Hospital-Acquired Infection (HAI) Prediction Dataset: HAI-release 1.0.1

🏥 Hospital-Acquired Infection (HAI) Prediction Dataset A comprehensive dataset of in-patient admission records from emergency and hospital wards, including demographic information, vital signs, laboratory results, and clinical outcomes. 📊 Overview This dataset includes 119,743 hospital admission records with 32 clinical and demographic features. The data is derived from real‑world hospital records spanning the years June 2023 – February 2026 at Imam-Reza Hospital, Kermanshah, Iran. The de-intefied dataset is included in a file, named Hospital_infection_data.csv. General Statistics Total Records: 119,743 Total Features: 32 Time Period: June 2023 – February 2026 Unique Complaints: 901 types 📋 Feature Description Demographic Information Column Description Data Type Missing % Age Patient's age (years) Float 0.01% Sex Gender (Male/Female) String 0% Admission and Ward Information Column Description Data Type Missing % Year Admission year (Solar Hijri) String 0% Department Hospital ward String 0.01% Triage level Triage level (1 to 5) Integer 0% Patient complaint Chief complaint String 4.1% diagnosis ICD‑10 diagnosis code String 63.2% Vital Signs Column Description Data Type Missing % SPo2 Blood oxygen saturation (%) Float 0.63% BPMin Diastolic blood pressure (mmHg) Float 2.27% BPMax Systolic blood pressure (mmHg) Float 2.27% PR Pulse rate (bpm) Float 0.75% RR Respiratory rate (per minute) Float 87.45% T Body temperature (°C) String 1.51% Blood Tests Column Description Data Type Missing % Unit BS Random blood sugar Float 74.02% mg/dL BS (second) Blood sugar (second measurement) Float 4.70% mg/dL WBC White blood cell count Float 2.48% ×10³/µL RBC Red blood cell count Float 93.21% — HB Hemoglobin Float 0.59% g/dL HCT Hematocrit Float 0.59% % PLT Platelet count Float 0.66% ×10³/µL ESR Erythrocyte sedimentation rate Float 81.73% mm/hr CRP C‑reactive protein String (Qualitative) 53.82% — Blood Biochemistry Column Description Data Type Missing % Unit UREA Blood urea Float 1.26% mg/dL CR Creatinine Float 1.17% mg/dL NA Sodium Float 16.12% mEq/L K Potassium Float 19.21% mEq/L Coagulation Column Description Data Type Missing % PT Prothrombin time String 99.25% Target Variable Column Description Data Type Missing % Label Clinical outcome (0 or 1) Integer 0% ⚠️ Data Quality Notes Missing Data Some columns have significant percentages of missing data: Status Missing % Columns Critical > 80% RBC (93.2%), PT (99.3%), ESR (81.7%), RR (87.5%) High 50–80% BS (74%), CRP (53.8%), diagnosis (63.2%) Medium 10–20% NA (16.1%), K (19.2%) Good < 10% Other columns Class Imbalance The target variable (Label) is imbalanced: Negative Class (0): 98.62% (118,083 cases) Positive Class (1): 1.38% (1,660 cases) 📝 Dataset Usage, Validation and Code We validated the dataset using classical machine learning methods to establish baseline performance for hospital-acquired infection (HAI) prediction on a subset of the dataset. The code is available in the CODE/ subdirectory, in Python. Requirements The scripts require the following Python packages: pandas numpy matplotlib seaborn scikit-learn xgboost Install them with: pip install pandas numpy matplotlib seaborn scikit-learn xgboost Running the Code Navigate to the CODES/ folder and run: python CODE.py The script will: Preprocess the data Train classifiers (Logistic Regression, Random Forest, AdaBoost, XGBoost) Evaluate performance using cross-validation Generate metrics and visualizations 🧑‍⚖️ License This repository contains two distinct components with different licenses: Source Code (Python scripts): Licensed under the MIT License – see the LICENSE file. Dataset (CSV files): Licensed under the Creative Commons Attribution‑NonCommercial 4.0 International License (CC BY‑NC 4.0) – see the LICENSE_DATA file. This license permits non‑commercial use, sharing, and adaptation, provided appropriate credit is given. Commercial use (including selling, licensing, or using the data in proprietary commercial products) is strictly prohibited without explicit written permission from the authors. Users must not attempt to re‑identify any patients, although all data has been de-identified. Additional Terms of Use: By downloading, accessing, or using this dataset, you agree to the following additional terms: Citation Requirement: You must cite the associated Data Descriptor paper in any publication, presentation, or product that uses or references this dataset: Bakhshi, Z., Ghasemi, V., & Zamanian, M. H. (2026). A Large‑Scale Hospital Dataset for Early Prediction of Hospital‑Acquired Infections Based on Admission and First‑Day Clinical Data. Submitted to The Journal of Biomedical Physics and Engineering. [DOI: To be Inserted] Dataset Citation (for data access): Bakhshi, Z., Ghasemi, V., & Zamanian, M. H. (2026). A Large‑Scale Hospital Dataset for Early Prediction of Hospital‑Acquired Infections Based on Admission and First‑Day Clinical Data (Version 1.0.0) [Dataset]. Available online: https://doi.org/10.5281/zenodo.21777046 Attribution: As required by the CC BY-NC 4.0 license, you must give appropriate credit to the dataset creators and provide a link to the license. 🛡️ Ethical Approval This study was conducted in compliance with the Declaration of Helsinki and approved by the Ethics Committee of the School of Medicine, Kermanshah University of Medical Sciences under approval code: IR.KUMS.MED.REC.1404.284 All patient data were fully anonymized prior to extraction, and the requirement for informed consent was waived by the Ethics Committee due to the retrospective nature of the study. 👥 Contributors Zahra Bakhshi, MSc – Department of Computer Engineering, Faculty of Information Technology, Kermanshah University of Technology, Kermanshah, Iran. Dr. Vahid Ghasemi, PhD – Department of Computer Engineering, Faculty of Information Technology, Kermanshah University of Technology, Kermanshah, Iran. Dr. Mohammadhossein Zamanian, PhD – Clinical Research Development Center, Imam-Reza Hospital, Kermanshah University of Medical Sciences, Kermanshah, Iran. 📧 Contact and Collaboration For questions, suggestions, or collaboration on research projects based on this data: ✉️ Email (Corresponding Author): [e-mail] 📧 Email (Supervisor): [e-mail] 📧 Email (Advisor): [e-mail]