J Accid Emerg Med: first published as 10.1136/emj.11.3.149 on 1 September 1994. Downloaded from Journal of Accident and Analysis of interpretation of Emergency 1994 electrocardiograms 11, 149-153 E.R.SNOEY,1'3 B. HOUSSET,1 P.GUYON,2 S. ELHADDAD,2 J.VALTY2 & P. HERICORD1

1Department of Emergency Medicine and 2Department of , Saint-Antoine, 184, rue du Faubourg Saint-Antoine, Paris, France and 3Department of Emergency Medicine, Highland General Hospital, 1411 East 31st Street, Oakland, California

SUMMARY INTRODUCTION The objective of the study was to determine the The ECG is an important element in the evaluation concordance of emergency ' and cardi- of a wide range of ED . The ED ologists' interpretations of emergency department often must interpret the ECG in difficult circum- (ED) electrocardiograms (ECG), to evaluate the stances: quickly, in isolation, without prior studies impact of ECG misinterpretation on patient manage- for comparison or a cardiologist's second opinion. ment, and to determine error rates as a function of Inaccuracy in the interpretation of the ECG by the level of physician training and the specific ECG emergency physicians may result in inappropriate diagnoses. ECG interpretations were registered management decisions regarding patient treatment prospectively using a programmed-response data or disposition. sheet. A second blinded interpretation by a staff Several large studies reviewing missed acute cardiologist was assumed to be correct. Only ECG (AMI) in the ED have suggested discrepancies with potential or probable clinical that failure to interpret the ECG accurately was a importance were considered as errors. The ED important factor in patient management errors.' 2 management of patients with ECG misinterpretations Two recent studies found a significant discordance

was reviewed by the investigators. The study was in the interpretation of ECGs between emergency http://emj.bmj.com/ performed at an urban university hospital using 300 medicine residents and cardiologists.3'4 Both consecutive ED ECGs. The analysis found 154 studies were retrospective, comparing open-ended errors of interpretation of which nine had probable ED interpretations with a follow-up interpretation by clinical significance, and 56 had indeterminant a cardiologist. This methodology is strongly oriented significance. The concordance was weak at 0.69 towards a comparison of diagnostic accuracy while

(Kappa = 0.32, weighted Kappa = 0.30) with a sig- ignoring a more fundamental evaluation of ECG on September 28, 2021 by guest. Protected copyright. nificant discordance (McNemar Chi 2:P < 0.05). interpretation skills. A prospective study format Error rates did not differ significantly between the permits a direct comparison of all elements of the diverse categories of physicians. In two cases, ECG interpretation process without the population interpretation errors impacted patient management bias inherent in retrospective studies. decisions but not patient outcomes. The most fre- Despite ECG misinterpretation rates in the ED quent errors involved repolarization abnormalities, of up to 58%, previous studies have found little ventricular hypertrophy and hemi-blocks. While clinical impact on patient management decisions. Correspondence: discordance was significant, errors in ECG inter- Although, based upon retrospective analyses, E.R. Snoey, pretation rarely impacted patient management. these results have raised questions regarding Department of Prospective evaluation of ECG interpretation may the relevance of quality assurance review of ECG Emergency Medicine, be a useful means of gauging physician skills. It can interpretation. Assuming that ED practice standards, Highland General also serve to focus educational activities on problem management strategies and patient populations Hospital 1411 East 31st Street, Oakland, areas in . may differ in France, the authors sought to pro- California 94602, Key words: Electrocardiogram, emergency de- spectively determine the concordance of emergency USA partment, interpretation physicians and cardiologists in the interpretation J Accid Emerg Med: first published as 10.1136/emj.11.3.149 on 1 September 1994. Downloaded from E.R. Snoey et al. of ED ECGs, and to measure the clinical impact of Table 1. Classification of ECG interpretation interpretation errors. discrepancies

