README.md

HL7v2 to FHIR using AWS Lambda & SQS

Development

Requirements

• Java 1.11+
• Apache Maven
• LinuxForHealth HL7v2-FHIR Converter
• Optional: AWS SAM (Serverless Application Model) CLI
• Optional: GNU Make
• Optional: Docker

Environment Configuration

• Add local AWS Credentials

Build & Deploy

# Build & Deploy CloudFormation stack with Make
make build-deploy
make delete-stack
make check-size

# Build with Maven
mvn clean install

# Run local unit & integration tests
mvn test
# Run test in docker
sam local invoke

# Deploy with AWS SAM CLI & CloudFormation template
sam build -t template.yaml
sam deploy -g -t template.yaml

TODO

• Extract reusable mapping logic into separate module
• Light dependency injection with something like Dagger at compile time

Architecture

Mapping Implementation

• LinuxForHealth HL7v2-FHIR Converter
  • HAPI HL7
  • HAPI FHIR

Event-Driven ETL

• The lambda function is triggered by incoming HL7 messages from an SQS queue
• The function transforms HL7 messages into FHIR JSON
• Individual failures are tracked and returned as a partial batch response
• Successful transformations are sent to another SQS queue holding FHIR JSON

Infrastructure as Code

An AWS CloudFormation template provides configuration for:

• AWS Lambda transformation function
• Incoming HL7 queue
• Outgoing FHIR queue
• IAM roles + policies for access to queues

Review

• Advantages
  • Reuse of stable HAPI HL7 & FHIR infrastructure
• Disadvantages
  • High memory usage from the mapping library used and Java generally
  • Startup is potentially too slow for certain use cases
  • Each queue should be limited to specific HL7 message types & FHIR resources from specific FHIR versions.