Calls to Action on Health Data Ecosystems - RECOMMENDATIONS FROM MULTI-STAKEHOLDER ROUND TABLES - ECHAlliance
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Calls to Action
on Health Data
Ecosystems
RECOMMENDATIONS FROM
MULTI-STAKEHOLDER ROUND TABLES
Raise the digital,
Digital Health literacy & skills of all
Society & the stakeholders
European Institute
for Innovation
Generate and value
Through Health trustworthy Real
Data Calls to Action World Evidence
on Health Data
Ecosystems Accelerate
This contribution summarises the outcomes interoperability across
of two recent multi-stakeholder consultations Europe and globally
to examine the acceptance criteria for societal
trust in the use of health data and a recipe for
trustworthy digital health: standards, architec-
ture and value. Demonstrate
benefits to society
The Round Tables were developed and con- from data access,
vened by DHS and i~HD neutrally and inde- use and reuse
pendently from the event sponsors, Johnson &
Johnson and Microsoft. Each meeting was at-
tended by around 27 online participants from
EU institutions, national governments, industry,
academia, hospital management,, healthcare
Adopt a risk
professionals, regulators and patient represen- stratification
tatives. DG Sante and Connect officials contrib- approach
uted to both events.
The recommendations and calls to action aris-
ing from these events were presented to large
online audience at the Digital Health Society
Build a trustworthy
Summit in November 2020 and discussed by framework for data
a multi-stakeholder panel. The recommenda- access and use
tions, the calls to action and round table sum-
maries are documented in the accompanying
uploaded report, also available at
www. Adopt a
transformational
The recommendations cover the following approach to
key themes. health data
01
Raise the
digital,
literacy &
skills of all
stakeholders
Member States should set Researchers, regulators, public
target standards for population health and political decision
and professional digital, health makers also need to be health
and data literacy and openly data science literate.
share these targets at a European
level.
Healthcare funders (ministries,
regions, insurers) should publicly
Literacy should cover, for the declare an annual budget they will
public: invest in patient/citizen literacy
• becoming fluent data users resources and initiatives, and
for their own health how they will cover age ranges,
• appreciating the importance ethnicities and other population
of the data they create subgroups and leave no one
• understanding their rights and behind.
protections over data held by
and used by others
• understanding the benefits Education providers targeting
their data can offer to society. public and health professional
education should be required
to share digital health curricula
Literacy should cover, for and learning objectives (not
existing and future health course delivery materials). Equally
professionals and managers: these points should be applied
• how to use digital health tools/ to curricula for health and data
data science for patients and literacy for the education of
citizens children.
• how to educate and support
patient/citizen users of health
data and digital health tools Industry should contribute to this
• how to respond to and mission by sharing educational
escalate issues, readings of resources and the selective
concern sponsorship of training places on
• the importance of RWE and its literacy programmes.
quality
• how to understand data
science and its contribution to
healthcare practice.
02
Generate
and value
trustworthy
Real World
Evidence
National and Regional Health • education to raise the skills of those who
data infrastructure providers need to generate real-world evidence, so
and coordinators, the research they ask the right questions and generate
community, public health comparable answers
agencies and European data • the kinds of research questions can be
infrastructure programmes answered by distributed analytics, and
should increase and co- which ones need to work on a dedicated
ordinate investments in: patient level data extract
• improving data quality, starting with
facilitating a more motivating culture within
healthcare professionals and better EHR
system user interfaces
• research into errors and statistical
corrections for low quality data, and the
generation of synthetic data e.g. for the
training and validation of AI
• audit processes and traceability of the
sources of data must be embedded into
policies and architectures to ensure
transparency.
03 Accelerate inter- operability across Europe and globally
Member States should Member States and the EC must
embrace an alignment of support patients and citizens to
standards adoption with become strong advocates of joined
other countries, such as on up (interoperable) health data
the EEHRxF, and reflect those balancing illness and wellbeing
as strong interoperability (prevention) needs.
demands within national and
regional procurement policy and
specifications. The EC should more strongly
encourage health data generated
through its funded projects to be
Future standards more widely reusable via the EHDS.
development strategies
should involve representative
data creators and Interoperability between
users, especially health consumer devices which
professionals and patients. generate health data and
EHRs will become increasingly
important as this type of data
Healthcare providers grows in volume and relevance
should demand, from their and must therefore be ensured
EHR suppliers, explicit and through regulation or soft law.
independently verified
interoperability against
prescribed standards through
procurement specifications and
renewal contracts.
The extent of the
interoperability a healthcare
organisation and its supplier
can deliver should be
measured and made public.
04 Demonstrate benefits to society from data access, use and reuse
Data Permit Authorities and Industry should support and data sharing intermediaries then adopt consensus practices should: on how best to communicate • publish lists of data uses they the benefits to society from will normally support, and their use of health data. those they would not • require the intended benefit of data use to be stated with each data request • define the terms and conditions they will require from data users • publish annually the benefits they have enabled, and lessons learned from reusing health data • consult with the public to define societal benefits and value • involve patients and citizens at decision making (board) levels • promote and oversee good models of data altruism.
05
Adopt a risk
stratification
approachThe GDPR places too strong an EU and national research
emphasis on the identifiability funders should invest in further
of individuals from data research on risk stratification
through explicit attributes methods for health data sets so
and does not give adequate that proportionate protections
recognition to unique data such as appropriate codes of
patterns that may enable conduct and suitable information
data subject identification. security measures and can be
applied consistently according to
purpose and risk and not, as at
At an EU level a specific health present, in a piecemeal way.
scientific and research basis
for reuse is needed.
