Autism Spectrum Disorders (ASD) are neurodevelopmental disorders with highly heterogeneous clinical profiles, outcomes and care pathways. Linking health administrative databases with cohort studies is an excellent opportunity to study healthcare utilisation by ASD patients. Optimisation of linkage should maximise the linkage rate and minimise potential errors and biases to ensure high quality analyses and the absence of bias.

This study aims to describe the linkage strategy between the French national health database (SNDS) and the ELENA cohort database, and to provide advice on optimizing linkage quality.

The ELENA cohort includes 876 children with a confirmed diagnosis of ASD between 2013 and 2019. The SNDS contains all healthcare services reimbursed by health insurance schemes. We performed deterministic matching on the reference sample and evaluated deterministic and probabilistic approaches on another sample. Performance was assessed using the matching rate and comparison between matched and unmatched groups.

We matched 96 % of the reference sample and 53 % of the other sample using either deterministic or probabilistic approach, resulting in a total match rate of 79 % with deterministic or mixed approach. The mixed approach minimized selection bias, producing a linked dataset of 692 children with ASD whose characteristics were comparable to those of the unmatched sample. Probabilistic weighting showed that hospitalization dates were the most discriminating variables for matching.

The mixed linkage strategy optimised match quality while limiting bias, confirming the feasibility of linking clinical ASD cohort data to the SNDS and strengthening future research on healthcare use among children with autism.

Trial registration number: NCT04292522