ASPIR-CLIN Dataset

Overview The ASPIR-CLIN Dataset is a clinical dataset designed to support research involving aspiration pneumonia, dysphagia assessment, swallowing disorders, and associated clinical risk factors. The dataset contains structured clinical information collected from patients evaluated through multiple clinical instruments and screening procedures, including swallowing assessment scales and hospitalization-related variables. The primary goal of this dataset is to facilitate: * Clinical research involving aspiration pneumonia;* Dysphagia risk analysis;* Statistical analyses of swallowing-related conditions;* Machine learning applications in healthcare;* Development of predictive clinical models. Dataset Contents The repository currently contains the following files: aspir_clin_dataset.csv = Main clinical datasetaspir_clin_data_dictionary.csv = Variable descriptions and coding informationaspir_clin_loader.py = Scikit-learn-compatible Python dataset loaderREADME.md = Dataset documentation Clinical Variables The dataset includes variables associated with: * Demographic information;* Feeding route;* Swallowing assessments;* Neurological conditions;* Respiratory conditions;* Cardiac conditions;* Nutritional indicators;* Functional limitations;* ICU and hospitalization information;* Aspiration pneumonia risk factors. Examples of included clinical instruments: * FEES;* FOIS;* EAT-10;* Glasgow Coma Scale. Data Collection Institution * Tuiuti University of Paraná Data Collection Period The sample was collected between May 2024 and March 2025 through a retrospective analysis of electronic medical records referring to the years 2022 and 2023. The data were extracted from the Tasy electronic system used by Hospital Nossa Senhora das Graças (HNSG), located in Curitiba, Paraná, Brazil, a reference institution for high-complexity clinical and surgical treatments. Population Hospitalized adults. Inclusion Criteria The study included patients of both sexes, aged 18 years or older, who underwent the institutional protocol for the investigation of bronchoaspiration and who underwent a Fiberoptic Endoscopic Evaluation of Swallowing (FEES) during hospitalization and/or a speech-language pathology clinical assessment. Exclusion Criteria The exclusion criteria included the absence of a speech-language pathology assessment record in the electronic medical record, incomplete data regarding the established variables, and age under 18 years. Ethical Considerations All patient information was anonymized prior to dataset publication. This study was conducted according to institutional ethical guidelines. Ethics Committee Approval Approved by the Research Ethics Committee under protocol number 75183123.9.3001.0269 Anonymization The following anonymization procedures were applied: * Removal of direct patient identifiers;* Replacement of medical record identifiers by synthetic patient IDs (`P001`, `P002`, etc.);* Removal of personally identifiable information. Dataset Structure Main Table Format * One row corresponds to one patient;* One column corresponds to one clinical variable. Python Loader The repository includes a Python loader (`aspir_clin_loader.py`) designed to provide a scikit-learn-compatible interface for loading the ASPIR-CLIN Dataset. The loader returns a `Bunch` object similar to datasets available in the scikit-learn library (e.g., the Iris dataset), providing direct access to: * `data`: feature matrix;* `target`: target variable;* `feature_names`: predictor variable names;* `target_names`: target class names;* `frame`: complete pandas DataFrame;* `DESCR`: dataset description. Example from aspir_clin_loader import load_aspir_clin dataset = load_aspir_clin("aspir_clin_dataset.csv") print(dataset.data.shape) print(dataset.target.shape) print(dataset.feature_names) print(dataset.target_names) print(dataset.DESCR) By default, the loader uses `penetration_risk` as the target variable, making the dataset readily applicable to machine learning, statistical analysis, and clinical decision support research. Missing Values No missing values were identified in Version 1.0 of the dataset. Future versions may include missing values represented as NA or empty cells. Variable Coding Detailed variable descriptions and codifications are available in: aspir_clin_data_dictionary.csv Potential Applications The ASPIR-CLIN Dataset may support studies involving: * Aspiration pneumonia prediction;* Dysphagia severity analysis;* Statistical analysis of clinical risk factors;* Explainable artificial intelligence in healthcare;* Clinical decision support systems;* Machine learning and data mining applications. Funding Coordination for the Improvement of Higher Education Personnel (CAPES) was responsible for granting the study scholarship to the lead author. Contact Information Rosane Sampaio Tuiuti University of Paraná[e-mail] Recommended Citation If you use the ASPIR-CLIN Dataset in your research, please cite it as: Ramos, M. E. M., Tomasi Junior, D. L., & Sampaio, R. (2026). ASPIR-CLIN Dataset: Aspiration Pneumonia Clinical Dataset (Version 1.0) [Data set]. Zenodo. https://10.5281/zenodo.20527814 Dataset Authors - Maria Eduarda Medeiros Ramos- Darci Luiz Tomasi Junior- Rosane Sampaio Acknowledgments We would like to thank the Health Artificial Intelligence Center (NIAS) for the support in the development of this study, and the Coordination for the Improvement of Higher Education Personnel (CAPES) for granting a study scholarship. License This dataset is distributed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). https://creativecommons.org/licenses/by/4.0/ Disclaimer This dataset is intended exclusively for scientific and research purposes. The authors are not responsible for clinical decisions derived from the use of this dataset.