Background:A clinician makes a decision about the treatment of patients with acute myocardial infarction (AMI) has to consider many clinical factors. One of the most difficult problems is how to distinguish patients with suspected AMI from those with acute chest pain.
Methods:For this issue, in this study, we present a decision support model based on univariate analysis (Chi-square and Student t-test, p< 0.05) and C5.0 decision tree algorithm to discover critical factors from low level data, 64 clinical laboratory findings. Patients with acute chest pain seen in the emergency medical center of Keimyung University Dongsan Hospital from July 2006 to June 2007 were identified retrospectively. The patient population of 592 consisted of AMI patients (n=225; mean age, 64.02 years) and discharged patients (n = 367; mean age, 56.26 years) with normal clinical findings, who came to the emergency medical center with chief complaints of acute chest pain.
Results: In univariate analysis, 31 clinical factors, Chief Complaints, S.G. in urinalysis, WBC, RBC, HGB, HCT, NEUT, LYMP, EOS, BASO, LUC, APTT, Fibrinogen, Na, K, Cl, LDH, Lipase, CPK, CK-MB, Total Calcium, Glucose, Total protein, Albumin, ALP, AST, Actual Ca, pH, HCO3, BE, and TCO2, were significant difference between two groups. On the basis of these variables, we obtained 10 clinical factors selected by C5.0 decision tree algorithm and its importance was defined as the following orders: CK-MB, APTT, AST, Age, LDH, Cl, Na, LYMP, Fibrinogen, and CPK. The performances of six measures, accuracy, sensitivity, specificity, positive predictive value, negative predictive value, area under ROC curve, were 97.0%, 91.1%, 94.6%, 91.1%, 94.6%, and 92.8% (95% CI: 90.4 ~ 94.8).
Conclusion: The results indicated that the decision model developed in our study can be applied to provide decision support to clinicians and increase vigilance before the occurrence to suspected AMI.
|