Doctor Know A Knowledge Commons in Health

Doctor Know
a knowledge commons
in health
John Loder, Laura Bunt and Jeremy C Wyatt
March 2013
2 Doctor Know: a knowledge commons in health
About The Young Foundation

We are The Young Foundation and we are determined to make
positive social change happen. We pioneered the field of social
innovation with The Open University, UpRising and Studio Schools.
We work closely with individuals, communities and partners
building relationships to ensure that our thinking does something,
our actions matter and the changes we make together will continue
to grow.
youngfoundation.org
About Nesta
Nesta is the UKs innovation foundation. We help people and
organisations bring great ideas to life. We do this by providing
investments and grants and mobilising research, networks and skills.
We are an independent charity and our work is enabled by an
endowment from the National Lottery.
Nesta Operating Company is a registered charity in England and Wales with company
number 7706036 and charity number 1144091. Registered as a charity in Scotland
number SC042833. Registered office: 1 Plough Place, London, EC4A 1DE
www.nesta.org.uk
Nesta 2013.
Doctor Know: a knowledge commons in health 3
Doctor Know
a knowledge commons
in health
CONTENTS
Introduction 4
1 A Brief Architecture of
a Health Knowledge System 6
2 Why a Health Knowledge Commons? 9
3 The potential for a Health Knowledge Commons 14
4 What next? Making a knowledge commons happen 30
Summary of recommendations 35
Conclusion and next steps 37
Acknowledgements 38
Bibliography 39
Endnotes 41
Introduction
The way we create, access and share information is changing rapidly. Every time we
look something up on Wikipedia, rate an experience on Tripadvisor or enter search
terms in Google, we are taking advantage of the increasingly sophisticated way in which
technology and digital tools are allowing us to capture, refine, synthesise and structure
our collective intelligence. With the ongoing advances of the semantic web, new sources
of and different applications for data and cultural shifts towards greater openness and
transparency, our capacity for creating and navigating complex knowledge grows.
These trends in the creation and application of knowledge have huge implications for
how we access, create and apply information in health, a field where knowledge held by
patients, doctors, medical researchers, nurses, carers, community providers, families and
others is all critical in improving our individual health and wellbeing. Where information
is vast and complex and the need for accuracy and reliability can be a matter of life and
death our ability to orchestrate knowledge in a useful way is a central concern for any
health system.
In this paper, we argue that societys growing ability to mobilise knowledge from different
fields and sources is beginning to show the potential of a knowledge commons in
healthcare: an open system of knowledge with researchers, practicing clinicians, patients,
their families and communities all involved in capturing, refining and utilising a common
body of knowledge in real time. Over the next few decades, this has the potential to
reframe our understanding of healthcare as a highly knowledgecentred system, with
informed patients able to take more responsibility and agency over their own health,
supported by an integrated network of healthcare providers and practitioners drawing on
the most uptodate and relevant knowledge to best care for our needs.
This raises questions: How does the health system tap into the collective intelligence of the
different actors involved in creating knowledge about health, and leverage different types
of knowledge? What are the opportunities presented by new technologies and digital tools
to orchestrate health knowledge in a coherent and useful way? And, how can we ensure
that health knowledge remains reliable and accessible to the right people at the right time?
Some components of this system are already in place, or will soon be. The UK Government
has committed to making all patient records accessible and available as anonymised data
for research by 2015. Cheap biometric sensors, smart phones, telehealth equipment and
increasingly wellstructured patient records will all contribute to a considerably larger

data set. This includes not only biomedical data, but data that is broader, more granular
and which takes account of patients whole lives. Advances in digital technology allows us
to gather and analyse much larger quantities of data, collaboratively structure and refine
knowledge in a more systematic way, and access relevant, useful knowledge in real time.
Cultural and political movement towards greater openness and transparency, campaigns
to encourage greater citizen participation in research and ownership of information and
the rise of commonsbased licensing agreements, consumption models and communities,
also have important application in health. Of course, the commons principle of openness
presents substantial issues around privacy, and there are legitimate questions about
the reliability of knowledge derived from outside a research context. There is also the
important aspect of accessibility. Digital inclusion and literacy cannot be a determinant of
whether someone can access information about their health. Finding ways to address these
issues is fundamental to realising the potential for a knowledge commons approach.
A health knowledge commons has the potential to allow us to better understand and react
to our own or anothers health in real time. It also has the potential to stimulate innovation
in open science and academia, as access to different sources and new combinations of
data generate new knowledge about the causes of disease. A knowledge commons will
have particular relevance for the kinds of knowledge which the system presently struggles
to create: applied knowledge, such as knowledge from ordinary practice, and knowledge
about how people respond to conditions in their daily lives. As longterm conditions are
the main challenge facing healthcare in the developed world, reliable information that helps
people pursue autonomous lives is critical.
This paper contains four sections: firstly, we set out a brief architecture of a knowledge
system and the different types of knowledge involved; secondly, we consider the issues
with the creation and application of knowledge in the current system, involving healthcare
providers, medical academia, patientheld and applied knowledge and considering how
this interacts; thirdly, we look in more detail at some key trends mediated by technology,
design, data management and presentation that have the potential to change how we
orchestrate knowledge in health, with some examples; finally, we consider what this means
for the development of a knowledge commons and the steps to get there.
1 A Brief Architecture of a
Health Knowledge System
One of the challenges in considering a knowledge commons in health is the complexity of

the many different types of knowledge involved, their relative utility and reliability. Any
attempt at structuring this kind of system would need to respect how different types of
knowledge were presented, and be explicit about different degrees of certainty what
knowledge is proven, what is only relevant in a particular context, what is a grey area etc.
This section considers what we mean by knowledge in this context, the different uses and
users of a knowledge commons in health, and how a knowledge system should work.
What do we mean by knowledge?
There are at least two complementary uses of the word knowledge. One is how
knowledge the understanding of how to do something, for example tie a shoe lace, speak
French, or stitch a wound. This is often referred to as procedural knowledge. The other is
what knowledge, which is our understanding of what is the case: that Paris is the capital
of France, or that a bacterium is the main cause of ulcers. This is propositional knowledge.
We hold propositional knowledge for a variety of purposes, but a chief purpose is to help
us make decisions about the world.
Here we are primarily concerned with propositional knowledge that informs decision
making about health, by anyone making a choice about their or anothers health (such
as patients, clinicians and managers). These decisions are about more than deciding on
the right course of treatment, and include diagnosis, testing, prognosis, referral, screening,
selfcare, resource allocation and strategic policy.
We also make a distinction between data and knowledge. The former describes a particular
state (e.g. a patients haemoglobin level or endoflife preference) while knowledge is
widely applicable (e.g. that transfusion will raise haemoglobin levels, but is not usually
indicated in someone who is at the end of their life). As well as being a necessary part of
an individual treatment decision, data are the raw material for the creation of knowledge,
and the job of a knowledge production system is transforming data into knowledge in the
form of theories or models drawn from research which guide our decisions and behaviour
and which can be applied in practice. This is described in functional terms in Figure 1.
Figure 1:
From data into knowledge
Derive
Guidance
Refine
Decision
Knowledge
and Action
Capture
Data
For example, how do we learn that diabetes is controlled by insulin, and how do we use
that knowledge to improve health outcomes in diabetics? As visualised in Figure 1, data
gathered from patient interactions and research is both used to help guide the care of that
patient and refined into knowledge, through analysis and distillation into evidence, which is
in turn translated into steps to guide treatment or action. This decision and action may be
carried out by a clinician, a patient or carer, or an alliance of all three.
What sort of knowledge system do we want
We want knowledge that guides us to the right decision as often as possible. That is to
say we want knowledge that is precise, applicable, accessible and relevant.
Firstly we want knowledge that is relevant to our concerns. There is more value to
knowledge about common conditions than rare ones, all other things being equal. Similarly
we would wish our knowledge to cover the range of possible influences on health. We
therefore need a variety of different kinds of knowledge, including:
Clinical knowledge tells us about how to make a diagnosis and what treatments cure
or ameliorate diseases. It is the body of knowledge that guides clinicians and patients
in predicting and altering the course of disease: which drugs, surgical techniques
or therapies are effective in improving outcomes. While this body of knowledge
is informed by biomedical knowledge, it is quite common to know how to treat
a condition without knowing exactly how the treatment works. Similarly, medical
understanding of a disease might not necessarily mean having a clear idea of how to do
anything about it.
Experiential knowledge is knowledge about how a condition manifests itself, such as

knowledge of pain or of particular symptoms, and the impact of disease on quality of
life.
Contextual knowledge is knowledge about how healthcare best happens in practice,

under uncertain conditions, with constrained resources, and against the backdrop
of peoples real and complex lives, geographies and social interactions. Contextual
knowledge requires a degree of judgement, interpretation and intuition.
Secondly a great deal of our health knowledge occupies a middle ground between
certainty and ignorance. Conditions and treatment are often understood in general terms,
but knowledge might be limited in relation to specific cases. A good knowledge system
gradually improves the reliability and precision of our knowledge, from knowing what
factors are relevant, to be able to predictably improve outcomes on average, to finally
being able to tailor a treatment to individual patients. A good knowledge system would
slowly increase the range, reliability, and precision of our knowledge, reasonably quickly
and cheaply.
Thirdly we want knowledge to be applicable in practical circumstances. This means that

our knowledge must contain not only general theories, but also knowledge of how to apply
these theories in practice the appropriate diagnostic tests, and management plans to
make treatment successful, and understanding of how treatment is affected by and works
in context.
There are many different individuals involved in creating and applying different types of
knowledge, including:
Highly engaged patients people living with a particular disease who are actively
involved in learning about their health, either as a consequence of the severity of a
particular condition and its degenerative factors, such as motor neurone disease, or
with a view to preventing a diseases development, such as cancer.1
Engaged patients people living with a longterm condition that requires ongoing
monitoring and treatment, such as diabetes, asthma, depression or heart conditions.
Families and carers people involved in supporting anothers condition who want to
find out more information about suggested tests or treatment.
Clinical practitioners specialist and general clinical practitioners working within a

hospital or general practice setting who need access to uptodate knowledge.
Community health practitioners community nurses, carers, specialist health workers,

health visitors, pharmacists, opticians and other service providers.
Researchers academics and researchers within academic institutions and universities,

medical colleges and teaching hospitals, and within R&D functions of pharmaceutical
companies, medical device developers and other businesses.
Businesses including clinical service providers, medical technology producers.
Commissioners clinical commissioners concerned with the efficient and effective

purchasing of drugs, treatments, services and so on to achieve health outcomes for a
specific population.
Citizens people not formally engaged with the health system but wanting to access
information about public health, lifestyle and preventative care, such as diet, exercise,
social and demographic factors, directory services and so on.
These various groups access and apply the types of health knowledge differently. For
example, a patient wanting to learn more about their particular condition would benefit
from access to experiential knowledge of others living with the same disease. This
knowledge, although not formalised within the medical arena, may be extremely beneficial
to a patient making a choice about a course of treatment, if only as a comparison. However,
a clinician might draw on experiential knowledge in prescribing treatment, but would not
want to rely on this for diagnosis. Any knowledge system must therefore make the relative
reliability of sources knowledge clear so as to guide decision making in a legitimate and
safe way.
In addition, we must be clear about the uses of knowledge. Health knowledge exists to
guide a whole range of clinical actions: diagnosis, testing, treatment, prognosis, referral,
screening, and so on. A knowledge system must also be structured so as to reflect who
needs what knowledge, at what points in time and in the most relevant format. What
is the right architecture for a system to ensure that the right people access the most
relevant, reliable and useful knowledge? How to account for different levels of confidence
in application and interpretation of knowledge? What are viable options for structuring
knowledge, acknowledging bias, and taking account for a variety of applications?
2 Why a Health
Knowledge Commons?
The success of modern medicine is one of the crowning human achievements of the 21st
century. Infectious diseases such as tuberculosis, diphtheria, cholera, measles, polio,
yellow fever and smallpox once common killers are now virtually eradicated in
developed countries. More recently, huge strides have been made against cancer, HIV/
Aids and forms of chronic disease such as stroke, diabetes, dyslipidaemia and ischaemic
heart disease.
The UK in particular benefits from an extremely sophisticated institutional system

supporting the health knowledge cycle, including research councils, universities, evidence
centres, centralised intermediaries such as the National Institute for Clinical Excellence
(NICE), clearing houses such as the Cochrane Collaboration and NHS Evidence, strong
professional development and teaching bodies, medical networks and a plethora of
established medical journals such as the British Medical Journal (BMJ), and distribution
channels such as the Medical and Nursing Royal Colleges and teaching hospitals. The NHS
as the national public health service brings huge advantage in coordinating this system at
a national scale.
However, all healthcare systems today face a new set of challenges and opportunities,
which offer the potential to address some of the limitations of the current ways in which
knowledge is generated and applied in health and respond to the way in which the health
system as whole needs to change in the future. Firstly, the changing scope of data and
health knowledge given the changing nature of disease; secondly, the process of refining
data into knowledge; and finally, applying this knowledge in practice.
The changing scope of data and health knowledge
Many have observed that there is a substantial need for innovation in the models of
healthcare delivery, given the changing nature of disease and the pressure on health
systems from the costs of managing longterm health conditions and providing more
complex treatments.2 In part as a result of such advances in curing disease and improved
longevity, the dominant demands on the system come from treating chronic diseases such
as chronic obstructive pulmonary disease (COPD), diabetes, asthma and heart failure.
Supporting patients living with these conditions to live autonomous lives requires a
different sort of healthcare supported by different sorts of engagement, as is summarised
in Table 1.
Table 1: The changing model of healthcare (based on a table in the Kerr

report into the future of the NHS in Scotland, 2005)3
Old model of care New model of care
Focus on acute conditions Focus on longterm conditions

