Synthetic Hospital
What it is
An open, fully synthetic longitudinal electronic health record (EHR) benchmark designed to test clinical AI models without using protected health information. It consists of 1,268 patients and 5,602 encounters grounded in standard medical ontologies (ICD-10-CM, SNOMED CT, LOINC) and served through a simulated hospital system with standard interoperability APIs.
The problem it solves
Real EHR data cannot be openly shared due to privacy and ethics constraints, and it lacks verifiable ground truth because it only reflects clinician documentation rather than objective facts.
What you could use it for
- Evaluating the performance of large language models on longitudinal patient care tasks
- Benchmarking clinical AI models against physician-level performance
- Developing and testing medical AI agents in a safe, privacy-preserving environment
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.