Do your research. Properly

Campaigning group Big Brother Watch have released a report entitled “NHS Data Breaches”. It purports to show the extent of such “breaches” within the NHS. However it fails properly to define its terms, and uses very questionable methodology. I think, most worryingly, this sort of flawed research could lead to a reluctance on the part of public sector data controllers to monitor and record data security incidents.

As I checked my news alerts over a mug of contemplative coffee last Friday morning, the first thing I noticed was an odd story from a Bedfordshire news outlet:

Bedford Hospital gets clean bill of health in new data protection breach report, unlike neighbouring counties…From 2011 to 2014 the hospital did not breach the data protection act once, unlike neighbours Northampton where the mental health facility recorded 346 breaches, and Cambridge University Hospitals which registered 535 (the third worst in the country).

Elsewhere I saw that one NHS Trust had apparently breached data protection law 869 times in the same period, but many others, like Bedford Hospital had not done so once. What was going on – are some NHS Trusts so much worse in terms of legal compliance than others? Are some staffed by people unaware and unconcerned about patient confidentiality? No. What was going on was that campaigning group Big Brother Watch had released a report with flawed methodology, a misrepresentation of the law and flawed conclusions, which I fear could actually lead to poorer data protection compliance in the future.

I have written before about the need for clear terminology when discussing data protection compliance, and of the confusion which can be caused by sloppiness. The data protection world is very found of the word “breach”, or “data breach”, and it can be a useful term to describe a data security incident involving compromise or potential compromise of personal data, but the confusion arises because it can also be used to describe, or assumed to apply to, a breach of the law, a breach of the Data Protection Act 1998 (DPA). But a data security incident is not necessarily a breach of a legal obligation in the DPA: the seventh data protection principle in Schedule One requires that

Appropriate technical and organisational measures shall be taken [by a data controller] against unauthorised or unlawful processing of personal data and against accidental loss or destruction of, or damage to, personal data

And section 4(4) of the DPA obliges a data controller to comply with the Schedule One data protection principles. This means that when appropriate technical and organisational measures are taken but unauthorised or unlawful processing, or accidental loss or destruction of, or damage to, personal data nonetheless occurs, the data controller is not in breach of its obligations (at least under the seventh principle). This distinction between a data security incident, and a breach, or contravention, of legal obligations, is one that the Information Commissioner’s Office (ICO) itself has sometimes failed to appreciate (as the First-tier Tribunal found in the Scottish Borders Council case EA/2012/0212). Confusion only increases when one takes into account that under The Privacy and Electronic Communications (EC Directive) Regulations 2003 (PECR) which are closely related to the DPA, and which deal with data security in – broadly – the telecoms arena, there is an actual legislative provision (regulation 2, as amended) which talks in terms of a “personal data breach”, which is

a breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorised disclosure of, or access to, personal data transmitted, stored or otherwise processed in connection with the provision of a public electronic communications service

and regulation 5A obliges a relevant data controller to inform the ICO when there has been a “personal data breach”. It is important to note, however, that a “personal data breach” under PECR will not be a breach, or contravention, of the seventh DPA data protection principle, provided the data controller took appropriate technical and organisational to safeguard the data.

Things get even more complex when one bears in mind that the draft European General Data Protection Regulation proposes a similar approach as PECR, and defines a “personal data breach” in similar terms as above (simply removing the words “in connection with the provision of a public electronic communications service“).

Notwithstanding this, the Big Brother Watch report is entitled “NHS Data Breaches”, so one would hope that it would have been clear about its own terms. It has led to a lot of coverage, with media outlets picking up on headline-grabbing claims of “7225 breaches” in the NHS between 2011 and 2014, which is the equivalent to “6 breaches a day”. But when one looks at the methodology used, serious questions are raised about the research. It used Freedom of Information requests to all NHS Trusts and Bodies, and the actual request was in the following terms

1. The number of a) medical personnel and b) non-medical personnel that have been convicted for breaches of the Data Protection Act.

2. The number of a) medical personnel and b) non-medical personnel that have had their employment terminated for breaches of the Data Protection Act.

3. The number of a) medical personnel and b) non-medical personnel that have been disciplined internally but have not been prosecuted for breaches of the Data Protection Act.

4. The number of a) medical personnel and b) non-medical personnel that have resigned during disciplinary procedures.

5. The number of instances where a breach has not led to any disciplinary action.

The first thing to note is that, in broad terms, the only way that an individual NHS employee can “breach the Data Protection Act” is by committing a criminal offence under section 55 of unlawfully obtaining personal data without the consent of the (employer) data controller. All the other relevant legal obligations under the DPA are ones attaching to the NHS body itself, as data controller. Thus, by section 4(4) the NHS body has an obligation to comply with the data protection principles in Schedule One of the DPA, not individual employees. And so, except in the most serious of cases, where an employee acts without the consent of the employer to unlawfully obtain personal data, individual employees, whether medical or non-medical personnel, cannot as a matter of law “breach the Data Protection Act”.

One might argue that it is easy to infer that what Big Brother Watch meant to ask for was information about the number of times when actions of individual employees meant that their employer NHS body had breached its obligations under the DPA, and, yes, that it probably what was meant, but the incorrect terms and lack of clarity vitiated the purported research from the start. This is because NHS bodies have to comply with the NHS/Department of Health Information Governance Toolkit. This toolkit actually requires NHS bodies to record serious data security incidents even where those incidents did not, in fact, constitute a breach of the body’s obligations under the DPA (i.e. incidents might be recorded which were “near misses” or which did not constitute a failure of the obligation to comply with the seventh, data security, principle).

The results Big Brother Watch got in response to their ambiguous and inaccurately termed FOI request show that some NHS bodies clearly interpreted it expansively, to encompass all data security incidents, while others – those with zero returns in any of the fields, for instance – clearly interpreted it restrictively. In fact, in at least one case an NHS Trust highlighted that its return included “near misses”, but these were still categorised by Big Brother Watch as “breaches”.

And this is not unimportant: data security and data protection are of immense importance in the NHS, which has to handle huge amounts of highly sensitive personal data, often under challenging circumstances. Awful contraventions of the DPA do occur, but so too do individual and unavoidable instances of human error. The best data controllers will record and act on the latter, even though they don’t give rise to liability under the DPA, and they should be applauded for doing so. Naming and shaming NHS bodies on the basis of such flawed research methodology might well achieve Big Brother Watch’s aim of publicising its call for greater sanctions for criminal offences, but I worry that it might lead to some data controllers being wary of recording incidents, for fear that they will be disclosed and misinterpreted in the pursuit of questionable research.