Validation of a rapid diagnostic strategy for determination of significant bacterial counts in bronchoalveolar lavage samples

Context. - Bacterial cultures of bronchoscopic samples require 1 to 2 days for results to be available for use in clinical decisions. We developed a rapid diagnostic testing strategy that is highly sensitive for screening bacteria in bronchoalveolar lavage (BAL) samples, with results available within hours of collection. Objective. - To validate the ability of a bacterial adenosine triphosphate (ATP) assay and routine Gram stain microscopy to detect significant bacterial counts in BAL samples. Design. - Four hundred seventy-seven BAL samples from 319 patients suspected of having pneumonia were tested using a rapid diagnostic strategy, consisting of Gram stain and a bacterial ATP assay. Rapid results were compared with quantitative cultures with a positive cutoff of 104 CFU/ml or higher. Results. - Significant bacterial counts were identified in 107 samples (22%). The most common etiologic agents were Staphylococcus aureus (25%), Haemophilus influenzae (17%), and Streptococcus pneumoniae (12%). The rapid test results were false negative in 5 cases (S aureus in 2, both Klebsiella pneumoniae and S aureus in 1, and Stenotrophomonas maltophilia and S pneumoniae in 1 case each). The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the rapid diagnostic strategy were 95.3%, 54.9%, 37.9%, 97.6%, and 63.9%, respectively. Conclusion. - A negative result with this rapid diagnostic testing strategy rules out significant bacterial counts in BAL samples with a high degree of certainty and may allow use of narrow-spectrum antimicrobial agents or withholding of empiric antimicrobial therapy in patients suspected of having ventilator-associated pneumonia.