Code inventory

Stage 7 of the pipeline writes code_inventory.csv next to dashboard_data.json. It lists every unique (vocabulary, code) pair seen anywhere in the bundle, along with how often it was referenced, how many distinct patients had it, the most-common display name observed, and which bundle categories it appeared in.

Columns

Column	Meaning
vocabulary	Source vocabulary as named by the pipeline (SNOMED-CT, ICD-10-CM, RxNorm, LOINC, CVX, CPT-4, FHIR encounter class, ...)
code	Source code value
display_name	Most-common display string seen for this code across all references
n_references	Total number of records (across all categories) referencing this code
n_unique_patients	Number of distinct patients who have at least one reference to this code
source_categories	Semicolon-separated list of bundle tabs this code appears in (e.g. problems;diagnostic_reports)

How codes are collected

For each record in the coded categories — problems, medications, procedures, allergies, immunizations, labs_vitals, diagnostic_reports, document_references, encounters — the pipeline walks each entry's all_codings array. A record with both a SNOMED concept and an ICD-10-CM translation contributes two rows to the inventory (one per vocabulary). When all_codings is missing, the pipeline falls back to the record's top-level code and code_system fields.

The split labs and vitals views are skipped because their rows are duplicated in the unified labs_vitals tab; including all three would triple-count the same observations.

Reading the inventory

The CSV is sorted by vocabulary, then by descending reference count, then by code. A typical first look:

import pandas as pd
inv = pd.read_csv('code_inventory.csv')
print(inv['vocabulary'].value_counts())               # codes per vocab
print(inv.sort_values('n_references', ascending=False).head(20))  # most-used codes
print(inv[inv['n_unique_patients'] >= 10].shape[0])   # codes hit by many patients

The inventory is the input to the OMOP ETL and is also useful on its own — for QA passes, to spot codes whose displays are inconsistent across records, or to size a vocabulary download from Athena.