG first, then use the computer analysis as a check on accuracy. Moreover, it is important to always consider alternative sources of an ECG abnormality (e.g. ST depression could be due to acute coronary syndrome or intracerebral hemorrhage, among other causes [24]).


The educational design of medicine is going through a rapid change with the advent of computer and internet based learning. The best way to teach ECG interpretation is still under debate. Many individuals enjoy the self-directed models; however, these may be less effective for long-term learning [25]. A caution should be made about online learning and ‘free open access medical’ (FOAM) education – be skeptical and check sources. One recent review of YouTube ECG interpretation videos found 13% to be misleading [26]. Reviewing 3,500 ECGs and memorizing the formulas for QTc, LVH, and the differentiation of early repolarization from ischemia [27] will make you a considerably more proficient ECG reader. But, these will not make you error proof. Understanding the heuristics we use and systematic errors they cause will greatly reduce the chance of ECG misinterpretation.

The author would like to thank Marlene Zacharia and Carl Schultz for their edits and contributions.


References / Further Reading

  1. Benner, J.P., et al., Impact of the 12-lead electrocardiogram on ED evaluation and management. Am J Emerg Med, 2007. 25(8): p. 942-8.
  2. Holmvang, L., et al., Differences between local investigator and core laboratory interpretation of the admission electrocardiogram in patients with unstable angina pectoris or non-Q-wave myocardial infarction (a Thrombin Inhibition in Myocardial Ischemia [TRIM] substudy). Am J Cardiol, 1998. 82(1): p. 54-60.
  3. Massel, D., J.A. Dawdy, and L.J. Melendez, Strict reliance on a computer algorithm or measurable ST segment criteria may lead to errors in thrombolytic therapy eligibility. Am Heart J, 2000. 140(2): p. 221-6.
  4. Myerburg, R.J., et al., Task force 2: training in electrocardiography, ambulatory electrocardiography, and exercise testing. J Am Coll Cardiol, 2008. 51(3): p. 348-54.
  5. Salerno, S.M., P.C. Alguire, and H.S. Waxman, Competency in interpretation of 12-lead electrocardiograms: a summary and appraisal of published evidence. Ann Intern Med, 2003. 138(9): p. 751-60.
  6. Heng, K.W., Teaching and evaluating multitasking ability in emergency medicine residents – what is the best practice? Int J Emerg Med, 2014. 7: p. 41.
  7. Fischer, R. and F. Plessow, Efficient multitasking: parallel versus serial processing of multiple tasks. Front Psychol, 2015. 6: p. 1366.
  8. Li, S.Y., F. Magrabi, and E. Coiera, A systematic review of the psychological literature on interruption and its patient safety implications. J Am Med Inform Assoc, 2012. 19(1): p. 6-12.
  9. Kahneman, D., Thinking, fast and slow. 1st pbk. ed. 2013, New York: Farrar, Straus and Giroux. 499 p.
  10. Lewis, A., The Cambridge handbook of psychology and economic behaviour. 2008, Cambridge University Press.
  11. Eva, K.W., et al., Teaching from the clinical reasoning literature: combined reasoning strategies help novice diagnosticians overcome misleading information. Med Educ, 2007. 41(12): p. 1152-8.
  12. Kumar, B., B. Kanna, and S. Kumar, The pitfalls of premature closure: clinical decision-making in a case of aortic dissection. BMJ case reports, 2011. 2011: p. bcr0820114594.
  13. Sibbald, M., A.B. de Bruin, and J.J. van Merrienboer, Checklists improve experts’ diagnostic decisions. Med Educ, 2013. 47(3): p. 301-8.
  14. Sibbald, M., et al., Why verifying diagnostic decisions with a checklist can help: insights from eye tracking. Adv Health Sci Educ Theory Pract, 2015. 20(4): p. 1053-60.
  15. Anh, D., S. Krishnan, and F. Bogun, Accuracy of electrocardiogram interpretation by cardiologists in the setting of incorrect computer analysis. J Electrocardiol, 2006. 39(3): p. 343-5.
  16. Cruz, M.F., et al., The effect of clinical history on accuracy of electrocardiograph interpretation among doctors working in emergency departments. Med J Aust, 2012. 197(3): p. 161-5.
  17. Hatala, R., G.R. Norman, and L.R. Brooks, Impact of a clinical scenario on accuracy of electrocardiogram interpretation. J Gen Intern Med, 1999. 14(2): p. 126-9.
  18. Tversky, A. and D. Kahneman, Judgment under Uncertainty: Heuristics and Biases. Science, 1974. 185(4157): p. 1124-31.
  19. Rubenstein, L.Z. and S. Greenfield, The baseline ECG in the evaluation of acute cardiac complaints. Jama, 1980. 244(22): p. 2536-9.
  20. DR Jr, J.A., et al., 2014 AHA/ACC Guideline for the Management of Patients With Non–ST-Elevation Acute Coronary Syndromes: Executive Summary. 2014.
  21. Hillson, S.D., D.P. Connelly, and Y. Liu, The effects of computer-assisted electrocardiographic interpretation on physicians’ diagnostic decisions. Med Decis Making, 1995. 15(2): p. 107-12.
  22. Tsai, T.L., D.B. Fridsma, and G. Gatti, Computer decision support as a source of interpretation error: the case of electrocardiograms. J Am Med Inform Assoc, 2003. 10(5): p. 478-83.
  23. Baron, J., J. Beattie, and J.C. Hershey, Heuristics and biases in diagnostic reasoning. Organizational Behavior and Human Decision Processes, 1988. 42(1): p. 88-110.
  24. Takeuchi, S., et al., Electrocardiograph abnormalities in intracerebral hemorrhage. J Clin Neurosci, 2015.
  25. Fent, G., J. Gosai, and M. Purva, Teaching the interpretation of electrocardiograms: which method is best? J Electrocardiol, 2015. 48(2): p. 190-3.
  26. Akgun, T., et al., Learning electrocardiogram on YouTube: how useful is it? J Electrocardiol, 2014. 47(1): p. 113-7.
  27. Smith, S.W., et al., Electrocardiographic Differentiation of Early Repolarization From Subtle Anterior ST-Segment Elevation Myocardial Infarction. Annals of Emergency Medicine. 60(1): p. 45-56.e2.

2 thoughts on “ECG Analysis: Where do we go wrong and how can we improve?”

Leave a Reply

Your email address will not be published. Required fields are marked *