In light of the increasing number of enforcement incidents under the General Data Protection Regulation, organisations active in the Health and Life Sciences sectors in the United Kingdom, the European Union and other European Economic Area countries are well-advised to review their obligations when it comes to the collection and processing of health data.
What is “Health Data”?
It is helpful to first understand what constitutes “health data” under the GDPR.
There is a common misconception that this term refers simply to medical records, but the definition is much broader.
It refers to all data concerning health including, significantly, “inference” data, which is data, which alone or with other data sources, would enable a third party to draw an inference about someone’s health.
This may include information about an individual’s diet, their exercise levels, and attendance at a clinic.
Data relating to a person’s health has always been seen as sensitive, even before the implementation of the GDPR in 2018.
It was protected by the previous Directive and national common laws, plus professional obligations, such as a doctor’s responsibility to keep a patient’s details confidential.
The result of this history is that the GDPR is not the only consideration when collecting and processing health data in Europe.
What Conditions Must be Met for Health Data to be Collected or Processed?
Consent is widely considered to be a key feature in the provision of health services, based on the requirement that patients give informed consent for treatment, or subjects granting consent for their involvement in clinical research.
As a result, many organisations default to thinking that consent is the only basis under which they can handle health data, or that using consent is the “gold-standard” to which they should aspire.
While this is the case under many national, common laws and professional duties, and is still the case in the United States, it is not the case under the GDPR.
Under Article 9 of the GDPR, consent is just one condition under which the collection and processing of health data is permitted. The other options are that:
- The data is being used for the provision of healthcare
- The data is being used in the interests of securing public health
- The data is being used for research purposes
Provision of Healthcare
For this condition to be met, the processing must be “necessary” for one of the following:
- The purposes of preventive or occupational medicine
- The assessment of the working capacity of the employee
- Making a medical diagnosis
- The provision of health or social care, or treatment
- The management of health or social care systems and services
In addition, the data must be processed by or under the responsibility of a professional who is subject to the obligation of professional secrecy or rules established by national competent bodies, or by someone subject to a legal obligation of secrecy or rules established by national competent bodies.
The professional could therefore be a medical doctor or other healthcare professional who is part of the wider healthcare team.
The public health condition covers health data processing required by a legal or regulatory provision.
Providing “high standards of quality and safety of health care and of medicinal products or medical devices” may, for example, cover the processing by a medical device manufacturer of health data for the purposes of ensuring the proper functioning of the device, or for reporting adverse events to the health authorities.
The research condition permits health data to be processed for scientific research purposes if based on an EU or a country’s law, provided that the processing is proportionate to the aim pursued, respects the essence of the right to data protection and provides for suitable and specific measures to safeguard the fundamental rights and the interests of the data subject.
The use of health data for research purposes is the most controversial of the GDPR’s conditions and has been supplemented slightly differently at the country level since GDPR requires a basis in an EU or a country’s law.
This means that practice is not consistent among European countries and their data protection authorities, so organisations that process health data for this purpose need to be particularly mindful of local requirements.
In Germany, for example, research must be in the public interest and that public interest must outweigh the data subject’s opposing data privacy rights.
Interestingly, the national data protection authorities are often reluctant to assume a public interest in a commercially driven research project.
For example, if a medical device company conducts research to improve its products, this might not be considered as fulfilling the research condition, even if it could be argued that improvements to the product are in the interests of the device’s users.
In France, where the research must also be justified by a public interest, the Commission nationale de l’informatique et des libertés (CNIL) and the Health Data Hub in charge of construing the concept of public interest, recognize that private research projects may be conducted in the public interest, as much as public research projects.
The CNIL takes a very broad view on what qualifies as being in the public interest (e.g. improving care, public health or the healthcare system, research and knowledge enhancement, etc.).
In the United Kingdom, the ICO has published draft guidance about research processing and there are also proposals to change data protection legislation to allow for greater flexibility for research.
The ICO’s approach is considered fairly flexible although it is worth noting that the research condition will not be considered satisfied if the data processing is likely to cause individuals substantial damage or substantial distress, or if it is carried out for the purposes of taking measures or decisions about particular individuals, except in the case of approved medical research.
