Making NHS Data Work For Everyone - 1544639862

Making NHS data
work for everyone
Eleonora Harwich
Rose Lasko-Skinner
December 2018 #reformNHSdata

Making NHS data
work for everyone
Eleonora Harwich
Rose Lasko-Skinner
December 2018
1

Acknowledgements
Advisory board
Reform is particularly grateful to the expert advisory board who supported the authors on
this project and provided feedback on the drafts of this paper.
Hugh Harvey, Clinical Lead Kheiron Medical. Dr Harvey is a board-certified radiologist
and clinical academic, trained in the NHS and Europe’s leading cancer research institute,
the ICR, where he was twice awarded Science Writer of the Year. He has worked at
Babylon Health, heading up the regulatory affairs team, gaining world-first CE marking for
an AI-supported triage service. Previously a consultant radiologist at Guys & St Thomas’,
he is now a Royal College of Radiologists informatics committee and AI advisory board
member, the Clinical Lead at Kheiron Medical an AI company focussed on deep learning
in breast cancer, and also co-chair of the government Topol Review into AI in healthcare.
Annemarie Naylor MBE is Director of Policy and Strategy at Future Care Capital
– a national charity that uses the insight gathered through evidence-based research to
advance ideas that will help shape future health and social care policy to deliver better
outcomes for society. She has worked with local authorities throughout England as well
as with the range of central government departments and has developed expertise in
policy and the practice of harnessing tangible and intangible assets for social benefit
whilst working for national third-sector organisations over the course of the past decade.
External reviewers
The authors would also like to thank Dr Natalie Banner, Lead, Understanding Patient
Data; Dr Rupert Dunbar-Rees, Founder and Chief Executive Officer, Outcomes Based
Healthcare; Rachel Dunscombe, Chief Executive Officer, NHS Digital Academy; Dr Jon
Fistein, Associate Professor in Clinical Informatics, University of Leeds, Chief Medical
Officer, Private Healthcare Information Network and Leadership and Management Theme
Lead, University of Cambridge School of Clinical Medicine; Dr Ed Fitzgerald, Global
Healthcare Practice Executive, KPMG International; Maxine Mackintosh, PhD student,
Alan Turing Institute; Andrew Morris, Director, Health Data Research UK; Sam Smith,
medConfidential; Dr Harpreet Sood, Associate Chief Clinical Information Officer, NHS
England; Dr Joshua Symons, Director, Big Data & Analytical Unit, Centre for Health Policy
Institute of Global Health Innovation, Imperial College London for helpful comments on an
earlier draft of this paper.
Interviewees
The authors would like to express their gratitude to the 48 individuals and organisations
who kindly agreed to be interviewed as part of the research for this paper and agreed to
be acknowledged:
Mohammad Al-Ubaydli, Founder and CEO, Patients Know Best
George Batchelor, Director, Edge Health
Guy Boersoma, Managing Director, Kent Surrey Sussex Academic Health Science
Network
Chris Boulton, Associate Director of Research and Governance, National Joint Registry
Chris Carrigan, Expert Data Adviser, use MY data
David Champeaux, Cognitive Digital Healthcare Solutions Leader, IPsoft
Andrew Chapman, Digital Health Sector Lead, Digital Catapult
2

Dr David Churchman, Deputy Head of Licensing & Ventures, Clinical Outcomes Business
Manager, Oxford University Innovation
Dr David Clifton, Associate Professor, Department of Engineering science, Oxford
University
Prof Mike Denis, Director of Information Strategy, Oxford Academic Health Science
Network
Hanan Drobiner, Director of Sales, MDClone
Rachel Dunscombe, Chief Executive Officer, NHS Digital Academy and Director of Digital
NCA/Salford Royal Group
Saira Ghafur, Lead for Digital Health Policy, Institute of Global Health Innovation, Imperial
College London
Russell Gundry, Director, 3RDOpinion (has since changed position)
Declan Hadley Digital Lead for Healthy Lancashire and South Cumbria Change
Programme, Lancashire STP
Nasrin Hafezparast, Co-Founder and Chief Technology Officer, Outcomes Based
Healthcare
Andreas Haimbock-Ticky, Director, Health and Life Sciences, IBM UKI
Prof Harry Hemingway, Professor of Clinical Epidemiology, University College London
Prof Aroon Hingorani, Chair of Genetic Epidemiology, Institute of Cardiovascular Science,
University College London
Dr Julian Huppert, Director of the Intellectual Forum, Jesus College, Cambridge, and
former Chair of the DeepMind Health Independent Review Panel.
Darren Jones MP, Member of the Science and Technology Select Committee
Prof Fred Kemp, Deputy Head of Licensing & Ventures, Life Sciences, Oxford University
Innovation
Susan Kennedy, Educationalist and Project Director, Health Education England
Rabia T. Khan, PhD, MBA, Associate Director, Strategy and Planning, BenevolentAI (has
since changed position)
David King, Cognitive Healthcare Solutions Architect, IPsoft (UK) Ltd
Dominic King, Health Lead, DeepMind
Norman Lamb MP, Chair, Science and Technology Select Committee
Prof Paul Leeson, Professor of Cardiovascular Medicine, University of Oxford and Co-
Founder, Ultromics
Lord Mitchell, Member, House of Lords
Prof Andrew Morris, Director, Health Data Research UK
Lord O’Shaughnessy, Parliamentary Under-Secretary, Department of Health and Social
Care
Dr Stephen Pavis, Head of Service, Public Health and Intelligence, NHS Scotland
Nicola Perrin, Lead, Understanding Patient Data (has since changed position)
Dr Navin Ramachandran, Radiology consultant, University College London Hospital
Dr Daniel Ray, Director of Data, NHS Digital
Dr Sobia Raza, Head of Science, PHG Foundation (has since changed position)
Nick Scott-Ram, Director of Strategic and Industry Partnerships, Oxford Academic Health
Science Network
Will Smart, Chief Information Officer, Health and Social Care, NHS England
3

Sam Smith, medConfidential

Harpreet Sood, Associate Chief Clinical Information Officer, NHS England
Alison Stone, Coordinator, use MY data
Nathan Suneeta, Professional Staff Team Lead, Institute of Cardiovascular Science,
Joshua Symons, Director, Big Data & Analytical Unit, Centre for Health Policy Institute of
Global Health Innovation, Imperial College London
Lydia Torne, Managing Associate, Simmons & Simmons
Dr Philip Xiu, Chief Medical Officer, MedicalChain
And 3 individuals who preferred to remain anonymous.
Events
Reform held two events on the topic of this paper, “Making NHS data work for everyone”.
A panel event on 5 June 20181 and a roundtable event on 24 September at Labour Party
Conference2 under Chatham House Rule. We would like to thank the panellists and
attendees for their contribution to the discussion.
Reform
Reform is established as the leading Westminster think tank for public service reform. We
are dedicated to achieving better and smarter public services. Our mission is to set out
ideas that will improve public services for all and deliver value for money.
We work on core sectors such as health and social care, education, home affairs and
justice, and work and pensions. Our work also covers issues that cut across these
sectors, including public service design and delivery and digital public services.
We are determinedly independent and strictly non-party political in our approach.
Reform is a registered charity, the Reform Research Trust, charity no.1103739. This
publication is the property of the Reform Research Trust.
This report was published independently and may not represent the views of our donors
and partners. Reform’s three largest corporate donors in 2017 were Accenture, Hewlett
Packard Enterprise, Baker McKenzie. For a full list of our corporate donors and further
information about Reform’s funding model see our webpage.
The arguments and any errors that remain are the authors’ and the authors’ alone.
1 Reform, ‘Making NHS Data Work for Everyone (Panel Event)’, Webpage, Reform, 5 June 2018.
2 Reform, ‘Labour Party Conference: “Making NHS Data Work for Everyone” (Roundtable Event)’, Webpage, Reform, 24
September 2018.
4
Contents
Executive summary 6
Recommendations 8
Introduction 9
1 Uses of healthcare data 11
1.1 Direct patient care 12
1.2 Secondary uses 13
1.2.1 Uses outside of the NHS 13
1.2.2 Access to data 14
2 The value exchange 15
2.1 Improvement for patients and the healthcare system 16
2.2 Data 17
2.2.1 What does value in data mean? 17
2.2.2 What affects value? 19
2.2.3 How to measure value 22
2.3 Expertise 23
2.4 Wider societal impact 24
3 Keeping society in the loop 25
3.1 Currently not in the loop 26
3.2 Engagement by design 28
4 The healthtech ecosystem 30
4.1 Letting a thousand flowers bloom? 31
4.2 The need for a national strategy 33
4.3 Access to good quality data 36
4.3.1 Digital innovation hubs 38
4.3.2 Synthetic data 39
4.4 Procurement, partnership, or somewhere in between? 40
4.5 Commercial and legal skills 42
Conclusion 44
Appendix 46
Glossary 52
Bibliography 54
5
Executive summary
This paper focuses on the use of NHS data by private sector companies. It highlights the
different elements that should be considered when the public and private sector partner
to build a product or service based on data – i.e. what each side is bringing to the table. It
looks at the challenges the NHS might face given the types of partnership that are
developing on the ground and recommends a strategic national approach to resolve
some of these issues.
Uses of healthcare data

Everyone and everything within the NHS generate data daily, from patients to doctors and
nurses, MRI scanners to appointment booking systems. The primary purpose of this data
is direct patient care which means that it is used to inform medical decisions about a
patient’s treatment. Data held and collected by NHS organisations can be used for other
purposes such as research by universities or by private sector companies, or for product
and service development.
The value exchange

The private sector is an important partner to the NHS and plays a crucial role in the
development of healthcare technologies that use data collected by hospitals or GP
practices. It provides the skills and know-how to develop data-driven tools which can be
used to improve patient care. However, this is not a one-sided exchange as the NHS
makes the data available to build these tools and offers medical expertise to make sense
of the data. This is known as the “value exchange”. The value exchange is specific to
each product and service. The research uncovered that there is a lack of clarity over what
a fair value exchange looks like. This is partly because healthcare data is an asset that no
one knows how to adequately value. The paper builds a list of factors that have an impact
on its value. However, it also highlights that the value exchange between the public and
private sector is about more than just data.
Keeping society in the loop

Engaging with the public about how data about them is being used within the NHS is
crucial. Various models have been developed to do so. There are, however, no public
sector models or examples that allow patients to participate in the conversation about
what fair value exchange with the private sector might be. The report recommends that
this should change, and people should be given an opportunity to engage in these
conversations. This is crucial to build a trustworthy system.
The healthtech ecosystem

To create a strong healthtech ecosystem whilst safeguarding the NHS’s constitution and
patients’ trust, strong guidance and leadership will be needed to make sure the value of
healthcare data is optimised.
Letting a thousand flowers bloom?

Research carried out for this report suggests that there is a wide variety of arrangements
between the NHS and industry when there is access to data for research and product or
service development purposes. However, it is difficult to gain an understanding of what
the national picture is as there is no national registry describing data sharing agreements
and the various types of commercial models that have developed on the ground. The
paper presents the first classification of the different models that can be found.
6

National strategy
It is crucial for the Department of Health and Social Care to ensure that local bodies are
not creating perverse incentives in the system. A commercial model might seem fair at GP
practice or Trust-level but could be detrimental at a national level. For example, a hospital
could receive a financial revenue from a partnership it has engaged in, because of the way
it was structured. This might be fair at a local level, but this increases the risk of some
Trusts becoming richer than others. The paper presents a table of various other
commercial models that could be explored. All models have advantages and
disadvantages, meaning that different models will suit different scenarios. There is no
one-size-fits-all solution. Nevertheless, it is crucial that national policy provides a
framework for the array of possible models that will not have an adverse effect at the
national level.
Access to good quality data

Most data held by NHS organisations is collected for the purposes of direct patient care.
This has an impact on its ability to be directly used for research or product or service
development purposes. In addition, data within the NHS is fragmented and there can be
issues around its quality. It is important for the private sector to have clarity over this
before engaging in a partnership. The paper suggests that transparency over data quality
could be increased. The speed of access to data can also be an issue for private sector
organisations, particularly smaller companies, however, accelerating this process should
never come at the price of protecting individuals’ privacy. The report suggests that
privacy-preserving techniques like synthetic data – a fake version of real data that is
sufficiently different so that it preserves a person’s anonymity, but sufficiently similar that it
can be used for analysis – should be used to speed up access. This would also allow a
better understanding of the commercial value of data-led innovations and a more
informed conversation about the appropriate type of commercial model that should be
developed.
Procurement, partnership, or somewhere in between?

Partnership models seem to offer more flexibility at the early stages of the innovation
process and might be a way to alleviate some of the concerns around the potentially
ill-fitted nature of current procurement processes for data-driven technologies. However,
there might be a need to clarify the boundaries between some commercial partnerships
and procurement processes.
Commercial and legal skills

To create a thriving healthcare ecosystem in which the private sector acts as a partner to
the public, the public sector will need the right skills to negotiate these partnerships.
Having an even distribution of commercial and legal skills at Trust-level would be
unfeasible as those skills are a scarce resource within the public sector. They are also an
expensive resource. The report recommends that the Department of Health and Social
Care should invest in creating a new independent unit with legal and business experts to
help NHS organisations negotiate fair and proportionate partnerships.
7

Recommendations
1 The Office for National Statistics should provide a framework and accounting
standards for measuring and reporting the value of knowledge assets such as
healthcare data.
2 The Department of Health and Social Care should in conjunction with Caldicott
Guardians, NHS organisations and industry representatives ensure that a dialogue
with the public is set at a local level to discuss commercial models. Proportionate
governance models provide an interesting avenue for public and patient engagement
that could be explored.
3 NHS England and NHS Digital should create a register of data sharing agreements
between the NHS and commercial organisations. It should include what type of data
are being shared and a description of the type of partnership model being used. This
would allow for clear understanding of what is happening on the ground and facilitate
public scrutiny.
4 The Department for Health and Social Care in conjunction with the Crown Commercial
Service, Office for Life Sciences and HM Treasury should make sure to include in its
formal review of commercial partnerships, a macroeconomic study of the impact that
different partnership models might have to avoid reinforcing a postcode lottery or
other negative externalities such as exclusive data access.
5 The Department of Health and Social Care should include the results of this
macroeconomic study in a clear national strategy which should seek to optimise the
value of data held by NHS organisations when it is accessed for commercial
purposes. It should consult with all stakeholders including industry, patients and NHS
organisations as to what a fair apportioning of value might be.
6 NHS England should create a ‘Data Quality Service’, with a tiered-fee system
dependent on factors such as company size and global profits, to provide bespoke
reports on data quality at the early stages of a partnership discussion between the
NHS and industry.
7 Procurement rules should include an agreement by digital providers that data
generated within clinical applications should also be freely available for and
interoperable with clinical information at the patient level, either via personal health
records or interchange with appropriate electronic health records.
8 Health Data Research UK in conjunction with NHS England, NHS Digital, the National
Data Guardian should work on developing the appropriate data governance
structures to ensure that Digital Innovation Hubs are safeguarding patient data. This
would include developing audit trails which track how data are used to ensure every
interaction with personal data is auditable, transparent and secure.
9 NHS organisations should offer synthetic datasets, which they can share with private
sector organisations for research and product or service development at the early
stages of the innovation process. This would enable a better understanding of the
commercial value of the innovation and a more informed conversation about the
appropriate type of commercial model that should be developed.
10 The Department of Health and Social Care should invest in creating a new
independent unit with legal and business experts to help NHS organisations negotiate
fair and proportionate partnerships. It would ensure that NHS organisations can have
access to consistent and necessary advice in order to negotiate fair partnership with
SMEs. The negotiation of partnerships should be done by this unit on behalf of Trusts
when dealing with Large Enterprise or SMEs whose parent companies are Large
Enterprise.
8

Introduction
There are many benefits that can emerge from the NHS partnering with the private sector
for the development of healthcare technologies, such as access to better treatments and
increased quality of care.3 However, as highlighted in a recent HMTreasury policy paper on
intangible assets, such as data and intellectual property (IP), the UK does not always have
a good history in optimising the value of public sector knowledge assets. The document
calls for a better understanding of the value of these assets as well as making sure that
the frameworks are in place to ensure that value is accrued by the UK.
The Secretary of State for Health and Social Care’s vision for the creation of a tech-driven
NHS seeks to take a radical approach “to technology across the system” and to lay the
foundations for the UK to become a world leader in healthcare technologies.4 A key theme
developed in the vision is the importance of collaboration and partnerships. The hope is to
create “an ecosystem where developers and vendors continuously compete on quality to
fill each niche”.5 There is a lot of debate around the commercial value of data controlled by
the NHS and how to ensure that partnerships are beneficial to patients, the NHS and
industry.6 Patients have a right to object to data being used for purposes beyond direct
care, which will increase if they do not have confidence in what will happen to data about
them. Protecting patients’ privacy and trust will be crucial.
The NHS “holds millions of electronic medical records on the health of the population from
birth to death”7 as well as extensive administrative datasets capturing, for example, the
number of operations being carried out in the operating theatres of a hospital. This wealth
of information is extremely valuable for the improvement of the quality of care and patient
outcomes.8 There is also an “explosion of health care data” being generated outside of
traditional healthcare settings thanks to the proliferation of smart healthcare devices and
trackers.9 However, the value of healthcare data is often trapped as “different kinds of
individual-patient data reside in disparate, unlinked silos”.10 Interviews carried out for this
paper revealed that accessing, cleaning, validating and linking data, known as data
engineering, represents most of the labour involved in creating data-driven technologies,
such as artificial intelligence, or new treatments that can be discovered through the
analysis of data.
There is a fear that as big tech corporations partner with the NHS they will eclipse SMEs,
further reinforcing their position as monopolists. The Department of Health and Social
Care (DoHSC) has responded to this by launching a draft code of conduct for technology
companies that use NHS data to create products.11 It presents ten broad principles that
are supposed to set the rules of engagement and expected behaviours. It is also currently
conducting a review of commercial models.12 Although this a promising start, there is a
lack of a national strategy and forum to engage with the public on commercial models.
The recent transfer of the Streams App developed for acute kidney injury by DeepMind to
3 Sir John Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector (HM
Government, 2017); Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, Report of Session
2017–19 (House of Lords, 2018).
4 Department of Health and Social Care, The Future of Healthcare: Our Vision for Digital, Data and Technology in Health
and Care, 2018.
5 Ibid.
6 Lord Parry Mitchell, ‘Protect Our Valuable NHS Data from Big Tech’, Financial Times, 7 February 2018; Select
Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, 88–93; Philip Aldrick, ‘If NHS Patient Data Is
Worth £10 Billion, Put It on the Balance Sheet and Save Lives Too’, The Times, 15 June 2018; Philip Aldrick, ‘Data Could
Be a Huge Source of Funding for the NHS and We Are about to Give It Away’, The Times, 3 March 2018; ibid.; Philip
Aldrick, ‘NHS Set to Be Offered a “Fair Share” of Data Profits’, The Times, 5 September 2018; Rebecca Hill, ‘Making
Good Use of NHS Data “Vital for UK Plc”, Says Royal Society’, Public Technology, 26 April 2017.
7 Peter Border, Big Data and Public Health (Parliamentary Office of Science and Technology, 2014), 1.
8 Michael E. Porter, ‘What Is Value in Health Care?’, New England Journal of Medicine 363, no. 26 (December 2010); The
Office for National Statistics, ‘Quality Adjustment’, in The ONS Productivity Handbook, 2016, 72–79.
9 Joachim Roski, George W. Bo-Linn, and Timothy A. Andrews, ‘Creating Value In Health Care Through Big Data:
Opportunities And Policy Implications’, Health Affairs 33, no. 7 (2014): 1115.
10 Alessandro Blasimme, Effy Vayena, and Ernst Hafen, ‘Democratizing Health Research Through Data Cooperatives’,
Philosophy & Technology 31, no. 3 (September 2018): 473.
11 Department of Health and Social Care, Initial Code of Conduct for Data-Driven Health and Care Technology, Guidance,
2018.
12 Ibid.
9

Making NHS data work for everyone / Introduction
its parent company Google, has sparked up fears around privacy13 and ownership of IP.14
Crucially, it has also brought to forefront the necessity to have a conversation about
commercial access to healthcare data and the type of partnership models that are
developing on the ground.
This paper will focus on the partnerships between NHS and industry when there is use of
data for research and development purposes or to create products and services. It will
highlight the importance of clarifying the NHS’s value proposition and will depict the
current partnership landscape. In the House of Lords report, AI in the UK: Ready, Willing
and Able?, concerns are raised about “the current piecemeal approach taken by NHS
Trusts”.15 There is a lack of an overall strategy considering whether the different
partnership models that are emerging on the ground lead to the best possible at a
national scale.16 In addition, it highlights the importance of engaging with the public on the
issue of commercial models and the use of data about them.
13 Charlotte Jee, ‘Google’s Decision to Absorb DeepMind’s Health Division Has Sparked Privacy Fears’, MIT Technology
Review, 14 November 2018; Madhumita Murgia, ‘DeepMind’s Move to Transfer Health Unit to Google Stirs Data Fears’,
Financial Times, 13 November 2018.
14 Alex Ralph and Philip Aldrick, ‘Google “Poised to Profit” from NHS Patient Records’, The Times, 24 November 2018.
15 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, 93.
16 Ibid.
10
1
Uses of healthcare data
1.1 Direct patient care 12

