Ninety-nine intravascular catheters were evaluated by a semiquantitative culture and Gram and acridine orange direct stains. A diagnosis of catheter-related infection was determined by a retrospective review of clinical records. Compared with the culture method, direct examination of catheters lacked sensitivity. Surprisingly, for some patients, a positive stain for yeasts not recovered by culture was considered significant. The culture correlated with bacteremia but failed to predict infection in 42% of patients.