Category I: No discrepancy or ECG with minor MATERIALS AND METHODS discrepancy

The institution is an urban, university-based teaching Sinus tachycardia hospital located in central Paris. The ED is licensed Sinus bradycardia to provide all levels of care and treats approximately Premature ventricular contractions 30000 adult patients per year. The department Premature atrial contractions is staffed 24-h a day by first year post-graduate First-degree atrial-ventricular block physicians, with a rotating pool of house officers Axis abnormality assuming the night duty. No house officer has Atrial enlargement Non-specific intraventricular conduction defect specific emergency medicine training. Attending Non-specific ST-T wave changes physician supervision is provided during day time and evening hours only. CateEory Il: ECG abnormalites with indeterminant clinical During a 2-month period, all patients, regardless significance of their ultimate disposition, for whom an ED ECG was performed were, included in the study. The Right bundle branch block most senior physician participating in the patient Left bundle branch block Left ventricular hypertrophy care was asked to complete a formal interpretation Right ventricular hypertrophy of the ECG using a programmed response data Left anterior hemi-block collection sheet. Each of the principal elements Left posterior hemi-block of the ECG interpretation (rate, rhythm, axis, QRS, ST abnormality hypertrophy and repolarization abnormalities) Repolarization abnormality required a response ranging from normal to one or Poor R wave progression more of a series of potential findings. A synthesis Q wave of the findings with a probable diagnosis was Category Ill: ECG abnormalites with probable clinical at the end. A second blinded interpre- registered significance tation using an identical data sheet was performed by a staff cardiologist. This interpretation was Ischemia assumed to be correct. Acute myocardial infarction Q-T interval ED and cardiologist interpretations were com- Prolonged http://emj.bmj.com/ pared. A misinterpretation was defined as any dis- Wolf-Parkinson-White syndrome crepancy between the cardiologist and the ED Abnormal rhythm Atrial fibrillation physician. The ECGs were subsequently classified Atrial flutter by the investigators into one of three categories Junctional rhythm based on the potential clinical importance of the Pacemaker - discrepancies: Category I ECGs with no dig- on September 28, 2021 by guest. Protected copyright. crepancy or with an error of little potential import-

