Why We Are Different
Unlike general-purpose LLMs (such as ChatGPT, Claude, etc.), we use specialized models trained specifically for clinical coding, achieving:- 77% higher accuracy than general LLMs in ICD-10 coding.
- A 92% reduction in medical hallucinations.
- Customization for your center: Personalized models trained on your nomenclatures and workflows.
- Truly multilingual: Trained on clinical Spanish, Catalan, Basque, Galician, and English.
What Is It For?
This API solves critical clinical documentation challenges:ICD-10 Medical Coding
- Automatic ICD-10 code extraction from clinical notes with AI-powered analysis.
- Confidence scores for each code (0-100%) to support coding decisions.
- Detailed justifications explaining why each code was selected.
- Discarded codes analysis showing rejected codes and reasoning.
- Measurable ROI: A 70% reduction in coding time.
Billing and Compliance
- Accurate code assignments for reimbursement.
- Audit trail with run_id tracking for each coding session.
- Doctor and patient tracking for regulatory compliance.
- Quality assurance through transparent AI decision-making.
Integration Ready
- PDF support for processing scanned documents (up to 5MB).
- Text input for direct EHR integration (up to 50KB).
- Custom AI models selection for specialized needs.
- RESTful API with consistent error handling.
How Does the Process Work?
Step 1: Prepare Your Clinical Note
Prepare your clinical note as either:- Plain text (up to 50KB) - from EHR, dictation, or manual entry
- PDF file (up to 5MB decoded) - scanned documents or reports
Step 2: Send to the Codify API
Send your clinical note to/v1/codify with optional tracking headers for audit purposes.
Our specialized clinical AI engine:
- Analyzes the medical context using models trained on millions of real clinical notes.
- Identifies diagnoses and medical conditions with high accuracy.
- Assigns appropriate ICD-10 codes with confidence scores.
- Provides detailed justifications for each code selection.
- Shows discarded codes and why they were rejected.
Step 3: Receive ICD-10 Code Assessments
The API returns:- Final code assessments: Selected ICD-10 codes with descriptions, justifications, and confidence scores
- Discarded assessments: Rejected codes showing the AI’s decision-making process
- Run ID: Unique identifier for tracking and debugging
Key advantage: Complete transparency with confidence scores and justifications helps coders make informed decisions and maintain quality standards.
Available Endpoints
POST /v1/codify
Codes clinical text into a JSON structure that you define.
GET /v1/health
Verifies that the service is operational.
Interactive API Playground
The documentation includes an interactive API playground powered by the OpenAPI specification. You can test API calls directly from the browser without writing any code.Navigate to the API Reference
Go to the API Reference section in the SofIA API tab, or click here.
Configure authentication
Enter your Bearer token in the Authorization field at the top of the playground.
Fill in the request
Use the form fields to build your request, or switch to the Body tab to paste raw JSON. The playground validates your input against the OpenAPI schema in real time.
API Versioning
Versioning Scheme
- Path versioning:
/v1/,/v2/, etc. - Backward-compatible changes: Are added to the current version without breaking the API.
- Breaking changes: Increment the major version (e.g., v1 → v2).
Current version: v1.0.0 — Initial release of the public endpoint.
Next Steps
1. Quickstart
Your first API call
2. Authentication
Configuring tokens and headers
3. Data Structure
Requests and responses
4. View Examples
Practical coding examples
5. Operations and Errors
Error handling and security
6. Complete Reference
OpenAPI specification