health-md-standard
EIR Format - Open Standard for LLM-Optimized Healthcare Data
| Version 1.0 | Website | Specification | Examples | MCP Server |
Named after Eir, the Norse goddess of healing and medicine
🎯 Mission
Healthcare data is complex, sensitive, and crucial. EIR Format defines an open YAML-based standard for structuring healthcare information optimized for Large Language Models while preserving privacy, accuracy, and clinical context.
🚀 Why Health.md?
- LLM-Optimized: Structured for perfect AI comprehension and reasoning
- Privacy-First: Built-in anonymization and data protection patterns
- Clinical-Accurate: Preserves medical context and relationships
- Human-Readable: Markdown format that clinicians and patients can read
- Interoperable: Works across systems, platforms, and use cases
📋 Quick Example
# Health Record - Anonymous Patient 001
## Demographics
- Age: 34
- Sex: Female
- Occupation: Software Engineer
## Current Medications
### Metformin 500mg
- **Indication:** Type 2 Diabetes Mellitus
- **Dosage:** 500mg twice daily with meals
- **Started:** 2024-01-15
- **Prescriber:** Dr. Smith (Endocrinology)
- **Notes:** Well tolerated, no gastrointestinal issues
## Lab Results
### Hemoglobin A1C
- **Date:** 2024-02-10
- **Value:** 6.8%
- **Reference:** <7.0% (target for diabetes)
- **Trend:** ↓ from 8.2% (2023-12-01)
- **Clinical Significance:** Improving glycemic control
## Clinical Timeline
### 2024-01-15: Initial Diabetes Diagnosis
- **HbA1c:** 8.2%
- **Fasting Glucose:** 180 mg/dL
- **Action:** Started Metformin 500mg BID
- **Goals:** HbA1c <7%, weight loss 5-10%
🛠️ Features
- Semantic Structure: Clear sections for demographics, medications, labs, timeline
- Temporal Context: Built-in date tracking and trend analysis
- Clinical Relationships: Links between conditions, medications, and outcomes
- Privacy Controls: Anonymization patterns and data sensitivity markers
- Validation: Schema validation and clinical accuracy checks
- LLM Integration: Parser library for seamless AI integration
📖 Specification
The complete specification is available in SPEC.md. Key sections include:
🔧 Parser & Tools
pip install health-md
from health_md import HealthRecord
# Parse a health.md file
record = HealthRecord.from_file('patient-001.health.md')
# Extract key information
medications = record.get_current_medications()
recent_labs = record.get_labs(days=30)
timeline = record.get_clinical_timeline()
# Generate LLM-optimized summary
summary = record.to_llm_context()
🌍 Use Cases
- Clinical Decision Support: Structured data for AI diagnostic tools
- Patient Portals: Human-readable health records
- Research: Standardized datasets for health AI training
- Health Apps: Interoperable data format across platforms
- Telemedicine: Efficient information sharing between providers
🤝 Contributing
We welcome contributions from:
- Clinicians: Medical accuracy and workflow insights
- Developers: Parser improvements and tooling
- Researchers: Validation studies and use cases
- Patients: Usability feedback and privacy requirements
See CONTRIBUTING.md for guidelines.
📊 Adoption
Organizations using Health.md:
- EIR Space - Health literacy platform
- Your organization here - submit a PR!
📄 License
MIT License - See LICENSE
🏆 Team
Created by:
- Birger Moëll - AI/NLP Researcher, Clinical Psychologist
- Uppsala University, Department of Linguistics and Philology
Special Thanks:
- Healthcare professionals who provided clinical input
- Privacy advocates who shaped our anonymization guidelines
- LLM researchers who validated our structured approach
Healthcare data deserves better standards. Let’s build them together. 🏥💙