Family Practice Advance Access originally published online on November 4, 2004
Family Practice 2005 22(1):109-113; doi:10.1093/fampra/cmh601
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Electrocardiogram interpretation in general practice
a Department and Research Unit of General Practice, and b Department of General Practice, University of Aarhus and c Department of Cardiology, Skejby University Hospital, Aarhus C, Denmark
Email: mj{at}alm.au.dk
Received 27 April 2004; Accepted 16 June 2004.
Jensen MSA, Thomsen JL, Jensen SE, Lauritzen T and Engberg M. Electrocardiogram interpretation in general practice. Family Practice 2005; 22: 109–113.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Background. The 12-lead electrocardiogram (ECG) is a common diagnostic test available to the GP in the evaluation of patients with cardiac complaints. In daily clinical practice it is important for GPs to know the sensitivity and specificity of their ECG interpretation skills.
Objectives. The purpose of the present study was to evaluate the ECG interpretation skills of GPs and the value of automatic ECG recorder interpretations in general practice.
Methods. A total of 902 ECGs were recorded in a random sample of the population aged 31–51 years in the district of Ebeltoft, Denmark, from December 1991 to June 1992. They were interpreted automatically by an interpretive ECG recorder and by the GPs in the clinic in Ebeltoft, with a cardiologists interpretation as a gold standard. Sensitivity, specificity and predictive values of diagnoses were calculated.
Results. Overall, the sensitivity of abnormal diagnoses made by the GPs (69.8%) was significantly lower (P < 0.001) than that of diagnoses made by the interpretive ECG recorder (84.4%). The overall specificity of abnormal diagnoses made by the GP (85.7%) was significantly higher (P < 0.001) than that achieved by the interpretive ECG recorder (75.6%).
Conclusions. GPs in this study were good at correcting false-positive diagnoses made by the interpretive ECG recorder. In order to avoid unfortunate reclassifications of true-positive to false-negative diagnoses, GPs are recommended to pay special attention to the diagnoses of ST-segment deviation, T-wave inversion or the presence of Q-waves made by interpretive ECG recorders, when ECGs are used in individual risk assessment.
Keywords. Cardiology, clinical diagnosis, predictive value of tests, primary health care, sensitivity and specificity.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
The 12-lead electrocardiogram (ECG) is a common diagnostic test available to the GP in the evaluation of patients with cardiac complaints.1 In daily clinical practice, it is important for GPs to know the sensitivity and specificity of their ECG interpretation skills in order to correctly exclude the presence of heart disease, treat their patients or refer them to a cardiologist.2 Only a few studies have been performed to evaluate the ECG interpretation skills of GPs and family practice residents.1,3–5 The studies were performed using a limited number of selected ECGs1,3,5 or evaluating the use of the ECG to select patients with suspected heart failure for referral to echocardiography.4 From three of these studies,1,3,5 it seems that GPs and family practice residents have difficulties interpreting ECGs compared with reference diagnoses made by cardiologists. The fourth study4, however, found a high degree of concordance between two GPs and a cardiologist on the overall interpretation of the ECGs of patients with suspected heart failure.
