Medication errors in a hospital in the United States and a hospital in the United Kingdom were compared. The study was conducted in wards with a high oral-drug-related workload in two large university hospitals. The U.S. hospital was studied in August 1993 and the U.K. hospital in May and June 1993. The U.S. hospital had a typical unit dose drug distribution system, and the U.K. hospital had the ward-based system commonly used in that country, in which a pharmacist visits each ward several times daily and reviews each patient's medication chart. The medication chart is used by the physician to order drugs and obviates the need for transcription of orders. A disguised-observation technique was used to determine frequencies and types of medication errors. Medication errors were identified retrospectively in the U.S. hospital by comparing the observer's notes with the original drug orders made in the patient's chart by the physician. In the U.K. hospital, identification of errors took place concurrently; as doses were administered, they were compared with the orders on the medication chart. In the U.S. and U.K. hospitals, 919 and 2756 opportunities for error were observed, respectively. The medication error rate in the U.S. hospital was 6.9% (95% confidence interval [CI], 5.2% to 8.5%), significantly higher than the 3.0% rate observed in the U.K. hospital (95% CI, 2.4% to 3.7%) (95% CI for the difference, 2.1% to 5.7%). Omitted doses and incorrect doses were the most common types of errors in the U.K. hospital; incorrect doses and unordered doses were the most common types in the U.S. hospital. An American hospital with a unit dose distribution system had a significantly higher medication error rate than a British hospital with a ward-based supply system.