Performance of the French national hospital discharge database algorithm to identify hospitalised Lyme borreliosis cases, France, 2017–2018

Résumé fourni par la source

BACKGROUND: In France, surveillance of Lyme borreliosis (LB) is based on general practitioners of a sentinel network and the national hospital discharge database (PMSI). Given the known limitations of the PMSI for epidemiological surveillance, we assessed the performance of its algorithm in identifying hospitalised LB cases in three university hospitals, in terms of sensitivity and positive predictive value (PPV). METHODS: We identified patients hospitalised during 2017-2018 with positive laboratory results for LB from hospital laboratory databases. Simultaneously, we screened in the PMSI LB hospitalised patients via the algorithm based on ICD-10 codes. Classifications were made by applying the EUCALB criteria. Confirmed and probable LB cases were classified as "LB+". We then calculated sensitivity (proportion of LB+ hospitalised cases identified by the PMSI algorithm) and PPV (proportion of LB+ cases among patients identified as LB by the PMSI algorithm). RESULTS: Among 541 patients with positive laboratory results, 54 were classified as LB+. The PMSI identified 62 cases, of which 38 were LB+. Overall sensitivity was 62%, varying by site (40-79%). Sensitivity was highest for paediatric cases (83%), Lyme arthritis (83%), and neuroborreliosis ( 61%). PPV was 61%, reaching 96% for neuroborreliosis and 83% for Lyme arthritis. CONCLUSION: The PMSI algorithm showed moderate sensitivity and PPV for identifying hospitalised LB cases, with higher performance for neuroborreliosis and Lyme arthritis. Our findings support focusing PMSI-based surveillance on neuroborreliosis and arthritis. Improvement of the PMSI algorithm is necessary but it remains a valid tool for assessing the burden and trends over time at the national and regional levels.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.