Decision making often relies on electrocardiographic results. Unfortunately, in most patients without acute myocardial infarction, electrocardiograms are not entirely normal and may show nonspecific ST-T changes [3, 4]. Classic findings of acute myocardial infarction—ST elevation with or without Q waves or new T-wave changes—are seen in less than one third of patients with acute myocardial infarction at the time of initial presentation to the emergency department [3-7]. In a few patients with acute myocardial infarction, initial electrocardiograms can also be entirely normal [3-7]. Because it takes several hours before creatine kinase levels can be detected in the blood after coronary occlusion [8], measurement of this enzyme at the time of initial presentation in the emergency department is usually also unhelpful [1].

For these reasons, physicians tend to hospitalize patients until acute myocardial infarction has been ruled out. Patients who progress to acute infarction either have electrocardiograms that indicate an infarction or have increased serum levels of creatine kinase (and the MB fraction) or both [8, 9]. Thus, if patients remain stable and an acute myocardial infarction has been ruled out, they can be discharged from the hospital; patients for whom an acute myocardial infarction has been "ruled in" can be managed accordingly.

Not surprisingly, only one third or less of all patients admitted to the coronary care unit with the diagnosis of no (ruled out) acute myocardial infarction, ultimately rule in for this condition [3, 4, 10]. Most of the others are discharged from the hospital, often without a cardiac diagnosis. Further, despite the cautious approach of having a relatively low threshold for admitting patients with chest pain, approximately 5% of those sent home from the emergency department—because they appear not to have acute ischemia—later have acute myocardial infarction outside the hospital [2, 3]. Lawsuits resulting from such missed diagnoses account for the largest settlements in the United States for failure to diagnose or treat a condition in the emergency department [11].

Several alternative approaches have been tried to decrease unnecessary hospital admissions and to improve the diagnosis of acute myocardial infarction in the emergency department. These include more expert evaluation of the patient and the electrocardiographic results and either the use of cardiac imaging to detect abnormal regional function with echocardiography [3, 12, 13] or the use of nuclear imaging techniques to detect abnormal regional perfusion [14, 15]. Although these imaging techniques have improved detection of acute myocardial infarction in the emergency department and may potentially decrease unnecessary hospital admissions and costs [3], these approaches are not routinely used in evaluating chest pain in the emergency department for several obvious reasons.

Emergency department physicians, by the nature of their training, are generalists and not experts in cardiovascular medicine. Consequently, their ability to differentiate chest pain of ischemic from that of nonischemic origin is not as good as that of cardiologists who evaluate patients with ischemic heart disease on a regular basis. Similarly, the skills of emergency department physicians in interpreting electrocardiograms, particularly when the findings are subtle, cannot be expected to equal those of cardiologists. Because echocardiography and nuclear imaging require a high degree of expertise in performance and interpretation, one cannot expect emergency department physicians to independently use these modalities to make important triage decisions in the emergency department.

Another approach reported to be useful in emergency department evaluation of chest pain is the use of decision-based computer models [10, 16] and probability analyses [17]. Although these models have been validated prospectively in large populations [16, 17] and are relatively easy to use, they have not gained widespread acceptance for four major reasons. First, most physicians do not have adequate training in biostatistics or mathematics, making computer models appear more abstract than tangible. Second, models reflect probability rather than certainty and are, therefore, unsatisfactory to most physicians who want conclusive proof to confirm a diagnosis. Third, although properly selected models can represent average experiences, their use to predict the likelihood of life-threatening events for a single person is highly questionable. Finally, physicians are aware that even a remote chance of acute myocardial infarction in a particular patient can lead to litigation. They will, therefore, take the least risky approach and hospitalize the patient.

Given these practical realities, what should the strategy be for chest pain of uncertain origin? We recommend two cost-effective and efficient approaches that do not place patients at any additional risk. One approach is to consult a cardiologist who can evaluate the patient and ancillary data and determine whether the chest pain is ischemic in origin. If appropriate, cardiologists can order tests, including echocardiography or nuclear perfusion imaging, to aid in decision making. This approach alone could substantially reduce hospital admissions. In addition, patients who truly have an evolving myocardial infarction are more likely to have early and more aggressive reperfusion (either thrombolysis or angioplasty) in the hands of cardiologists. It may not always be logistically possible, however, to have cardiologists available in the emergency department.

The other approach, which is more likely to have widespread application, is to observe patients in special chest pain units located in or near the emergency department that are staffed by coronary care-trained nurses. Monitoring tests—electrocardiography and measurement of cardiac enzyme levels—can be repeated every hour [18], and if acute myocardial infarction is ruled in, the patient can be treated accordingly and possibly admitted to the hospital. Blood levels of newly described cardiac enzymes, including subforms of the MB fraction, myoglobin [19], and troponin T [20], increase early after the onset of myocardial necrosis and are positive markers for myocardial infarction with a high degree of specificity. Increased levels of these cardiac enzymes could make the diagnosis of acute myocardial infarction more certain, and earlier detection of this increase might even lead to earlier therapy to reestablish perfusion. If, for 8 hours or some other acceptable interval after onset of chest pain, electrocardiographic results and cardiac enzyme levels remain negative for myocardial infarction and if chest pain does not recur, the patient can be sent home with appropriate follow-up. In this manner, low-risk patients for whom acute myocardial infarction is ruled out are not admitted to the hospital, translating into considerable cost saving. Currently, this approach is being successfully implemented in several institutions [21].

A substantial proportion of patients who present to the emergency department with cardiac-related symptoms and for whom an acute myocardial infarction is ruled out are still at risk for future events, including nonfatal infarction and death [4, 12]. These high-risk patients can be identified in the emergency department on the basis of historical data (such as history of hypertension) and physical findings (such as presence of congestive heart failure). The electrocardiogram is not helpful in these patients [4]. In comparison, evaluation of left ventricular function by two-dimensional echocardiography is extremely useful, and this technique distinguishes between high and low-risk patients [12]. Therefore, even when sent home from the emergency department after an acute myocardial infarction is ruled out, these patients need aggressive follow-up.

It is unlikely that these recommendations will detect all patients with acute myocardial infarction or avoid unnecessary hospital admissions. They will, however, decrease hospital admissions and costs without putting patients at additional risk. Because all patients with cardiac-related symptoms cannot be hospitalized, we will always miss some patients with acute infarctions. We believe the approaches discussed above are reasonable and can form the basis for more detailed guidelines drawn up by expert panels.