{"id":174,"date":"2026-09-28T10:29:24","date_gmt":"2026-09-28T09:29:24","guid":{"rendered":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/?p=174"},"modified":"2026-09-29T16:46:01","modified_gmt":"2026-09-29T15:46:01","slug":"alls-fair-in-love-and-raw-healthcare-data","status":"publish","type":"post","link":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/alls-fair-in-love-and-raw-healthcare-data\/","title":{"rendered":"All\u2019s FAIR in Love and Raw (Healthcare Data)"},"content":{"rendered":"<h4>By James Ellaway the Data Lead at Cardiff Experimental Cancer Medicine Centre (ECMC)<\/h4>\n<h2>What is FAIR data?<\/h2>\n<p>We love an acronym in healthcare. So, why would healthcare data be any different? FAIR data means:<\/p>\n<p style=\"text-align: center\"><img decoding=\"async\" class=\"aligncenter\" style=\"width: 338px;height: 338px\" src=\"http:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-content\/uploads\/sites\/681\/2026\/09\/05-James_01_Acronym.jpg\" alt=\"\" \/><\/p>\n<p>Each of us who has worked with not only healthcare data, but any kind of \u2018scholarly\u2019 data \u2013 probably knows and understands all-too-well, the trials and tribulations of navigating different datasets in different formats and across different organisations at different times under different management structures under different levels of stewardship (I could go on\u2026) \u2013 and the all-too-real impact this can (and does) have upon the robustness of the data, and the workflows, which inform our research and operational decision making.<\/p>\n<figure id=\"post-188 media-188\" class=\"image align-right\"><img decoding=\"async\" style=\"width: 336px;height: 336px\" src=\"http:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-content\/uploads\/sites\/681\/2026\/09\/06-James_Clip-Art_Softer-Rework.jpg\" alt=\"\" \/><\/figure>\n<p>Whilst the FAIR principles are not a manifesto for the standardisation of data as such, they are a set of guiding \u2018best practices\u2019 which, if everyone followed, would lead to a potential standardisation of embedded processes relating to the management and stewardship of data, at least. For each principle, there are a set of \u2018sub-principles\u2019 attached.<\/p>\n<p>Moreover, with the ever-proliferating use of machines in data creation, collation, analysis, etc, the FAIR principles were designed predominantly with automated machine-use, or \u2018computational stakeholders\u20191 in mind (futureproofing, you could say).<\/p>\n<p>This was the intention of the FAIR principles founding members; a consortium of academics, industry folk, research funders, and publishing organisations, who published The FAIR Guiding Principles for scientific data management and stewardship in 2016 in Scientific Data, citing:<\/p>\n<p style=\"padding-left: 40px\"><em><span style=\"color: #993300\">\u201can urgent need to improve the infrastructure supporting the reuse of scholarly data\u2026[and that] Good data management is not a goal in itself, but rather is the key conduit leading to knowledge discovery and innovation, and to subsequent data and knowledge integration and reuse by the community after the data publication process\u2026\u201d \u00a0Ref2<\/span><\/em><\/p>\n<p>Whilst the FAIR principles ultimately assist the human-user as well, the consortium were keen to differentiate from other peer initiatives at the time that put the \u2018human scholar\u2019 at the forefront.<\/p>\n<p>The desired outcomes might be summarised as the following:<\/p>\n<ul>\n<li>High quality digital publications and resources.<\/li>\n<li>A simplified, ongoing process of data (and knowledge) discovery, evaluation, and \u2018downstream\u2019 reuse.<\/li>\n<li>Maximising added value of data and the tools and workflows used that led to the data.<\/li>\n<li>All scholarly digital research objects benefit from increased transparency, reproducibility, and reusability.<\/li>\n<\/ul>\n<p>So, let\u2019s get into the specifics\u2026<\/p>\n<hr \/>\n<p style=\"text-align: center\"><img loading=\"lazy\" decoding=\"async\" class=\"\" src=\"http:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-content\/uploads\/sites\/681\/2026\/09\/05A-Jamesword-01-Findability-scaled.jpg\" alt=\"\" width=\"300\" height=\"113\" \/><\/p>\n<p>What good is data if we can\u2019t find it in the first place? This is all about making your data (and metadata!) as easy as possible for machine-kind and humankind alike to find. Making metadata machine-legible = easy automated discovery of datasets. Essential to the core ideas behind these principles. Findability comprises the following sub-principles:<\/p>\n<h4>F1. Data and all metadata are assigned globally unique and persistent identifiers (GUPIs).