Metadata and Reference Data Management Strategy

1Summary

The trust will consume national content rather than recreate it, and keep local content only where local variation is legitimate.

Most trusts do not lack data. They lack agreement on what their codes mean, who may change them and which version a system is using. The visible symptoms are familiar: wards and clinics set up differently in the EPR and the PAS feed, specialty codes that fail national submissions, outdated SNOMED CT or dm+d content in departmental systems, and reports that disagree because each team maps local values in its own spreadsheet.

This strategy is deliberately modest. It does not require a new platform or a large team. It requires a small number of decisions about ownership, a catalogue, a change process, and discipline in applying national releases. A data steward resource of roughly 1 to 2 whole-time equivalents, drawn largely from existing information and EPR teams, is assumed.

2Scope and definitions

The strategy covers four related types of content. They are governed differently, so the trust must keep them distinct even where one system stores them together.

In scope: reference data and metadata used by the EPR, PAS, departmental systems, the data warehouse, national submissions and integration engines.

Out of scope: authoring of national terminology, patient-level data quality (covered by the data quality policy), and records retention. Patient master data is covered only where local reference data depends on it.

3Why this matters now

The trust carries statutory duties to apply national information standards, and the release cycles of the content it depends on are now too frequent to manage informally.

National obligations and cadences

The local problem

• Reference data is created inside each system by whoever administers it, with no record of meaning or owner.
• The same ward, clinic or specialty exists under different codes in the EPR, PAS, theatres, pharmacy and the warehouse.
• National updates are applied late, partially or not at all in departmental systems.
• Mappings between local and national codes live in spreadsheets and SQL scripts that only one or two people understand.
• When those people leave, the knowledge goes with them.

The Single Patient Record, shared care records and federated data platform work all depend on trust data that means what it says. Poor local reference data becomes a regional and national problem once it is shared.

4Principles

Eight principles govern decisions under this strategy. Where a proposal conflicts with one, the proposer must explain why.

1. Use national before local. If a national code system, dataset or definition exists, the trust uses it. Local lists are created only for genuinely local things, such as wards and rooms.
2. One authoritative source per list. Each reference dataset has one system or catalogue entry that is the master. Every other system receives it from there.
3. Every list has a named owner. No reference dataset or data definition is live without an accountable owner and a working steward.
4. Meaning before code. A new value is not added until its definition is written down.
5. Retire, do not delete. Values are end-dated, never removed, so historical records stay interpretable.
6. Mappings are directional and purposeful. A local to national mapping states its purpose (for example, national submission) and is not assumed to work in reverse.
7. Proportionate control. Clinically significant and nationally reported lists get formal change control. Low-risk administrative lists get a lighter route.
8. Separate meaning from presentation. Display names, local abbreviations and screen layouts can vary by service. The underlying code and definition cannot.

5Classes of reference data

Every reference dataset is placed in one of four classes, and the class determines who may change it and how. Classification is the first job of the catalogue.

Two rules follow. Class 1 content is never edited locally; a local need is raised with the national owner. A Class 3 value cannot go live until its national mapping is complete, because unmapped values are a frequent cause of rejected national submissions.

6Governance and roles

Accountability sits with the Senior Information Risk Owner through the existing data governance structure. The strategy adds named roles rather than new committees.

Decision rights in short: stewards decide routine changes; the CSO must agree anything clinically significant; the data governance group decides anything that crosses divisions or changes a national submission; the executive sponsor settles disputes.

7Core processes

Four processes cover almost all the work. Each is written down once, kept short and run by the stewards.

7.1 National release management

1. The terminology and reference data lead keeps a release calendar for every Class 1 source the trust uses, with the target time to apply each.
2. On each release, the lead reviews the change summary and lists affected local items: value sets, order sets, mappings, reports and forms.
3. Inactivated or changed clinical concepts in use are sent to the CSO for a safety view where they affect clinical decisions or displays.
4. Updates are applied in test, checked, then promoted to live, with the version recorded in the catalogue.
5. Suppliers that apply updates on the trust's behalf must report the version installed. The lead checks this monthly.

7.2 Local value change requests

1. Requester submits a short form: proposed value, definition, reason, class, and national mapping if Class 3.
2. Steward checks for duplicates and national equivalents.
3. Steward approves, or refers to the CSO or data governance group under the decision rights in section 6.
4. System administrators apply the change in the master source; downstream systems receive it through the agreed feed.
5. The catalogue is updated and the requester told.

Service target: five working days for Class 4 changes, fifteen for changes needing CSO or group review. Emergency changes may be made first and reviewed within five working days.

7.3 Mapping management

Each mapping records source list and version, target list and version, direction, purpose, author, reviewer, date and known limitations. Mappings are held in the catalogue or a governed table, not in report code. They are reviewed whenever either side changes.

7.4 Data definitions

Every item that appears in a board report, national submission or shared care feed has a catalogue entry giving its definition, source, derivation, owner and refresh. National definitions from the NHS Data Model and Dictionary are referenced, not copied and edited. New reports cannot go live without catalogue entries for their measures.

8Tooling

The trust should start with tools it already has and buy only when a measured need appears. A catalogue that people use matters more than its platform.

When buying, require that the product can export its content in an open format, records versions and effective dates, and exposes an API. These requirements also apply to EPR and departmental system procurements: suppliers must state which terminology versions they support, how updates reach live, and how local values can be exported.

9Roadmap

Delivery runs in three phases over 24 months. Each phase ends with a check by the data governance group before the next begins.

The priority list in phase 2 should be chosen from evidence: national submission rejection reports, data quality dashboards and known incidents. Start where errors are already costing the trust.

10Measures

Six measures are enough to show whether the strategy is working. Baselines are taken in phase 1, and targets set by the data governance group once baselines are known.

The measures are reported to the data governance group quarterly, and to the board through the annual Data Security and Protection Toolkit and data quality reporting.

11Risks

The largest risk is that the work is treated as a one-off clean-up and decays once the project team moves on.

12Decisions required and first 90 days

The executive sponsor is asked to approve the strategy and four decisions:

1. Designate or fund the terminology and reference data lead.
2. Extend the data governance group's terms of reference to cover reference data and metadata.
3. Confirm the CSO's role in terminology and value set change.
4. Require catalogue entries and change control for priority lists from the end of phase 1.

First 90 days

• Appoint the terminology and reference data lead and name stewards for the priority lists
• Agree the catalogue schema and set it up in an existing tool
• Inventory reference datasets in the EPR, PAS and three highest-volume departmental systems
• Classify each dataset into the four classes and name an owner
• Record current dm+d, SNOMED CT and ODS versions in every clinical system
• Build the national release calendar
• Launch the change request form on the service desk
• Pull 12 months of national submission rejections to choose the phase 2 priority lists

SSources

Accessed 3 October 2026.

• NHS England, dm+d and SNOMED CT UK Drug Extension release cycles
• NHS England, DAPB0084 OPCS Classification of Interventions and Procedures
• NHS England, national review of DCB0129 and DCB0160
• Data (Use and Access) Act 2025 commencement regulations, SI 2026/82