Changelog

Tip

The RD-CDM paper has now been published at Nature Scientific Data. You can read it here!

This changelog provides a history of the changes to the RD-CDM.

Note

Previous versions (v1.0 - v1.7) were developed in a German context and are not publicly available.

v2.0.3 (2026-03-24)

Code System Version Updates

All ontology and code system versions have been updated to their latest releases as reported by BioPortal. The following version drifts were resolved:

Code System

Previous Version

Updated Version

SNOMEDCT

SNOMEDCT_US_2024_09_01

2025AB

LOINC

LNC278

281

HP

2025-05-06

2026-02-16

NCIT

24.01e

26.02d

NCBITAXON

NCBI2024_04_02

2025_04_10

GENO

2023-10-08

2026-02-02

UO

2023-05-25

2026-01-16

ECO

2025-06-23

releases/2025-06-23

ICD10CM

ICD10CM_2025

2026

MONDO

2025-06-03

2026-03-03

ORDO

4.7

4.8

Data Element Update: Sex at Birth (2.2)

The code for data element 2.2 Sex at Birth has been updated. The previous SNOMED CT concept 281053000 | Sex of baby at delivery (observable entity) was identified as inactive (deprecated) in the SNOMED CT browser.

The element is now coded using the LOINC concept:

  • LOINC 76689-9 — Sex assigned at birth

This aligns with HL7 FHIR’s Patient resource and is semantically precise as a question/observable code. The value set choices (Female, Male, Unknown, etc.) remain SNOMED CT encoded as before.

Repository and Package Structure Refactor

The versioned folder structure inside src/rd_cdm/instances/ has been removed. Previously, each data model version occupied a dedicated subdirectory (e.g., instances/v2_0_2/) with the version repeated in filenames (e.g., rd_cdm_v2_0_2.yaml). This created redundancy since Git tags and PyPI versioning already serve as the version archive.

New structure:

src/rd_cdm/
├── schema/
│   └── rd_cdm.yaml          # LinkML schema — version defined here
├── instances/
│   ├── code_systems.yaml
│   ├── data_elements.yaml
│   ├── value_sets.yaml
│   └── rd_cdm.yaml          # merged, version-stamped output
│   ├── jsons/
│   │   └── rd_cdm.json
│   └── csvs/
│       ├── code_systems.csv
│       ├── data_elements.csv
│       ├── value_sets.csv
│       └── rd_cdm.csv

Version is now embedded in every exported file. The fields rd_cdm_version and rd_cdm_date are defined in the LinkML schema (schema/rd_cdm.yaml) and written into every merged and exported file by rd-cdm-merge. Every standalone YAML, JSON, or CSV file is therefore self-describing.

To use an older version of the model, use the corresponding Git tag or pin the PyPI package:

pip install rd-cdm==2.0.2

CLI Naming Update

All CLI commands have been renamed from rdcdm-* to rd-cdm-* for consistency with the PyPI package name rd-cdm:

Old command

New command

rdcdm-merge

rd-cdm-merge

rdcdm-json

rd-cdm-json

rdcdm-csv

rd-cdm-csv

rdcdm-validate

rd-cdm-validate

The --version / -v argument has been removed from all CLI tools since version resolution via subdirectories is no longer needed.

The recommended workflow after updating instance files is:

rd-cdm-merge && rd-cdm-json && rd-cdm-csv && rd-cdm-validate

Validation Improvements

The rd-cdm-validate command now shows progress bars (via tqdm) for each of the three validation phases: code system version checking, data element validation, and value set code validation. The current element being checked is shown in the progress bar postfix.

The validation summary now reports the data model version at the top:

=== RD-CDM VALIDATION SUMMARY (model version: 2.0.3) ===

Dependency Changes

The following dependencies were removed as they were not used by the package:

  • numpy

  • requests-cache

  • jsonschema

  • oaklib

linkml has been moved from a core runtime dependency to an optional dev extra, since it is only needed to regenerate the Python classes from the schema. Users installing rd-cdm for data access do not require it:

pip install rd-cdm        # no linkml
pip install rd-cdm[dev]   # includes linkml for schema development

tqdm has been added as a core dependency for validation progress reporting.