1.2 Secondary uses 13
1.2.1 Uses outside of the NHS 13
1.2.2 Access to data 14
11
1
Making NHS data work for everyone / Uses of healthcare data
The data collected and controlled by NHS organisations is set in the context of a
perpetual evolution.17 New data are being continuously generated by patients and
healthcare organisations through appointment booking systems or the use of X-ray
machines. There are improvements in the quality of data collected within the clinical
setting thanks to the introduction of better scanners or data from trackers. It is not a static
space. However, to produce benefits for the individual and the healthcare system, data
needs to be used and shared.18 This can be done in several ways within the healthcare
ecosystem; and there is robust legislation and guidance safeguarding and regulating what
can be done to healthcare data.19 The primary purpose of data collected by NHS
organisations is direct patient care. The benefits of granting access to data for this
purpose are generally clear: improving the quality of care and the patient experience.
Healthcare data can also be used for other purposes, namely secondary uses. These can
occur within the NHS – for example monitoring population outcomes (e.g. reduction in the
prevalence of diabetes) – or outside the health service, by third-party organisations – like
universities or the private sector. Healthcare data might be used for research purposes by
both universities and commercial organisations.20 Generally speaking, the purpose of
research is to advance knowledge, however, it can sometimes lead to the development of
a product or service (see Figure 1). Data collected and controlled by NHS organisations
can, within clearly specified parameters, be used directly for product or service
development by commercial organisations. There can be clear benefits for the patients
and the healthcare system to use data in this way, however, this type of use is more
contentious in the eyes of the public21 and sometimes rightfully so.22
1.1 Direct patient care

NHS organisations currently hold millions of electronic healthcare records on the UK
population from cradle to grave.23 Most of these data are produced when individuals
interact with healthcare services and are primarily collected for the purposes of direct
patient care – meaning that medical professionals use this information to make decisions
about care.24 The value of using data in this way is undeniable and immediately
measurable as it directly increases the quality of care.25 For example, the type of care a
secondary care specialist will be able to deliver if they are able to access a person’s GP
records will be faster and more accurate than if they are solely relying on the patient’s
recall.
Although there are clear benefits to using healthcare data to inform medical decision-
making, it is not always possible to access these data in a timely fashion. Patients often
assume that data between primary and secondary care are digitally shared, however this
17 See Glossary for definition of data controller.
18 Hugh Whittall, ‘The Use and Reuse of Data’, in Connecting Debates on the Governance of Data and Its Use (The Royal
Society and British Academy, 2016), 14.
19 Information Commissioner’s Office, ‘Overview of the General Data Protection Regulation (GDPR)’, Webpage, 2017;
Information Commissioner’s Office, Guide to the General Data Protection Regulation (GDPR), 2018; Information
Commissioner’s Office, Guide to Data Protection, 2017; National Data Guardian for Health and Care, Information: To
Share or Not to Share? Information Governance Review (Department of Health, 2013); National Data Guardian for Health
and Care, Review of Data Security, Consent and Opt-Outs, 2016; National Data Guardian for Health and Care, Impact
and Influence for Patients and Service Users (Department of Health, 2017); Department of Health and Social Care and
NHS, ‘Information Governance Toolkit’, Webpage, n.d.; Department for Digital, Culture, Media and Sport, Data
Protection Bill: Factsheet – Overview, 2017.
20 Chris Smith, NHS and Healthcare Data (House of Lord’s Research Briefing, 2018), 2.
21 Wellcome Trust, The One-Way Mirror: Public Attitudes to Commercial Access to Health Data (Ipsos MORI, 2016).
22 BBC Two, ‘Diagnosis on Demand? The Computer Will See You Now’, Horizons, 5 November 2018; Timothy Revell,
‘Google DeepMind NHS Data Deal Was “Legally Inappropriate”’, New Scientist, 16 May 2017; Jee, ‘Google’s Decision to
Absorb DeepMind’s Health Division Has Sparked Privacy Fears’; Naureen Bhatti, ‘Seeing a GP on a Smartphone
Sounds Wonderful – but It’s Not’, The Guardian, 16 November 2017; Alex P. Burns, ‘Where Is the Evidence for
Automated Triage Apps?’, British Medical Journal 360 (February 2018); Margaret McCartney, ‘Could Babylon Please
Supply Evidence ?’, BMJ, no. 358 (September 2017).
23 Smith, NHS and Healthcare Data, 1.
24 Ibid., 2.
25 Erin Dettrey, ‘The Value of Healthcare Data Analytics in an EMR’, Webpage, Modernizing Medicine, 13 September 2017;
HIMSS, ‘The Value of Quality Healthcare Data’, Webpage, HIMSS, 16 May 2016; Dave deBronkart, ‘The Value of Sharing
Data: What Healthcare Can Learn from Oncology’, Future Health Index Blog, 3 March 2017; Sophie Clark et al.,
‘Patient-Level Data Linkage across Ambulance Services and Acute Trusts: Assessing the Potential for Improving Patient
Care’, International Journal for Population Data Science 1, no. 1 (April 2017); Dame Fiona Caldicott, ‘A Data Opt-out Will
Reassure Patients and Improve Care’, The Times, 13 July 2017.
12
1
is often not the case.26 Healthcare records are siloed by the activities (e.g. radiology,
laboratory, etc.) performed within NHS organisations (e.g. Trusts, GP surgery etc.).27
Having a single consolidated view of a person’s entire healthcare record is the exception
rather than the rule. This is not only the result of how data are collected but also due to
the lack of interoperability of healthcare IT systems – meaning that systems cannot easily
“exchange and use electronic health information” making it difficult to link data together.28
It is important to highlight that individuals can also produce health data outside of the
medical setting using wearables and other type of sensors. Measurements with these
types of device are continuous and not just confined to a single point in time. In addition,
they can be used as a way of gathering data about good health. However, these types of
data are not currently integrated in a systematic way to the patient’s healthcare record.
1.2 Secondary uses

Secondary uses of healthcare data that happen within the NHS tend to be less
contentious in the eyes of the public than those happening outside and the public benefit
can be clearly expressed. For example, clinical commissioning groups (CCGs) use
healthcare data to effectively design and deliver services. The Clinical Effectiveness Group
in East London, which covers three CCGs, has been using a data-driven approach to
systematically improve clinical standards and reduce variation in primary care.29 This has
had a positive impact on patient outcomes such as the improvement “in blood pressure
control” for patients with certain types of chronic health conditions such as diabetes.30
However, as highlighted by medConfidential the use of healthcare data for secondary
purposes is not devoid of contention. It reported that NHS England have granted a legal
exemption to pass personal identifiable (see Glossary) data to various commissioning
bodies to use for administrative purposes when there is arguably no need to pass
identifiable data.31
1.2.1 Uses outside of the NHS

Accessing data collected and controlled by NHS organisations for secondary uses
outside of the NHS can be hugely beneficial and is also the key to developing a
competitive life sciences industry from data-led innovation.32 Harnessing the power of
data will mean building relationships with various stakeholders from academia and
industry.33
As shown in Figure 1, a stakeholder’s objective will influence the type of output. Research
carried out for the purpose of peer-reviewed publication seeks to advance knowledge and
is mostly carried out by universities. It is generally transparent in its processes and
validated through a rigorous peer-review system. Research carried out for product or
service development is rarely openly published due to commercial sensitivity. In addition,
product or service development research might not always be carried out with the same
rigour as peer-reviewed fundamental or applied research.34 However, there can be
interactions between published peer-reviewed research and product development, as one
can lead to the other. This paper will focus on the uses of NHS data by commercial
organisations.
26 Rebecca Fisher, ‘Data Sharing to Improve Care’, Webpage, The Health Foundation, 18 May 2018.
27 Eleonora Harwich and Kate Laycock, Thinking on Its Own: AI in the NHS (Reform, 2018), 33.
28 Robert M. Wachter, Making IT Work: Harnessing the Power of Health Information Technology to Improve Care in
England (Department of Health and Social Care, 2016), 49.
29 Rebecca Fisher, Ruth Thorlby, and Will Warburton, ‘Briefing: Sharing to Improve: Four Case Studies of Data Sharing in
General Practice’, The Health Foundation, May 2018, 4.
30 Ibid., 6.
31 MedConfidential, ‘More Information’, Web Page, MedConfidential, 2017.
32 Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector; Department for
Digital, Culture, Media and Sport, UK Digital Strategy 2017, 2017; House of Lords, Select Committee on Digital Skills.
Summary and Summary of Conclusions and Recommendations, 2015; Jon Hoeksma, ‘NHS England to Establish New
Regional Digital Innovation Hubs’, Webpage, Digital Health, 12 September 2017.
33 Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector, 50.
34 BBC Two, ‘BBC Two – Horizon, 2018, Diagnosis on Demand?’
13
1
Figure 1: Stages of research, development and product/service development
Idea/ Patient/user Proof of

Inception engagement concept
Stakeholder objective:
Fundamental research
Applied research
Product development
Publication and
Data access Research Peer review dissemination
of research
Final Pre-market Market

product/service launch launch
Product/service
development
Product/ Compliance Commercial

service testing/ with models
trialling/ standards,
sandbox regulation and
accreditation
Source: Reform interviews and research.
Discussions about commercial models arise in the later stages of product development
(see Figure 1). It is important to understand the array of available models that could
benefit patients, the NHS as a whole – and not just parts of its thousands of local and
national organisations – and the private sector.
1.2.2 Access to data

Accessing data held by NHS organisations can be a complex and costly endeavor.
Interviews carried out for this paper uncovered that it is often complicated to know what
data are being collected at Trust level and consequently what other types of data might
need to be collected in order to complete a research project or develop a given data-
driven technology or service.
At national level or even at regional level, only a minority of data are organised in neat or
‘clean’ datasets ready for external providers to request access to.35 In addition, these
“sources, whilst they may be linked in some cases, do not all currently provide deep, near
real-time data for research across multiple care settings as standard.”36 The fragmentation
of data across the system and the lack of interoperability of IT systems makes it difficult to
access data.37
Interviews also revealed that even when data access requests are compliant with
legislation and information governance frameworks (see Appendix) and projects have
gone through ethical approvals, there can be delays in obtaining access. Circumstances
such as these hinder the entry of smaller businesses with fewer resources from entering
the market. NHS organisations, such as NHS Digital have recognised this issue and have
“reduced time to completion through the Data Access Request Service to an average of
60 days”.38 However, this is for initial approval once an application is submitted and is not
reflective of actual end to end process, which can take much longer.
35 Cliff Saran, ‘NHS Data Not Fit for AI, Lords Select Committee Told’, Webpage, ComputerWeekly.com, n.d.
37 Wachter, Making IT Work: Harnessing the Power of Health Information Technology to Improve Care in England; Harwich
and Laycock, Thinking on Its Own: AI in the NHS, 29–38; Health Data Research UK, ‘Digital Innovation Hub Programme’,
Webpage, Health Data Research UK, n.d.
14
2
The value exchange
2.1 Improvement for patients and the healthcare system 16

2.2 Data 17
2.2.1 What does value in data mean? 17
2.2.2 What affects value? 19
2.2.3 How to measure value 22
2.3 Expertise 23
2.4 Wider societal impact 24
15
2
Making NHS data work for everyone / The value exchange
The DoHSC’s code of conduct for data-driven health and care technology highlights that
partnerships between NHS organisations and the private sector can and should “deliver
benefits to patients, clinicians, industry and the health and care system as a whole.”39 The
private sector could be a helpful partner by bringing its expertise and capabilities for
research or to develop data-driven technologies and services – including the cost of
compliance with standards and regulation for medical devices – as NHS organisations
often do not have the necessary expertise or funds to acquire the skills needed to fully
develop these in-house.40 Despite this the NHS has a lot to bring to the negotiation table
such as medical expertise and data. It is therefore crucial that it understands what its
value proposition is – defined as the “positioning statement that explains what benefit you
provide for who and how you do it uniquely well”.41 A mutually beneficial value exchange
is only possible if there is a discussion and clarity over “what contribution is expected of
each party”.42 The private sector plays an integral role in the provision of healthcare
services in the UK. There are clear advantages for procuring products and services from
the private sector such as CT scanners or other medical equipment. However, as revealed
by interviews carried out for this paper, the value exchange is not always clear, and the
healthcare system lacks people with the commercial skills to engage in these
conversations.43
2.1 Improvement for patients and the healthcare system

Arguably, the most important part of the value exchange in a partnership between the
NHS and the private sector is demonstrating “how and where the product will add value
to people and the health and care system”.44 In other words, how will these advances
improve patient outcomes whilst safeguarding privacy, population health and the
productivity of the healthcare system. This is the legal basis used by most organisations
to access data controlled by NHS organisations.
Data-driven technologies can provide support for clinical decision-making by, for
example, helping to reduce the risk of morbidity and complications through improved
diagnostics. Machine learning algorithms (see Glossary) can interpret mammography
scans with high accuracy rates when screening for breast cancer.45 This can improve early
detection and help doctors better triage patients. Connected Yorkshire has created a new
Electronic Frailty Index (eFI), based on routine patient data collected over time.46 It helps
doctors measure what support elderly people need as they reach old age. Based on the
eFI clinicians can make more accurate interventions to avoid elderly patients from falling
over or experiencing more acute health problems.
Operational improvements can also be derived from the development and implementation
of data-driven products or services on the ground. Predictive analytics can be used to
help accurately forecast bed occupancy rates and plan accordingly.47 Bradford hospital in
collaboration with General Electric is set to create a hub monitoring patients, allowing the
hospital to “make better use of scarce resources”. 48
39 Department of Health and Social Care, Initial Code of Conduct for Data-Driven Health and Care Technology.
40 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, p.88.
41 Michael Skok, ‘4 Steps To Building A Compelling Value Proposition’, Forbes, 14 June 2013.
43 Ralph and Aldrick, ‘Google “Poised to Profit” from NHS Patient Records’.
45 Nehmat Houssami et al., ‘Artificial Intelligence for Breast Cancer Screening: Opportunity or Hype?’, Breast 36
(December 2017); Alejandro Rodriguez-Ruiz et al., ‘Can Radiologists Improve Their Breast Cancer Detection in
Mammography When Using a Deep Learning Based Computer System as Decision Support?’, vol. 10718 (14th
International Workshop on Breast Imaging (IWBI 2018), International Society for Optics and Photonics, 2018); Tejal A.
Patel et al., ‘Correlating Mammographic and Pathologic Findings in Clinical Decision Support Using Natural Language
Processing and Data Mining Methods’, Cancer 123, no. 1 (January 2017); L. Hussain et al., ‘Automated Breast Cancer
Detection Using Machine Learning Techniques by Extracting Different Feature Extracting Strategies’ (2018 17th IEEE
International Conference On Trust, Security And Privacy In Computing And Communications/ 12th IEEE International
Conference On Big Data Science And Engineering (TrustCom/BigDataSE), IEEE, 2018).
46 Connected Health Cities, ‘Connected Yorkshire’, Webpage, Connected Health Cities, 2016.
47 Edge Health, ‘Prediction Tool for Critical Care Bed Occupancy – “Critical Planner”’, Webpage, Edge Health, 18 October
2017.
48 Sarah Neville, ‘Bradford Hospital Launches AI Powered Command Centre’, Financial Times, 15 October 2018.
16
2
At a national level, it is important to consider what the deployment of these data-driven

technologies means for the future of the NHS and how these integrate with the NHS’s
objectives of reducing the health and wellbeing gap (i.e. focus on prevention to improve
healthy life expectancy);49and the care and quality gap (i.e. standardise high-quality
care).50 There might be trade-offs to consider when deciding whether to incentivise
data-led technologies in prevention (i.e. avoiding the ailment and need for health system
interaction), treatment (i.e. addressing or mitigating the symptoms) or cure. Data-led
innovations that focus on prevention may be less immediately profitable than treatment,
but have the potential to reduce burden and cost on healthcare systems and might be
more socially desirable long-term.
Privacy concerns should also be considered as part of value exchange. Research has
found that “machine learning models are vulnerable to a range of cybersecurity attacks
that cause breaches of confidentiality”.51 This has led to some discussion over the nature
of these algorithms and how they should be treated (i.e. should they be treated as
personal data).52 The vulnerability of models to confidentiality attacks raises legitimate
questions over the commercialisation of ‘vulnerable’ machine learning algorithms based
on UK patient data abroad. Finding the appropriate governance models for the
commercialisation of these products or services will be crucial.
2.2 Data
Data collected and controlled by NHS organisations are often described as ‘a gold
mine’,53 because the NHS is one of the oldest health services of its kind in the world, with
– in theory – a unique national patient identifier.54 Some have argued that these data could
be harnessed to offer an additional revenue stream for the NHS55 or invested into a
sovereign wealth fund for the UK.56 These ideas can be attractive, however, it is important
to remember the financial value of data controlled by the NHS has not yet been
established. Valuing data could allow the Government to better understand the relative
returns from partnerships with the private sector and therefore ensure that it is optimising
the value between patients, the NHS and industry. Several features of data “make them
difficult to value within a traditional balance-sheet accounting framework.”57 Nonetheless,
HM Treasury has recently released a paper on valuation techniques for intangible and
knowledge assets (see Glossary) – like data – which makes a few recommendations on
what should be considered in future guidance.58
2.2.1 What does value in data mean?

There is no single definition of the value of NHS data and no single way of measuring.59
This is because the value that is ascribed to healthcare data is dependent on both “the
intrinsic characteristics of the data and the environment in which they are used.”60 A data
49 NHS England, Five Year Forward View, 2014, 10; NHS England, Next Steps on the NHS Five Year Forward View, 2017, 9;
Department of Health and Social Care, The Future of Healthcare; Duncan Selbie, ‘Prevention and the NHS Long Term
Plan: 3 Ways We Can Save More Lives’, Press Release, Public Health England, 6 September 2018; Anna Coote,
Prevention Rather Than Cure (The King’s Fund, 2004).
50 NHS England, Next Steps on the NHS Five Year Forward View.
51 Michael Veale, Reuben Binns, and Lilian Edwards, ‘Algorithms That Remember: Model Inversion Attacks and Data
Protection Law’, Phil. Trans. R. Soc. A, no. 376 (July 2018): 4.
52 Ibid., 12.
53 Aldrick, ‘Data Could Be a Huge Source of Funding for the NHS and We Are about to Give It Away’; Select Committee on
Artificial Intelligence, AI in the UK: Ready, Willing and Able, 88–89; Mitchell, ‘Protect Our Valuable NHS Data from Big
Tech’.
54 In practice, there are duplicates of NHS number and wrong attributions of NHS numbers. See: Sarah Timmis, Luke
Heselwood, and Eleonora Harwich, Sharing the Benefits: How to Use Data Effectively in the Public Sector (Reform,
2018), 19.
55 Aldrick, ‘Data Could Be a Huge Source of Funding for the NHS and We Are about to Give It Away’; Aldrick, ‘If NHS
Patient Data Is Worth £10 Billion, Put It on the Balance Sheet and Save Lives Too’; Mitchell, ‘Protect Our Valuable NHS
Data from Big Tech’.
56 Annemarie Naylor, ‘Parliamentary Debate Briefing: This House Takes Note of the Value of NHS and Healthcare Data and
How It Could Be Used to Improve the Health of Our Nation.’, Future Care Capital, 6 September 2018.
57 Centre for Economics and Business Research, Data on the Balance Sheet (SAS, 2013).
58 HM Treasury, Getting Smart about Intellectual Property and Other Intangibles in the Public Sector: Budget 2018, 2018.
59 Ibid., 8–10; Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, 88–89.
60 Centre for Economics and Business Research, Data on the Balance Sheet, 3.
17
2
value chain exists “whereby the value increases as data are transformed into information,
knowledge and ultimately action”.61 The exact nature of that value might depend on the
type of application and type of data used.62 The nature and amount of value ascribed to a
particular data-driven product or service “always lie in the eye of the beholder”63 (see
Figure 2).
Figure 2: Value(s) of NHS data
Source: Reform interviews and research. Gif by Zuhura Plummer.
61 Alan Walker and Philippa Westbury, ‘Towards a Successful Data-Enabled Economy: Promoting Trust in Data and
Data-Driven Systems’, in Connecting Debates on the Governance of Data and Its Use, 2017, 40.
62 Ibid.
63 Eric Almquist, John Senior, and Nicolas Bloch, ‘The Elements of Value’, Harvard Business Review, 1 September 2016.
18
2
Value is highly complex to define because as shown in Figure 2 different stakeholders

might have a different perspective on what value means to them. Value can also be added
by cleaning and curating data.
2.2.2 What affects value?

