OHC UHI — Secure Inter-Hospital Data Exchange

A hackathon demo showing how patient health records can be securely shared between hospitals using encrypted FHIR bundles, patient-controlled consent, and a zero-data-storage broker. It also demonstrates how the MedGemma AI model works inside the EMR platform to analyze reports and provide constructive insights.

🏗️ System Architecture

The system is purposely built for mock purposes, consisting of three main environments:

┌─── System 1 (Hospital A) ───┐     ┌──── INTERNET ────┐     ┌─── System 2 (Hospital B) ───┐
│ CityCare Hospital           │     │  UHI SWITCH      │     │ Metro Radiology             │
│ Local Network               │◄───►│  Consent Broker  │◄───►│ Local Network               │
│ CARE EMR + MedGemma         │     │  No Data Storage │     │ CARE EMR + MedGemma         │
│ :9001 / :4001               │     │  :8080 (Cloud)   │     │ :9002 / :4002               │
└─────────────────────────────┘     └──────────────────┘     └─────────────────────────────┘
                                             ▲
                                             │
                                  ┌──────────┴─────────┐
                                  │ Patient Mobile App │
                                  │ (Flutter)          │
                                  └────────────────────┘

Important: Hospital A and Hospital B are on completely separate internal networks and only communicate with each other through the UHI-Switch.

📂 Repository Structure

├── care_medgemma/       # Django plugin — MedGemma AI + FHIR export + Consent
├── hospital-a/          # Hospital A deployment (docker-compose + setup)
├── hospital-b/          # Hospital B deployment (docker-compose + setup)
├── devaganesh-reports-hospital-1/  # Patient reports (Hospital A) - Months 1-10 Progress, Months 1-2 Imaging
├── devaganesh-reports-hospital-2/  # Patient reports (Hospital B) - Months 3-8 Imaging
├── uhi-switch/          # Central consent broker server
├── DEPLOYMENT.md        # Detailed deployment guide for 2 systems
├── CareConnect/         # Flutter mobile app 

🚀 Quick Start — Deploy on 2 Systems

Step 1: UHI Switch

The UHI Switch application is hosted on a VPS at:

• 🌐 Server: https://uhi-switch-production.up.railway.app
• 📖 API Docs: https://uhi-switch-production.up.railway.app/docs

The python server maintains a consistent connection with the hospitals' backends and sends notifications to the admin in case of any outage.

Step 2: Deploy Hospital A (On System 1)

Clone this repository on your first machine:

cd hospital-a
docker compose up -d

# Frontend: http://localhost:4001 (Login: admin / admin OR doctor_a / doctor_a)
# Backend: http://localhost:9001

Step 3: Deploy Hospital B (On System 2)

Clone this repository on your second machine:

cd hospital-b
docker compose up -d

# Frontend: http://localhost:4002 (Login: admin / admin OR doctor_b / doctor_b)
# Backend: http://localhost:9002

Step 4: The Mobile App

The Flutter application (CareConnect/) can be run on a mobile emulator to represent the patient. It connects to the UHI Switch to manage consents and view summary data.

🔐 Core Demo Flow

1. Patient Data Initialization: Hospital A holds mock data from devaganesh-reports-hospital-1/ and Hospital B holds mock data from devaganesh-reports-hospital-2/.
2. Analysis Request: A doctor wants to analyze the medical reports from Month 2-8 from a single patient (ABHA ID: 91-1234-5678-9012).
3. Patient Consent: This creates a request in the patient application (HealthWallet). The patient reviews and approves the data access.
4. Data Routing: The UHI Switch brokers the consent and routes the encryption keys. The required data is fetched from the respective hospitals.
5. MedGemma Dashboard: The data shows up in the MedGemma dashboard within the EMR platform.
6. AI Analysis: The MedGemma model analyzes the longitudinal data (Months 2-8) and gives a constructive output (this functionality is mocked to run efficiently within RAM constraints).

The UHI Switch NEVER stores patient data. It only brokers consent and routes encryption keys.

See DEPLOYMENT.md for full reproduction steps.