Potent antimicrobial agents have been developed as a response to the development of antibiotic-resistant bacteria, which especially affect patients with prolonged hospitalization in Intensive Care Units (ICU) and who had been previously treated with antimicrobials, especially third-generation cephalosporins. This study was to determine how changes in the empirical treatment of infections in ICU patients affect the incidence of Gram-negative bacteria species and their susceptibility to antimicrobials, and examine the impact of these changes on nosocomial infections. A prospective interventional study was performed in a university hospital during two periods: 1) First period (September 1999 to February 2000); and 2) Second period (August 2000 to December 2000); empirical treatment was changed from ceftriaxone and/or ceftazidime in the first period to piperacillin/tazobactam in the second. ICU epidemiological and infection control rates, as well as bacterial isolates from upper airways were analyzed. Ceftazidime consumption dropped from 34.83 to 0.85 DDD/1000 patients per day (p=0.004). Piperacillin/tazobactam was originally not available; its consumption reached 157.07 DDD/1000 patients per day in the second period (p=0.0002). Eighty-seven patients and 66 patients were evaluated for upper airway colonization in the first and second periods, respectively. There was a significant decrease in the incidence of K. pneumoniae (p=0.004) and P. mirabilis (p=0.036), restoration of K. pneumoniae susceptibility to cephalosporins (p<0.0001) and reduction of ventilator-associated pneumonia rates (p<0.0001). However, there was an increase in P. aeruginosa incidence (p=0.005) and increases in ceftazidime (p=0.003) and meropenem (p<0.0001) susceptibilities. Changing antimicrobial selective pressure on multi-resistant Gram-negative bacteria helps control ventilator-associated pneumonia and decreases antimicrobial resistance.