<\/h4>\n<ul>\n<li>Identifiers here refers to a URL (internet hyperlink) \u2013 sometimes called a Digital Object Identifier (DOI).<\/li>\n<li>This means that each and every aspect of your data (and metadata) can be compiled in an easily-accessible\/findable way \u2013 and removes any barrier of uncertainty over the meaning of your data, whether they be specific measurements, concepts, etc, the DOI \u2018points\u2019 you to the exact definition.<\/li>\n<li>Moreover, this exactness allows machines to interpret the data accurately when given a directive; e.g., automatic data parsing\/integration. Embedding a level of trust in the outputs of machines when using these data.<\/li>\n<li>This human-machine interplay is a core aspect of open science3 methodology in the findability, reusability, and adaptability of data.<\/li>\n<li>Identifiers also allow others to accurately cite work when reusing data.<\/li>\n<li>ALL identifiers must abide these two conditions to meet the mark:\n<ul>\n<li>It must be globally unique, i.e., nobody else can reuse or reassign the same identifier. Registry services such as http:\/\/www.doi.org and http:\/\/www.identifiers.org ensure this uniqueness using algorithms.<\/li>\n<li>The identifier must persist. That is to say, it can never expire (or at the very least, needs to be resolvable).<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p>F2. Data are described with rich metadata.<\/p>\n<ul>\n<li>Metadata should be substantial in its descriptive content and include information as to the context, condition, quality, and properties of the data to which it is attached.\n<ul>\n<li>This allows for users to find data based on the descriptive content of the metadata alone, without knowing the unique identifier of the data required, for example.<\/li>\n<li>This increases findability and reuse of data overall.<\/li>\n<\/ul>\n<\/li>\n<li>Metadata comprises both intrinsic and contextual subtypes:\n<ul>\n<li>Intrinsic: data captured automatically by machines (think the kind of metadata embedded in a digital photographic image, e.g., the file type, date and time it was taken, the aperture and shutter speed used, etc.)<\/li>\n<li>Contextual: additional metadata supporting the former intrinsic data, for example, the photographer\u2019s details, the location of the \u2018photoshoot\u2019, the campaign the photographs were made for.<\/li>\n<\/ul>\n<\/li>\n<li>All of which can be searchable and assist in the retrieval and reusability of data.<\/li>\n<li>As a general rule of thumb, more is more!<\/li>\n<\/ul>\n<h4>F3. Metadata clearly and explicitly contain the identifier of the data they describe.<\/h4>\n<ul>\n<li>\u00a0The data and metadata are usually separate files in themselves, and therefore it is vital that the metadata refer specifically to the data it is associated with by including details of the data\u2019s globally unique and persistent identifier.<\/li>\n<\/ul>\n<h4>F4. Metadata are registered or indexed in a searchable resource.<\/h4>\n<ul>\n<li>Perfectly \u2018good\u2019 data resources may go unused if nobody knows that they exist. Therefore, we should ensure that data are \u2018organised\u2019 as efficiently as possible:\n<ul>\n<li>Digital resources, such as DOIs can be made more easily \u2018discoverable\u2019 by internet searches for example, by using techniques such as indexing.<\/li>\n<li>Indexing is the process of structuring and tagging (categorising) data and metadata so that it is more readily searchable and retrievable by machines and humans alike.<\/li>\n<li>The more organised and better categorised the data, the easier it will be to find in search engines and repositories, etc.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<hr \/>\n<p style=\"text-align: center\"><img loading=\"lazy\" decoding=\"async\" class=\"\" src=\"http:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-content\/uploads\/sites\/681\/2026\/09\/05B-Jamesword-02-Accessibility-scaled.jpg\" alt=\"\" width=\"300\" height=\"113\" \/><\/p>\n<p>So, we\u2019ve found our data, but now we need to access it. Accessibility can vary between repositories, often requiring authorisation or authentication. Certain steps can be made to make sure that data are as easily accessible as possible. The following sub-principles ensure this:<\/p>\n<h4>A1. (Meta)data are retrievable by their identifier using a standardised communications protocol.