Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Health data anonymisation is the process of removing the ability to identify individuals from clinical data, so that it can be used for research and planning without remaining personal data.

What Is Health Data Anonymisation?
Anonymisation is often described as removing names. It is not. Names are rarely what identifies someone in a clinical data set; the combination of attributes is. Latanya Sweeney demonstrated in the 1990s that date of birth, sex, and postal code alone uniquely identified a large majority of the United States population, and used that observation to re-identify a state governor in a supposedly anonymous hospital discharge file that had been released publicly. Every subsequent failure in this field is a variation on the same theme.
The legal distinction decides whether data protection law applies at all. Under UK GDPR, genuinely anonymous information falls outside the regulation. Pseudonymised data, where identifiers have been replaced with a key that could be reversed by someone holding the mapping, remains personal data and remains fully in scope. Most health data described in projects as anonymised is in fact pseudonymised, and the distinction is worth insisting on because the obligations that follow are entirely different.
The ICO’s position is that anonymisation is assessed against the risk of re-identification by a motivated intruder: a reasonably competent person with access to public resources and an interest in identifying someone, but no specialist skills or criminal intent. That is a contextual test, not a property of a file. The same data set can be anonymous when held in a controlled environment and personal data when published openly, which is why release decisions depend on the environment as much as the transformation.
Re-identification research has been consistent and uncomfortable. The Netflix Prize data set was de-anonymised by cross-referencing public film ratings. An Australian medical billing data set released openly in 2016 was shown by researchers at the University of Melbourne to permit re-identification of providers and, in combination with other sources, of patients, and was withdrawn.
The common factor is auxiliary information. Anonymisation is evaluated against the data available today, and new sources keep appearing: social media posts, published registries, other released data sets, and the individual’s own knowledge of their care. A release that was defensible in 2015 may not be in 2026, and there is no mechanism to recall an open data set.
This is why health systems have moved away from releasing data and towards bringing researchers to it. Trusted research environments, called secure data environments in NHS England’s policy and strongly endorsed by the Goldacre review in 2022, keep data in a controlled setting where analysts work on it without taking it away. Outputs are checked before release, access is approved and logged, and the environment itself is assessed. The Five Safes framework, long used by the Office for National Statistics, structures the decision around safe people, safe projects, safe settings, safe data, and safe outputs, which is a better way to think about the problem than a transformation applied to a file.
In Turkey, anonymisation is addressed directly in Personal Data Protection Law No. 6698 and the implementing regulation on deletion, destruction, and anonymisation, with guidance from the Turkish authority setting out accepted techniques including masking, aggregation, generalisation, derivation, and noise addition. Health data is special category data, and the Ministry of Health’s own rules govern secondary use of clinical information, including research access to the national health data described under e-Nabiz and MHRS. A UK sponsor planning Turkish research should expect approval routes that differ from the UK’s, with the Ministry as a gatekeeper alongside the data protection regime.
One final practical point. Whoever produces an anonymised extract should not be the only person who judges whether it is anonymous. Independent review, a documented motivated intruder assessment, and a re-evaluation when new data sets become public are what separate a defensible release from one that merely has not been tested yet. The controller and processor roles should also be settled before any extract moves, because an anonymisation performed by a supplier is still processing of personal data until it succeeds.
At Privia Security, we review anonymisation designs and the environments they depend on through our regulation and compliance services.
Yes. Under UK GDPR, replacing identifiers with a key that could be reversed by anyone holding the mapping leaves the data within scope of the regulation, with all the obligations that follow. Only genuinely anonymous information, where re-identification is not reasonably likely for any party in the relevant context, falls outside. Most health data described as anonymised is in fact pseudonymised.
The ICO’s standard for assessing whether data is anonymous. It asks whether a reasonably competent person, motivated to identify someone and willing to use public resources and reasonable effort, but without specialist skills or unlawful methods, could succeed. It is contextual: the same data set may be anonymous inside a controlled environment and personal data if published openly.
No. Synthetic records preserve statistical properties of the original, and a generative model trained on a small cohort can reproduce real individuals or allow inference about whether someone was in the training set. Synthetic data is useful for development, testing, and method prototyping, but it needs the same privacy evaluation as any other release before it leaves a controlled environment.
Because anonymisation strong enough to survive uncontrolled release usually destroys the research value, and because a released data set cannot be recalled when new auxiliary information appears. Bringing researchers to the data in a controlled environment, with approved projects, logged access, and checked outputs, preserves utility while keeping the release decision reversible and the risk manageable.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.