Reactive management Prevention and continuing care
Hospital centred Embedded in homes and communities
Disjointed episodes Integrated with peoples lives
Doctor dependent Team based, shared record
Patient as passive recipient Patient as partner
Selfcare infrequent Selfcare encouraged and supported
Use of information and communication Dependent on ICT and devices
technology (ICT) rare
Table 1 contrasts the reactive, doctorcentred model of old with the emerging model in
which the doctor is supported by a wide range of other professionals and increasingly,
by the informed, autonomous patient. Information and communication technology play an
important part in this new model, but need to be accompanied by a shift in attitude and
professional culture towards what we could call clinical subsidiarity devolving clinical
decisions to the person closest to the patient, which may often be the patient. This has
been enshrined in recent health policy in the phrase no decision about me without me,
which should perhaps evolve to decisions about me made by me.4 This has significant
implications for the way in which knowledge is orchestrated in health.
Firstly, the move to promoting wellness and the longterm management of chronic
conditions changes the objective of medical knowledge, exposing the fact that health
can in many cases be defined by the ability to live an autonomous life rather than by
physiology alone. When an absolute cure is possible, this distinction is not particular
important restoring physical condition maximises autonomy as far as the clinician can
affect it. However, when this is not possible (such as with chronic disease), maximising
autonomy becomes central. Clinicians and patients need to work together to help elicit
the patients preferences, set appropriate goals and responsibilities in the management
and treatment plan, codesign health services and software applications that respond
dynamically to changing circumstances and unforeseen events, and equip patients to make
informed and effective decisions in their daytoday lives.
For example, an accurate prognosis that allows a patient to know if they will be in good or
poor shape tomorrow would be invaluable, even if it made no actual difference to health
outcomes, as it would allow people to plan their lives, rather than be prisoners of their
condition. In making decisions such as these, it is important to note that the experts on
coping with the complexity of life with a longterm condition are not doctors but patients;
so gathering, validating and sharing their knowledge becomes critical.
Secondly, more account needs to be taken of complex nonphysiological factors that have
a significant effect on health outcomes. Social, emotional, behavioural and practical factors
all combine in a complex web in the management of longterm conditions. Consider a
diagnosis of childhood diabetes; for both the child and their whole family, this can require
complex and profound adjustments. Figure 2 shows some of these.
Figure 2: The complex interacting system of adolescent diabetes management
Friends Family Blood

Glucose
Independence Puberty
& autonomy
Body
image
Exercise Lifestyle School
Health
& illness
Blood
glucose
tests
Alcohol Diet Emotions Insulin
Thirdly, medicine currently fails to learn from routine management. For many diseases,
development of the best management approach is an iterative process. Patients are
individuals, who often have multiple comorbidities. Further, standard medical knowledge
is far from a complete recipe, and doctors often go beyond established protocols.
Treatment effects are somewhat uncertain, and there is a considerable need for adjustment
and iteration before the correct course is found. Indeed the negotiation of this process
is one of the key skills of the physician. What this means is that there are thousands of
smallscale undocumented experiments taking place in surgeries and hospitals across the
country every day. However, we currently do not extract any learning from these for the
system, which stays with the individual physician, if they are able to make use of it.5
Reliability
As is often acknowledged, the process of refining data into knowledge even when it
concerns the formal research process is somewhat unreliable. In a now famous paper in
JAMA, John Ioannidis examined the most cited papers (1,000 plus citations) in the best
regarded journals in the world largely drawn from The Lancet, the New England Journal
of Medicine and the Journal of the American Medical Association. Of those with claims
of efficacy whose results had been tested, 41 per cent were either found to be wrong, or
the impact was much smaller than the original study suggested.6 Sponsorship of trials by
the pharmaceutical industry may be part of the explanation, but given that randomised
controlled trials with hundreds or thousands of participants represent the gold standard in
research, the reliability of smaller studies published elsewhere is likely to be lower.
Whether Ioannidis results are explained by the nature of science or bias, there does seem
considerable opportunity to improve. And indeed there is considerable effort to improve
the quality of research results, by the use of higher evidentiary standards, reporting
guidelines and more emphasis on metaanalyses (for example the Equator network).7
However, this more careful approach seems likely to exacerbate a second problem.
The progress of an idea from inception to the publication of an influential article that
demonstrates clinical efficacy is slow a median 24 years.8 While larger trials with more
careful followup may improve the reliability of results, they might also slow down the
already glacial pace of generating research results.
Further to this, the creation of applied knowledge has been comparatively neglected.9 As
mentioned above, applied knowledge is knowledge about how treatment is implemented
in practice, under uncertainty, constrained resources, and against the backdrop of real and
complex lives. It includes knowledge about how to diagnose a condition, how to balance
risks in deciding on a course of treatment, and how patient and clinician can work together
to deliver that treatment in an effective chronic disease management programme. This
encompasses understanding of organisational processes. Healthcare is not a craft practiced
by individuals, but a service delivered by a complex system involving many individuals
and institutions. The way this system is arranged has profound consequences for health
outcomes.
However, while this area is vitally important to health systems worldwide probably more
so than uncovering the molecular mechanisms of disease it is challenging to fundamental
research, and consequently has only received a fraction of the investment of biomedical
science. Improving how a condition is managed often requires a complex service redesign,
composed of many individuals and interventions whose effect together creates the impact.
This can make a full RCT challenging,10 though the Medical Research Council (MRC) has led
work on understanding this challenge.11
Applicability
A pragmatist would argue that clinical knowledge is only useful when it guides patient or
professional decision making, and thus improves health.12 For professional knowledge to be
applied successfully to an individual patient, four things have to be in place:
A clear description of the current status of the patient, their problem, their capabilities
and preferences for diagnostic and treatment options usually held as data in the
medical record;
Access to the relevant medical knowledge;
Application of that knowledge to the specific patient to create a management plan,

including where appropriate a course of treatment;
Successful execution of the management plan over time.

All of these stages can be problematic. First, clinicians still often do not have access to all
the relevant patient data when they need to make a decision. Second, the body of medical
knowledge is now so vast and growing so quickly, that it is becoming humanly impossible
for unaided healthcare professionals to deliver patient care with the efficacy, consistency
and safety that the full range of current knowledge could support. PubMed, the online
database of medical research, has 21 million articles, with 500,000 being added every year
and a doubling time of 19 years.13 The capacity of individuals to access the most current,
highquality knowledge pertaining to their practice or specific cases remains a challenge
(if not an insurmountable one).14 Third, application to the individual patient sometimes
requires complex logic to be followed (e.g. which combination of chemotherapy drugs is
most likely to be effective and minimise side effects) or calculations (e.g. of drug dose)
to be carried out, and often remains a challenging expert judgement. Further, even when
the knowledge exists and is well synthesised and understood, it is extremely easy in the
course of a complex activity for mistakes to be made for the knowledge to be applied
incorrectly.
As effective selfcare becomes increasingly important, knowledge needs to be made

useful to support patients to care for themselves. Patients with chronic conditions typically
spend the vast majority of their time away from a clinician, so their health outcomes will
be critically affected by lifestyle and behaviour. Anything that highlights early warning
signs, allowing patients to take action to avoid a deterioration or crisis in their condition,
would be very significant. Motivational help e.g. to support medication adherence is also
important in this context; a longterm condition can test ones emotional capability and
resilience, and active and determined selfcare of a serious health condition is a challenge
for anyone.15
Of course medical knowledge is complex, and patients will continue to rely on clinical
interaction to manage their treatment to a certain extent. However, it would be complacent
to assume that people are as informed and as active in their own selfcare as they could
be. When longterm conditions are the main medical issue for around 17 million people
in the UK, a passive patient population and overreliance on a medicalised model of
knowledge cannot be the answer.
The paradox of health knowledge
This list of issues, while perhaps not wholly consensual, is not particularly controversial.
Our more substantial point is that, despite the sophistication of the UKs current health
knowledge system, we are not drawing on this knowledge as best we might to improve
our health. It is perhaps a paradox that as our knowledge base has grown, it has become
increasingly difficult to navigate and make use of in order to improve care and outcomes.
How could a commons approach address these issues? How in the future can the health
system draw on new sources of data, new modes of distillation of data into knowledge,
and new ways of applying that knowledge in practice? What sort of system could make
different kinds of knowledge available for different applications?
3 The potential for a Health

Knowledge Commons
In essence, a knowledge commons is defined by the principle that knowledge is a shared

