Key words: ECG interpretation, primary health care, medical skills.

Electrocardiography (ECG) is still regarded as the basic tool in the evaluation of cardiac diseases. It is performed in approximately 2 % of all office visits, and 30 % to 38 % of these ECGs will be abnormal [1]. ECG may be an important tool in primary care and it can considerably reduce the number of unnecessary referrals [2].

However, studies have revealed insufficiencies in the ECG interpretation skills of primary care physicians. In a study of Sur et al. 21% of the US family practice residents could not identify ECG findings of acute myocardial infarction [3]. Margolis et al. obtained similar results from family practice residents in the United Arab Emirates [4]. Although difficulties in ECG interpretation seem to transcend geographical boundaries, it is not clear whether the capabilities of family physicians are lower than those of other specialties.

This study compared ECG interpretation skills of family physicians, untrained physicians, and internists from two different geographical locations in Turkey and investigated the effecting factors.

Setting
The study was conducted between March and June 2002. During that period, it was not necessary to have a special training (postgraduate training) in order to work in primary care; family physicians as well as untrained medical school graduates were working in primary care positions in Turkey; medical students had two months of internal medicine clerkship and 12 months of internship during the six years of undergraduate medical education; family medicine residents were receiving 9 months of internal medicine rotation (this training did not have any cardiology components); and internal medicine residency included 5 months of cardiology rotation. On the other hand, continuous medical education is not an obligation for either untrained physicians or specialists in Turkey.

Sample
The samples in this study consisted of five different groups: family practice specialists (SFM), family practice residents (RFM), internal medicine specialists (SIM), internal medicine residents (RIM), and untrained physicians (UP). Untrained physicians are just graduates from medical schools in Turkey, who do not have any residency education or other vocational training, but still can work in primary care facilities.

Eighty family physicians were selected randomly from the registry of Turkish Association of Family Physicians (total 793 family physicians) and asked to join the study. 59 family physicians (73.8 %) have accepted to join. Of the family physicians, 37 were specialists and 22 residents. Their mean time (mean ± SD) of being in that current position was 2.68 ± 2.07 years (min. 1, max. 10) and 2.05 ± 1.17 years (min. 1, max. 5) respectively.

Nine primary healthcare offices with 38 physicians (all without postgraduate education) are providing primary care to 140.000 inhabitants in Edirne. All untrained physicians working in Edirne, were asked to join the study. Thirty (78.9 %) out of 38 accepted to join the study. The mean time (mean ± SD) of the untrained physicians for being in the current position was 6.10 ± 3.50 years (min 1, max. 12).

All internal medicine residents and specialists from a teaching hospital and a medical faculty were invited to join the study. Out of 68 physicians invited, 51 (70.0 %) have accepted to join. Twenty of the internists were specialists and 31 were residents. Mean time (mean ± SD) of the participants for being in their current positions was 3.29 ± 3.64 years (min. 1, max. 15) and 2.71 ± 1.04 (min 1, max. 4) years respectively. Demographic features of the different groups are presented in Table 1.

Measurement instrument
A measurement instrument was developed similar to that of Margolis et al. [4] consisting of 10 standard ECGs with 12 leads. Each ECG contained one to three clinical diagnoses with a total of 16 diagnoses in the ECG set. Two of the ECGs were normal. The measurement instrument was applied in a comfortable atmosphere without giving any clinical information. Three researchers were trained and used for this purpose. The participants had 20 minutes to complete the instrument. They were not allowed to use any additional tool during this process. Before measuring the ECG interpretation skills, a questionnaire was given to each participant to obtain demographic data such as age, sex, date of receiving medical license, and current position. No other discriminating question was asked to prevent violation of anonymity.

Statistical analysis
One of the researchers was the final rater. He checked the responses of the participants to the ECG sets and scored correct diagnoses (CD) and false diagnoses (FD) for each ECG. By counting the CD and FD of each ECG, correct scores (CS) and false scores (FS) were calculated for each individual. The results were evaluated by a computer using the SPSS package program (SPSS for Windows release 10.0.5, standard version, SPSS, Inc, Chicago, 1989-1999). Comparisons were done with chi-square, Pearson's bivariate correlation, Kruskal-Wallis, one-way ANOVA, and Tukey's post hoc analysis.

