is a Knowledge Graph.
It is part of these collections: Translator, OMOP
Open Health Data @ Carolina (OHD@Carolina) is a statistical association knowledge graph of concept prevalence and co-occurrence in electronic health records, derived from UNC Health's OMOP database on a five-year cohort of about 6 million patients (2018 through 2022). Edges link conditions, drugs, procedures and phenotypes with positive or negative correlation predicates and carry chi-squared p-values, log odds ratios and sample sizes. Concepts and pairs with counts of 10 or fewer were excluded and counts were randomized with a Poisson distribution to protect patient privacy. RENCI builds it with ORION and serves it as an Automat TRAPI knowledge provider for the NCATS Biomedical Data Translator.
MIT (inferred)
Unknown
URL: https://renci.org/
Github: YaphetKG
| ID | Name | URL | Category | Format | Description |
|---|---|---|---|---|---|
| ohd-carolina.graph | OHD@Carolina Automat KGX Graph | f627ebbefd242454 | Graph | kgx-jsonl | KGX JSONL nodes and edges files for t... |
| ohd-carolina.neo4j | OHD@Carolina Automat Neo4j Dump | graph_f627ebbefd242454.db.dump (1.6 GB) | Graph | neo4j | Neo4j database dump of the OHD@Caroli... |
| ohd-carolina.trapi | OHD@Carolina Automat TRAPI API | docs | Programming Interface | http | Automat TRAPI endpoint for OHD@Caroli... |
| ohd-carolina.metadata | OHD@Carolina Meta Knowledge Graph | meta_knowledge_graph.json (547.1 KB) | Product | json | Meta knowledge graph JSON listing the... |
| ohd-carolina.source-edges | OHD@Carolina Source Edge Table | unc_omop_2018_2022_kg.zip (1.9 GB) | Product | csv | Source edge table (unc_omop_2018_2022... |
| ID | Name | URL | Category | Format | Relation | Description |
|---|---|---|---|---|---|---|
| automat.ohd-carolina | OHD_Carolina_Automat | OHD_Carolina_Automat | Graph | kgx-jsonl | had primary source | OHD Carolina Automat: per-source KGX ... |
Open Health Data @ Carolina (OHD@Carolina) exposes counts, frequencies and co-occurrence statistics of clinical concepts from UNC Health’s OMOP database. The cohort covers about 6 million UNC Health patients over 2018 through 2022, including inpatient and outpatient visits. Concepts are coded by their OMOP standard concept IDs and normalized to Biolink-compatible identifiers (RxNorm, UMLS, MONDO, ChEBI, HP and others) in the knowledge graph.
It follows the approach of COHD, the Columbia Open Health Data service, applied to UNC Health data. Counts for each concept include patients from all descendant concepts. To protect patient privacy, concepts and concept pairs with counts of 10 or fewer were excluded and counts were randomized with a Poisson distribution.
The Automat graph built on 2025-09-28 (build f627ebbefd242454, published 2025-10-06) holds 27,356 nodes and 22,732,570 edges, all from infores:openhealthdata-carolina. Every edge is a statistical association (biolink:positively_correlated_with or biolink:negatively_correlated_with) with p_value, log_odds_ratio, log_odds_ratio_95_ci and total_sample_size attributes.
The source infores is infores:openhealthdata-carolina. The Automat graph is infores:automat-openhealthdata-carolina, consumed by ARAGORN.
Created: October 04, 2026 | Last modified: October 04, 2026