resource made accessible and intentionally open rather than subject to ever greater
restrictions through intellectual property legislation, patenting and licensing. The concept
of a knowledge commons has developed in line with the expansion of digital information,
as digital platforms offer unprecedented access to information through the Internet
and the tools to create and share knowledge openly online. Pioneers of the knowledge
commons movement include Wikipedia as an online, collaborativelybuilt encyclopaedia,
Linux and the open source software movement, Creative Commons licensed art and open
scientific collections such as the Public Library of Science or the Science Commons.16
The notion of a knowledge commons in health implies a system where participants share
data and knowledge openly with a view to accelerating improvements in health outcomes,
through discovery of new drugs or pathways of selfcare and clinical research, better
application of knowledge and more selfcare and prevention through more informed
citizens taking more responsibility for their own health. The components of a commons
might include databases on the results of clinical trials, libraries of drug and chemical
compounds, computational models predicting drug efficacy or side effects, shared tools or
methods for research, practice and innovations in service design, personalised data analysis
based on demographic, behavioural or health data and comparisons with others, with a
personalised interface and interactive system.17
Given the highly sensitive and complex nature of health, a health knowledge commons would
need to have the right regulatory and governance system, and the means of sanctioning and
highlighting the varying reliability of knowledge and degrees of uncertainty. The quantity
of medical information that is produced is huge, and beyond the practical ability of most
clinicians to keep abreast. Some degree of aggregating and filtering is therefore necessary to
ensure the most relevant information is available to the person who needs it and the means
of illustrating varying interpretations of data according to perspective.18
Though perhaps still difficult to realise as a whole, there are a number of key current
trends that begin to illustrate what a knowledge commons in health might achieve.
Firstly, our increasing ability to capture and process big data that can illuminate different
perspectives on what affects health and wellbeing; secondly, collaborative analysis tools
that facilitate distributed problem solving and the generation of new knowledge through
combination; finally, a commitment to openness and transparency.
Capturing and assembling big data
According to Moores Law, the number of transistors in integrated circuits has doubled
every two years over the history of computing hardware. Thats to say, the ability of
our computer systems to process and hold information is growing exponentially, and is
continuing to do so. As memory and processing speed gets faster, so the amount of data
we can interpret gets bigger, leading to new advances in areas such as genomics, nuclear
physics and biomedical research. The data science historian George Dyson describes the
big data phenomenon as when the human cost of throwing away data became higher than
the machine cost of storing it.19
Presently most formal knowledge in healthcare is based on data gathered in the research
setting. However, while more data does not necessarily equate to more knowledge, there is
a clear opportunity to learn from the routine practice of medicine, and more broadly from
patients whole lives. Progress in understanding has often grown alongside changes in our
ability to gather and analyse data. This is perhaps best seen as an increase in resolution; we
are looking at the same phenomena, but seeing in more detail.
With advances in mobile phones and homebased sensing devices, we also have access
to much richer sources of longitudinal data, allowing us to capture more granular data in
real time relating to peoples lived experience of disease, behaviours, physiology, social
networks or health outcomes.20 As a result, much of the opportunity in future will be to
extract knowledge from information gathered outside the research context, in a practical
setting. This data will come from a variety of sources:
Firstly, from existing data that is made more accessible. In the course of their ordinary
clinical interactions, patients and doctors already generate large amounts of data.
Historically much of this has been inaccessible, either held on paper or in databases that
cannot effectively share the information. However this is rapidly changing, with existing
data being able to move around the health system in a much more effective way via open
patient records (anonymised where appropriate). The direction of NHS policy is to open
up far more data and information. The NHS Information Centre has already published data
including presenting and prescription data from all GP practices in England data that was
previously proprietary with the purpose of making this data more widely available for
improvement and innovation.21
Secondly, from rich data recorded as part of ordinary clinical practice. Following promising
results from largescale telehealth trials we know that health outcomes for some people
can be improved through real time monitoring.22 At the moment this is a fairly expensive
exercise, due to the cost of the equipment and support. However if this cost can be
reduced, then real time monitoring of some patients with chronic conditions could become
part of normal practice. For patients at risk of stroke or heart attack, it seems plausible that
real time data could provide crucial early warning signs, as we develop more sophisticated
ways to analyse the data.
Thirdly, patients will record data for their own reasons. PatientsLikeMe is a wellrehearsed
example of this, but it deserves revisiting.
PatientsLikeMe
PatientsLikeMe (PLM) is an online platform where patients with lifealtering
conditions share structured information about their symptoms, treatments and
outcomes. Members can view this data as individuallevel graphical health profiles
and aggregated reports, discuss health and offer and receive support on forums and
through private messages. As of August 2012, PatientsLikeMe has 160,000 users with
over 1,000 conditions being recorded. Similar to other online communities, members
of PatientsLikeMe offer one another support based on their own personal experience,
and advise each other on how to improve daily life and longterm health outcomes.
What distinguishes this patient platform from others is that members tailor questions
and consult each other by referring to concrete data displayed for each member.
The members of PatientsLikeMe dont simply share their experiences anecdotally;
they quantify them, breaking down their symptoms into hard data, inputting their
condition and treatment details including dosage, efficacy and side effects, along
with information on symptom history, tracking and a variety of biological information.
All this data is entered onto simple dataentry forms and turned into graphs and
charts via the sites software. Patients are able to share their experience using
patientreported outcomes, find other patients like them matched to demographic
and clinical characteristics, and learn from the aggregated data reports of others
to improve their outcomes. The site offers two types of data both individual and
aggregated thus giving patients insight into both the specificities and full variety of
experiences associated with interventions, and not only what happens on average,
as is often the case.
PatientsLikeMe also publish this data in peerreviewed journals, such as in a study

published in Nature Biotechnology to explore the impact of the drug lithium on
ALS patients using data collected from the open platform and a patientmatching
algorithm.
PatientsLikeMe allows patients to record the progression of their condition, and compare
that to a projected prognosis. This puts them in the position of being able to judge whether
changes in their treatment, diet or lifestyle or other factors are likely to be making any
difference to the progression of their condition. While the validity of these judgements
does depend crucially on the quality of the prognostic model, and the good sense of
patients, PatientsLikeMe does clearly demonstrate the enthusiasm of patients for recording
quite detailed data about their treatments, lives, and progress, where that data can be put
to use by them. In the areas where PatientsLikeMe is strongest (for example amyotrophic
lateral sclerosis (ALS)) it now has more data on patient progression than any clinical trial
that has ever been conducted, though with the caveat that there may be an element of bias
with a subset of patients who enjoy recording their data.
Similar sites allow individuals to record their mood (e.g. Moodscope), mobility (e.g.
MyClinical Outcomes), and a host of other details. And indeed for some individuals there
seems almost no limit to what they are prepared to record.
Quantified Self
The Quantified Self or movement refers to a growing grassroots network of
individuals who monitor, track and quantify their health and wellbeing using a
variety of metrics and conducted through monitoring devices. Adherents utilise
embedded devices, such as the FitBit activity monitor, to collect detailed data on
their health and mood with the goal of optimising their wellbeing. This information
is then uploaded to online platforms which contextualise the data and provide
advice based on it. These emerging technologies are prototypical models of how
individuals can accumulate and share information on their health independent of
formal institutions, and demonstrate the democratising impact of personalised health
technologies on individual health and wellbeing.
As an example of a Quantified Self technology, Jawbone UP is a wristband which

interacts with an iPhone app to record information on an individuals activity, sleep
quality and eating habits. In utilising the online platform, users can communicate
with each other and share their quantifying experiences. Jawbone UP and other
devices of this sort in combination with smart phones, represent a newly affordable
purchase for those interested in monitoring their health. They also represent the
advent of widespread datasharing which could be hugely significant for healthcare.
Patients will also gather data that is not directly related to a health condition, but
concerns their health or wellbeing in a general sense. An interesting example of this is
Myfitnesspal.23 This app allows the user to search for the calorie content of everything they
eat, and to record their exercise regime, giving a net calorie intake or deficit for the day.
This can be compared with their weight loss goals to help them reach a target weight.
This application is backed up by a large database of food products (more than 1.9 million).
The app can be combined with a wearable sensor (FitBit) which tracks how many steps
you take per day, altitude climbed, and sleep cycles. This site is therefore gathering a rich
dataset about peoples diet and exercise. Correctly analysed, for example using Mendelian
randomisation methods, this database has the potential to get closer to an answer about
what works with respect to maintaining a healthy weight, a question that presently lacks a
rigorous answer.24 Imagine dietary advice based on real successes and failures, rather than
the subjective, biased and unsupported advice that tends to dominate the dieting industry
at the moment.
Another interesting dimension to big data in healthcare will be the social connections
of patients. We now know that certain health risks are affected by the social networks
in which individuals are embedded, for example obesity.25 Facebook and other social
networking sites can give us a clear picture of the social connectedness of individuals.
Geotagging and apps such as foursquare can tell us how people are interacting
geographically.26 Personal finance apps such as Mint can let us know how people spend
their money.27 Presently the data to draw these sorts of conclusions about social influences
comes from a small number of very detailed cohort studies, such as the Marmot study on
the impact of stress at work on health. Big data could produce information of even greater
richness, on a far larger scale that could inform new methods for analysing cause and
effect, such as Mendelian randomisation or wellcalibrated prediction rules (albeit with
challenging applications in practice).
Overall, then, we are now building datasets that are extremely large and contain granular
data about real patients, including a much fuller picture of their lives, including diet,
exercise, and other lifestyle factors, and showing both the treatment they received and the
outcomes they achieved. Of course it must be noted that these data relate to treatments
that are not randomly allocated, which present considerable issues to which we return
below.
Collaborative structuring and analysis to refine data into knowledge
People have always collaborated as a means to solve problems and create new knowledge.
Yet todays digital tools and processes make it possible to collaborate faster and at a much
greater scale than ever before. Online platforms allow people from all over the world to
come together and solve problems or contribute to a growing knowledge base in a more
distributed way, allowing us to work together to refine, interpret and make use of complex
information.28
Though there are still relatively few examples, collaborative platforms that allow individuals
within or beyond the health system to input structured information relating to health
outcomes and that facilitate problem solving in relation to health research challenges,
demonstrate the opportunity for a commons approach to health knowledge.
An example of Internetbased professional collaborative rating in healthcare is MORE, the

McMaster Online Rating of Evidence tool used to select the most appropriate content for
the EvidenceBased series of journals and the McMaster PLUS evidence service.29
McMaster Online Rating

The McMaster Online Rating system is an online information management system that
assesses the quality of evidence resulting from medical research, and tailors its alerts
to newly published evidence sent to clinicians according to their clinical disciplines.
The system hierarchically organises medical information using a qualitative metric,
and alerts physicians to the latest and most relevant developments in their field.
Within the context of the health commons and the future of healthcare more broadly,
its significance lies in the fact that it represents a means of applying precision
medicine. By utilising the specified evidence provided by the system, physicians can
personalise their treatments and apply more specialised research results to treat
individual patients a step away from populationbased medicine and towards
personalised healthcare.
In MORE, the central office prefilters the huge number of new publications in the 110
highest yielding journals potentially relevant to informing clinical decisions, using relevance
and methodology criteria. The results of this stage usually about 6 per cent of each
months publications are then divided up by clinical specialty and extracts sent to
selected individuals drawn from a global network of over 8,000 practitioners for final rating
(including 4,000 physicians across 61 disciplines, 3,000 nurses across 36 care specialties,
and 1,000 rehabilitation professionals in 21 areas of practice).
The rating criteria are novelty and relevance to practice, and each article is judged
independently by between three and 20 clinicians (i.e. three to four per relevant discipline).
The central office only writes its carefully worded, structured abstract summarising
the key study methods and findings if there is close agreement of all raters on both of
these criteria. Articles that score less highly but retain clinical relevance and interest are
disseminated e.g. through Evidence Updates. Those using MORE to rate articles describing
primary research rate them high for novelty but lower for relevance to practice; for
systematic reviews the reverse is true.
To be successful, other collaborative models will need to build on infrastructure already in

the public domain, complementing and supporting existing efforts at sharing information
whilst drawing relevant data from them. One fundamental characteristic to create an open
and collaborative model is a clear code of rules for accessing and sharing data; a founding
commons consensus, set out in a constitution (such as through social chartering).30 Pre
negotiation of the terms and conditions of access to data, but with enough flexibility to
permit the adaptation of prices and conditions of access to changes in circumstance and
demand, is a critical part of this process. Clarity in this area would provide a major public
and private benefit of the commons model: liberating patients, researchers and health
care practitioners from the legal blocks which hinder progress in traditional collaborative
medicine.31
FoldIt
FoldIt is an online computer game that engages its players in a collaborative
research project about the dynamics of protein folding prediction and the degree to
which human protein folders are more effective than computers at protein structure
prediction. Puzzles on FoldIt ask players to predict the structure of a protein based
on its amino acid structure, with a goal of eventually having human folders working
on unknown protein structures and automating the strategies humans come up with
to improve software. Data from the game is captured and used to drive research
into processes for analysing protein structure, in essence becoming an effective
collaborative research platform.
Effective collaboration is made easier by knowledge being structured, with clear and
unambiguous statements of what the relevant data is, and exactly how to draw conclusions
from that data. Algorithms which underlie risk stratification tools are a form of structured
knowledge. Patient histories are fed into it, and a risk rating comes out which helps make
decisions with respect to that patient or group of patients. As long as the prediction
made is accurate (both discriminating and well calibrated) this prognosis might help them
make judgements about whether any changes in their lifestyle or treatment are genuinely
helping.
Another example of structured knowledge is clinical decision support. This has been
an active area of research and innovation for many years, with welldeveloped ways of
translating clinical guidelines into systems which automatically read off patient data and
apply relevant guidelines to produce contextual and helpful advice for the clinician in
deciding on treatment. Through sophisticated languages such as PROFORMA we are
already able to represent clinical guidelines in this way, but usage is still not mainstream.32
Effective collaboration depends on knowledge being structured in a way that means

