Clinical Investigations Imaging/Diagnostic Testing Efficient utilization of echocardiography for the assessment of left ventricular systolic function Jkepak Taheja, MD, Carol Gruver, MD, Jiri Sklenar, PhD, John Dent, MD, and Sanjiv Kaul, MD Charlottesville, Vu Background We hypothesized thot patients could be selected for echocardiogrophic evaluation of left ventricular (LV) systolic function on the basis of historic, clinical, radiographic, and electrocardiographic criteria. Methods and Results We prospectively evaluated 300 consecutive inpatients referred for the echocardiographic assessment of LV function, of whom 124 (41%) had LV systolic dysfunction (LVSD) (IV ejection fraction ~0.45). Among the historic variables, male sex was the only predictor of LVSD, whereas of the abnormal physical and radiographic findings, cardiomegaly on chest radiography was the only predictor. Among the electrocardiographic findings, the presence of left bundle branch block was positively correlated with the presence of LVSD, whereas a normal electrocardiogram was nega- tively correlated with this finding. Only 2 patients with LVSD had o normal electrocardiogram. The addition of significant predictors on physical examination and chest radiography doubled the predictive value of the historic variables for determining LVSD. The addition of electrocardiographic findings further doubled the predictive value of the model. Almost 45% of the predictive power of the final multivariate model (chi-square of 48 of the total chi-square of 108) was based on the absence of normal electrocardiogram in patients with LVSD. Wh en chest radiographic findings were excluded from the model, the overall predictive power of the model did not change, with the normal electrocardiogram gaining greater prominence: Full 56% of the predictive power of the model (chi-square of 60 of the total chi-square of 108) resided in the ability of a normal electrocardiogram to discriminate between patients with and those without LVSD. Conclusions Historic, chest radiographic, and electrocardiographic variables can be used to predict low likelihood of LVSD on echocardiography. In particular, when the electrocardiogram is normal, it is extremely unlikely to have LVSD. It can be argued that such patients should not be referred for echocardiography. (Am Heart J 2000; 139:394-8.) See related Editorial on page 388. Unnecessary tests and procedures account for about one sixth of the $1 trillion health care cost in the United States1 One of the fastest-growing tests in health care, and definitely the fastest-growing in cardiology, is echocardiography.* It is estimated that more than 15 mil- lion echocardiograms were performed in the United States in 1997 alone.* One of the most common indica- tions for echocardiography is the assessment of left ventric- ular (LV) systolic function, which accounts for approxi- mately one third of the echocardiograms ordered at our institution. Physicians of all backgrounds and in all clini- cal contexts request the test. We hypothesized that patients could be selected for echocardiographic evaluation of LV systolic function on From Ihe Cordiovosculor Dwsion, Unwersiiy of Virginlo Medical Center. Presented in port 01 the 70fh Annual Scienhfic Sesrron of the Amerlcon Heari ASSOCI- ohon, November 1998, Orlando, Flo, Grculofion I997;96(suppl I ):620. Submitted April 5, 1999; accepted July 2 I, 1999. Reprint requests: Sonjiv Koul, MD, Cordiovosculor Dwis~on, Box 158, University of Virginia Medical Center, Chorlo&wille, VA 22908 E-mail: sL@virginio.edu Copyright 8 2000 by Mosby, Inc. 0002.8703/2000/$I2.00 + 0 4/l/103224 the basis of historic, clinical, radiographic, and electro- cardiographic criteria. To test this hypothesis, we per- formed a prospective study of inpatients referred to our echocardiography laboratory for the evaluation of LV systolic function. Methods Three hundred thirty consecutive inpatients referred to the echocardiography laboratory for evaluation of LV systolic function were initially evaluated. The referring physicians were directly contacted to confirm that the primary indication for echocardiography was in fact the assessment of LV systolic func- tion. Thirty patients who did not have an electrocardiogram per- formed within 1 week before echocardiognphy were excluded from the study. The study population, therefore, comprised 300 patients. LV systolic dysfunction (LVSD) was considered to be present if the visually estimated LV ejection fraction was ~0.45 as assessed by an experienced echocardiognpher blinded to all other information. Further grading of LVSD was not performed. Variables analyzed All patients underwent chart review at the time of echocar- diography. The historic variables assessed were history of anterior myocardial infarction, valvular heart disease, essential hypertension, and diabetes mellitus. The admission physical