Objective: The purpose of this study was to determine, with CT as the reference standard, the ability of radiologists to detect pleural effusions on bedside chest radiographs.
Materials and methods: Images of 200 hemithoraces in 100 ICU patients undergoing chest radiography and CT within 24 hours were reviewed. Four readers with varying levels of experience reviewed the chest radiographs and predicted the likelihood of the presence of an effusion or parenchymal opacity on independent 5-point scales. The results were compared with the CT findings.
Results: All readers regardless of experience had similar accuracy in detecting pleural effusions. Among 117 pleural effusions, 66% were detected on chest radiographs (53%, 71%, and 92% of small, moderate, and large effusions) with 89% specificity. Similarly, 65% of all parenchymal opacities were detected on chest radiographs, also with 89% specificity. Most (93%) of the misdiagnosed pulmonary opacities were simply not seen. Meniscus, apical cap, lateral band, and subpulmonic opacity were highly specific findings but had low individual sensitivity for effusions. The finding of homogeneous opacity, including both layering and gradient opacities, was the most sensitive sign of effusion. Atelectasis can occasionally mimic the pleural veil sign of effusion, accounting for most false-positive findings.
Conclusion: Radiologists interpreting bedside chest radiographs of ICU patients detect large pleural effusions 92% of the time and can exclude large effusions with high confidence. However, small and medium effusions often are misdiagnosed as parenchymal opacities (45%) or are not seen (55%). Pulmonary opacities often are missed (34%) but are rarely misdiagnosed as pleural effusions (7%).