Skip to content
2026-06-23

Onboarding Overview

Onboarding to the Cohort Discovery Service requires completing three parallel workstreams. An introductory meeting with the HDR UK Technology Team will be held at the start to define tasks, responsibilities, and timelines.

The Three Workstreams

graph TD
    G[Governance Workstream] --> G1[G1: Obtain Data Controller Consent]
    G --> G2[G2: Local Approvals & Processes]
    G --> G3[G3: Technical Risk Assessment]

    D[Data Workstream] --> D1[D1: Extract Data Subset]
    D --> D2[D2: Map to OMOP]
    D --> D3[D3: Create Synthetic Data]
    D --> D4[D4: ETL to OMOP Database]

    I[Infrastructure Workstream] --> I1[I1: Set Up Secure Network Area]
    I --> I2[I2: Install Query Retrieving Software]
    I --> I3[I3: Connect OMOP Data to Software]

    G1 -->|Required before live data| I3
    D2 --> D4
    I1 --> I2
    I2 --> I3

Figure 3 — Cohort Discovery workstreams and tasks

The three parallel onboarding workstreams: Governance (G1–G3), Data (D1–D4), and Infrastructure (I1–I3), with arrows showing task dependencies

Figure 4 — Interdependencies between tasks for onboarding to Cohort Discovery

Task interdependency diagram showing that G1 (Data Controller Consent) must precede live data connection, and that D2 feeds D4 which feeds I3

Key dependency

G1 (Data Controller Consent) must be completed before real (non-synthetic) data is connected to the Cohort Discovery Service. D1, D2, and D3 can proceed before G1 is complete.

Workstream Summary

Step Description Responsible
G1 Obtain Data Controller consent Data Custodian
G2 Carry out local approvals and processes Data Custodian
G3 Technical risk assessment (e.g. SSP) Data Custodian

Governance steps are not always linear and can occur in any order. Start as early as possible.

Governance Workstream

Step Description Responsible
D1 Extract a subset of raw identifiable data Data Custodian
D2 Map extracted subset to OMOP CDM Data Custodian (HDR UK can assist)
D3 Create synthetic data for testing Data Custodian (optional but recommended)
D4 ETL extracted data to create OMOP database Data Custodian

Data Workstream

Step Description Responsible
I1 Set up secure network area Data Custodian
I2 Install query retrieving software Data Custodian
I3 Connect OMOP data to query retrieving software Data Custodian

Infrastructure Workstream

Iterative approach

Cohort Discovery supports an iterative onboarding model. You do not need to onboard all your data at once.

  • Multiple datasets: Bunny requires two instances per OMOP database. Multiple sets can run on the same server.
  • New fields: Additional data fields can be added over time following the same governance process.

Tip

Focus first on the key fields needed to answer pressing research questions. Additional fields can be added iteratively as community needs evolve.