Objective: To describe the practice pattern of a group of primary care providers in ordering echocardiography for healthy patients and to derive a decision aid to identify patients with systolic murmurs who have significant cardiac lesions.
Design: Retrospective case review of clinical variables, including the demographic, historical, and physical examination features obtained from chart review of the physician evaluation prior to referral for echocardiography and the echocardiographic results, with both univariate and multivariate regression analyses used to identify variables predictive of positive echocardiographic outcomes and to derive a regression model.
Setting: General internal medicine clinic in an urban major teaching hospital.
Patients/participants: 169 patients between the ages of 18 and 55 years selected retrospectively from medical records.
Measurements and main results: Symptoms and auscultatory findings were not predictive of positive echocardiographic outcomes among the patients referred for the test. For the patients who had systolic murmurs, a logistic regression model had three significant predictors of positive echocardiographic results: additional year of age (odds ratio: 1.08, 95% CI: 1.02-1.13; p = 0.007); male gender (odds ratio: 5.87, 95% CI: 1.99-17.3; p = 0.002); and murmur grade > or = 3 (odds ratio: 4.99, 95% CI: 1.27-19.6; p = 0.02). The receiver-operating characteristic (ROC) curve area for this model was 0.741. If women aged 35 years and less with murmur grades < or = 2 had not been referred, 47% of the echocardiographies could have been avoided while retaining a sensitivity of 90%.
Conclusions: The study suggested that healthy patients with systolic murmurs are least likely to have positive echocardiographic results when they are young and female and have murmur grades < or = 2. More careful screening of this group might make unnecessary almost half of the echocardiographies yielding negative results ordered for healthy patients with systolic murmurs.