- Email: [email protected]
Causes of errors in the electrocardiographic diagnosis of atrial fibrillation by physicians
Journal of Electrocardiology 40 (2007) 450 – 456 www.jecgonline.com
Causes of errors in the electrocardiographic diagnosis of atrial fibrillation by physicians Jorge Mario Davidenko, MD, FACC,a,b,c,4 Lisa Simonetta Snyder, RN, MSN, FNPb a
Department of Cardiology, St Joseph Hospital Health Center, Syracuse, NY, USA b New York Heart Center, Syracuse, NY, USA c SUNY Upstate Medical University, Syracuse, NY, USA Received 26 November 2006; accepted 11 January 2007
Abstract
Background: The emphasis of most large studies has been placed on the treatment and prevention of atrial fibrillation (AF) and its complications. Little is known about the accuracy of physicians in the electrocardiographic (ECG) diagnosis of AF and the possible causes of the diagnostic errors. Methods: Over a period of 10 months, a total of 35 508 ECGs (28 356 patients) were overread in a 385-bed community hospital within 24 hours of the initial reading. Corrected ECGs were returned to the patient file. The gold standard for the final diagnosis was based on the consensus by the cardiologist readers. Results: In all, 35 508 ECGs were reviewed. A total of 2809 cases of AF were studied. Incorrect diagnoses related to AF were found in 219 cases. Type I errors (overdiagnosis) occurred in 137 cases. Rhythms with irregular R-R intervals (sinus rhythm with premature atrial contractions and atrial tachycardia or flutter with variable atrioventricular conduction) were misdiagnosed as AF. The presence of low-amplitude atrial activity and/or baseline artifact significantly increased the likelihood of the erroneous diagnosis, whereas ventricular rates of 130 beats/min did not influence the rate of error. Type II errors (missed AF) occurred in 82 cases where AF was either missed or confused with atrial tachycardia/flutter. Finally, ventricular pacing significantly increased the likelihood of type II errors. Conclusions: In our institution, about 900 ECGs are read each week and 5 of them carry a wrong interpretation related to AF. More attention to common sources of errors as reinforced by an ongoing quality improvement program may reduce the rate of mistakes and thus prevent serious consequences. D 2007 Elsevier Inc. All rights reserved.
Keywords:
Atrial fibrillation; Electrocardiographic errors; Arrhythmias
Introduction Atrial fibrillation is recognized as one of the major challenges in modern cardiology both in terms of the treatment of arrhythmia and prevention of stroke. Efforts have been made to define the best therapeutic approaches such as rate vs rhythm control or pharmacologic vs nonpharmacologic methods. 1-3 A few studies have addressed the misinterpretations of the 12-lead ECG4-6 and the potential consequences of diagnostic errors related
Abbreviations: AF, Atrial fibrillation; Afl/tach, Atrial flutter or atrial tachycardia; SR/PACs, Sinus rhythm and premature atrial contractions. 4 Corresponding author. 1000 East Genesee Street, Syracuse, NY 13210, USA. Tel.: +1 315 471 1044; fax: +1 315 474 4312. E-mail address: [email protected] 0022-0736/$ – see front matter D 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jelectrocard.2007.01.003
to AF.4,7,8 A recent article by Anh et al7 has determined that familiarity with the patient may improve the accuracy of the ECG interpretation. In the current study, we analyzed the influence of some ECG features as possible reasons for the erroneous interpretation of AF by physicians. The gold standard for the final diagnosis was based on the consensus by cardiologist readers. Methods A total of 35 508 consecutive ECGs were overread during the months of March through December 2005 at St. Joseph’s Hospital Health Center (Syracuse, NY). The study was approved by the hospital’s institutional review board. The ECGs were obtained with Marquette machines and the interpretation software used was MAC 5000 6A, last
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
upgraded on May 16, 2004. The tracings were displayed to show 2.5 seconds of each of the 12 leads and 10-second simultaneous strips of leads V1, lead II, and V5. The simultaneous recordings of all 12 leads were available upon the request of the reader. The ECGs were presented to the primary readers with the computer-generated report (GE Marquette Medical Systems 12SL MAC Rhythm, Milwaukee, WI, USA). The reports were modified and confirmed by 1 of the 54 primary readers of the hospital. In our hospital, ECG readers are board certified in internal medicine; 35 of them also have a certification in cardiovascular diseases and 2 of them are board-certified electrophysiologists. A copy of the confirmed ECG was delivered to the patient’s chart. A second copy was saved for analysis. The initial screening of all confirmed copies and the separation by diagnosis was done with the cooperation of a highly trained nurse. The review process was performed on a daily basis. The incorrect interpretations were corrected and returned to the patient’s chart within a 24-hour period from the initial reading. This effort was aimed at minimizing any potential consequences derived from the diagnostic errors. In addition, at the end of each month all corrected ECGs were sent back to the primary reader as part of a quality improvement program. Finally, all corrected ECGs were discussed during the monthly ECG conference. If— after a careful review and a discussion with the primary reader, as well as later with the rest of the physicians during the ECG conference—there was still uncertainty regarding the diagnosis, then those ECGs were excluded from the study. Electrocardiograms were separated into 2 groups: those showing spontaneous QRS complexes and those in which 100% of the QRS complexes were paced. The groups were subdivided according to the actual supraventricular rhythm. Two types of errors were analyzed: type I or overdiagnosis of AF, and type II or missing the diagnosis of AF. Definitions Atrial fibrillation was defined as a supraventricular tachyarrhythmia characterized by the replacement of consistent P waves by rapid oscillations or fibrillatory waves (ie, f waves) that vary in size, shape, and timing, associated with an irregular, frequently rapid ventricular response when the atrioventricular (AV) conduction is intact.9 Regular R-R intervals, however, are possible in the presence of AV block with ventricular or junctional rhythms and ventricular pacing. High-amplitude AF was defined as AF in which the amplitude of the f waves is more than 0.2 mV in at least 1 lead. High ventricular rate was defined as a ventricular rate of 130 beats per minute (beats/min) or higher. Atrial flutter and atrial tachycardia were grouped together and were defined as distinct atrial activity (P waves or flutter waves) observed in 1 or more leads, with regular intervals, and with a rate of 100 to 350 beats/min. Statistics Frequencies of correctly and incorrectly diagnosed ECGs were compared by using a v 2 test with 1 df. Detection accuracy of AF in cases with and without 100% ventricular
451
