Administrative claims databases: from epidemiology to pharmacoviligance ; Bases médico-administratives : de l'épidemiologie à la pharmacovigilance

The availability and the volume of data in administrative claims databases could offer a powerful tool for research. However, the origin and the financial purpose of these data require an adaptation of the classical epidemiological approaches. Indeed, the secondary use of these data raises several issues that need to be carefully addressed. The aim of this thesis work was to assess the contribution of claims databases in epidemiology and in pharmacovigilance. The first section discusses the definition of the administrative claims databases and their scope through international examples and then focuses on French claims databases that collect inpatient care data (Programme de médicalisation des systèmes d’information, PMSI) and outpatient care data (datamart de consommation inter-régime, DCIR). The second section discusses methodological issues to take into account the specificities of administrative claims databases: definition and assessment of a prognostic score (Charlson’s comorbidity index) applied to the SNDS database, and the management of missing data. Two approaches were used: i) the development of a proxy applied to the concept of planned hospital stay, and ii) the enhancement of claims data with clinical data from a hospital warehouse. The third section illustrates, through use cases, how claims data could be useful in epidemiology and pharmacovigilance: wide exposure window (maternal comorbidities collected in the 6 years before the conception date), exhaustive recruitment especially for rare pathologies (patients with inflammatory bowel disease and myocardial infarction), and improved screening of eligible patients and data collection especially during a health crisis (patients admitted to the intensive care unit for COVID-19). Finally, the limitations due to the secondary use of these data are discussed: their volume that might lead to statistical overpowered studies, their delayed access due to technical and legal restrictions, the effort required to understand these data that is often ...