the most relevant data is inputted and produced. Structured knowledge is analogous to
structured data. For example, the algorithms which underlie risk stratification tools are
a form of structured knowledge. Patient histories are fed into the algorithm, and a risk
rating comes out which helps make decisions with respect to that patient or group of
patients. As long as the prediction is accurate (both discriminating and well calibrated) this
prognosis might help them make judgements about whether any changes in their lifestyle
or treatment are genuinely helping.
As this knowledge takes the form of explicit links between input and output, the logic and
probabilities involved can be exposed for anyone to see, and more importantly, to adapt.
Structuring knowledge in this way makes largescale collaboration possible, opening out
the process of encoding clinical knowledge into treatment to a wider range of participants
to propose improvements to the model, or adapt it to a slightly different patient group.
There are already examples of this in practice; PatientsLikeMe has published over 27
research studies based on data collected by patients.33
Openness and transparency
We are experiencing a substantial cultural and technological shift towards much greater
openness and transparency in what data we share and expect to be shared. Ubiquitous
social networks and social media platforms make more information more openly available.
Governments, businesses, research institutions and public services are all increasingly
making commitments to transparency and opening up their data for others to use.34
Confidentiality of medical data and privacy concerns about identifiable data are an
important consideration in health, concerns which are at odds with the concept of a
knowledge commons that is defined by openness. By law in the UK all medical information
that can be linked with a specific individual is treated as highly confidential, and cannot
be processed (held or used) without the explicit and informed consent of the individual.35
Less confidential but still identifiable data cannot be used or shared outside of its limited
original purpose. Where the data cannot be linked to a specific individual, typically
because it has been anonymised in some way, the rules on the research use of this data are
becoming more relaxed.
The NHS Information Strategy revealed this Governments intention to work towards
greater transparency with health data, and the NHS Information Centre is obliged to
publish increasing amounts of data on performance, clinical outcomes, prescriptions and
complaints. This agenda is tightly bound to the intention to make each patient records
information available to that patient by the end of this parliamentary term, and to make
anonymised patient data open to researchers. However, openness is more than just
the release of data. This needs to be matched with a cultural shift in the way in which
healthcare professionals and patients can access information, with the right incentives to
record relevant data at the point of care and the right skills to interpret data correctly.36
Such a culture of openness has significant implications for research. As suggested above,
one of the issues in assuring the reliability of medical knowledge is publication bias, and
the commercial and academic tendency to promote positive rather than negative results.
This lack of openness limits our ability to collectively learn from both positive and negative
findings, and the currency of publication and citation creates the incentive for researchers
to withhold rather than share data and discovery. Open Clinical is an example of a research
project based on more of a commons approach, where knowledge bases for protocol
based decision support systems is shared and learning is collaborative, with a goal of
optimising knowledge management in medical practice.37
Open Clinical
Open Clinical is an online collaborative effort involving organisations, companies
and individuals operating through open access websites with the goal of optimising
knowledge management through decision support in medical practice. It operates
primarily through a portal providing access to developments in medical decision
making, journals, directories of protocol representation languages and relevant
products, and encourages discussion and contribution from all members. Open
Clinical is an excellent example of a collaborative effort involving researchers,
industry and healthcare professionals, and provides a model for future efforts in this
area. Important characteristics include simplicity and easeofaccess, encouraging
contribution and perpetual refinement of its information, and its niche of delivering
specialised information to a diverse audience of both patients and professionals.
What might a health knowledge commons mean in practice?

If more and better data, collaboration tools and a principle of openness are the
components of a knowledge commons in health, what are the outcomes? Where might a
commons approach hold particular value and what innovation would it lead to? We think
there are four important areas where this approach to orchestrating knowledge in health
will create value:
Research and discovery both in evaluation of the significance of clinical findings for
diagnosis, prognosis and the efficacy of treatment and in generating hypotheses and
prompts for new discoveries.
Diagnosis and prognosis to more accurately predict the progression of disease, and
diagnosis and treatment based on real time, dynamic information.
Changing relationships when people feel better informed, they are able to take
more control. The ability to not only access but also generate useful knowledge about
health is an important prerequisite for a more balanced and equal relationship between
patients and clinicians, and the spread of decision support tools.
Prevention and wellness more accurate knowledge about which habits and risk
factors affect our health should encourage more selfcare and personal responsibility,
shifting the focus of health from illness and treatment to wellness and prevention.
Research and discovery
The simplest way to use data to improve our knowledge is to confirm that new treatments
are fulfilling the potential they showed in clinical trials. The generalisability of the results
of even wellregarded clinical trials is far from perfect. While this is to be expected, it does
mean that drugs and treatments can be adopted incorrectly. Take the example of SSRIs
such as Prozac. The most authoritative metaanalysis suggests that these have no clinically
significant effect on any but the most depressed patients.38 If this is true, many billions of
pounds have been wasted globally over the last few decades. In a knowledge commons, it
would be possible to know much sooner what was happening to patients prescribed SSRIs,
and to spot early on that trial results were not being replicated in practice.
More and better data as generated through a knowledge commons could also generate
new hypotheses or provide prompts for new discoveries. The history of medicine is
replete with examples of clinical discoveries which result from a correlation being spotted
somewhat serendipitously. For example, the initial insight that led to the discovery
of steroids happened when Philip Hench of the Mayo Clinic noticed the remission of
rheumatoid arthritis in those who were suffering from jaundice.39 This led to a hunt over
many decades for the agent that caused this remission, and eventually to the discovery
of steroids. Medical treatments have often emerged from examining a correlation that
seems surprising and potentially beneficial, and isolating the drug that causes it; with the
treatment coming in advance of any understanding of the underlying mechanism. In fact
examples of new treatments being designed, based on a fundamental understanding of
disease processes are relatively few and far between. The promise of genomics was to
facilitate exactly this process, but progress has not lived up to expectations.40
With the increasing specialisation of medicine and research, there are a declining number
of clinicians who both carry out research and treat patients.41 People who are both in a
position to spot an interesting correlation, and are primed to recognise its significance are
relatively low in number. The medical profession is increasingly split between researchers
and practitioners, and research into basic medical science is incentivised by the publication
of articles, not the creation of new treatments.
Open and linked data could help the research community and patients to identify any
number of surprising correlations which are presently hidden from them, and will probably
do so more effectively than the previous serendipitous system. It could also include data on
a broader range of factors, including social connection, diet, exercise and stress, which are
hard to include in a trial setting. These correlations will then need to be confirmed in more
controlled settings, but even if only a fraction prove to be robust, this will be useful.
The points made hitherto are about prompts for further research, rather than anything that
can provide evidence of a causal link. A more ambitious possibility is to add to learning
about the causes of disease from routine data. As has been mentioned, treatment is often
a trial and error process. Patients are individuals, who often have multiple comorbidities.
Further, standard medical knowledge is far from a complete recipe, and in some conditions
doctors often go beyond established protocols. Treatment effects are somewhat uncertain,
and there is a considerable need for adjustment and iteration before the correct course is
found. Indeed the negotiation of this process is one of the key skills of the physician. What
this means is that there are thousands of informal, undocumented microexperiments
going on in surgeries and hospitals across the country every day. However, we do not
extract any learning from these for society, so any learning stays with the individual
physician. As this routine information is more frequently recorded in a digital format
that can be shared and interrogated, potentially this learning can be released. However,
routine data means that treatment allocation is of course not randomised, leading to
biases, including placebo effects, confounding by indication, Simpsons paradox, and other
problems.42
There are two approaches that could be adopted to draw more robust conclusions from
routine practice. The first is to use alternative methods which do not require randomisation,
such as instrumental variable methods (i.e. finding a variable which affects the availability
of the intervention of interest, but cannot affect the dependent variable in any other way
also called Mendelian randomisation) or using welldeveloped predictive models to
compare against the actual outcome in a particular condition.
An example of instrumental variable method is a study of the effect of bone marrow

transplant on the survival of children with AML (leukaemia). This was tested by examining
the correlation between having a live related sibling and the childs survival. This is because
having a live related sibling affects the availability of a bone marrow transplant, but cannot
effect a childs survival in any other way. It is therefore legitimate to conclude that any
correlation between having a live related sibling and a childs survival is due to the efficacy
of bone marrow transplants.
The alternative approach involves integrating a research approach with routine practice.
One way of doing this is N of 1 trials.43 In an N of 1 trial, the clinician and a patient with a
longterm condition agree on the two drugs (or drug and placebo) to compare, and on
the main symptoms or other outcome measures to be used as a metric of success. The
drugs are then changed in a randomised alternating pattern unknown to both patient
and doctor over a period of between five and ten weeks, while the patient records their
symptoms as often as needed. The trial is then unblinded and the drug which results in
the best outcome for that patient is identified. An app has been developed to support this
process, and there is work on synthesising the results of many N of 1 trials to create new
insights about the probability that a drug will help an individual patient reach their goals.
If many more clinicians collectively promoted and participated in N of 1 trials, thus earning
them membership of the knowledge commons, the results could be aggregated into a
statistically respectable and significant whole. This much more disciplined approach to off
label treatment could result in significantly faster progress in addressing clinical research
questions.
One advantage of this approach is that it would cope much more efficiently with diseases
that are much more variable between patients in their responses to treatment (or are really
a bundle of diseases, disguised under a single name). Consider the following quote from
Cancer Commons, a knowledge commons for cancer.
Modern molecular biology supports the hypothesis that cancer is actually hundreds
or thousands of rare diseases, and that every patients tumour is, to some extent,
unique. Although there is a rapidly growing arsenal of targeted cancer therapies that
can be highly effective in specific subpopulations, especially when used in rational
combinations to block complementary pathways, the pharmaceutical industry
continues to rely on largescale randomised clinical trials that test drugs individually
in heterogeneous populations. Such trials are an extremely inefficient strategy for
searching the combinational treatment space, and capture only a small portion of the
data needed to predict individual treatment responses. On the other hand, an estimated
70 per cent of all cancer drugs are used offlabel in cocktails based on each individual
physicians experience, as if the nations 30,000 oncologists are engaged in a gigantic
uncontrolled and unobserved experiment, involving hundreds of thousands of patients
suffering from an undetermined number of diseases. These informal experiments could
provide the basis for what amounts to a giant adaptive search for better treatments, if
only the genomic and outcomes data could be captured and analysed, and the findings
integrated and disseminated.44
While the extent to which this description of cancer is accurate is certainly debatable, it
perfectly plausible that some cancers and other conditions are of this type. If this is the
case, we need to learn from routine treatment if medical progress is to be sustained.
Cancer Commons
Cancer Commons is an online service designed to provide a more specialised
approach to the treatment of cancer. The goals of the programme are to provide a
personalised approach to cancer treatment, to accumulate knowledge on individual
cancers, and to disseminate the knowledge through an open source collaborative
framework. It is structured so as to connect advisory boards of specialists
drawing on collaboratively obtained knowledge with patients through Webbased
applications. It demonstrates that specialist advice can be accessed by patients
without direct institutional access, effectively; by utilising the Internet, Cancer
Commons provides an additional medium of support and expertise to patients
through bypassing the traditional patientdoctor relationship. This is a model which
could be emulated for other medical problems.
Prognostic models and citizen clinical science
In certain fields of applied knowledge there is the opportunity to use a knowledge