Although healthcare data are part of a broader value chain, research and interviews
carried out for this paper have uncovered 10 broad determinants that affect its value (see
Figure 3 and Figure 4). The elements presented in Figure 3 describe the supply factors
affecting the value of healthcare data – in other words, what makes data in isolation
valuable.
19
2
Figure 3: Determinants of the value of data – the supply side 64 65 66 67 68 69 70 71 72 73
Category Definition
The quality of the data The better the quality of the data, the greater the value.64 This is because of the
accuracy of the insights derived from the data. There are several factors which affect the
quality of healthcare data: integrity, timeliness, validity, completeness and coverage.65
Integrity ensures that data are accurate (e.g. a patient’s date of birth is rightly inputted).
Timeliness ensures that data are recorded at the time of the event or as close to the
event as possible and that a timestamp is added to the “metadata”.66 Validity ensures
that collected data satisfy a set of standards (e.g. making sure that the accepted clinical
coding standards are used when coding a diagnostic or procedure). In the NHS coverage
is reached if data “have been received from all expected data suppliers.”67
In addition, greater quality data can reduce the cost of the data cleaning process. Data
curation is the biggest drain on time when doing research or trying to develop a product
or a service using healthcare data. Several interviewees estimated that about 60 to 70
per cent of the time spent on a project is dedicated to data curation. As highlighted by
Neil Lawrence, professor at the University of Sheffield, there are different levels of data
readiness and stages that need to be passed before data can be analysed by a machine
learning algorithm.68 Data cleaning processes are bespoke to every use and it is an
inevitable process.
The format of data The format of the data will have an impact on its value. Data in a machine-readable
format and accompanied by robust “metadata”69 describing the data content is more
value than information held in paper files. Secondary care (e.g. hospital Trusts) still
SUPPLY FACTORS
heavily relies on paper files.70
The ability to link data The value of a dataset increases when it is linked with another. For example, by
combining different datasets on different disease types, new patterns of comorbidity can
be discovered.71 The capacity to link genomic, clinical and diagnostic, medicines, and
lifestyle data “forms the powerhouse for personalised medicine.”72
The type of data The type of data can have an impact on its value. For example, the value of patient
records can be different to the value of data generated by appointments and booking
systems. The value of personal identifiable data (e.g. data including name, date of birth,
contact details, etc.) can be different from the value of anonymised data (see Appendix
for definition).
Although they are not currently common practice, personal healthcare records might be
more valuable than medical health records. Instead of collecting information through the
silo of healthcare activities (i.e. GP records, lab results, etc.), personal healthcare records
allow for the patient to act as the integrator of the data. The data follows them and is no
longer siloed by activity. Patients are effectively given an active role as data curators and
validators which might have an impact on the quality of this type of data.
The reason for data collection Most data controlled by NHS organisations is primarily collected for the purposes of
direct patient care. It is not collected for other purposes.73 This affects its value for
secondary uses as it might not be the exact type of data needed for research or product/
service development purposes.
Interviews carried out forf this paper highlighted that some research projects include an
entire process of data collection within an NHS setting as a very specific type of data
might be needed for the purposes of research. In several cases, data collected as part of
research projects can become long-term databases which can then be used for clinical
care, such as disease registries or system-level dashboards.
65 Data Services for Commissioners, Data Quality Guidance for Providers and Commissioners (NHS England, 2016);
Nicole Gray Weiskopf and Chunhua Weng, ‘Methods and Dimensions of Electronic Health Record Data Quality
Assessment: Enabling Reuse for Clinical Research’, Journal of the American Medical Informatics Association 20, no. 1
(January 2013).
66 Walker and Westbury, ‘Towards a Successful Data-Enabled Economy: Promoting Trust in Data and Data-Driven
Systems’, 40.
67 Data Services for Commissioners, Data Quality Guidance for Providers and Commissioners, 6.
68 Neil D. Lawrence, ‘Data Readiness Levels’, ArXiv, May 2017.
Systems’, 40.
70 Wachter, Making IT Work: Harnessing the Power of Health Information Technology to Improve Care in England.
71 National Institute for Health and Care Excellence, Data Science for Health and Care Excellence, 2016.
72 NHS England, Improving Outcomes through Personalised Medicine, 2016, 7.
73 Timmis, Luke Heselwood, and Harwich, Sharing the Benefits: How to Use Data Effectively in the Public Sector, 17.
20
2
Category Definition
The quantity of data The quantity of healthcare data has an impact on its value. For example, the value of a
single patient record is different to the value of multiple patient records. This is because
the smaller the sample size the harder it is to find robust results or to develop system-
level products. In the context of machine learning, Posner and Weyl’s research provides
an interesting visualisation of the value of data, showing that it increases with the amount
following a step-wise function.74 More data becomes more valuable as it allows for the
analysis of more complicated and potentially more valuable problems.75
Simple level Moderate level High level

of complexity of complexity of complexity
Value of data
SUPPLY FACTORS
Number of samples
According to Posner and Weyl, the capacity of a machine learning algorithm to solve
issues of varying complexity is dependent on the number of samples available.
Source: Eric A. Posner and E. Glen Weyl, ‘Data as Labor’, in Radical Markets
(Princeton University Press, 2018), 227-228
Currently, trade-offs need to be made between the quality and quantity of data as high-
quality data are a scarce resource.
The actionability of data The contrast in the objectives of the stakeholders highlighted in Figure 1 becomes
apparent when looking at the actionability of data. For example, academic researchers
might produce excellent research, but it might not have immediate, obvious or realistic
real-world applications. However, this type of research might produce value in real-world
applications in an indirect way. By advancing fundamental research, this might in turn
lead to some practical real-world applications. Another example would be the tension
between a data scientist wanting to develop a high-performing model and what might be
realistic for the NHS. For example, if the model is good at finding true positives (i.e. the
proportion of people with a specific condition who are correctly identified as having that
condition by the algorithms) but produces many false positives (i.e. healthy individuals
who are told they have a condition when they do not), it can have severe financial
implications for the NHS.
The ability to produce change on the ground thanks to insights derived from data are
essential to derive value. If the insights provided by the eFI did not lead to action, there
would be no value generated from data in terms of improved patient outcomes. Making
sure that insights are converted into best practice on the ground is not an easy feat.76
The production of new clinical evidence around technologies and medicine “does not
guarantee its implementation”.77
74 75 76 77
74 Eric A. Posner and E. Glen Weyl, ‘Data as Labor’, in Radical Markets (Princeton University Press, 2018), 227.
75 Ibid., 227–28.
76 L. A. Bero et al., ‘Closing the Gap between Research and Practice: An Overview of Systematic Reviews of Interventions
to Promote the Implementation of Research Findings.’, British Medical Journal 317, no. 7156 (August 1998); Mark J.
Johnson and Carl R. May, ‘Promoting Professional Behaviour Change in Healthcare: What Interventions Work, and
Why? A Theory-Led Overview of Systematic Reviews’, BMJ Open 5, no. 9 (September 2015); Stephan U. Dombrowski et
al., ‘Interventions for Sustained Healthcare Professional Behaviour Change: A Protocol for an Overview of Reviews’,
Systematic Reviews 5 (October 2016); Bhupendrasinh F. Chauhan et al., ‘Behavior Change Interventions and Policies
Influencing Primary Healthcare Professionals’ Practice—an Overview of Reviews’, Implementation Science 12 (January
2017); Heather L. Colquhoun et al., ‘Methods for Designing Interventions to Change Healthcare Professionals’
Behaviour: A Systematic Review’, Implementation Science 12, no. 1 (March 2017).
77 Dombrowski et al., ‘Interventions for Sustained Healthcare Professional Behaviour Change: A Protocol for an Overview
of Reviews’, 2.
21
2
Figure 4 presents some of the factors which might have an impact on the value of data
from a demand-side perspective. In other words what factors might affect an
organisation’s decision to request data from NHS organisations.
Figure 4: Determinants of the value of data – the demand side 78
Category Definition
The use of data The same data could have a different value depending on what it is used for and applied
to. Information contained in clinical audits and registries can be both used for monitoring
public health outcomes or by commercial organisations for post-market surveillance. In
addition, data governance might affect what the data might be used for which might have
an impact in its value.78 There can be legislative constraints around what can be done to
DEMAND FACTORS
data.
The market capitalisation of The value of a publicly traded private sector organisation might affect its willingness to
organisations access different types of data. Organisations with lower market caps, might not be able
to afford the time and cost of accessing and cleaning data, thus affecting their willingness
to incur the costs of accessing the data.
The relative cost of getting It is important to acknowledge that many healthcare systems in the world are preparing
data elsewhere for data to be shared with private companies for fundamental and applied research
and product/service development. The relative ease and cost with which private sector
organisations will be able to access data in other countries might affect the value of the
UK knowledge asset.
2.2.3 How to measure value

HM Treasury’s recent report on intellectual property and other intangible assets in the
public sector defines public sector data – including healthcare data – as a knowledge
asset (e.g. intellectual property, software, data, technological expertise etc).79 It argues
that better management of knowledge assets is required to release the full value of public
sector data.80 Ascertaining the value of healthcare data would ensure that this ‘national
asset’ is harnessed for the benefit of society.
Knowledge assets and public sector intellectual property should be protected.81 In the
context of healthcare this should mean benefits are optimised between patients, the NHS
and industry. The healthcare system, however, does not have a history of fully harnessing
the benefits of innovations it has helped develop. The NHS played a key role in the
development of computed tomography (CT), but failed to protect this asset by securing
“meaningful financial interest” in the exploitation of that technology.82
However, it is a challenge to measure as data are an intangible asset83 and one for which
there is no clear market – although some private sector companies are entering that
space and allowing patients to be remunerated for sharing data.84 The acquisitions of
some healthcare companies have been used as indications for the financial value of NHS
data. IBM’s acquisition of Merge Healthcare “in the USA for $1 billion, which netted them
five to six million patients’ records, might be indicative of the value of the data held by the
NHS.”85 Another example would be Roche who paid $1.8 billion for Flatiron Health’s
79 HM Treasury, Getting Smart about Intellectual Property and Other Intangibles in the Public Sector: Budget 2018, 3,6.
80 Ibid., 12.
81 Ibid., 16.
82 Ibid.
Systems’, 41.
84 Medicalchain, Medicalchain Whitepaper 2.1, 2018; Datum Foundation Ltd, ‘Datum Returns Data Ownership to
Individuals’, PR Newswire, 2017; Vasilis Gkatzelis, Christina Aperjis, and Bernardo A. Huberman, ‘Pricing Private Data’,
Electronic Markets 25, no. 2 (June 2015); K. T. Pickard, ‘Exploring Markets of Data for Personal Health Information’, in
2014 IEEE International Conference on Data Mining Workshop, 2014.
22
2
health data.86 However, healthcare data from the USA is very different to that held by the
NHS as it is mostly collected for payment purposes and is very fragmented. An interview
carried out for this paper estimated that on average the value of healthcare data can be
approximately estimated at £0.45 per feature (e.g. date of birth, weight…) per individual.
Financial estimates vary widely and the usefulness of this type of ‘market value’ measure
is limited by the various determinants of value described in Figure 3 and Figure 4.
Healthcare data could be valued by considering the cost of collection and curation.87
Nevertheless, this type of valuation assumes that the “cost of data collection can be
clearly identified” 88 and would be a gross underestimation of the value that might be
generated from it. Data could be valued “based on a quantification of the earnings and
profits from their use”. 89 This is similar to the method suggested in the Treasury’s report:
“the capitalised net present value of past investment in knowledge generating activities
such as research.”90 Measuring value in this way is difficult because it is complex to
estimate what potential earnings data could generate.
In addition, there are both non-monetary and monetary aspects to value. The non-
monetary value such as increased quality of care and better health and care outcomes for
patients could be measured using traditional instruments such as Quality-Adjusted Life
Year (QALYS), 91 which is a measure of the quality and quantity of life generated by a
healthcare intervention.92
Given all these measurement pitfalls, the Treasury’s report recommends the development
of new standards and approaches for measuring and reporting the value of knowledge
assets.93 It also suggests registering intellectual property assets with the most commercial
potential so that “their value to the UK is maximised”.94
Recommendation 1
The Office for National Statistics should provide a framework and accounting standards
for measuring and reporting the value of knowledge assets such as healthcare data.
The NHS holds a record from cradle to grave of a diverse population. This is a truly unique
asset, which could give the UK a competitive advantage on the global stage.95 As
highlighted, by the House of Lords review on AI, the public are very concerned about
companies “making a profit at the expense of both the NHS and patients”.96 However,
there might be a lack of awareness amongst the public about the key role that the private
sector plays in the provision of goods and services in the NHS. A fair value exchange
should be about optimising the benefits between patients, the NHS and industry.
2.3 Expertise
An important part of the value exchange in the relationship between the public and private
sector is expertise. The NHS can provide medical expertise and the insights needed for
the research and development of data-driven technologies. It can also provide guidance
86 Roche, ‘Roche to Acquire Flatiron Health to Accelerate Industry-Wide Development and Delivery of Breakthrough
Medicines for Patients with Cancer’, Press Release, n.d.
88 Ibid.
89 Ibid.
91 HM Treasury, The Green Book, Central Government Guidance on Appraisal and Evaluation, 2018, 72.
92 National Institute for Health and Care Excellence, ‘Glossary’, Webpage, NICE, n.d.
93 HM Treasury, Getting Smart about Intellectual Property and Other Intangibles in the Public Sector: Budget 2018, 30–31.
94 Ibid.
95 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, Report of Session 2017–19 (House of
Lords, 2018) p.88; Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector; ‘A
Revolution in Health Care Is Coming’, p.50, The Economist, 1 February 2018;. Aldrick, ‘Data Could Be a Huge Source of
Funding for the NHS and We Are about to Give It Away’.
23
2
on how to make sense of data that is not primarily collected for the purposes of the
research and development of a data-driven technology. In turn, the private sector lends its
expertise in fields such as data engineering, artificial intelligence or user centred design to
create data-driven technologies. It can also lend its experience and capability in taking a
new product or service through development stages, the required regulation,
accreditation and standards process (see 3rd arrow at the bottom of Figure 1).
Matt Hancock’s vision for the future of the NHS stressed the importance of skills. It raised
the need “to recruit and retain” medical and non-clinical professions.97 Nevertheless, as
highlighted by a few interviews and the research events carried out for this paper, the
private sector is acquiring medical expertise by attracting the NHS workforce with higher
salaries and has been attracting data scientists for a long time. Putting the NHS’s financial
constraints aside, it would impossible for the healthcare system to compete with the
salaries in the private sector. The question of staff retention thus will be a crucial one to
solve.
2.4 Wider societal impact

Another important element to consider when thinking about the value exchange is the
wider societal benefits. Interviews carried out for this paper argued that there is a ‘tacit’
understanding that when a private sector organisation engages in a partnership with the
NHS there is a wider societal benefit expressed through the creation of employment, the
payment of taxes and the generations of new technologies and solutions for the market.
There is potentially a need to make a distinction between organisations who are
headquartered and pay taxes in the UK and those overseas. It cannot be forgotten that the
race for data-driven technologies is a global one.98 Having headquarters overseas might
allow certain companies to shift profits and not pay their fair share of taxes in the UK.
DeepMind’s former independent Review Board had called upon the company to clarify its
relationship with its parent company Alphabet partly for this purpose.99 The company pays
its taxes in the UK, however, now that the Streams app (see Figure 12) they developed is
moving to Google,100 the revenue streams generated by this product might not be
captured by the UK economy. The Chancellor announced in his 2018 budget a
consultation about a UK Digital Services Tax which would seek to target “UK-generated
revenues of specific digital platform business models” to make sure that “global giants,
with profitable businesses in the UK, pay their fair share towards supporting our public
services”.101
97 Department of Health and Social Care, The Future of Healthcare.

98 Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector; Ed Fitzgerald,
‘International Lessons to Help the NHS Harness Health Data’, The Reformer Blog, 17 July 2018.
99 Martin Bromiley OBE et al., DeepMind Health Independent Review Panel Annual Report, 2018.
100 DeepMind Health, ‘Scaling Streams with Google’, Website, DeepMind, 13 November 2018; Jee, ‘Google’s Decision to
Absorb DeepMind’s Health Division Has Sparked Privacy Fears’.
101 Rt Hon Philip Hammond MP, ‘Budget 2018: Philip Hammond’s Speech’, Speech, HM Treasury, 29 October 2018.
24
3
Keeping society
in the loop
3.1 Currently not in the loop 26

3.2 Engagement by design 28
25
3
Making NHS data work for everyone / Keeping society in the loop
MIT Professor Iyad Rahman has defined the concept of ‘keeping society in the loop’ as
enabling “society to influence and shape decisions about technology and innovation”.102
Although his concept specifically applies to the design of AI systems,103 it is a valuable
concept to explore when looking at the types of partnerships people would deem
acceptable between the NHS and private sector companies in the data-driven
technologies and services industries. Patients are ultimately the producers of that data104
and will also be on the receiving end as consumers of the products or services being
designed.105 Private sector and NHS organisations need to work on earning patients’
trust.106 Patients and the public need to feel comfortable and able to trust that data is not
being exploited107 and that their privacy is safeguarded.108 Building a trustworthy system
will be key to ensuring patients, the NHS and industry can work collaboratively when
using data for research or product and service development purposes.
3.1 Currently not in the loop

While patients have the ability to object to data about them being used beyond direct
care,109 at present patients are not actively involved in partnership discussions. They are
given the opportunity to opt-out of data about them being accessed for purposes beyond
direct care. In some instances, patients might be asked for their explicit consent when
identifiable data about them is being used by NHS organisations, universities or
commercial companies for secondary uses. These are the only institutional opportunities
for patients to have a ‘say’. The model is rather limited by only giving patients a binary
option.110 This might not be a sufficient way of engaging on complex issues as there are a
number of nuances to patient and public opinion about data uses and in particular
commercial access (see Figure 5). These include ‘who, what, why, when and how’,111
indicating that patient opinion will vary from partnership to partnership.
Figure 5: Public opinion

A 2016 study by Ipsos Mori and Wellcome Trust found that people were less inclined to
support commercial access to healthcare data, in particular in marketing and insurance
industries.112 However, people were generally in favour of commercial access if this was
necessary for health benefit.113 Sixty-one per cent of people in the study would rather
commercial companies access data than miss out on research benefits,114 while 25 per
cent did not want commercial organisations to access their data under any
circumstances.115 It also found that views varied depending on the interaction the
102 Iyad Rahwan, ‘Society-in-the-Loop’, MIT MEDIA LAB (blog), 12 August 2016.
103 Iyad Rahwan, ‘Society-in-the-Loop: Programming the Algorithmic Social Contract’, Ethics and Information Technology
20, no. 1 (March 2018).
104 Imanol Arrieta Ibarra et al., ‘Should We Treat Data as Labor? Moving Beyond “Free”’, American Economic Association
Papers and Proceedings 1, no. 1 (December 2017); Posner and Weyl, ‘Data as Labor’.
105 Department of Health and Social Care, ‘New Guidance to Help NHS Patients Benefit from Digital Technology’, GOV.UK,
n.d.; Kaela Scott, Data for Public Benefit: Balancing the Risks and Benefits of Data Sharing (Involve, The Carnegie UK
Trust’s, Understanding Patient Data, 2018); Accenture, Digital Consumer Health Engagement 2016 – Global Report,
2016.
106 Timmis, Heselwood, and Harwich, Sharing the Benefits: How to Use Data Effectively in the Public Sector; Scott, Data for
Public Benefit: Balancing the Risks and Benefits of Data Sharing; The Caldicott Committee, Report on the Review of
Patient-Identifiable Information (Department of Health, 1997).
107 Nick Couldry and Ulises A. Mejias, ‘Data Colonialism: Rethinking Big Data’s Relation to the Contemporary Subject’,
Television & New Media, September 2018.
108 Alan F. T. Winfield and Marina Jirotka, ‘Ethical Governance Is Essential to Building Trust in Robotics and Artificial
Intelligence Systems’, Phil. Trans. R. Soc. A 376, no. 2133 (November 2018); Graeme Laurie and Emily Postan, ‘Rhetoric
or Reality: What Is the Legal Status of the Consent Form in Health-Related Research?’, Medical Law Review 21, no. 3
(September 2013).
109 National Data Guardian for Health and Care, Review of Data Security, Consent and Opt-Outs.
110 Isabelle Budin-Ljøsne et al., ‘Dynamic Consent: A Potential Solution to Some of the Challenges of Modern Biomedical
Research’, BMC Medical Ethics 18, no. 1 (January 2017); Hawys Williams et al., ‘Dynamic Consent: A Possible Solution
to Improve Patient Confidence and Trust in How Electronic Patient Records Are Used in Medical Research’, JMIR
Medical Informatics 3, no. 1 (13 January 2015).
111 Wellcome Trust, The One-Way Mirror: Public Attitudes to Commercial Access to Health Data.
112 Ibid., 91.
113 Ibid., 97.
114 Ibid., 15.
115 Ibid.
26
3
individuals had with the NHS. In general, those who interacted more with the NHS were
more likely to advocate for data sharing.116 Although, those with severe conditions that are
easily identifiable were also worried about the ramifications of sharing data about them.117
Acceptability
Why Mix of public and Solely for

Clear public benefit
private benefit private benefit
Who For profit but in No link to

Clear public outcomes improving
health sector
public benefit
What Aggregate passively Aggregate but risk Identifiable

collected reidentification personal details
How Secure storage & regulation is assumed
Source: What drives acceptability: in summary, adapted from Ipsos MORI and Wellcome,
The One-way Mirror, 2016.
Public buy-in will be essential to forming partnerships.118 The care.data programme

designed to allow anonymised primary care health records to be shared outside the NHS,
is often cited as an example of how policy can go wrong, demonstrating a clear moment
where public opinion overturned policy.119 The policy was controversial among the public,
who was deeply unsettled by the idea that data about them would be commercialised.120
The policy’s failure was closely related to a lack of communication and transparency121 as
well as a lack of public confidence in institutions to build fair commercial models.
As shown in Figure 5, on average the public is generally comfortable with commercial
access to data if there is an undeniable public benefit. However, when presented with
various scenarios levels of acceptability vary.122 An interview carried out for this paper
described that most commercial partnerships will be along the amber line – partnerships
will never be solely for private profit or solely for public benefit – they will be a mix of the
two, further demonstrating the need for patients and the public to engage on this issue.
117 Ibid., 74.
118 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able; Lord Freyberg, ‘NHS: Healthcare Data
– House of Lords Debate Hansard’, 2018; Pam Carter, Graeme T. Laurie, and Mary Dixon-Woods, ‘The Social Licence
for Research: Why Care.Data Ran into Trouble’, Journal of Medical Ethics, (January 2015); National Data Guardian for
Health and Care, Review of Data Security, Consent and Opt-Outs.
119 Budin-Ljøsne et al., ‘Dynamic Consent’; Carter, Laurie, and Dixon-Woods, ‘The Social Licence for Research: Why Care.
Data Ran into Trouble’; Roski, Bo-Linn, and Andrews, ‘Creating Value In Health Care Through Big Data: Opportunities
And Policy Implications’.
120 Carter, Laurie, and Dixon-Woods, ‘The Social Licence for Research: Why Care.Data Ran into Trouble’; Colin Marrs,
‘Care.Data – An In-Depth Check-up on NHS England’s Controversial Bid to Join up Health Data’, Civil Service World, 18
September 2015; Paraskevas Vezyridis and Stephen Timmons, ‘Understanding the Care.Data Conundrum: New
Information Flows for Economic Growth’, Big Data & Society 4, no. 1 (June 2017).
121 Budin-Ljøsne et al., ‘Dynamic Consent’; Carter, Laurie, and Dixon-Woods, ‘The Social Licence for Research: Why Care.
Data Ran into Trouble’; Roski, Bo-Linn, and Andrews, ‘Creating Value In Health Care Through Big Data: Opportunities
And Policy Implications’.
122 Wellcome Trust, The One-Way Mirror: Public Attitudes to Commercial Access to Health Data, 9–13.
27
3
3.2 Engagement by design

