Accurate diagnosis of the causes of chest pain and dyspnea remain challenging. In this preliminary observational study with a 5-year follow-up, we attempted to find a simplified approach to selecting patients with chest pain needing immediate care based on the initial evaluation in ED. During a 24-month period were randomly selected 301 patients and a conditional inference tree (CIT) was used as the basis of the prognostic rule. Common diagnoses were musculoskeletal chest pain (27%), ACS (19%) and panic attack (12%). Using variables of ACS symptoms we estimated the likelihood of ACS based on a CIT to be high at 91% (32), low at 4% (198) and intermediate at 20.5–40% in (71) patients. Coronary catheterization was performed within 24 hours in 91% of the patients with ACS. A culprit lesion was found in 79%. Follow-up (median 4.2 years) information was available for 70% of the patients. Of the 164 patients without ACS who were followed up, 5 were treated with revascularization for stable angina pectoris, 2 were treated with revascularization for myocardial infarction, and 25 died. Although a simple triage decision tree could theoretically help to efficient select patients needing immediate care we need also to be vigilant for those presenting with atypical symptoms.
Chest pain is one of the most common causes for referral to emergency departments (ED), accounting for several million visits annually [
Recent updated NICE guidance recommends 12-lead electrocardiography (ECG) as soon as possible in patients with acute chest pain but warns not to exclude ACS if the ECG is normal [
This was a prospective observational study. To prevent selection bias, which has been shown to influence hospital admissions for myocardial infarction [
The following symptoms on presentation were assessed: oppressive retrosternal chest pain radiating to arms or neck, oppressive retrosternal chest pain without radiation, respiration-dependent chest pain, stabbing chest pain, burning chest pain, dyspnea, cough, expectoration, heartburn, and the time of start of symptoms.
Definition of events: dyspnea was defined as the perception of an inability to breathe comfortably [
ACS was defined as acute myocardial infarction or unstable angina pectoris. Acute myocardial infarction was defined as an increase in troponin T to >0.01 ug/L, either initially or 6 hours later, with at least one of the following: ischaemic symptoms (chest pain, dyspnea), ECG changes indicative of ischaemia (ST-segment elevation or depression, or new left bundle-branch block), development of pathological Q waves in the ECG, imaging evidence of new loss of viable myocardium, or a new regional wall motion abnormality. ST-segment depression was defined as new horizontal or downsloping ST-segment depression ≥0.05 mV in two contiguous leads and/or T inversion ≥0.1 mV in two contiguous leads with a prominent R wave or R/S ratio >1. ST-segment elevation myocardial infarction (STEMI) was defined as an acute myocardial infarction with new ST-segment elevation at the J point in two contiguous leads with the cut-off points: ≥0.2 mV in men or ≥0.15 mV in women in leads V2 and V3 and/or ≥0.1 mV in other leads [
Other defined differential diagnoses were musculoskeletal chest pain (consistently reproducible chest wall tenderness or pain) and panic attack, defined according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-PC). Results of coronary catheterization were collected from medical records. Followup was performed in April 2010 by contacting the patients by phone. Secondary outcome measures were coronary catheterization, revascularization, myocardial infarction, and death before followup. The study protocol was approved by the local ethics committee. Patients signed an informed consent form.
Separate logistic regression models for each variable provided odds ratios, 95% CIs, and the corresponding
All analyses were performed using R version 2.9.2 (R Development Core Team (2009). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, available from:
From December 2005 through November 2007 on 60 randomly selected days, 301 patients presented at our ED with chest pain or dyspnea. Patient characteristics are shown in Table
Characteristics of 301 Patients.