Data Protection Authorities
and the European Data
Pseudonymisation should Protection Board should indicate
not always be considered as willingness to develop and adopt
personal data without taking risk stratification guidance on the
into account the safeguards use of data protection safeguards.
including the protection of
linkage keys. Member States and the EC
should balance risks with the
opportunity costs of not sharing
health data.06 Build a trustworthy framework for data access and use
Data Permit Authorities All public and private should: stakeholder should support • promote the development the adoption of standards and and adoption of multi- Compacts for how data access stakeholder Compacts requests are formulated and regarding responsible transparently reported on. data use, transparency, accountability, communication, by including the public (patient and civil society organisations) health funders, providers and health data organisations (public bodies and industry) • hold open public consultation when developing governance frameworks and decision- making rules for health data uses and reuses • include members of the public in the constitution of the European, national or regional decision making bodies themselves • publish inventories of data use requests received, accepted, declined and of any investigations into misconduct • conduct public awareness campaigns to explain to the public the research uses and benefits of using health data.
07
Adopt a trans-
formational
approach to
health dataAll stakeholders should support Synthetic data sandboxes
and promote treating repositories should be developed to enable
of pooled anonymised health research into novel security
data as a societal good. approaches and the training of
AI algorithms.
Investments should promote
the uptake of federated A transformation towards
data models to facilitate cross-organisational and
interoperability, connectivity independently run health data
and FAIR data access while repositories will require radical
upholding GDPR compliance. change in ICT products and
procurement, for which policy
enablers must now be enacted.
Europe should now consolidate
efforts on one or a small number
of common data models so that Regional and national early
data harmonisation methods, adopters should be encouraged
tools and skills can be scaled up to collaborate across borders to
to become a readily available and develop best practices, lessons
affordable resource. learned and accelerate the reuse
of data and the development of
benefits from it, sharing with other
Stakeholders should focus Member States and stimulating
eHealth governance models, European competitiveness.
trust mechanisms and research
infrastructures to contribute data
to large-scale independent health
data repositories that provide
real-time continuity of data
access for individuals, healthcare
delivery and for population
level analysis, with appropriate
governance.DHS Summit
Feedback
The above calls to action were presented and
discussed by an expert panel which com-
prised Nicola Bedlington, Chair of Data Saves
Lives, Jesper Kjaer, Danish Medicines Agen-
cy and Nigel Hughes Project Lead EHDEN &
Janssen and Ioana-Marie Gligor, Head of Unit
DG Sante. They agreed the two most import-
ant calls were upskilling digital, data and
health literacy, and generating and valu-
ing trustworthy Real World Evidence.
We conducted a poll of Summit attendees
and the results were:
50% 82%
thought the most thought if health
important call data is to be a
was adopting a societal good it
transformational should be defined
approach to by a group formed
health data. of multiple
stakeholders.
72% 73%
thought a list of thought that, to
data uses that would develop trust in
normally be data access and use,
supported and those they would prefer to
that would not be see a combination
supported should be of written laws/
developed by a group regulations and
formed of multiple multi-stakeholder
stakeholders. codes of conduct.REPORT OF ROUND TABLE VIRTUAL MEETING ON Acceptance criteria for societal trust in the use of health data
Round Table
Summary
This report summarises the topics, discussions ever, there is less public understanding
and conclusions of a multi-stakeholder Round and therefore trust for uses of data that
Table held on Thursday 3rd September 2020 are not directly applicable to the indi-
on acceptance criteria for societal trust in the vidual and performed by organisations
use of health data. Its aim was to propose who seem less familiar within the health
criteria for building and retaining socie- ecosystem. Participants considered the chal-
tal trust in the uses and reuses of health lenge of societal acceptance criteria from three
data, across a spectrum across direct care, perspectives: the who, what and why of data
public health, health system improvement and use and reuse; technical and organisational
research. 27 participants, comprising patient safeguards; transparency and trust about use
organisations, healthcare providers, payers, and value. Through breakout group and plena-
ministries, data protection authorities, industry ry discussions, the following themes emerged.
and industry associations and representatives
from the European Commission participated in There is a big difference conceptually between
a highly interactive half-day meeting designed data use to benefit the individual and larg-
and run by the Digital Health Society and the er scale data reuse that has the potential to
European Institute for Innovation through benefit many (but might not include individual
Health Data, sponsored by Microsoft and data subjects). The public, and individual
Johnson and Johnson. The Round Table sought data subjects generally support data re-
to consolidate what society and decision mak- use if it clearly explained to them what
ers would regard as acceptable conditions the beneficial objective is. It was recognised
and terms for access to large scale data that there is no universally accepted definition
resources. It has been timed and offered as of beneficial use, but that illustrative lists of
input to the scope, design and governance purposes that would normally be support-
framework being developed for the European ed by decision making bodies, and pur-
Health Data Space. poses that would not be supported, are
helpful for guiding the public and guiding
Right across the learning and innovation eco- decision makers. It would additionally be
system, there is a growing need for large scale reassuring and strengthen public support if
access to health data. A momentum for Eu- list of approved uses, and denied uses, would
ropean cohesion on data access, harmonised periodically be published. This is in effect a
criteria and governance, has been accelerated combination of transparency of intention
by the European Health Data Space. How- and transparency of action.It has been found by a number of patient and Decision-making bodies, and the governance
public perception studies that ensuring that frameworks that they operate under, are more
the use purpose will deliver a benefit to health likely to operate at Member State and/or re-
systems and that this benefit will be afford- gional level rather that EU level, for legal, polit-
able, without excessive profit taken, is the ical and practical reasons, but the governance
most important criterion for support. Whether frameworks they utilise should be as consistent
the bodies involved are commercial or public, as possible, across Europe. The public must
whether they are classically associated with be involved in developing their gover-
healthcare or not, are less important factors. nance frameworks and decision-making
COVID -19 has demonstrated the level of public rules and should be included in the constitu-
engagement and support that is possible if the tion of the bodies themselves.