National Laws on Medical Confidentiality
All EEA countries and the United Kingdom have national laws concerning medical confidentiality, which apply to the professionals handling information relating to their patients’ medical histories.
The fundamental principle behind data protection law in the United Kingdom is that information shared by individuals with a professional should not be used or shared except as originally understood by the individual or with their subsequent consent, which can be expressed or implied.
National Health Service legislation overrides confidentiality for certain approved research, but patients may choose to opt-out of this use to protect their privacy.
In Germany, doctors and other healthcare professionals are generally prohibited from sharing patient data with third parties without the patient’s consent.
Notably, a breach of professional secrecy obligations is not just sanctionable by the German medical association, the Bundesärztekammer, it is also a criminal offence.
Doctors and healthcare professionals are, however, permitted to share health data with “contributing persons”, such as service providers who act in the interest of and on behalf of medical professionals, such as IT service providers, or providers of practice management software.
France takes an extremely strict line on healthcare data.
Medical confidentiality is an irrevocable duty imposed on healthcare professionals and even patients cannot consent to release them from their obligations if the law does not expressly permit it.
Similarly, patients cannot grant consent for their healthcare data to be transferred to a third party.
The main exception is in the context of secondary data processing for research, in which case the relevant organisation must first obtain an authorisation from the CNIL.
Although the Article 9 alternatives to consent outlined above are available, it is often the case that the conditions attached to them cannot be met and obtaining consent is the best option.
In these situations, organisations need to understand what constitutes valid consent.
First and foremost, data subjects cannot be forced to give consent.
They must be given free and ongoing choice in when and how their data is collected and processed, and they must actively choose to grant consent; they must “opt-in” rather than “opt-out”.
Any request for consent must be prominent, independent from other terms and conditions, concise, and easy to understand.
It must also be “granular”, which means it needs to be very specific about how the data will be used and for what purposes.
It must also cover the name of the controller, i.e., the relevant legal entity that will process the data, and any other relevant data controllers.
When the processing has multiple purposes, consent must be given for each of them, which in practice means multiple opt-in check boxes and, because the data subject can remove consent for any one or more of these purposes at any time, there must be systems in place to manage every aspect of the consent.
The control given to data subjects over how and when their data is used presents the biggest challenge inherent in relying on consent as a condition for health data processing.
Under the GDPR, the data subject must be able to withdraw consent at any time, which forces the organisation to stop processing their personal data immediately.
In the case of clinical trials, for example, if there is no other lawful basis and Article 9 condition for justifying the retention of the data for further processing, it must be deleted by the controller, which could have an impact on the reporting of serious adverse effects.
Consent Form Best Practice
The key to obtaining effective and legal consent is a well-drafted consent form.
It should be written in clear language that does not overwhelm nor confuse the data subject, whilst also providing all the information necessary for the consent to be valid.
At least, the following points should be considered:
- It is easier to obtain, easier to document, and easier to manage consent if it is obtained electronically.
- Because data subjects must actively opt-in, do not use any pre-ticked boxes.
- Even if you have limited space on the form, give some context. The data subject needs to know what you plan to do with their data. If necessary, use a second page, another text box, or a weblink, and make clear reference to it.
- Name the data controller. If health data will be shared with another legal entity, state who that is and the specific purposes for which it will be used.
- Assess whether or not additional, national medical confidentiality requirements apply. An additional check box or consent process may be needed to cover these.
Does Anonymising Data Avoid the Need for Consent?
Data that has been anonymised is not subject to the GDPR because it is not “personal data”.
However, the line between anonymised and pseudonymised data, which is still personal data and therefore regulated under the GDPR, is grey.
To a certain extent, the question hinges on whether an individual could be “singled out” even if their name, address, and patient code have been removed from a data set.
For example, singling out might be possible in small populations of patients who suffer from rare diseases or when a patient who has undergone a certain type of treatment in a specialised hospital.
In addition, identifiability exists on a spectrum, so it depends on how the data is processed or who holds it.
