We prospectively studied the safety of replacing intravenous delivery systems, including those used in total parenteral nutrition, at 72- compared with 48-hour intervals in 487 patients. Although the prevalence of contamination of intravenous fluid was higher in administration sets replaced at 72-hour intervals (10/664, 1.5%) than in sets replaced every 48 hours (6/710, 0.8%), the difference is not statistically significant. Contamination in both groups was almost exclusively with small numbers of coagulase-negative staphylococci (range, 1 to 27 colony-forming units/mL); no contaminated infusion was associated with clinical signs of sepsis or concordant bacteremia. Contaminants were recovered less frequently from peripheral venous infusions (0.6%) than from infusions used for central venous access or hemodynamic monitoring (1.5%) or total parenteral nutrition (3.6%); infusions in an intensive care unit were more frequently contaminated (2.5%) than infusions on medical and surgical wards (0.9%). These data indicate that extrinsic contamination of intravenous fluid is a rare cause of endemic nosocomial septicemia, and for most infusions it is unnecessary to routinely replace delivery systems more frequently than every 72 hours.