The validity of ECG interpretations made by GPs in health screenings of the middle-aged population has not been investigated. It has, however, been suggested that more correct interpretations of ECGs may be achieved by using interpretive ECG recorders in general practice.3,6 In this context, it is important to know if reliable diagnoses are actually given by interpretive ECG recorders in general practice. Therefore, the aim of the present study is to determine the sensitivity and specificity of ECG interpretation by both GPs and interpretive ECG recorders in the setting of a randomly selected middle-aged population invited to a health screening.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
The Ebeltoft Health Promotion Project is a randomized, controlled, 5 year population-based follow-up study conducted during the years 1991–1997 aiming to study the impact of general health screening and health discussion in general practice. The study population was a random sample from the 3464 subjects aged 30–49 years (1 January 1991) who were registered with a GP and resided in the district of Ebeltoft. The sample was largely representative of the Danish population.7 A random sample of 2000 of the 3464 subjects was invited to enter the study and, of these, 1507 subjects agreed to participate. The 1507 subjects were randomized to three study groups: two different intervention groups and a control group. A total of 1006 subjects were randomized to the two intervention groups, and were offered a general health screening including ECG recording. The remaining 501 subjects were randomized to the control group and did not receive a health screening, and therefore had no ECG recorded. Of the 1006 subjects offered a health check, 905 accepted and had an ECG recorded.8 The baseline general health screenings were carried out from December 1991 to June 1992. The design and methodology of the Ebeltoft project have been described in detail elsewhere.8
All ECGs are standard 12-lead ECGs recorded with participants in a supine position using either a Schiller Cardiovit AT-3/1 or a Schiller Cardiovit AT-1 ECG recorder. The recorder was set at a paper speed of 25 mm/s and a standard sensitivity of 10 mm/mV. The ECGs were interpreted by the GPs in the clinic in Ebeltoft at the baseline general health screening. GPs were not blinded to the interpretation of the interpretive ECG recorder and had access to both patient history and other clinical data. All ECGs were re-evaluated in the period from December 2000 to June 2001 by a cardiologist using a set of clinical diagnoses.9 It was not possible to blind the cardiologist, as the interpretations made by the interpretive ECG recorder were printed on the ECG, but the cardiologist had no access to patient history or other clinical data. The cardiologists' electrocardiographic diagnoses were used as a reference interpretation (gold standard) for the determination of sensitivity, specificity and predictive values of diagnoses made by GPs and the interpretive ECG recorder. If an ECG was given more than one diagnosis by the cardiologist, the clinically more important diagnosis was selected and chosen for the analysis of data, i.e. one ECG can only contribute one diagnosis in the analysis.
A 10% random sample of the ECGs was evaluated by another cardiologist for the purpose of determining the interobserver agreement on the diagnoses used as the reference.
Ethics
Permission to conduct the study was given by the Scientific Ethical Committee of Aarhus County (1990/1996) and the Danish Data Protection Agency (2001-41-0738).
Statistics
Data were entered in SPSS Data Entry Station® v.1.0 and analysed using SPSS® v.10.0. Double data entry was performed.
Sensitivity, specificity, predictive value of a positive test and predictive value of a negative test of diagnoses were calculated for both GPs and the interpretive ECG recorder. For these calculations, the clinical diagnoses9 made by the cardiologist were used as reference. For the predictive value of a positive test and the predictive value of a negative test, 95% confidence intervals (CIs) were computed using an electronic statistical table of the binomial distribution.
Sensitivity and specificity of diagnoses made by the GPs and the interpretive ECG recorder were treated as paired data, and comparisons of these proportions were made using McNemar's test. Significance for all measures is considered at the 5% level.
Kappa statistics10 were calculated to evaluate interobserver agreement on the diagnoses made by the two cardiologists in a 10% random sample of ECGs.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
A total of 905 ECGs were available. Of the 905 ECGs, three had to be discarded: two because of inadequate identification and one because the ECG had been destroyed by accident. The remaining 902 ECGs, representing 89.7% of participants randomized to general health screenings, originated from 429 men (47.6%) and 473 women (52.4%). Median age was 41.0 years for both men and women.
The evaluations of the two cardiologists showed an interobserver agreement of 97.0%. Kappa statistics were calculated with kappa = 0.856 (95% CI 0.742– 0.970), which is defined as a very good agreement.
In total, 96 ECGs were classified as abnormal by the cardiologist, whereas 182 and 278 were classified as abnormal by the GPs and the interpretive ECG recorder, respectively. The cardiologist found 31 ECGs to have signs of ischaemia or prior myocardial infarction. The corresponding number of electrocardiographic diagnoses made by the GPs was 58, whereas the interpretive ECG recorder made 155 diagnoses of signs of ischaemia or prior myocardial infarction.