purpose for data collection and use is clear and
in society’s interest. Public fears about misuse, including fears
that information will be used in some way
The importance of transparency to the to disadvantage or discriminate against
public as well as to individuals whose data individuals or minority sub-populations,
might be reused was a dominant theme must also be addressed by such bodies.
throughout the event, across all of the three These fears are very powerful and if they are
breakout groups. This was perhaps considered not addressed, they risk dominating over the
to be the most important success factor. It was perceived benefits of using health data. And
emphasised that this transparency must be important mitigation for this fear is, again,
inclusive, including vulnerable groups of peo- transparency. The public have to know how
ple whose data are equally important and who their data is being used, and how it is not
should be inclusive beneficiaries of the out- used. Even when it is not feasible to give in-
comes from using data (a potential adaptation dividual level control over all possible uses of
of the concept of reasonable accommodation data, the public then need to have confi-
was discussed). Inclusivity may have economic dence in the organisations that are mak-
challenges in a single market, but on the Euro- ing decisions on their behalf. In cases where
pean scale and through the use of European data have been anonymised it is not easy to
standards, inclusivity can be made economical- provide personalised feedback, but collective
ly viable. Greater public and health work- published feedback about how data have been
force education, including data literacy, used to populations may prove sufficient.
digital literacy and health literacy, are
therefore essential success factors as well.Many of the reuses of data, especially for Synthetic data, in which noise (perturbation) is
research, public health and health service added to the data in order to prevent individu-
improvement do not need identifiable als being recognised even from very rich data
data, but they do often need fine-grained, patterns, is a method that is gaining recogni-
close to real-time, data including longitudi- tion as a method for some kind of population
nal histories and increasingly including special- level research.
ised data types such as genomics. The biggest
concern for citizens is whether they could be It is usually fruitful to think about the inter-
identified from a dataset that is being shared play between what we construct by means of
or accessed. The GDPR strongly distinguish- technology and our social constructions (laws,
es pseudonymised from anonymised data, codes of conduct, organisations, behaviour).
but it was argued that fine-grained data Law should not be made without thinking
can never be truly anonymous. about how technical constructs can help en-
able compliance and enforcement. Technology
initiatives should not be developed without
considering how and by whom these initiatives
will be governed. A third, psycho-social (people
oriented), dimension is also important.
This historic distinction is perhaps
There was considerable discussion about a
no longer viable, and a risk code of conduct. Although a formal and pos-
sibly legally enforceable code might be devel-
stratification approach which takes oped at a European level, there was support
into account the way in which the for organisations, especially health and
health-related companies, to come togeth-
data are being processed and er and to develop voluntary codes of prac-
protected and the benefits of tice that they agree to adopt: known as a
compact. The public would be most assured
use through information security if this is a single code developed through
measures is more appropriate. multi stakeholder engagement including
patients and the public, and was adopted by
all health data user organisations (commercial
and public, and including patient and civil soci-
ety organisations themselves when they collect
and use data).Introduction
This report summarises the topics, discussions The Round Table was an invitation-only,
and conclusions of a multi-stakeholder Round multi-stakeholder and highly interactive half-
Table held on Thursday 3rd September 2020 day online event with 27 participants, dividing
on Acceptance criteria for societal trust in for some of the time into three virtual break-
the use of health data. out rooms for deep dive topics. The agenda is
given in Appendix 1. The participants included
Its aim was to propose criteria for building and patient organisations, healthcare providers,
retaining societal trust in the uses and reuses payers, ministries, data protection authorities,
of health data, across a spectrum from direct industry and industry associations and rep-
care, public health and health system improve- resentatives from the European Commission
ment to research. The Round Table sought who are architecting the European Health Data
to identify what society and decision makers Space. The list of meeting participants is given
would regard as acceptable conditions and in Appendix 2.
terms for access to large scale data resources.
This report is therefore intended to help frame The event was jointly run by the Digital Health
future European initiatives to develop better Society (represented by Bleddyn Rees) and
formalised models for data provision, use and the European Institute for Innovation through
governance, to better position new actors Health Data (represented by Dipak Kalra). It
(e.g. industry) in roles such as healthcare de- built on the Digital Health Society’s Summit in
livery partner, care pathway redesign partner, Helsinki with the Finnish Presidency last De-
analytics partner and knowledge partner. In cember when both organisations collaborated
particular, this Round Table and report have on the data and digital content.
been timed and offered as inputs to the scope,
design and governance framework being de- It was sponsored by Microsoft and Johnson and
veloped for the European Health Data Space. It Johnson, who contributed financially for pre-
may guide the development of any necessary paring and running the event but did not con-
enabling legislation and policy instruments, trol the structure, hosting, content or reporting
industry promoted standards or codes and of the event.
innovations in information security safeguards.
