ICD-10 API for Automated Medical Coding
Send clinical notes to the ICD-10 API and get back accurate ICD-10-CM diagnosis codes with confidence scores in seconds. Reduce coding errors, speed up claims, and let your team focus on patient care.
How It Works
The coding pipeline reads clinical text, extracts medical entities, and matches each one to the most relevant ICD-10-CM codes.
Clinical text is analyzed by a specialized medical NER model to identify diagnoses, symptoms, and conditions.
Post-processing detects negation, expands abbreviations, corrects spelling, and classifies clinical context.
Each entity is matched against 74,000+ ICD-10-CM codes using medical embeddings enhanced with SNOMED CT and UMLS synonyms.
Results are returned with ranked code candidates, confidence scores, and context flags.
Code Example
Three lines of code to go from clinical text to ICD-10 codes.
import { AutoICD } from "autoicd-js";
const client = new AutoICD({ apiKey: "sk_..." });
const result = await client.code(
"Type 2 diabetes with chronic kidney disease stage 3. " +
"Patient denies chest pain."
);
// result.entities[0].entity_text → "Type 2 diabetes"
// result.entities[0].codes[0].code → "E11.65"
// result.entities[0].codes[0].confidence → "high"
// result.entities[1].negated → true (chest pain denied)Capabilities
Built to handle the complexity of real clinical documentation.
Entity Extraction
The AI reads clinical text and identifies every medical condition, symptom, and diagnosis, just like an experienced coder.
ICD-10-CM Code Matching
Each entity is matched against the complete 2025 ICD-10-CM code set (74,000+ codes) using medical embeddings enriched with SNOMED CT and UMLS synonyms.
Confidence Scoring
Every code comes with a similarity score (0–1) and confidence level (high or moderate) so you can prioritize review.
Negation Detection
Understands phrases like "no evidence of heart failure" or "denies chest pain" and flags negated findings so only confirmed diagnoses are coded.
Clinical Context
Distinguishes between current diagnoses, past medical history, family history, and uncertain findings for precise, context-aware coding.
Spell Correction
Handles misspelled medical terms and shorthand, common in fast-paced clinical documentation, so nothing gets missed.
Integration Options
Use the coding endpoint from any language or platform.
Common ICD-10 Codes
The API covers all 74,000+ ICD-10-CM codes. Here are some of the most frequently coded conditions:
Start coding diagnoses in minutes
Free 7-day trial with 100 requests per day. No credit card required.