Freeing the power of administrative data:
can Scotland become a global leader?
Dr Stephen Pavis
Programme Director
Farr Institute (Scotland)
NSS, NHS
Structure of the talk
1.
2.
3.
4.
Characteristics of Administrative data
Why it’s a valuable resource
Some of challenges in ‘freeing its power’
Progress which has been made and future
direction
As we go through I’ll tell you about some empirical
findings to keep your interest
What do we mean by Administrative data?
• Routinely collected data normally during the provision of
a service (health, education, criminal justice, benefits
and taxation, social care etc)
• Commonly have National coverage and over a relatively
long time period
• Relatively consistent data definitions and collection
processes
• Important to planning and providing services
– Limited control over creation processes
– Must undermine primary purpose/use
Public sector collects & analyse lots of
person level data
Health
Schools
Justice
Outcomes
Outcomes
Outcomes
What Health
Service does
What Schools do
What happens in
justics system
About service
users
About service
users
About service
users
Workforce
Workforce
Workforce
Funding
Funding
Funding
Further
Education
Social care
Higher
Education
Children’s
services
But it is in
silos
NHS data from cradle to grave
Mental Health
Dental
Neonatal Record
BIRTH
Maternity
Out patients
Community care
Hospital Admissions
GP consultations
Immunisation
Child health surveillance
Prescribing
A&E
Screening
DEATH
Suicide
Cancer registrations
Ways we can use Administrative data
Social mobility – by linking data on education, training,
employment, unemployment, incomes and benefits
Causal pathways over the life course – linking data on
education, health, employment, incomes and wealth
Informing policies designed to tackle poverty – linking
data on incomes and benefits, housing conditions,
(re)offending behaviour, exploring the role of poor
physical and/or mental health.
Report from the Administrative Data Taskforce (ESRC, MRC, Wellcome
Trust)
The human perspective
‘Cycles of deprivation and neglect overlap. The pressure of
circumstances, of chronic housing poverty, unemployment,
low income, poor wellbeing and poor education all
undermine resilience. It’s when these structural conditions
combine with other impoverishing experiences – such as
violence, crime, isolation, an unhappy childhood,
separation and poor mental health – that problems become
insurmountable’
Action for Children: Deprivation and risk: the case for early intervention
Dame ClareTickell
Chief Executive
There’s an economic case too
‘The public service bill for the 46,000 most deprived families
is over £4billion a year, almost £100k per family.’
‘Services tend to focus on a single problem of a single
person. Treating problems in isolation increases the risk of
relapse and creates a costly cycle of managed deprivation.
Breaking this cycle will mean…better value for taxpayers’.
https://www.gov.uk/government/news
26th August 2011
Administrative data:
‘win’, ‘win’, ‘win’ ‘win’
• For Researchers/statisticians
– It’s time efficient and cost effective– compared with
empirical data collection
• Policy makers and practitioners need to move beyond
the traditional silos
• Citizens deserve better opportunities and services
• Economic growth – developing and attracting industry to
Scotland
Anziolytic benzodiazepines (AB’s)
and RTAs
• Barbone et al used linked prescriptions and
Tayside police data to look at AB’s and road
traffic accidents
• The researchers found a dose related
relationship and estimate that non driving while
on AB’s would prevent 1577 accidents and 110
deaths annually across the UK
• The research led to a re-labeling of
the medication
(Barbone et al Lancet 1998)
Where are Scotland’s
competitive advantages?
• A small country with relatively stable
population
• Strong administrative data – eg NHS single
supplier
• Strong universities with significant expertise
• Cross sector and cross discipline
networks
What are the challenges?
1. We must ensure the public support the use of their
data and contribute to debates about what is in the
‘public interest’
2. Individual privacy must be protected and the law
complied with
3. Permission processes must be efficient and not
overly bureaucratic
4. We link data efficiently and have high end
computing
5. We need multi-disciplinary working
What are the Scottish public’s views?
• Methodologically a difficult area, there are:
– multiple ‘publics’ and various opinions
– people have different levels of knowledge
– we’re discussing multiple datasets
– different people want to access data
– for different reasons
• Methods and framing of questions is
crucial
But some information does exist…
Scotland has been leading on this area :
•Scottish Health Informatics Programme
(Wellcome Trust)
• Data Linkage Framework
(Scottish Government)
Scottish Health Informatics
Programme
Least trusted
Most trusted
NHS
Academics
Scottish Government
Purpose of use
Wellcome Trust; University of Edinburgh
Private sector
Data sharing for research
purposes
Least trusted
Most trusted
Public Sector
3rd Sector
Benefit sharing a key factor
Sara Davidson, et al Ipsos Mori and University of Edinburgh
Private sector
Future areas for Public Engagement
• Moving forward we need to have iterative
cycles of:
– listening to the public
– educating on benefits and processes
– modifying our procedures and processes
Recognised by the ESRC, MRC and Scottish
Government
Suicide in Scotland
• Linkage of death records with psychiatric and general
hospital admissions
>15 years, 1981 to 2010
16,411 people had died; 10,907 had a hospital record
• 24% of deaths within 3 months of a hospital discharge
• People who had died were 3.1 times more likely to
have last visited a general rather than a psychiatric
hospital
• Dougall, et al University of Stirling , British Journal of Psychiatry
(in press)
What are the challenges?