No individual stakeholder is able to solve the
challenges and now more than ever we need a
deep collaboration which strikes fair balances
for all to enable the common good.OP ENI NG PLENARY SESSION :
Scene setting
Bleddyn Rees and Dipak Kalra welcomed partici- taking a deeper dive on societal acceptance
pants. factors for data reuse that might be taken on
board when developing the EHDS governance
Bleddyn set the scene for this Round Table, framework. A future event is planned by Digital
which has built on prior related events over Health Europe on the perspectives of patient
the past several months, starting with a Digi- organisation representatives on this topic.
tal Health Society Summit in December 2019, These events have some organisers and partic-
which highlighted many of the issues and chal- ipants in common, and are sharing outputs so
lenges that impact on how the public and pa- that their progression is complementary and
tients understand and indicate preferences for, additive.
or control over, the uses made of health data.
In the spring of this year DG Santé has run a Dipak reminded the audience that there is an
series of consultation workshops with strong explosion of the opportunity space to learn
DPA participation, moderated by Petra Wilson. more from health data, as more and more
In May, the DigitalHealthEurope project ran a kinds of data are being captured about and by
virtual focus group for industry about compa- patients and citizens, and are potentially com-
ny aspirations and potential contributions to binable if this is permitted to build up rich pic-
the European Health Data Space (EHDS). The tures of healthcare, health outcomes, wellness
industry participants highlighted the special op- and wider influences on health etc. We must
portunity for architects of the EHDS to develop take a future looking vision on the availability of
a coherent governance framework that could data. We should especially note the most rapid
be adopted by other European data initiatives, growth area will be citizen generated data, and
thereby helping to harmonise approaches our approaches to governing data use must
adopted across Europe. This Round Table was actively win citizens’ trust in sharing their data.
designed to contribute to that aspiration, by
Population registries,Clinical
Geomic data trials databasees Care apthways, decision
support, trends and alerts
The
Transport, Ditigal
Mobile devices
environment ect. Citizen
Bio-sensors Social networks Clinical applications
“Personal sensor data is expected to grow to 90% “By 2021 there will be almost as many personal
of all stored information within the next decade” assistant bots on the planet as people”
“90% of the data in the world today has “> 1billion have access to mobile
been created in the last 2 years” broadband internet”Right across the learning and innovation eco- This has included important for the collection
system, there is a growing need for large scale of personal data by a range of organisations
access to health data. Many of the innovations (such as restaurants) that would not normally
we are developing or foreseeing need to ben- have this and allowing location data to be used
efit from vast volumes of health data. This may by agencies that do not normally have this
be from conventional healthcare sources (e.g. either. However, it would be wrong to assume
detailed EHRs), patient and citizen generated, that this societal goodwill is going to be perma-
medical devices and non-health sources such nent or can now be relied upon for many other
as pollution. The analysis needs are for this to desired uses of health data.
be fine grained, individual level data (normally
anonymised) so that precise and novel analy- A momentum for European cohesion on data
ses can be undertaken. Pre-compiled aggre- access has been accelerated by the European
gated data sets, or data warehouses refreshing Health Data Space (EHDS). The details of how
their data every few months, are no longer this space will be designed, what data sources
adequate. There is also an increasing demand it will contain and which ones it will network to,
for the data to be close to real time (so that and how it will be governed including its uses
real time feedback systems, for example driv- as well as the terms of use are still being de-
en by AI, can be developed), and for this to be fined. However, the following concept diagram,
longitudinal, reflecting health, wellness, disease developed by DigitalHealthEurope, seems to
trajectories and outcomes. be a plausible concept model. Europe already
has several existing data network infrastruc-
This growing data need has stimulated national tures that might be interconnected through the
and EU level (and international) investments in space, which could then offer a single portal
large scale data resources and networks that for access to permitted data extract from these
offer these opportunities. The different exist- networks. This includes the eHealth Digital Ser-
ing and emerging infrastructures comprise a vice Infrastructure (which presently exchanges
mixture of eHealth (digital health) services and patient summaries and electronic prescrip-
research infrastructures. They are implement- tions between some member states, but has
ed via a mixture of centralised and federated a roadmap to extend the range of electronic
architectures. These different infrastructures health record data sets to be communicated),
are often set up quite differently. They may the DARWIN network being developed by the
process different kinds of data, from different EMA and the national medicines agencies for
sources, serve different purposes and user the exchange of medicines information in-
types and have different governance frame- cluding for pharmacovigilance, the European
works. This makes it difficult to win public trust Reference Networks that especially connect
at a European level. centres across Europe caring for patients or
conducting research in rare diseases, and the
The COVID-19 pandemic has stimulated public life sciences research infrastructures. There are
awareness and support for generating rapid additionally, stakeholder groups which have
insights from data, and the attitude to data col- data accumulations that could be contributed
laboration amongst data using organisations. into the space, as physical data or as networkconnections, subject to suitable agreement and to scale up. Personalised medicine and AI, for
terms. This includes industry, such as Pharma, example, will become increasingly important.
MedTech, Telecos and large ICT companies,
and public health agencies that are starting to However, when shifting from left to right on
accumulate data in response to emergencies the diagram there are several challenges that
such as COVID-19. Patients and citizens are an are faced by the public when it comes to ac-
important potential data contributor, as well as cepting and supporting these data uses. This
data user, as mentioned earlier. These different includes the more limited widescale public
data sources span healthcare, research, tech understanding of the right-hand side uses,
and regulation, and the uses of the space may how they are undertaken, with what kinds of
cover any or all of these areas. Data quality, data. Additionally, the kinds of organisation that
interoperability, collaboration and governance become involved in undertaking those uses
have to be right. The public have to be on are less familiar as health stakeholders to the
board for this to succeed. public. Knowledge derived from populations of
patients may take a long time to feed back to
The diagram below illustrates, in the upper visible public benefit, and the benefit may be
portion of the diagram, many examples of the perceived by different people from the ones
potential agreed uses of health data that occur whose data was needed to generate a knowl-
at an individual level (close to the patient or citi- edge. This all creates a progressive disconnect
zen), at the level of regional and national health between patients and the public and the uses
systems (for public health or health service and uses of data, with reducing direct engage-
improvement purposes) and at an even larger ment and choice, making it harder to win public
scale for the conduct of research. All of these trust and provide public assurance. We need
example uses occur today, but some of them to develop a new consensus on acceptance
are relatively local and are only just beginning criteria for the uses of data.
