Introduction: Reimbursement databases are potentially invaluable tools to develop and conduct pharmacoepidemiological studies on drug use. However, two types of factors that may influence the performance of a database can be distinguished: firstly, factors related to the constitution of the database, and secondly, factors related to the data. For the latter, we think that two are important: the presence of the drug in the database and the capacity to capture real-life use, both of which are influenced by the marketing status (e.g., OTC) and whether reimbursement is possible.
Objectives: To illustrate and discuss to what extent reimbursement databases are relevant tools to conduct drug utilization studies with regard to the data on drugs.
Methods: In order to illustrate the reliability of data in reimbursement databases, data from the main French national health insurance database (55 million individuals) were compared to national drug sales in France (units) during the same year.
Results: Depending on the ATC class, the capture in the database of drugs actually sold ranged from 32 to 81%.
Discussion: Capture of classes of drugs in the database may be explained by the specific characteristics of the French health insurance system (reimbursable drugs, OTC market share). These characteristics influence the studies that can be performed both in terms of the topic but also the methodology. This problem probably exists for the other reimbursement databases used worldwide.
Conclusion: Studies should be designed according to the strengths and weaknesses of reimbursement databases that were not originally developed for pharmacoepidemiology.