ance, Category II - ECGs of indeterminate clinical

importance, and Category IlIl - ECGs with probable Overall concordance of interpretations was deter- clinical importance (Table 1). These groupings mined using the Chi2 McNemar test and Kappa test. define the potential for a specific ECG finding to Error rates based on the level of physician training change ED management (treatment, disposition were compared using the Chi2 test. etc.) if misinterpreted. In the situation where more than one discrepancy was identified, the most signi- RESULTS ficant error defined the category designation. For Category Ill patients (probable clinical import- A total of 300 consecutive ECG interpretations were ance), the ED charts were reviewed for evidence analysed, of which 102 were ultimately read as that ECG interpretations errors may have impacted normal and 198 abnormal. The analysis found 154 patient management decisions. Patient outcomes (51%) errors of interpretation of which 56 (19%) were reviewed at the time of hospital discharge or were of little clinical significance (Category I), 89 150 within 10 days of discharge from the ED. (30%) were of indeterminate significance (Category J Accid Emerg Med: first published as 10.1136/emj.11.3.149 on 1 September 1994. Downloaded from Emergency II) and 9 (3%) were of probable clinical significance to a ward bed for intravenous rehydration. The department ECG (Category Ill). cardiologist reported a short P-R interval with interpretation If the analysis is limited to indeterminate or import- repolarization changes characteristic of Wolf- ant errors (Categories II and 111) only, one finds a Parkinson-White syndrome. The authors believe total of 98 interpretation discrepancies. In 61 cases, this patient merited and a cardi- the discrepant interpretation was the result of a ology evaluation, given the context of a near syn- missed finding by the ED (false negative). In 37 copal event. She was discharged 2 days later cases, the errors related to an overread by the ED and was well at a 1-week follow-up by telephone. not confirmed by the cardiologist (false positive). All A second case involved a 66-year-old male hospi- but two of the nine Category IlIl patients were hospi- talized for an exacerbation of chronic obstructive talized at the time of presentation. pulmonary . Initial ED interpretation of the The overall misinterpretation rate for all three ECG suggested a left anterior superior hemi-block. categories was 51 %. If the analysis is limited only to The cardiologist reading noted S-T segment changes those errors in Categories 11 and Ill (indeterminate suggestive of anterolateral ischaemia. The patient or probable clinical significance), the rate improves was admitted to a ward bed for standard bron- to 32.7% with a Kappa = 0.32, and weighted Kappa chodilator . The ischaemic changes were = 0.30. The discordance was significant (P < 0.05). recognized by the admitting team and the patient No significant difference in error rates was found was ruled-out for myocardial infarction. The patient between the various levels of physician training nor was discharged in good condition 7 days later with a consistent pattern of errors for a particular phy- cardiology follow-up. The authors felt that admission sician. The concordance rates for the ten most to a monitored unit with cautious use of broncho- commons errors are listed in Table 2. dilator therapy would have been more appropriate A total of 12 patients (4%) in our study exhibited give the initial ECG findings. ECG evidence of AMI. All were identified correctly Using a review of hospital discharge summaries by the ED physician. By protocol, all patients had a together with telephone follow-up for those dis- cardiology consultation prior to initiating thrombolytic charged directly from the ED, it was determined that therapy. none of the nine Category IlIl patients experienced a ED records were reviewed for each of the nine near-term complication. patients with Category IlIl ECG errors. In two cases, interpretation errors may have impacted patient DISCUSSION management decisions. The first case concerned a 76-year-old female who presented with weakness The analysis confirms the significant discordance http://emj.bmj.com/ and near syncope. The ED interpreted her ECG as between ED and cardiologist interpretations of non-specific S-T segment abnormalities. The patient ECGs previously described. The importance of was given a diagnosis of dehydration and admitted accurate ECG interpretation has been brought into greater focus recently with the more routine use of thrombolytics in the ED and the unmitigating on September 28, 2021 by guest. Protected copyright. Table 2. Ten most common ECG elements with medicol-legal environment. Time to thrombolytic concordance results: Kappa Coefficient varies from therapy has been confirmed as an independent 0 (no concordance) to 1.0 (maximum concordance) variable in improving survival for patients with AMI. As such, a timely initiation of thrombolytic therapy ECG Finding Kappa coefficient (0-1.0) may preclude a concurrent reading of an ECG by a cardiologist. Several studies have evaluated Left ventricular hypertrophy 0.41 Left anterior hemi-block 0.43 accuracy in identifying patient ST segment abnormality 0.45 candidates for thrombolytic therapy. In a study by Bundle branch block 0.54 Ho et al., 13 out of 236 patients with ECGs indicative Right ventricular hypertrophy 0.59 of acute myocardial infarction (AMI) were missed Pathologic Q-wave 0.64 by emergency physicians.1 Mistakes most often Left atrial enlargement 0.72 involved small S-T segment abnormalities or bundle Premature atrial contraction 0.92 branch blocks. A second study by Lee et al. found Premature ventricular 0.92 that 21 out of 445 ECGs indicating AMI had been contraction Heart rate 0.96 misinterpreted by emergency physicians.2 In a 151 related study population of patients with missed J Accid Emerg Med: first published as 10.1136/emj.11.3.149 on 1 September 1994. Downloaded from E.R. Snoey et al. AMI, 16 of 34 ECGs (47%) with evidence of AMI or polarization abnormalities has been emphasized in ischaemia were interpreted as normal or showing a recent series of articles reviewing in-hospital nonspecific ST-T wave abnormalities.2'5 Other complications from AMI.9'10 Bundle branch blocks studies reviewing the problem of missed ED AMI and left ventricular hypertrophy, in addition to and litigation have found ECG misinterpretation to myocardial ischaemia, were found to be prognostic be a contributing cause in up to 47% of cases.6'7 In for in-hospital cardiac complications. The authors the USA, missed AMI continues to occupy the suggest that suspected AMI patients without these number one position for compensation awarded in ECG findings (patients with normal, non-specific, malpractice litigation. It comprises 10% of all cases or unchanged ED ECGs) may be safely admitted to that go to litigation and between 20 and 45% of an intermediate care unit. Cardiac care unit overload all dollar losses in the field of emergency medicine.8 and hospital expense could thereby be reduced. This current study addressed the issue of ECG Given the high misinterpretation rate for bundle interpretation in the setting of all forms of ED pathol- branch blocks and left ventricular hypertrophy found ogy without focusing specifically on ischaemic heart in this and other studies, it is conceivable that the disease. In a recent retrospective study of ECG clinical relevance of these findings may change interpretation at an emergency medicine teaching as patient admission protocols evolve. Because programme, Kuhn et al. found error rates of 8.6 and no intermediate cardiac care unit exists at our 8.3% for ECGs with indeterminate and probable institution, no patient care was affected. clinical importance respectively.3 In only three cases This study addresses not only the diagnostic (0.9%) was patient ED management felt to be accuracy of ED ECG interpretation, but also basic affected. Westdorp et al. had similar results in a interpretation skills as defined by a direct com- review of 716 ECGs in patients discharged from the parison of each element of the ECG. While retro- ED.4 Errors rates of 40.6 and 17.5% were reported spective studies of ECG concordance provide useful for indeterminate and probable errors respectively. quality assurance information in terms of patient ECG misinterpretation was judged to have changed outcomes, they are less precise in their ability to patient care in two cases (0.5%). While this study evaluate physician electrocardiographic skills or found a much smaller rate of errors with probable deficiencies. The evaluation of individual physicians clinical importance (3%), overall discordance or the orientation of educational activities in the of ECG interpretation remained significant. The domain of ECG interpretation may be well served most common errors involved the identification of by this kind of analysis. repolarization abnormalities, bundle branch blocks Despite a significant overall discordance, ED

