Breast ultrasound utilization in a safety net emergency department.

Abstract

The purpose of this study is to (1) assess the diagnostic yield of ultrasounds performed in the emergency department for suspected breast abscess and determine the rates of reimaging, discordance, and emergent intervention in a large, busy safety net hospital and (2) determine clinical factors significantly associated with abscess as a way to improve patient selection for emergent breast ultrasounds. A total of 581 consecutive breast ultrasounds performed in the emergency department for suspected abscess over 15months were retrospectively reviewed for imaging, demographics, laboratory data, and physical exam findings. Breast abscess was confirmed by combining imaging, clinical, and laboratory data. Linear logistic regression analysis estimated the likelihood of abscess, and the cross-validated area under the receiver operating characteristic curve (AUC) evaluated diagnostic performance. Final diagnoses included abscess (153/581, 26%), cancer (29/581, 5%), granulomatous mastitis (41/581, 7%), normal (120/581, 21%), and other/indeterminate (238/581, 41%). Factors associated with abscess included induration, fluctuance, erythema, drainage, smoking, diabetes, and Black race. Based on these factors, the AUC of the characteristics predictive of abscess was 0.77 (CI, 0.72-0.81). Six breast cancers were not diagnosed on ultrasound. 40% of ultrasounds (231/581) were considered incomplete/inadequate. 74% (428/581) of emergent breast ultrasounds in our population were negative for abscess, while 21% (6/29) of cancers were not diagnosed, and 40% (231/581) of exams were incomplete/inadequate. Patient selection for emergent ultrasounds can be improved, allowing patients with a low likelihood of abscess to be imaged in a more optimal setting.