There was statistically significant difference between groups with regard to correctly diagnosing right and left bundle branch block, past myocardial infarction, sinus bradycardia, acute myocardial infarction, left ventricle hypertrophy, atrial flutter, first-degree atrioventricular block, and pace maker. Family practice residents and untrained physicians were significantly less successful in correctly identifying acute myocardial infarction when compared with other groups (p=0.003; Table 2).

There was no significant correlation between the work experience (the mean duration of the current position) and mean CS and FS (p>0.05).

Specialists of internal medicine and untrained physicians have received the highest and lowest mean correct scores respectively (9.85 ± 2.06 vs. 4.73 ± 1.84). Mean CS values for family medicine specialists, family practice residents, and internal medicine residents were 7.05 ± 2.30, 6.59 ± 2.46, and 8.16 ± 2.19 respectively. In accordance with this, total response rates (CS + FS) of SIM, RIM, SFM, RFM, UP were 11.05, 9.87, 9.59, 9.32, and 7.13 respectively (Figure 1). There was a statistically significant difference between these values (F=18.983, p=0.000). Tukey's post hoc analysis was performed in order to search for the groups creating the difference (Table 3). This analysis revealed that untrained physicians were less successful compared with family practice specialists and internal medicine specialists, whereas family practice residents were less successful compared with internal medicine specialists and internal medicine residents.

Family practice residents received the highest false scores whereas internal medicine specialists received the lowest false scores (mean ± SD: 2.73 ± 1.98 vs. 1.20 ± 1.15). The mean false scores for family practice specialists, internal medicine residents, and untrained physicians were 2.54 ± 1.63, 1.71 ± 1.07, and 2.40 ± 1.54 respectively (Figure 1). There was a statistically significant difference between these values (F = 4.284, p = 0.003). Tukey's post hoc analysis was performed in order to determine the groups responsible from the difference (Table 4).

The highest error rate was in the ECG set with pacemaker and atrial fibrillation diagnoses. 32.8 % (n = 46) of the participants received false scores from this ECG set. The lowest error rate was in the ECG set with acute myocardial infarction and sinus bradycardia. Only 7.9 % of the participants (n = 11) got false scores from this ECG set.

Normal ECG and acute MI were selected as the most important key diagnoses important for all specialties There was no statistically significant difference between groups with regard to correctly identifying two normal ECG sets plus acute MI (X2=3.54; p=0.471) (Table 5). Of the total sample, 54.2 % (n = 76) could correctly identify these three diagnoses. On the other hand, correct identification rates of the normal ECG's and acute MI were 67.1 % (n = 94) and 85 % (n = 119) respectively.
Median non-response rates of the groups were calculated. UP's had the highest non-response rates followed by RI, SFP, RFP, and SI (Median non-response rates 3.5, 2, 1, 1, and 0 respectively) (Kruskal-Wallis X2=25.7, p<0.001). The highest non-response rate was observed in the ECG 9 (n=79, 56.4%), which contained the diagnoses left bundle branch block and atrial flutter whereas the lowest non-response rate was in ECG 4 (n=12, 8.6 %) (Table 6).

ECG interpretation skills are important for all clinicians and many studies from different countries have revealed that the problem is universal [3-6]. This study examines the problem by focusing on primary care physicians and comparing them with internists.

These results should be interpreted in the view of the undergraduate and residency education curriculum of Turkey. Currently, there is a common curriculum for family practice residency education throughout Turkey, which does not contain any place for cardiology rotations [7]. In our opinion, lack of cardiology education during family practice residency is the main reason why family physicians scored less than internists in this study. Family physicians gain their skills probably during their internal medicine rotations and from patient encounters in their practices. However, although not at the desired level, it is clear that specialisation makes a difference. Family physicians are in an intermediate place between internists and untrained physicians.
The differences between family physicians and untrained physicians raise concerns about a potential lack of sufficient ECG training in undergraduate medical education. curriculum. This finding, while limited to Edirne, has implications for the rest of the nation.