pacing was measured with signal detection theory using the discriminability (dV) statistic.10 A dV value exceeding 1 indicates that true cases can be reliably discriminated from false cases; a dV of more than 3 indicates excellent discriminability. Results Of a total of 35 508 ECGs reviewed, there were 2809 cases (7.9%) of true AF. A total of 219 ECG carried a wrong diagnostic interpretation related to AF, representing an error rate of 7.8%. Errors in the diagnosis of AF were of 2 types: type I when the diagnosis of AF was made, but other rhythms were actually present (typically rhythms with irregular R-R intervals); and type II when the diagnosis of other rhythms was made, but AF actually was present (typically AF with high-amplitude f waves or with regular R-R intervals). To facilitate the description of the results, ECG with and without 100% ventricular pacing were analyzed separately. Electrocardiograms without 100% ventricular pacing Atrial fibrillation was found in 2402 of 34 008 ECGs, representing an incidence of 7.1%. Type I error (overdiagnosis of AF) Atrial fibrillation was wrongly diagnosed in 108 cases, representing 4.5% of all true cases of AF. Rhythms confused with AF were sinus rhythm with premature atrial contractions (SR/PACs, 42 cases, see Table 1 and Fig. 1A and B) and atrial flutter or atrial tachycardia with variable AV block (Afl/tach, 66 cases, see Table 2 and Fig. 1 C and D). Comparisons were made between features associated with correct and incorrect diagnoses in SR/PACs and Afl/tach cases. Table 1 shows that, in cases of SR/PACs, low-amplitude atrial activity and baseline artifact were more prevalent in cases with overdiagnosis of AF (47.6% and 31.0%, respectively) than in cases with correct diagnosis of SR/PACs (15.7% and 7.1%, respectively; P b .001 for both comparisons using v 2 test with 1 df). Table 2 shows that, in cases of Afl/tach with variable block, frequencies of low-amplitude atrial activity, baseline artifact, and ventricular rates of more than 130 beats/min were not statistically different between cases with overdiagnosis of AF and cases with correct diagnosis of Afl/tach, although there was nonsignificant trend toward higher frequencies of low-amplitude atrial activity and baseline artifact in cases of overdiagnosis of AF. Fig. 1 shows examples of type I errors in the absence of 100% ventricular pacing. Panel A shows SR with sinus Table 1 Type I error: cases of SR/PAC in the absence of 100% ventricular pacing
Low-amplitude atrial activity Baseline artifact All of the above None of the above Total
True diagnosis of SR/PAC
False diagnosis of AF
259 (15.7%) 117 (7.1%) 10 (0.6%) 1261 (76.6%) 1647
204 (47.6%) 13* (31.0%) 3* (7.1%) 6* (14.3%) 42
4 Significantly different at P b.001 using v 2 with df = 1.
452
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
Fig. 1. (A) Sinus rhythm with sinus arrhythmia and PACs giving rise to an irregular rhythm initially diagnosed as AF. Low-amplitude P waves can be seen preceding each QRS. (B) Sinus rhythm and PACs with baseline artifact initially diagnosed as AF. (C) Rhythm strip obtained in lead VI, lead II, and V5 showing a supraventricular tachycardia with grouping of the QRS complexes that is perfectly repetitive as shown by the brackets. There is a stable 8:3 ratio. (D) Irregularly irregular narrow complex tachycardia initially diagnosed as AF. When carefully measured, the atrial waves march perfectly (as indicated by the asterisk).
arrhythmia and PACs giving rise to irregularly irregular R-R intervals. The P waves show very low amplitude. This led to the wrong diagnosis of AF. Panel B shows SR and PACs with baseline artifact. The baseline artifact interferes with visualization of the P waves and gives the wrong impression of f waves. Panel C shows Afl/tach with variable AV block. Atrial waves show low voltage and are not clearly visualized in all leads (arrows). The R-R intervals are indeed irregular. A more careful analysis of the tracing, however, reveals a repetitive pattern in the R-R intervals (see the perfect grouping of the QRS complexes as marked by the brackets). The AV conduction ratio in this case is 8:3, and results from alternate Wenckebach block at the AV nodal level. This repetitive pattern reflects the presence of a perfectly repetitive supraventricular source suggesting the diagnosis of atrial flutter even if the flutter waves are not well visualized. Panel D shows another example of Afl/tach with variable block. In this case, there seems to be no pattern in the irregularity of the R-R interval. The
differential diagnosis with AF is based only on the finding of the repetitive atrial activity as indicated by the asterisks. Type II errors (missed AF) Table 3 shows that AF was missed in 57 cases (2.4%). These ECGs were compared with those in which the diagnosis of AF was correct. High-amplitude AF was more Table 2 Type I error: cases of atrial flutter/atrial tachycardia with variable AV block in the absence of 100% ventricular pacing
Low-amplitude atrial activity Baseline artifact Ventricular rate, 130 beats/min More than 1 of the above None of the above Total
True diagnosis of Afl/tach
False diagnosis of AF
279 47 121 17 177 641
32 8 12 54 94 66
(43.5%) (7.3%) (18.9%) (2.7%) (27.6%)
(48.5%) (12.1%) (18.2%) (7.6%) (13.6%)
4 Significantly different with P b.05 using v 2 with df = 1.
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456 Table 3 Type II error: cases of AF in the absence of 100% ventricular pacing
High-amplitude AF Regular Ventricular Activity Ventricular rate, 130 beats/min More than 1 of the above None of the above Total
True diagnosis of AF
Missed diagnosis of AF
309 49 290 34 1663 2345
274 444 9 2 15* 57
(13.2%) (2.1%) (12.4%) (1.4%) (70.9%)
Table 4 Summary of AF-related diagnoses in cases without 100% ventricular pacing
(47.4%) (15.8%) (3.5%) (26.3%)
4 Significantly different at P b .001. 44 Significantly different at P b .05 using v 2 with df = 1.
prevalent in the group with incorrect diagnosis (47.4% vs 13.2%, respectively; P b .001). Regular ventricular rate observed in cases of relatively high degree of AV block and a competing junctional rhythm also was more prevalent in cases in which the diagnosis of AF was missed (7.0% vs 2.1%, respectively; P b .05). Finally, frequencies of ventricular rates of more than 130 beats/min in the 2 groups were not statistically different. Fig. 2 shows examples of type II errors (missed AF). Panel A shows a typical case of coarse AF. A careful measurement of the baseline waves (shown by the brackets) highlights the irregularity of the atrial activity characteristic of AF. Panel B shows a case of AF with regular R-R interval. The baseline shows erratic undulations; yet, the diagnosis of AF was missed. Panel C shows a recording of the same patient a few hours later. Here, the junctional rhythm is interrupted by sporadic ventricular paced com-
453
True AF Other rhythms Total
Diagnosed as AF
Diagnosed as other rhythms
Total
2345 108 2453
57 2288 2345
2402 2396 4798
Percentage of missed AF, 2.3%; percentage of overdiagnosed AF, 4.5%; dV = 3.677.
plexes, and the R-R interval shows irregularity. This time, the diagnosis of AF was correctly identified. In summary, as shown in Table 4, the overall rate of overdiagnosis of AF (type I error) was 2.4%, and the rate of missed AF (type II error) was 4.5%. Electrocardiograms with 100% ventricular pacing Atrial fibrillation was found in 407 of 1500 ECGs with 100% ventricular pacing, representing an incidence of 27.1% (nearly 4 times the incidence of AF in the whole population). As shown in Table 5, the presence of 100% ventricular pacing was associated with a 6.1% type I error rate (missed AF) and a 3.2% type II error rate (overdiagnosis of AF). Type I error (overdiagnosis of AF) Atrial fibrillation was incorrectly diagnosed in 29 cases. Rhythms confused with AF were Afl/tach (18 cases) and SR or retrograde P waves (11 cases). In general, the atrial
Fig. 2. (A) Tracing showing V1 originally diagnosed as atrial flutter with variable AV conduction. Yet a more careful analysis shows that the atrial activity is not regular (see the brackets). (B) Recording obtained at 4:56 pm with an initial report indicating the presence of junctional tachycardia and nonspecific ST and T wave abnormalities. There was no mention about the atrial activity. (C) Recording from the same patient in (B) obtained at 5:30 pm showing btypicalQ AF pattern with varying R-R intervals (as well as sporadic ventricular pacing). Note that in both (B) and (C), the baseline shows the same type of f waves.
