Syntha — Hybrid Synthetic Patient Record Generator Released
— by Ariorad Moniri
Vivax CTO Ariorad Moniri has released Syntha, a Synthea-inspired hybrid synthetic patient record generator that learns the joint distribution of real…
Syntha is a Python library, command-line tool, and signed cross-platform desktop app for generating realistic synthetic patient records — flat CSVs and FHIR R4 transaction Bundles — that match the statistical structure of an anonymized Turkish-cohort EHR while staying physiologically valid and clinically coded.
The pipeline is hybrid: a Gaussian copula fitted on real anonymized episodes preserves marginal distributions (age, labs, vitals, comorbidity prevalence) and their joint correlation structure; a physiologic filter rejects samples that violate pulse-pressure, Friedewald lipid coherence, or eGFR ↔ creatinine constraints; and nine Synthea-style clinical modules emit Encounters, MedicationRequests (RxNorm-coded), Procedures, and CarePlans matching each patient's comorbidity profile.
Syntha ships as a Python package on PyPI (`pip install syntha`), a CLI, and signed Tauri-based desktop apps for macOS and Windows. Licensed under Apache 2.0 with 91% test coverage. FHIR R4 export includes Patient, Observation, Condition, Encounter, MedicationRequest, Procedure, CarePlan, DiagnosticReport, RiskAssessment, and FamilyMemberHistory — dual-coded LOINC / SNOMED CT / ICD-10 / RxNorm with the Turkish (tr-TR) locale.