It is essential that society is kept in the loop and feels involved in the data-driven
innovation landscape. This goes further than patients engaging in what data are used and
by whom, but also involves the partnership and commercial models that exist. Reform
research suggests that there have been limited attempts to engage the public over this.
As Nicola Perrin, then lead of the Understanding Patient Data initiative, commented in the
House of Lord’s report, the public “do not like the idea of the NHS selling data, but they
are even more concerned if companies are making a profit at the expense of both the
NHS and patients”.123 This means patients might potentially be less in favour of models
such as ‘cost-recovery’ (see Figure 7) that do not achieve financial value back. The Life
Sciences Industrial Strategy highlighted that a “strong patient and clinician engagement
and involvement, alongside clear permissions and controls, are vital to the success of any
health data initiative.”124
To build a trustworthy system, where patients and the public are engaged, academics
such as Mary Dixon-Woods have argued actors, either processing or controlling health
data, would need a social licence.125 This is different from a legal licence, as it ensures the
data processor (see Glossary) or controller consistently acts according to socially defined
ethics. A social licence, she argues, would be achieved by acting with ‘reciprocity’ and
‘fairness’.126 The former, because the public receives a return from the use of health data
(e.g. access to better diagnostic tools) and the latter because the return is equally enjoyed
by the public. In order to understand what a fair return might be from the patients’ and
public’s perspectives, private sector and NHS organisations should engage with them to
learn about their views. A critical part of this is ensuring that the conversation between
patients and private sector stakeholders is continuous. In a field as dynamic as data-
driven innovation in the NHS — with constant new patients, new technological potential
and a variety of external players, including other government departments127—it is
essential that models of engagement are more than a one-off occurrence.128
There are various methods for ensuring the continuous nature of the dialogue between
patients, the NHS and industry. Connected Health Cities, in association with the National
Data Guardian, have used Citizens’ Juries to ask “to what extent should patients control
access to patient records for secondary use of data”.129 They found jurors changed their
mind over the three days, becoming more positive towards data sharing for public health
benefits.130 The benefits of this model is that it gives time for patients to form an opinion
and have direct contact with experts in the field, while those holding the research can take
on patients’ opinion and feed it back to governing institutions. To reach more rural areas
of the country, such as Cumbria, ‘Chatty Vans’ have been deployed to explore people’s
views on various issues to do with health and care.131 While both examples have been
used to engage the public over the governance of health data, these could be more
specifically tailored towards commercial models.
There are other ways to platform better governance of data. Until recently, DeepMind
Health had an independent review panel, which brought together a group of experts and
patient voices to scrutinise the company’s behaviour and publish a statement.132 Part of
this involved assessing the company’s approach to engaging patients, as well as
125 Carter, Laurie, and Dixon-Woods, ‘The Social Licence for Research: Why Care.Data Ran into Trouble’.
126 Ibid.
127 Timmis, Heselwood, and Harwich, Sharing the Benefits: How to Use Data Effectively in the Public Sector.
128 Carter, Laurie, and Dixon-Woods, ‘The Social Licence for Research: Why Care.Data Ran into Trouble’; Sylvie Delacroix,
‘Pervasive Data Profiling, Moral Equality and Civic Responsibility’, UCL Laws and UCL Computer Science, 17 August
2017; Williams et al., ‘Dynamic Consent: A Possible Solution to Improve Patient Confidence and Trust in How Electronic
Patient Records Are Used in Medical Research’.
129 Malcolm Oswald et al., ‘Health Data on Public Trial: To What Extent Should Patients Control Access to Patient
Records?’, International Journal of Population Data Science 1, no. 1 (April 2017).
130 Ibid.
131 HealthWatch Cumbria, Annual Report 2018, 2018.
132 DeepMind Health, ‘Independent Reviewers’, Webpage, DeepMind, 2018.
28
3
publishing reports with principles for corporate action.133 However, this review panel has
been dismantled since DeepMind transferred the “control of its health subsidiary to its
parent company Google”.134
Proportionate governance offers another alternative to meaningfully engage patients and
the public (see Figure 6). This system-wide change, which is government led, could
include practical ways of assessing and creating bespoke agreements that use public
engagement.
Figure 6: Proportionate governance

In Scotland, the Public Benefit and Privacy Panel for Health and Social Care has been set
up to scrutinise applications for access to NHS Scotland data that go beyond direct
care.135 The idea is to construct a governance model that can be adaptive to each
individual research and development application.136 Similar to the Independent Group
Advising on the Release of Data, which considers external requests for data held by NHS
organisations, one of its pivotal assessment criteria for project proposals is whether it has
demonstrable public benefit.137 The model follows a three-tiered system based on risk. If
commercial organisations are involved, the application will immediately go to the second
tier where the application is reviewed by a panel of senior experts and Caldicot
representatives. If an agreement is not reached about the public benefit, the decision is
taken to a third panel made up of members of the public.
As highlighted in the British Academy’s work in partnership with the Royal Society on data
governance for the 21st century, access to data as well the governance structures “must
consider who reaps the most benefit from capturing, analysing and acting on different
types of data”.138 Proportionate governance could provide a framework for engaging with
the public and keeping ‘society in the loop’ on the types of commercial models that would
be deemed fair when there is access to data for research and development or for product
or service development. Patient voice and opinion would be included in the governance
model by design. This would ensure that commercial agreements are co-produced
between industry, health experts, public sector leaders and public representatives through
the review process.
Recommendation 2
The Department of Health and Social Care should in conjunction with Caldicott
Guardians, NHS organisations and industry representatives ensure that a dialogue with
the public is set at a local level to discuss commercial models. Proportionate governance
models provide an interesting avenue for public and patient engagement that could be
explored.
133 Bromiley OBE et al., DeepMind Health Independent Review Panel Annual Report.
134 Murgia, ‘DeepMind’s Move to Transfer Health Unit to Google Stirs Data Fears’.
135 Information Assurance and Governance Team, ‘About the Panel Public Benefit and Privacy Panel for Health and Social
Care’, Webpage, Information Assurance and Governance Team, 2015.
136 Nayha Sethi and Graeme T. Laurie, ‘Delivering Proportionate Governance in the Era of EHealth’, Medical Law
International 13, no. 2–3 (June 2013).
137 NHS Scotland, Public Benefit and Privacy Panel for Health and Social Care: Guiding Principles and Policy for Decision
Making, 1st ed., 2015.
138 British Academy and Royal Society, Data Management and Use: Governance in the 21st Century, 2017, 27.
29
4
The healthtech
ecosystem
4.1 Letting a thousand flowers bloom? 31

4.2 The need for a national strategy 33
4.3 Access to good quality data 36
4.3.1 Digital innovation hubs 38
4.3.2 Synthetic data 39
4.4 Procurement, partnership, or somewhere in between? 40
4.5 Commercial and legal skills 42
30
To create a strong healthtech ecosystem in which a “tech company can…have an equal
opportunity to deliver”,139 whilst safeguarding the NHS’s constitution140 and patients’ trust,
strong guidance and leadership will be needed to make sure the value of healthcare data
is optimised. Efforts have been made by the DoHSC to develop a code of conduct to
ensure that partnerships “deliver benefits to patients, clinicians, industry and the health
and care system as a whole”.141 The DoHSC is also conducting a formal review to assess
“commercial models used in technology partnerships”.142 However, prior to the code of
conduct there was no guidance on what partnerships should look like when there is
access to data to create a product or service. This has led to a lack of clarity as to what is
happening on the ground and a lack of strategic understanding at a national level about
the consequences of the current commercial models developing on the ground.
4.1 Letting a thousand flowers bloom?

Research carried out for this report suggests that there is a wide variety of arrangements
between the NHS and industry when there is access to data for research and product or
service development purposes.143 However, it is difficult to gain an understanding of what
the national picture is as there is no national registry describing the various types of
commercial models that have developed on the ground. There are registries at Trust-level
documenting data access requests, but these are generally not made available online.
Figure 7 presents the different models that can be currently found on the ground – more
details can be found in Appendix.
Figure 7: Current models 144
Data sharing Grant-funded Licencing Cost-recovery Commercial

agreements collaboration arrangements
Various Grants can be An external This model often Commercial
organisations allocated either organisation is operates using arrangements
including private by academic or given or pays a licence. In this include any kind
sector companies public sector a fee to hold a case the data of arrangement
can access NHS organisations for licence to use controller cleans where there is
data for a specific specific purposes or access data. the data and an exchange
purpose with to other In most cases, provides services of assets
patient-consent organisations this fee will in exchange or valuable
for secondary including the cover the cost of for a fee to resources. For
uses or without private sector maintaining the organisations example, access
patient consent as long as they data. The licence hoping to do to data might
for direct care. can demonstrate might also research or be granted to
Profits generated evidence of stipulate privacy create products an external
from the innovative restrictions or services. There organisation
secondary uses potential.144 and ethical are variations in for research
of that data are regulations. cost recovery or commercial
retained by the Profits generated models. purposes and
private sector. from the use in return the
of that data is NHS might get
retained by the a cleaned higher
private sector. quality dataset
back.
Source: Reform research and interviews.

140 Department of Health and Social Care, The NHS Constitution for England, 2015.
142 Ibid.
143 Alexander J Martin, ‘NHS Patients’ Data Was Illegally Transferred to Google DeepMind’, Webpage, Sky News, 3 July
2017; Sarah Knapton, ‘How the NHS Got It so Wrong with Care.Data’, The Telegraph, 7 July 2016.
144 Innovate UK and UK Research and Innovation, ‘Business Innovation: What Funding You Can Get and How to Apply’,
GOV.UK, 16 August 2018; Vicki Sellick, Solder Amy, and Rosalinda Baker, Funding for Innovation (Nesta, 2018).
31
4
Making NHS data work for everyone / The healthtech ecosystem
The NHS is not a homogenous body,145 it is a highly fragmented system and, as a result,
each Trust has its own protocol for partnerships.146 Partnerships might differ as some
products or services require different types of data. For example, products or services
that support the organisation and management of hospitals may need real-time access to
admin data.147 Other products or services that seek to improve direct patient care might
need access to patient records. Relationships between private sector companies in the
data-driven technology space and NHS organisations can be mostly characterised by
data sharing agreements with no definition of a commercial model involving the NHS and/
or patients (see Figure 7).
There are, however, examples of datasets held by central NHS organisations that can
share data at a national level, with datasets that cover the entire population. For example,
NHS Digital has 12 datasets currently available for external research.148 However, some
interviews carried out for this paper have suggested that most access requests for
datasets held by NHS Digital have been from other NHS Trusts and academic institutions.
In addition, there are also independent data repositories which have been established by
a variety of non-profit actors and funded by research, charity or council funding (see
Figure 8 and Appendix).
Figure 8: UK Biobank
The UK Biobank is a registered charity that was set up in 2006 by the Wellcome Trust, the
Medical Research Council, the Department of Health and Social Care, the Scottish
Government and the Northwest Regional Development Agency, with the aim to be a
major national and international health resource for improving the diagnosis, treatment
and prevention of diseases such as cancer, heart disease, depression and forms of
dementia.149 The data collected thus far has come from 500,000 volunteers between the
ages of 40-69 years and is accessible to any organisation that can prove a research need.
The only contingency is that the research must be made publicly available.
There are benefits to an ‘open access’ model as it offers ‘value back’ to society through
delivering positive externalities such as furthering knowledge and innovation.150 It increases
the opportunities to conduct academic research as well as research and development, thus
potentially leading to better health outcomes and economic growth. The model is based on
a “notion of altruistic donation and the notion that biobanks serve the scientific and public
good”.151 This model is not cost-free and requires finance from often a variety of actors from
government, third sector donors and corporate donors. The UK Biobank receives finance
from government and the Wellcome Trust. In addition, the cost of the cleaning up of UK
Biobank data for pharmaceutical research, for example, has required a number of corporate
sponsors, who will get exclusive rights to the data for the first two years – see Appendix.
However, as highlighted in the Treasury’s discussion paper on the economic value of data,
this ‘open access model’ may not be desirable in every instance.152
The only example known to date of a partnership between the NHS and industry offering
a form of direct ‘financial’ value back to NHS Trusts for product development purposes is
the partnership between Sensyne Health, the University of Oxford and the Oxford
University Hospitals NHS Foundation Trust, South Warwickshire NHS Foundation Trust,
and Chelsea & Westminster NHS Foundation Trust (see Figure 9).
145 Wachter, Making IT Work: Harnessing the Power of Health Information Technology to Improve Care in England.
147 Ibid., 50; Edge Health, ‘Improving Theatre Productivity with Machine Learning – “Space Finder”’, Webpage, Edge
Health, 2017; Edge Health, ‘Surgical Pathway Demand and Capacity Planning’, Webpage, Edge Health, July 2017.
148 NHS Digital, ‘Data Sets’, Webpage, NHS Digital, n.d.
149 UK Biobank, ‘About’, Webpage, UK Biobank, n.d.
150 Andrew Turner, Clara Dallaire-Fortier, and Madeleine J. Murtagh, ‘Biobank Economics and the “Commercialization
Problem”’, Spontaneous Generations: A Journal for the History and Philosophy of Science 7, no. 1 (2013).
151 Ibid., 72.
152 HM Treasury, ‘The Economic Value of Data: Discussion Paper’, August 2018, 11.
32
4
Figure 9: Sensyne health

Sensyne Health partnered with three separate NHS Trusts – Oxford University Hospitals
NHS Foundation Trust, South Warwickshire NHS Foundation Trust, and Chelsea &
Westminster NHS Foundation Trust – which have equity shares of 6 per cent, 3.7 per cent
and 3.7 per cent respectively.153 NHS Trusts, however, will never own above 10 per cent
collectively,154 so each time an NHS Trust joins, the other Trusts’ shares reduce in size.
Alongside this, each Trust is offered a “double bottom-line return” which is £5 million
worth of shares and a royalty on any product developed using its data.155 In 2018, the
company was floated on the London Stock Exchange and has since raised £60 million,
according to the Oxford Academic Health Science Network (AHSN) review.156
The company uses health data to create a variety of health-tech products; from “validated
software applications powered by artificial intelligence including prescribed digital
therapeutics [to] hospital systems for clinical care”.157 One example is the GDm-Health
App that helps manage gestational diabetes at home. Thus far, the product has achieved
a 25 per cent reduction in clinic visits and a 50 per cent reduction in time spent by the
diabetes midwives on clerical administrative tasks. The App uses NHS data as well as
collecting more data for future research.
The Sensyne example is commendable for sharing equity with the Trusts involved,
meaning the NHS Trusts have a stake in the company and the potential to benefit
financially.158 Nonetheless, it is important to consider the macroeconomic effect of such
partnerships.
4.2 The need for a national strategy

There is a risk that if all partnerships operate at Trust-level only the Trusts in question will
benefit, leaving other Trusts to miss out on products or services and potentially monetary
benefits.159 In the long run this can create imbalances in the healthcare system as a whole,
contradicting the principle that the NHS is meant to “maximise [its] resources for the
benefit of the whole community, and make sure nobody is excluded”.160 This situation
echoes the one explored in the Naylor Review: that Trusts built on more expensive land
have higher value capital assets leading to inequality between NHS Trusts.161 With the
potential for the value of NHS data assets to be similarly variable between Trusts –
because of varying quality of data, for example – there is a risk that keeping value at a
local level could increase inequality. As shown in Figure 10, which presents a typology of
commercial models, there are many options that could be explored with varying
advantages and disadvantages.
153 London Stock Exchange, ‘Sch 1 – Sensyne Health Limited – RNS -’, Webpage, London Stock Exchange, 3 August 2018.
154 Clive Cookson, ‘Drayson Floats Medical AI Group Sensyne Health on Aim’, Financial Times, 14 August 2018.
155 Ibid.
156 Oxford Academic Health Science Network, Oxford AHSN Year 6 Q2 Report (Oxford Academic Health Science Network,
2018).
157 ‘Sensyne Health: AI Powered Clinical Solutions’, accessed 6 August 2018.
158 London Stock Exchange, ‘Sch 1 – Sensyne Health Limited – RNS -’.
159 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, 93–94.
160 Department of Health and Social Care, The NHS Constitution for England.
161 Sir Robert Naylor, NHS Property and Estates, 2017.
33
4
Figure 10: Typology of commercial models 162 163 164
Name Description Brief evaluation

Consortium This would be a joint venture The company and the NHS organisation will be the recipients of financial
agreements (contractual or incorporated) benefits through ownership of the company’s shares. The amount will
between an NHS organisation, depend on the terms of the profit share and the success of the consortium.
industry, research institutes and/or If an NHS organisation has a say in the consortium’s activities this may
academic institutes to collaborate be very valuable. However, there is a risk at a national level of creating
as partners to develop and inequalities between Trusts as the NHS organisation involved in the
commercialise a product. company will be the only one receiving financial benefits.
Feasibility: There could be various joint venture entities in which the NHS
participates and, as part of which, it would need to contribute its own
offering (e.g. funding, IP, know-how) to the consortium and engage in the
operation and running of the consortium. This would likely mean a radical
change in the way the NHS currently operates. Managing this could also
require a significant amount of NHS manpower, which might be unfeasible
given the NHS mandate.
Equity shares An NHS organisation holds a The company will be the main recipient of the financial benefits from any
minority stake in the company dividend and capital appreciation in the value of the shares crystallised
it collaborates with. The NHS on a sale of the company. The size of that benefit will depend on the size
organisation holding the shares of equity share and the success of the products or services. The ability to
would have the right to receive a negotiate preferential arrangements for the provision of goods/services to
dividend/capital appreciation and equity investors may also mean better (or cheaper) access to products/
to vote at shareholder meetings. services. Individual NHS Trusts who have equity shares will be the only
ones that receive financial benefits and there is therefore a risk of creating
financial inequalities at a national level.
Feasibility: This model will be most appropriate for a company that is
young or relatively new. A larger or more established company might
struggle to convince existing shareholders to adopt this model.
Grant funding Funding from the academic There are various models within grant funding, some of which might have
models or public sectors is a well- conditions that stipulate an equity stake for a government department,
established method for research independent body or third sector organisation or discounts on products or
and development (R&D).162 This services. The recipient of the grant will need to prove that their research,
model can be used for data- product or service has clear societal benefits. Many grants such as those
driven innovation and might from the Industrial Strategy Challenge Fund (ISCF) will also give finance
involve competition for funding directly to the NHS or to an SME for them to build the necessary data
opportunities and include infrastructure.
commercial conditions such as
Feasibility: The model has been used in the life sciences space for a long
royalties or equity stakes.
time. By providing upfront funding, this helps platform R&D and enables
organisations to take the risks needed to build data-driven products and
services. However, it is unrealistic to expect grants to be the only way
data-driven products and services are created. The donor may be a central
government entity (as opposed to the NHS or a Trust) and, as a result, the
NHS would not directly receive the financial benefit.
Golden share In a typical scenario a public The NHS organisation – this may be NHS Digital or an individual Trust – that
sector organisation owns 51 per gets the ultimate say will receive a lot of the benefits,163 e.g. profit share,
cent of shares giving it majority control over the direction of the business and its products. The other
voting rights, or a much smaller shareholders in the company will also receive financial benefits.
percentage of shares but still has
Feasibility: Despite being suggested in the Life and Science Industrial
veto or majority voting rights over
Strategy,164 this model may be unattractive for commercial organisations
certain types of decisions. This is
who would not want the NHS to own the majority of the company and/
the key defining feature of golden
or have veto over every decision made. This model could be perceived as
share agreements.
restricting industry from being able to pursue innovation and its business as
it sees fit.
162 Sellick, Solder, and Baker, Funding for Innovation, 5.

