SpCN-Synth - Spanish Synthetic Clinical Notes

The dataset is comprised of Spanish synthetic clinical notes. Each one of them has their primary diagnoses annotated with corresponding character offsets, along with to their International Statistical Classification of Diseases (ICD-10-CM) [1] code. There is a total of 1000 clinical notes divided into two file types: .txt Here are all the artificial notes as plain text files. .ann Here are the annotated entities in BRAT style. The terminology used is the Unified Medical Language System (UMLS) [2], where the diagnoses are given the DISO label. All of them have specific character offsets that correspond to the .txt file, in addition to the corresponding ICD-10-CM code. Each clinical note has one of each file type. Both .txt and .ann files were made by prompting the OpenEvidence model [3] during March/April of 2026. The task was developed by generating one text with the offsets of their diagnoses marked up. In the prompting phase, a total of 1 to 9 different ICD-10-CM codes were inputted in order for the model to generate an artificial clinical note composed of those codes. Access to the data is granted upon request to the authors and by accepting OpenEvidence terms and conditions [4]. [1] Ministerio de Sanidad, “Clasificación estadística internacional de enfermedades y problemas de salud relacionados, 10a revisión, modificación española (CIE-10-ES),” Jan. 2026. [Online]. Available: https://www.eciemaps.sanidad.gob.es/browser/metabuscador [2] O. Bodenreider, “The Unified Medical Language System (UMLS): integrating biomedical terminology,” Nucleic Acids Research, vol. 32, no. suppl 1, pp. D267–D270, Jan. 2004. [Online]. Available: https://doi.org/10.1093/nar/gkh061 [3] S. Philip and R. Kurain, “OpenEvidence,” Journal of the Medical Library Association, vol. 114, no. 1, pp. 86–87, Mar. 2026. [Online]. Available: https://jmla.pitt.edu/ojs/jmla/article/view/2247 [4] OpenEvidence, Inc., “Acceptable Use Policy.” [Online]. Available: https://patients.openevidence.com/unavailable?reason=eu#openevidence-inc-acceptable-use-policy.