For example, if a clinical trial company holds personal information about participants and uses an encryption method to anonymise that data, the data will not be anonymous within that company as it holds the decryption key.
The same data may, however, be anonymous in the hands of a controller to whom that data is transferred, because the controller does not have the key.
For the CNIL, which takes a strict view of the EU case law, if the data is not anonymous in the hands of one organisation, the data will not be considered as anonymised even if held under an anonymous form by another organisation.
Similarly, identifiability is subject to technological developments.
If decryption techniques advance to the point where current safeguards are no longer effective, the data may no longer be anonymous.
Regulators have recognised that it may not be possible to achieve absolute anonymity forever; they state that what is important is that anonymisation is effective.
It is worth noting, however, that “effective” has multiple definitions.
For the UK Information Commissioner’s Office, it means reducing identification risk down to a sufficiently remote level; for the CNIL, it means there must be zero chance of identification.
Regardless of whether an organisation has applied one of the GDPR’s conditions to its data collection, or has opted for obtaining consent, unless health data is entirely anonymised, it should always be protected with the application of pseudonymisation techniques, encrypted servers, and additional access controls and safeguards.
Re-Using Health Data
The re-use of health data can be challenging for a number of reasons.
The first challenge relates to how the data was initially obtained and under which condition.
If, for example, the data was originally collected for research purposes, the secondary use is also for legitimate research purposes, and no national laws are triggered by the secondary use, it may be appropriate to re-use the data.
If, however, consent was obtained for the primary purpose, the organisation will need to check whether or not the scope and content of the initial consent covers the secondary purpose.
This is an excellent example of the impact of granular consent, and why consent should only be relied upon for the use of health data where absolutely necessary.
It also shows the value of foresight when drafting a consent form.
The second challenge relates to data security.
The GDPR requires additional safeguards when health data is used for secondary purposes, such as anonymisation or pseudonymisation to the extent compatible with the relevant research project.
The UK Government and ICO have committed to changing the UK legal provisions on research in order to simplify data protection compliance in the context of research.
The ICO has issued draft guidance on using the Article 9 GDPR research condition in the United Kingdom.
In the guidance, the ICO emphasises the purpose limitation exemption for consent in Article 5, which states that existing personal data can be re-used for research related purposes, as long as appropriate safeguards are in place, because this is considered to be “compatible” data processing.
The situation is different if consent was the original lawful basis for processing the data, in which case, new consent is needed for each new use of the data.
The ICO also set out helpful guidance and criteria for what would be considered research in the public interest.
For example, research activities that are peer reviewed, published, subject to ethics guidance or committee approval, compliant with rules on research, and are published are, according to the guidance, likely to be considered legitimate research.
In relation to re-using data obtained from another organisation, the ICO states that the recipient organisation is essentially collecting new data, not re-using or repurposing it.
The recipient organisation cannot therefore rely on the original organisation’s purpose and must instead identify its own lawful basis for processing.
Data subjects should be informed that the data has been passed on and provided with the recipient organisation’s privacy policies, unless doing so is impossible or involves disproportionate effort.
After the ICO’s guidance was published, a draft bill was introduced in July 2022 but the bill has not yet progressed through parliament.
Across all sectors, GDPR enforcement has been increasing since 2019, in terms of both the number of fines and their size.
Looking at general trends, it appears that most fines have been imposed as a result of processing data with an insufficient legal basis, either because companies did not apply one, misinterpreted the legal basis, or obtained invalid consent because they were not familiar with the consent requirements.
The second most common reason for fines was a lack of data security, which increases the risk of cyber-attacks or third parties accessing the data.
The third most common reason was companies simply not complying with general data protection principles, e.g., by processing more data than was actually needed for their purposes, or not deleting data when a specific project was concluded.
Specifically in the health sector, there appears to be an increasing number of enforcement actions relating to data security incidents.
It is apparent that data protection authorities across Europe are looking closely at the appropriateness of technical and organisational security measures, such as pseudonymisation and anonymisation.
Although Italy leads the field in imposing fines for GDPR violations, Germany and France are not far behind, suggesting that the data protection authorities in all three countries have limited tolerance for health data being compromised.
