icdlookup

A Python package for looking up ICD codes and mapping them to human-readable descriptions.

Ships with a built-in ICD-10-CM 2026 dataset sourced from the official CMS ICD-10 files. The hierarchy approach was inspired by rjake/ICD10-hierarchy.

The package name is version-agnostic — future releases may extend support to ICD-9, ICD-11, and other coding systems.

Installation

pip install icdlookup

For DataFrame annotation support:

pip install icdlookup[pandas]

Quick Start

Look up a single code

import icdlookup

icdlookup.lookup("A00.1")
# → "cholera due to vibrio cholerae 01, biovar eltor"

Get full hierarchy detail

icdlookup.lookup("A00.1", detail=True)
# → {
#     "icd10_code": "A00.1",
#     "description": "cholera due to vibrio cholerae 01, biovar eltor",
#     "chapter_desc": "Certain infectious and parasitic diseases (A00-B99)",
#     "section_desc": "Intestinal infectious diseases (A00-A09)",
#     "category_desc": "cholera"
#   }

Search by keyword

icdlookup.search("cholera")
# → [("A00", "cholera"), ("A00.0", "cholera due to ..."), ...]

Annotate a pandas DataFrame

import pandas as pd
import icdlookup

df = pd.DataFrame({"icd_code": ["A00.1", "J06.9", "E11.9"]})
df = icdlookup.annotate(df, code_column="icd_code")
# df now has an "icd_description" column

Use a custom reference CSV

Any CSV with at least icd10_code and description columns works:

icdlookup.lookup("A00.1", reference="/path/to/custom.csv")
icdlookup.annotate(df, code_column="icd_code", reference="/path/to/custom.csv")

CLI

# Single lookup
icdlookup A00.1

# Full detail
icdlookup A00.1 --detail

# Search by keyword
icdlookup --search cholera

# Annotate a CSV file
icdlookup --file data.csv --column icd_code --output annotated.csv

# Use custom reference
icdlookup A00.1 --reference /path/to/custom.csv

Design

• No pandas required for core functions — lookup() and search() use only the standard library (csv + gzip). pandas is only needed for annotate().
• Case & format insensitive — lookups strip whitespace, are case-insensitive, and work with or without dots (A001 and A00.1 both match).
• Lazy-loaded & cached — the built-in CSV is only parsed on first use, then kept in memory for the session.
• Custom reference support — pass any CSV with icd10_code + description columns.
• 48,600+ ICD-10-CM codes — the bundled 2026 dataset includes full hierarchy info (chapter, section, category descriptions).

Data Source

The bundled dataset is derived from the official CMS ICD-10-CM 2026 Code Tables, specifically the tabular XML file. Each code includes:

Rebuilding the Dataset

To regenerate the bundled data from the latest CMS release:

python scripts/build_data.py

This downloads the CMS XML zip, parses the hierarchy, and writes the compressed CSV to src/icdlookup/_data/.

Development

git clone https://github.com/icdlookup/icdlookup.git
cd icdlookup
pip install -e ".[dev]"
pytest

License

MIT