This retrospective multicenter study compared microorganisms documented by quantitative cultures from bronchoscopic samples in episodes of ventilator-associated pneumonia (VAP) from three different institutions in Barcelona (B), Montevideo (M), and Seville (S). The observations were compared with the findings reported by Trouillet and coworkers (AJRCCM 1998;157:531-539) in Paris (P). The objective was to evaluate whether a classification of etiologies of VAP in four groups, based on the number of ventilation days and previous antimicrobial use, might contribute to establishing generalized guidelines for empirical therapy. Significant variations in etiologies (p < 0.05) were found in all of the microorganisms isolated from VAP episodes across three treatment sites when compared with the reference site (P). In Group 1 (< 7 d and absence of antibiotics), Pseudomonas aeruginosa remained extremely infrequent (3 of 89, 3.3%) in the joint category, whereas the incidence of Acinetobacter baumannii was significantly higher, owing to M findings. On the other hand, one site (B) had a significantly lower incidence of multiresistant pathogens (Methicillin-resistant Staphylococcus aureus [MRSA] and nonfermenters other than P. aeruginosa), even in Group 2 (< 7 d and antibiotics), Group 3 (>/= 7 d and absence of antibiotics), and Group 4 (antibiotics and >/= 7 days). Similar findings were documented when episodes were grouped according to Groups 1 and 3 of the ATS guidelines. We conclude that causes of VAP varied markedly across four treatment sites, resulting in the need for large-scale variations in antimicrobial prescribing practices. Instead of following general recommendations, antimicrobial prescribing practices for VAP should be based on up-to-date information of the pattern of multiresistant isolates from each institution.