The radiologic diagnosis of autopsy-proven ventilator-associated pneumonia

An abnormal chest roentgenogram is essential for the diagnosis of ventilator-associated pneumonia. The diagnostic accuracy of various roentgenographic signs of pneumonia has not been assessed previously in the portable anteroposterior roentgenograms obtained in ventilated patients. Seven roentgenographic signs (air bronchograms, alveolar infiltrates, silhouette sign, cavities, fissure abutment, atelectasis, and asymmetric infiltrates superimposed on diffuse bilateral infiltrates) were evaluated for their accuracy in predicting pneumonia alone, in combination with other signs, or in combination with clinical parameters. The last roentgenogram prior to autopsy of 69 ventilated patients was interpreted by three reviewers and the above signs were correlated with autopsy evidence of pneumonia. Pneumonia was present in 24 (35 percent) of the 69 autopsies. No roentgenographic sign had a diagnostic efficiency of greater than 68 percent. By stepwise logistic regression, the presence of air bronchograms was the only roentgenographic sign that correlated with pneumonia in the total group, correctly predicting 64 percent of pneumonias. In patients without adult respiratory distress syndrome (ARDS), the presence of air bronchograms or alveolar infiltrates correlated with pneumonia, while in patients with ARDS, no roentgenographic sign and only the clinical parameter of purulent sputum correlated with pneumonia. Only a minority (7/22) of worsening alveolar infiltrates in all groups were due to pneumonia and were often confused with ARDS. Alveolar hemorrhage occurred with a surprising frequency (38 percent of autopsies), including 13/45 (29 percent) patients without pneumonia. Alveolar hemorrhage was associated with 29 percent of multiple air bronchograms and 30 percent of bilateral alveolar infiltrates in patients without pneumonia. We conclude that in intubated patients with diffuse bilateral roentgenographic infiltrates, no roentgenographic sign correlates well with pneumonia. No clinical parameter added to the accuracy of either an alveolar infiltrate or an air bronchogram in patients without diffuse infiltrates. Pulmonary hemorrhage and/or infarction are frequent autopsy findings in intubated patients and may be confused radiologically with pneumonia.