Journal Articles Multiple Sclerosis and Related Disorders Year : 2026

External validation and improvement of an algorithm that identifies multiple sclerosis relapses in the French national healthcare claims database

Laure Carcaillon-Bentata
Abdelilah Abouelfath
  • Function : Author
Séverine Lignot-Maleyran
  • Function : Author
Régis Lassalle
Jérémy Jové
  • Function : Author
Patrick Blin
Eric Berger
  • Function : Author
Bertrand Bourre
Jean-Philippe Camdessanché
Olivier Casez
Jonathan Ciron
Jérôme de Sèze
Gilles-Louis Defer
Inès Doghri
Amélie dos Santos
Karolina Hankiewicz
Olivier Heinzlef
Pierre M. Labauge
Emmanuelle Le Page
Alexandre Leclancher
  • Function : Author
Céline Louapre
Éric Manchon
Guillaume Mathey
Alexis Montcuquet
Caroline Papeix
Jean Pelletier
Corinne Potier
  • Function : Author
Aurélie Ruet
Mariana Sarov-Riviere
  • Function : Author
Sandra Vukusic
Laurent Magy
Maximilien Moulin
Romain Casey

Abstract

Background : Multiple sclerosis (MS) is a frequent neurological condition affecting young adults with acute disabling neurological episodes (relapses). MS relapses are not recorded in claims databases despite their importance for (pharmaco)epidemiological studies. This study aimed to validate and improve an algorithm identifying relapses in relapsing-remitting MS initiating disease-modifying therapy (DMT) within the French nationwide claims database (SNDS). Methods : Clinical data from the French MS registry (OFSEP) linked to the SNDS were used. The cohort included MS patients with a first DMT claim between July 2015 and December 2017, naive to any MS treatment, followed until December 2018 (n=1,640). The initial relapse algorithm combined high-dose corticosteroid prescriptions and hospitalization duration. Incidence of the first relapse identified in the SNDS was compared with OFSEP confirmed relapses that have been treated by corticosteroid or hospitalized (gold standard). Performances were estimated using Sensitivity, Specificity, PPV, NPV. Algorithm were reevaluated after revision of its criteria based on experts’ review of false positive and negative cases’ claims, and finally after adding non-naive MS patients (n=9,966). Results : The performances of the initial algorithm were Specificity 85.2%, Sensitivity 74.0%, NPV 89.8%, PPV 65.3%. After revision of corticosteroid dosages and hospitalization durations thresholds, performance slightly improved: 85.0%, 75.4%, 90.2%, 65.3%, respectively. When considering naive and non-naive MS patients, Sensitivity and NPV stayed similar (75.1% and 91.6%) while Specificity and PPV decreased (81.8% and 55.4%). Conclusion : The final algorithm of treated/hospitalized relapses can be applied accurately in naïve MS patients initiating a DMT, in the SNDS and likely in other claims databases after adapting to each country’s reimbursement and care practices.

No file

Dates and versions

hal-05607088 , version 1 (29-04-2026)

Identifiers

Cite

Laure Carcaillon-Bentata, Abdelilah Abouelfath, Séverine Lignot-Maleyran, Régis Lassalle, Jérémy Jové, et al.. External validation and improvement of an algorithm that identifies multiple sclerosis relapses in the French national healthcare claims database. Multiple Sclerosis and Related Disorders, In press, pp.107209. ⟨10.1016/j.msard.2026.107209⟩. ⟨hal-05607088⟩
8 View
0 Download

Altmetric

Share

  • More