commons approach to establish useful understanding without the need for randomisation.
One example is prognostic models that accurately predict the progression of disease or
response to therapy.
Prognostic models are essentially equations or algorithms that link relevant facts about
the patient to an outcome, such as the trajectory of their conditions. They can thus be
encoded as small pieces of software. A genuine commons would mean that the logic and
code behind these models was shared, allowing the models to be continually updated
and iterated as new data and insights come in, analogous to the open source software
movement. In a big data era we are in possession of a large amount of granular data about
patients recovery/progression, as well as relevant background data such as age, time
since diagnosis etc. We can mine this data to build a model of disease progression, and to
predict individualised results. As can be seen from the PatientsLikeMe example, there is a
clear appetite for this among certain patient groups for their own use.
Accurate, wellcalibrated prognostic models would be useful to help patients to plan their
lives, and also to check if they are responding to treatment as expected. As these models
evolve, they could begin to offer predictions about response to treatment for individual
patients. Health systems and patients would find it very useful to be able to identify who
will respond to standard treatment, who will need a more customised approach, and who
needs to understand that little can be done for them. This would effectively be a much
more sophisticated development of present predictive risk models such as PARR (which
only predicts risk of hospital admission), but with the addition identifying those most likely
to be helped, and how. Clearly this could enable and impact the targeting and rationing
of services, and is thus politically sensitive. However, given real and growing financial
constraints, health systems must put their resources where they do the most good.
Disagreements about what outcomes to predict and which modelling methods to use
might result in several competing versions, until data decided on the best. A similar process
could be used for diagnostic models, and for treatment and selfmanagement protocols.
An important potential research contribution of prognostic models is to reduce our

reliance on clinical trials in some situations. If we can develop a wellcalibrated model
that accurately predicts outcomes in a defined patient group, we can then potentially use
its predictions to substitute for controls in a randomised trial. The reasoning is that in a
trial, we compare outcomes in patients exposed to the drug against those not exposed
but otherwise identical. Randomisation is the best way to achieve this, though Mendelian
randomisation (instrumental variable analysis) is an emerging alternative. If, however, we
can accurately predict what would have happened to each patient had they not been
treated with the new drug, we have a kind of evaluation machine which eliminates the
need for a controlled experiment. This kind of approach has rarely been used in the past
due to the lack of wellcalibrated prediction rules, but could become more common in
future with the increasing availability of comprehensive high quality longitudinal datasets
on people living with longterm conditions.45
Large collaborative platforms could also allow patients themselves to systematically

research ways of coping with their own health in the real world. Patients would experiment
with explicit protocols for managing their condition, and in groups create results that were
at least suggestive of a causal relationship. The relationship between research and selfcare
here would be very close, and would amount to an iterative citizen science of longterm
conditions, embedded in and directly benefitting the global patient community. Clinicians
and central organisations might have quite a small part to play, with most decisions
devolved to local patient groups, who then propose apparently successful new models to
the wider patient community for further randomised testing.
Given the numbers of people with the top five longterm conditions, it might take only
days to recruit the thousands of patients needed to conduct a confirmatory randomised
trial in such a network, though followup of each patient might take several months or a
year. However, since the data will be collected electronically and quality assured by instant
querying of missing values, analysis could start immediately after the trial ends and the
extended cycle time for conventional trials would be shortened to just a few weeks longer
than the followup period. This kind of model already underlies existing platforms such as
PatientsLikeMe, and has clear traction with patients. As they become more exact, they would
be very useful in allocating resources. If they reached a sufficiently high level of exactitude,
they could be compared with outcome data in an evaluation machine approach, and used
to make causal claims about the efficacy of novel forms of treatment compared to a normal
outcome.
Another area of interest would be shortterm predictive models for use by patients. A
good example of information that is not valued clinically but which could be immensely
useful to patients is Will I be ill or well tomorrow? Even if there is little that can done
about the answer, the autonomy and control of ones daytoday life that comes from
knowing if tomorrow is going to be a bad day is hugely valuable. Reducing the anxiety
and uncertainty of ill health, and planning a life around their illness are of enormous value
to the patient, perhaps more so than many of the outcomes on which the healthcare
system is incentivised.
Diagnostic models
Diagnosis has usually been based on a thorough history and examination of the patient.
It is an underrated triumph of the medical profession that diagnosis can be done so
well, given the relatively thin information on which it is based: we do not have conscious
awareness of much of what is going on in our bodies; we often cannot recall that which
we are aware of when the time comes to speak to a doctor; it is hard to communicate this
sort of information with any precision; and finally our recollection is inaccurate. Diagnosis is
therefore a highly expert skill.
Richer data could provide a productive basis for building diagnostic models, based on a
host of uptodate information about the patient (building on Sacketts Rational Clinical
Exam model).46 Initially these models might be an opportunity for a clinician to clarify a
difficult diagnosis wear this sensor for a week and the data will help me decide what the
problem is. As these models reached a higher level of sophistication and precision they
could become a new form of screening, run over large datasets to spot health problems
early.
Taking this processing of real time information a step forward would involve moving
from real time prognosis to dynamic treatment selection. Real time information about
the patient would inform which treatment is selected and when. This is essentially
what home telemonitoring or telehealth should enable, however we are still at the early
stages of understanding who will use telehealth, what kinds of data to gather, and how
to process and respond to it. Again, the protocols here could be evolved communally,
with their efficacy compared. Richer data can enhance this process. A simple example
is the Met Offices Health Outlook app. This alerts COPD patients to upcoming changes
in temperature, which are known to trigger exacerbations, allowing patients to take
preventative actions.
As with many of the trends identified in this paper, we see this approach being as useful
for the patient community as for the clinical one. Digital technology gives patients who are
struggling with a longterm condition the opportunity to take a collaborative approach to
understanding what steps they can take to improve their lives and to live better with their
conditions. While online forums and discussion groups represent a significant improvement
over what was available to patients previously, they are still long on anecdote and relatively
short on fact. Patients who are recording how their diet, exercise, environment, social
life, or treatment is affecting their health are in a position to move beyond exchanging
anecdotes and begin to establish facts. By aggregating and analysing their data in the ways
we describe they have a chance to understand what is helping them and what is not.
26 Doctor Know: a knowledge commons in health Doctor Know: a knowledge commons in health 27
Figure 3: COPD Care Checklist from the Chronic Obstructive Pulmonary Disease (COPD)
project47
Annual COPD You should see your doctor or nurse at least This chart shows Contact your practice 1. Annual
1 review
once a year for a COPD review. During your
review, you should be asked how well your
medicines are helping with your symptoms
the date you
should book
your next annual
Last Review(Now)
for your next annual
review in Nov 2012
COPD review
___________
___________
You have read a review in and whether you have had any side effects. review for ___________
the last 12 months. ___________
___________
Now
2. Spirometry
Spirometry test Your diagnosis of COPD should be confirmed by What is a
2 a postbronchodilator spirometry, also known as spirometry test? test

___________
You have not had this a blowtest. This test checks how well your lungs It is when you take an
___________
test and should contact work by measuring the amount of air you can blow inhaler to open up
out. This helps to decide upon the treatment your your airways, before ___________
your surgery at the first ___________
doctor should offer. taking a deep breath
opportunity. 1. Take an 2. Blow into 3. Measures
and blowing as hard as ___________
inhaler sensor lung capacity
you can into a sensor.
If you are a smoker and you 3. Stopping

Stopping smoking Giving up smoking and sticking to it is extremely
3
tried stopping but couldnt, Talk to Phone Ask your NHS smoking
important if you have COPD. Your doctor should ___________
your surgery should be the your your chemist Smoking
Your records show you encourage and help you to do this.
surgery Local about a helpline ___________
first point of contact for
are currently a smoker deciding which method of to find out NHS nicotine 0800 ___________
and are trying to quit. quitting would most help you. what help RED quit alternative 022 4332 ___________
They can direct you to the you can on: 0208
___________
appropriate service. receive 926 5275
Inhaler technique The medicines you use depend on how severe your There are different You may consider: 4. Inhaler
4 Your records show you

are currently a smoker
COPD is, how it is effecting your everyday life, and
what side effects you may experience. You should
only be given an inhaler once youve been shown
types of inhalers.
Please make sure
your doctor/nurse
technique
___________
___________
and are trying to quit. how to use it and you are confident that you can has shown you the ___________
use it properly. Your technique should be checked right way of using it. ___________
annually. If you need a refresher, please contact Should you How hard How long should
___________
your surgery. breath out should you be you hold your
before use? breathing in? breath?
Pulmonary Certain patients could benefit from a pulmonary What is a pulmonary 5. Pulmonary
5 rehabilitation
rehabilitation course. It is a programme of care
designed for your individual needs. During the
twice weekly session, for 68 weeks, you work with
rehabilitation programme?
It helps to support you to
live better with COPD.
Ask your GP
if you are
suitable for a
If you are
suitable
your GP will
rehabilitation
___________
___________
You havent been a healthcare professional in your local area to help It includes breathing ___________
pulmonary refer you
referred to a course you to make the most of your physical abilities and techniques, exercise, rehabilitation ___________
in your local area. to become as independent as possible. useful information and programme ___________
advice on your diet.
Support with Sometimes your symptoms may become Whats a rescue pack? 6. Self
6 selfmanagement
particularly severe. These are called exacerbations
or flareups. You should be given advice about
how to spot these early and prevent them from
It contains antibodies and
steroid tablets so that you
can start these as soon as
-management
___________
___________
Contact your GP to getting worse. You may be given a rescue pack to possible when your COPD ___________
talk about getting a keep at home to help prevent exacerbations. starts getting worse. Bad Rescue Talk with ___________
selfmanagement plan breathing pack your GP ___________
with a rescue pack.
Changing relationships with patients
When people feel better informed, they are able to take more control. The ability to not
only access but also generate useful knowledge about health is an important prerequisite
to a more balanced and equal relationship between patients and clinicians, and the spread
of decision support tools.
Pilot projects focusing on coproduction in health, such as Nestas People Powered

Health programme, have shown how creating an equal relationship between patient and
professional in the design and delivery of care for longterm conditions brings with it a
number of challenges to the system. In particular, challenges around sharing information
and responsibility for care with patients, changing professional culture and helping
clinicians to see the value of working in partnership with patients, and capturing data
relevant both to the person as well as to the health system.
For example, the Year of Care (YOC) diabetes programme introduced a personal care plan
and the sharing of results and other information with the patient prior to consultation. This
has demonstrated the value in engaging patients actively in the decision making process
around their conditions, and providing the relevant information. This value comes from
the perspective of both patients and the professionals in both primary and specialist care
settings.
The dashboard illustrated in Figure 3 which was codesigned with COPD patients to
enable them to better manage their condition and prepare for consultations builds on the
same premise as the Year of Care care plan, but takes this one step further by providing
patients with personalised information about their condition, for example how to work
an inhaler, what services are available or how to make the most of a GP consultation. To
encourage patients to engage with preventative services, the guide goes as far as breaking
down the cost of these and emergency services by for example outlining the cost
between an inhaler (38) and GP emergency callout (128). However, whilst this points
to the potential value in sharing data and engaging patients more in the management
of their own care, it also demonstrates the limits of current approaches to engaging
patients actively in managing their own condition. Firstly, the Year of Care dashboard is
paper based, making it hard to capture, track and share usage data between people and
organisations over time. This also still requires a consultation to take effect, rather than a
digitalised version facilitating new kinds of interaction between patient and professional.
Secondly, this does not capture the whole variety of information relevant to a patients life.
Patients engage with multiple services and relationships that directly or indirectly affect
their condition and care. This might include primary and secondary care, communitybased
options such as time banks, walking groups or peer networks, as well as nonhealth related
services such as social care, housing or debt advice. Health data is rarely captured in all of
these settings and is rarely shared and integrated between agencies.
Therefore, an ambitious but important area for development on the basis of a health
knowledge commons is decision aids, to facilitate more informed and equal interaction
between patients and clinicians. Clinical decision support systems analyse the patient
record, help elicit patient preferences and suggest treatment options to the clinician and
patient, based on encoded guideline recommendations. Decision support tools can use
software to allow access to the most uptodate research about a particular issue, guide
treatment decisions based on specific data relevant to the patient and can facilitate shared
decision making by surfacing and translating different interpretations and knowledge.
The lack of relevant digital infrastructure including a lack of electronic patient records,
tools for electronic ordering of tests or drugs and encoded guideline knowledge mean
that decision support systems remain difficult to achieve. Trials show that we do not yet
know how to design decision support systems for patient benefits to accrue consistently,
not to mention the changes in clinical practice that would need to sit alongside. However,
larger quantities of real time information and growing demand from increasingly articulate
patients to have an opportunity to analyse and make choices based on their own data,
demonstrate the potential for innovation in this area.
Prevention and wellness
Access to more reliable, specific and actionable health knowledge could also have
significant potential for encouraging prevention of disease, both in preventing the
development of an existing condition or in encouraging positive choices to maintain
wellness. As prevention is necessary before an illness is fully manifested, relevant
knowledge in this instance relates to potential precursors of dysfunction of health or risk
and protective factors, respectively.48
A health knowledge commons should therefore be relevant and accessible to people not
currently within the health system, whether in informing diet or exercise, tracking health
improvements or personal health risks, or in simply encouraging a mindset of taking care
of ones wellness and mental and physical health. Taltioni is one example of a knowledge
platform that allows citizens to monitor their health in this way.
Taltioni
Taltioni is citizenled system for holding health information developed by the
Finnish Innovation Fund Sitra, to provide a new health knowledge system for the
population of Finland. Information is owned entirely by citizens, who have the choice
to upload their own data in a structured way and the choice to share this information
with relevant health providers or researchers as they please. The infrastructure is
being developed as a cooperative by Sitra, a Finnish telecommunications agency
and health service providers, and the ambition is to have over 300,000 citizens
contributing to the platform in the next two years.
Taltioni is intentionally for citizens, not patients. In that sense, it is a platform that
encourages wellness rather than manages health. In encouraging people to record
and review their demographic, health and lifestyle data, the aim is to support people
to take better care of their own health and fitness. Yet even if only a minority of
Taltionis users choose to share their information for research and with health service
providers, this nonetheless provides considerable resource for furthering research
and innovation in the market of health services in Finland.
4 What next? Making a

