9  Analytics on FHIR

9.1 Current standards for analytics

  • Explain Jupyter notebook for interactive computing
  • We want to support different languages
  • Federated analytics
    • Most simple way: we mount the different data containers, which currently are ADLS Gen2 containers
    • In future: federated analytics libraries
  • Option: computation in the blind

Interactive notebooks are the de-facto standard for analysts and researchers. Many open source solutions which can be self-hosted, most importantly Jupyter Project Many commercial offerings: All cloud vendors have own version of Jupyter already integrated in their environment To save engineering work, for now we opt for Azure ML Studio

The paradigm shift that we introduce is federated analytics Federated analytics = queries over decentralised databases Federated learning = training algorithms over decentralised data We use federated analytics as a term to cover both. For LMICs, starting with federated analytics (queries) is positioned on short term We add multi-party computation as optional component that can be used through interactive notebook environment

Impacts following components: Data Visualization and User Interaction Data Analytics Data processing architectures Standards Considerations for LRS