EDS AP-HP — Entrepôt de Données de Santé

Overview

The EDS AP-HP (Entrepôt de Données de Santé de l’Assistance Publique – Hôpitaux de Paris) is one of the largest hospital clinical data warehouses in Europe, integrating administrative and medical data from the 38 establishments of AP-HP. Authorised by CNIL since 19 January 2017 (authorisation n°1980120), it provides a unified, standardised research infrastructure for observational studies, cohort building, algorithm development, and AI applications in clinical medicine.

Scale

The EDS AP-HP covers 19 million distinct patients treated at AP-HP hospitals since the 2017 authorisation, confirmed by AP-HP. The following more granular figures are as published at the source URL and may be updated as the EDS expands:

  • 190 million medical reports
  • 1,372 million biology results
  • 42 million imaging exams
  • 83 million diagnoses
  • 93 million acts Source

Data Content

The EDS integrates structured and unstructured data from across AP-HP’s hospital information systems. Administrative data covers hospitalisations, consultations, diagnoses (PMSI), procedures (CCAM), and billing. Biological data contains laboratory results coded with LOINC. Prescriptions and dispensing carry drug data using ATC classification. Clinical notes are free-text medical reports processed with NLP tools. Imaging metadata links to DICOM imaging data. Diagnostic coding in source data uses ICD-10 / CIM-10 (PMSI); within the OMOP CDM layer, conditions and procedures are mapped to SNOMED CT as the standard target vocabulary.

Standards and Interoperability

Data is standardised to international standards: OMOP CDM as the primary common data model, HL7 FHIR for API access, and medical terminologies including LOINC, CIM-10 (ICD-10), and ATC. This standardisation enabled AP-HP to join the international 4CE consortium (Consortium for Clinical Characterization of COVID-19 by EHR) for federated international analyses during the pandemic.

Tools and Access

Cohort360

Cohort360 is the primary web interface for the EDS, allowing AP-HP clinicians and researchers to build patient cohorts using inclusion/exclusion criteria. It is open source, built on a HL7 FHIR API over the OMOP CDM backend, and is in active production use.

i2b2

The EDS also exposes data through an i2b2 interface, enabling federated queries compatible with the broader i2b2/SHRINE network.

Datalab

The Datalab is a secure private workspace for data analysis, providing up to 5 CPU, 16 GB RAM, and 100 GB storage per project. It includes Jupyter notebooks, GitLab, Slurm, and HDFS/Hive, and is accessible remotely from anywhere via double authentication. Additional AI tools are available including EDS-NLP and EDS-Scikit for medical text and tabular data analysis. Described at the EDS products page.

Governance and Access

  • The Comité de pilotage stratégique defines the EDS strategy and roadmap, meets bimonthly, and includes medical, paramedical, scientific, executive, and patient representatives.
  • The CSE (Comité Scientifique et Éthique) reviews and approves all research projects. External partner access decisions are made within 2 months of a complete dossier submission.
  • CNIL authorisation for research reuse has been in place since January 2017.
  • Patients can opt out of research reuse at any time via the opt-out form.
  • External access requires an AP-HP medical sponsor, URC feasibility review, CSE approval, and contractualisation before data access is granted.

Connections

  • Operated by: AP-HP
  • Regulatory authorisation: CNIL (since January 2017)
  • Data model: OMOP CDM (primary), HL7 FHIR (API layer)
  • Query tools: Cohort360 (open source), i2b2

Resources