The object of this study was to evaluate the appropriateness of antibiotic use in relation to diagnosis and bacteriological findings in the intensive care units (ICUs) of a 1100-bed referral and tertiary care hospital with an antibiotic restriction policy in Turkey. Between June and December 2002, patients who received antibiotics in the medical and surgical ICUs were evaluated prospectively. Two infectious diseases (ID) specialists assessed the antibiotics ordered daily. Of the 368 patients admitted to the ICUs, 223 (60.6%) received 440 antibiotics. The most frequently prescribed antibiotics were first-generation cephalosporins (16.1%), third-generation cephalosporins (15.2%), aminoglycosides (12.1%), carbapenems (10.7%) and ampicillin-sulbactam (8.7%). Antibiotic use was inappropriate in 47.3% of antibiotics. ID specialists recommended the use of 47% of all antibiotics. An antibiotic order without an ID consultation was more likely to be inappropriate [odds ratio (OR)=13.2, P<0.001, confidence intervals (CI)=4.4-39.5]. Antibiotics ordered empirically were found to be less appropriate than those ordered with evidence of culture and susceptibility results (OR=3.8, P=0.038, CI=1.1-13.1). Inappropriate antibiotic use was significantly higher in patients who had surgical interventions (OR=3.6, P=0.025, CI=1.2-10.8). Irrational antibiotic use was high for unrestricted antibiotics. In particular, antibiotic use was inappropriate in surgical ICUs. Additional interventions such as postgraduate training programmes and elaboration of local guidelines could be beneficial.