Background/Purpose: DESIR is the French cohort of patients with symptoms suggestive of early axial Spondyloarthritis (axSpA). Our objective was to enrich this cohort with claims and hospitalization data from the French national healthcare database (Système National des Données de Santé, SNDS) in order to provide opportunities to answer new research questions related to care consumption, the occurrence of health outcomes and costs and enhance data completeness.

Methods: We carried out a probabilistic linkage of the DESIR cohort database (N=703) and the SNDS (N= 387,463 individuals, sharing the same sex, birth month and year with DESIR  patients). DESIR is a cohort of 708 individuals aged 18 to 60 years, included from 2007 to 2010 in 25 centers in France, of whom 703 were not opposed to this linkage. The probabilistic linkage was performed with the Fellegi and Sunter method, using 29 variables collected between baseline and the 24^th month of follow-up. Matching variables included area of residence, attending physician, sacroiliac radiographs and MRI dates, blood test dates and treatment by a Disease Modifying Anti-Rheumatic Drug. We calculated for each DESIR-SNDS potential pair of patients a matching score that summed the matching probabilities calculated across all selected variables. Distributions of the highest matching scores for each DESIR patient and for other matching scores were displayed graphically and a cutoff was chosen visually. Pairs with highest matching score for each individual were matched if they exceed this cutoff. To assess the quality of the linkage we compared individuals’ characteristics between matched and unmatched individuals. P values were computed with T.test, Chi-square test, Wilcoxon test or Fisher exact test.

Results: A total of 594 (84%) were matched. They were, on average, 3 years older than unmatched individuals, were more likely men, and less likely to be HLA-B27 positive. No significant difference was observed between matched and unmatched on baseline clinical or radiographic characteristics: BASDAI and BASFI scores, MRI sacroiliac joint inflammation (ASAS criteria) neither among anti-TNF treatment during the first 5 years of follow-up. Finally, rate of loss to follow up by year 5 were similar in both groups while rate of exit from the DESIR cohort due to other diagnosis than axSpA was slightly higher but not significantly so, among matched vs. unmatched.

Conclusion: It was possible to perform a matching for 84% of patients. No difference was observed on clinical or radiographic characteristics nor treatment between matched and unmatched individuals, thus this cohort issued from the linkage can be used to study outcomes among patients with symptoms suggestive of early axSpA. A possible bias related to both male, older and HLA-B27 positivity over-representation should be discussed in any future study using this linked cohort. In a next step, a deterministic linkage will be performed using a unique individual identifier to validate this linkage and capture remaining unmatched individuals. 

« Back to ACR Convergence 2020

ACR Meeting Abstracts - https://acrabstracts.org/abstract/probabilistic-linkage-of-a-cohort-of-individuals-with-symptoms-suggestive-of-early-spondyloarthritis-and-the-french-national-healthcare-database/