To identify bacteraemic children who are at increased risk of inappropriate empiric antibiotic therapy, we performed univariate and multivariate analyses of prospectively-studied bacteraemic episodes. Appropriateness of therapy was defined according to the in vitro susceptibility of the isolate. Inappropriate empiric therapy was found in 38% of 516 bacteraemic episodes and was associated with higher mortality. The rate of inappropriate treatment was lower in neonates and infants (28% and 33%, respectively) but higher in children 1- to 5-years old (51%, P = 0.0029). The rate was dependent on the source of bacteraemia (range, 18%-70%, P = 0.0092), underlying conditions (range, 26%-53%, P = 0.0001), the specific paediatric section in which the child was hospitalized (range, 24%-70%, P = 0.0002), and the causative micro-organism (range, 15%-75%, P < 0.0001). Four clinical variables that independently and significantly affected the rate of inappropriate antibiotic treatment were identified by multivariate stepwise logistic regression analysis (odds ratios in parentheses): hospital-acquired bacteraemia (2.3), age of 1- to 5-years (2.1), cytotoxic therapy (1.8) and presence of central i.v. line (1.6). We defined bacteraemic children who are at risk of inappropriate empiric antibiotic therapy. Special efforts are needed to improve their treatment and consequently their outcome.