knowledge commons happen
Knowledge and knowledge systems are not easy concepts to grasp. The issues involved in
developing a commons approach to health knowledge also mean contending with difficult
issues around patient confidentiality and the sensitivity of health data; the question of
who has access to data and for what intent. The nature of a commons is that it is open to
everyone to generate value from shared resources, prompting questions about the ethics
of using patient data or health knowledge for private profit or controversial trials. Public
concerns regarding the use of health data and the deepset culture of clinical practice are
not trivial challenges to overcome either.
However, the reality is that much of this is already happening. Some patients are
already actively participating in systems like PatientsLikeMe and MyFitnessPal which are
capturing and using data for clinical research and development. There is a burgeoning
market in lifestyle technologies that monitor and feedback our behaviour, such as the
FitBit or Smartphone apps that record our diet, exercise routine and sleep cycle. As our
experiences and expectations of accessing and sharing knowledge in other fields grow
more personalised and synchronous, so our expectations around interaction with the health
system will continue to rise.
Though still far from universal implementation, the premise of open patient records
and of routinely using patient data for research is becoming more widely accepted. In
the context of the NHS, clinical research and disease management systems are already
at a considerable advantage in that research practice is already relatively collaborative
and there exists a sophisticated infrastructure of intermediaries active in disseminating
structured information, such as MyDex.
MyDex
MyDex establishes a personalised system of data ownership and sharing, allowing
individuals to gain control of their own data. It creates a central hub of personal
data for an individual, and has an extensive system of privacy and sharing controls
that facilitate selective direct sharing of information with organisations or other
individuals. This allows the user to maintain more efficient ownership of data, and
also to more comprehensively manage their interactions in terms of information
sharing with external agents and organisations. It is significant in that it provides
an excellent model of the personal ownership of data although it is focused around
confidential personal information, such as passwords, it is not difficult to foresee its
application to medicine and patient records.
The prospect of a health knowledge commons builds on all of this work, and brings it
together into a vision for the health system that is centred on knowledge, and that draws
on the collective intelligence of us all to improve health, wellbeing and productivity.
Developing a health knowledge commons is therefore a systemic innovation, and requires
iterative action in a range of different domains and contexts rather than a centralised
strategy to implement a static approach. Such a systemic innovation requires:
New technologies, products and services e.g. apps, sensors, quantified self
technologies, shared decision making and decision support tools, better automation;
New policies and regulations such as for better structuring of data and clinical
guidelines, open patient records, using anonymised patient data for research, and new
governance frameworks around data access;
New business models, and forms of organisation e.g. collaborative research platforms
such as PatientsLikeMe, open research networks and citizenled models like Taltioni;
Behaviour and culture change e.g. greater openness and transparency, a shift in
relationship between patients and professionals, new skills such as translation and
interpretation of knowledge and aligning of incentives.
The health knowledge commons is not a shortterm objective, but an ambition for how
health will look in 20 or 30 years time. However, there are a number of current initiatives
and actions that relate to this longterm vision that should be supported. This section
will consider what needs to be in place to make a knowledge commons happen, both
in relation to action in the short term and more speculative recommendations for future
innovation.
New technologies, products and services
There has been a flourishing of new healthcare technologies and products that allow for
more widespread creation and application of health knowledge, many of which are detailed
throughout this paper. Technologies such as sensors capable of recording rich data
accurately and effortlessly greatly enhance the opportunity for automated data capture,
overcoming the challenge of incentivising regular checkins. Consumer devices such as
smart phones and games consoles contain sensitive devices which are already being
adapted for health purposes, and healthspecific sensors are also being designed.
However, this remains a new field, and many questions still exist such as when and how
patients are happy to record their own data and to participate in selfmonitoring activity.
Key to the knowledge commons is understanding the personal drive of individuals in
entering their data.
One purpose may be motivational: setting goals, tracking them, and sharing your success,
is an effective motivational tool. A second opportunity would be knowledge about how to
carry out a complex task. For example, motionsensitive games consoles such as the Wii
and Microsoft Kinect can detect if a series of rehabilitation exercises are being carried out
correctly, and track improvement in range of motion. Patients whose condition imposes
complex requirements on their lifestyle, such as those with cystic fibrosis, may find a
combination of these two useful. Tracking of diet and physiotherapy may help a patient
stick to a demanding regimen, and advice on how to carry out complex exercises could
be very useful. An important area for research and development is to experiment with
developing communities that successfully collect large amounts of data.
Advances in digital technology combined with minor changes in routine data collection
also offer the opportunity for generating more robust data. N of 1 trials have the potential
to both generate valuable research, and improve daytoday patient experience. They
might even help both patients and primary care clinicians better understand the value
of randomisation and increase recruitment to clinical trials. We would like to see active
support for demonstration projects around N of 1 trials.
Another important area for development is in decision support tools, for both clinicians
and patients. Active user communities such as PatientsLikeMe and MyFitnessPal translate
uptodate individual data into guidance that patients believe is useful in their daytoday
lives. This guidance can be prognostic, or it can suggest dynamic selfcare options. The
data can be used to iteratively improve the prognostic or treatment options. Similarly,
clinicians and managers willingly enter data when that data gives them something useful
in return. Finding out what kinds of information are useful to them, when and how to
generate it, should be an active area of research and innovation. Simple prognostic models,
of which several are available (for example in stroke), could motivate patients to share their
own data, and allow for iterative improvement of the prognostic model.
As an area for future enquiry, we should explore how these prognostic models could
become more reliable, to the extent that they might be useable to judge the efficacy of
new treatments. This would open up a very different and much more rapid treatment
discovery process. We would like to see more research funding for prognosis, both for the
development of specific prognostic models and to develop better understanding of the
psychology of prognosis and the methodology of developing reliable, useful prognostic
models/clinical prediction rules.
New policies and regulations
Realising the potential of data from more dispersed sources means developing common
standards and formats to allow data to be combined, shared and analysed effectively
(almost as a precondition). If the data recorded on different GP systems, pieces of
telehealth equipment or on websites is incompatible in format or structure then the
potential for integration is lost. The NHS is currently promoting common standards through
the interoperability toolkit; however there is still much ground to cover here, with huge
amounts of legacy data held in systems which are not compliant.
Data and knowledge from research can also be structured in a much more helpful way. In
many areas of science (e.g. genetics), it is standard practice for researchers to deposit their
data into repositories that other researchers can access. However, this is rarely the case
with clinical research and should become part of standard practice. This should include
data from trials that were unsuccessful and were not published, helping to counteract
publication bias. Trial results could also benefit from structuring. It should be easy and
quick to call up the conclusion of a range of different trials on a given drug or treatment,
rather than comb through large sets of individual papers to compare results. The ESRC
already insists on data archiving as a condition of funding. Medical research funders such
as the Wellcome Trust and the Medical Research Council should insist on data archiving as
a condition of funding.
There is also an opportunity to structure knowledge, which is presently not well recorded.
For example, case studies of successful implementations of new techniques or technologies
are not centrally collected, and are hard to find. One way to promote knowledge sharing
would be for the National Institute for Health Research (NIHR) to incentivise researchers
to systematically structure and archive implementation case studies, so we can learn more
about how innovations are applied in practice.
Clinical guidelines would also benefit from common standards for structuring their
recommendations. This would allow integration into decision support systems, and
potentially allow patients to understand when their care had departed from best practice.
We recommend that all publically funded clinical guideline producers such as NICE and
SIGN move towards capturing their recommendation in a standard structured, coded
knowledge base form, rather than as text.
A knowledge commons also requires that people have the right to access their own data in
a standardised format, and to share it as they chose. Where data belongs to the individual,
this is about ensuring that they have the right to use it as they wish, bar exceptional
circumstances. This right is only meaningful if the data can be accessed in a way that is
useable by individuals, and in the digital age this means structured, machinereadable data.
This creates room for innovation by allowing those with a new way of making data useful
to gather the data, independently of existing structures. There should be a policy incentive
to promote patient data being structured in a way that allows it to be shared and analysed.
We suggest this right be enshrined in the NHS constitution, with a target that all patients
should have access to their data in a standard electronic format by 2018.
Some versions of the knowledge commons will involve commons members changing their
disease management methodologies and observing the results. Often this could be no
more dramatic than what normally happens to a clinicians practice, as they become more
experienced. Or it could be a more radical change that requires careful consideration and
evaluation. It is important that this distinction is observed, and that the mere fact that data
is being systematically recorded should not trigger a lengthy ethics process. Part of the
opportunity in this model is to iterate based on continuous feedback, which this process
would make impossible. Oversight is necessary, but must be proportionate. The National
Research Ethics Services Proportionate Review Service is an important development in
this regard.
Finally, without undermining the gold standard of an RCT in most circumstances,