EHDS Inno
vation Research Regulation Continuity of
car
e
concept
diagram
Cross EU RD
Puplic sectional Registries
health crisis EDS Platform
solution
creators
Distributed
Industry Common European research
research and Health Data Space infrastructures
(e.g. ELIXIR)
innovation - Data Governance
- Data Interoperability and Quality
- Infrastructural Building Blocks
National
Non-Federated Patients & & regional
Netwoks citizens networks (e.g.
Trygge)
eHDSI Regulators
Federated
NetwoksIndividual level health data Population level health data Big health data
EHR systems, apps, sensors, genomics, EHR systems, regional & national national & international research infra-
Clinical Decision Support, AI guidance + eHealth infrastructures structures, federated query research
health impacting data e.g. pollution platforms + cross-sectoral infrastruc-
tures & services
Used for: Reused for: Reused for:
• Health status monitoring • Healthcare provider performance and • Epidemiology
• Continuity of care (including the patient planning • Digital innovation: devices, sensors,
and caregivers) • Quality and safety, care pathway opti- apps
• Care pathway tracking, clinical workflow misation • AI development
management • Medical device and algorith refinement • Personalised medicine and bio-marker
• Real-time feedback and guidance to • Pharmacovigilance research
patients and clinicians • Public health surveillance • Diagnostics development
• Personalised medicine • Public health strategy • Drug development
• Disease interception, prevention and • Health services and resources planning • Disease understanding and
wellness stratification
• Healtcare provider reimbursement
Decreasing public understanding of why and how data are used
Increasingly unfamiliar data users
Increasing distance from data results from the patient
Increasing time from data use to demostrated value
Perceived lessening choice and greater cybersecurity risk = harder trust
Europe has seen many public attitude and identify some ground rules for building and
patient attitude surveys in recent years, which retaining societal trust in the uses and reuses
have been conducted by many different or- of data. This should be across the spectrum of
ganisations, using different methods and espe- purposes and users, consider different access
cially using differently framed questions (some and governance models, how transparency
of which have not been well worded to yield should be demonstrated and what acceptable
precise answers). The result is possibly a more societal benefit should look like. It is hoped that
confused picture of public opinion than a help- an eventual set of endorsed acceptance criteria
ful one, and we need to recognise that public would give greater confidence to data provid-
confidence and trust in the uses of data they ers and data users about data availability and
understand less well, such as genomics infor- what access arrangements are permissible,
mation, is lower than for classical clinical data. acceptable and serve to catalyse greater data
However, there is a general consensus that the availability and data use.
public does support the use of health data for
quality improvement at research that is target- Participants were divided into three breakout
ing new or better solutions for diagnosis, treat- groups to discuss specific facets of this chal-
ment and prevention, but not for non-health lenge:
purposes.
1. The who, what and why of data use and
This was the starting point for the Round Table. reuse
Its objective, through three breakout groups 2. Technical and organisational safeguards
and subsequent plenary discussion, was to 3. Transparency and trust about use and valueS U MM A RY O F DISCUSSIO N S I N B RE AK O UT 1:
The who, what and why of
data use and reuse
Moderators: Dipak Kalra and Zoi Kolitsi » It was stressed that, as experienced
through COVID-19, people understand
1. Individual level health that putting data into a pool contributes to
the greater good. They do not need to get
data v. Big health data personal benefit in order to participate (a
crowd-sourcing spirit). They are nervous
General observations and much less trusting if the benefits can-
not be clearly explained.
• There is a big difference conceptually be-
tween data use and big data reuse (“is my • The paradigm of rare diseases (RD), which
data being used for my care or in a research was agreed to be a situation where patients
setting?”). Dividing the data use landscape do not see a hard divide between care and
in more ways, as in the figure shown earli- research. They want their data used to
er, might not be helpful when it comes to improve diagnosis and treatment. This gave
acceptable data use. rise to the following questions:
• It is very important for people to be able to » How far are we from extending the ap-
trust the people who use the data. Lines get proaches we adopt for RD patients, in which
blurred when citizens don’t know what will research is an integral part of care, to other
happen to their data → transparency was patients and conditions? Might RD patients
agreed to be key in order to build the nec- help to champion the research agenda for
essary trust. other patient groups?
» Can these RD concepts be easily under-
stood as we present the benefits to society?
