An open data standard and a permanent open catalogue for workplace health.

Two things, one project

The standard says what a workplace-health record is. The catalogue says what the measures inside it are worth, and where they come from. Each is published at three depths, so a first-time reader and a statistician each have a door.

Find your way in

Nobody needs all of this. Pick the door that matches why you came.

Why this exists

Payroll data moves between systems because standards exist. Banking got Open Banking. Healthcare got FHIR. But ask what a company's sickness absence actually is, and you get a spreadsheet export whose columns mean whatever the last HR system decided. The function holding the most sensitive data about working people runs on the least structured data in the building.

A national unit is now being stood up to collect standardised workplace-health data from employers and providers. OWHS is an open, independent proposal for what that data should look like at the level of an employer's records: aligned to the definitions the UK already trusts, implementable by the smallest employer, and versioned so that when official definitions arrive, they slot in rather than start over.

The full argument, including the survey of every existing standard and why none covers this ground, is in the case for a standard. How the specification was designed, and what was deliberately left open, is in the design notes.

Held as an open, versioned, forkable public record

Reference material in this field has a habit of disappearing. The What Works Centre for Wellbeing, which published the 2020 Workplace Wellbeing Question Bank, closed in April 2024; that bank now survives only as an archive. OWHS is built so the same cannot happen here.

1. Everything is versioned and downloadable. Each release of the specification, the registry and the question bank is a dated dataset anyone can keep a copy of.
2. Corrections are made in the open. The registry carries a public corrections log and a right of reply for instrument authors; the question bank carries a changelog. A grade or an item changes with a reason attached, never quietly.
3. Nothing is silently withdrawn. An item that loses its licence, or an instrument that leaves the field, stays catalogued with its provenance and the reason recorded.
4. The licences outlive the steward. Prose is CC-BY 4.0, structure and tools are Apache 2.0, and third-party wording stays with its source under its own terms. If stewardship changes hands, the record does not.

Privacy is conformance, not a promise

Workplace-health data can identify and harm people. So the privacy rules are part of the standard itself: a technically valid payload that violates them is non-conformant, and a conformant producer refuses to emit it.

1. No direct identifiers, ever. Pseudonymous worker references and banded demographics only, in every payload. Pseudonymised records are still personal data under UK GDPR: the standard reduces disclosure and re-identification risk, it does not make data anonymous.
2. Small groups are never reported in aggregates. Minimum aggregation of five for employer-visible aggregates, rising to ten for severe-distress measures, consistent with statistical disclosure control practice in official statistics. The separately enumerated individual-employer fields, such as a released occupational-health opinion, sit outside those floors where an independent legal basis permits the employer to hold them.
3. Individual results belong to the individual. Instrument results are never employer-visible at person level, by definition rather than by configuration.
4. Safeguarding signals are excluded entirely. Bullying, harassment, discrimination and crisis signals never appear in employer-visible outputs, at any group size.
5. Every number carries its context. Completion rates and suppression records travel with every aggregate, so consumers can see what is not being said.

What this is not

• Not a product. There is nothing to buy here. The specification, schemas, code lists and validator are open, under CC-BY 4.0 and Apache 2.0.
• Not a survey or a score. The standard defines the shape of the data, not who collects it or how it is interpreted. The catalogue describes instruments; it never reproduces them.
• Not a rival to the national effort. Version 0.x exists to be useful now and to be overwritten gracefully: the specification reserves space for the Workplace Health Intelligence Unit's definitions and will adopt them when they exist.
• Not clinical. It carries no diagnosis semantics, and it treats clinical coding (SNOMED CT) as an optional extension for those licensed to use it.

Get involved

This is version 0.2, published for review, with version 0.1 archived beside it. The full specification, schemas and change process (open issues, RFCs, a public decision log) live at github.com/openworkplacehealth/OWHS. Tell us where it breaks. The contribute page sets out the three small asks, the call for independent registry raters, and the interest declaration every contribution carries. You do not need to endorse the standard or join anything.