Surgical 1K: Synthetic Medical Records for Model Training

WitnessDataFactory provides a 1,000-record sample of enterprise-grade synthetic medical data. The dataset is described as HIPAA-aligned and contains zero Protected Health Information (PHI). It was last updated on 2026-05-12.

Use Cases

• Training predictive models for surgical outcomes based on synthetic clinical features.
• Benchmarking data generation and anonymization techniques using the described quality metrics.
• Developing and validating healthcare NLP models on HIPAA-aligned synthetic text data.

Strengths

• The dataset is explicitly free of PHI, addressing a major privacy concern.
• Quality metrics are provided, including a Trinity Consensus Score of 98.0% and a Macro F1 score of 0.97.
• Data is generated via a 3-LLM Trinity Ensemble method, which may indicate higher synthetic fidelity.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is limited to the 1,000-record sample, which may restrict model training scale.

Provenance

Source

WitnessDataFactory

Collection Method

Synthetic generation via 3-LLM Trinity Ensemble.

Freshness

Last updated 2026-05-12 21:08:14; freshness should be verified.