Despite the potential benefits of using prescription claims databases for pharmacoepidemiological research, little work has been reported on the nature of available information or its accuracy. The purpose of this study was to describe information contained within the prescription claims database in Québec, and to assess the accuracy of drug information that might be used to monitor drug exposure and physician prescribing. The comprehensiveness of the prescriptions claims database was assessed by examining 1,917,214 records of dispensed prescriptions for a regionally stratified random sample of 65,349 Québec elderly in 1990. We found that values in key fields (individual identifiers, drug, quantity, date dispensed and duration) were missing or out of range in 0-0.4% of records. The accuracy of data were examined in 723 prescriptions filled by 306 elderly patients attending one internal medicine clinic. Of these prescriptions, 83% were filled by the patient and correctly identified the patient and drug and in 89% of these 599 records, the prescribing physician was correctly identified. The quantity and duration of the prescriptions were accurate in 69.1% and 72.1% of records, respectively. We conclude that the prescription claims database in Québec may represent one of the most accurate means of determining drugs dispensed to individuals. There may be limitations in using this database for dosing information.