the validity and applicability of alternative methods for extracting firm conclusions
from routine data (e.g. the instrumental variable approach) should be examined in
comparison to trials, to build their credibility in the medical community. Support for such
methodology development by a credible funder such as the MRC or Wellcome Trust
would greatly assist this.
New business models and forms of organisation
Provided the underlying data and knowledge standards and potential software applications
exist, the next development is to create the right incentives and market conditions to
encourage new business models and organisations to take advantage of this potential. We
would suggest that there needs to be some earlystage research funding in the key areas
outlined above:
Creating sensors which effortlessly gather relevant information;
Designing apps that encourage patients, clinicians and managers to enter and share
data;
Perfecting statistical techniques for extracting information from data;
Working with clinicians and patients to gather real time data within a research
methodology;
Better understanding of how to present data to individuals to help them find it quickly
and interpret it without error.
There is also a need to invest in the kind of products for patients which are not directly
connected to biomedical or financial outcomes, but do result in a patient who is
informed and autonomous. There is an opportunity for NIHR working with the National
Commissioning Board (NCB) and other funders such as TSB or EPSRC to commission
a number of knowledge commons projects which gather and process information for
patients, with patient autonomy as a goal.
A softer and cheaper form of influence would be the NHS accrediting certain apps and
communities as being interoperable and clinically sound in their advice. An NHS App
Store would also promote products that support patient autonomy.
Finally, the issue of confidentiality and consent around the use of patient data is currently
under review, but suggests the necessity of a trusted intermediary or third party who
can hold individuals data, collected from various sources, and make that data available
to authorised agents in a way that preserves privacy. Large datasets can be made
available for research purposes with a fresh anonymisation each time, thereby making
deanonymisation more difficult, or they can be analysed only on a virtual workstation
running on a computer with zero export facilities for the raw data, as used by the UK Social
Science data archive,49 Services like Microsoft Health Vault and Mydex are already capable
of performing this role. To accelerate this, standards for compatibility of data and systems
with sites such as these should be enforced by the National Commissioning Board.
Behavioural and cultural change
For clinicians to spend time helping patients understand their condition, making choices
with them, and trusting them to take more responsibility for themselves will require a shift
in mindset amongst clinical professionals. It will involve giving patients more power and
autonomy. While this sounds reasonable, in practice it is likely to encounter significant
resistance, not least as there still remains some scepticism in this area. Clinicians may
filter and edit the information they give to patients. For example, when the possibility of
making a full recovery is low, a frank assessment of the situation can be demotivating for
patients. With patients able, for example, to access a prognosis on their own, such filtering
is no longer possible. While a knowledge commons approach does have some scope for
discretion, it does involve trusting patients with more unfiltered information, and this
may make many clinicians nervous. We recommend that the GMC, Royal Colleges and
other leadership organisations examine the implications for professional roles, education
and training of greater patient autonomy, selfreliance and access to patient data and
knowledge, and seek ways to bring about the necessary culture change.
Similarly patients will take a more active role in determining the research direction. Online
communities of patients have already decided on their own research questions and
gathered data on them, and indeed had the results published in good quality journals. This
has concerned many, as these may not be randomised or controlled trials, and patients
may be exposing themselves to particular risks.
Showing that patients can handle more autonomy safely, and understanding what the
proper limits are can be explored through demonstration projects which are carefully
designed to handle these issues and sensitivities, but which start from a position of
positivity about patient capabilities. Coordinated leadership and vision will be needed
to make this a reality; changing minds and culture are as important as providing new
techniques, tools and platforms. We recommend that NIHR, other medical research
funders and patient groups collaborate to investigate the benefits and limitations of
patientinitiated and patientrun studies and develop funding programmes to support
those models that appear more fruitful than clinicianinitiated studies.
Summary of recommendations
Recommendation Audience Timescale
Experiment with approaches that Research funders, Shortterm

candevelop communities that successfully academic community (02yrs)
collect large amounts of data.
Find and develop opportunities for Research funders, academic Shortterm
demonstration projects around N of 1 trials. community, clinical
professionals, patients
Invest in further research and development Research funders, academic Shortterm
for prognosis, both for the development of community
specific prognostic models and to develop
better understanding of the psychology
of prognosis and the methodology of
developing reliable, useful prognostic
models/clinical prediction rules.
Support the development and Research funders, academic Shortterm
implementation of decision support tools community, health
for patients and clinicians. providers, NHS, National
Commissioning Board, local
commissioners (GPs)
Test the reliability of prognostic models Research funders, academic Midterm
through trials to explore whether they community, NICE (25yrs)
might be useable to judge the efficacy of
new treatments.
Research funders should insist on publishing Research funders, academic Shortterm
underlying data of trials as a condition of community
funding, and explore other opportunities for
content to be authored collaboratively in an
open environment.
Results from a range of trials on a similar Research funders, academic Midterm
treatment should be made readily available community
to compare results of trials.
Invest in research to better understand the Research funders, Shortterm
translation of knowledge into practice, and academic community,
how knowledge is applied, such as through National Institute for
implementation case studies. Health Research (NIHR)
research networks e.g. NHS
Confederation
Develop common standards on structuring Research funders, academic Midterm
for clinical guidelines and all publiclyfunded community, NICE, SIGN,
clinical guideline producers move towards NHS, Department of Health
capturing their recommendation in a
standard structured, coded knowledge
base form, rather than as text.
Enshrine the right of patients to access Department of Health, NHS, Midterm

data in the NHS constitution, with a target patients
that all patients should have access to their
data in a standard format by 2018.
Research and experimentation to Research funders, academic Shortterm
understand patient incentives for community, patients,
capturing data and participating in a clinical professionals,
knowledge commons approach, including Department of Health
characteristics of who this works for.
Trials to examine further iterations of Research funders, academic Shortterm
methods for extracting firm conclusions community, NHS
from routine data (e.g. the instrument
variable approach).
Research and development funding for new Public bodies (e.g. TSB), Midterm
business models and product propositions Department of Health,
for knowledge commons, including possible health providers, NHS,
design and interaction. National Commissioning
Board, EPSRC
NHS accredited apps and communities as Department of Health, Shortterm
being interoperable and clinically sound in health providers, NHS,
their advice. National Commissioning
Board
Develop standards for compatibility of Department of Health, Shortterm
data and systems with external sites as health providers, NHS,
gatekeepers enforced by the National National Commissioning
Commissioning Board. Board
Leadership organisations should examine Royal Colleges, leadership Midterm
the implications for professional roles, and training providers,
education and training of greater patient professional, patient groups
autonomy, selfreliance and access to and advocacy organisation
patient data and knowledge, and seek ways
to bring about culture change.
Identify and develop demonstration Research funders, NIHR Shortterm
projects to investigate the benefits Department of Health,
and limitations of patientinitiated and health providers, NHS,
patientrun studies and develop funding National Commissioning
programmes to support those models that Board, local commissioners
appear more fruitful than clinicianinitiated (GPs)
studies.
Conclusion and next steps
Greater openness and collaboration in society combined with advances in technology, in

our understanding of human psychology, drug discovery and communication tools present
new opportunities for research, clinical practice, selfmanagement and interaction with the
health system. Big and open data and more precise analytical tools allows us to trial and
model the benefits of new models of care faster, more cheaply and on a larger scale, with
more granular data than we previously had access to.
The growing supply of useful tools, methods and platforms through which we can create
and apply knowledge is changing expectations of what it is possible to achieve in health,
and of the balance between patient and professional inputs to health knowledge. This
paper attempts to frame debate and action in moving towards a health knowledge
commons, and explores the potential for a more dynamic, distributed, collective system of
knowledge creation and application in health. We think this issue needs to be at the heart
of health policy and the focus of research and innovation to realise the potential of a more
knowledge centred system of health and care in ten to 20 years time.
Although there is much already happening in this area, we think more rapid development
is needed in a number of key areas to make a knowledge commons happen: new policies
and regulations, new technologies, products and services, new business models and
institutions and new behaviours and culture change. We want to play a role in this by
forming and mobilising alliances for change, working with others to develop key parts of
the infrastructure such as common standards and access, commons-building models and
practices, and promoting and shaping further research and development to demonstrate
the potential of a knowledge commons in practice.
We know that such a campaign is inevitably informed by the experience and efforts of
many others and that, going forward, our work needs to build on this, and develop an
alliance of support and action towards a more effective health knowledge system. If you
would like to be involved, please do get in touch with us at: healthcommons@nesta.org.uk
Acknowledgements
We are extremely grateful for the thoughtful engagement and input from many experts in
the development of this paper. We would especially like to thank those who attended our
expert seminars held at Nesta in October 2012 as well as the feedback we have received
outside of that session. In particular, we would like to thank Professor Brian Haynes at
McMaster University, Sir Muir Grey, Jeremy Heywood and Paul Wicks at PatientsLikeMe,
Professor John Fox at Oxford University, John de Pury at the NHS Confederation, Dr Vivek
Muthu of Bazian, William Heath of MyDex, Bob Gann from NHS Choices, Graham Dover
at the Mindset Foundation and Martin Feuz at Goldsmiths University for their substantial
contributions. We would also like to thank Cassie Robinson for her design work, Peter
Gerry at The Young Foundation, and Geoff Mulgan, Halima Khan, Louise Marston and Henri
Rapson at Nesta. All errors and omissions remain our own.
Bibliography
Balas, E.A. and Boren, S.A. (2000) Managing clinical knowledge for health care improvement. In van Bemmel, J.
H. and McCray, A.T. (2000) Yearbook of Medical Informatics 2000: PatientCentered Systems. 6570. Stuttgart,
Germany: Schattauer Verlagsgesellschaft mbH, 2000.
Byar, D.P. (1980) Why data bases should not replace randomized clinical trials. Biometrics. 36, no. 2 (June 1980):
33742.
Christakis, N. A. and Fowler, J.H. (2007) The Spread of Obesity in a Large Social Network Over 32 Years. New
England Journal of Medicine. No. 357 (2007): 3709.
ContopoulosIoannidis, D., Alexiou, G.A., Gouvias, T.C. and Ioannidis, J.P.A. (2008) Life Cycle of Translational
Research for Medical Interventions. Science. 321, no. 5894 (September 2008): 12981299.
Department of Health (2012) The Power of Information. London: Department of Health.
GarciaAlamino, J.M. et al., (2010) Selfmonitoring and selfmanagement of oral anticoagulation. Cochrane
Database of Systematic Reviews. No. 4 (2010).
Glyn, J. (1998)The discovery and early use of cortisone. Journal of the Royal Society of Medicine. 91 (December
1998): 513517.
Goldstein, J.L. and Brown, M.S. (1997) The Clinical Investigator: Bewitched, Bothered, and BewilderedBut Still
Beloved. The Journal of Clinical Investigation. 99, no. 12 (June 1997): 28032812.
Greenes, R. and Shortliffe, E. (1990) Medical Infomatics. The Journal of the American Medical Association. 263,
no. 4 (1990): 11141120.
Guyatt, G. et al.(1988) A clinicians guide for conducting randomized trials in individual patients. Canadian Medical
Association Journal. 139, no. 6 (Sept 1988): 497503.
Haynes, A.B. et al. (2009) A Surgical Safety Checklist to Reduce Morbidity and Mortality in a Global Population.
The New England Journal of Medicine. 2009: 491499.
Haynes, R.B. et al. (2006) SecondOrder Peer Review of the Medical Literature for Clinical Practitioners. The
Journal of the American Medical Association. 295, no. 15 (2006): 18011808.
Ioannidis, J.A. (2005) Contraindicated and Initially Stronger Effects in highly Cited Clinical Research. The Journal
of the American Medical Association. 294(2) (2005): 218228.
Kirsch, I., Deacon, B. J., HuedoMedina, T. B., Scoboria, A., Moore, T.J. and Johnson, B.T. (2008) Initial severity and
antidepressant benefits: a metaanalysis of data submitted to the food and drug administration. PLOS Medicine.
2008: 5(2), e45 EP.
Kosciely, S. (2010) Why most gene expression signatures of tumours have not been useful in the clinic. Science
Translational Medicine. 2, no. 14 (Jan 2010).
Medical Research Council (2008) Developing and Evaluating Complex Interventions. London: Medical Research
Council.
Morris, J.K. (2002) Screening for neuroblastoma in children. Journal of Medical Screening. 9, no. 2 (2002): 56.
National Audit Office (2010) Management of NHS hospital productivity. London: HMSO.
Paul et al. (2010) How to improve R&D productivity: the pharmaceutical industrys grand challenge. Nature
Reviews: Drug Discovery. 9, 203214.
Ratner, M. and Gura, T. (2008) Offlabel or offlimits? Nature Biotechnology. 26 (2008): 867 875.
Steventon, A. et al., (2012) Effect of telehealth on use of secondary care and mortality: findings from the Whole
System Demonstrator cluster randomised trial. BMJ. No. 344 (June 2012).
Tenenbaum, J.M. and Wilbanks, J. (2008) Health Commons: Therapy Development in a Networked Worlds.
Treweek, S. and Zwarenstein, M. (2009) Making trials matter: pragmatic and explanatory trials and the problem of
applicability. Trials. 10, no. 37 (June 2009).
Wyatt, J. (2001) Knowledge for the clinician 10. Management of explicit and tacit knowledge. (2001) Journal of
the Royal Society of Medicine. 94 (2001): 69.
Wyatt, J. and Taylor, P. Decision support systems and clinical innovation. In Haines, A. and Donald, A. (2001)
Getting research findings into practice. 12337. London: BMJ Publishing.
Wyatt, J., PatersonBrown, S., Johanson, R., Altman, D.G., Bradburn, M.J. and Fisk, N.M. (1998) Randomised trial of
educational visits to enhance use of systematic reviews in 25 obstetric units. BMJ. 1998; 317:1041.
Wyatt, J. (1991) Use and Sources of Medical Knowledge. The Lancet. No. 338 (1991): 13681378.
Endnotes
1. By engagement, we refer to actions individuals must take to obtain the greatest benefit from the healthcare services available
to them. See Center for Advancing Health (2010)A New Definition of Patient Engagement: What is Engagement and Why is it
Important? Washington DC: Center for Advancing Health.
2. See for example, Naylor, C., Parsonage, M., McDaid, D., Knapp, M., Fossey, M., Galea, A. (2012) Longterm Conditions and Mental
Health: the cost of comorbidities. London: The Kings Fund.
3. Kerr, D. (2005) Building a Health Service Fit for the Future: a report on the future of the NHS in Scotland. Edinburgh: NHS
Scotland.
4. Department of Health (2012) Liberating the NHS: no decision about me without me. London: Department of Health.
5. See for example the Institute of Medicines series The Learning Health System Series: Continuous improvement and innovation
in health and health care.
6. Ioannidis, J.A. (2005) Contraindications and Initially Stronger Effects in Highly Cited Clinical Research. The Journal of the
American Medical Association. 294(2) (2005): 218228.
7. See: http://www.equatornetwork.org/
8. ContopoulosIoannidis, D. G., Alexiou, G.A., Gouvias, T.C. and Ioannidis, J.P.A. (2008) Life Cycle of Translational Research for
Medical Interventions. Science. 321, no. 5894 (September 2008): 12981299.
9. See research on knowledge translation and the barriers of research into practice, for example in Lang, E. S., Wyer, P. C. and
Haynes, B. R. (2007) Knowledge Translation: Closing the EvidencetoPractice Gap. Annals of Emergency Medicine. Volume
39, No.3.
10. For example, it is often not sensible to randomise patients as staff involved in delivering the new service may accidentally
transfer the new practice to control patients. Whereas patients in a drug trial would continue to attend the same clinic, served
by the same staff, for an RCT of a service intervention it is often necessary to randomise staff, or even hospitals (Wyatt, et al.
1998), effectively running two discrete services simultaneously. This imposes significant cost. Patients may have existing long
term clinical relationships and not wish to be allocated to a new service with unknown staff. This is not to say that RCTs of a
process improvement are impossible, but that it is a complex issue that can require ingenious study designs.
11. Medical Research Council (2008) Developing and Evaluating Complex Interventions. London: Medical Research Council.
12. Treweek, S. and Zwarenstein, M. (2009) Making trials matter: pragmatic and explanatory trials and the problem of applicability.
Trials. 10, no. 37 (June 2009).
13. Wyatt, J. (1991) Use and Sources of Medical Knowledge. The Lancet. No. 338 (1991): 13681378.
14. For an account of this as a wider phenomenon, see Weinberger, D. (2012) Too Big to Know: Rethinking Knowledge Now That
the Facts Arent the Facts, Experts Are Everywhere, and the Smartest Person in the Room Is the Room.
15. See for example, Clifford, S., Garfield, S., Eliasson, L. and Barber N. (2010) Medication adherence and community pharmacy: A
review of education, policy and research in England. Pharmacy Practice. 8, pp. 7788.
16. See for example Weber, S. (2004) The Success of Open Source. Boston, MA: Harvard University Press; Bricklin, D. (2006)
The Cornucopia of the Commons. OReilly and Associates, Inc; Ostrum, E. and Hess, C. (2006) Understanding Knowledge as
a Commons: from theory to practice. Boston, MA: Harvard MIT Press; Institute for the Future (2008) Healthcare 2020 Signals
and forecasts map. Palo Alto CA: Institute for the Future.
17. Tenenbaum, J.M. and Wilbanks, J. (2008) Health Commons: Therapy Development in a Networked World.
18. Frydman, G. (2009) Patient-Driven Research: rich opportunities and real risks. Society for Participatory Medicine.
19. Dyson, G. (2012) Turings Cathedral: the origins of the digital universe. London: Allen Lane.
20. See for example, McKinsey Global Institute (2011) Big Data, the Next Frontier for Innovation, Competition and Productivity.
McKinsey Global Institute; Bollier, D. (2010) The Promise and Peril of Big Data. Queenstown MD: Aspen Institute.
21. Department of Health (2012) The Power of Information: Putting us all in control of the health and care information we need.
London: Department of Health.
22. Steventon, A. et al. (2012) Effect of telehealth on use of secondary care and mortality: findings from the Whole System
Demonstrator cluster randomised trial. BMJ. no. 344 (June 2012).
23. See: www.myfitnesspal.com
24. Harrison, S. C., Holmes, M. V. and Humphries, S. E. (2012) Mendelian randomisation, lipids, and cardiovascular disease. The
Lancet. Volume 380, Issue 9841, Pages 543 545, 11 August 2012.
25. Christakis, N. A. and Fowler, J.H. (2007) The Spread of Obesity in a Large Social Network Over 32 Years. New England Journal
of Medicine. No. 357 (2007): 3709.
26. See: https://foursquare.com/

