Note NLP

Epic equivalent: No direct equivalent — this is NLP pipeline output derived from clinical notes

The note_nlp table contains structured outputs from NLP models applied to clinical notes. Each row is a single extracted entity, concept, assertion, or relationship found in the free text stored in the note table. Think of it as the bridge between narrative documentation and structured analytics.

Field Reference

Field reference (click to expand)

OMOP Field	What It Captures
note_nlp_id	Unique identifier for the NLP output row
note_id	Links back to the source note
section_concept_id	Note section (e.g., HPI, Assessment) for context
snippet	Short text window around the concept mention
offset	Character position within the note
lexical_variant	Actual words used (e.g., "SOB" or "shortness of breath")
note_nlp_concept_id	Standard OMOP concept inferred from the text
note_nlp_source_concept_id	Source vocabulary concept (UMLS, MedLEE, etc.)
nlp_system	NLP engine used (cTAKES, MedLEE, BERT, etc.)
nlp_date	When the NLP pipeline was run
term_exists	Affirmed (1), negated (0), or uncertain
term_temporal	Timing: current, history of, planned
term_modifiers	Additional qualifiers (severity, etc.)

What to Watch For

Common pitfalls

One concept per row

Multiple entities in one sentence produce multiple rows. Don't assume 1:1 with notes.

Tool-dependent structure

Different NLP pipelines populate different fields. Know what your pipeline supports.

Limited population at Emory

This table is project-specific at present. Check Known Issues for current status.

Research Patterns

Question	Tables Involved
Symptom mentions in COVID-era ER notes	note_nlp.note_nlp_concept_id + note.note_type_concept_id + date filter
Undocumented comorbidities found in text	note_nlp (term_exists = 1) cross-checked against condition_occurrence
Goals-of-care documentation frequency	note_nlp.lexical_variant or note_nlp_concept_id for palliative terms
NLP pain assessments vs. structured scores	note_nlp + measurement (pain scale)
Functional decline trends across progress notes	note_nlp_concept_id (functional status) + nlp_date