BRATECA (Brazilian Tertiary Care Dataset): a Clinical Information Dataset for the Portuguese Language

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5609–5616 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5609 BRATECA (Brazilian Tertiary Care Dataset): a Clinical Information Dataset for the Portuguese Language Bernardo S. Consoli1, Henrique D. P. dos Santos2, Ana Helena D. P. S. Ulbrich2, Renata Vieira3, Rafael H. Bordini1 1Pontifical Catholic University of Rio Grande do Sul (PUCRS), Brazil 2Institute for Artificial Intelligence in Healthcare, Brazil 3CIDEHUS, University of ´Evora, Portugal bernardo.consoli@acad.pucrs.br, {henrique, anahelena}@noharm.ai, renatav@uevora.pt, rafael.bordini@pucrs.br Abstract Computational medicine research requires clinical data for training and testing purposes, so the development of datasets composed of real hospital data is of utmost importance in this field. Most such data collections are in the English language, were collected in anglophone countries, and do not reflect other clinical realities, which increases the importance of national datasets for projects that hope to positively impact public health. This paper presents a new Brazilian Clinical Dataset containing over 70,000 admissions from 10 hospitals in two Brazilian states, composed of a sum total of over 2.5 million free-text clinical notes alongside data pertaining to patient information, prescription information, and exam results. This data was collected, organized, deidentified, and is being distributed via credentialed access for the use of the research community. In the course of presenting the new dataset, this paper will explore the new dataset’s structure, population, and potential benefits of using this dataset in clinical AI tasks. Keywords: computational medicine, portuguese language, clinical data, tertiary care 1. Introduction Decision making in healthcare scenarios has been a topic of increasing interest in the field of artificial intel- ligence (Shamout et al., 2021; Si et al., 2021). Studies into methods for predicting diseases (Song et al., 2018; Xu et al., 2018), mortality (Xu et al., 2018), length- of-stay (Song et al., 2018; Xu et al., 2018), admis- sion (Liu et al., 2019), and interventions (Suresh et al., 2017) have become more common with the widespread use of EHRs (Electronic Health Records) in hospitals worldwide, which in turn led to efforts to deidentify this information, as it is sensitive and extremely per- sonal, in order to make it available for use in related research (Shamout et al., 2021). These efforts have resulted in several clinical datasets, most of which were constructed from sources in the English language. The MIMIC (Medical Information Mart for Intensive Care) collection, composed of data extracted from the Beth Israel Deaconess Medical Cen- ter in the United States, is a part of PhysioNet (Gold- berger et al., 2000) and is the foremost example of such datasets in computational medicine and focuses on in- tensive care patients. Many papers (Song et al., 2018; Xu et al., 2018; Liu et al., 2019; Suresh et al., 2017) make use of this collection to build training and test- ing datasets for deep learning architectures. MIMIC- IV (Johnson et al., 2020), the latest edition of the collection, possesses comprehensive information about each patient, including laboratory measurements, ad- ministered …

Content truncated to 500 words