The utility of clinical, electrocardiographic, and roentgenographic variables in the prediction of left ventricular function.

To determine the clinical value of simple, widely available variables in estimating left ventricular (LV) function, we performed an analysis on 14,507 patients presenting with chest pain who were enrolled in the Coronary Artery Surgery Study registry. Of these patients, 4,034 had a normal electrocardiogram, and of these, 91.8% had an LV ejection fraction (EF) > 0.50, 7.6% had an EF of 0.36 to 0.50, and only 0.6% had an EF < or = 0.35. The presence of T-wave abnormalities (with normal QRS), left bundle branch block, electrocardiographic evidence of LV hypertrophy or myocardial infarction, cardiomegaly on chest roentgenogram, basilar rales, or third heart sound significantly decreased the likelihood of normal LVEF. Based on these clinical variables, a logistic regression model with a sensitivity of 68% and a specificity of 74% for identifying subjects with normal EF was developed. It was concluded that in patients with chest pain, consideration of such readily available clinical data provides useful information and may decrease the need for more expensive imaging methods.