A common viewpoint participants » Is it productive still to separate primary
had encountered is that if patients and secondary uses as we do today (care
vs. research) or can we make things easier
are told clearly how a particular
by looking at the bigger picture? Would it
data use will lead to a direct benefit be realistic to progressively blur the bound-
to them (e.g. for a new type of aries between primary and secondary uses
for commoner conditions?
treatment for them) or a benefit to
others (e.g. treatments for other » It was agreed that the area of rare dis-
eases shines a positive light on the use of
patients, but might not include
data for research, given that the medical
the patient him or herself), they need is so high. Patients are much more
willingly give their consent. willing to make their data available, but also
are contributors to finding the solutions.» RD should not be completely separated The fear that data is being held to a
from commoner disease research. These patient’s disadvantage
should be linked: common disease research
will progressively be based on smaller pop-
• Testimonies of patients have shown that
in spite of various explanations offered to
ulations of patients, which might eventually
them, they fear that data they have shared
be applicable to a single patient.
will then be used to their personal disad-
» We all agree there has to be a link be- vantage (whether this be penalising by in-
tween care and research but there are sep- surance companies, for marketing, discrimi-
arate demarcation lines – a single patient is nation, preventing their career progress e.g.
of no interest to the research data user, but at work, cyber-criminality: blackmail after
care needs to have nominative (identifiable) the hacking of private information).
data.
• It was agreed that it is difficult to fight these
• Can we use these different groupings of fears about health data misuse due to a
purpose to better understand the worries lack of faith in policy making and political
that patients have about how their data is decisions, as well as a deep-rooted fear
used? Would it be productive to look at the about big private companies/industry tak-
groupings their acceptance in a different ing advantage of the data but then pushing
way, to understand the worries and the excessive pricing of products onto health-
solutions to those worries? care systems. There may be value in listing
as prohibited those misuses that the public
Concerns expressed have greatest concern about.
Misuse
Monetising data
• Preventing the misuse of data, having trans-
parency about proper uses and the reason- • The topic of monetising data isn’t something
ing behind these uses is important. There that is spoken about much these days but
must be understanding and awareness was rumoured in the past to be lucrative.
about what it means to have access to a However, it was agreed that this issue has
patient’s data and for this to be used for the not gone away. It was agreed that this also
broader good. links to data ownership. Some ICT compa-
nies (e.g. app developers) offer their cus-
“Fear is the strongest emotion” tomers free services in exchange for per-
mission to use their health data. Another
• As seen and concluded from the enquires point was made that many hospitals don’t
run at the European Commission over the understand – but sometimes overestimate
last three years in data sharing, fear is the - the value of their data. (A comparison was
strongest of the emotions relating to data drawn with the data of an individual being
protection, rather than the potential for of little value in that regard).
good.• We need to use a lot more data efficiently » The public has to understand why, how
and effectively and safely, and to bring the and by whom data can be usefully used.
public along. We can only overcome the
fears mentioned above by showing the
• The issue of trust enablers and transpar-
ency was discussed alongside a criticism of
public the benefits from making good uses
GDPR.
of data.
» Although GDPR is a means towards
Assurance principles/core achieving trust, it was discussed that the
principles Regulation is still in infancy when it comes
to helping to ensure that citizens can find
• It was agreed that it is important to illustrate out “what the system knows about me”.
the positive examples of how using the data
of large populations can lead to more effec-
tive treatments, with concrete case studies.
With regard to matters of control,
» There was a recent Belgium consultation
on genomic data undertaken with engage- citizens should be able to decide
ment of students and teachers and a wider (authorise) who may use data and
engagement of citizens which highlighted for what purpose. Citizens cannot
what the fears are. We need comparative
qualitative studies to better understand
easily find this out. Although this
what the issues are, and then to see what is an area that GDPR tackles, it is
the impact of education is on those fears. pragmatically not happening in real
Patients may then become actors, not just
passive data donors.
life. If data is used under a public
interest legal basis the citizen is
» The example of Citizens’ Juries, conduct-
ed by the UK Connected Health Cities pro-
not asked, does not decide, does
gramme, was discussed1. In order to under- not control and is not normally
stand the scenarios for which participants informed about that use.
would agree to the use their data for re-
search, they were offered a three-day edu- A third important trust enabler
cation course, which encouraged people to
is the need for greater public
be engaged in understanding data use and
to voice informed decisions. This helped to knowledge and access to education
highlight what aspects of a proposed data and literacy (health, data and
use influenced decision making the most. digital) – from primary school
However, this brought up the question as
to how feasible it would be to carry out a upwards, about how to treat your
similar exercise on a whole population. own health data, about how to deal
with common health conditions.
1 Tully, MP, Hassan, L, Oswald, M, Ainsworth, J. Commercial use of health data—A public “trial” by citizens’ jury. Learn Health Sys. 2019; 3:e10200.
https://doi.org/10.1002/lrh2.10200» These are three empowering enablers.
• However, how realistic is it for patients and We need education showing the
the public to be able to read and digest
public how their data can be useful
about all of the different possible uses of
health data, and to understand these well without their identity ever being
enough to be able to make informed deci- divulged to researchers. This could
sions? We cannot ignore the role of govern- be through lists of example uses
ments here, to create integrated and co-or-
dinated governance frameworks, where and more detailed case study
citizens can see the principles, the values examples. With the evolution of
that governments respect, and what are the analytical methodologies, we can
monitoring measures being taken.
do virtually everything researchers
• It was agreed that there is an interesting need with data without making
dynamic between citizen responsibility and
government assistance. It was concluded
personally-identifiable data
that a citizen needs to be active and to lo- accessible.
cate/find if there is a governance framework
that they can be part of designing or influ-
encing, part of which is transparency.
• On the other hand, we must get away from B RE AK /////////////
the notion that everything must be put onto
the shoulders of the citizen. There must be 2. Purposes and people
an agency that they can trust to make deci-
sions about data use on a collective behalf, • It was agreed that the ways forward for
and that they can monitor. encouraging people to be fear-free and to
share their data is through transparency
• Some of these ideas were seen to be mostly about having done the right things with
applicable to situations in which personally data. This is paramount. The conversation
identifiable information is being used. We then focused in that regard on the purpos-
also need to make more transparent use of es and players.
solutions that don’t include identifiable data
or patient level data. Distributed networks Purposes
and analytics hold promise for removing
that need. • An increasing number of organisations are
becoming part of the health data ecosys-
tem.
