dorsaffer/Data-Meditation-Layer

This project demonstrates a secure mediation layer for cross-organization health data access, built to support Universal Health Coverage (UHC) planning in Sierra Leone. # Data Mediation Layer A secure data mediation layer for cross-organisation health data access, focused on supporting Universal Health Coverage (UHC) analysis in Sierra Leone. ## Problem Health data needed for UHC analysis can come from different organisations and systems. In this project, we work with: - DHIS2 for aggregate health service activity data - Statistics Sierra Leone for district population data These sources have different data structures and formats, making it difficult to use them together directly. The data also needs to be assessed before it can be shared. This includes checking its quality, privacy risks, semantic meaning, and compliance with a standard such as FHIR. Different users also need different levels of access, and their actions need to be traceable. ## Solution A pipeline that takes raw source data through acquisition → canonical modelling → quality/privacy screening → standards-based (FHIR R4) export, with role-based access control and an audit trail at every step: 1. **Acquire** aggregate indicators from the Sierra Leone DHIS2 demo instance, unmodified (`docs/rbac.md` covers who can read raw records). 2. **Transform** into a canonical model (District/Indicator/Observation) plus self-describing governance metadata per dataset (`DataProduct`). 3. **Integrate a second source** — 2021 census district population — reconciled against DHIS2's district boundaries, to derive a UHC-relevant metric (service activity as a share of population) that neither source provides alone (`docs/population-integration.md`). 4. **Assess quality and privacy** automatically: completeness, validity, outliers, staleness, identifier leakage, small-cell disclosure risk — resolved into a visible `publishable` / `publishable_with_warnings` / `blocked` decision that always requires a human governance sign-off (`docs/data-quality.md`, `docs/privacy.md`). 5. **Map to a recognised terminology** (ICD-10), proposal separated from human-reviewed acceptance (`docs/ter …