Gram stain of bronchoalveolar lavage fluid in the early diagnosis of ventilator-associated pneumonia

Abstract

To assess the usefulness of the Gram stain in the early diagnosis of ventilator-associated pneumonia (VAP), we performed 146 protected specimen brushings (PSB) and bronchoalveolar lavages (BAL) in 118 patients suspected of having nosocomial pneumonia. Gram stain and counts of infected cells were performed in all samples from BAL fluid. A final diagnosis of pneumonia was established in 51 patients and there was no infection in 95 cases. A threshold of 2% of infected cells was used to distinguish between VAP and the group without VAP (sensitivity 86.3%, specificity 78.9%, positive predictive value 68.7% and negative predictive value 91.4%); there was good agreement with the final diagnosis (kappa statistic 0.616; concordance 81.5%). Regarding detection of bacteria using the Gram stain, we found a sensitivity of 90.2%, specificity 73.7%, positive predictive value 64.8% and negative predictive value 93.3%; there was moderate agreement with the final diagnosis (kappa statistic 0.586; concordance 79.4%). In the VAP group, we analysed the degree of qualitative agreement between Gram stain and PSB quantitative cultures: the correlation was complete in 51% (26 of 51 VAP), partial in 39.2% (20 of 51 VAP) and there was no correlation in 9.8% (five of 51 VAP). We conclude that the Gram stain is useful for rapid diagnosis of VAP but unreliable for early adaptation of empiric therapy.