The chest radiograph in critically ill surgical patients is inaccurate in predicting ventilator-associated pneumonia.

Chest radiographs (CXRs) are frequently obtained in surgical intensive care unit (SICU) patients when a diagnosis of ventilator-associated pneumonia (VAP) is suspected. The purpose of this study was to determine if the interpretation of the CXR correlated with a diagnosis of VAP in SICU patients. Prospective evaluation of 20 SICU patients clinically suspected of VAP was performed from July 1997 through December 1998. All patients required mechanical ventilation for at least 48 hours, and antibiotic use was discontinued 24 hours before entry into the study. Bronchoscopy with protected specimen brush (PSB) sampling of secretions from the right and left lung was performed. A positive PSB was present if quantitative analysis yielded > or = 10(4) colony-forming units/mL of bacteria. VAP was diagnosed if either the right or left PSB was positive and ruled-out if both the right and left PSB yielded < 10(4) colony-forming units/mL. Twelve of 20 patients (60%) were diagnosed to have VAP by PSB criteria. Eight of 20 patients (40%) had CXRs interpreted as negative for infiltrates; four patients had VAP by PSB criteria. There were four patients with focal infiltrates; two patients had VAP. The remaining eight patients had radiographs interpreted as bilateral infiltrates (one) or pulmonary edema (seven); of these, six patients (75%) had VAP. The sensitivity of the CXR in determining the presence of VAP was 25 per cent, the specificity was 75 per cent, and the accuracy was 0.45. The CXR does not improve the clinician's ability to diagnose VAP: a normal CXR does not exclude the presence of VAP and the finding of a focal infiltrate does not confirm the diagnosis of VAP.