Development of a Statistical Model for Prediction of Acute Myocardial Infarction by Biochemical Markers

Abstract

Diagnosis and management of potential acute myocardial infarction (AMI) patients presenting to the hospital emergency department (ED) with chest pain are difficult. To aid in clinical decision making, a multivariate logistic regression model was developed using the patient's presenting history, electrocardiogram (ECG), and biochemical marker (cardiac Troponin T, cTnT) to predict the probability of AMI. Data from 102 AMI and 619 non-AMI patients from a multicenter clinical trial were used for model building. The World Health Organization criteria were used for AMI diagnosis. Univariate analyses were performed to assess the effect of individual factors. Based on the significance of univariate effect and clinical importance, the following variables were included in a risk factor only model: age, sex, chest pain, systolic blood pressure, smoking, and history of myocardial infarction and hypertension with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.66 (95% confidence interval [CI]: 0.60–0.71) and an R² of 0.09. The final model, which included the risk factors and results of ECG and cTnT, increased the AUC and R² values to 0.93 (95% CI: 0.90–0.96) and 0.83, respectively. At the probability level of 0.10 for AMI, the model showed a sensitivity of 78.4% and a specificity of 92.6%. Statistical modeling is a useful tool in predicting risk of chest pain patients in the ED.