The sensitivity, specificity and predictive values are defined in Table 1.
|
Overall, the sensitivity of abnormal diagnoses made by the GPs (69.8%) was significantly lower (P < 0.001) than that of the interpretive ECG recorder (84.4%). The overall specificity of abnormal diagnoses made by the GPs (85.7%) was significantly higher (P < 0.001) than that achieved by the interpretive ECG recorder (75.6%) (Table 2).
|
The predictive values of a positive test and of a negative test were not different for the GPs and the interpretive ECG recorder for any of the diagnoses reported (Table 2).
Considering prior myocardial infarction or ischaemia, sensitivities were significantly higher (P < 0.001) for the interpretive ECG recorder than for the GPs, whereas the specificities of the interpretive ECG recorder on these diagnoses were significantly lower (P < 0.001) than for the GPs. Sensitivity and specificity of diagnoses of any bundle branch blocks and complete bundle branch blocks were not significantly different (P-values ranging from 0.50 to 1.00) for the GPs and the interpretive ECG recorder.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
The higher sensitivity of diagnoses made by the interpretive ECG recorder compared with that achieved by the GPs was expected. The level of sensitivity is, however, less than desirable for both the interpretive ECG recorder and, in particular, the GPs. That GPs achieved a higher specificity of diagnoses than the interpretive ECG recorder was also expected, and the level of specificity achieved was generally acceptable. A high level of sensitivity of an interpretive ECG recorder in clinical practice is very important to keep the false-negative rate as low as possible. It is then left to the GP to attempt to achieve a higher specificity by identifying the false–positives. Sensitivities of diagnoses made by the interpretive ECG recorder and the GPs were especially low for electrocardiographic signs of ischaemia or of prior myocardial infarction, whereas the sensitivities of diagnoses of electrocardiographic signs of bundle branch blocks were much higher and reached acceptable levels.
The low predictive value of a positive test, and high predictive value of a negative test observed in the present study may be explained by the levels of sensitivity and specificity reached by the interpretive ECG recorder and the GPs, but also by the low prevalence of abnormal electrocardiographic findings in the study population.9 For a test with a given sensitivity and specificity, the predictive value of a positive test decreases with the prevalence of disease,11 i.e. a positive test will on more occasions be false–positive when the prevalence of disease is very low.
It is important to stress that the topic of the present study is the presence or absence of abnormal electrocardiographic findings and not the actual presence or absence of heart disease. Furthermore, failure to interpret an ECG correctly may not always be a hazard to the patient's health, as White et al. found a misinterpretation rate of 33% to result in serious management error in only 3.3% of cases.12 This is especially interesting when confronted with the apparent lower sensitivity and higher specificity of diagnoses made by the GPs compared with the sensitivity and specificity of diagnoses achieved by the interpretive ECG recorder. As mentioned above, the GPs, in contrast to the cardiologists, had access to patient history and other clinical data. As a consequence, the GPs might not have taken notice of certain electrocardiographic findings, for instance electro-cardiographic changes due to ischaemia, when these findings occurred in a patient otherwise free of symptoms and risk factors for coronary heart disease. It has, however, been reported that the absolute risk of coronary heart disease associated with serious electrocardiographic findings in subjects without symptoms of coronary heart disease is almost the same as that of subjects with symptoms of coronary heart disease and a normal electrocardiogram.13 This absolute risk is, however, two to three times the risk of subjects with normal ECGs and without symptoms of coronary heart disease, but only half that of subjects with symptoms of coronary heart disease and serious electrocardiographic findings.13 Therefore, a decision to ignore a serious electrocardiographic finding is unfortunate, even if the patient lacks symptoms of coronary heart disease.
Lack of blinding with respect to the diagnoses made by GPs and the ECG recorder introduces the possibility that the cardiologist, knowingly or not, considers diagnoses made by the GPs and the interpretive ECG recorder in a different way and thus biases the comparison between the two in favour of one or the other.
