Use of the 'historical' National Health Data System to study infectious diseases

Identifying infectious diseases through French National Health Data System (SNDS), a medico-administrative database, presents specific challenges due to their intrinsic characteristics, such as curability or highly variable prevalence. This study assesses the capacity of the SNDS to accurately identify pathologies and to ensure follow-up of affected individuals.

The study focuses on "Certain infectious and parasitic diseases" (ICD-10 Chapters A and B). Eligible conditions were required to correspond to a three-character ICD-10 sub-chapter accounting for fewer than 10,000 discharge summaries (MCO) within the Programme for Medicalisation of Information Systems (PMSI) between 2006 and 2024. It evaluates the influence of different indicators, such as the extraction sources, the quality of record linkage using the national registration number (NIR) and the follow-up of subjects.

A total of 155,419 individuals across 77 selected ICD-10 sub-chapters were included. On average, 6% of individuals were non-linkable. Subjects identification was primarily achieved through the beneficiary registry (87%). By sub-chapter, the PMSI was the main source of identification, with an average of 95%. The average proportion of individuals still present in the SNDS five years after the initial occurrence was 58%.

These findings are closely linked to the core concept of the SNDS, which relies on health insurance data collection. Data linkage is fundamentally tied to insurance affiliation; consequently, while high-prevalence tropical or sub-tropical diseases are identifiable, longitudinal tracking is more challenging to achieve.