PURPOSE: To determine the diagnostic accuracy of computed tomography (CT) for pneumonia in patients with adult respiratory distress syndrome (ARDS). MATERIALS AND METHODS: CT scans were obtained within 1 week of bronchoscopic sampling in 31 patients receiving mechanical ventilation for ARDS for more than 48 hours. Of 11 patients with pneumonia, five developed symptoms less than 11 days after the onset of ARDS (early ARDS). CT scans were rated for pneumonia independently by four radiologists who were unaware of the clinical diagnosis. Diagnostic accuracy was defined by means of the area under the receiver operating characteristic curve, or A2. RESULTS: Diagnostic accuracy for pneumonia was fair (A2 = 0.69 +/- 0.04 [standard error]) owing to 70% true-negative ratings (vs 59% true-positive ratings). The generalizability coefficient was good (0.79). No single CT finding was significantly different for the presence of pneumonia. Nondependent opacities predominated in 10 (91%) of 11 patients with pneumonia and 12 (60%) of 20 without pneumonia. Nondependent opacities predominated in nine (56%) of 16 patients with early ARDS and 13 (87%) of 15 with late ARDS. CONCLUSION: CT has fair diagnostic accuracy for ventilator-associated pneumonia in patients with ARDS owing primarily to identification of patients without pneumonia. No single CT sign was significantly different for pneumonia, but dependent atelectasis was more common in patients with early ARDS without pneumonia.

Article History

Published in print: 1998