Synthetic OMOP Healthcare Dataset for Data Quality and Degradation Experiments

A deliberately degraded version of a synthetic population of 1,181 persons, generated with Synthea (v3.3.0) and converted to the OMOP CDM v5.4 (Synthea ETL v2.1.1). The original dataset is version 1 of this record. No real patient data are included. The dataset is intended for studying how data quality issues affect analyses and how they can be reported, not for analyses that assume valid data. Files synthea_omop_degraded.dump is a database dump of the OMOP CDM with three introduced quality problems: unmapped diabetes diagnoses, mislabelled HbA1c units and placeholder birth dates. report-dbb3f37f-beaa-4160-bb8f-a9c161eb3e67.ttl is a quality report for this dataset in the W3C Data Quality Vocabulary (Turtle format). It was produced with the Federated Data Quality Framework and gives the results of ten checks, protected with differential privacy (ε = 3.0), so the values are approximate. Check definitions are available at https://github.com/RadovanTomik/fair-but-unfit. Accompanies the manuscript “FAIR but unfit: privacy-preserving quality control for sensitive scientific datasets” (in preparation).