From medico-administrative databases analysis to care trajectories analytics: an example with the French SNDS

International audience; Medico-administrative data like SNDS (Système National de Données de Santé) are not collected initially for epidemiological purposes. Moreover, the data model and the tools proposed to SNDS users make their in-depth exploitation difficult. We propose a data model, called the ePEPS model, based on healthcare trajectories to provide a medical view of raw data. A data abstraction process enables the clini-cian to have an intuitive medical view of raw data and to design a study-specific view. This view is based on a generic model of care trajectory, that is a sequence of time stamped medical events for a given patient. This model is combined with tools to manipulate care trajectories efficiently. I N T R O D U C T I O N Medico-administrative databases hold rich information about healthcare trajectories (or healthcare pathways) at an individual level. Such data are very valuable for carrying out pharmaco-epidemiological studies on large representative cohorts of patients in real-life conditions. Moreover, historical data are readily available for longitudinal analysis of care trajectories. These opportunities are given by the use of the database of the French healthcare system, so called SNDS (Syst eme National de Donn ees de Sant e) database, which covers 98.8% of the French population, with a sliding period of 3 years. A classical pharmaco-epidemiological study from medico-administrative databases consists of three main steps: (i) defining inclusion and exclusion criteria of a cohort, (ii) specifying proxies for events of interest, and (iii) analyzing the transformed data. Practically, these three steps are closely intertwined and make use of digital data management tools (e.g., SQL databases, R, or SAS). The study outcomes depend on the available data at hand as much as on the tools to manage and process them. But the data model, 1 designed for administrative purposes , is not suitable for pharmaco-epidemiological studies without careful data preparation. It leads to difficulties for epidemiologists to access the useful information and even to know what is reachable with such databases. For instance, the SNDS database is a rela-tional database with hundreds of tables with very complex join relations. The set of prescribed drugs of a patient is accessible with a query containing 10 join relations involving attributes with unintuitive names. Mastering the data management with such complex models requires a lot of time, good knowledge of its content, and some technical skills. It is a practical bottleneck to exploit the potential of the database. 1 A data model is an abstract model that describes the organization of the data. In relational database, it is the description of tables, their attributes, and their relations. ª 2017 Soci et e