Open Data in Public-Health Research: Standards

Public health data describe populations and the events that affect them, and increasingly they are expected to be shared so that findings can be verified, combined and reused. Open data means data made available for others to access, use and redistribute under clear terms, while the FAIR principles set out that data should be Findable, Accessible, Interoperable and Reusable. Together they define a standards-based approach to managing research data responsibly, balancing the value of openness against the obligations that attach to data about people.

These ideas have moved from aspiration to expectation: funders, journals and institutions increasingly ask researchers to describe how their data will be managed and shared. Understanding the standards involved is now part of doing public-health research well.

FAIR data principles

The FAIR principles are a widely adopted framework for good data stewardship, and each letter carries a specific meaning. Findable means data carry rich metadata and a persistent identifier so they can be located by both people and machines. Accessible means the data, or at least their metadata, can be retrieved through a standard, open protocol, with any access conditions clearly stated. Interoperable means data use shared vocabularies, standards and formats so they can be combined with other datasets without bespoke translation. Reusable means data are richly described, with clear provenance and an explicit usage licence, so that others can determine whether and how the data may be reused. A crucial point is that FAIR is about machine-actionable description; it does not require data to be fully open, which matters for sensitive public-health data that cannot simply be published. This is a common source of confusion: a dataset held under strict access controls can still be fully FAIR if it is well described, carries a persistent identifier, uses interoperable formats and states clear terms of reuse. The principles describe how data should be prepared and exposed to the world, not whether the data themselves must be freely downloadable by anyone.

Data-sharing standards

Responsible sharing rests on a few recurring components that turn the FAIR principles into practice. Datasets are described with structured metadata, assigned persistent identifiers, deposited in trustworthy repositories, and released under explicit licences. These practices map directly onto FAIR and are reinforced across the wider data infrastructure community and the controlled vocabulary in the CASRAI dictionary.

De-identification and anonymisation

Because public-health data can relate to identifiable individuals, sharing must protect privacy before any wider release. At a high level, de-identification removes or transforms information that could identify a person, while anonymisation aims to reduce the risk of re-identification to an acceptable level given the data and the context. The specific techniques, thresholds and acceptable residual risk are governed by law and institutional policy rather than by a single universal rule, and they are applied before data are opened. The key methodological point, framed here as governance rather than clinical guidance, is that openness and privacy protection are managed together as a deliberate design choice, not traded off blindly.

Interoperability and controlled vocabularies

The Interoperable element of FAIR depends heavily on shared controlled vocabularies and standard formats, which let data from different sources be combined without ambiguity. When two datasets use the same defined term to mean the same thing, they can be linked and analysed together; when they use different words, or the same word for different concepts, integration becomes error-prone. Public-health data are particularly affected, because the same measure can be defined in subtly different ways across systems, as the distinction between crude and standardised rates illustrates. Adopting recognised vocabularies and classifications, and recording exactly which version was used, turns interoperability from an aspiration into a practical property of the data. This is also where definitional infrastructure earns its keep: a stable, shared set of terms reduces the translation effort every time datasets are combined, and it makes automated reuse far more reliable.

Metadata and persistent identifiers

Metadata are what make a dataset interpretable by someone who did not create it: they record what was measured, how, when, over which population and under what definitions, and they capture the provenance of the data. Without good metadata, even openly available data are difficult to reuse correctly. Persistent identifiers, such as a DOI assigned to a dataset, give it a stable address so that it can be cited and tracked over time even if its location changes. Together, rich metadata and persistent identifiers make a dataset Findable and Reusable, and they let derived measures such as incidence and prevalence be traced back to the exact source data that produced them.

Data management plans and the research lifecycle

Standards-based sharing is most effective when it is planned from the start rather than bolted on at publication. A data management plan sets out, early in a project, how data will be collected, documented, stored, protected and ultimately shared, and many funders now require one. Building FAIR practice into the plan means deciding in advance which repository will hold the data, what metadata standard will describe it, how identifiers will be assigned, and what governance will control access. This lifecycle view avoids a common failure mode in which data that could have been shared are effectively lost because they were never documented well enough to be understood later. It also makes de-identification and licensing decisions deliberate rather than rushed, and it connects neatly to reproducibility, because a dataset that is well managed throughout its life is far easier for others to verify and reuse.

Governance and reuse

Open does not mean unmanaged. Governance frameworks set out who may access data, under what conditions, and for what purposes, balancing the goal of transparency against privacy, consent and legal obligations. For sensitive data, this often means controlled access through an application and approval process rather than unrestricted download, which can still be FAIR. Clear governance, combined with transparent reporting guidelines such as STROBE, supports trustworthy reproducibility and responsible reuse, because it lets others understand both how the data were produced and how they may legitimately be used. Researchers can consult the guidance for authors to align their data-sharing practice with these standards from the start of a project.

Frequently asked questions

Does FAIR mean the same thing as open?

No. FAIR concerns how well data are described and structured for discovery and reuse, including by machines. Data can be FAIR while access remains controlled, which is common for sensitive public-health datasets where full openness is constrained by privacy and legal duties.

Why are persistent identifiers important?

A persistent identifier gives a dataset a stable, citable reference that does not break over time even if the data move. It supports findability, enables proper data citation and credit, and lets analyses be traced back to the exact source data used.

How are privacy and openness reconciled?

Through de-identification or anonymisation applied before release, combined with governance frameworks that define access conditions. The aim is to maximise responsible reuse while keeping the risk of re-identification within legally and ethically acceptable limits.