454
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
Table 5 Summary of AF-related diagnoses in cases with 100% ventricular pacing
True AF Other rhythms Total
Diagnosed as AF
Diagnosed as other rhythms
Total
382 29 411
25 889 914
407 918 1325
Percentage of missed AF, 6.1%; percentage of overdiagnosed, AF: 3.2%; dV = 3.401.
activity had low amplitude or was obscured by the ventricular activity. Fig. 3 shows typical examples with ventricular pacing in which the underlying rhythm was wrongly interpreted as AF. Panel A shows Afl/tach. Regular negative flutter waves (asterisks) are seen in only segments of the cardiac cycle, but can be tracked throughout the tracing (arrows) with a constant cycle length. Panel B shows isolated P waves with low amplitude. Yet, they show a constant cycle length. Because the sinus rate is very slow
and there is no synchronization with the ventricular pacing (AV dissociation), the reader assumed an absence of P waves and made the incorrect diagnosis of AF. Panel C shows P waves that are difficult to see because they occur just after the QRS (ie, retrograde P waves). Here, again, the primary reader overlooked the P waves and made the wrong diagnosis of AF. Finally, panel D shows a case in which the P waves are nearly unrecognizable. The interesting feature of this ECG is the presence of premature ventricular stimuli followed by narrow QRS complexes (first, fifth, and ninth complexes) that represent fusion between paced and native QRS. This can be explained by the presence of a hidden P wave that is both tracked by the generator and conducted to the ventricles. A more careful look at the tracing reveals the presence of a very low voltage deflection at the end of the T wave preceding the premature stimulus. Also, there is a consistent low-amplitude deflection preceding the ventricular stimulus after the premature beats (second, sixth, and 10th complexes). If we divide the
Fig. 3. Four examples showing 100% ventricular pacing. (A) Initially reported as ventricular pacing with undetermined underlying supraventricular rhythm, likely AF. A more careful analysis shows the presence of very rhythmic atrial activity (asterisks), some of which are concealed within the QRS. The correct diagnosis is atrial tachycardia with no evidence of AV conduction and nonsynchronized ventricular pacing. (B) The initial diagnosis was absence of P waves, AF. P waves are indeed present (arrows in V1) at a slow and very regular rate. The correct diagnosis is sinus bradycardia with high degree AV block and nonsynchronized ventricular pacing. (C) Again, the diagnosis was underlying AF. P waves, however, are retrograde. They can be seen after each QRS (arrows) with a fixed RP interval. (D) Atrial activity is not obvious so the initial report suggested the diagnosis of AF. There are 3 premature ventricular stimuli. The inset in lead V1 shows a very low-amplitude P wave occurring at the end of the T wave that precedes the premature stimuli. In addition, there is a bpseudo r waveQ preceding the next QRS in all 3 instances. If we assume that this is indeed the cycle of the atrial activity, we find that there is a very regular sinus bradycardia (asterisks and arrows) underlying ventricular pacing. The correct diagnosis is sinus bradycardia with preserved AV conduction (the premature QRS is a fusion between paced and native activation) and ventricular pacing. There seems to be appropriate atrial sensing, but there is no evidence of atrial pacing.
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
interval between these deflections in 2, we notice that there may be indeed a regular sinus rhythm underlying ventricular pacing, thus ruling out AF. Type II errors (missing AF) Atrial fibrillation was missed in 25 cases with 100% ventricular pacing. In most cases, there was no actual mention of the underlying rhythm. In 2 cases, AF had relatively high amplitude waves and was confused with atrial flutter. Computer interpretation Although the scope of the present study was to determine the accuracy of physician-generated ECG reports, we also assessed the rate of accuracy of the computer generated report. From 3596 cases read in 1 month, the computer overdiagnosed AF in 30. Sixteen of these cases were indeed sinus rhythm with PACs, and 14 were cases Afl/tach. Of the 250 cases of true AF, the computer report missed the diagnosis in 44 cases (17.6%). Eighteen of 44 cases showed 100% ventricular pacing and the computer offered no comments on the underlying rhythm. Other diagnosis generated by the computer were: undetermined rhythm (18 cases), sinus tachycardia (7 cases), atrial flutter (3 cases), sinus rhythm (3 cases), atrial rhythm (1 case), junctional rhythm (1 case), and wide complex rhythm (1 case). It is interesting to note that although the rate of inaccuracy of the computer reports was relatively high, it did not seem to significantly influence the reports generated by the physicians. In fact, most of the ECGs were modified by the primary reader and in some cases the reader wrongly overruled the computer diagnosis. Discussion Both confirming and discarding the diagnosis of AF may greatly influence the management of a patient. The 12-lead ECG remains the most common tool in the diagnosis of AF. The present study shows that over a period of 10 months there were 219 incorrect diagnoses related to AF, representing 5 errors per week. The main objective of the study was to determine the most likely reasons for the erroneous diagnoses. Type I error Irregularly irregular ventricular rhythm The definition of AF has been classically associated with the words birregularly irregularQ rhythm. Although irregularly irregular ventricular rhythm is indeed a very common feature of AF, it is neither exclusive of AF nor necessary for its diagnosis. Sinus rhythm with PACs and or sinus arrhythmia may lead to a very irregular rhythm. If the P waves are not well visualized because of a low-amplitude and/or the presence of artifact, there is high likelihood of an erroneous diagnosis of AF. When a regular atrial rhythm (ie, atrial flutter) drives the ventricles, the ventricular rhythm may either be regular or irregular depending on the rate of
455
the atrial rhythm and the electrophysiologic properties of the AV node. Wenckebach type of conduction with varying periodicities may result in organized irregularity (ie, patterns such as those shown in Fig. 1C) or a totally irregular rhythm (Fig. 1D) thus mimicking AF. Type II error Regular R-R interval ventricular rhythms It is difficult to associate AF with a regular ventricular rhythm. The presence of a high-degree AV block with or without ventricular pacing may set the basis for this combination. Our study shows that the presence of regular ventricular rhythm is very likely to be associated with a missed diagnosis of the underlying AF. High-amplitude AF (bpseudoQ atrial flutter) Atrial fibrillation is disorganized and unpredictable. On the other hand, atrial flutter and other forms of atrial tachycardia in general show a fixed interval with little or no variations over the 10-second period of the ECG recording.11 The random nature of activation during AF makes it unlikely for operators to find identical cycle lengths in the 10-second tracing of the ECG. Random activation during AF (f waves) may still appear as bPQ waves separated