TinyEHR: A 100-Patient Synthetic Electronic Health Record Dataset for Prototyping

TinyEHR provides a 100-patient synthetic dataset of Electronic Health Records designed for learning and prototyping healthcare AI tools. The dataset is derived from the MIMIC-IV Demo source and is free to use, bypassing typical credentialing and data access agreements required for real healthcare data. It was created by vidulpanickan and last updated on April 3, 2026.

Use Cases

• Prototyping AI agentic systems for healthcare based on synthetic patient records.
• Learning to handle and process Electronic Health Record data structures.
• Experimenting with clinical data analysis tools without requiring real data access agreements.
• Developing models for table question answering tasks using clinical tabular data.

Strengths

• Contains records for 100 synthetic patients, providing a manageable scale for experimentation.
• Derived from the MIMIC-IV Demo source, suggesting a realistic structure.
• Explicitly intended for free use without credentialing barriers.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known (100 patients), but the number of records or events per patient is unspecified.
• The dataset is synthetic and explicitly not for clinical analysis or medical decision-making.

Provenance

Source

Derived from the PhysioNet MIMIC-IV Demo dataset.

Collection Method

Synthetically built for learning and prototyping.

Time Range

null

Freshness

Last updated 2026-04-03 18:58:57.

Geography

Region is tagged as 'us'.