The Commissioner, with the support of the Court of Session, had decided that, as a matter of fact, a set of anonymised medical statistics were not personal data and had ordered their release under Freedom of Information (FOI) laws.
According to one expert, the Lords' judgment shows that the Commissioner had failed to appreciate that the degree of anonymisation employed was very likely to be insufficient to protect anonyminity.
The Lords overturned decisions of the Court of Session and the Scottish Information Commissioner and required the Commissioner to re-examine the original request.
The Lords conducted a two day hearing in April on a case which directly pitted the Scottish Freedom of Information Act against the Data Protection Act.
FOI legislation is designed to release information generated by public bodies while data protection laws restrict people's access to information which can be classed as personal data.
The Lords said that the information that Scottish Parliamentary researcher Michael Collie requested from NHS agency the Common Services Agency (CSA) counted as personal data and should not be released.
Collie had asked the CSA to provide him with "details of all incidents of leukaemia for both sexes in the age range 0–14 by year from 1990–2003 for all the DG [Dumfries and Galloway] postal area by census ward".
When figures are so low as to be capable of identifying individuals they can be 'barnardised'. This is a method designed by statistician Professor George Barnard that helps to disguise people's identities when cells of information contain low numbers. It randomly adds zero, one, or minus one to the numbers two, three and four; and adds zero or one to the number one, when they appear in a database. Zeros are left at zero.
The Scottish Information Commissioner (SIC) Kevin Dunion had originally ruled that the anonymised data was not personal and should be released.
This has been overturned, and Dunion came in for stiff criticism in the Lords ruling.
"It has to be said, with respect, that the approach which he took … suffers from a number of defects," the Lords said. "He did not ask himself whether the barnardised data would be personal data within the meaning of section 1(1) DPA" and, if so, whether its disclosure would breach a data protection principles. Indeed "he did not find it necessary to consider whether release of the data in that form would be in accordance with the data protection principles," said the ruling.
"The effect of the Commissioner’s decision was to require the Agency to release information to Mr Collie, not just to give him advice or assistance," said the judgment. "[The SIC] did not pursue the point to its proper conclusion" and "this was an error of law".
The Scottish court decisions that supported the SIC were based on an error, the Lords said. The House of Lords judgment states that in an attempt to resolve the case, the "First Division the Lord President looked for guidance as to how to approach the problem to the decision of the Court of Appeal in Durant v Financial Services Authority".
The Durant case is a landmark in data protection law. Michael Durant had requested all the information held on him by Barclays Bank, but the case ruled that this did not include every single document that mentioned him. It narrowed the definition of personal data.
This not only restricted what should be released under Data Protection Act subject access requests, but opened up the amount of data which could be released under FOI legislation. The FOI Act bars the release of personal data, but if that is defined narrowly as in the Durant case then the volume of data that is releasable grows.
The Lords, though, said that the Durant case was irrelevant. "[Durant's] was a case where the person who was seeking disclosure of the information was the data subject, as he was the individual who was the subject of the personal data to which he request related". However in this case "the only element in question is the identification of the individual to whom the data relate". It follows that "there is no need in this case to consider the kinds of issue which the Court of Appeal addressed in Durant".
Dr Chris Pounder, an information law expert with Pinsent Masons and editor of Data Protection Quarterly, said that the Lords were worried about the impact of their decision. "The real issue which taxed the Court for two days was that if it decided in favour of the CSA, whether its decision would halt the ability of the individuals to use FOI to obtain anonymous statistics in general."
"The Court readily accepted that there is a significant public interest in disclosing statistical data under FOI regimes and was very concerned that its judgment might make other similar disclosures of sensitive personal data very unlikely," he said. "But I think they're basically saying that, on the facts of this case, the Commissioner failed to recognise that the degree of anonymisation employed would fail to protect anonyminity."
Pounder said that those in the data protection field will be disappointed that the House of Lords did not take the opportunity to rule on the issues covered so controversially in the Court of Appeal in the Durant case, but that the ruling clears up some crucial issues.
"I think the House of Lords have crafted a decision that will have minimal impact on the public and the FOISA and FOI regime," he said. "However, people who work in data protection will find the consideration of the definition of 'data', 'personal data' and 'sensitive personal data' are well worth the effort."
The Lords have remitted the case back to the Commissioner to see whether the leukaemia statistics in question are capable of being rendered fully anonymous. This establishes the principle that public authorities can be required to anonymise data provided the cost to the authority does not exceed a certain limit (currently £600 in Scotland) and that, in so doing, they are not to be viewed as creating new information. If the statistics are not capable of being rendered fully anonymous then the Commissioner has to consider whether the disclosure of the sensitive health personal data would breach any of the data protection principles if they were to be published.
Editor's note, 10/07/2008: Kevin Dunion, the Scottish Information Commissioner, wrote to OUT-LAW today regarding the original version of this story:
"The heading to the article states that 'A set of anonymised medical statistics about a region is still personal data and should not be released.' No such finding was made by the Lords. Instead, the Lords have remitted the decision to me to satisfy myself whether the leukaemia statistics in question are capable of being rendered fully anonymous. If they are capable of being rendered fully anonymous, then the anonymised statistics will not, according to the Lords, be personal data. If I am satisfied that the statistics are not capable of being rendered fully anonymous, then I will go on to consider whether the disclosure of the data would breach any of the data protection principles.
"I am concerned that your readers, who may not have had an opportunity to read the judgement in full, may come to the wrong conclusions about the outcome in this case, which importantly also confirmed that public authorities can be required to anonymise data and that, in so doing, they are not to be viewed as creating new information. Clearly, this is important in deciding whether information is 'held' by a public authority for the purposes of the Freedom of Information (Scotland) Act."
We have amended the story to make this clear. We apologise to Mr Dunion and our readers for the inaccuracy in our original intro.
Training for your organisation: Pinsent Masons offers Data protection training and also Freedom of Information training.
Footnote: Dr Chris Pounder was a consultant with Pinsent Masons until September 2008. He now runs a new training business, Amberhawk.