<\/h4>\n<ul>\n<li>In plain language: all data should be accessible without the need for any specialist tools or communication methods, i.e., minimise barriers to access e.g., by common methods such as standard http(s) weblinks accessible by any internet browser (e.g., Google Chrome, Apple Safari, etc). In essence, anyone with a computer and internet connection should be able to access the metadata at the very least.<\/li>\n<li>The conditions to method of access (protocol) should be of no-cost and open-sourced to ensure maximum potential for data retrieval, reuse, and implementation.<\/li>\n<li>With the exception of highly sensitive or personalised data (see my blog on section 251 exemptions) which may require a level of access authorisation, in which case it would be encouraged to provide the contact details for the data owner to discuss access, for example.<\/li>\n<li>This contact information and access process must be made explicit in the metadata.<\/li>\n<\/ul>\n<h4>A2. Metadata should be accessible even when the data is no longer available.<\/h4>\n<ul>\n<li>Whilst every measure should be put in place to ensure the longevity of data (and their identifiers), occasionally data will disappear or decay due to maintenance costs, e.g., web-hosting upkeep, etc.<\/li>\n<li>Links to data can sometimes resultantly become dead ends in cases like these, and there is argument that storage of metadata is considered easier and cheaper. Efforts should be made to ensure their continued existence and accessibility, even if the data to which they are associated is no longer available.<\/li>\n<li>These metadata can be invaluable in themselves, as they often (and should often) contain details as to the researchers, institutions, outputs such as publications associated with the original data and research undertaken.<\/li>\n<\/ul>\n<hr \/>\n<p style=\"text-align: center\"><img loading=\"lazy\" decoding=\"async\" class=\"\" src=\"http:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-content\/uploads\/sites\/681\/2026\/09\/05C-Jamesword-03-Interoperability-scaled.jpg\" alt=\"\" width=\"301\" height=\"113\" \/><\/p>\n<p>Data -usually- need to be combined with, or at least be compatible with, other data. This interoperability needs to expand across applications, systems, and workflows for all manner of storage, analysis, or processing needs. Interoperable sub-principles are as follows:<\/p>\n<h4>I1. (Meta)data use a formal, accessible, shared, and broadly applicable language for knowledge representation.<\/h4>\n<ul>\n<li>Or: people need to be able to share and understand each other\u2019s data. Moreover, this also applies to machines; there should be no need for specialist skills, or intervention such as translation, to make the data legible.<\/li>\n<li>Interoperability ensures a basis of pre-determined \u2018common understanding\u2019, or universality of reading, of digital objects and the language used to represent them.<\/li>\n<li>Whatever language used, it should be chosen based on the following specifications:\n<ul>\n<li>Its syntax and grammar are precisely defined.<\/li>\n<li>These definitions should be shared in an accessible way that allows other users to learn the language.<\/li>\n<li>It should be designed to be used in more than one, singular, setting\/application.<\/li>\n<li>One such example is the JSON LD4<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h4>I2. (Meta)data use vocabularies that follow FAIR principles.<\/h4>\n<ul>\n<li>Whatever vocabularies are used to describe our data and metadata, these should also follow FAIR principles, so that like the data themselves, others \u2013 whether human or machine \u2013 can find, access, interoperate and reuse these also.<\/li>\n<li>Vocabulary therefore needs to be controlled, i.e., documented and resolvable themselves using GUPIs.<\/li>\n<li>This documentation, like the data, also need to be easily findable and accessible.<\/li>\n<\/ul>\n<h4>I3. (Meta)data include qualified references to other (meta)data.<\/h4>\n<ul>\n<li>E.g., a cross-reference describing its purpose.<\/li>\n<li>This is a process of referential \u2018linking\u2019 of (meta)data which explains clearly their relationship to each other. This enhances the robustness of contextual information about the data.<\/li>\n<li>Scientific links between datasets must be described, and all datasets need proper citation (including their GUPIs).