Direct observation was used to detect medication errors and Bar Code Medication Administration (BCMA) workarounds on two pediatric units and one neonatal unit at UCSF Benioff Children's Hospital. The study (1) measured the frequency of nursing medication administration-related errors, (2) characterized the types of medication errors, (3) assessed compliance with the institution's six medication administration safety processes, and (4) identified observed workarounds following BCMA implementation. The results of the direct observation were compared to medication administration-related incident reports (IRs) for the same period. The frequency of medication errors was 5% for the three units. Compliance with the process measures was achieved 86% of the time (range 23–100%). Seven medication administration-related IRs were submitted during the same observation period. Three BCMA workarounds were identified; (1) failure to visually confirm patient's identification, (2) failure to compare the medication to the electronic medication administration record at least twice before administration, and (3) charting administration of medication before actual administration. The direct observation methodology identified a low frequency of medication administration errors (MAEs) consistent with post-BCMA implementation. The incident reporting system identified different MAEs than direct observation suggesting that both methods should be used to better characterize the scope of MAEs.