A decision tree for the early diagnosis of acute myocardial infarction in nontraumatic chest pain patients at hospital admission

Abstract

Study objective: To find an accurate algorithm for the diagnosis of acute myocardial infarction in nontraumatic chest pain patients on presentation to the emergency department.

Design: In a prospective clinical study, we compared the diagnostic performances of clinical symptoms, presenting ECG, creatinine kinase, creatine kinase MB activity and mass concentration, myoglobin, and cardiac troponin T test results of hospital admission blood samples. By classification and regression trees, a decision tree for the diagnosis of acute myocardial infarction was developed.

Setting: Emergency room of a Department of Internal Medicine (University Hospital).

Patients: One hundred fourteen nontraumatic chest pain patients (median delay from onset of chest pain to hospital admission, 3 h; range, 0.33 to 22): 26 Q-wave and 19 non-Q-wave myocardial infarctions, 49 patients with unstable angina pectoris, and 20 patients with chest pain caused by other diseases.

Measurements and results: Of each parameter taken by itself, the ECG was tendentiously most informative (areas under receiver operating characteristic plots: 0.87 +/- 0.04 [ECG], 0.80 +/- 0.08 [myoglobin], 0.80 +/- 0.04 [creatine kinase MB mass], 0.77 +/- 0.04 [creatine kinase activity], 0.69 +/- 0.06 [clinical symptoms] 0.67 +/- 0.06 [creatine kinase MB activity], 0.67 +/- 0.05 [troponin T]). In patients presenting 3 h or less after the onset of chest pain, ECG signs of acute transmural myocardial ischemia were the best discriminator between patients with and without myocardial infarction. In patients presenting more than 3 h, however, creatine kinase MB mass concentrations (discriminator value, 6.7 micrograms/L) were superior to the ECG, clinical symptoms, and all other biochemical markers tested. This algorithm for diagnosing acute myocardial infarction was superior to each parameter by itself and was characterized by 0.91 sensitivity, a 0.90 specificity, a 0.90 positive and negative predictive value, and a 0.90 efficiency.

Conclusions: We found an algorithm that could accurately separate the myocardial infarction patients from the others on admission to the emergency department. Therefore, this classifier could be a valuable diagnostic aid for rapid confirmation of a suspected myocardial infarction.