by relatively irregular intervals of nearly absence activity. It is necessary to determine if the morphology of the ballegedQ P is consistent in leads obtained simultaneously as one may expect in cases of real P waves or PACs originated in a single focus. The distinction between AF and some forms of multifocal atrial tachycardia may be very difficult. The search for reasonable association between the P waves and the QRS complexes may help distinguish between these 2 arrhythmias. In AF, higher amplitude waves (simulating P waves) are not necessarily more likely to conduct through the AV node and thus be followed by QRS complexes. On the other hand, in multifocal atrial tachycardia, QRS are usually preceded by a discernible P wave. High ventricular rates ( N130 beats/min) Although rapid ventricular rate was found both in cases of Afl/tach and AF, it did not increase the rates of either type I or type II error. Ventricular pacing In our study, cases with 100% ventricular pacing were associated with an incidence of AF that was 4 times higher than in cases with an absence of 100% ventricular pacing. Importantly, the presence of 100% ventricular pacing significantly increased the risk of missing AF. A signal detection theory analysis revealed a significantly higher dVfor these cases (dV = 3.40 vs dV = 3.68; P b .05). Thus, patients with pacemakers seem to be at higher risk of having undiagnosed AF and slightly lower risk of having overdiagnosis of AF. In many cases, missing the diagnosis of AF was the result of omitting the description of the supraventricular rhythm in the presence of ventricular pacing. On the other hand, type I error occurred when the P waves were not
456
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
obvious. The apparent absence of P waves led to the wrong assumption of the presence of AF. A careful search for hidden P waves may help avoid these mistakes. In other cases, the presence of irregular ventricular pacing should alert the reader to the possibility of unrecognized organized atrial activity that has been tracked by the generator. In some cases, the interrogation of the pacemaker (ie, lowering the pacing rate or acquiring the intracardiac electrograms) may be necessary to rule out the presence of AF. Computer interpretations The scope of the current study was to analyze the reading abilities of the physician. Yet, we observed that our computer program has a relatively high inaccuracy rate (from a total of 250 true AF, the computer missed 44 [18%]). The most common reason for this type of mistake was the presence of 100% ventricular pacing in which the computer made no comments on the underlying rhythm. In addition, of 3597 reports the computer overdiagnosed AF in 30 (0.8%). Similar to what was observed with the physicians, the most common rhythms confused with AF were sinus rhythm with PACs and atrial tachycardia/atrial flutter with variable AV conduction. In spite of the similarities between computer and human sources for error, the computer was not found to be the reason for human mistakes because most of the confirmed reports in our series included corrections of the original computer reading. Limitations The study was based on consecutive ECGs read. Thus, the incidence of AF may not represent the actual incidence of the arrhythmia in our institution, because several tracings may have been obtained from the same patient. The main goal of the study, however, was to determine the rate of error and how to reduce it. Because in some cases ECGs from the same patient with the same rhythm had different interpretations, we decided not to exclude them from the study. There was no attempt at determining if the initial error in the ECG interpretation led to changes in the care of the patient. The corrections were done and delivered within a 24-hour period, thus reducing the chances of complications derived from the initial erroneous ECG interpretation. Although the number of ECG readers was relatively high (54), the results of this study may only reflect the quality of reading in our hospital. Until similar quality assessment studies are performed in other centers, we will not know if our results reflect a more general problem. Implications This study conducted in a midsize community hospital indicates that the ECG diagnosis of AF is not accurate. About 5 ECGs per week shows are misdiagnosed as AF.
Type I errors (overdiagnosis AF) may lead to the unnecessary use of anticoagulation or antiarrhythmic drugs when the real rhythm is simply sinus rhythm and PACs. When the real rhythm is atrial flutter or atrial tachycardia, the diagnostic error use may result in lack of appropriate use of well-established and readily available ablation techniques. Type II errors (missed AF) may result in inadequate measures for the prevention of stroke. Because the 12-lead ECG is still the main diagnostic tool of AF, efforts should be placed in improving the quality of ECG reading. In our institution, a quality improvement program is in place by which corrected ECGs are sent to the primary readers on a monthly basis for review and discussion. In addition, misinterpreted ECGs are presented during the monthly ECG conference. Future studies may determine the usefulness of such a program in reducing diagnostic errors. Acknowledgments The authors thank Dr. Nicolas Davidenko, who performed the statistical analysis and contributed with important ideas for the design of the study, and Dr. Dennis Ehrich, Vice President of Medical Affairs, St Joseph Hospital, for financial support and encouragement. References 1. Wyse DG, Waldo AL, Di Marco JP. The Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) Investigators. A comparison of rate control and rhythm control in patients with atrial fibrillation. N Engl J Med 2002;347:1825. 2. Van Gelder IC, Hagens VE, Bosker HA, et al. A comparison of rate control and rhythm control in patients with recurrent persistent atrial fibrillation. N Engl J Med 2002;347:1834. 3. Haissaguerre M, Jais P, Shah DC, et al. Spontaneous initiation of atrial fibrillation by ectopic beats originating in the pulmonary veins. N Engl J Med 1998;339:695, 1997;20:337 4. Bogun F, Anh D, Kalahasty G, et al. Misdiagnosis of atrial fibrillation and its clinical consequences. Am J Med 2004;117:706. 5. Knight BP, Michaud GF, Strickberger SA, Morady F. Electrocardiographic differentiation of atrial flutter from atrial fibrillation by physicians. J Electrocardiol 1999;32:315. 6. Poon K, Okin PM, Kligfield P. Diagnostic performance of a computer based ECG rhythm algorithm. J Electrocardiol 2005;38:235. 7. Anh D, Krishnan S, Bogun F. Accuracy of electrocardiogram interpretation by cardiologists in the setting of incorrect computer analysis. J Electrocardiol 2006;39:343. 8. Finsterer J, Stollberger C, Gatterer E. Oral anticoagulation for ECG tremor artifact simulating atrial fibrillation. Acta Cardiol 2003;58:425. 9. Bellet S. Clinical disorders of the heart beat. 3rd ed. Philadelphia7 Lea & Febiger; 1971. 10. Dorfman DD, Alf E. Maximum-likelihood estimation of parameters of signal-detection theory and determination of confidence intervals— rating method data. J Math Psychol 1969;6:487. 11. Prystowsky EN, Katz AM. Atrial fibrillation. In: Topol ES, editor. Textbook of cardiovascular medicine. Philadelphia7 Lippincott-Raven; 1998. p. 1827.