• Similarly to the growing number of food in-
gredients, a list of acceptable purposes may
2 https://www.snds.gouv.fr/SNDS/Finalites-autoriseeshelp to address the need for transparency Organisations/Players
and trust building. It may be illustrative.
• An illustrative list was shown through the
• It was agreed that there is no need for a slides and the discussion showed that pub-
comprehensive list of purposes. It may be lic surveys have come down against having
too generic (or would need to be infinite!). a fixed list of approved organisations.
• The need remains for ethic committees • The conversation led to questions such as
and local government bodies to approve or whether any kind of organisation should not
disapprove a research data use, as well as be granted access to health data and how
educating citizens about the role of such the public perceives their data is being used
committees.
• To address this, the findings of a recent
• A blacklist of the purposes for which data Citizens’ Jury in Scotland were touched on.
will never be used would also be useful. The most important decision-making factor
was the purpose for which the data would
» Attention was brought on the Guidance
be used, but not with whom the data would
note for researchers and evaluators for
be shared. There was support for improve-
social sciences and humanities research
ments to their own health, the heath of
2010 – which sets boundaries through a list
others or the health system, but not use for
of what data may not be used for.
a purely financial purpose.
» It was also mentioned that a list such as
» Specifically, it was mentioned that par-
this exists in France2. This includes prohib-
ticipants asked, “Why is my data not already
ited examples such as marketing about a
used (like in banking, to improve experi-
pharma product and insurance purposes.
ence)”.
In France public interest is also a ground
for allowing data use (though this is not » On the topic of data donations, if the
enacted by written law) – but this should be purpose is to save a life or to serve public
formally defined and ideally then enacted/ good, the common answer in the Jury was
adopted into written law. There is patient definitely yes, we would be very keen to
organisational involvement in that decision share our data.
making.
» Within this context, it was also men-
tioned that perhaps the Declaration of Hel-
sinki could also contribute to determining a
blacklist.
» A list of purposes that would be prohib-
ited is also useful as it is illustrative of the
boundaries a decision-making body would
not cross.S U MM A RY O F DISCUSSIO N S I N B RE AK O UT 2:
Organizational approaches:
Safety and Acceptance
Moderators: Paul Timmers and Nathan Lea • Main risk is of citizens handing over control
of their data if they can be identified again.
1. What standards should » Maybe many citizens would like to do-
be set around anonymisa- nate fully personal data – perhaps altruistic
plus personal benefits (different motives).
tion, given the challenges » Citizens would likely be more ready to
with genomics, fine grained hand over detailed data if they knew how it
location data, rich clinical was protected, where it was going and who
would see it.
profiles, rare diseases…?
• If citizens knew it could be used for advanc-
• Are we getting to the stage where tech is ing medical innovations, they may be happi-
moving so fast that it is difficult to make er to share, but they may be worried about
anything pseudonymous/anonymous? insurance company interests and how data
may be used against their interests.
» Rare disease
» Is this true of synthetic data?
» But each of us is different – what level
of risk will you tolerate? May have a generic
» Avoiding the wicked question: how do risk band we opt into, for health data shar-
we define a consent-based model? ing. Anything out of someone’s band needs
• Need an open and honest dialogue special permission.
with citizens
• They raise different concerns to med-
» Near real time use – real time man-
agement decisions. Recognition of social
ical professionals
determinants of health may be increasingly
• Transparency of how citizens’ data is used important for guiding care.
» If you can give them a basis to consent » Decisions are made on trusted data –
and be aware of data use it builds trust currently about 16% of that which is avail-
able.
» Need to have a risk framework around
different kinds of data – e.g. time limited » Blending of citizen data with health ser-
trust, certain data like family history of rare vice data in a single environment is key.
disease. If shared more widely can cause
issues and have implications for others.
» Truly synthetic data3.
3 The term synthetic data is used variably, at present. In this context the term was used to describe real health data extracts that have been modified to a sufficient extent that
individuals could not be identified from the data. Techniques include blurring, rounding, date shifting, making coded terms less granular… An alternative meaning is when data is
generated entirely by a computer algorithm that has parameters which ensure the fictitious data value ranges resemble real population norms but resemble no specific individuals.2. When and how (even, if) • GDPR created to develop common rules
within the digital single market (and beyond)
pseudonymised data can be
» Rule harmonisations
safeguarded enough to be
» Definitions of trust – are these culturally
considered effectively to be bound to different jurisdictions?
anonymised to a data recipi- » Can we have common rules?
ent/user? • The wild west – trading privacy for conve-
• Risk stratification, where there is no binary nience
yes / no. Unlikely to hard code anything in » Some discomfort with what convenient
particular because all data is so different. use implies for wider data use.
» Could we develop tool sets to test that: » How can you - the citizen - use your data
e.g. could the data be reassembled? as well as for the common good? Can you
• Aggregated data implies the need for less yourself use your data to negotiate insur-
protection. GDPR still applies to pseud- ance? Trade your data for health? You can
onymized data, so that determines how you defend your own interests, since you define
proceed. them.
• This is hard – an issue is that anonymous » COVID-19 Apps – to what extent is that
data has a use but in many cases it is next choice enforced?
to useless. This is about a world that is » Look for game changers in the privacy /
evolving – the big games are now in person- convenience trade-off
alised prevention where anonymous data is
not useful
• Who makes the determination?
