A Curated Multimodal Dataset for Sleep Apnea and Cardiometabolic Comorbidities (Healthcare)
Description
This dataset represents the Gold Layer of a Sleep Apnea Data Warehouse developed using a Medallion Architecture (Bronze–Silver–Gold) in Microsoft SQL Server. It contains 10,044 unique patient records and 47 curated analytical features integrating demographic, physiological, clinical, lifestyle, and diagnostic data for research on obstructive sleep apnea (OSA) and metabolic comorbidities. The dataset was generated by integrating multiple structured health data sources through ETL processes, dimensional modeling, and feature engineering into a unified star-schema warehouse. Each record corresponds to a single patient and includes demographics (age, gender, occupation), metabolic indicators (BMI, glucose, insulin, HbA1c, cholesterol), cardiovascular variables (blood pressure, heart rate), sleep-related physiological measurements (AHI, oxygen saturation, EEG sleep stage, nasal airflow, chest movement), lifestyle indicators (physical activity, stress, diet, alcohol use), and diagnostic labels for sleep apnea, hypertension, and diabetes. The Gold Layer includes engineered variables such as age bands, BMI categories, comorbidity profiles, binary health flags, and standardized analytical features optimized for machine learning and clinical analytics. The repository was designed to support predictive modeling, multi-label classification, risk stratification, clustering, and healthcare business intelligence applications. Exported in CSV format with UTF-8 encoding, the dataset is compatible with Python, R, SQL Server, Power BI, Tableau, and statistical analysis tools. Synthetic composite identifiers are used, and no personally identifiable information is included, supporting ethical data sharing for research and educational purposes. Potential applications include OSA diagnosis prediction, comorbidity risk scoring, explainable machine learning, patient segmentation, feature engineering research, and demonstration of Medallion Architecture implementation in healthcare data warehousing. This dataset also serves as a reproducible benchmark for integrating data engineering and medical analytics workflows. Keywords: Sleep Apnea, OSA, Healthcare Analytics, Data Warehouse, Gold Layer, Medallion Architecture, Predictive Modeling, Multi-label Classification, Clinical Data Engineering.
Files
Steps to reproduce
The Gold Layer dataset was produced through a multi-stage Medallion Architecture pipeline: Bronze Layer (Raw Ingestion) Collected raw structured datasets from multiple sleep-health and metabolic source repositories (CSV files). Loaded source files into SQL Server staging tables without altering original data. Preserved raw schema, source lineage, and ingestion timestamps for traceability. Silver Layer (Cleaning and Integration) 4. Performed data profiling and quality assessment. 5. Removed duplicates, standardized formats, corrected inconsistencies, and handled missing values. 6. Harmonized variable names, units, and data types across sources. 7. Integrated sources into a relational warehouse using dimensional modeling and star schema design. 8. Created fact and dimension tables for patients, clinical metrics, physiological signals, and lifestyle attributes. 9. Generated surrogate/composite keys and established relationships among tables. Gold Layer (Curated Analytics Layer) 10. Joined cleaned dimensional and fact tables into a denormalized analytical feature mart. 11. Engineered derived variables such as age_band, bmi_category, comorbidity_profile, diagnostic flags, and encoded predictors. 12. Applied business and clinical transformation rules for feature enrichment and standardization. 13. Selected high-value analytical attributes and removed redundant operational fields. 14. Validated consistency, integrity, and feature distributions. 15. Exported the final Gold Layer as a curated flat dataset containing 10,044 records and 47 features optimized for analytics, machine learning, and research. The resulting Gold Layer represents the final trusted analytical dataset, balancing data quality, integration, reproducibility, and modeling readiness.
Institutions
- New Mansoura UniversityDakahlia, Al Mansurah
- Mansoura UniversityDakahlia, Al Mansurah