How to intermediate · 4 min read

ICD code extraction with AI

Q: ICD code extraction with AI

Use a large language model like gpt-4o to extract ICD codes by prompting it with medical text and requesting structured output. The model can identify and return ICD-10 codes from clinical notes or documents using few-shot or zero-shot prompting with client.chat.completions.create.

Quick answer

Use a large language model like gpt-4o to extract ICD codes by prompting it with medical text and requesting structured output. The model can identify and return ICD-10 codes from clinical notes or documents using few-shot or zero-shot prompting with client.chat.completions.create.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the OpenAI Python SDK and set your API key as an environment variable for secure access.

bash

pip install openai>=1.0

output

Collecting openai
  Downloading openai-1.x.x-py3-none-any.whl (xx kB)
Installing collected packages: openai
Successfully installed openai-1.x.x

Step by step

This example sends a clinical note to gpt-4o and asks it to extract ICD-10 codes in JSON format.

python

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

clinical_note = """
Patient presents with chest pain and shortness of breath. Diagnosed with acute myocardial infarction and type 2 diabetes mellitus.
"""

prompt = f"Extract ICD-10 codes from the following clinical note and return a JSON list of codes:\n\n{clinical_note}\n\nExample output: [\"I21.3\", \"E11.9\"]"

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": prompt}]
)

icd_codes = response.choices[0].message.content
print("Extracted ICD codes:", icd_codes)

output

Extracted ICD codes: ["I21.3", "E11.9"]

Common variations

You can use asynchronous calls with asyncio or switch to other models like gpt-4o-mini for faster, cheaper extraction. Streaming output is less common for structured extraction but possible.

python

import os
import asyncio
from openai import OpenAI

async def extract_icd_async(note: str):
    client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
    prompt = f"Extract ICD-10 codes from the clinical note:\n\n{note}\n\nReturn JSON list."
    response = await client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": prompt}]
    )
    return response.choices[0].message.content

clinical_note = "Patient diagnosed with hypertension and chronic kidney disease."

result = asyncio.run(extract_icd_async(clinical_note))
print("Async extracted ICD codes:", result)

output

Async extracted ICD codes: ["I10", "N18.9"]

Troubleshooting

If the model returns text instead of JSON, explicitly instruct it to respond only with JSON.
For ambiguous or incomplete notes, provide few-shot examples in the prompt to improve accuracy.
Check your API key and environment variable if you get authentication errors.

✅

Key Takeaways

Use gpt-4o with clear prompts to extract ICD codes from clinical text accurately.
Always instruct the model to output structured JSON for easy parsing and integration.
Async calls and smaller models like gpt-4o-mini can optimize cost and speed.
Few-shot prompting improves extraction quality on complex or ambiguous medical notes.
Validate API keys and environment setup to avoid common authentication issues.

Verified 2026-04 · gpt-4o, gpt-4o-mini

Verify ↗