To determine both the global Incident, and the Incident for stages of medication errors in six Catalonian hospitals, the types of error and the consequences. A prospective design, with the global variable of the medication error. Potential errors have been excluded. The patients admitted to each hospital were studied in 2 groups of up to 300 patients and 1,500 administrations were observed. The NCCMERP taxonomy was applied. The prescription error was detected through the review of prescriptions, checking the patient, medication, adherence to protocols, interactions, contraindications, omission, duplicated therapy, doses, frequency, method, and lack of follow-up. During the transcription/validation, it was verified that the prescription matched the original order. In the dispensing process, the content of the drawers was checked, comparing it to the computer generated list, before sending out the single dose trolley. The transcription, preparation and administration were observed on the wards. The information for all the procedures was registered in a specific data sheet. There was moderate concordance amongst the inspectors (kappa = 0.525). 16.94 errors were detected per 100 patients-day and 0.98 errors per patient: 16 % in prescription, 27 % in transcription/validation, 48 % in dispensing, and 9 % in administration. 84.47 % were category B errors (they did not reach the patient), and < 0.5 % of the errors were harmful. The population, with an average age of 65, had a male/female ratio of 60/40. The principal therapeutic groups were: agents against peptic ulcer and GERD, antithrombotic agents, and other analgesics and antipyretics, principally in a solid oral drug form (58 %). The medications per patient-day were 5.5 and the units of medication were on average 11.21, varying greatly among the institutions. The adjustment of 10 units made the results more uniform. In all the stages, omission was the most frequent error. The different methods used and different areas of the investigations make comparisons difficult. This is evident in the harmful errors, the proportion of which is affected by the detection procedure. The number of mistakes avoided during the execution of this project demonstrates the need to improve the planning of the work systems and to establish safety measures.