164 Ibid.
34
4
Name Description Brief evaluation

Intellectual IP generated from using the Financial benefits could be generated from any foreground IP – this is the
property NHS data would be owned IP generated in collaboration during a partnership – that is produced by a
sharing (IP) (entirely or in part) by the NHS partnership with the health sector. This would allow the NHS and the private
organisation entering into the sector organisation to own parts of the IP.
data sharing agreement. Such
Feasibility: In order to own (in whole or part) IP generated using NHS data,
NHS organisations would then
it is likely that the NHS would need to contribute more than just raw data.
license this IP to the third-party
This would need to be provided alongside know-how/clinical expertise
collaborator in return for payment
to add value to justify IP ownership. Equally joint ownership of IP may be
(e.g. one-off fees, milestones and/
resisted on the basis that it can be legally quite complex.
or royalties).
Open-access In this model, the results from Indirect financial benefits would be created through a thriving market for
the research produced from innovation. Some interviewees have argued that this model could generate
healthcare data have to be made a lot of value to the wider economy.
‘open access’ or aggregate
Feasibility: Making the results from research transparent is key to building
healthcare data could be made
good evidence. However, this model does not allow for information to
openly available (see Glossary).
remain proprietary, which might disincentivise the private sector. There
is a limit on the type of data that could made open access because of
compliance with regulation which means that the benefits of an open data
model in healthcare might be limited.
Profit-sharing/ In return for providing data, the The NHS organisation will receive financial benefits from whatever
revenue- NHS organisation receives a product(s) or services reach market – the exact apportioning of the profit
sharing share of the profits of any product or revenue share might vary. Financial benefits will depend on whether the
developed using the data (without company is a UK-based tax paying company.
any ownership interest in the
Feasibility: This is dependent on the share the NHS organisation demands
company or the IP in the product).
and what the NHS’s contribution is to justify such a share. It should be
This would provide an alternative
recognised that multiple arrangements of this nature, each with small
fee structure to a one-off up-front
percentage shares may generate significant revenue.
fee.
Royalty An NHS organisation and a private Direct financial benefits from whatever products or services reach market
payment sector organisation negotiate and there is demand for– the exact apportioning of the royalty might vary.
a right to a fixed percentage This will not capture the value of any ‘background IP’ – any IP owned
of profits from a specific asset before the partnership. Financial benefits will depend on whether the
or a fixed price per unit that is company is a UK-based tax paying company. In practice there is a risk that
sold. There can be other ways of one NHS Trust could be receiving royalties for another Trust purchasing the
licencing royalties. product or service.
Feasibility: This is dependent on what royalties the NHS organisation
demands as it would need to be realistic for it to be feasible. It ensures that
a company does not pay any up-front costs for data.
Local Asset- This is a joint commercial venture The NHS organisation and its corporate partner will benefit 50-50. Financial
backed between NHS organisations benefits will also be received as the venture is commercial and would be
vehicles and a private sector partner. In dependent on how the venture is structured, but there is high potential for
(LABV) this model, the public sector NHS organisations to profit both from new health innovation and in financial
organisation provides an asset and teams.
the private partner matches the
Feasibility: There are no examples of data being used as an asset in this
contribution in liquid assets, skills
way. As discussed in Chapter 2, valuing NHS data is a very complicated
and experience (generally there is
process and might make LABV difficult to implement.
a 50-50 apportioning).
It is crucial for national bodies to ensure that local bodies are not creating perverse
incentives in the system. A commercial model might seem fair at GP practice or Trust-
level, but on aggregate could be detrimental. Concerns have been raised about these
types of perverse incentives already happening on the ground with certain healthtech
partnerships.165 The code of conduct argues that the “benefits of the partnerships
between technology companies and health and care providers [should be] shared
fairly”.166 However, there seems to be a lack of clarity about whether fairness should be
considered at a micro or macro scale, or both.
165 Bhatti, ‘Seeing a GP on a Smartphone Sounds Wonderful – but It’s Not’.

35
4
The lack of guidance from central government on the type of appropriate commercial
arrangements in the data-driven space has also been found to be a key challenge by
parts of the private sector. In a recent Academic Health Science Network survey about AI
technologies in health and care, 34 per cent of respondents said the lack of clarity over
‘appropriate business models for AI development and deployment’ was a key factor
affecting AI’s potential.167
All models have advantages and disadvantages, demonstrated by Figure 10, meaning
different models will suit different scenarios. It is therefore impractical to recommend one
model for all circumstances. Nevertheless, it is crucial that national policy provides a
framework for the array of possible models that will not have an adverse effect on the
national level.
Recommendation 3
NHS England and NHS Digital should create a register of data sharing agreements
between the NHS and commercial organisations. It should include what type of data are
being shared and a description of the type of partnership model being used. This would
allow for clear understanding of what is happening on the ground and facilitate public
scrutiny.
Recommendation 4
The Department for Health and Social Care in conjunction with the Crown Commercial
Service, Office for Life Sciences and HM Treasury should make sure to include in its formal
review of commercial partnerships, a macroeconomic study of the impact that different
partnership models might have to avoid reinforcing a postcode lottery or other negative
externalities such as exclusive data access.
Recommendation 5
The Department of Health and Social Care should include the results of this
macroeconomic study in a clear national strategy which should seek to optimise the value
of data held by NHS organisations when it is accessed for commercial purposes. It should
consult with all stakeholders including industry, patients and NHS organisations as to
what a fair apportioning of value might be.
4.3 Access to good quality data

Access to good quality data that is fair and for clear public benefit will be key to creating a
competitive life and sciences industry that is both innovative and dynamic.168 It is also
crucial for different sized companies operating in the life sciences industry to have a level
playing field to “compete and have an equal opportunity to deliver”.169 However, as
highlighted in Chapters 1 and 2, data within NHS organisations are currently fragmented
and of varying quality.
Most data held by NHS organisations is collected for the purposes of direct patient care
and not for research or product or service development purposes. When using this data
for purposes beyond direct care it is essential to look at how it was generated to
understand how it should be cleaned and avoid “using inappropriate data to make
decisions”.170 Evidence shows that using data from electronic health records “without
consideration to context, can easily lead to biases or nonsensical findings, making it
167 Melissa Ream et al., Accelerating Artificial Intelligence in Health and Care: Results from a State of the Nation Survey
(The AHSN Network, Department of Health and Social Care and NHS England, 2018), 32.
168 HM Treasury, ‘The Economic Value of Data: Discussion Paper’.
170 British Academy and Royal Society, Connecting Debates on the Governance of Data and Its Uses, 2017.
36
4
unsuitable for many research questions”.171 Data curation is thus an inevitable part of the
process for secondary uses of NHS data.
Nonetheless, there is a strong case to be made for increasing data quality to provide better
care for patients172 and potentially provide a more solid basis of data quality for secondary
uses. This can be done in various ways such as minimising error at the data collection
point173 and embedding data quality by design.174 The design of electronic health record
systems can have an impact on the quality of data collected.175 IT systems used for data
collection could flag for potential mistakes when data are being entered. Similarly, systems
that have a greater focus on data visualisation “can reveal data quality problems”,176 which
can then be corrected. The use of AI to translate text to clinical code could be explored to
reduce the burden on clinical coders.177
Once data are collected the monitoring of data quality is crucial.178 NHS Digital has a Data
Quality Maturity Index “quarterly publication about data quality in the NHS”.179 It provides
submitters with transparent information about the quality of the data they collect and does
so for specific datasets. Understanding the state of data quality at the early stages of
partnership discussions with commercial organisations is crucial. This will enable
organisations to understand what data engineering capabilities they will need in order to
get the data into an acceptable state for analysis.
Recommendation 6
NHS England should create a ‘Data Quality Service’, with a tiered-fee system dependent
on factors such as company size and global profits, to provide bespoke reports on data
quality at the early stages of a partnership discussion between the NHS and industry.
In addition, the lack of interoperability of IT systems within the NHS creates a situation in
which data are locked into multiple fragmented systems. This can have a negative impact
on direct patient care as records cannot always be rapidly shared in a digital format. In
addition, the difficulty in linking datasets together can put brakes on innovation and limits
patient benefits. NHS Trusts in the North East and North Cumbria, the Darlington Borough
Council, and GP practices under the umbrella of the Connected Health Cities programme
have decided to collaborate in a research project to tackle this problem. They have
managed to link-up data from all the stakeholders and provide it in an anonymised and
pseudo-anonymised form to researchers to develop evidence on optimal planning and
management for urgent care.180 This project could have not been completed without the
linking of datasets.
It is also crucial that when developing data-driven technologies companies ensure that
data generated within those applications are made freely available and interoperable. This
would avoid vendor lock-in situations and promote competition by allowing NHS
organisations to choose different providers.
171 Denis Agniel, Isaac S. Kohane, and Griffin M. Weber, ‘Biases in Electronic Health Record Data Due to Processes within
the Healthcare System: Retrospective Observational Study’, BMJ 361 (April 2018).
172 Tim Yates, ‘Improving Clinical Data Quality: The Digital Health Challenge’, BMJ Quality Blog, 13 December 2016.
173 Amy P. Abernethy et al., ‘Use of Electronic Health Record Data for Quality Reporting’, Journal of Oncology Practice 13,
no. 8 (July 2017): 530–34.
174 Timmis, Heselwood, and Harwich, Sharing the Benefits: How to Use Data Effectively in the Public Sector, 17; Harwich
and Laycock, Thinking on Its Own: AI in the NHS, 31.
175 Wendy Rollason, Kamlesh Khunti, and Simon de Lusignan, ‘Variation in the Recording of Diabetes Diagnostic Data in
Primary Care Computer Systems: Implications for the Quality of Care’, Informatics in Primary Care 17, no. 2 (October
2009); Rui Mendes and Pedro Pereira Rodrigues, ‘Main Barriers for Quality Data Collection in EHR – A Review’,
Conference: HEALTHINF 2011 – Proceedings of the International Conference on Health Informatics, January 2011.
176 Ben Shneiderman, Catherine Plaisant, and Bradford W Hesse, ‘Improving Healthcare with Interactive Visualization’,
Computer 46, no. 5 (May 2013): 61.
177 Fathom, ‘Fathom – AI to Automate Medical Coding’, Webpage, Fathom, n.d.; dictate.it, ‘Welcome to Dictate.It’,
Webpage, dictate.it, n.d.
178 Timmis, Heselwood, and Harwich, Sharing the Benefits: How to Use Data Effectively in the Public Sector, 18.
179 NHS Digital, ‘Data Quality’, Webpage, NHS Digital, 2018.
180 Connected Health Cities, ‘Predictive Modelling for Unplanned Care in the North East and North Cumbria’, Webpage,
Connected Health Cities, 2018.
37
4
Recommendation 7
Procurement rules should include an agreement by digital providers that data generated
within clinical applications should also be freely available for and interoperable with clinical
information at the patient level, either via personal health records or interchange with
appropriate electronic health records.
In the Royal Society and the British Academy’s seminal work on data management and
use in the 21st century, a key concern is the tension between improving public services
through the use of data whilst protecting privacy.181 Risks to patient privacy such as
reidentification can compromise the trustworthiness of private companies and NHS
organisations seeking to use data for research or product development. Ensuring the
auditability, transparency and security of a data access system are not easy feats. Whilst
proving compliance is essential, it might also slow down access to data. A few interviews
for this paper highlighted that there is a reticence within the NHS to share data with
private sector organisations because their objectives are often viewed as antithetical with
the idea of public benefit. This does not need to be the case. The Life Sciences Industrial
Strategy argued that collaboration between NHS and industry can facilitate better care
“for patients through better adoption of innovative treatments and technologies”.182
However, compliance with information governance models is key and clarity over the type
of partnerships and commercial agreements is needed.
4.3.1 Digital innovation hubs

The Life Sciences Industrial Strategy highlighted the use of ‘Digital Innovation Hubs’ as an
opportunity to access healthcare data for research and innovation and reduce the
variation “in digital maturity”.183 The delivery of this programme is being led by Health Data
Research UK.184 The programme has received £37.5 million in funding from the Industrial
Strategy’s Challenge Fund.185 The Hubs are meant to connect data from GP practices,
hospitals, social and community care providers, alongside genetic and biomedical
information and other datasets across regions of 3-5 million people safely and securely.186
The programme aims to improve access to data for “researchers and industry
innovators…to help deliver better care and improve health for patients, society and for
future generations.”187 The governance of these hubs needs to be given serious thought
but could also present the opportunity to involve patients or patient representatives188 to
engage with decisions about how data about them is being used and what type of
commercial arrangements should be put in place.
Recommendation 8
Health Data Research UK in conjunction with NHS England, NHS Digital and the National
Data Guardian should work on developing the appropriate data governance structures to
ensure that Digital Innovation Hubs are safeguarding patient data. This would include
developing audit trails which track how data are used to ensure every interaction with
personal data is auditable, transparent and secure.
181 British Academy and Royal Society, ‘B: Priorities for Governance’, in Data Management and Use: Governance in the 21st
Century: Priorities for Data Governance: Discussions at the British Academy and Royal Socity Seminar on October 2017,
2017, 5.
183 Ibid., 59.
184 Health Data Research UK, ‘Digital Innovation Hub Programme’.
185 Ibid.
186 Health Data Research UK, ‘£37.5m Investment in Digital Innovation Hubs to Tackle Britain’s Biggest Health Challenges’,
Webpage, Health Data Research UK, 16 July 2018.
187 Ibid.
188 Delacroix, ‘Pervasive Data Profiling, Moral Equality and Civic Responsibility’; Sethi and Laurie, ‘Delivering Proportionate
Governance in the Era of EHealth’.
38
4
4.3.2 Synthetic data

Speed of access to data might have an impact on the type of health tech market that will
develop, with smaller businesses being more greatly impacted by delays in access
compared to bigger companies. Nevertheless, that should never come at the price of
good data governance and protection of an individual’s privacy. There are however some
interesting advances in the privacy-preserving space.189
Advances in processing capacity have allowed for the production of high-quality synthetic
data.190 By selecting a specific sample of ‘real’ data, it is possible to create a much larger
sample of very realistic fake data.191 This fake data can then be used to train another
machine learning model to read and appropriately classify mammography scans, for
example. This offers huge future potential in terms of facilitating access to healthcare data.
Synthetic data offers exciting new possibilities in terms of a privacy-preserving technique
and lowering barriers to access data. 192 NHS Digital in partnership with the Medicines and
Healthcare products Regulatory Agency (MHRA) has received funding from government
to develop synthetic datasets.193 The fake datasets will “be aimed at increasing the
capability to accurately measure the effectiveness of new artificial intelligence algorithms
in medical devices in order to validate them and get them to the market”. 194 This is an
interesting step as one as of the regulatory challenges of validating machine learning
algorithms is the fact that datasets used to train the algorithm need to be different from
those used to validate it. Synthetic data offers the possibility of creating infinite amounts of
slightly different datasets.
Public Health England has also recently funded the release of a synthetic dataset for the
cancer registry.195 Having done the initial work around what distribution system for this
type of data works others should follow suit. The Hospital Episode Statistics (HES)
dataset is linked to the cancer registry,196 however NHS Digital is the controller of that
data, which means they could easily replicate what Public Health England has done and
create a synthetic version of HES to be published on their website.
Other NHS organisations should explore the potential of synthetic datasets, by releasing
synthetic data that could be used by private sector companies for research and product
or services development purposes. This would ensure that patient privacy is preserved as
patient data would never be revealed or shared at that stage. In addition, it could allow for
a more informed conversation about commercial models and for the potential value from
data-driven innovations in healthcare to be realised without having to use patient data in
the initial stages of innovation.
Recommendation 9
NHS organisations should offer synthetic datasets, which they can share with private
sector organisations for research and product or service development at the early stages
of the innovation process. This would enable a better understanding of the commercial
value of the innovation and a more informed conversation about the appropriate type of
commercial model that should be developed.
189 MDClone, ‘Synthetic Data’, Webpage, MDClone, n.d.

190 Dimitrios Korkinof et al., ‘High-Resolution Mammogram Synthesis Using Progressive Generative Adversarial Networks’,
ArXiv, July 2018; Arvind Arasu, Kaushik Shriraghav, and Jian Li, ‘Synthetic Data Generation’, 37th International
Conference on Very Large Data Bases, December 2012.
191 Stefanie Koperniak, ‘Artificial Data Give the Same Results as Real Data — without Compromising Privacy’, MIT News, 3
March 2017.
192 R. Shamsuddin et al., ‘Virtual Patient Model: An Approach for Generating Synthetic Healthcare Time Series Data’, in
2018 IEEE International Conference on Healthcare Informatics (ICHI), 2018, 208–18.
193 NHS Digital, ‘NHS Digital Welcomes Funding for Joint Project with MHRA on Creating Synthetic Devices’, Webpage,
NHS Digital, 5 October 2018.
194 Ibid.
195 Public Health England, National Cancer Registration and Analysis Service, and National Disease Registration Service,
‘Background on the Simulacrum’, Webpage, Simulacrum, 2018.
196 NHS Health Research Authority, ‘Cancer Survival Using Further Data from Audits, RTDS and HES – V2’, Webpage, NHS
Health Research Authority, 2017.
39
4
4.4 Procurement, partnership, or somewhere in between?

The healthtech market operates by using two main models: a partnership model and a
procurement model. The procurement model is underpinned by the delivery of a clearly
defined good or a service from an external provider, which is bought by the public sector
(see Figure 11). The provider must have validated its innovation and be able to establish a
business case during competitive tender processes.197
Figure 11: Example of procurement

DrDoctor has created an application that helps to schedule out-patient appointments
which has gone through procurement. The application organises appointments and
platforms virtual appointments between patients and doctors. Industry evidence on users
of this App shows that users have enjoyed the ease of access to appointment services
and healthcare professionals. DrDoctor requires real-time patient data which an algorithm
analyses to make suggestions.
In contrast, a partnership will encompass a more complex relationship that does

necessitate a transfer of finance (see Figure 12). The latter is more likely to be used for the
creation of products or services that need research and trials, while procurement would
be used for a finished products or services, such as the DrDoctor App.
Figure 12: Example of partnership

The Royal Free Trust and DeepMind have partnered to create an App known as Streams.
The App was meant to alert clinicians, “up to 7 hours earlier”, to acute kidney injury (AKI) in
patients. Enabling them to respond faster and avoid kidney damage.198 The App has a
‘service agreement’,199 which stipulates that the NHS uses the App for free unless
DeepMind’s running costs surpass £15,000.200 This has meant that the App has not gone
through a public procurement process as the App is used for free. In 2015, the partners
signed an Information Sharing Agreement, and then later a Memorandum of
Understanding in 2016, which transferred identifiable patient data from the Royal Free to
DeepMind. This transfer was deemed unlawful by the Information Commissioner’s Office.201
Recently, DeepMind has announced it will be passing over the Streams App to its parent
company Google.202 The company has pledged that patient data remains under the strict
control of NHS Trusts “and all decisions about its use will continue to lie with them.”203
However, the company has been criticised for “betraying patient trust”, and dismantling
its Independent Review Panel.204
However, there remain some blurred boundaries. It is unclear how a procurement process
could have been triggered in the partnership model presented in Figure 12 if running costs
surpassed £15,000 as the Royal Free would have already been using the product.
It is also important to consider at what stage data are being used for products or services
development, as it will impact whether procurement or partnership is necessary. Some
197 Department of Health and Social Care, Procurement Guide for Commissioners of NHS-Funded Services, 2010.
198 Natasha Lomas, ‘DeepMind Health Inks New Deal with UK’s NHS to Deploy Streams App in Early 2017’, TechCrunch,
21 November 2016.
199 DeepMind Health and Royal Free London NHS Foundation, ‘Redacted: Services Agreement between DeepMind
Technologies Limited and Royal Free London NHS Foundation Trust’, Contract, 2016.
200 Sam Shead, ‘Google DeepMind Is Giving the NHS Free Access to Its Patient Monitoring App’, Business Insider, 24 June
2017.
201 Information Commissioner’s Office, ‘Royal Free – Google DeepMind Trial Failed to Comply with Data Protection Law’,
Press Release, 3 July 2017.
202 DeepMind Health, ‘Scaling Streams with Google’.
203 Ibid.
204 Jee, ‘Google’s Decision to Absorb DeepMind’s Health Division Has Sparked Privacy Fears’.
40
4
products or services require data for research and development, while other products or
services that have been created will only rely on data to run i.e. a product or service that
requires the input of live patient data to make real-time decisions. The latter is more suited
to procurement, demonstrated by Figure 11. However, often products will require data at
both stages, meaning that partnerships might eventually need to go through
procurement, once the research and development stage has been finished. In addition,
the Life Sciences Industrial Strategy highlights that NHS procurement processes can
“make it very difficult for SMEs to find a route to market”,205 thus creating an uneven
playing field.
There are, however, more structured innovation processes such as the Small Business
Research Initiative (SBRI) (see Figure 13). It offers a pathway to connect public sector
needs with innovative ideas and solutions from the private sector.206 These are divided by
different sectors, the healthcare SBRI is led by the Academic Health and Sciences
Networks (AHSN) that have been set up to encourage health innovation across the UK.207
Figure 13: SBRI process
AHSN led – AHSN led – PHASE 1: PHASE 2: PHASE 3:

typically workshops Typically Typically Typically
undertaken by with industry 6 months – 12 months – 12 months –
clinicians – to support max of £100k milestones agreed milestones agreed
service driven understanding & monitored & monitored
Open procurement
Assessment
Assessment
Prototype
Problem Open call Feasibility development
identification to industry testing Pathway testing
& Proof of value
Due diligence & contracts
Under this initiative, the NHS can identify a market need and run a competition for
businesses to compete for funding in return for providing solutions to health and care
needs.208 The programme is aimed at SMEs, as it provides vital assistance to smaller
companies in the early stages of creation but is, despite the name, open to all business.209
The advantage of this more structured innovation is that government can have more
control over the products or services created with the private sector, ensuring they cater
to the demands of health and care. Once a product or service has been created, the
public sector has the right to license the subsequent technology, however its IP remains
with the company enabling growth and wealth creation for the UK economy.210 Thus far,
the programme has awarded 211 contracts and managed to bring 22 products to
market. 211
However, there are some disadvantages to the SBRI model. As shown in Figure 13 the
maximum amount of funding is limited, which might constrain the type of innovation being
developed or might necessitate further private finance. In addition, the competition is a
one-off, meaning that most other companies that apply will not be supported in any way
and will most likely not survive post-competition, as highlighted by one of the interviews
for this paper.
206 Innovate UK and UK Research and Innovation, ‘SBRI: The Small Business Research Initiative’, Webpage, GOV.UK, 26
September 2018.
207 SBRI Healthcare, ‘About Us’, Webpage, SBRI Healthcare, 2017.
208 Ibid.
209 Ibid.
210 Ibid.
211 Martin Leaver, AHSN Network Impact Report 2017 (The AHSN Network, 2017).
41
4
The Government has launched the Industrial Strategy Challenge Fund (ISCF) to support
UK science and business innovation and help research and product or service
development. Part of this fund makes NHS data available to industry competitors who
can offer product solutions for: early diagnosis and precision medicine, healthy ageing,
and leading-edge healthcare.212 Up to £50 million is available for the creation of Centres of
Excellence for healthcare which will be tailored towards helping SMEs test new products
that use new technologies such as artificial intelligence within the NHS. 213 Alongside the
ISCF, as part of the wider Industrial Strategy in the UK,214 the Government has recently
announced an AI Sector Deal worth £1 billion, £300 million of which comes from private-
sector finance, some of which comes from Japan and Canada.215 The deal puts into
action some of the recommendations in the House of Lords report ‘AI in the UK: ready,
willing and able?’ and is focused on creating the right ecosystem in the UK for a thriving
AI industry.
Partnership models seem to offer more flexibility at the early stages of the innovation
process and might be a way to alleviate some of the concerns around the potentially
ill-fitted nature of current procurement processes for data-driven technologies. However,
there might be a need to clarify the boundaries between some commercial partnerships
and procurement processes. In addition, Figure 12 shows the importance of having the
legal and commercial skills to negotiate a ‘change of control clause’ – a provision in an
agreement giving “a party certain rights in connection with a change in ownership or
management of the other party to the agreement” – in commercial contracts.216
4.5 Commercial and legal skills