We assume that ECG reading capabilities certainly play some role in the referral rates to secondary and tertiary levels. The referral rate in the primary health care is currently around 14.4 % for Turkey [8]. Insufficiencies in ECG reading probably can be regarded as a factor that makes primary care physicians fear cardiac symptoms, but this topic needs investigation by other studies.

There was statistically significant difference between the groups with regard to most of the diagnoses. It is interesting to note that all diagnoses that reveal no difference between the groups are normal ECGs or those related to cardiac rhythm disturbances. These are relatively easy to diagnose just by measurement of cardiac rate. On the other hand, specialists of internal medicine were more successful than other groups in identifying diagnoses that are more difficult.

Correct recognition of an acute MI strip is one of the important skills primary care physicians should have [2]. While only 66.7 % of the untrained physicians could identify acute MI, this percentage increases to 91.9 % for specialists of family medicine. This finding supports our assumption on the insufficiency of undergraduate medical education to cover the clinical requirements of the graduates.

The total of CS plus FS did not approach the total number of diagnoses (i.e. 16) in any group. This reveals that all groups, but especially the untrained physicians, have doubts in making a certain decision on the ECG sets. Although we expect somehow reverse findings in the results of false scores, untrained physicians represent an exception in this picture. Pacemaker is a rare diagnosis encountered in primary care [1]. We assume that the current undergraduate as well as postgraduate curricula of all groups in this study should be questioned with regard to teaching ECG reading skills. We conclude that especially physicians working at primary care positions have less chance to have patients with pacemakers, and are thus less successful in correctly identifying this diagnosis.

Aspects Concerning Education
These findings show that, graduates of medical faculties do not have the necessary qualifications for ECG interpretation. However, there is an agreement that the aim of medical education is to train graduates for some kind of specialisation, including family practice [9]. Hence, it is understandable that medical graduates are not ready to practice in primary care settings. The fact that medical graduates can work in primary care is the side of this problem, which might be a disadvantage for patients. Information seeking among primary care physicians is a problem for coping with the growing knowledge of medicine [10] Lack of obligations for continuous medical education may be an explanation for lower scores of untrained physicians. Another reason may be the relatively low availability of diagnostic tools including ECG in the Turkish primary health centres.

Although internists have performed better than family physicians, both groups seem to have problems with ECG interpretation. One striking result of this study is that even the specialists of internal medicine could get a score of only 9.85 out of 16 (61.6 %). We suggest cardiology rotations of certain durations for all residency trainings where ECG reading skills are important. A curriculum focused especially on diagnoses prevalent for primary care should be applied to family practice residents, taking place in cardiology departments and primary care offices together.

Structured training programs of ECG interpretation skills are in fact necessary for all clinical specialties dealing with the patient.

Work experience should contribute to knowledge enhancement. However, given the fact that there is no correlation between work experience and total correct and false scores, we assume that work circumstances have no effect in improving the diagnostic skills,

Limitations
Although we tried to include the maximum sample size from each group, it was not possible to reach a nationally representative sample. Hence, this is just a study comparing different medical specialties. The samples of different groups were selected with similar characteristics with regard to sex, but the same is not true for age; specialists are older due to the time elapsed in residency education. On the other hand, there are some non-respondents, who might be assumed as different than the study population, but it is not possible to clarify this issue in this study setting.

There is currently no standardised instrument to measure ECG reading skills. We developed our own instrument in the light of the literature and with the counselling of a cardiologist.

Although particularly important for branches such as internal medicine, paediatrics, and family medicine, ECG reading is a skill necessary for all clinicians. To reveal a whole picture exposing the different factors on the degree of ECG reading skills, studies covering all physician groups with higher sampling rates should be conducted. A standardised ECG set can be developed for this purpose in order to help researchers from different nations to plan studies enabling international comparison.

Conclusion
This study has demonstrated that family practice residency education contributes to ECG interpretation skills. Beyond that the skills are not satisfactory enough, we do not think that formal education has much contribution to this effect. Personal efforts and the efforts of individual educators probably have much more effect in this matter. We think that the addition of formal cardiology training in family practice residency education will help to close the gap between internists and family practitioners. This study also supports the fact that primary care physicians must have special training in accordance with international standards in order to work in the field. There is a need to improve the ECG interpretation skills of medical undergraduates supported by appropriate postgraduate education.