RESUMEN
BACKGROUND: Ventilator-associated pneumonia (VAP) is diagnosed in about 30% to 50% of critically ill postsurgical and trauma patients. Early appropriate antibiotic therapy has been associated with improved survival rates. The diagnosis, however, continues to be a challenge. We routinely employ clinical pulmonary infection scores to warrant a bronchoalveolar lavage (BAL) quantitative culture to subsequently diagnose VAP. Presumptive antibiotic therapy for the first 48 to 72 hours is based on the sputum Gram stain, obtained at the time of BAL. This study was conducted to analyze the predictive value of sputum Gram stain for selecting appropriate early antibiotic therapy for VAP as confirmed by a BAL quantitative culture (>10 CFU/mL considered diagnostic). METHODS: The retrospective analysis included 124 consecutive intensive care unit patients with 186 identified episodes of presumed VAP from December 2002 to June 2006. VAP episodes were identified by a clinical pulmonary infection score > or =6, availability of a sputum Gram stain, and a corresponding quantitative culture result from a BAL sample. RESULTS: The overall correlation between Gram stain and subsequent organism identified on the BAL quantitative culture was only fair with a kappa score of 0.314. The best predictive value calculated was for the category of negative Gram stain. However, in 10 of 45 episodes where the sputum Gram stain did not identify a predominant organism, the BAL culture isolated pathogenic strains. Pseudomonas sp. was the most common bacteria isolated from the BAL samples. CONCLUSIONS: Irrespective of sputum Gram stain, presumptive triple antibiotic coverage should be instituted to provide dual antibiotic coverage for gram-negative bacilli, and vancomycin for gram-positive cocci. Additionally, identification of no organisms in the sputum Gram stain should still prompt broad-spectrum antibiotic coverage until the final results of the BAL quantitative culture are available.