» Data Trust – a third arbiter that rep-
resents that patients’ say. The right con-
struct. Political cycles make them less ap-
propriate.
» Can Data Trusts/arbiters keep big plat-
forms under control?
» Trust brokers – independent of all play-
ers but ensuring data subjects’ concerns are
addressed in their determinations. Charities
and patient organisations may play a role
here.
» Data Permit Authorities…?3. Rules regarding authori- » Is individual more secure if they know
how data is secured and across a more har-
sation and access controls, monized framework?
restricting indiscriminate » Avoid shopping around where there are
“internal reuse” of the data variations in protection.
by large (public & private)
organisations? 4. Concluding points
• Endgame is that the citizen controls their • Architecture is the enabler if supports ev-
data, or it is controlled on their behalf with a erything we have covered, and it is feasible.
full audit trail
» Tech is malleable
» There would be exceptions under
consent • Codified real time data for life – clean. Obli-
• Study consent – use only as long as gation to citizens to make this data available
consent and participation permits to support their care and care of others
• Where are these scoped by contract? • Keen to see a more harmonious and har-
» Dynamically withdraw data for wider monized health regime in Europe.
healthcare access
• The nature of a study is different. Right to
be forgotten may not be realistic given the Great potential especially with AI to
seemingly contractual nature of studies revolutionize medicine – find a way
and participant responsibilities as an active
participant.
to strike the balance with concerns
around privacy and sensitive data.
» Find mechanisms for changing minds as
well
• Ethical determinants around non-punitive • This is about better tech and organisational
opting out and cannot force people to give safeguards. What is “better” in this case? Is it
their data faster?
» Issues are often covered by GDPR » Make it possible
» Is it more useful to have something at » Tying ourselves in knots in the status
EU level harmonizing the selection of legal quo
bases?
» Opportunity to make things smoother
» Harmonized interpretations? Probably…
There are variations on use of consent, pub- » Fairness relates to how data is acquired
lic task or others for collection. and its accuracy, but it is hard to find what
is fair.S U MM A RY O F DISCUSSIO N S I N B RE AK O UT 3:
Transparency & Trust
about data use and value
Moderators: Bleddyn Rees and Carina Dantas • Health and societal value should be seen
in the perspective of “common good” – But
Context for the session does common good apply only to the entire
Many studies have suggested that the population or specific groups/segments
overriding discriminating characteristics of the population (e.g. rare diseases). How
in the eyes of the public about the use large does a segment have to be or serious/
of health data mainly refer that the use valuable the challenge being tackled?
should deliver clear public benefit, espe- • Information should be provided to citizens /
cially benefits to the health system and patients not only at the beginning and end
to citizens, and should not primarily be of a study but throughout its duration.
used for personal or organisational gain of
the organisation using the data. Although • When a study fails, the results should also
deeper dive research has danced around be shared – lessons learned has real value
this topic, it has so far proved difficult to and failure can avoid future failings. Failure
formally specify ways of defining public should not be stigmatised and can be as
benefit and health benefit. important as success in terms of learning.
• The value of published research, product
and services was discussed; they are all
1. Demonstrating health or valuable and understood as diverse but all
societal value forming different parts of the chain of inno-
vation. Value will vary but as long as passes
• Do we need a definition of health or societal a minimum value of benefit to patients/peo-
benefit? A definition is challenging. ple that is acceptable.
Engaging patients/citizens for Value is also seen differently
co creation and public-private depending on the type of
partnerships (extending IMI to organisations (e.g. research,
other industries e.g. tech, MedTech, industry and patient). Criteria to
automotive and energy etc.) on judge value might be helpful e.g.
projects to meet unmet need has reducing, inequalities, improving
real value. outcomes, improving financial
stability of health systems.2. Earning trust
• Can we, should we, incentivise citizens or Messaging needs to be clear and
patients to share data and build trust? Pa- balanced – clear information,
tients are usually more willing to share data
than a “common” citizen – this distinction
transparency and consents in
needs to be addressed. Also, incentive is a order to develop data trust and
term that may be misunderstood. Need to altruism.
consider what incentives or benefits might
be possible and desirable.
• It is important to know/understand main- • Trust can be guaranteed by which mea-
stream opinions but also the different sures? Legislative, information, (some
segments/minorities. Can the concept of countries have evolved systems e.g. data
“reasonable accommodation” (taken from altruism, data protection), others – is there
disability legislation) be adapted for data the need for concerted actions in multiple
sharing? Use of facial recognition might not perspectives? Insurers use of health data
be 100% compliant with GDPR but its use appears to raise concerns especially in
for persons with disabilities/vulnerable peo- countries with insurance funded systems
ple may be justified as “reasonable accom- which could be controlled via legislation.
modation” and an alternative to consent. A
form of exemption allowing societal benefit 3. Missing further informa-
to override compliance (think public interest
justification). tion citizen perspective on
data sharing
• Information to citizens, that are not only
Literacy (digital & health) is misinformed but sometimes also suspicious
essential to empower people (they – this is a key element to be addressed
need to understand) and it is also • Why health data is secure and its use safe.
key to provide strong safeguards • What data is being shared and the benefits.
that assure people their data will
• Dynamic consent was discussed as a poten-
not be misused. More work needed tial tool. It was highlighted it is necessary to
on these possible safeguards. understand what it implies, so that it is not
a burden to citizens/patients [ very short
discussion as ran out of time].You can also read