Age, years, median (range) | 56 (17–92) |
Male/female | 190 (63%)/111 (37%) |
Coronary heart disease | 64 (21%) |
Other cardiovascular disease | 32 (10%) |
Other heart disease | 57 (19%) |
Chronic lung disease | 38 (13%) |
Diabetes mellitus | 39 (13%) |
Arterial hypertension | 119 (40%) |
Cigarette smoker/history of smoking | 129 (43%) |
Dyslipidemia | 36 (12%) |
Familial history of coronary heart disease | 49 (16%) |
History of thromboembolic disease | 18 (6%) |
Cancer | 22 (7%) |
Chest pain | 234 (78%) |
Oppressive chest pain radiating to arms or neck | 42 (14%) |
Nonradiating oppressive chest pain | 50 (17%) |
Respiration-dependent chest pain | 55 (18%) |
Stabbing chest pain | 55 (18%) |
Nonspecific chest pain | 89 (30%) |
Burning chest pain | 8 (3%) |
Dyspnoea | 74 (25%) |
Cough | 67 (22%) |
Expectoration | 19 (6%) |
Heartburn | 14 (5%) |
Troponin T was measured in 235 (78%) patients. Troponin was not determined if the cause of chest pain could be attributed to noncardiac symptoms based on the clinical findings of the attending ED physician. In 42 (14%) patients, troponin T was measured a second time 6 hours later, because the chest pain lasted less than 6 hours. An ECG was performed in 279 (93%) patients. If the cause of chest pain could be determined by a method other than ECG (e.g., pleuritic chest pain caused by pneumonia diagnosed by chest X-ray), an ECG was not performed. The final diagnoses in the ER are shown in Table
Final diagnosis at the ED of 301 Patients†.
Musculoskeletal chest pain | 82 (27%) |
Acute coronary syndrome | 56 (19%) |
ST-elevation myocardial infarction | 16 (5%) |
Non-ST-elevation myocardial infarction | 28 (9%) |
Unstable angina pectoris | 12 (4%) |
Panic attack | 38 (12%) |
Cardiac arrhythmia | 31 (10%) |
Chronic lung disease | 22 (7%) |
Heart failure | 21 (7%) |
Lower respiratory tract infection | 16 (5%) |
Dyspepsia | 15 (5%) |
Pulmonary embolism | 10 (3%) |
Hypertensive crisis | 7 (2%) |
Upper respiratory tract infection | 6 (2%) |
Vasovagal Syncope | 5 (2%) |
Pericarditis | 3 (1%) |
Abdominal disease | 3 (1%) |
Pneumothorax | 1 (0.5%) |
Other ‡ | 5 (2%) |
†More than one diagnosis per patient possible.
‡Other: Newly diagnosed cancer, haematoma in the pouch of an implantable cardioverter/defibrillator, instent stenosis of the subclavian artery, skin infection, cerebral haemorrhage.
Out of 56 patients with ACS (16 with STEMI, 28 with non-STEMI), 23 had signs of myocardial ischaemia in the ECG. Coronary catheterization with or without PCI was performed within 24 hours in 51 (91%) patients with ACS. A culprit lesion explaining ischaemic symptoms was found in 44 (79%) patients. Vasospasm was found in 1 patient, 3 patients showed normal coronary arteries, and 3 patients showed good results after earlier PCI. Of the 5 patients without coronary catheterization, one had non-STEMI because of cocaine abuse, one died because of STEMI and cardiogenic shock, one with a non-STEMI received conservative management, one with unstable angina pectoris was sent home for an outpatient stress test, and one patient with non-STEMI did not undergo PCI for unknown reasons.
The results of the logistic regression models are outlined in Table
Logistic regression models for each variable separately on ACS.
