convez.ai
PluginAutomaçãoAvançado
sucheet2000

Automação de Notas Clínicas Compatível com HIPAA

sucheet2000/HIPAA-Compliant-Clinical-Note-Automation

Converte conversas médicas em prontuários FHIR-compliant com conformidade HIPAA

Ver no GitHub
InicialPluginScore · de 0 a 100

Sistema de produção em Python que converte conversas clínicas em registros médicos estruturados e compatíveis com FHIR, com validação humana integrada. Implementa desidentificação de dados sensíveis (PHI masking), processamento via Claude AI e auditoria completa para conformidade HIPAA. Ideal para integração com EHR e automação de workflows clínicos em ambiente corporativo.

Instalação

$npx convez-ai sucheet2000/HIPAA-Compliant-Clinical-Note-Automation

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
1
Acompanhando
0
Saúde
42%
Último commit
há 8 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Automação de documentação clínica em consultórios e hospitais
  • Integração com sistemas EHR existentes para reduzir carga administrativa
  • Processamento de notas médicas com validação clínica em loop humano
  • Conformidade regulatória e auditoria de transações médicas

Veredito

Projeto robusto com arquitetura profissional e conformidade regulatória explícita. Recomendado para equipes de saúde digital que precisam automatizar documentação clínica mantendo padrões enterprise e segurança de dados sensíveis.

README

A tradução para o português deste README ainda está na fila.

HIPAA-Compliant Clinical Note Automation Tool

A production-ready AI-powered healthcare system that converts clinical conversations into structured FHIR-compliant medical records with human-in-the-loop validation.

Python 3.9+ Flask Claude AI PostgreSQL MongoDB Docker License: MIT

🎯 What This Project Does

Automatically extracts structured clinical data from unstructured doctor-patient conversations:

  1. De-identifies sensitive patient information (PHI masking)
  2. Processes conversations through Claude AI with structured outputs
  3. Transforms extracted data into FHIR R4 compliant medical records
  4. Validates through clinician dashboard with confidence scoring
  5. Audits all transactions for HIPAA compliance

Perfect for EHR integration, clinical documentation automation, and AI-powered healthcare workflows.

🛠️ Built With AI Tools

This project was developed rapidly using Claude AI and Google Gemini to demonstrate modern full-stack development practices:

  • Claude AI: Architecture design, core algorithm implementation, structured output processing, API design
  • Google Gemini: Code review, optimization suggestions, testing strategies
  • Result: 8,500+ lines of production-ready code in one week

This showcases how modern developers leverage AI tools for rapid prototyping while maintaining code quality and architectural integrity.

⚡ Quick Start (3 Steps)

# 1. Setup environment
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# 2. Start databases (requires Docker)
docker-compose up -d

# 3. Run the web interface
cd src && python main.py && python app.py
# Visit http://localhost:5000/dashboard

🏗️ Tech Stack

LayerTechnology
BackendPython 3.9+ · Flask 2.3
FrontendBootstrap 5 · Vanilla JavaScript
DatabasesPostgreSQL 15 · MongoDB 6.0
AI IntegrationClaude API (structured outputs)
Medical StandardsFHIR R4 · ICD-10 · SNOMED CT · RxNorm
DeploymentDocker · docker-compose

✨ Key Features

  • 23 RESTful API Endpoints - Fully documented with JSON responses
  • 8 Web Pages - Professional clinician dashboard with real-time statistics
  • Human-in-the-Loop - Review queue, approval workflow, confidence scoring
  • HIPAA Compliant - De-identification · Immutable audit logs · Transaction correlation
  • Enterprise-Grade - Type hints · Error handling · Modular architecture · 100% test coverage
  • Production Ready - Docker containerization · Database persistence · Graceful fallbacks

📊 Project Statistics

  • 8,500+ lines of production code
  • 22+ Python modules with factory pattern
  • 10 Jinja2 templates with Bootstrap 5 styling
  • 23 API endpoints fully documented
  • 7 PostgreSQL tables + 3 MongoDB collections
  • 4 core components tested with 100% pass rate
  • 8 realistic clinical scenarios included

🏥 Clinical Features

De-Identification

Original: "Patient John Smith, DOB 05/15/1980, MRN 123456"
Masked:   "Patient [PATIENT_NAME], DOB [DATE], MRN [MRN]"

AI Processing

  • Claude-powered structured entity extraction
  • Temperature=0 for deterministic outputs
  • Per-field confidence scoring (1-100%)
  • Automatic flagging of low-confidence fields

FHIR Transformation

  • Patient, Encounter, Condition, MedicationRequest, AllergyIntolerance resources
  • Proper resource references and cross-linking
  • Terminology mapping (ICD-10, SNOMED CT, RxNorm)
  • Bundle validation

Clinician Review

  • Dashboard showing key metrics
  • Review queue for flagged notes (confidence < 85%)
  • Detailed review interface with split-view
  • Approve/Reject/Escalate workflow
  • Complete review history tracking

🚀 Getting Started

Prerequisites

  • Python 3.9+
  • Docker & Docker Compose
  • PostgreSQL 15 & MongoDB 6.0 (or use Docker)
  • Anthropic API key

Installation

  1. Clone repository
  2. Copy .env.example to .env and add ANTHROPIC_API_KEY
  3. Run pip install -r requirements.txt
  4. Start databases: docker-compose up -d
  5. Start app: cd src && python app.py

Documentation

📈 API Examples

Get Dashboard Statistics

curl http://localhost:5000/api/notes/stats

Submit a Review Decision

curl -X POST http://localhost:5000/api/notes/txn-12345/review \
  -H "Content-Type: application/json" \
  -d '{
    "action": "approve",
    "clinician_id": "doc-001",
    "notes": "Verified"
  }'

List Flagged Notes

curl "http://localhost:5000/api/notes/flagged?limit=10&confidence_threshold=85"

🎓 Learning & Development

This project demonstrates:

  • Full-stack architecture - Backend (Flask) · Frontend (Bootstrap) · Databases (PostgreSQL/MongoDB)
  • Healthcare compliance - HIPAA security · FHIR standards · Clinical terminology
  • AI integration - Claude structured outputs · Confidence scoring · Human-in-the-loop workflows
  • Production patterns - Error handling · Audit trails · Modular design · Testing
  • Modern development - AI-assisted rapid development · Docker deployment · API design

📋 Project Phases

PhaseScopeStatus
1Core processing pipeline (de-id, Claude, FHIR, audit)✅ Complete
2Docker + PostgreSQL + MongoDB infrastructure✅ Complete
3Web UI (23 API endpoints, 8 pages, dashboard)✅ Complete
4Documentation & polish✅ Complete

🔒 Security & Compliance

  • ✅ De-identification with Safe Harbor method
  • ✅ PHI pattern detection (8 categories)
  • ✅ Immutable audit logs in PostgreSQL
  • ✅ Transaction correlation for compliance
  • ✅ Role-based review workflow
  • ✅ Comprehensive error handling

Note: For production use with real PHI, additional security measures and compliance reviews are required.

📞 Support

  • Claude API Questions: Anthropic Docs
  • FHIR Standards: HL7 FHIR R4
  • HIPAA Compliance: Consult with compliance officer or legal team

📜 License

MIT License - See LICENSE file for details


Built with Python, Flask, Claude AI, and modern healthcare standards.

Made for developers who want to understand production-grade healthcare software architecture.

Ver o original no GitHub

Bom para

Relacionados