deadeyedan/synthetic-patient-records

---
license: apache-2.0
tags:
- healthcare
- synthetic
- medical
- ehr
- fhir
- icd-10
- patient records
- billing
pretty_name: Synthetic Patient Records
size_categories:
- 1K<n<10K

task_categories:
- text-classification
- token-classification

---

# Synthetic Patient Records — PatientDatasets.com

Pre-built synthetic patient datasets for healthcare software developers, billing teams, and ML
researchers. Instantly downloadable, commercially licensed, no IRB required.

## What's included

- **FHIR R4** — Patient, Encounter, Condition, Observation, MedicationRequest bundles
- **HL7 v2** — ADT, ORM, ORU, DFT message types
- **ICD-10-CM** — Coded diagnoses across 60+ specialties with comorbidities
- **CSV / JSON** — Flat files ready for import into any system
- **Apache Parquet** — Columnar format for Spark, Pandas, DuckDB, BigQuery
- **Physical Therapy** — CPT-coded PT episodes with functional outcome measures
- **Chiropractic** — CMT codes 98940–98943, Medicare AT modifier, spinal diagnoses

## Why PatientDatasets vs. MIMIC or Synthea

MIMIC-III requires PhysioNet credentialing — a 2 to 4 week wait, research-only license, no
commercial use. Synthea requires Java, a local install, and generates generic data with no
billing structure. Neither gives you ready-to-use, commercially licensed patient records you
can drop into a billing system or EMR today.

PatientDatasets is instant download, commercially licensed, and built around the coding and
billing structure your software actually processes — CPT codes, ICD-10-CM, 837P EDI, FHIR R4
bundles, HL7 v2 messages.

## Access

Full datasets at **[patientdatasets.com](https://patientdatasets.com)** — free 5-patient
sample included. Paid tiers from $49, one-time purchase, no subscription.