Current issues of ACP Journal Club are published in Annals of Internal Medicine


Radiologic diagnosis of ventilator-associated pneumonia had low accuracy

ACP J Club. 1992 Sept-Oct;117:54. doi:10.7326/ACPJC-1992-117-2-054

Source Citation

Wunderink RG, Woldenberg LS, Zeiss J, et al. The radiologic diagnosis of autopsy-proven ventilator-associated pneumonia. Chest. 1992 Feb;101;458-63.



To assess the accuracy of 7 roentgenographic signs (air bronchograms, alveolar infiltrates, silhouette sign, cavities, fissure abutment, atelectasis, and asymmetry of diffuse bilateral infiltrates) for diagnosing ventilator-associated pneumonia.


Comparison of portable anteroposterior chest roentgenogram findings with autopsy evidence of pneumonia.


Not stated.


69 patients who received ≥ 12 hours of mechanical ventilation before death, who had ≥ 2 available chest roentgenograms while receiving ventilation, and who had a complete autopsy.

Description of test and diagnostic standard

The last roentgenogram before death was interpreted independently by 3 reviewers using the previous day's and baseline films for comparison. The presence of the 7 roentgenographic signs was recorded. Disagreements on the presence of ≥ 2 signs were resolved by consensus. Pneumonia was defined as fibrinopurulent exudate in bronchioles or alveoli, or both, at autopsy.

Main outcome measures

Sensitivity, specificity, and accuracy of each roentgenographic sign for pneumonia.

Main results

24 of the 69 patients (35%) had pneumonia present at autopsy. All radiologic signs had an accuracy of ≤ 68%. Single air bronchograms and fissure abutment were highly specific (95%) but had low sensitivity (< 20%). {For single air bronchogram the likelihood ratio for a positive test result (+LR) was 3.69 and for a negative test result (-LR) was 0.87. For fissure abutment the +LR was 1.89 and the -LR was 0.96.}* Alveolar infiltrates and any air bronchograms were the most sensitive (88% and 83%, respectively) but were not specific (< 60%). {For alveolar infiltrates the +LR was 1.18 and the -LR was 0.49; for any air bronchograms the +LR and -LR were 1.97 and 0.29.}* In logistic regression analysis, only air bronchograms were predictive of pneumonia, but this model correctly identified only 65% of pneumonias. When both radiologic and clinical signs (fever, leukocytosis, and pathogens in sputum) were included in the analysis, both air bronchograms and pathogens in sputum were predictive of pneumonia, but the accuracy of the model increased minimally (72%). When the 16 patients with the adult respiratory distress syndrome were analyzed separately, no radiologic sign was predictive of pneumonia.


7 roentgenographic signs used in the diagnosis of pneumonia (air bronchograms, alveolar infiltrates, silhouette sign, cavities, fissure abutment, atelectasis, and asymmetry of diffuse bilateral infiltrates) were of limited value when applied to roentgenograms of patients with ventilator-associated pneumonia who ultimately died and were autopsied.

Source of funding: Not stated.

Address for article reprint: Dr. R.G. Wunderink, 956 Court Avenue, Memphis, TN 38163, USA.

*Numbers calculated from data in article.


Despite significant advances in patient management, a clinically simple, reliable, and accurate technique to diagnose ventilator-associated pneumonia remains elusive. Analysis of sputum has a low specificity for distinguishing colonization from pneumonia. Several recently described bronchoscopic techniques including quantitative cultures of protected brush specimens and bronchoalveolar lavage have improved diagnostic specificity but may not be suitable for all patients or readily available. Further, their sensitivities drop precipitously if empiric antibiotics are given before doing the procedures. Clinical criteria for pneumonia, such as purulent sputum, fever, and leukocytosis, although frequently seen with true ventilator-associated pneumonia, often are associated with tracheobronchitis and other distant nonpulmonary infections. Finally, treatment with broad spectrum antibiotics without a confirming diagnosis has risks: emergence of antibiotic-resistant organisms, fungal overgrowth, antibiotic-associated colitis, and direct toxicity.

Thus, clinicians still rely heavily on daily roentgenograms to detect ventilator-associated pneumonia. Wunderink and colleagues attempted to determine the usefulness of roentgenographic signs for diagnosis of ventilator-associated pneumonia. Unfortunately, they found no radiographic sign, alone or combined with clinical features, that was sufficiently sensitive or specific to be definitive. These results, although disappointing, are not surprising because roentgenographic expression of alveolar filling is similar whether the cause is pneumonia, acute lung injury, alveolar hemorrhage, or even cardiogenic pulmonary edema.

The authors use autopsy findings as the "gold standard" for ventilator-associated pneumonia. Because all patients were terminally ill, these findings may not necessarily be applicable to a more general intensive care unit population, especially to survivors of nosocomial pneumonia. Documentation of alveolar filling in nondependent lung regions that do not expand with recruitment techniques (positioning, suctioning, and transient hyperinflation) may improve the diagnostic accuracy of chest roentgenograms, although this approach will be more time consuming.

Adelaida Miro, MD
Michael R. Pinsky, MDUniversity of PittsburghPittsburgh, Pennsylvania, USA