To create a thriving healthcare ecosystem in which the private sector acts as a partner to
the public, Government will need the right skills to negotiate these partnerships. 217
However, this will require a significant amount of upskilling. This has been recognised in
Matt Hancock’s vision for the future of the NHS which highlights the need to “build
in-house capability to procure data-driven technologies and manage commercial
arrangements”.218
NHS Digital’s Capability 2020 review found that there were “gaps in expertise […and] legal
capacity”, compared to the banking sector, which can offer higher salaries, being a key
competitor for this labour.219 It also found “the structure of the workforce is unsustainable
with significantly more senior people than junior in commercial, and in procurement not
enough senior people.”220 The review concluded that a “bigger inhouse capability” would
be necessary to bridge the skills gap in NHS digital for working with the private sector.221
NHS England has sought to address this gap in its workforce through the NHS Digital
Academy, a virtual organisation set up to develop a new generation of excellent digital
leaders.222 In partnership with Imperial College London, the University of Edinburgh and
Harvard Medical School, it “provides a year-long world class learning programme”.223 One of
its modules specifically seeks to address the gap around commercial skills and
procurement. However, it may take some time for this programme to impact the prevalence
of commercial and legal skills on the ground, as the first cohort only started in April 2018.
212 Innovate UK, ‘Industrial Strategy Challenge Fund – What Is It and How Is It Being Formed?’, Webpage, GOV.UK,, 3
February 2017.
213 UK Research and Innovation, ‘Up to £50 Million for New Centres of Excellence to Deliver 21st Century Healthcare – Call
Open’, Web Page, UK Research and Innovation, 2018.
214 HM Government, Industrial Strategy: Building a Britain Fit for the Future, 2018.
215 HM Government, Industrial Strategy: Artificial Intelligence Sector Deal, 2018.
216 Practical Law, ‘Change of Control Clause’, Webpage, Practical Law, n.d.
217 Department for Digital, Culture, Media & Sport, Department for Business, Energy & Industrial Strategy, The Rt Hon Greg
Clark MP, and The Rt Hon Matt Hancock MP, ‘Tech Sector Backs British AI Industry with Multi Million Pound
Investment’, Press Release, GOV.UK, 26 April 2018
219 NHS Digital, ‘Fit for 2020 – Report from the NHS Digital Capability Review’, July 2017, 37.
220 Ibid.
221 Ibid.
222 NHS England, ‘NHS Digital Academy’, Webpage, n.d.
223 Ibid.
42
4
At the early stage of partnership discussions, NHS organisations will need to be savvy in
drawing up contracts, applying the most suitable and proportionate models when
negotiating clauses with the private sector.224 There was a near consensus among
interviewees that these frameworks and models will need to be flexible and tailored to a
specific context. These might vary depending on the application and size of the company
as it is crucial that SMEs benefit from a level playing field.
Having an even distribution of commercial skills at Trust-level would be unfeasible as
those skills are a scarce resource within the public sector. They are also a very expensive
resource. The Treasury’s report on knowledge assets recommends the establishment of a
“centre of expertise within government to provide advice and support on the technical,
legal and financial aspects for generating and exploiting knowledge assets.”225
Establishing this centre at a national level would avoid issues around vacant positions at a
local level because attracting that talent is not easy. Nevertheless, when setting up this
‘centre of expertise’ it is important that policy makers are aware of the risk of regulatory
capture – when a regulatory agency advances the concerns of commercial interest
groups rather than acting in the public interest.226
Recommendation 10
The Department of Health and Social Care should invest in creating a new independent
unit with legal and business experts to help NHS organisations negotiate fair and
proportionate partnerships. It would ensure that NHS organisations can have access to
consistent and necessary advice in order to negotiate fair partnership with SMEs. The
negotiation of partnerships should be done by this unit on behalf of Trusts when dealing
with Large Enterprise or SMEs whose parent companies are Large Enterprise.
226 Michael E. Levine and Jennifer L. Forrence, ‘Regulatory Capture, Public Interest, and the Public Agenda: Toward a
Synthesis’, Journal of Law, Economics, & Organization 6 (April 1990); Ernesto Dal Bó, ‘Regulatory Capture: A Review’,
Oxford Review of Economic Policy 22, no. 2 (July 2006).
43
Conclusion
The wealth of data collected and controlled by NHS organisations is extremely valuable
for the improvement of direct patient care and the productivity of the healthcare system.227
States around the world, from Israel228 to South Korea,229 are heavily investing in digital
health programmes as there is huge potential for data-driven innovation in healthcare.230
With one of the oldest national health services in the world, the UK has a uniquely large
and diverse population health-data resource which could give the country a competitive
advantage on the global stage.231
Healthcare data are primarily used for the purposes of patient care but can sometimes be
accessed by organisations outside of the NHS for research and development or product
and service development purposes. Industry access to healthcare data are more
contentious in the eyes of the public than access by academic institutions for example.232
Nevertheless, partnerships with industry are necessary to reap the full benefits of data-
driven innovation.
The DoHSC’s code of conduct for data-driven health and care technology stipulates that
“any benefits from partnerships between technology companies and health and care
providers are shared fairly”.233 However, prior to this code there was no guidance on what
partnerships should look like when there is access to data to create a product or service.
This has meant that Trusts have been independently building partnerships with the private
sector, creating a patchwork on the ground.234
At a national level, there is both a lack of clarity on developments at a local level as well as
a lack of a strategic understanding about the consequences of the current commercial
models. What might seem fair at a local level might lead to a poor outcome on a national
level such as further entrenching healthcare inequalities. In addition, finding the
appropriate ways to engage with patients on the types of commercial models that might
be deemed fair is crucial. Patients need to have trust that their data are being used
ethically for the public benefit whilst preserving their privacy.
As highlighted by the British Academy and Royal Society “societies can act in advance to
create well-founded responses that contribute to bringing the benefits of disruptive
technologies into being.” 235 The DoHSC needs to develop a clear national strategy that
seeks to optimise the value of data held by NHS organisations when it is accessed for
commercial purposes. This will ensure that the value of data held by NHS organisations is
optimised between patients, NHS organisations and industry for public benefit.
227 Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector; Roski, Bo-Linn, and
Andrews, ‘Creating Value In Health Care Through Big Data: Opportunities And Policy Implications’; Border, Big Data
and Public Health; House of Commons Science and Technology Committee, The Big Data Dilemma – Fourth Report of
Session 2015-16, HC 468 (London: The Stationary Office, 2016); Department of Health and Social Care, The Future of
Healthcare; Harvey, ‘Ready…set…AI — Preparing NHS Medical Imaging Data for the Future’; Harwich and Laycock,
Thinking on Its Own: AI in the NHS.
228 Yaacov Benmeleh, ‘Israel to Invest $275 Million in Digital Health Project’, Bloomberg, 25 March 2018; Rahul Dutta, ‘A Bit
of Israeli Chutzpah — a Way Forward for AI Applications in Healthcare!’, Times of Israel, 28 October 2018; Max Schindler
and Judy Siegel-Itzkovich, ‘Israel to Launch Big Data Health Project amid Privacy Concerns’, Jerusalem Post, 25 March
2018.
229 Richard Staines, ‘South Korea Hopes Big Data Will Boost Life Sciences Industry’, Pharmaphorum, 20 February 2018.
230 Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector; Roski, Bo-Linn, and
Andrews, ‘Creating Value In Health Care Through Big Data: Opportunities And Policy Implications’; Border, Big Data
and Public Health; House of Commons Science and Technology Committee, The Big Data Dilemma – Fourth Report of
Session 2015-16; Department of Health and Social Care, The Future of Healthcare; Harvey, ‘Ready…set…AI —
Preparing NHS Medical Imaging Data for the Future’; Harwich and Laycock, Thinking on Its Own: AI in the NHS.
231 Select Committee on Artificial Intelligence, AI in the UK: Ready, Willing and Able, Report of Session 2017–19 (House of
Lords, 2018) p.88; Bell, Life Sciences Industrial Strategy – A Report to the Government from the Life Sciences Sector; ‘A
Revolution in Health Care Is Coming’, p.50, The Economist, 1 February 2018;. Aldrick, ‘Data Could Be a Huge Source of
Funding for the NHS and We Are about to Give It Away’.
234 Ream et al., Accelerating Artificial Intelligence in Health and Care: Results from a State of the Nation Survey, 6.
235 British Academy and Royal Society, Data Management and Use: Governance in the 21st Century, 27.
44
Appendix 236 237 238 239 240 241
Summary of research
project or product/
Year Parties involved service Terms of agreement
2015 - DeepMind and Royal The Streams App can alert A redacted version of the ‘Service Agreement’ can be found
present Free NHS Foundation clinicians to acute kidney online:239
Trusts injury in patients “up to 7
>> The Trust is the data controller acquiring data processing
hours earlier”,237 enabling
Has since been rolled services from DeepMind to ‘support direct care of patients’
them to respond faster and
out in Imperial; Tauton
avoid kidney damage. >> DeepMind own all the IP of any product, as well as a licence to
and Somerset; and
use the Trusts’ background IP
Yeovil Trusts. 236
Eventually, it is supposed
to become an AI assistant >> The Trust is entitled to no commercial benefit; it is being supplied
DeepMind has
to help with diagnosis.238 a service, not a product
recently announced
However, it is currently
the App would be >> The App provides ‘direct care’ and therefore there is ‘implied
used to help doctors
scaled up by Google, patient consent’
communicate by sharing
out of the company’s
real-time patient data with >> If DeepMind’s running costs surpass a threshold of £15,000, the
control.
doctors. NHS trust may have to cover extra costs for the service.240
2016 - Moorfields Eye The partnership is a Specifics of the agreement:

present Hospital NHS research collaboration
>> Data security is ensured by DeepMind
Foundation Trust and between DeepMind and
DeepMind241 Moorfields Eye Hospital. >> The agreement lasts five years with the possibility of renewal
AI is used to diagnose
>> DeepMind accesses pseudonymised and/or anonymised data,
and detect eye disease,
potentially faster and more therefore they do not need direct patient consent
accurately than doctors >> Oncethe agreement comes to an end, all the data collected will
can. be destroyed – unless a new agreement is negotiated
Results found that the >> The algorithms used are owned by DeepMind
machine learning algorithm
made the correct
diagnoses 94.5 per cent of
the time.
236 DeepMind Health, ‘DeepMind Health and the Royal Free’, Webpage, DeepMind, 2017.
237 Ibid.
238 Jamie Condliffe, ‘Is DeepMind’s Health-Care App a Solution, or a Problem?’, MIT Technology Review, 22 November
2016.
239 DeepMind Health and Royal Free London NHS Foundation, ‘Redacted: Services Agreement between DeepMind
Technologies Limited and Royal Free London NHS Foundation Trust’.
240 Shead, ‘Google DeepMind is giving the NHS free access to its patient monitoring app.’
241 Aliya Ram, ‘DeepMind Develops AI to Diagnose Eye Diseases’, Financial Times, 4 February 2018.
45
Making NHS data work for everyone / Appendix
Summary of research
project or product/
2006 - UK Biobank and The UK Biobank is a major Data are licenced to external organisations who can prove that
present various commercial national and international their research will have a positive impact on health. The terms of
and non-commercial resource open to a such agreements:
actors 242 variety of stakeholders
>> Applicants will be expected to pay for access to the resource
(e.g. universities, private
on a cost-recovery basis, with a fixed charge for managing the
sector). Its aim is to
application review process and a variable charge depending
improve prevention,
on how many samples, tests and/or data are required for the
diagnosis and treatment of
approved research project
life-threatening illnesses.
>> UK Biobank will remain the controller of the database and
The data from various
samples but will have no financial claim over any inventions that
blood, urine, and saliva
are developed by researchers using the resource
samples, and detailed
information from 500,000 >> All users will be required to publish their findings and return
people, is used primarily their results to UK Biobank so that they are available for other
for research.243 researchers to use for health-related research that is in the public
interest244
2018 - UK Biobank and Regeneron et al. and UK Terms for partnership:

2020 a consortium of Biobank data are being
>> All contributing $10 million each to the sequencing project
pharmaceutical cleaned and used for
companies such research and drug testing. >> The dataset will eventually be open access, but the consortium
as: Regeneron, will get exclusive access until the end of 2020.
The pre-competitive
AbbVie, Alnylam,
consortium will fund the
AstraZeneca, Biogen
creation of genetic exome
and Pfizer245
sequence data (500,000
genetic data records
linked with other measures
such as brain, heart, body
imaging) for research.
The main aim is to find and
create new treatments.
242 243 244 245
242 UK Biobank, ‘About’.

243 Ibid.
244 UK Biobank, ‘Researchers’, Webpage, UK Biobank, 28 November 2018.
245 UK Biobank, ‘About’; Richard Staines, ‘Regeneron: Biobank Project Offers Hope for Alzheimer’s R&D’, Pharmaphorum,
15 January 2018.
46
Summary of research
project or product/
2016 - Babylon Health The GP at Hand App is for The App has been commissioned by the NHS Hammersmith and
present patients to seek medical Fulham CCG under a General Medical Services (GMS) contractual
help online. The platform agreement:247
connects GPs with
>> The App is free at the point of access, but patients will need to
patients; so they can either
live within proximity to the GP at Hand’s surgeries or live and
have a consultation over
work in London zones 1-3
the phone or book one
online. This is supposed to >> The App has other restrictions on access e.g. people who are
cut waiting times and save under 16 or with complex needs
GP time.
>> Patient data they collect can be shared back with the NHS
It is also hoped that
>> Data are also kept for Babylon’s own research. They write: “We
the App will be able to
diagnose using AI. The may also retain your data for medical regulatory purposes, as
company claimed that its legally required. On a fully anonymised basis, we may use the
diagnosing systems were data for research purposes and to improve the services we
on par with doctors. 246 deliver to all patients”
>> TheApp is still under trial, and progress in the Hammersmith and
Fulham CCG will be monitored before thinking of scaling-up use
The expansion of the App outside of London was recently blocked
by NHS leaders for safety concerns.248
The App exists outside of the NHS, and here Babylon collects
health data through the App. It also has a relationship with Bupa,
were it charges them £25 per appointment.249
246 247 248 249
246 BBC Two, ‘BBC Two – Horizon, 2018, Diagnosis on Demand?’

247 NHS England, ‘GP at Hand – Fact Sheet’, Webpage, NHS England, n.d.
248 National Health Executive, ‘NHS England Blocks Expansion Plans for “GP at Hand” App Citing Safety Fears’, Webpage,
National Health Executive, August 2018.
249 Data Services for Commissioners, Data Quality Guidance for Providers and Commission
47
250 251 252 253 254 255 256 257 258 259 260
Summary of research
project or product/
2017 – Sensyne Health, Sensyne Health is a The company has a partnership and a Memorandum of
present University of Oxford healthcare technology Understanding with NHS Trusts – the first being Oxford University
and Oxford Hospital company using AI and Hospital Foundation Trusts and the University of Oxford.
Foundation Trust, data science to develop
The commercial model involves equity shares:
initially – Chelsea and new medicines and
Westminster Hospital improve patient care. >> OxfordUniversity Hospitals NHS Foundation Trust, South
NHS Foundation Warwickshire NHS Foundation Trust, and Chelsea & Westminster
For example, the GDm-
Trust and South NHS Foundation Trust, have equity shares of 6 per cent, 3.7 per
Health App, which
Warwickshire NHS cent and 3.7 per cent respectively253
is providing digital
Foundation Trust are
therapeutic management >> Itappears that NHS Trusts will own 10 per cent collectively,254 so
now also involved250
for diabetes during it is assumed that each time an NHS Trust joins, the other Trusts’
pregnancy, has been shares reduce in size
particularly successful.
>> Each Trust is also offered a “double bottom-line return” which
98 per cent would
recommend it to friends is £5 million worth of shares, as well as a royalty on any product
and family. It has also developed using its data 255
significantly reduced Sensyne was originally known as Drayson Health when it first
caesarean sections. The started in 2017 – it became Sensyne Health only in 2018, when it
company estimates a registered on the London Stock Exchange.
cost saving of £230 per
patient.251
The company hopes
to become an example
of how a commercial
company can partner with
an NHS organisation.252
2016 – Adler Hey Children’s In 2016 Alder Hey IBM Watson services have been procured through the Technologic
present NHS Foundation announced a new Facilities Council for Adler Hey Children’s NHS Foundation Trust.
Trust, IBM Watson ‘Cognitive Hospital’ which
>> IBM Watson technology is used by the hospital to exploit its own
and Science and would use IBM Watson
data asset, meaning IBM has no need to access NHS data
Technology Facilities technology to harness
Council (STFC) the power of Big Data. 256
>> IBM is contributing £200m, and the government £115m, for
According to IBM, the research into ‘Big Data’ applications259
partnership will improve
>> The Alder Play App collects data to identify patterns such as
the patient experience,
direct care and potentially food queries, but does not collect any personal data other than
create efficiency savings. 257 email addresses. 260
One recent product
development is the Alder
Play App, an AI powered
patient App and webpage.
It allows incoming patients
to ask a chatbot questions
and helps improve
children’s experiences of
hospitals by helping them
earn rewards for visiting
parts of the hospital.258
250 Sensyne Health, ‘Sensyne Health: AI Powered Clinical Solutions’.

251 Sensyne Health, ‘Research & Discovery Powered by AI’, Webpage, Sensyne Health, n.d.
252 Sensyne Health, ‘Sensyne Health – Placing and Proposed Admission to Trading on AIM’, Webpage, London Stock
Exchange, August 2018.
253 Ibid
254 Cookson, ‘Drayson Floats Medical AI Group Sensyne Health on Aim’.
255 Ibid.
256 IBM, ‘Alder Hey Children’s Hospital Set to Become UK’s First “Cognitive” Hospital’, Webpage, IBM, 11 May 2016.
257 Ibid.
258 Stephen Armstrong, ‘Alder Hey’s “Cognitive Hospital” Aims to Turn NHS Use of AI on Its Head’, BMJ 362 (10 September
2018).
259 Clive Cookson, ‘Alder Hey Hospital to Use IBM’s Watson to Comfort Patients’, Financial Times, 10 May 2016.
260 Armstrong, ‘Alder Hey’s “Cognitive Hospital” Aims to Turn NHS Use of AI on Its Head’, 1.
48
Summary of research
project or product/
2017 - NHS Digital and Each company will be Data sharing agreements effectively give a third party (e.g.
2018 various commercial using different datasets university or private-sector company) a licence to use the data:
companies such as: for different purposes,
>> Users pay a fee for the licence to cover the cost for NHS Digital,
Northgate Public such as research and
which costs £1,000 for set-up and the first year’s service. There
Services Ltd; Capita; development.
is an Annual Service Charge of £1,000
IQVIA Solutions UK
Ltd; HEALTH IQ Ltd; >> The owner of the licence can sublet this agreement
Method Analytics
>> If
NHS Digital feels that the agreement has been breached, it will
Ltd; Meditrends Ltd;
Lightfoot Solutions; charge the user up to £15,000 for an audit if it is found to have
Harvey Walsh Ltd; breached the terms.
Device Access UK
Ltd261
2012 - Clinical Practice CPRD collects and links A licence to use data is granted by the Independent Scientific
present Research Datalink de-identified patient data Advisory Committee which independently reviews research
(CPRD) from a network of GP requests to access data from CPRD.
practices and other health-
The CPRD has two main types of licences:
related longitudinal data
such as HES controlled >> An Individual Study Licence for one-off study
by NHS Digital. Data are
>> An Unlimited Licence for an annual subscription
provided for research that
will bring about health It has a three-tiered pricing model for:
benefits and medical
advancement. >> Non-commercial customers (academic, government, charity)
>> Commercial mid-tier customers (single entity commercial
organisations), and
>> Commercial multi-affiliate organisations
The price of an Individual Study Licence for most of CPRD’s link
data is £4,000 irrespective of the customer sector. The Individual
Study Licence for primary care data is £15,000, £30,000 or
£60,000 for each tier, respectively. A Multi-Study Licence for
primary care data is approximately 5 times the price of the
Individual Study Licence.
The review process for each licence has multiple stages. The
priority is ensuring the owner of the license is a reputable bona
fide research organisation which is going to use the data to
conduct research for public benefit – it is expected that all
research and results will be published.
The terms of the licence include:
>> Restrictions preventing the owner of the licence from merging the
data with other data sources; transferring the data to third parties
other than as strictly permitted through the customer’s licence;
or selling the data or incorporating it in to a product. Crucially,
however, the results of the research using the data are owned by
the licensee and not further restricted
>> All
CPRD data must be destroyed on the completion of the
study and/or licence term; the data cannot be used to measure
effectiveness of sales or advertising campaigns.
261
261 For more information, http://theysolditanyway.com/ website has a further examples of NHS Digital data release register.
49
Summary of research
project or product/
2015 - DrDoctor and various An App that helps patients This example follows procurement via the G-Cloud framework.
present health Trusts, such make appointments and Alongside this, it has had to make separate data sharing
as Guy’s Foundation reminders. The Trusts agreements with each Trust, sometimes a costly experience for
Trust262 are expected to make a DrDoctor.
£2.5 million to £3.5 million
return from the App.
Outcomes thus far:
Improve referral-to-
treatment time; increase
attendance by over 15
per cent; reduce non-
attendance by 47 per cent.
2017 - Ultromics263 a spin out Creating new diagnostic The trial agreement:
present from the University aids for cardiologists using
>> Ultromics have access to NHS data across various wards and
of Oxford and now AI.
use data to run algorithms to diagnose heart disease
partnered with 20
Initial findings indicate
NHS hospitals264 >> It has said publicly that whatever diagnostic products it creates
the AI could increase the
would be shared for free with the NHS
accuracy of diagnoses by
80-90 per cent. >> The company have said that the products could potentially save
the NHS £300 million a year by reducing the number of people

who are incorrectly diagnosed265
The company intends to bring technology to the USA and UK
markets.266
2002 – National Joint Most companies use The NJR has different costings for different stakeholders. The
present Registry (NJR) and the data to judge the principle of the subscription rate is that ‘you pay what you can’. It
various commercial success of their hip or joint is a cost-recovery model:
companies replacements and other
>> The data comes from the NHS web portal. The NJR will clean
health products/services
and analyse the data – which the NHS will get back for free
associated with joints.267
>> Commercial companies will pay a subscription of £3,500 per
year, large companies pay more

>> Data comes with analytical report – but a report or bespoke data
will cost extra.
262 263 264 265 266 267
262 DrDoctor, ‘Home’, Webpage, 2018.