Variables | Odds ratio | 95% CI | |
---|---|---|---|
Age | 2 | 1.29–3.09 | 0.002 |
Gender (female : male) | 0.4 | 0.20–0.80 | 0.0094 |
Coronary heart disease | 2.82 | 1.50–5.31 | 0.0013 |
Other heart disease | 0.29 | 0.10–0.83 | 0.021 |
Other cardiovascular disease | 1.54 | 0.65–3.62 | 0.328 |
History of thromboembolic disease | 0.87 | 0.24–3.11 | 0.828 |
Chronic lung disease | 0.63 | 0.23–1.70 | 0.36 |
Diabetes mellitus | 1.51 | 0.57–4.01 | 0.404 |
Arterial hypertension | 0.68 | 0.34–1.37 | 0.281 |
Cigarette smoker | 0.76 | 0.32–1.79 | 0.522 |
History of cigarette smoking | 0.86 | 0.36–2.04 | 0.728 |
Dyslipidemia | 1.24 | 0.44–3.49 | 0.688 |
Familial history of coronary heart disease | 0.73 | 0.27–1.98 | 0.537 |
Cancer | 0.24 | 0.03–1.87 | 0.175 |
Referred with chest pain | 0.68 | 0.37–1.24 | 0.212 |
Referred with dyspnoea | 0.27 | 0.10–0.70 | 0.0069 |
Referred with suspected ACS | 10.8 | 4.86–23.9 | <0.001 |
Symptoms lasting <1 hour | 0.87 | 0.10–7.62 | 0.902 |
Symptoms lasting 1–6 hours | 0.67 | 0.29–1.58 | 0.362 |
Symptoms lasting >6 hours | 2.03 | 0.96–4.29 | 0.063 |
Symptom lasting one day or more | 0.26 | 0.11–0.60 | 0.0015 |
Symptom chest pain | 3.46 | 1.32–9.05 | 0.012 |
Oppressive retrosternal chest pain radiating to arms or neck | 37.6 | 16.0–88.2 | <0.001 |
Nonradiating oppressive retrosternal chest pain | 1.29 | 0.61–2.71 | 0.5 |
Burning chest pain | 0.62 | 0.075–5.13 | 0.656 |
Respiration-dependent chest pain | 0.13 | 0.03–0.57 | 0.0064 |
Stabbing chest pain | 0.29 | 0.10–0.85 | 0.023 |
Nonspecific chest pain | 0.11 | 0.032–0.35 | 0.0002 |
Symptom dyspnoea | 0.91 | 0.46–1.81 | 0.792 |
Cough | 0.16 | 0.05–0.53 | 0.0027 |
Heartburn | 0.72 | 0.16–3.31 | 0.672 |
Figure
Conditional inference tree based on all clinical relevant variables associated with ACS: The boxes show the likelihood of ACS to be high (91%,
Median followup time was 4.2 years (2.5–4.4). Followup information was available for 210 patients (70%).
Followup was available for 46 (82%) patients diagnosed with ACS: 33 had stable CHD, three underwent revascularization, two suffered myocardial infarction, and 8 died.
Followup was available for 164 (67%) patients not diagnosed with ACS: 125 developed no CHD, 6 had coronary catheterization showing normal coronary arteries, one had coronary catheterization showing good results after earlier stenting, 7 underwent revascularization, 5 because of stable angina pectoris and 2 because of myocardial infarction, and 25 died. Information concerning the exact cause of death (e.g., cardiogenic or noncardiogenic causes) was not available. Median time between discharge from the ED to revascularization was 1.1 years (1 day–3 years).
Followup was available for 59 (72%) patients diagnosed with musculoskeletal chest pain: 51 patients developed no CHD, 4 underwent revascularization, one of whom had myocardial infarction, and 4 died.
Followup was available for 27 (71%) patients diagnosed with panic attack: none of the patients developed CHD, underwent revascularization, or suffered myocardial infarction, three underwent coronary catheterization showing normal coronary arteries, and one died.