and hemi-blocks. ED physicians were most adept at interpretation errors rarely affect patient manage- http://emj.bmj.com/ identifying premature contractions, axis abnormalities ment decisions. Kuhn et a!. argue that systematic and rhythm disturbances. Of the nine potentially review of ED ECG interpretation by a staff cardi- significant errors classified in Category Ill, six were ologist is not indicated based on these findings.3 represented by three diagnoses: prolonged Q-T Clearly, if a quality assurance activity is to be judged syndrome, Wolf-Parkinson-White syndrome and according to its impact on patient management or

atrial fibrillation (Table 3). In two cases, interpret- outcome, the effort involved in a systematic review on September 28, 2021 by guest. Protected copyright. ation errors impacted management decisions but of all tracings seems unwarranted. However, ED not patient outcomes. management decisions depend on a host of different The importance of ancillary findings such as factors of which the ECG is but one. Retrospective intraventricular conduction disturbances and re- review of patient management and outcome is there- fore likely to be an insensitive means of determining the clinical relevance of an ECG interpretation error. If one's goal is to identify ED management mistakes Table 3. Category Ill ECG Diagnostic Errors or poor patient outcomes, other quality assurance strategies may prove more effective, such as call- Wolf-Parkinson-White syndrome back of predetermined high-risk patients. Prolonged Q-T syndrome There were several potential limitations to this Atrial fibrillation study. The ECG 'gold standard' used was the Junctional tachycardia interpretation of one staff cardiologist. We did Atrial flutter Anterolateral ischaemia by S-T segment not develop a consensus process to review interpret- 152 ation discrepancies. The evaluation of patient Emergency management, based on a retrospective review of Joumal of the American College of Cardiology 15, J Accid Emerg Med: first published as 10.1136/emj.11.3.149 on 1 September 1994. Downloaded from department ECG patient charts and clinical outcomes, was limited 192A. interpretation to Category Ill patients only. It was, however, per- 2. Lee T., Weisberg M., Brand D. etal. (1987) Candidates formed on a consensus basis between two exper- for thrombolysis among emergency room patients ienced ED physicians. Lastly, the author's goal with acute . Annals of was to evaluate physician accuracy in ECG inter- 110, 957-962. M., Morgan M. & Hoffman J. (1992) Quality pretation. The impact of using pre-programmed 3. Kuhn assurance in the emergency department: Evaluation ED interpretations data sheets on the quality of ECG of the ECG review process. Annals of Emergency was not tested a priori. It is likely that physicians in Medicine 21, 10-15. this study were more fastidious in their interpret- 4. Westdorp E., Gratton M. & Watson W. (1992) Emerg- ations given the prospective study format. This ency department interpretation of electrocardiograms. is supported by the relatively low error rate in com- Annals of Emergency Medicine 21, 541-544. parison to previous studies of this kind. 5. Lee T., Rouan G. & Weisberg M. (1987) Clinical Characteristics and natural history of patients with acute myocardial infarction sent home from the emerg- CONCLUSION ency room. American Journal of Cardiology 60, This analysis found a significant discordance 219-224. between ED physicians and cardiologists in the 6. Karcz A., Holbrook J., Auerbach B. et al. (1990) of malpractice claims in emergency interpretation of ED ECGs. Despite this discordance, Preventability medicine; A closed claims study. Annals ofEmergency errors were deemed unlikely to impact patient Medicine 19, 865-873. management decisions. The authors of this study 7. Rusnak R., Stair T., Hansen K. et al. (1989) Litigation support periodic prospective reviews of ECG inter- against emergency physicians: Common features in pretation skills as a means of evaluating overall ED cases of missed myocardial infarction. Annals of performance and individual physician competence. Emergency Medicine, 1029-1034. The ECG remains one of the most common, use- 8. Hill M. (1989) Failure to diagnose myocardial infarc- ful and cost effective tests available to the ED tion: The emergency physician's greatest liability in physician. The skillful interpretation of the ECG practice. QRC Advisor 6, 1-8. should continue to be a focus of quality assurance 9. Stark M. & Vacek J. (1987) The initial electrocardiogram and educational activity. during admission for myocardial infarction. Archives of Intemal Medicine, 843-846. 10. Villanueva F., Sabia P., Afrookteh A., Pollock S., REFERENCES Hwang L. & Kaul S. (1992) Value and limitations of current methods of evaluating patients presenting to http://emj.bmj.com/ 1. Ho M., Kudenchuk P., Eisenberg M. et al. (1990) the emergency room with cardiac-related symptoms Patient selection for thrombolytic therapy: Emergency for determining long-term prognosis. American physicians versus electrocardiographer (abstract). Journal of Cardiology 69, 746-750. on September 28, 2021 by guest. Protected copyright.

153