OBJECTIVE: This prospective trial examined the efficacy of using bronchoalveolar lavage (BAL) for the diagnosis of pneumonia (PN) and the utility of Gram's stain (GS) for dictating empiric therapy. SUMMARY BACKGROUND DATA: Posttraumatic nosocomial PN remains a significant cause of morbidity and mortality. However, its diagnosis is elusive, especially in multiply injured patients. The systemic inflammatory response syndrome of fever, leukocytosis, and a hyperdynamic state is common in trauma patients, especially patients with pulmonary contusion. Bronchoscopy with BAL with quantitative cultures of the lavage effluent may distinguish between PN and systemic inflammatory response syndrome, and GS of the lavage effluent may guide empiric therapy before quantitative culture results. METHODS: Mechanically ventilated trauma patients with a clinical diagnosis of PN (fever, leukocytosis, purulent sputum, and new or changing infiltrate on chest radiograph) underwent bronchoscopy with BAL. Effluent was sent for GS and quantitative cultures. The diagnostic threshold for PN was > or =10(5) colony-forming units (CFU)/mL, and antibiotics were continued. Antibiotics were stopped for < 10(5) CFU/mL and the diagnosis of systemic inflammatory response syndrome was made. Causative organisms for PN were compared to GS. RESULTS: Over a 2-year period, 232 patients underwent 443 bronchoscopies with BAL (71% men, 29% women; mean age, 41). The mean injury severity score was 30. Sixty percent of the patients had pulmonary contusion, and 59% were cigarette smokers. The overall incidence of PN was 39% and was no different regardless of the number of BALs a patient had. The false-negative rate of BAL was 7%. GS identified gram-positive organisms in 80% of patients with gram-positive PN and 40% of patients with gram-negative PN. GS identified gram-negative organisms in 52% of patients with gram-positive PN and 77% with gram-negative PN. The duration of the intensive care unit stay relative to the timing of BAL was beneficial for guiding empiric therapy. BAL in week 1 primarily identified Haemophilus influenzae and gram-positive organisms; Acinetobacter sp. and Pseudomonas sp. were more common after week 1. CONCLUSIONS: Bronchoscopy with BAL is an effective method to diagnose PN and avoids prolonged, unnecessary antibiotic therapy. Empiric therapy should be adjusted to the duration of the intensive care unit stay because the causative bacteria flora changes over time. GS of BAL effluent correlates poorly with quantitative cultures and is not reliable for dictating empiric therapy.