Causes of errors in the electrocardiographic diagnosis of atrial fibrillation by physicians Jorge Mario Davidenko, MD, FACC,a,b,c,4 Lisa Simonetta Snyder, RN, MSN, FNPb a
Department of Cardiology, St Joseph Hospital Health Center, Syracuse, NY, USA b New York Heart Center, Syracuse, NY, USA c SUNY Upstate Medical University, Syracuse, NY, USA Received 26 November 2006; accepted 11 January 2007
Abstract
Background: The emphasis of most large studies has been placed on the treatment and prevention of atrial fibrillation (AF) and its complications. Little is known about the accuracy of physicians in the electrocardiographic (ECG) diagnosis of AF and the possible causes of the diagnostic errors. Methods: Over a period of 10 months, a total of 35 508 ECGs (28 356 patients) were overread in a 385-bed community hospital within 24 hours of the initial reading. Corrected ECGs were returned to the patient file. The gold standard for the final diagnosis was based on the consensus by the cardiologist readers. Results: In all, 35 508 ECGs were reviewed. A total of 2809 cases of AF were studied. Incorrect diagnoses related to AF were found in 219 cases. Type I errors (overdiagnosis) occurred in 137 cases. Rhythms with irregular R-R intervals (sinus rhythm with premature atrial contractions and atrial tachycardia or flutter with variable atrioventricular conduction) were misdiagnosed as AF. The presence of low-amplitude atrial activity and/or baseline artifact significantly increased the likelihood of the erroneous diagnosis, whereas ventricular rates of 130 beats/min did not influence the rate of error. Type II errors (missed AF) occurred in 82 cases where AF was either missed or confused with atrial tachycardia/flutter. Finally, ventricular pacing significantly increased the likelihood of type II errors. Conclusions: In our institution, about 900 ECGs are read each week and 5 of them carry a wrong interpretation related to AF. More attention to common sources of errors as reinforced by an ongoing quality improvement program may reduce the rate of mistakes and thus prevent serious consequences. D 2007 Elsevier Inc. All rights reserved.
Keywords:
Atrial fibrillation; Electrocardiographic errors; Arrhythmias
Introduction Atrial fibrillation is recognized as one of the major challenges in modern cardiology both in terms of the treatment of arrhythmia and prevention of stroke. Efforts have been made to define the best therapeutic approaches such as rate vs rhythm control or pharmacologic vs nonpharmacologic methods. 1-3 A few studies have addressed the misinterpretations of the 12-lead ECG4-6 and the potential consequences of diagnostic errors related
Abbreviations: AF, Atrial fibrillation; Afl/tach, Atrial flutter or atrial tachycardia; SR/PACs, Sinus rhythm and premature atrial contractions. 4 Corresponding author. 1000 East Genesee Street, Syracuse, NY 13210, USA. Tel.: +1 315 471 1044; fax: +1 315 474 4312. E-mail address: [email protected] 0022-0736/$ – see front matter D 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jelectrocard.2007.01.003
to AF.4,7,8 A recent article by Anh et al7 has determined that familiarity with the patient may improve the accuracy of the ECG interpretation. In the current study, we analyzed the influence of some ECG features as possible reasons for the erroneous interpretation of AF by physicians. The gold standard for the final diagnosis was based on the consensus by cardiologist readers. Methods A total of 35 508 consecutive ECGs were overread during the months of March through December 2005 at St. Joseph’s Hospital Health Center (Syracuse, NY). The study was approved by the hospital’s institutional review board. The ECGs were obtained with Marquette machines and the interpretation software used was MAC 5000 6A, last
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
upgraded on May 16, 2004. The tracings were displayed to show 2.5 seconds of each of the 12 leads and 10-second simultaneous strips of leads V1, lead II, and V5. The simultaneous recordings of all 12 leads were available upon the request of the reader. The ECGs were presented to the primary readers with the computer-generated report (GE Marquette Medical Systems 12SL MAC Rhythm, Milwaukee, WI, USA). The reports were modified and confirmed by 1 of the 54 primary readers of the hospital. In our hospital, ECG readers are board certified in internal medicine; 35 of them also have a certification in cardiovascular diseases and 2 of them are board-certified electrophysiologists. A copy of the confirmed ECG was delivered to the patient’s chart. A second copy was saved for analysis. The initial screening of all confirmed copies and the separation by diagnosis was done with the cooperation of a highly trained nurse. The review process was performed on a daily basis. The incorrect interpretations were corrected and returned to the patient’s chart within a 24-hour period from the initial reading. This effort was aimed at minimizing any potential consequences derived from the diagnostic errors. In addition, at the end of each month all corrected ECGs were sent back to the primary reader as part of a quality improvement program. Finally, all corrected ECGs were discussed during the monthly ECG conference. If— after a careful review and a discussion with the primary reader, as well as later with the rest of the physicians during the ECG conference—there was still uncertainty regarding the diagnosis, then those ECGs were excluded from the study. Electrocardiograms were separated into 2 groups: those showing spontaneous QRS complexes and those in which 100% of the QRS complexes were paced. The groups were subdivided according to the actual supraventricular rhythm. Two types of errors were analyzed: type I or overdiagnosis of AF, and type II or missing the diagnosis of AF. Definitions Atrial fibrillation was defined as a supraventricular tachyarrhythmia characterized by the replacement of consistent P waves by rapid oscillations or fibrillatory waves (ie, f waves) that vary in size, shape, and timing, associated with an irregular, frequently rapid ventricular response when the atrioventricular (AV) conduction is intact.9 Regular R-R intervals, however, are possible in the presence of AV block with ventricular or junctional rhythms and ventricular pacing. High-amplitude AF was defined as AF in which the amplitude of the f waves is more than 0.2 mV in at least 1 lead. High ventricular rate was defined as a ventricular rate of 130 beats per minute (beats/min) or higher. Atrial flutter and atrial tachycardia were grouped together and were defined as distinct atrial activity (P waves or flutter waves) observed in 1 or more leads, with regular intervals, and with a rate of 100 to 350 beats/min. Statistics Frequencies of correctly and incorrectly diagnosed ECGs were compared by using a v 2 test with 1 df. Detection accuracy of AF in cases with and without 100% ventricular
451
pacing was measured with signal detection theory using the discriminability (dV) statistic.10 A dV value exceeding 1 indicates that true cases can be reliably discriminated from false cases; a dV of more than 3 indicates excellent discriminability. Results Of a total of 35 508 ECGs reviewed, there were 2809 cases (7.9%) of true AF. A total of 219 ECG carried a wrong diagnostic interpretation related to AF, representing an error rate of 7.8%. Errors in the diagnosis of AF were of 2 types: type I when the diagnosis of AF was made, but other rhythms were actually present (typically rhythms with irregular R-R intervals); and type II when the diagnosis of other rhythms was made, but AF actually was present (typically AF with high-amplitude f waves or with regular R-R intervals). To facilitate the description of the results, ECG with and without 100% ventricular pacing were analyzed separately. Electrocardiograms without 100% ventricular pacing Atrial fibrillation was found in 2402 of 34 008 ECGs, representing an incidence of 7.1%. Type I error (overdiagnosis of AF) Atrial fibrillation was wrongly diagnosed in 108 cases, representing 4.5% of all true cases of AF. Rhythms confused with AF were sinus rhythm with premature atrial contractions (SR/PACs, 42 cases, see Table 1 and Fig. 1A and B) and atrial flutter or atrial tachycardia with variable AV block (Afl/tach, 66 cases, see Table 2 and Fig. 1 C and D). Comparisons were made between features associated with correct and incorrect diagnoses in SR/PACs and Afl/tach cases. Table 1 shows that, in cases of SR/PACs, low-amplitude atrial activity and baseline artifact were more prevalent in cases with overdiagnosis of AF (47.6% and 31.0%, respectively) than in cases with correct diagnosis of SR/PACs (15.7% and 7.1%, respectively; P b .001 for both comparisons using v 2 test with 1 df). Table 2 shows that, in cases of Afl/tach with variable block, frequencies of low-amplitude atrial activity, baseline artifact, and ventricular rates of more than 130 beats/min were not statistically different between cases with overdiagnosis of AF and cases with correct diagnosis of Afl/tach, although there was nonsignificant trend toward higher frequencies of low-amplitude atrial activity and baseline artifact in cases of overdiagnosis of AF. Fig. 1 shows examples of type I errors in the absence of 100% ventricular pacing. Panel A shows SR with sinus Table 1 Type I error: cases of SR/PAC in the absence of 100% ventricular pacing
Low-amplitude atrial activity Baseline artifact All of the above None of the above Total
