Bosco-Levy P1, Blin P2, Lignot-Maleyran S2, Lassalle R2, Abouelfath A2, Diez P3, Debouverie M4, Brochet B5, Louapre C6, Heinzlef O7, Maillart E8, Moore N9, Droz-Perroteau C2
1Bordeaux PharmacoEpi, INSERM CIC1401, Bordeaux University ; Bordeaux population health research centre, INSERM UMR 1219, Bordeaux University, Bordeaux, France, 2Bordeaux PharmacoEpi, INSERM CIC1401, Bordeaux University, Bordeaux, 33, France, 3Bordeaux PharmacoEpi, INSERM CIC1401, Bordeaux University, BORDEAUX, France, 4CHU de Nancy - Hôpital Central, Nancy, France, 5CRC SEP, service de neurologie, CHU de Bordeaux, Bordeaux, France, 6Sorbonne Université, institut du cerveau de la moelle épinière, ICM, Hôpital de la Pitié Salpétrière, INSERM UMR S 1121, CNRS UMR 7225 ; Service de Neuroligie, CHU de Paris Salpétrière, PARIS, France, 7service de neurologie, CHU de Poissy, POISSY, France, 8service de neurologie, CHU de Paris Salpétrière, PARIS, France, 9Bordeaux PharmacoEpi, INSERM CIC1401, Bordeaux University ; CHU de Bordeaux, Bordeaux, France
BACKGROUND Although the main limitation of the French nationwide claims database (SNDS) is the absence of clinical information, relapse is an outcome that can be identified to assess effects of disease modifying therapies in multiple sclerosis (MS) in real word setting. OBJECTIVES The objective of this study was to assess the validity of an algorithm identifying relapses in MS patients in SNDS. METHODS A random sample of 200 patients - 100 with at least one relapse and 100 without relapse screened by the algorithm - were randomly selected from a cohort of 37,986 MS patients previously identified in the SNDS. For each case, all data available in the SNDS, in particular those related to the dispensing of corticosteroids, hospitalizations for potential MS relapse or for administration of high dose of steroids and plasmapheresis procedures were examined by 2 neurologists to assess the presence or absence of relapses, blind to the result of the algorithm. In the event of an inter-expert discrepancy, the summary sheets were reviewed in a collegiate manner, in order to reach a consensus. Algorithm performance was estimated using the positive and negative predictive values (PPV, NPV). RESULTS Among the 200 patients randomly selected- 100 with at least one relapse and 100 without relapse - the algorithm correctly detected 95% patients with relapses (PPV) and 96% of patients without relapses (NPV). CONCLUSIONS This claim-based algorithm appeared to successfully detect MS relapse and could thus be applied to future observational MS studies in SNDS. It could be secondarily revised to include all changes proposed by the experts in order to optimize its performance.
Conference/Value in Health Info
2019-11, ISPOR Europe 2019, Copenhagen, Denmark
Epidemiology & Public Health
Disease Classification & Coding