Discriminant Methods as Aids to the Diagnosis of Acute Coronary Heart Disease in the Emergency Room

Summary Multivariate stepwise logistic regression and discriminant analyses were used to reduce emergency room data to the most relevant subset of variables that could improve the sensitivity and specificity of the diagnosis of acute coronary heart disease. Both procedures selected 5 out of 29 variables collected from 154 patients : abnormality of the ST segment in the electrocardiogram (ECG), shortness of breath, history of sublingual glyceryltrinitrate intake, chest pain duration >15 minutes, and previous history of a heart attack. The stepwise discriminant function selected additional variables: chest pain as the most important symptom, retrosternal site of chest pain and T wave abnormality on the ECG. The performance of the two methods was assessed by finding the classification matrix.