True diagnosis of SR/PAC
False diagnosis of AF
259 (15.7%) 117 (7.1%) 10 (0.6%) 1261 (76.6%) 1647
204 (47.6%) 13* (31.0%) 3* (7.1%) 6* (14.3%) 42
4 Significantly different at P b.001 using v 2 with df = 1.
452
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
Fig. 1. (A) Sinus rhythm with sinus arrhythmia and PACs giving rise to an irregular rhythm initially diagnosed as AF. Low-amplitude P waves can be seen preceding each QRS. (B) Sinus rhythm and PACs with baseline artifact initially diagnosed as AF. (C) Rhythm strip obtained in lead VI, lead II, and V5 showing a supraventricular tachycardia with grouping of the QRS complexes that is perfectly repetitive as shown by the brackets. There is a stable 8:3 ratio. (D) Irregularly irregular narrow complex tachycardia initially diagnosed as AF. When carefully measured, the atrial waves march perfectly (as indicated by the asterisk).
arrhythmia and PACs giving rise to irregularly irregular R-R intervals. The P waves show very low amplitude. This led to the wrong diagnosis of AF. Panel B shows SR and PACs with baseline artifact. The baseline artifact interferes with visualization of the P waves and gives the wrong impression of f waves. Panel C shows Afl/tach with variable AV block. Atrial waves show low voltage and are not clearly visualized in all leads (arrows). The R-R intervals are indeed irregular. A more careful analysis of the tracing, however, reveals a repetitive pattern in the R-R intervals (see the perfect grouping of the QRS complexes as marked by the brackets). The AV conduction ratio in this case is 8:3, and results from alternate Wenckebach block at the AV nodal level. This repetitive pattern reflects the presence of a perfectly repetitive supraventricular source suggesting the diagnosis of atrial flutter even if the flutter waves are not well visualized. Panel D shows another example of Afl/tach with variable block. In this case, there seems to be no pattern in the irregularity of the R-R interval. The
differential diagnosis with AF is based only on the finding of the repetitive atrial activity as indicated by the asterisks. Type II errors (missed AF) Table 3 shows that AF was missed in 57 cases (2.4%). These ECGs were compared with those in which the diagnosis of AF was correct. High-amplitude AF was more Table 2 Type I error: cases of atrial flutter/atrial tachycardia with variable AV block in the absence of 100% ventricular pacing
Low-amplitude atrial activity Baseline artifact Ventricular rate, 130 beats/min More than 1 of the above None of the above Total
True diagnosis of Afl/tach
False diagnosis of AF
279 47 121 17 177 641
32 8 12 54 94 66
(43.5%) (7.3%) (18.9%) (2.7%) (27.6%)
(48.5%) (12.1%) (18.2%) (7.6%) (13.6%)
4 Significantly different with P b.05 using v 2 with df = 1.
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456 Table 3 Type II error: cases of AF in the absence of 100% ventricular pacing
High-amplitude AF Regular Ventricular Activity Ventricular rate, 130 beats/min More than 1 of the above None of the above Total
True diagnosis of AF
Missed diagnosis of AF
309 49 290 34 1663 2345
274 444 9 2 15* 57
(13.2%) (2.1%) (12.4%) (1.4%) (70.9%)
Table 4 Summary of AF-related diagnoses in cases without 100% ventricular pacing
(47.4%) (15.8%) (3.5%) (26.3%)
4 Significantly different at P b .001. 44 Significantly different at P b .05 using v 2 with df = 1.
prevalent in the group with incorrect diagnosis (47.4% vs 13.2%, respectively; P b .001). Regular ventricular rate observed in cases of relatively high degree of AV block and a competing junctional rhythm also was more prevalent in cases in which the diagnosis of AF was missed (7.0% vs 2.1%, respectively; P b .05). Finally, frequencies of ventricular rates of more than 130 beats/min in the 2 groups were not statistically different. Fig. 2 shows examples of type II errors (missed AF). Panel A shows a typical case of coarse AF. A careful measurement of the baseline waves (shown by the brackets) highlights the irregularity of the atrial activity characteristic of AF. Panel B shows a case of AF with regular R-R interval. The baseline shows erratic undulations; yet, the diagnosis of AF was missed. Panel C shows a recording of the same patient a few hours later. Here, the junctional rhythm is interrupted by sporadic ventricular paced com-
453
True AF Other rhythms Total
Diagnosed as AF
Diagnosed as other rhythms
Total
2345 108 2453
57 2288 2345
2402 2396 4798
Percentage of missed AF, 2.3%; percentage of overdiagnosed AF, 4.5%; dV = 3.677.
plexes, and the R-R interval shows irregularity. This time, the diagnosis of AF was correctly identified. In summary, as shown in Table 4, the overall rate of overdiagnosis of AF (type I error) was 2.4%, and the rate of missed AF (type II error) was 4.5%. Electrocardiograms with 100% ventricular pacing Atrial fibrillation was found in 407 of 1500 ECGs with 100% ventricular pacing, representing an incidence of 27.1% (nearly 4 times the incidence of AF in the whole population). As shown in Table 5, the presence of 100% ventricular pacing was associated with a 6.1% type I error rate (missed AF) and a 3.2% type II error rate (overdiagnosis of AF). Type I error (overdiagnosis of AF) Atrial fibrillation was incorrectly diagnosed in 29 cases. Rhythms confused with AF were Afl/tach (18 cases) and SR or retrograde P waves (11 cases). In general, the atrial
Fig. 2. (A) Tracing showing V1 originally diagnosed as atrial flutter with variable AV conduction. Yet a more careful analysis shows that the atrial activity is not regular (see the brackets). (B) Recording obtained at 4:56 pm with an initial report indicating the presence of junctional tachycardia and nonspecific ST and T wave abnormalities. There was no mention about the atrial activity. (C) Recording from the same patient in (B) obtained at 5:30 pm showing btypicalQ AF pattern with varying R-R intervals (as well as sporadic ventricular pacing). Note that in both (B) and (C), the baseline shows the same type of f waves.
