hie-synthetic-data

This repo contains:

• A set of clinical terms for substance abuse and mental health, mapped into a knowledge graph. These terms are called Part 2 data, in reference to 42 Code of Federal Regulations Part 2, which is intended to safeguard this kind of sensitive health information. More about HIPAA and Part 2 here.
• A Medical Frailty knowledge graph (MedicalFrailty.ttl) containing value sets and codes (CPT, HCPCS, ICD-10-CM) used to identify medical frailty, sourced from NCQA and IPRO value sets.
• Scripts to generate synthetic claims data. The data is output in ttl format, and can then be linked to the Part 2 graph on SNOMED-CT terms.

All knowledge graph namespaces use https://medicaidpolygraph.com as the base URI.

HIPAA Part 2 Knowledge Graph

We classified the NIH value sets to identify clinical terms associated with Substance Abuse and Mental Health as shown in the table. The complete list of terms is available here in Excel.

Clinical Terminology	Diagnosis	Finding	History	Lab	Medication	Procedure	Total
CPT1	0	0	0	0	0	48	48
HCPCS2	0	0	0	0	0	59	59
ICD-10-CM3	893	0	0	0	0	0	893
ICD-9-CM4	522	0	0	0	0	0	522
LOINC5	0	0	0	1311	0	0	1311
RXNORM6	0	0	0	0	129	0	129
SNOMED-CT7	1538	493	17	0	0	0	2048
Grand Total	2953	493	17	1311	129	107	5010

[1]: Current Procedural Terminology
[2]: Healthcare Common Procedure Coding System
[3]: International Classification of Diseases, 10th Revision, Clinical Modification
[4]: International Classification of Diseases, 9th Revision, Clinical Modification
[5]: Logical Observation Identifiers Names and Codes
[6]: RxNorm
[7]: Systematized Nomenclature of Medicine - Clinical Terms

We translated this categorization into the knowledge graph:

This results in the ability to link to claims that contain clinical terms:

An example:

Medical Frailty Knowledge Graph

The MedicalFrailty.ttl file contains a knowledge graph of medical frailty value sets and codes, generated from the following source spreadsheets:

Value Set	OID	Code System	Type	Steward	Codes
Frailty Device	2.16.840.1.113883.3.464.1003.118.11.1114	HCPCS	Extensional	NCQA	83
Frailty Encounter	2.16.840.1.113883.3.464.1003.101.11.1279	HCPCS	Extensional	NCQA	21
Frailty Encounter	2.16.840.1.113883.3.464.1003.101.12.1088	CPT / HCPCS	Grouping	NCQA	23
ICP Frailty	2.16.840.1.113762.1.4.1290.108	HCPCS / ICD-10-CM	Grouping	IPRO	383
ICP Frailty (HCPCS)	2.16.840.1.113762.1.4.1290.107	HCPCS	Extensional	IPRO	98

The graph uses the namespace mf: (https://medicaidpolygraph.com/medical_frailty#) and models:

• Value Sets with metadata (OID, steward, FHIR URI, version, purpose)
• Codes with description, code system, version, OID, and links back to parent value sets
• Codes appearing in multiple value sets are deduplicated with multiple mf:belongsToValueSet relationships

Total: 391 unique codes, 3,977 RDF triples.

Generating Ad Hoc Claims data in RDF Format

In the ad-hoc-claims folder are shell and python scripts for generating data in Synthea-RDF format.

• Shell scripts for generating synthetic data for six hospitals
• Python scripts to convert synthetic data to RDF format