Applied Data Harmonization, Quality, Analytics and Federated Node Deployment

1 · Programme purpose and positioning

Why this certificate

The EpiScientia OMOP Data Engineer Certificate is an applied competency programme designed to develop data engineers who can lead local health-data harmonization at African data-owner institutions. The Rwanda implementation is the initial training setting, while the curriculum is intentionally portable to institutions using OpenMRS, OpenClinic GA, eFiche, eBuzima, MediSoft and registry datasets. The programme follows the source-to-target sequence: source inspection and profiling, structural mapping, semantic standardization, ETL implementation, data-quality validation, cohort construction, analytics, containerization and Vantage6 node operation. Certification is based on demonstrated engineering competence rather than attendance alone: a participant must be able to explain mapping decisions, implement and troubleshoot an ETL, evaluate data quality, construct an executable cohort, and demonstrate a controlled federated computation.

Teaching staff

Course directors & assistants

2 · Award and programme specifications

Specifications

3 · Entry prerequisites

Who can join

4 · Programme-level learning outcomes

On successful completion, the certificate holder will be able to

1. LO1Explain the OHDSI ecosystem, OMOP CDM architecture, domains, conventions and standardized vocabulary model.
2. LO2Inspect an unfamiliar EHR or registry schema and produce a source-data inventory and profiling report.
3. LO3Design structural source-to-target mappings and document transformation logic before coding.
4. LO4Map local clinical codes and terms to appropriate standard concepts, documenting uncertainty and expert review.
5. LO5Implement a reproducible ETL that preserves source provenance and populates a selected OMOP CDM release.
6. LO6Run and interpret OMOP data-quality and characterization outputs and iteratively correct the ETL.
7. LO7Configure/use WebAPI and ATLAS sufficiently to create concept sets, define cohorts and execute basic standardized analyses.
8. LO8Critically use alternative and AI-assisted mapping approaches without delegating semantic accountability to automation.
9. LO9Containerize/configure a local data-owner environment and connect an approved Vantage6 node to a training central server.
10. LO10Independently demonstrate the end-to-end pathway: source → profile → map → ETL → OMOP → quality → cohort/analytics → federated node.

Teaching model — Each online week uses a fixed 60-minute live pattern: 25 minutes of focused theory followed immediately by 35 minutes of supervised practice. The practical is performed on the learner's own approved data where feasible, or on a common source-like training dataset. A 20-minute asynchronous quiz is linked to the same week's competency and is completed later at the learner's pace. Weekly assessment combines short knowledge items with evidence from the practical exercise so the certificate does not become a knowledge-only award.

6 · Ten-week curriculum map

Week by week

8–9 · In-person workshops

Two engineering sprints in Kigali

10 · Assessment framework

Continuous assessment 60% · final practical examination 40%

Final practical examination rubric (40%)

Performance descriptors

11 · Certification criteria

EpiScientia Certificate in OMOP Data Engineering

• Overall weighted mark ≥ 60%.
• Final practical examination mark ≥ 50% (competency gate).
• Submission of all mandatory capstone artefacts or approved equivalents.
• Participation in both in-person workshops unless a formally approved equivalent supervised practical assessment is provided.
• No critical failure in data-governance, security or semantic-mapping practice that would make the demonstrated node unsafe or analytically invalid.
• Participants who meet the overall mark but fail the practical gate receive a completion statement rather than the OMOP Data Engineer Certificate and may undertake a supervised reassessment.

Statement of competence — A successful EpiScientia OMOP Data Engineer Certificate holder has demonstrated the ability to independently take an approved clinical or registry source through source profiling, structural and semantic mapping, reproducible ETL implementation, OMOP quality validation, cohort construction and local analytics, and to configure the resulting data-owner environment for controlled participation in Vantage6-mediated federated analytics.

Trainer and assessor requirements

13 · Technical and training infrastructure

Software stack & environments

Data governance, security and responsible AI

• Use only data for which the participant and training team have explicit institutional authorization.
• Prefer synthetic or de-identified datasets for cross-country teaching demonstrations; real patient-level data must not be copied into shared course infrastructure without a lawful and approved basis.
• Maintain separation between training and production credentials, servers and Vantage6 organizations.
• Record source-to-standard mapping decisions, including rejected and uncertain mappings, so that semantic choices are auditable.
• AI/LLM tools may propose mappings, transformation code or documentation, but their outputs must be validated against the selected OMOP vocabulary/CDM release and reviewed by a competent human.
• Patient-level data must not be entered into external generative-AI services unless explicitly authorized under applicable governance.
• Federated execution does not by itself guarantee privacy: algorithms, output disclosure rules, permissions and node configuration require governance and technical review.

15 · Mandatory learner portfolio

What every certificate holder hands over

1. Source-system inventory and WhiteRabbit profiling report.
2. Rabbit-in-a-Hat/source-to-target ETL design.
3. Versioned vocabulary mapping file with review status.
4. Executable ETL package and documentation.
5. OMOP data-quality/characterization report and remediation log.
6. Saved/exported ATLAS concept set and cohort definition.
7. Docker/Vantage6 node configuration evidence with secrets removed.
8. Federated test execution evidence and troubleshooting log.
9. Final technical handover note describing how the local institution can maintain and rerun the pipeline.

Quality assurance — Each cohort freezes the OMOP CDM release, vocabulary release, tool versions and assessment dataset/environment before teaching begins. Assessment rubrics are calibrated by trainers before Week 5 and moderated after Week 10. Mapping disputes are escalated to a designated terminology/clinical reviewer rather than resolved by convenience. Course feedback distinguishes teaching quality from technical environment failures so that infrastructure problems do not distort competency judgments.

17 · Reference framework

Core references

• The Book of OHDSI — core conceptual and methodological reference, particularly the Extract-Transform-Load chapter.
• OHDSI OMOP Common Data Model documentation — current CDM specification and conventions.
• OHDSI standardized vocabularies and associated mapping tooling/documentation.
• DataQualityDashboard and Achilles documentation for quality and characterization.
• ATLAS/WebAPI documentation for cohort and analytical workflows.
• Vantage6 documentation for server, node, client, Dockerized algorithm execution and node configuration.