454
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
Table 5 Summary of AF-related diagnoses in cases with 100% ventricular pacing
True AF Other rhythms Total
Diagnosed as AF
Diagnosed as other rhythms
Total
382 29 411
25 889 914
407 918 1325
Percentage of missed AF, 6.1%; percentage of overdiagnosed, AF: 3.2%; dV = 3.401.
activity had low amplitude or was obscured by the ventricular activity. Fig. 3 shows typical examples with ventricular pacing in which the underlying rhythm was wrongly interpreted as AF. Panel A shows Afl/tach. Regular negative flutter waves (asterisks) are seen in only segments of the cardiac cycle, but can be tracked throughout the tracing (arrows) with a constant cycle length. Panel B shows isolated P waves with low amplitude. Yet, they show a constant cycle length. Because the sinus rate is very slow
and there is no synchronization with the ventricular pacing (AV dissociation), the reader assumed an absence of P waves and made the incorrect diagnosis of AF. Panel C shows P waves that are difficult to see because they occur just after the QRS (ie, retrograde P waves). Here, again, the primary reader overlooked the P waves and made the wrong diagnosis of AF. Finally, panel D shows a case in which the P waves are nearly unrecognizable. The interesting feature of this ECG is the presence of premature ventricular stimuli followed by narrow QRS complexes (first, fifth, and ninth complexes) that represent fusion between paced and native QRS. This can be explained by the presence of a hidden P wave that is both tracked by the generator and conducted to the ventricles. A more careful look at the tracing reveals the presence of a very low voltage deflection at the end of the T wave preceding the premature stimulus. Also, there is a consistent low-amplitude deflection preceding the ventricular stimulus after the premature beats (second, sixth, and 10th complexes). If we divide the
Fig. 3. Four examples showing 100% ventricular pacing. (A) Initially reported as ventricular pacing with undetermined underlying supraventricular rhythm, likely AF. A more careful analysis shows the presence of very rhythmic atrial activity (asterisks), some of which are concealed within the QRS. The correct diagnosis is atrial tachycardia with no evidence of AV conduction and nonsynchronized ventricular pacing. (B) The initial diagnosis was absence of P waves, AF. P waves are indeed present (arrows in V1) at a slow and very regular rate. The correct diagnosis is sinus bradycardia with high degree AV block and nonsynchronized ventricular pacing. (C) Again, the diagnosis was underlying AF. P waves, however, are retrograde. They can be seen after each QRS (arrows) with a fixed RP interval. (D) Atrial activity is not obvious so the initial report suggested the diagnosis of AF. There are 3 premature ventricular stimuli. The inset in lead V1 shows a very low-amplitude P wave occurring at the end of the T wave that precedes the premature stimuli. In addition, there is a bpseudo r waveQ preceding the next QRS in all 3 instances. If we assume that this is indeed the cycle of the atrial activity, we find that there is a very regular sinus bradycardia (asterisks and arrows) underlying ventricular pacing. The correct diagnosis is sinus bradycardia with preserved AV conduction (the premature QRS is a fusion between paced and native activation) and ventricular pacing. There seems to be appropriate atrial sensing, but there is no evidence of atrial pacing.
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
interval between these deflections in 2, we notice that there may be indeed a regular sinus rhythm underlying ventricular pacing, thus ruling out AF. Type II errors (missing AF) Atrial fibrillation was missed in 25 cases with 100% ventricular pacing. In most cases, there was no actual mention of the underlying rhythm. In 2 cases, AF had relatively high amplitude waves and was confused with atrial flutter. Computer interpretation Although the scope of the present study was to determine the accuracy of physician-generated ECG reports, we also assessed the rate of accuracy of the computer generated report. From 3596 cases read in 1 month, the computer overdiagnosed AF in 30. Sixteen of these cases were indeed sinus rhythm with PACs, and 14 were cases Afl/tach. Of the 250 cases of true AF, the computer report missed the diagnosis in 44 cases (17.6%). Eighteen of 44 cases showed 100% ventricular pacing and the computer offered no comments on the underlying rhythm. Other diagnosis generated by the computer were: undetermined rhythm (18 cases), sinus tachycardia (7 cases), atrial flutter (3 cases), sinus rhythm (3 cases), atrial rhythm (1 case), junctional rhythm (1 case), and wide complex rhythm (1 case). It is interesting to note that although the rate of inaccuracy of the computer reports was relatively high, it did not seem to significantly influence the reports generated by the physicians. In fact, most of the ECGs were modified by the primary reader and in some cases the reader wrongly overruled the computer diagnosis. Discussion Both confirming and discarding the diagnosis of AF may greatly influence the management of a patient. The 12-lead ECG remains the most common tool in the diagnosis of AF. The present study shows that over a period of 10 months there were 219 incorrect diagnoses related to AF, representing 5 errors per week. The main objective of the study was to determine the most likely reasons for the erroneous diagnoses. Type I error Irregularly irregular ventricular rhythm The definition of AF has been classically associated with the words birregularly irregularQ rhythm. Although irregularly irregular ventricular rhythm is indeed a very common feature of AF, it is neither exclusive of AF nor necessary for its diagnosis. Sinus rhythm with PACs and or sinus arrhythmia may lead to a very irregular rhythm. If the P waves are not well visualized because of a low-amplitude and/or the presence of artifact, there is high likelihood of an erroneous diagnosis of AF. When a regular atrial rhythm (ie, atrial flutter) drives the ventricles, the ventricular rhythm may either be regular or irregular depending on the rate of
455
the atrial rhythm and the electrophysiologic properties of the AV node. Wenckebach type of conduction with varying periodicities may result in organized irregularity (ie, patterns such as those shown in Fig. 1C) or a totally irregular rhythm (Fig. 1D) thus mimicking AF. Type II error Regular R-R interval ventricular rhythms It is difficult to associate AF with a regular ventricular rhythm. The presence of a high-degree AV block with or without ventricular pacing may set the basis for this combination. Our study shows that the presence of regular ventricular rhythm is very likely to be associated with a missed diagnosis of the underlying AF. High-amplitude AF (bpseudoQ atrial flutter) Atrial fibrillation is disorganized and unpredictable. On the other hand, atrial flutter and other forms of atrial tachycardia in general show a fixed interval with little or no variations over the 10-second period of the ECG recording.11 The random nature of activation during AF makes it unlikely for operators to find identical cycle lengths in the 10-second tracing of the ECG. Random activation during AF (f waves) may still appear as bPQ waves separated by