<\/li>\n<li>This is particularly important when building from existing datasets, or if additional data are required to complete the data.<\/li>\n<\/ul>\n<hr \/>\n<p style=\"text-align: center\"><img loading=\"lazy\" decoding=\"async\" class=\"\" src=\"http:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-content\/uploads\/sites\/681\/2026\/09\/05D-Jamesword-04-Reusability-scaled.jpg\" alt=\"\" width=\"301\" height=\"113\" \/><\/p>\n<p>Ultimately, FAIR principles seek to create data optimised for continued reuse and reproducibility. Meta data and data therefore need to be well described for reproducibility or combination in many different settings or applications. This can be achieved by following the sub-principles outlined below:<\/p>\n<h4>R1. (Meta)data are richly described with a plurality of accurate and relevant attributes.<\/h4>\n<ul>\n<li>Whilst related to data findability, this principle is focused more on the relevance or usefulness of the data in a particular chosen context. This needs to be determinable by human and machine.<\/li>\n<li>Metadata therefore needs to describe the context in which the data to which it is attached was created: e.g., scope and purpose of the data\/research, particularities such as protocols, machine brand\/manufacturer information, species type, any noted limitations of the data, whether it has been processed or is raw data, clear documentation of the version\/iteration of the dataset, etc.<\/li>\n<li>This metadata should be unbiased and factual, not anticipating or precluding the data user\u2019s need for the data nor their identity.<\/li>\n<li>Make clear what usage rights are attached to the data, e.g., licensing status. E.g., the (meta)data\u2019s legal interoperability (and limitations).<\/li>\n<li>It should be made clear where the data came from (its origin and history), who to cite for publication and terms of acknowledgement, whether the data have been published before, how the data have been processed, etc.<\/li>\n<li>Follow community standards or best practices of data storage, protection, and sharing if they exist, e.g., MIAME standards. If such standards can\u2019t be met, this should be explained in the metadata for other user\u2019s understanding.<\/li>\n<\/ul>\n<hr \/>\n<p>So, there you have it. At its core, FAIR data provides a framework to ensure a minimum level of standardisation and onward data utilisation through its Findability, Accessibility, Interoperability, and Reusability.<\/p>\n<p>&nbsp;<\/p>\n<hr \/>\n<h6>References<br \/>\n1. Wilkinson et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3, 160018 (2016). https:\/\/doi.org\/10.1038\/sdata.2016.18<br \/>\n2. Wilkinson et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3, 160018 (2016). https:\/\/doi.org\/10.1038\/sdata.2016.18<br \/>\n3. www.dtls.nl (2016). Vision on Open Science. https:\/\/www.dtls.nl\/2016\/04\/14\/5825-2\/<br \/>\n4. https:\/\/json-ld.org (2026). JSON for Linking Data. JSON-LD &#8211; JSON for Linked Data<\/h6>\n<hr \/>\n<h6><em>\u00a0<\/em><em>Note: Header image imagined from the text by artist <a href=\"https:\/\/www.chrisglynn.net\" target=\"_blank\" rel=\"noopener\">Chris Glynn<\/a><\/em><\/h6>\n","protected":false},"excerpt":{"rendered":"By James Ellaway the Data Lead at Cardiff Experimental Cancer Medicine Centre (ECMC)","protected":false},"author":7176,"featured_media":183,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5],"tags":[],"class_list":["post-174","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-accessing-data"],"meta_box":[],"_links":{"self":[{"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/posts\/174","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/users\/7176"}],"replies":[{"embeddable":true,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/comments?post=174"}],"version-history":[{"count":22,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/posts\/174\/revisions"}],"predecessor-version":[{"id":302,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/posts\/174\/revisions\/302"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/media\/183"}],"wp:attachment":[{"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/media?parent=174"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/categories?post=174"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blogs.cardiff.ac.uk\/bioresource-data-accelerator\/wp-json\/wp\/v2\/tags?post=174"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}