263 Ultromics, ‘Revolutionising Cardiovascular Diagnosis’, Webpage, Ultromics, n.d.
264 Leontina Postelnicu, ‘Start-Ups Leveraging AI to Improve Outcomes and Put Patients in Control of Their Care’,
Webpage, MobiHealthNews, 2 October 2018; Pete Hughes, ‘JR AI to Detect Heart Disease Could Be Rolled out across
the UK’, Oxford Mail, 4 January 2018.
265 Owen Hughes, ‘Ultromics AI Platform to Expand Trials to 20 NHS Hospitals’, Webpage, Digital Health, May 2018.
266 Ibid.
267 Understanding Patient Data, ‘Monitoring Joint Replacement Surgery’, Webpage, Understanding Patient Data, n.d.
50
Summary of research
project or product/
2016 - Kheiron Medical, Kheiron Medical, a Kheiron has a partnership with the NHS platformed by grant
present several universities, machine learning company funding and data sharing agreements:
the NHS and seed focused on supporting
>> Grants have been provided by various UK and EU organisations.
funders radiologists to tackle
One example is the NHS England Wave 2 Test Bed project
breast cancer. The
with the East Midlands Radiology Consortium (EMRAD), which
company brings together
Kheiron was awarded in a competition269
experts from academia
(Oxford, Cambridge, >> Kheironis also involved in two out of the five ISCF Centres of
Edinburgh, McGill, KTH, Excellence for AI in digital radiology and pathology
[UCL] and Imperial Some of these public-sector grants also provide money to the
College), and clinical NHS partner to produce the resources the SME needs to develop
experts.268 products or services
The company has also published on its website that it hopes to
create a commercial strategy with the NHS.270
Spring Skin Analytics Ltd Currently in development, The partnership is platformed by the SBRI healthcare initiative.
2017 and Imperial College the partnership is
>> SkinAnalytics has won competition as part of the ISCF data
Health Partners pioneering two new
to early programme. The funding will help them support the
AHSN271 technologies which
development of two products for the diagnosis of skin cancer.272
should enable cost-
effective population- >> Theproject has thus far received £99,941 from government
based screening for skin funding
cancer. The company
has already built two AI Prior to the partnership, the company had received private funding
melanoma screening tools for a start-up. One product was initially priced at £40 and aimed at
for assessing pigmented families and consumers, rather than medical professionals.273
lesions in primary care The company is also conducting research with 7 NHS hospitals,
and earlier. Some can led by the Royal Free. It is the first prospective study for AI and
be used at home on a melanoma, and the company promises to share the research
personal mobile. The two results in early 2019.274
extra products under
development will be used
in addition to the others.
268 269 270 271 272 273 274
268 Kheiron Medical, ‘About Us’, Webpage, Kheiron Medical, 2018.

269 Kheiron Medical, ‘NHS Code of Conduct for Data-Driven Health and Care Technology’, Webpage, Kheiron Medical,
2018.
270 Ibid.
271 Skin Analytics, ‘Pioneering Research’, Webpage, Skin Analytics, 2018.
272 SBRI Healthcare, ‘Skin Analytics Ltd’, Webpage, SBRI Healthcare, 2017
273 L aura Rosbrow-Telem, ‘Skin Analytics’ Smartphone Lens Screens for Skin Cancer’, Webpage, Geektime, 16 October
2015.
274 Skin Analytics, ‘Pioneering Research’
51
Glossary
Anonymised data: data “about individuals but with identifying details removed”.275
Artificial intelligence: any manmade agent (i.e. computer programme or robot) that
exhibits intelligence. Intelligence is defined as an “agent’s ability to achieve goals in a wide
range of environments.”276
Audit trail: a record of everyone who has looked at or changed a record, why and when
they did so and what changes they made.277
Consent (and its different forms): “approval or agreement for something to happen
after consideration. For consent to be legally valid, the individual must be informed, must
have the capacity to make the decision in question and must give consent voluntarily.”278
Explicit Consent: “It can be given in writing or verbally, or conveyed through another
form of communication such as signing.” 279 Implied consent: “applicable only within the
context of direct care of individuals. It refers to instances where the consent of the
individual patient can be implied without having to make any positive action, such as
giving their verbal agreement for a specific aspect of sharing information to proceed.” 280
Data architecture: this is the overall environment of data and includes frameworks,
models, standards, policies, data management, data quality, unstructured data etc.
Data controller: “a person who (either alone or jointly or in common with other persons)
determines the purposes for which and the manner in which any personal data are, or are
to be, processed.”281
Data-driven technologies: in this paper the term will describe any technology that uses
data as part of its creation and not just or solely as part of its functioning. An example
would be a machine learning algorithm which needs data to be trained on.
Data processor: “any person (other than an employee of the data controller) who
processes the data on behalf of the data controller.”282
Data subject: “an individual who is the subject of personal data”.283
Data standards: the rules by which data are described and recorded. In order to share
data properly, the format and the meaning of the data must be standardised.
Information governance: the way in which organisations management the way
information and data are handled to ensure it is legal, secure, efficient and effective.284
Intangible asset: a “non-monetary asset without a physical substance”. An asset is
defined as “any resource controlled by a company which generate future economic
benefits and has an associated cost or value which can be reliably measured.”285
Knowledge asset: a type of intangible asset. It defines a wide-range of intellectual
resources such as intellectual property (IP), software, data, technological expertise,
organisational know-how, etc.286
275 NHS Digital, ‘How We Look after Information’, Webpage, NHS Digital, 2017.
276 Shane Legg and Marcus Hutter, ‘A Collection of Definitions of Intelligence’, Frontiers in Artificial Intelligence and
Applications 157 (June 2007): 8.
277 National Data Guardian for Health and Care, Information: To Share or Not to Share? Information Governance Review.
278 Health and Social Care Information Centre, A Guide to Confidentiality in Health and Social Care, 2013, 7.
279 Ibid.
280 Ibid.
281 Information Commissioner’s Office, Guide to Data Protection.
282 Ibid.
283 Ibid.
284 National Data Guardian for Health and Care, Information: To Share or Not to Share? Information Governance Review.
286 HM Treasury, Getting Smart about Intellectual Property and Other Intangibles in the Public Sector: Budget 2018.
52
Machine learning: a subset of AI that allows computer systems to learn by analyzing huge
amounts of data and drawing insights from it rather than following pre-programmed rules.287
Metadata: information that describes other data by providing a description of its
content.288 For example, a digital image may include information on its size, resolution,
when the image was created etc.
Personal identifiable data: data “containing details that identify individuals”.289
Pseudonymised data: data “about individuals but with identifying details (such as name
or NHS number) replaced with a unique code.”290
Population/aggregate data: “anonymised information grouped together so that it
doesn’t identify” individuals. 291
287 The Royal Society, Machine Learning: The Power and Promise of Computers That Learn by Example, 2017, 20.
288 Richard Kemp, Paul Hinton, and Paul Garland, ‘Legal Rights in Data’, Computer Law & Security Review 27, no. 2 (April
2011): 142.
289 NHS Digital, ‘How We Look after Information’.
290 Ibid.
291 Ibid.
53
Bibliography
Abernethy, Amy P., James Gippetti, Rohit Parulkar, and Cindy Revol. ‘Use of Electronic
Health Record Data for Quality Reporting’. Journal of Oncology Practice 13, no. 8
(July 2017).
Accenture. Digital Consumer Health Engagement 2016 – Global Report, 2016.
Agniel, Denis, Isaac S. Kohane, and Griffin M. Weber. ‘Biases in Electronic Health Record
Data Due to Processes within the Healthcare System: Retrospective
Observational Study’. BMJ 361 (April 2018).
Aldrick, Philip. ‘Data Could Be a Huge Source of Funding for the NHS and We Are about
to Give It Away’. The Times, 3 March 2018.
———. ‘If NHS Patient Data Is Worth £10 Billion, Put It on the Balance Sheet and Save
Lives Too’. The Times, 15 June 2018.
———. ‘NHS Set to Be Offered a “Fair Share” of Data Profits’. The Times, 5 September
2018.
Almquist, Eric, John Senior, and Nicolas Bloch. ‘The Elements of Value’. Harvard
Business Review, 1 September 2016.
Arasu, Arvind, Kaushik Shriraghav, and Jian Li. ‘Synthetic Data Generation’. 37th
International Conference on Very Large Data Bases, December 2012.
Armstrong, Stephen. ‘Alder Hey’s “Cognitive Hospital” Aims to Turn NHS Use of AI on Its
Head’. BMJ 362 (September 2018).
BBC Two. ‘Diagnosis on Demand? The Computer Will See You Now’. Horizons, 5
November 2018.
Bell, Sir John. Life Sciences Industrial Strategy – A Report to the Government from the
Life Sciences Sector. HM Government, 2017.
Benmeleh, Yaacov. ‘Israel to Invest $275 Million in Digital Health Project’. Bloomberg, 25
March 2018.
Bero, L. A., R. Grilli, J. M. Grimshaw, E. Harvey, A. D. Oxman, and M. A. Thomson.
‘Closing the Gap between Research and Practice: An Overview of Systematic
Reviews of Interventions to Promote the Implementation of Research Findings.’
BMJ 317, no. 7156 (August 1998).
Bhatti, Naureen. ‘Seeing a GP on a Smartphone Sounds Wonderful – but It’s Not’. The
Guardian, 16 November 2017.
Blasimme, Alessandro, Effy Vayena, and Ernst Hafen. ‘Democratizing Health Research
Through Data Cooperatives’. Philosophy & Technology 31, no. 3 (September
2018).
Border, Peter. Big Data and Public Health. Parliamentary Office of Science and
Technology, 2014.
Bromiley OBE, Martin, Elisabeth Buggins CBE, Eileen Burbidge MBE, Richard Horton, Dr
Julian Huppert, Professor Donal O’Donoghue, Matthew Taylor, and Professor Sir
John Tooke. DeepMind Health Independent Review Panel Annual Report, 2018.
British Academy, and Royal Society. ‘B: Priorities for Governance’. In Data Management
and Use: Governance in the 21st Century: Priorities for Data Governance:
Discussions at the British Academy and Royal Socity Seminar on October 2017,
2017.
54
———. Data Management and Use: Governance in the 21st Century, 2017.
British Academy, and Royal Society. Connecting Debates on the Governance of Data and
Its Uses, 2017.
Budin-Ljøsne, Isabelle, Harriet J. A. Teare, Jane Kaye, Stephan Beck, Heidi Beate
Bentzen, Luciana Caenazzo, Clive Collett, et al. ‘Dynamic Consent: A Potential
Solution to Some of the Challenges of Modern Biomedical Research’. BMC
Medical Ethics 18, no. 1 (January 2017).
Burns, Alex P. ‘Where Is the Evidence for Automated Triage Apps?’ British Medical
Journal 360 (February 2018).
Caldicott, Dame Fiona. ‘A Data Opt-out Will Reassure Patients and Improve Care’. The
Times, 13 July 2017.
Carter, Pam, Graeme T. Laurie, and Mary Dixon-Woods. ‘The Social Licence for
Research: Why Care.Data Ran into Trouble’. Journal of Medical Ethics, January
2015.
Centre for Economics and Business Research. Data on the Balance Sheet. SAS, 2013.
Chauhan, Bhupendrasinh F., Maya Jeyaraman, Amrinder Singh Mann, Justin Lys, Becky
Skidmore, Kathryn M. Sibley, Ahmed Abou-Setta, and Ryan Zarychanksi.
‘Behavior Change Interventions and Policies Influencing Primary Healthcare
Professionals’ Practice—an Overview of Reviews’. Implementation Science 12
(January 2017).
Clark, Sophie, Michael Damiani, Holly Dorning, Mary Halter, and Alison Porter. ‘Patient-
Level Data Linkage across Ambulance Services and Acute Trusts: Assessing the
Potential for Improving Patient Care’. International Journal for Population Data
Science 1, no. 1 (April 2017).
Colquhoun, Heather L., Janet E. Squires, Niina Kolehmainen, Cynthia Fraser, and Jeremy
M. Grimshaw. ‘Methods for Designing Interventions to Change Healthcare
Professionals’ Behaviour: A Systematic Review’. Implementation Science 12, no.
1 (March 2017).
Condliffe, Jamie. ‘Is DeepMind’s Health-Care App a Solution, or a Problem?’ MIT
Technology Review, 22 November 2016.
Connected Health Cities. ‘Connected Yorkshire’. Webpage. Connected Health Cities,
2016.
———. ‘Predictive Modelling for Unplanned Care in the North East and North Cumbria’.
Webpage. Connected Health Cities, 2018.
Cookson, Clive. ‘Alder Hey Hospital to Use IBM’s Watson to Comfort Patients’. Financial
Times, 10 May 2016.
———. ‘Drayson Floats Medical AI Group Sensyne Health on Aim’. Financial Times, 14
August 2018.
Coote, Anna. Prevention Rather Than Cure. The King’s Fund, 2004.
Couldry, Nick, and Ulises A. Mejias. ‘Data Colonialism: Rethinking Big Data’s Relation to
the Contemporary Subject’. Television & New Media, September 2018.
Dal Bó, Ernesto. ‘Regulatory Capture: A Review’. Oxford Review of Economic Policy 22,
no. 2 (July 2006).
Data Services for Commissioners. Data Quality Guidance for Providers and
Commissioners. NHS England, 2016.
55
Making NHS data work for everyone / Bibliography
Datum Foundation Ltd. ‘Datum Returns Data Ownership to Individuals’. PR Newswire.

2017.
deBronkart, Dave. ‘The Value of Sharing Data: What Healthcare Can Learn from
Oncology’. Future Health Index Blog, 3 March 2017.
DeepMind Health. ‘DeepMind Health and the Royal Free’. Webpage. DeepMind, 2017.
———. ‘Independent Reviewers’. Webpage. DeepMind, 2018.
———. ‘Scaling Streams with Google’. Webpage. DeepMind, 13 November 2018.
DeepMind Health, and Royal Free London NHS Foundation. ‘Redacted: Services
Agreement between DeepMind Technologies Limited and Royal Free London
NHS Foundation Trust’. Contract, 2016.
Delacroix, Sylvie. ‘Pervasive Data Profiling, Moral Equality and Civic Responsibility’. UCL
Laws and UCL Computer Science, 17 August 2017.
Department for Digital, Culture, Media and Sport. Data Protection Bill: Factsheet –
Overview, 2017.
———. UK Digital Strategy 2017, 2017.
Department for Digital, Culture, Media & Sport, Department for Business, Energy &
Industrial Strategy, The Rt Hon Greg Clark MP, and The Rt Hon Matt Hancock MP,
‘Tech Sector Backs British AI Industry with Multi Million Pound Investment’, Press
Release, GOV.UK, 26 April 2018
Department of Health and Social Care. Initial Code of Conduct for Data-Driven Health
and Care Technology. Guidance, 2018.
———. ‘New Guidance to Help NHS Patients Benefit from Digital Technology’. GOV.UK,
n.d.
———. Procurement Guide for Commissioners of NHS-Funded Services, 2010.
———. The Future of Healthcare: Our Vision for Digital, Data and Technology in Health
and Care, 2018.
———. The NHS Constitution for England, 2015.
Department of Health and Social Care, and NHS. ‘Information Governance Toolkit’.
Webpage, n.d.
Dettrey, Erin. ‘The Value of Healthcare Data Analytics in an EMR’. Webpage. Modernizing
Medicine, 13 September 2017.
dictate.it. ‘Welcome to Dictate.It’. Webpage. dictate.it, n.d.
Dombrowski, Stephan U., Pauline Campbell, Helen Frost, Alex Pollock, Julie McLellan,
Steve MacGillivray, Anna Gavine, et al. ‘Interventions for Sustained Healthcare
Professional Behaviour Change: A Protocol for an Overview of Reviews’.
Systematic Reviews 5 (October 2016).
DrDoctor. ‘Home’. Webpage, 2018.
Dutta, Rahul. ‘A Bit of Israeli Chutzpah — a Way Forward for AI Applications in
Healthcare!’ Webpage. Times of Israel, 28 October 2018.
Edge Health. ‘Improving Theatre Productivity with Machine Learning – “Space Finder”’.
Webpage. Edge Health, 2017.
———. ‘Prediction Tool for Critical Care Bed Occupancy – “Critical Planner”’. Webpage.
Edge Health, 18 October 2017.
56
———. ‘Surgical Pathway Demand and Capacity Planning’. Webpage. Edge Health, July
2017.
Fathom. ‘Fathom – AI to Automate Medical Coding’. Webpage. Fathom, n.d.
Fisher, Rebecca. ‘Data Sharing to Improve Care’. Webpage. The Health Foundation, 18
May 2018.
Fisher, Rebecca, Ruth Thorlby, and Will Warburton. ‘Briefing: Sharing to Improve: Four
Case Studies of Data Sharing in General Practice’. The Health Foundation, May
2018.
Fitzgerald, Ed. ‘International Lessons to Help the NHS Harness Health Data’. The
Reformer Blog, 17 July 2018.
Gkatzelis, Vasilis, Christina Aperjis, and Bernardo A. Huberman. ‘Pricing Private Data’.
Electronic Markets 25, no. 2 (June 2015).
Hammond MP, Rt Hon Philip. ‘Budget 2018: Philip Hammond’s Speech’. Speech. HM
Treasury, 29 October 2018.
Harvey, Hugh. ‘Ready…set…AI — Preparing NHS Medical Imaging Data for the Future’.
Webpage. Towards Data Science, 24 October 2017.
Harwich, Eleonora, and Kate Laycock. Thinking on Its Own: AI in the NHS. Reform, 2018.
Health and Social Care Information Centre. A Guide to Confidentiality in Health and Social
Care, 2013.
Health Data Research UK. ‘£37.5m Investment in Digital Innovation Hubs to Tackle
Britain’s Biggest Health Challenges’. Webpage. Health Data Research UK, 16 July
2018.
———. ‘Digital Innovation Hub Programme’. Webpage. Health Data Research UK, n.d.
HealthWatch Cumbria. Annual Report 2018, 2018.
Hill, Rebecca. ‘Making Good Use of NHS Data “Vital for UK Plc”, Says Royal Society’.
Public Technology, 26 April 2017.
HIMSS. ‘The Value of Quality Healthcare Data’. Webpage. HIMSS, 16 May 2016.
HM Government. Industrial Strategy: Artificial Intelligence Sector Deal, 2018.
———. Industrial Strategy: Building a Britain Fit for the Future, 2018.
HM Treasury. Getting Smart about Intellectual Property and Other Intangibles in the
Public Sector: Budget 2018, 2018.
———. ‘The Economic Value of Data: Discussion Paper’, August 2018.
———. The Green Book, Central Government Guidance on Appraisal and Evaluation,
2018.
Hoeksma, Jon. ‘NHS England to Establish New Regional Digital Innovation Hubs’.
Webpage. Digital Health, 12 September 2017.
House of Commons Science and Technology Committee. The Big Data Dilemma – Fourth
Report of Session 2015-16. HC 468. London: The Stationary Office, 2016.
House of Lords. Select Committee on Digital Skills. Summary and Summary of
Conclusions and Recommendations, 2015.
Houssami, Nehmat, Christoph I. Lee, Diana S. M. Buist, and Dacheng Tao. ‘Artificial
Intelligence for Breast Cancer Screening: Opportunity or Hype?’ Breast 36
(December 2017).
57
Hughes, Owen. ‘Ultromics AI Platform to Expand Trials to 20 NHS Hospitals’. Webpage.

