NOMAD taxonomy classifies 22K+ medical disease names
ICD-10-CM disease names analyzed with 99.1% classification accuracy
Deep Dive
Researchers built NOMAD (Nomenclature Ontology for Medical And Disease names), a meta-taxonomy classifying 22,548 ICD-10-CM disease names into 9 top-level categories and 20 subcategories. Their ML pipeline classified 99.1% of terms with a mean of 2.12 labels per entry. Manual review of 2,255 entries showed 70% full agreement and 26% partial agreement (Cohen's Kappa 0.832). The system exposes gender bias in disease naming—only 57 (2.6%) of eponymous diseases were named after a female person.
Key Points
- NOMAD classifies 22,548 ICD-10-CM disease names into 9 categories and 20 subcategories with 99.1% accuracy
- Anatomical (63.8%), Descriptive (48.4%), and Pathophysiological (40.2%) labels dominate; only 57 diseases (2.6%) eponyms named after females
- ML pipeline achieves 70% full agreement and 26% partial agreement (Cohen's Kappa 0.832) on 10% manual review
Why It Matters
Standardizing disease naming improves clinical data interoperability and exposes systemic naming biases in medical literature.