BACKGROUND: While current guidelines recommend consideration of local microbiologic data when selecting empiric treatment for hospital-acquired pneumonia (HAP), few specifics of how to do this have been offered. METHODS: We conducted a retrospective analysis of HAP pathogens in 111 consecutive patients who acquired HAP during July to December 2004 and had a corresponding positive culture finding for a bacterial pathogen. These data were used to develop institution-specific guidelines. RESULTS: The most common bacteria identified were Staphylococcus aureus, Acinetobacter baumannii, and Pseudomonas aeruginosa, which were associated with 38%, 25%, and 19% of pneumonias, respectively. Susceptibility of Gram-negative bacteria to piperacillin-tazobactam and cefepime was 80% and 81%, respectively. The isolation of organisms resistant to piperacillin-tazobactam or cefepime was significantly more frequent in patients who had been hospitalized > or = 10 days. Of Gram-negative isolates resistant to piperacillin-tazobactam or cefepime, ciprofloxacin was active against < 10%, while amikacin was active against > 80%. New treatment guidelines were developed that divided the American Thoracic Society/Infectious Diseases Society of America "late onset/risk of multidrug-resistant pathogens" group of patients into two subcategories: "early-late" pneumonias (< 10 days of hospitalization) and "late-late" pneumonias (> or = 10 days of hospitalization). Guideline-directed treatment regimens would be predicted to provide adequate initial therapy for > 90% of late-onset pneumonias at our institution. CONCLUSIONS: Current guidelines suggest adding either an aminoglycoside or a fluoroquinolone to beta-lactam therapy for empiric Gram-negative coverage. However, in our institution, adding ciprofloxacin would not appreciably enhance the likelihood of providing initial appropriate antibiotic coverage. This underscores the importance of employing a systematic analysis of local data when developing treatment guidelines.