Identification des maladies neurodégénératives dans les bases de données médicoadministratives en France : revue systématique de la littérature

Abstract

BackgroundGiven the health, social and economic burden of neurodegenerative diseases (ND), the development of epidemiologic studies is required. Administrative databases, such as the French national health insurance database (SNIIRAM) could represent an opportunity for researchers. ND could be presumed from drug reimbursement data, hospital stays or registration of a chronic condition. The aim of this study was to describe, in French administrative databases, algorithms used to identify Alzheimer's disease and associated disorders (ADAD), Parkinson's disease and associated disorders (PDAD), multiple sclerosis (MS), and amyotrophic lateral sclerosis (ALS). MethodsA systematic literature review was performed in Medline and gray literature through December 31th, 2015. French studies focusing on ADAD, PDAD, MS or ALS as a primary health outcome, conducted among one of the SNIIRAM data sources (outpatient reimbursements, chronic condition registration, hospital discharge) were included. ResultsThirty-four studies were included (ADAD, n=18, PDAD, n=9, MS, n=4, ALS, n=3), leading to 36 algorithms. For each studied ND, there was an important variability in the algorithms, concerning (i) the type of criteria used (administrative database versus multi-source systems); (ii) the number of criteria used; (iii) the definition used for each criteria. The extent and level of drug exposure highly varied. Identification through hospitalizations showed variations in terms of type of stay (short stay, long-term stay, psychiatric ward…), extent of diagnosis codes used, diagnosis type (principal, related, associated diagnosis) and period used. A validation study was conducted for 2 out of 36 algorithms (PDAD), and criteria completeness was estimated for 3 algorithms (MS, ALS). ConclusionDespite the increase in ND identification among French administrative databases, few algorithms have been validated. Validation studies should be encouraged.