Whether the GPs in the present study are representative of the average GP in Denmark can be discussed. We have no reason to believe, however, that their ECG interpretation skills are worse or better than that of the average GP in Denmark.
A strength of the present study is that it is based on secondary data. Thus, GPs never knew that they were going to be evaluated on their ECG interpretation skills. Therefore, the ECG interpretations made by the GPs represent their everyday skills when an interpretive ECG recorder is used for screening purposes in general practice.
GPs in this study are very good at correcting the false–positive diagnoses made by the interpretive ECG recorder; however, the gain in specificity is paid for with a loss of sensitivity. Therefore, GPs should be careful and thorough when deciding to over-ride a diagnosis made by an interpretive ECG recorder. For some diagnoses made by the interpretive ECG recorder, such as signs of ischaemia or of prior myocardial infarction, it may be appropriate for ECGs to be thoroughly checked by an experienced ECG interpreter. This would probably avoid the unfortunate reclassification of true–positives to false–negatives sometimes made by the GPs, explaining some of the low sensitivity of their diagnoses. Relying heavily on the diagnoses of the interpretive ECG recorders alone for referral of patients with suspected ischaemia-like electrocardiographic changes to a cardiologist service would be insufficient. The predictive value of a positive test of the interpretive ECG recorder for these changes is poor, as the sensitivity for these changes is reached through severe overdiagnosing. Thus, the referral rate would become too high, subjecting false–positives to further investigations and possible concomitant psychological strain. This unnecessary high referral rate would also put a strain on the secondary health care system. What is even worse is that the interpretive ECG recorder manages to identify only 64.5% of subjects with electrocardiographic signs of ischaemia or of prior myocardial infarction.
The sensitivity, specificity and predictive values of diagnoses found in the present study may not be generalizable. The present study is, however, the only study performed evaluating ECG interpretation skills of GPs using ECGs obtained from a sample of the general population. Other studies beside the present one have noted the difficulties which GPs have in diagnosing ST- and/or T-wave changes or prior myocardial infarction14 or even acute myocardial infarction,1,3 but only on limited numbers of selected ECGs. Even though the ECG is known to be an imperfect reflection of existing coronary heart disease and a poor predictor of future heart disease,15 it would probably be of value for GPs to devote special attention to the diagnoses of ST-segment deviation, T-wave inversion or the presence of Q-waves made by the interpretive ECG recorder, when the ECG is used in individual risk assessment. When there is any doubt about whether the automated interpretation is correct or not, it is advisable to consult an experienced ECG interpreter before over-riding the automated interpretation.
Since the GPs had access to the diagnoses made by the interpretive ECG recorder, it is, however, not possible to say whether the diagnoses made by the latter should preferably be completely ignored or always taken into consideration when the GP makes his diagnosis. The diagnostic performance of the GPs with and without access to diagnoses made by an interpretive ECG recorder calls for further studies.
|
Declaration |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Funding: Financial support for the investigations in relation to Health Promotion Project Ebeltoft was given by the County Health Insurance Office Aarhus, the Danish College of General Practitioners, the Sara Krabbe scholarship, the Lundbeck scholarship, the Danish Research Foundation for General Practice, the General Practitioners' Education and Development Fund, the Health Promotion Council of Aarhus, the Health Insurance Fund, the Lundbeck's Foundation scientific research grant to GPs, the Ministry of Health Foundation for Research and Development, The Danish Medical Research Council (9801336) and The Danish Heart Foundation (97-2-F-22515).
Ethical approval: was given by the Scientific Ethical Committee of Aarhus County (J. no. 1990/1996).
