OBJECTIVE: Patients are at risk of harm from medication errors. Barcode medication administration (BCMA) systems are recommended to mitigate preventable adverse drug events (ADEs). Our hypothesis was that a BCMA system would reduce preventable ADEs by 45% in a neonatal intensive care unit. STUDY DESIGN: We conducted a prospective, observational, cohort study of a BCMA system intervention in a neonatal intensive care unit. Participants were admitted neonates during 50 weeks. Medication errors and potential or preventable ADEs were detected by a daily structured audit of each subject's medical record, with assignment of an event as a preventable ADE made by blinded assessors. The generalized estimating equation method was used in modeling the targeted, preventable ADE rate with covariates. RESULTS: A total of 92,398 medication doses were administered to 958 subjects. The generalized estimating equation method yielded a relative risk of preventable ADE when the system was implemented of 0.53 (95% confidence limits 0.29 to 0.91, P = .04), adjusted for log(10)doses of medication/subject/day, a significant predictive covariate (P < .001), as well as for birth weight, sex, Caucasian race, birth cohort number, and nursing hours/subject/day. CONCLUSION: The BCMA system reduced the risk of targeted, preventable ADEs by 47%, controlling for the number of medication doses/subject/day, an important risk exposure.