Dataset of anonymized clinical events and notes from obstetric electronic health records for NLP

Interoperability remains a global challenge in healthcare systems. Addressing the need for standardized health data to ensure compatibility would enable large-scale clinical data analysis and systematic studies across heterogeneous databases, generating statistical information to support the analysis of health conditions. Healthcare data documented in Spanish face a critical limitation: the limited availability of standardized, representative, and publicly accessible real-world clinical datasets to support the development and evaluation of data-driven approaches, including Natural Language Processing (NLP) applications for treatment support, diagnosis, patient monitoring, and observational research. Analyses conducted on proprietary hospital data models are often subject to population-specific biases. Moreover, artificial intelligence applications, such as NLP using large language models (LLMs), face significant constraints due to heterogeneous clinical workflows and non-standardized storage schemas. Adherence to standardized data models, supported by syntactic and semantic standardization processes, facilitates multicenter and large-scale observational research. In this work, we present a Spanish-language dataset comprising 23,879 electronic health records (EHRs), each corresponding to a unique patient treated in a gynecology and obstetrics setting at the Clínica Universitaria Bolivariana, a national reference institution in Colombia. The dataset was standardized according to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (version 5.4). It includes both structured and unstructured patient data from deliveries attended between 2015 and 2017. The dataset contains >3.7 million anonymized clinical notes in Spanish natural language, preserving confidentiality and privacy. It integrates 89,893 clinical events from which 200,070 observations related to diagnoses and procedures are derived, together with associated clinical documentation. Standardization under the OMOP framework enables harmonization with additional data sources to address research questions and to generate clinical evidence for large-scale observational studies. This dataset supports reuse in research focused on the development, training, and evaluation of tools-particularly in clinical NLP in Spanish-as well as comparative studies between structured data and free-text clinical documentation in large-scale healthcare settings, thereby contributing evidence grounded in real-world medical practice.