Python Class Generation

The gen_pydantic.py utility now generates both output files from the schema in a single run:

  • src/rd_cdm/python_classes/rd_cdm.py — LinkML runtime dataclasses (via PythonGenerator)

  • src/rd_cdm/python_classes/rd_cdm_pydantic.py — Pydantic v2 models (via PydanticGenerator)

Custom top-level schema fields (date) that are not valid SchemaDefinition fields are stripped into a temporary file before generation to avoid SchemaDefinition.__init__() errors.

Version 2.0.1 (2025-08-07)

What’s new

  • Complete LinkML model definition polished and consolidated for RD-CDM v2.0.1. - Coding, ValueSet, DataElement, and CodeSystem clarified and aligned. - elementCode.system and CodeSystem.id now consistently use ontology acronyms (e.g., SNOMEDCT, LOINC, HP, NCIT).

  • Automated validation against BioPortal (new CLI: ``rd-cdm-validate``). - Validates all DataElement.elementCode and ValueSet.codes entries against the latest BioPortal content. - Summarizes results: number of data elements & value set members checked, valid/missing/skipped terms, label-drift warnings. - Label drift (model label ≠ live prefLabel) is reported as a warning, not a failure. - Composite SNOMED expressions (codes containing =) are skipped on purpose. - Handles LOINC part/answer codes (e.g., LA26406-1) and NCIt IRIs via the EVS Thesaurus mapping. - Uses an explicit ontology mapping (HP/MONDO/OBO, NCIT/EVS, SNOMEDCT, LOINC, etc.) + CURIE and IRI fallbacks.

  • Version checks (live vs. model) - By default checks every CodeSystem in the instance directory against BioPortal’s latest_submission. - A configurable skip list excludes non-ontology systems (e.g., CustomCode, GA4GH, HL7FHIR) from version drift checks. - Environment variable: BIOPORTAL_API_KEY is required.

  • Dynamic instance version resolution - Validation and merge tooling now auto-locate the latest instances directory:

    • src/rd_cdm/instances/{version}/rd_cdm_full.yaml is chosen from --version, then pyproject.toml (tool.poetry.version), then the newest folder on disk.

    • Works across future releases without changing hard-coded paths.

  • Merge improvements - merge_instances.py updated to reliably rebuild rd_cdm_full.yaml from code_systems.yaml, data_elements.yaml, and value_sets.yaml in the resolved version directory.

  • Export utilities - Added helpers to export LinkML instances to JSON and CSV for downstream processing (via LinkML dumpers), improving round-tripping and interoperability.

Data & label consistency updates

  • Adjusted several labels to match BioPortal prefLabel (reported previously as label drift), e.g.: - SNOMED CT: 410605003 → “Confirmed present” (capitalization). - HPO onset labels simplified to BioPortal’s canonical forms (e.g., “Embryonal onset”).

  • Ensured validation uses ValueSet.codes (the members) rather than the ValueSet id itself.

CodeSystem version alignment

  • Code system versions in the schema updated to BioPortal’s current latest_submission: - HP → hp/releases/2025-05-06 - SNOMEDCT → SNOMEDCT_US_2024_09_01 - LOINC → LNC278 - NCIT → 24.01e

    (Version drift is now reported as a warning during validation.)

Breaking/behavioral notes

  • elementCode.system must match a CodeSystem.id (e.g., SNOMEDCT, LOINC).

  • Validation of SNOMED CT post-coordination / ECL (codes containing =) is skipped.

  • Version checks intentionally exclude systems in the skip list (configurable).

How to run

export BIOPORTAL_API_KEY=...   # required for live checks
rd-cdm-validate                # validates the resolved instance version

# Optional: validate a specific instance version folder
rd-cdm-validate --version 2.0.1

Version 2.0.0 (2025-02-08)

The RD-CDM has been updated to version 2.0.0 as the corresponding manuscript was published.

Version 2.0.0.dev0 (2024-09-30)

  • Initial release of the RD-CDM in development and review.