Can we reduce the use of echocardiography by using other clinical information to identify patients at very low risk for left ventricular systolic dysfunction?

BACKGROUND: About one third of all echocardiograms are ordered to evaluate left ventricular (LV) dysfunction. If other clinical markers could be used to identify patients at very low risk for LV dysfunction, some of these echocardiograms would be unnecessary. Eliminating unnecessary echocardiograms might help control health care costs without reducing the quality of care.

POPULATION STUDIED: The study sample consisted of 330 consecutive inpatients who underwent echocardiography to evaluate LV dysfunction. Of these subjects, 300 had an electrocardiogram (EKG) within 1 week before the echocardiogram, and the other 30 subjects who did not have EKGs were eliminated from the sample.

STUDY DESIGN AND VALIDITY: Variables related to medical history, physical examination, chest radiography, and EKG findings were identified using chart review. These variables were analyzed using a logistic regression model to determine which variables could be used to identify patients at very low risk for LV dysfunction on subsequent echocardiogram. EKGs were evaluated for the following abnormalities: Q waves, poor R-wave progression, LV hypertrophy, S-T segment abnormalities, left bundle branch block, or a paced rhythm. Subjects who had none of these abnormalities were classified as having normal EKGs. A patient with a right bundle branch block or a rate disturbances could have been classified as having a normal EKG. The regression model and sample size are appropriate for this type of analysis. The model could be strengthened by validating it using similar subjects not involved in the initial development of the model.

OUTCOMES MEASURED: The primary outcome was LV systolic dysfunction defined as an echocardiogram with an LV ejection fraction less than 0.45 as assessed by an echocardiographer blinded to other clinical information.

RESULTS: Of the 300 patients, 124 (41%) had LV systolic dysfunction on echocardiogram. The variables found to be important predictors of LV dysfunction were male sex, cardiomegaly on chest radiograph, and left bundle branch block on EKG. Of the 300 subjects, 118 (39%) had normal EKGs. Only 2 subjects with normal EKG findings had LV dysfunction on echocardiography. One had a right bundle branch block and known valvular disease. The other patient had undergone cardiac bypass surgery between the time the EKG was recorded and the time the echocardiograph was performed.