Digital Health, May 2018
Hughes, Pete. ‘JR AI to Detect Heart Disease Could Be Rolled out across the UK’. Oxford
Mail, 4 January 2018.
Hussain, L., W. Aziz, S. Saeed, S. Rathore, and M. Rafique. ‘Automated Breast Cancer
Detection Using Machine Learning Techniques by Extracting Different Feature
Extracting Strategies’. IEEE, 2018.
Ibarra, Imanol Arrieta, Leonard Goff, Diego Jiménez Hernández, Jaron Lanier, and E. Glen
Weyl. ‘Should We Treat Data as Labor? Moving Beyond “Free”’. American
Economic Association Papers and Proceedings 1, no. 1 (December 2017).
IBM. ‘Alder Hey Children’s Hospital Set to Become UK’s First “Cognitive” Hospital’.
Webpage. IBM, 11 May 2016.
Information Assurance and Governance Team. ‘About the Panel Public Benefit and
Privacy Panel for Health and Social Care’. Webpage. Information Assurance and
Governance Team, 2015.
Information Commissioner’s Office. Guide to Data Protection, 2017.
———. Guide to the General Data Protection Regulation (GDPR), 2018.
———. ‘Overview of the General Data Protection Regulation (GDPR)’. Webpage, 2017.
———. ‘Royal Free – Google DeepMind Trial Failed to Comply with Data Protection Law’.
Press Release, 3 July 2017.
Innovate UK. ‘Industrial Strategy Challenge Fund – What Is It and How Is It Being
Formed?’ Webpage. GOV.UK, 3 February 2017.
Innovate UK, and UK Research and Innovation. ‘Business Innovation: What Funding You
Can Get and How to Apply’. Webpage. GOV.UK, 16 August 2018.
———. ‘SBRI: The Small Business Research Initiative’. Webpage. GOV.UK, 26
September 2018.
Jee, Charlotte. ‘Google’s Decision to Absorb DeepMind’s Health Division Has Sparked
Privacy Fears’. MIT Technology Review, 14 November 2018.
Johnson, Mark J., and Carl R. May. ‘Promoting Professional Behaviour Change in
Healthcare: What Interventions Work, and Why? A Theory-Led Overview of
Systematic Reviews’. BMJ Open 5, no. 9 (September 2015).
Kemp, Richard, Paul Hinton, and Paul Garland. ‘Legal Rights in Data’. Computer Law &
Security Review 27, no. 2 (April 2011).
Kheiron Medical. ‘About Us’. Webpage. Kheiron Medical, 2018.
———. ‘NHS Code of Conduct for Data-Driven Health and Care Technology’. Webpage.
Kheiron Medical, 2018.
Knapton, Sarah. ‘How the NHS Got It so Wrong with Care.Data’. The Telegraph, 7 July
2016.
Koperniak, Stefanie. ‘Artificial Data Give the Same Results as Real Data — without
Compromising Privacy’. MIT News, 3 March 2017.
Korkinof, Dimitrios, Tobias Rijken, Michael O’Neill, Joseph Yearsley, Hugh Harvey, and
Ben Glocker. ‘High-Resolution Mammogram Synthesis Using Progressive
Generative Adversarial Networks’. ArXiv, July 2018.
58
Laurie, Graeme, and Emily Postan. ‘Rhetoric or Reality: What Is the Legal Status of the
Consent Form in Health-Related Research?’ Medical Law Review 21, no. 3
(September 2013).
Lawrence, Neil D. ‘Data Readiness Levels’. ArXiv, May 2017.
Leaver, Martin. AHSN Network Impact Report 2017. The AHSN Network, 2017.
Legg, Shane, and Marcus Hutter. ‘A Collection of Definitions of Intelligence’. Frontiers in
Artificial Intelligence and Applications 157 (June 2007).
Levine, Michael E., and Jennifer L. Forrence. ‘Regulatory Capture, Public Interest, and the
Public Agenda: Toward a Synthesis’. Journal of Law, Economics, & Organization
6 (April 1990).
Lomas, Natasha. ‘DeepMind Health Inks New Deal with UK’s NHS to Deploy Streams
App in Early 2017’. TechCrunch, 21 November 2016.
London Stock Exchange. ‘Sch 1 – Sensyne Health Limited – RNS -’. Webpage. London
Stock Exchange, 3 August 2018.
Lord Freyberg. ‘NHS: Healthcare Data – House of Lords Debate Hansard’, 2018.
Marrs, Colin. ‘Care.Data – An In-Depth Check-up on NHS England’s Controversial Bid to
Join up Health Data’. Civil Service World, 18 September 2015.
Martin, Alexander J. ‘NHS Patients’ Data Was Illegally Transferred to Google DeepMind’.
Webpage. Sky News, 3 July 2017.
McCartney, Margaret. ‘Could Babylon Please Supply Evidence ?’ BMJ, no. 358
(September 2017).
MDClone. ‘Synthetic Data’. Webpage. MDClone, n.d.
MedConfidential. ‘More Information’. Webpage. MedConfidential, 2017.
Medicalchain. Medicalchain Whitepaper 2.1, 2018.
Mendes, Rui, and Pedro Pereira Rodrigues. ‘Main Barriers for Quality Data Collection in
EHR – A Review’. Conference: HEALTHINF 2011 – Proceedings of the
International Conference on Health Informatics, January 2011.
Mitchell, Lord Parry. ‘Protect Our Valuable NHS Data from Big Tech’. Financial Times, 7
February 2018.
Murgia, Madhumita. ‘DeepMind’s Move to Transfer Health Unit to Google Stirs Data
Fears’. Financial Times, 13 November 2018.
National Data Guardian for Health and Care. Impact and Influence for Patients and
Service Users. Department of Health, 2017.
———. Information: To Share or Not to Share? Information Governance Review.
Department of Health, 2013.
———. Review of Data Security, Consent and Opt-Outs, 2016.
National Institute for Health and Care Excellence. Data Science for Health and Care
Excellence, 2016.
———. ‘Glossary’. Webpage. NICE, n.d.
National Joint Registry. ‘Home’. Webpage. National Joint Registry, n.d.
Naylor, Annemarie. ‘Parliamentary Debate Briefing: This House Takes Note of the Value of
NHS and Healthcare Data and How It Could Be Used to Improve the Health of
Our Nation.’ Future Care Capital, 6 September 2018.
59
Naylor, Sir Robert. NHS Property and Estates, 2017.

Neville, Sarah. ‘Bradford Hospital Launches AI Powered Command Centre’. Financial
Times, 15 October 2018.
NHS Digital. ‘Data Quality’. Webpage. NHS Digital, 2018.
———. ‘Data Sets’. Webpage. NHS Digital, n.d.
———. ‘Fit for 2020 – Report from the NHS Digital Capability Review’, July 2017.
———. ‘How We Look after Information’. Webpage. NHS Digital, 2017.
———. ‘NHS Digital Welcomes Funding for Joint Project with MHRA on Creating
Synthetic Devices’. Webpage. NHS Digital, 5 October 2018.
NHS England. Five Year Forward View, 2014.
———. ‘GP at Hand – Fact Sheet’. Webpage. NHS England, n.d.
———. Improving Outcomes through Personalised Medicine, 2016.
———. Next Steps on the NHS Five Year Forward View, 2017.
———. ‘NHS Digital Academy’. Webpage, n.d.
National Health Executive. ‘NHS England Blocks Expansion Plans for “GP at Hand” App
Citing Safety Fears’. Webpage. National Health Executive, August 2018.
NHS Health Research Authority. ‘Cancer Survival Using Further Data from Audits, RTDS
and HES – V2’. Webpage. NHS Health Research Authority, 2017.
NHS Scotland. Public Benefit and Privacy Panel for Health and Social Care: Guiding
Principles and Policy for Decision Making. 1st ed., 2015.
Oswald, Malcolm, Ruth Norris, Lamiece Hassan, Niels Peek, and Mary Tully. ‘Health Data
on Public Trial: To What Extent Should Patients Control Access to Patient
Records?’ International Journal of Population Data Science 1, no. 1 (April 2017).
Oxford Academic Health Science Network. Oxford AHSN Year 6 Q2 Report. Oxford
Academic Health Science Network, 2018.
Patel, Tejal A., Mamta Puppala, Richard O. Ogunti, Joe E. Ensor, Tiancheng He, Jitesh B.
Shewale, Donna P. Ankerst, et al. ‘Correlating Mammographic and Pathologic
Findings in Clinical Decision Support Using Natural Language Processing and
Data Mining Methods’. Cancer 123, no. 1 (January 2017).
Pickard, K. T. ‘Exploring Markets of Data for Personal Health Information’. In 2014 IEEE
International Conference on Data Mining Workshop, 2014.
Porter, Michael E. ‘What Is Value in Health Care?’ New England Journal of Medicine 363,
no. 26 (December 2010).
Posner, Eric A., and E. Glen Weyl. ‘Data as Labor’. In Radical Markets. Princeton
University Press, 2018.
Postelnicu, Leontina. ‘Start-Ups Leveraging AI to Improve Outcomes and Put Patients in
Control of Their Care’. Webpage. MobiHealthNews, 2 October 2018.
Powles, Julia, and Hal Hodson. ‘Google DeepMind and Healthcare in an Age of
Algorithms’. Health and Technology 7, no. 4 (December 2017).
Practical Law. ‘Change of Control Clause’. Webpage. Practical Law, n.d.
60
Public Health England, National Cancer Registration and Analysis Service, and National
Disease Registration Service. ‘Background on the Simulacrum’. Webpage.
Simulacrum, 2018.
Rahwan, Iyad. ‘Society-in-the-Loop’. MIT MEDIA LAB (blog), 12 August 2016.
———. ‘Society-in-the-Loop: Programming the Algorithmic Social Contract’. Ethics and
Information Technology 20, no. 1 (March 2018).
Ralph, Alex, and Philip Aldrick. ‘Google “Poised to Profit” from NHS Patient Records’. The
Times, 24 November 2018.
Ram, Aliya. ‘Babylon Signs Digital Health Deal with Samsung’. Financial Times, May
2018.
———. ‘DeepMind Develops AI to Diagnose Eye Diseases’. Financial Times, 4 February
2018.
Ream, Melissa, Tina Woods, Indra Joshi, and Laura Day. Accelerating Artificial
Intelligence in Health and Care: Results from a State of the Nation Survey. The
AHSN Network, Department of Health and Social Care and NHS England, 2018.
Reform. ‘Labour Party Conference: “Making NHS Data Work for Everyone” (Roundtable
Event)’. Webpage. Reform, 24 September 2018.
———. ‘Making NHS Data Work for Everyone (Panel Event)’. Webpage. Reform, 5 June
2018.
Roche. ‘Roche to Acquire Flatiron Health to Accelerate Industry-Wide Development and
Delivery of Breakthrough Medicines for Patients with Cancer’. Press Release, n.d.
Rodriguez-Ruiz, Alejandro, Jan-Jurre Mordang, Nico Karssemeijer, Ioannis Sechopoulos,
and Ritse M. Mann. ‘Can Radiologists Improve Their Breast Cancer Detection in
Mammography When Using a Deep Learning Based Computer System as
Decision Support?’, Vol. 10718. International Society for Optics and Photonics,
2018.
Rollason, Wendy, Kamlesh Khunti, and Simon de Lusignan. ‘Variation in the Recording of
Diabetes Diagnostic Data in Primary Care Computer Systems: Implications for the
Quality of Care’. Informatics in Primary Care 17, no. 2 (October 2009).
Rosbrow-Telem, Laura. ‘Skin Analytics’ Smartphone Lens Screens for Skin Cancer’.
Webpage. Geektime, 16 October 2015.
Roski, Joachim, George W. Bo-Linn, and Timothy A. Andrews. ‘Creating Value In Health
Care Through Big Data: Opportunities And Policy Implications’. Health Affairs 33,
no. 7 (2014).
Saran, Cliff. ‘NHS Data Not Fit for AI, Lords Select Committee Told’. Webpage.
ComputerWeekly.com, n.d.
SBRI Healthcare. ‘About Us’. Webpage. SBRI Healthcare, 2017.
SBRI Healthcare. ‘Skin Analytics Ltd’. Webpage. SBRI Healthcare, 2017.
Schindler, Max, and Judy Siegel-Itzkovich. ‘Israel to Launch Big Data Health Project amid
Privacy Concerns’. Jerusalem Post, 25 March 2018.
Scott, Kaela. Data for Public Benefit: Balancing the Risks and Benefits of Data Sharing.
Involve, The Carnegie UK Trust’s, Understanding Patient Data, 2018.
Selbie, Duncan. ‘Prevention and the NHS Long Term Plan: 3 Ways We Can Save More
Lives’. Press Release. Public Health England, 6 September 2018.
61
Select Committee on Artificial Intelligence. AI in the UK: Ready, Willing and Able. Report
of Session 2017–19. House of Lords, 2018.
Sellick, Vicki, Solder Amy, and Rosalinda Baker. Funding for Innovation. Nesta, 2018.
Sensyne Health. ‘Research & Discovery Powered by AI’. Webpage. Sensyne Health, n.d.
———. ‘Sensyne Health: AI Powered Clinical Solutions’. Webpage. Sensyne Health, n.d.
———. ‘Sensyne Health – Placing and Proposed Admission to Trading on AIM’.
Webpage. London Stock Exchange, August 2018.
Sethi, Nayha, and Graeme T. Laurie. ‘Delivering Proportionate Governance in the Era of
EHealth’. Medical Law International 13, no. 2–3 (June 2013).
Shamsuddin, R., B. M. Maweu, M. Li, and B. Prabhakaran. ‘Virtual Patient Model: An
Approach for Generating Synthetic Healthcare Time Series Data’. In 2018 IEEE
International Conference on Healthcare Informatics (ICHI), 208–18, 2018.
Shead, Sam. ‘Google DeepMind Is Giving the NHS Free Access to Its Patient Monitoring
App’. Business Insider, 24 June 2017.
Shneiderman, Ben, Catherine Plaisant, and Bradford W Hesse. ‘Improving Healthcare
with Interactive Visualization’. Computer 46, no. 5 (May 2013).
Skin Analytics. ‘Pioneering Research’. Webpage. Skin Analytics, 2018.
Skok, Michael. ‘4 Steps To Building A Compelling Value Proposition’. Forbes, 14 June
2013.
Smith, Chris. NHS and Healthcare Data. House of Lord’s Research Briefing, 2018.
Staines, Richard. ‘Regeneron: Biobank Project Offers Hope for Alzheimer’s R&D’.
Pharmaphorum, 15 January 2018.
———. ‘South Korea Hopes Big Data Will Boost Life Sciences Industry’. Pharmaphorum,
20 February 2018.
The Caldicott Committee. Report on the Review of Patient-Identifiable Information.
Department of Health, 1997.
The Economist. ‘A Revolution in Health Care Is Coming’. The Economist, 1 February 2018.
The Office for National Statistics. ‘Quality Adjustment’. In The ONS Productivity
Handbook, 2016.
The Royal Society. Machine Learning: The Power and Promise of Computers That Learn
by Example, 2017.
Timmis, Sarah, Luke Heselwood, and Eleonora Harwich. Sharing the Benefits: How to
Use Data Effectively in the Public Sector. Reform, 2018.
Timothy Revell. ‘Google DeepMind NHS Data Deal Was “Legally Inappropriate”’. New
Scientist, 16 May 2017.
Turner, Andrew, Clara Dallaire-Fortier, and Madeleine J. Murtagh. ‘Biobank Economics
and the “Commercialization Problem”’. Spontaneous Generations: A Journal for
the History and Philosophy of Science 7, no. 1 (2013).
UK Biobank. ‘About’. Webpage. UK Biobank, n.d.
UK Biobank. ‘Researchers’. Webpage. UK Biobank, 28 November 2018.
UK Research and Innovation. ‘Up to £50 Million for New Centres of Excellence to Deliver
21st Century Healthcare – Call Open’. Web Page. UK Research and Innovation,
2018.
62
Ultromics. ‘Revolutionising Cardiovascular Diagnosis’. Webpage. Ultromics, n.d.

Understanding Patient Data. ‘Monitoring Joint Replacement Surgery’. Webpage.
Understanding Patient Data, n.d.
Veale, Michael, Reuben Binns, and Lilian Edwards. ‘Algorithms That Remember: Model
Inversion Attacks and Data Protection Law’. Phil. Trans. R. Soc. A, no. 376 (July
2018).
Vezyridis, Paraskevas, and Stephen Timmons. ‘Understanding the Care.Data
Conundrum: New Information Flows for Economic Growth’. Big Data & Society 4,
no. 1 (1 June 2017).
Wachter, Robert M. Making IT Work: Harnessing the Power of Health Information
Technology to Improve Care in England. Department of Health and Social Care,
2016.
Walker, Alan, and Philippa Westbury. ‘Towards a Successful Data-Enabled Economy:
Promoting Trust in Data and Data-Driven Systems’. In Connecting Debates on the
Governance of Data and Its Use, 2017.
Weiskopf, Nicole Gray, and Chunhua Weng. ‘Methods and Dimensions of Electronic
Health Record Data Quality Assessment: Enabling Reuse for Clinical Research’.
Journal of the American Medical Informatics Association 20, no. 1 (January
2013).
Wellcome Trust. The One-Way Mirror: Public Attitudes to Commercial Access to Health
Data. Ipsos MORI, 2016.
Whittall, Hugh. ‘The Use and Reuse of Data’. In Connecting Debates on the Governance
of Data and Its Use. The Royal Society and British Academy, 2016.
Williams, Hawys, Karen Spencer, Caroline Sanders, David Lund, Edgar A Whitley, Jane
Kaye, and William G Dixon. ‘Dynamic Consent: A Possible Solution to Improve
Patient Confidence and Trust in How Electronic Patient Records Are Used in
Medical Research’. JMIR Medical Informatics 3, no. 1 (January 2015).
Winfield, Alan F. T., and Marina Jirotka. ‘Ethical Governance Is Essential to Building Trust
in Robotics and Artificial Intelligence Systems’. Phil. Trans. R. Soc. A 376, no.
2133 (November 2018).
Yates, Tim. ‘Improving Clinical Data Quality: The Digital Health Challenge’. BMJ Quality
Blog, 13 December 2016.
63
64
Reform T 020 7799 6699
5-6 St Matthew Street [email protected]
London www.reform.uk
SW1P 2JT
ISBN 978-1-910850-26-8

Making NHS Data Work For Everyone - 1544639862

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Making NHS Data Work For Everyone - 1544639862

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Making NHS Data Work For Everyone - 1544639862

Uploaded by

Copyright:

Available Formats

Making NHS data

work for everyone

December 2018 #reformNHSdata

Sam Smith, medConfidential

Uses of healthcare data

The value exchange

Keeping society in the loop

The healthtech ecosystem

Letting a thousand flowers bloom?

Access to good quality data

Procurement, partnership, or somewhere in between?

Commercial and legal skills

1.1 Direct patient care  12

1.1 Direct patient care

1.2 Secondary uses

1.2.1 Uses outside of the NHS

Figure 1: Stages of research, development and product/service development

Idea/ Patient/user Proof of

Final Pre-market Market

Product/ Compliance Commercial

Source: Reform interviews and research.

1.2.2 Access to data

2.1 Improvement for patients and the healthcare system  16

2.1 Improvement for patients and the healthcare system

At a national level, it is important to consider what the deployment of these data-driven

2.2.1 What does value in data mean?

Figure 2: Value(s) of NHS data

Source: Reform interviews and research. Gif by Zuhura Plummer.

Value is highly complex to define because as shown in Figure 2 different stakeholders

2.2.2 What affects value?

Figure 3: Determinants of the value of data – the supply side 64 65 66 67 68 69 70 71 72 73

Simple level Moderate level High level

Figure 4: Determinants of the value of data – the demand side 78

2.2.3 How to measure value

2.4 Wider societal impact

97 Department of Health and Social Care, The Future of Healthcare.

3.1 Currently not in the loop  26

3.1 Currently not in the loop

Figure 5: Public opinion

Why Mix of public and Solely for

Who For profit but in No link to

What Aggregate passively Aggregate but risk Identifiable

How Secure storage & regulation is assumed

Public buy-in will be essential to forming partnerships.118 The care.data programme

3.2 Engagement by design

Figure 6: Proportionate governance

4.1 Letting a thousand flowers bloom?  31

4.1 Letting a thousand flowers bloom?

Figure 7: Current models 144

Data sharing Grant-funded Licencing Cost-recovery Commercial

Source: Reform research and interviews.

139 Department of Health and Social Care, The Future of Healthcare.

Figure 9: Sensyne health

4.2 The need for a national strategy

Figure 10: Typology of commercial models 162 163 164

Name Description Brief evaluation

162 Sellick, Solder, and Baker, Funding for Innovation, 5.

1.1 Direct patient care 12

2.1 Improvement for patients and the healthcare system 16

97 Department of Health and Social Care, The Future of Healthcare.

3.1 Currently not in the loop 26

4.1 Letting a thousand flowers bloom? 31

139 Department of Health and Social Care, The Future of Healthcare.

162 Sellick, Solder, and Baker, Funding for Innovation, 5.

165 Bhatti, ‘Seeing a GP on a Smartphone Sounds Wonderful – but It’s Not’.

189 MDClone, ‘Synthetic Data’, Webpage, MDClone, n.d.

242 UK Biobank, ‘About’.

246 BBC Two, ‘BBC Two – Horizon, 2018, Diagnosis on Demand?’

250 Sensyne Health, ‘Sensyne Health: AI Powered Clinical Solutions’.

262 DrDoctor, ‘Home’, Webpage, 2018.

268 Kheiron Medical, ‘About Us’, Webpage, Kheiron Medical, 2018.