OBJECTIVE: Minimizing unwarranted prescription of antibiotics remains an important objective. Because of the heterogeneity between units regarding patient mix and other characteristics, site-specific targets for reduction must be identified. Here we present a model to address the issue by means of an observational cohort study. SETTING: A tertiary, multidisciplinary, neonatal, and pediatric intensive care unit of a university teaching hospital. PATIENTS: All newborns and children present in the unit (n = 456) between September 1998 and March 1999. Reasons for admission included postoperative care after cardiac surgery, major neonatal or pediatric surgery, severe trauma, and medical conditions requiring critical care. METHODS: Daily recording of antibiotics given and of indications for initiation. After discontinuation, each treatment episode was assessed as to the presence or absence of infection. RESULTS: Of the 456 patients 258 (56.6%) received systemic antibiotics, amounting to 1815 exposure days (54.6%) during 3322 hospitalization days. Of these, 512 (28%) were prescribed as prophylaxis and 1303 for suspected infection. Treatment for suspected ventilator-associated pneumonia accounted for 616 (47%) of 1303 treatment days and suspected sepsis for 255 days (20%). Patients were classified as having no infection or viral infection during 552 (40%) treatment days. The average weekly exposure rate in the unit varied considerably during the 29-week study period (range: 40-77/100 hospitalization days). Patient characteristics did not explain this variation. CONCLUSION: In this unit the largest reduction in antibiotic treatment would result from measures assisting suspected ventilator-associated pneumonia to be ruled out and from curtailing extended prophylaxis.