It is interesting to note that authorities are not just targeting big pharmaceutical or medical technology companies.
Fines are also being imposed on hospitals, doctors, and other players in the industry.
What The Future Holds: The European Health Data Space
The European Commission has published a draft European Health Data Space (EHDS) Regulation with the aim of creating a common space where researchers, innovators and policy makers can use electronic health data in a trusted and secure way, preserving privacy and the rights of data subjects to control their data.
The draft EHDS Regulation addresses the challenge of establishing such a space by promoting the digital transformation of the use and access to health data in healthcare (primary use), and regulating electronic health records (EHR); and accelerating the re-use (i.e., secondary use) of individuals’ health data.
The draft Regulation also builds on the requirements that have been imposed on software through the Medical Devices Regulation and the proposed Artificial Intelligence (AI) Act.
In order to avoid any regulatory gap, where manufacturers of medical devices (which need to be certified under the Medical Devices Regulation) and high-risk AI systems (which should be subject to risk management, security requirements, and certification under the draft AI Act) will need to comply with interoperability requirements, to the extent they claim interoperability with EHR systems.
United States Perspective
Unlike the United Kingdom and the EEA, the United States has multiple federal and state privacy laws specifically focused on individual sectors, including the health sector.
Some of the laws regulate types of organisations within specific sectors, such as mental health facilities or subcategories of health data, such as genetic test or HIV test results.
As a result, privacy protections often depend not only on whether the data is health-related, but also on who holds the data.
The primary law regulating personal health data in the US on the federal level is the Health Insurance Portability and Accountability Act of 1996 (HIPAA).
HIPAA regulates certain health care providers; governmental and private health plans; and healthcare clearinghouses, which are intermediaries between providers and plans.
HIPAA defines “protected health information” broadly (with certain exceptions) as information that meets the following two criteria.
- The information must relate either to the past, present, or future physical or mental health or condition of an individual; or the provision of health care to an individual; or the past, present, or future payment for the provision of health care to an individual.
- The information must also identify an individual, or there is a reasonable basis to believe the information can be used to identify the individual.
In the United States, consent is generally seen as the gold standard for processing data.
HIPAA does not require a legal basis for the use of de-identified data and it requires re-identification risk to be reduced to a “very small,” but not zero, level of risk.
HIPAA’s willingness to allow data sets that involve only a very small residual risk of re-identification to be considered de-identified and outside of HIPAA privacy requirements reflects the HIPAA Privacy Rule’s policy goal of reasonably balancing the competing goals of data utility and privacy.
Two de-identification methods are permitted under HIPAA: “safe harbor;” and expert determination.
Under the “safe harbor” method, a HIPAA-regulated entity must remove 18 identifiers, which include direct identifiers, e.g., name and email address; and indirect identifiers, e.g., birth dates and other dates more specific than year.
This method is comparable to the singling-out approach described above under GDPR.
Health information is considered de-identified under the expert determination method if someone with expertise in generally accepted methods for rendering information not individually identifiable determines that the risk the information could be used to identify an individual is very small, and documents the methods and results used to make this determination.
Urgent work needed to tackle ‘substantial’ digital health inequality
eSight: “The technology has the potential to change someone’s life”
Microsoft invests £2.5 billion in UK AI
TMS shows promise in tackling depression ‘epidemic’
AI depression app set for NHS clinical trial
UK Biobank releases world’s largest single set of sequencing data
Listen: Longevity, Eastern wisdom and Western science
Fundamental principles of healthcare digital twins
Tackling the diagnostic testing sustainability problem
Gym-going men ‘unaware’ of protein risk to fertility
- News3 weeks ago
Why a leading healthcare CEO sees recombinant DNA as a metaphor for developing breakthrough technologies
- AI5 days ago
AI model predicts breast cancer risk without racial bias
- Medtech4 weeks ago
Surtex Instruments to unveil game-changing Infinex microsurgery instruments at MEDICA
- AI2 weeks ago
Humans make better cancer treatment decisions than AI, study finds