Clinical notes hold rich yet unstructured details about diagnoses, treatments, and outcomes that are vital to precision medicine but hard to exploit at scale. We introduce a method that represents each patient as a matrix built from aggregated embeddings of all their notes, enabling robust patient similarity computation based on their latent low-rank representations. Using clinical notes of 4,267 Czech breast-cancer patients and expert similarity labels from Masaryk Memorial Cancer Institute, we evaluate several matrix-based similarity measures and analyze their strengths and limitations across different similarity facets, such as clinical history, treatment, and adverse events. The results demonstrate the usefulness of the presented method for downstream tasks, such as personalized therapy recommendations or toxicity warnings.
@article{arxiv.2601.07385,
title = {Computing patient similarity based on unstructured clinical notes},
author = {Petr Zelina and Marko Řeháček and Jana Halámková and Lucia Bohovicová and Martin Rusinko and Vít Nováček},
journal= {arXiv preprint arXiv:2601.07385},
year = {2026}
}
Comments
This is a preprint and has not undergone peer review. Final version was presented at the Text, Speech, and Dialogue 2025 conference. The Version of Record is available at https://doi.org/10.1007/978-3-032-02551-7_13