Overview

The Federated Data Quality Framework (FDQF) is open-source tooling for assessing data quality in federated environments — networks where data remains distributed across independent locations for legal, organizational, or technical reasons.

Instead of moving data to a central location, the FDQF runs quality checks locally at each site and reports only aggregated, privacy-preserving results.

Components

The FDQF consists of two main components:

Data Quality Agent (DQA)

Runs locally at each data-holding site. It:

• Executes data quality checks against the site's data source
• Supports two data source types:
  • FHIR — a FHIR R4 server queried with HL7 CQL (Clinical Quality Language)
  • SQL — a relational database or directory of CSV files queried with SQL (PostgreSQL, MySQL, SQLite, Apache Calcite/CSV)
• Aggregates results and applies differential privacy before anything leaves the site
• Pushes quality reports to the central server (one-way only; the server cannot pull data)

See Data Sources for configuration details and Privacy and Security for how the reported results are anonymized.

Data Quality Server (DQS)

Runs centrally to collect and present results. It:

• Receives quality reports from multiple agents
• Aggregates results across the network
• Provides dashboards for end users
• Exposes a REST API for integrating reports into external dataset catalogues

This lets researchers and study investigators compare data quality across sites without accessing raw data.

Why Federated Data Quality?

Traditional data quality validation requires central access to raw datasets, which is often impossible in healthcare and other regulated environments. The FDQF enables:

• Local processing at each data site
• Privacy-preserving quality metrics
• Cross-site comparison without data sharing
• Standardized and reproducible quality evaluation

Who Is It For?

The FDQF is designed for:

• Research networks and data consortia
• Healthcare institutions
• Biobanking and cohort infrastructures
• Privacy-sensitive or regulated environments

Typical roles include data stewards, study principal investigators, clinical researchers, and data engineers or IT operators.

Summary

Concept	Description
Federated approach	Data stays at each site; checks run locally
Local agent	Executes CQL or SQL checks and reports privacy-preserving metrics
Central server	Collects, aggregates, and displays results
Goal	Assess data quality across sites without sharing raw data

See it in action

Follow the Hands-On Guide to deploy an agent and run your first quality checks.