Background : Patient information in rare disease registries is generally collected from numerous data sources, necessitating the data to be federated. In addition, data for research purposes must be de-identified. Transforming nominative data into de-identified data is thus a key issue, while minimizing the number of identity duplicates. We propose a method enabling patient identity federation and rare disease data de-identification while preserving the pertinence of the provided data. Results : We developed a rare disease patient identifier. The IdMR generation process is a three-phased algorithm involving a hash function to irreversibly de-identify nominative patient data, including those of foetuses. This process minimizes collision risks and reduces variability for the purpose of identity federation. The IdMR was generated for 360,000 patients of the CEMARA database. It allowed identity federation of 1771 duplicated files. No collisions were introduced. Conclusion : We examined and discussed the risks of collisions and the creation of duplicates as well as the risks of patient re-identification. We discussed our choice of nominative input information in light of that used by other patient identification solutions. The IdMR is a patient identifier that enables identity federation and file linkage. The simplicity of the algorithm and the universality and stability of the input data make it a good candidate for European cross-border rare disease projects.