rapid-elastic 1.0.1

Rapidly query clinical notes in Elasticsearch for mentions of diseases

rapid-elastic

Disease Synonyms Curation

• Disease names and synonyms were HUMAN expert curated and assisted via ChatGPT with additional checks within NLM GeneReviews, OrphaNet, and Google searches.
• disease_names_expanded.json is pre-built and recommended for use without modification.
• disease_names.py assists curation if changes need to be made: read spreadsheet CSV, write disease names JSON, recommend GPT4 prompts, find duplicates, and merges curated JSON.

Quickstart

1. configure ENV variables query your Elasticsearch server (within your VPN/firewall)

   1. ELASTIC_HOST : default http://localhost:9200
   2. ELASTIC_USER : basic auth
   3. ELASTIC_PASS : basic auth

2. pipx install rapid-elastic

3. rapid-elastic

Elasticsearch server

Read the server setup docs.

KQL (Kibana Query Language)

• By default, synonyms used to match any "exact phrase" in the note.
• see kql_syntax.py for alternate methods of building Elasticsearch queries.

Each Elasticsearch hit is saved to "output" folder

• subject_ref (FHIR Patient.id)
• encounter_ref (FHIR Encounter.id)
• document_ref (FHIR DocumentReference.id)
• group_name (optional)
• document_title (optional)

SQL Athena

SQL tables are generated from spreadsheet CSV files

• disease_name
• orpha_code
• icd10_code

FHIR resources are referenced using

• subject_ref (FHIR Patient.id)
• encounter_ref (FHIR Encounter.id)
• documentreference_ref (FHIR DocumentReference.id)

FHIR Condition.code is selected using

• icd10_code