relatively irregular intervals of nearly absence activity. It is necessary to determine if the morphology of the ballegedQ P is consistent in leads obtained simultaneously as one may expect in cases of real P waves or PACs originated in a single focus. The distinction between AF and some forms of multifocal atrial tachycardia may be very difficult. The search for reasonable association between the P waves and the QRS complexes may help distinguish between these 2 arrhythmias. In AF, higher amplitude waves (simulating P waves) are not necessarily more likely to conduct through the AV node and thus be followed by QRS complexes. On the other hand, in multifocal atrial tachycardia, QRS are usually preceded by a discernible P wave. High ventricular rates ( N130 beats/min) Although rapid ventricular rate was found both in cases of Afl/tach and AF, it did not increase the rates of either type I or type II error. Ventricular pacing In our study, cases with 100% ventricular pacing were associated with an incidence of AF that was 4 times higher than in cases with an absence of 100% ventricular pacing. Importantly, the presence of 100% ventricular pacing significantly increased the risk of missing AF. A signal detection theory analysis revealed a significantly higher dVfor these cases (dV = 3.40 vs dV = 3.68; P b .05). Thus, patients with pacemakers seem to be at higher risk of having undiagnosed AF and slightly lower risk of having overdiagnosis of AF. In many cases, missing the diagnosis of AF was the result of omitting the description of the supraventricular rhythm in the presence of ventricular pacing. On the other hand, type I error occurred when the P waves were not
456
J.M. Davidenko, L.S. Snyder / Journal of Electrocardiology 40 (2007) 450 – 456
obvious. The apparent absence of P waves led to the wrong assumption of the presence of AF. A careful search for hidden P waves may help avoid these mistakes. In other cases, the presence of irregular ventricular pacing should alert the reader to the possibility of unrecognized organized atrial activity that has been tracked by the generator. In some cases, the interrogation of the pacemaker (ie, lowering the pacing rate or acquiring the intracardiac electrograms) may be necessary to rule out the presence of AF. Computer interpretations The scope of the current study was to analyze the reading abilities of the physician. Yet, we observed that our computer program has a relatively high inaccuracy rate (from a total of 250 true AF, the computer missed 44 [18%]). The most common reason for this type of mistake was the presence of 100% ventricular pacing in which the computer made no comments on the underlying rhythm. In addition, of 3597 reports the computer overdiagnosed AF in 30 (0.8%). Similar to what was observed with the physicians, the most common rhythms confused with AF were sinus rhythm with PACs and atrial tachycardia/atrial flutter with variable AV conduction. In spite of the similarities between computer and human sources for error, the computer was not found to be the reason for human mistakes because most of the confirmed reports in our series included corrections of the original computer reading. Limitations The study was based on consecutive ECGs read. Thus, the incidence of AF may not represent the actual incidence of the arrhythmia in our institution, because several tracings may have been obtained from the same patient. The main goal of the study, however, was to determine the rate of error and how to reduce it. Because in some cases ECGs from the same patient with the same rhythm had different interpretations, we decided not to exclude them from the study. There was no attempt at determining if the initial error in the ECG interpretation led to changes in the care of the patient. The corrections were done and delivered within a 24-hour period, thus reducing the chances of complications derived from the initial erroneous ECG interpretation. Although the number of ECG readers was relatively high (54), the results of this study may only reflect the quality of reading in our hospital. Until similar quality assessment studies are performed in other centers, we will not know if our results reflect a more general problem. Implications This study conducted in a midsize community hospital indicates that the ECG diagnosis of AF is not accurate. About 5 ECGs per week shows are misdiagnosed as AF.
Type I errors (overdiagnosis AF) may lead to the unnecessary use of anticoagulation or antiarrhythmic drugs when the real rhythm is simply sinus rhythm and PACs. When the real rhythm is atrial flutter or atrial tachycardia, the diagnostic error use may result in lack of appropriate use of well-established and readily available ablation techniques. Type II errors (missed AF) may result in inadequate measures for the prevention of stroke. Because the 12-lead ECG is still the main diagnostic tool of AF, efforts should be placed in improving the quality of ECG reading. In our institution, a quality improvement program is in place by which corrected ECGs are sent to the primary readers on a monthly basis for review and discussion. In addition, misinterpreted ECGs are presented during the monthly ECG conference. Future studies may determine the usefulness of such a program in reducing diagnostic errors. Acknowledgments The authors thank Dr. Nicolas Davidenko, who performed the statistical analysis and contributed with important ideas for the design of the study, and Dr. Dennis Ehrich, Vice President of Medical Affairs, St Joseph Hospital, for financial support and encouragement. References 1. Wyse DG, Waldo AL, Di Marco JP. The Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) Investigators. A comparison of rate control and rhythm control in patients with atrial fibrillation. N Engl J Med 2002;347:1825. 2. Van Gelder IC, Hagens VE, Bosker HA, et al. A comparison of rate control and rhythm control in patients with recurrent persistent atrial fibrillation. N Engl J Med 2002;347:1834. 3. Haissaguerre M, Jais P, Shah DC, et al. Spontaneous initiation of atrial fibrillation by ectopic beats originating in the pulmonary veins. N Engl J Med 1998;339:695, 1997;20:337 4. Bogun F, Anh D, Kalahasty G, et al. Misdiagnosis of atrial fibrillation and its clinical consequences. Am J Med 2004;117:706. 5. Knight BP, Michaud GF, Strickberger SA, Morady F. Electrocardiographic differentiation of atrial flutter from atrial fibrillation by physicians. J Electrocardiol 1999;32:315. 6. Poon K, Okin PM, Kligfield P. Diagnostic performance of a computer based ECG rhythm algorithm. J Electrocardiol 2005;38:235. 7. Anh D, Krishnan S, Bogun F. Accuracy of electrocardiogram interpretation by cardiologists in the setting of incorrect computer analysis. J Electrocardiol 2006;39:343. 8. Finsterer J, Stollberger C, Gatterer E. Oral anticoagulation for ECG tremor artifact simulating atrial fibrillation. Acta Cardiol 2003;58:425. 9. Bellet S. Clinical disorders of the heart beat. 3rd ed. Philadelphia7 Lea & Febiger; 1971. 10. Dorfman DD, Alf E. Maximum-likelihood estimation of parameters of signal-detection theory and determination of confidence intervals— rating method data. J Math Psychol 1969;6:487. 11. Prystowsky EN, Katz AM. Atrial fibrillation. In: Topol ES, editor. Textbook of cardiovascular medicine. Philadelphia7 Lippincott-Raven; 1998. p. 1827.