The frequency of antibiotic resistance among bacteria in 4 intensive care units (ICUs) at a university hospital in Sweden was investigated annually from 1993 to 1996. An increase in ampicillin-resistant enterococci from 1993 to 1995 was seen which was due to a shift from Enterococcus faecalis to Enterococcus faecium. After a special infection control programme was instituted, the rate of ampicillin resistance among enterococci and the number of E. faecium isolates declined during 1996. The oxacillin resistance rates for Staphylococcus aureus were < or = 2%, while most of the coagulase-negative staphylococci (CNS) were oxacillin resistant. No vancomycin-resistant enterococci or staphylococci were seen. The ciprofloxacin resistance rate for CNS and Enterococci spp. were high. Relatively, high levels of resistance to cefotaxime and piperacillin/tazobactam among Enterobacter spp. were also seen. During 1995 and 1996 Pseudomonas aeruginosa showed increasing resistance to ceftazidime, ciprofloxacin and piperacillin/tazobactam. This was due to an outbreak among rather few patients. The overall resistance rates for Gram-negative bacteria were low for aminoglycosides and imipenem. From 1993 to 1996 the total antibiotic consumption decreased by 27% in the whole hospital and 16.5% in the ICUs. However, the reduced antibiotic consumption was paralleled with a 23% decrease in the total number of patients treated in the hospital from 1993 to 1996. In contrast there was an 11.5% increase in the number of ICU patients treated during this period. The conclusion is that all ICUs within a hospital should have a programme for 'on-line' antibiotic resistance surveillance of drugs used in that unit in order to change the empiric treatment when there is an increase in antibiotic resistance. It is also important to survey the antibiotic consumption in the ICUs in order to avoid further selective pressure on bacteria showing increased resistance rates.