27. See: https://www.mint.com/
28. Benkler, Y. (2002) Coases Penguin, or, Linux and the Nature of the Firm. Yale Law Journal. Vol. 112 (20022003).
29. Haynes, R. B. et al. (2006) SecondOrder Peer Review of the Medical Literature for Clinical Practitioners. The Journal of the
American Medical Association. 295, no. 15 (2006): 18011808.
30. See: www.schoolofcommoning.com
32. See: http://www.openclinical.org/gmm_proforma.html
33. See for example Wicks, P., Vaughan, T. E., Massagli M. P. and Heywood, J. (2011) Accelerated clinical discovery via open data
platform and patientmatching algorithm. Nature Biotechnology. 29(5):411 See for example recommendations from Tim Kesley,
the NHS National Director for Patients and Information, on the value of openness of data and information in healthcare, as
reported in http://www.hsj.co.uk/news/exclusivenhsinfochiefpredictsusstylerevolution/5049828.article (accessed 26
November 2012).
34. UK Data protection Act 1996.
35. See: www.openclinical.org
36. Gigerenzen, G., and Gray, M. (2011) Better Doctors, Better Patients, Better Decisions: envisioning healthcare 2020. Boston, MA:
MIT Press.
37. Kirsch, I., Deacon, B. J., HuedoMedina, T. B., Scoboria, A., Moore, T.J. and Johnson, B. T. (2008) Initial severity and
antidepressant benefits: a metaanalysis of data submitted to the food and drug administration. PLOS Medicine. 2008: 5(2),
e45 EP.
38. Glyn, J. (1998) The discovery and early use of cortisone. Journal of the Royal Society of Medicine. 91 (December 1998): 513517.
39. Kosciely, S. (2010) Why most gene expression signatures of tumours have not been useful in the clinic. Science Translational
Medicine. 2, no. 14 (Jan 2010).
40. Goldstein, J.L. and Brown, M.S. (1997) The Clinical Investigator: Bewitched, Bothered, and BewilderedBut Still Beloved. The
Journal of Clinical Investigation. 99, no. 12 (June 1997): 28032812.
41. Byar, D.P. (1980) Why data bases should not replace randomized clinical trials. Biometrics. 36, no. 2 (June 1980): 33742.
42. Guyatt, G. et al. (1988) A clinicians guide for conducting randomized trials in individual patients. Canadian Medical Association
Journal. 139, no. 6 (Sept 1988): 497503.
44. Spindler, K. P., Parker, R. D., Andrish, J. T., Kaeding, C. C., Wright, R. W., Marx, R. G., McCarty, E. C., Amendola, A., Dunn, W. R.,
Huston, L. J., Harrell, F. E. Jr; MOON Group. (Epub ahead of print) (2012) Prognosis and predictors of ACL reconstructions using
the MOON cohort: A model for comparative effectiveness studies. Journal of Orthopaedic Research. 2012 Aug 21. doi: 10.1002/
jor.22201.
45. Sackett, D.L. (1992) The rational clinical examination. A primer on the precision and accuracy of the clinical examination. The
Journal of the American Medical Association. 1992 May 20;267(19):263844.
46. See: http://www.implementationscience.com/series/CCDSS
47. Chronic Obstructive Pulmonary Disease (COPD) project Empowering COPD patients to manage their own condition, led
by the Innovation Unit with NHS Redbridge, UCL Partners, Outer North East London NHS, North East London, North Central
London and Essex Health Innovation Cluster (NECLES) and the Health Foundation, see Innovation Unit (2012) COPD Care
Checklist: Putting data in the hands of patients to drive care quality and reduce hospital admissions. London: Innovation Unit.
48. Coie, J.D., Watt, N.F., West, S.G., Hawkins, J.D., Asarnow, J.R., Markman, H.J., Ramey, S.L., Shure, M.B. and Long, B. (1993)
The Science of Prevention: A conceptual framework and some direction for a national research programme. American
Psychologist. Vol. 48, no. 10, pp. 1013 1022.
49. See: http://dataarchive.ac.uk/
Nesta
1 Plough Place
London EC4A 1DE
research@nesta.org.uk
www.twitter.com/nesta_uk
www.facebook.com/nesta.uk
www.nesta.org.uk
March 2013
Nesta Operating Company. English charity no. 7706036.

Scottish charity no. SC042833.

Doctor Know A Knowledge Commons in Health

Uploaded by

Copyright:

Available Formats

Doctor Know A Knowledge Commons in Health

Uploaded by

Document Information

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

Doctor Know A Knowledge Commons in Health

Uploaded by

Copyright:

Available Formats

Doctor Know

About The Young Foundation

2 Why a Health Knowledge Commons? 9

3 The potential for a Health Knowledge Commons 14

4 What next? Making a knowledge commons happen 30

Conclusion and next steps 37

increasingly wellstructured patient records will all contribute to a considerably larger

One of the challenges in considering a knowledge commons in health is the complexity of

What do we mean by knowledge?

What sort of knowledge system do we want

Experiential knowledge is knowledge about how a condition manifests itself, such as

Contextual knowledge is knowledge about how healthcare best happens in practice,

Thirdly we want knowledge to be applicable in practical circumstances. This means that

Clinical practitioners specialist and general clinical practitioners working within a

Community health practitioners community nurses, carers, specialist health workers,

Researchers academics and researchers within academic institutions and universities,

Businesses including clinical service providers, medical technology producers.

Commissioners clinical commissioners concerned with the efficient and effective

The UK in particular benefits from an extremely sophisticated institutional system

The changing scope of data and health knowledge

Table 1: The changing model of healthcare (based on a table in the Kerr

Old model of care New model of care

Focus on acute conditions Focus on longterm conditions

Figure 2: The complex interacting system of adolescent diabetes management

Friends Family Blood

Exercise Lifestyle School

Alcohol Diet Emotions Insulin

Access to the relevant medical knowledge;

Application of that knowledge to the specific patient to create a management plan,

Successful execution of the management plan over time.

As effective selfcare becomes increasingly important, knowledge needs to be made

The paradox of health knowledge

3 The potential for a Health

In essence, a knowledge commons is defined by the principle that knowledge is a shared

Capturing and assembling big data

PatientsLikeMe also publish this data in peerreviewed journals, such as in a study

As an example of a Quantified Self technology, Jawbone UP is a wristband which

Collaborative structuring and analysis to refine data into knowledge

An example of Internetbased professional collaborative rating in healthcare is MORE, the

McMaster Online Rating

To be successful, other collaborative models will need to build on infrastructure already in

Effective collaboration depends on knowledge being structured in a way that means

Openness and transparency

What might a health knowledge commons mean in practice?

Research and discovery

An example of instrumental variable method is a study of the effect of bone marrow

Prognostic models and citizen clinical science

In certain fields of applied knowledge there is the opportunity to use a knowledge

An important potential research contribution of prognostic models is to reduce our

Large collaborative platforms could also allow patients themselves to systematically

2 a postbronchodilator spirometry, also known as spirometry test? test

If you are a smoker and you 3. Stopping

4 Your records show you

Changing relationships with patients

Pilot projects focusing on coproduction in health, such as Nestas People Powered

Prevention and wellness

4 What next? Making a

New technologies, products and services

New policies and regulations