The aims of this prospective study were to determine the incidence and types of prescribing medication errors and ways to prevent them from reaching patients. Data were collected from 4951 prescriptions over a 25 week period in 2002. Medication errors were classified as: incorrect dose, incorrect dose interval, duplication of therapy and drug interactions. The medical record analysis was used to compare prescribing with Croatian literature drug data and AHFS first Web version 2 (American Society of Health System Pharmacists). The incidence of medication errors in the entire sample, including all potential drug interactions, was 14.7%. However, as only 8 interactions (out of 356 potentially possible interactions) were assessed as clinically significant, then the total number of all types of medication errors equals 379. This resulted in an incidence of 7.7%. Dosage errors were the most frequent errors, followed by incorrect interval, drug duplication and drug interaction. The difference between the incidence of potentially possible and clinically significant drug interactions was quite large (7.2 vs. 0.2%). Thus, a critical attitude is necessary when evaluating data on drug interactions. Our findings point to the need of systematic control of prescribed therapies, which could be ensured by the application of the Unit Dose Drug Distribution System. A medication errors reporting program should be established both at hospital and at national levels in Croatia.