In this study in patients admitted to the ED with chest pain or dyspnea, the largest group of patients (27%) had musculoskeletal pain, 19% were diagnosed with ACS before leaving the ED, and 12% with panic attack. These findings are similar to earlier reports [
A further interesting finding of our study is that 43% of patients were self-admissions to the ED with chest pain or dyspnea. Statutory health insurance in Switzerland covers visits to any emergency healthcare facility which means that patients are free to seek medical advice without being referred by a primary care physician. Comparisons with other parts of Europe are therefore difficult, because the emergency services in Switzerland are often used instead of physicians in office practice, especially in towns and cities at night and during the weekend, which increases the burden on EDs at these times. This healthcare system-specific consultation pattern is a typical finding in Switzerland. For example in a recent analysis of a stratified sample of 1173 patients among 11258 ED admissions in our surgical emergency department concerning referral practice among Swiss and non-Swiss walk-in patients we found that nationality was associated with greater use of ED services for nonurgent problems. From all Swiss patients 67% referred themselves (walk-in patients) without a previous GP visit, compared to 79% self-referrals among foreigners. 83% of Swiss patients visited a GP regularly, compared with 57% of non-Swiss patients (498/598 Swiss versus 331/575 non-Swiss,
In addition, 42% of our patients had had chest pain for >24 hours. Some of the patients in our catchment area live in remote valleys or in mountainous areas, up to 1900 meters above sea level, which means that some of them cannot consult a doctor immediately. The severity of the symptoms on presentation might also have played a role in the referral time point. A further explanation may be the specialty and experience level of the referring physician and the selection bias in referral rates of men and women with suspected angina, since angina pectoris symptoms tend to be more subtle in women than in men [
Based on the present study, after analysis of variables significantly associated with ACS, we selected four quickly assessable variables which can be used to create a conditional inference tree (Figure
In our study, 91% of the patients with ACS underwent coronary catheterization within 24 hours with identification of a culprit lesion confirming the diagnosis in 79% of cases. In other studies designed to establish clinical rules to predict the presence of ACS, coronary catheterization was either performed in only 19% of patients with ACS [
Based on our own findings, we feel that it would be interesting to set up a prospective study to validate this triage system and investigate the extent to which overtriaging might be reduced, together with patient outcome. Since some of our non-ACS patients later underwent coronary catheterization, after triaging patients on presentation to the ED, further assessment is necessary based on clinical skills and other risk scores and guidelines for the management of patients with ACS [
Our study had some limitations: firstly, troponin T, an important element in the definition of the primary outcome variable ACS, was measured only in 78% of patients. Secondly, 30% of patients were lost to followup because early follow-up was not performed. Thirdly, our sample size was small and represents experience at a single institution in a tertiary care ED setting and may therefore not be generalizable to other populations. Only 37% of our patients were women. Women with myocardial ischaemia can present with atypical symptoms [
Estimation of the incidence of ACS and other cardiovascular events in emergency medicine is a dynamic field, and the accurate diagnosis of the causes of chest pain and dyspnea remains challenging. In this preliminary observational study, we attempted to find a simplified approach to selecting patients needing immediate care based on the initial evaluation of the ED physician and patient-derived information. Although a simple triage decision tree could theoretically help to select patients needing immediate care and thus further enhance the quality of care and optimize the allocation of limited resources in ED, since out of 91% patients primarily diagnosed with ACS 79% did have lesions at coronary catheterization; a minority of non-ACS patients also underwent coronary catheterization at a later timepoint after ED presentation. This suggests that despite using decision trees in the ED based on screening rules, we not only need to be vigilant for those presenting with typical symptoms, but also need to use our clinical skills and experience to identify and follow up patients who appear primarily not to have ACS.
Conditional inference tree analysis was applied to elicit possible combinations of two or more risk factors. A tree based model is a good exploratory tool to approximate a complex model. The advantages are clearness of interpretation and visualisation of complex interactions which are not covered by regression modelling. The disadvantage is the possible instability using strong correlated predictors and the splitting of continuous variables into classes. Details are described here: statistical learning from a regression perspective available from [ original research on inference trees available from:
The authors cordially thank Rodney Yeates and Alistair Reeves for language revisions.