Conflicts of interest: none declared.
|
Acknowledgments |
|---|
The following GPs participated in the study: A Bøgedal, P Grønbæk, L Jørgensen, PT Jørgensen, H Lundberg, JM Nielsen, GS Pedersen, JC Rahbek and N Bie. We thank the staff at the GPs clinic in Ebeltoft for their extraordinary effort, including the extensive administrative assistance given by A Hilligsøe and E Therkildsen.
Contributors: TL originated and designed the Ebeltoft Health Promotion Project, and was project leader until 1997, where ME became project leader. MSAJ, JLT, SEJ and ME all contributed to the protocol for the investigations. MSAJ adapted electrocardiographic data collected by TL. SEJ interpreted the electrocardiograms. MSAJ conducted the data analyses and MSAJ and JLT drafted the text of the paper with the support of the other authors. MSAJ is the guarantor.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
1 Sur DK, Kaye L, Mikus M, Goad J, Morena A. Accuracy of electrocardiogram reading by family practice residents. Fam Med 2000; 32: 315–319.[Web of Science][Medline]
2 Zwietering P, Knottnerus A, Gorgels T, Rinkens P. Occurrence of arrhythmias in general practice. Scand J Prim Health Care 1996; 14: 244–250.[Web of Science][Medline]
3 Macallan DC, Bell JA, Braddick M, Endersby K, Rizzo-Naudi J. The electrocardiogram in general practice: its use and its interpretation. J R Soc Med 1990; 83: 559–562.[Abstract]
4 Houghton AR, Sparrow NJ, Toms E, Cowley AJ. Should general practitioners use the electrocardiogram to select patients with suspected heart failure for echocardiography? Int J Cardiol 1997; 62: 31–36.[CrossRef][Web of Science][Medline]
5 Margolis S, Reed R. EKG analysis skills of family practice residents in the United Arab Emirates: a comparison with US data. Fam Med 2001; 33: 447–452.[Medline]
6 Khunti K, McKinley RK. Value of ECGs in identifying heart failure due to left ventricular systolic dysfunction. Courses on interpreting ECGs would improve general practitioners' skills. Br Med J 1996; 312: 1161.
7 Lauritzen T, Leboeuf-Yde C, Lunde IM, Nielsen KD. Ebeltoft project: baseline data from a five-year randomized, controlled, prospective health promotion study in a Danish population. Br J Gen Pract 1995; 45: 542–547.[Web of Science][Medline]
8 Engberg M, Christensen B, Karlsmose B, Lous J, Lauritzen T. General health screenings to improve cardiovascular risk profiles: a randomized controlled trial in general practice with 5-year follow-up. J Fam Pract 2002; 51: 546–552.[Web of Science][Medline]
9 Jensen MS, Thomsen JL, Jensen SE, Lauritzen T, Engberg M. Low prevalence of ischemic electrocardiographic findings in a Danish population. Scand Cardiovasc J 2003; 37: 49–57.[CrossRef][Web of Science][Medline]
10 Altman DG. Inter-rater agreement. In Practical Statistics for Medical Research. London: Chapman & Hall; 1991: 403–409.
11 Morrison AS. Screening. In Rothman KJ, Greenland S (eds). Modern Epidemiology. Philadelphia: Lippincott Williams & Wilkins; 1998: 499–518.
12 White T, Woodmansey P, Ferguson DG, Channer KS. Improving the interpretation of electrocardiographs in an accident and emergency department. Postgrad Med J 1995; 71: 132–135.
13 Whincup PH, Wannamethee G, Macfarlane PW, Walker M, Shaper AG. Resting electrocardiogram and risk of coronary heart disease in middle-aged British men. J Cardiovasc Risk 1995; 2: 533–543.[CrossRef][Medline]
14 Rutten FH, Kessels AG, Willems FF, Hoes AW. Electrocardiography in primary care; is it useful? Int J Cardiol 2000; 74: 199–205.[Medline]
15 Sox HC, Garber AM, Littenberg B. The resting electrocardiogram as a screening test. A clinical analysis. Ann Intern Med 1989; 111: 489–502.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||