1. We must ensure the public support the use of their
data and contribute to debates about what is in the
‘public interest’
2. Individual privacy must be protected and the law
complied with
3. Permission processes must be efficient and not
overly bureaucratic
4. We link data efficiently and have high end
computing
5. We need multi-disciplinary working
Permissions to access data
• Inconsistency across sectors and data
controllers
• Multiple processes and forms often requiring
similar information
• In health research it is common to need ethics
committee, PAC, data controllers, and R&D from
each Health Board,
Privacy Impact Assessment
The Information Governance Review
(Caldicott 2)
‘(we) heard from researchers that
complexity, confusion and lack of
consistency… hamper research. … data
controllers tend to be risk-averse, erring on
the side of caution rather than public benefit’
(p62)
The Information Governance Review
Caldicott 2
‘We recommend that the linkage of deidentified but still potentially identifiable
information from more than one organisation
should be done in specialist, well governed
independently scrutinised environments
known as ‘accredited safe havens’
Future work around permissions
• The Scottish Government’s Health Information
Research Advisory Group is likely to recommend
a review of permissions processes to ensure
efficiency
• The Farr Institute and the Administrative Data
Research Centre will support a network of
accredited safe havens
Proportionate Information Governance
• Safe People
– Accredited researchers
• Safe Data
– Linking minimum data to answer question
• Safe Locations
– Controlling access, limiting data travel (unless
consented)
University of Edinburgh School of Law and NSS
have led the way
Non-experimental evaluation (policy)
Effect of smoking legislation in Scotland
Admissions fell by 17% - 67% of
reduction was in non-smokers
Fall in England 4% (no legislation);
long term trend 3%
Acute Coronary syndrome
Pell et al, N Eng J Med (2008) 359; 482-491
Before ban 5.2% increase per annum
After ban 18.2% decrease per annum
Childhood asthma
Pell et al New Engl J M 201o, 363 . pp. 1139-1145
What are the challenges?
1. We must ensure the public support the use of their
data and contribute to debates about what is in the
‘public interest’
2. Individual privacy must be protected and the law
complied with
3. Permission processes must be efficient and not
overly bureaucratic
4. We link data efficiently and have high end
computing
5. We need multi-disciplinary working
Linking data and high performance
computing
• Community Health Index as a population
spine gives Scotland a competitive
advantage
• Linkage process: separation of functions
– personal information split from ‘payload’ data
as early as possible
– separate organisations ‘Index’ and ‘link’ data
• Can be used beyond health to link data
SHIP IT infrastructure
Farr Institute of Health
Informatics Research
Harnessing Data for Health Science and e-Health
Innovation
Health Informatics Research Centres
Scotland
Dundee, Glasgow, Edinburgh,
St Andrews, Aberdeen,
Strathclyde, MRC HGU, NHS
NSS
HeRC N8
Manchester, York, Lancaster,
Liverpool, Sheffield, AHSNs
CIPHER
Swansea, Bristol, Cardiff,
Exeter, Leicester, Sussex,
NWIS, Public Health Wales
UCL Partners
UCL, LSHTM, Queen
Mary, Public Health
England
Map Source: www.m62.net
The Farr Institute in the UK
MRC & 9 other funders
£39M investment
Brings together the
Health Informatics
Research Centres with
additional
capital/infrastructure
resource
Who is William Farr?
“Diseases are more easily prevented
than cured and the first step to their
prevention is the discovery of their
exciting causes.”
William Farr
Farr UK Objectives
• E-Infrastructure: To establish an outstanding UK einfrastructure across the Centres.
• Research: To enhance research productivity across the
UK by widening access to well-described datasets
through a prominent UK-wide portal, bringing new
datasets to the research community and enhancing
communication.
• Capacity: To develop a UK-wide co-ordinated offering for
training and capacity development to address the acute
skills shortage in health informatics research.
• Public engagement: To engage the public and patients
across the UK and locally, in a novel range of activities to
enhance public trust in the use of health records for
research.
How the £20m capital investment
supports the objectives
There are five core components of the institute for
capital investments
1. Physical centres
2. Safe havens
3. e-infrastructure
4. New data access
5. Communication
Intended impact of the Farr Institute
• New Science
• Larger Scale
• Better connections between centres of excellence
and datasets
• New Partnerships – academic, NHS and industry
• Increase UK skill base
• Public and patient advocacy
The Farr Health Informatics
Research Institute Scotland
£9M MRC and nine other funders
£2M SGHD
£0.5M NHS, NSS
Scottish Enterprise
Farr Institute, Scotland
• New, innovative Science
• High performance computing
– within a University
– a private cloud
• New datasets
– General Practice
– Laboratory data – clinical tests
– Clinical Images
Farr Scotland Safe Havens and IT resources
Farr accredited Regional
Safe Havens (x4)
Farr National Safe Haven
Farr Safe Havens manage
dataset access and
support researchers
Site
National datasets
NSS
Site
Farr datastore/ university
hosted
Server
Server
Farr analytic private
cloud /university
environment
SHIP/Farr analytic
environment / ATOS
hosted
Server
Server
User
Site
Local/regional datasets
Server
Data can move from any storage
location to any analytic platform
(with appropriate data controller
permission)
Regional analytic
environments
Server
Accredited user can access any of analytic
environment securely for a specific project
depending on compute needs
(assuming permissions in place )
Making it happen in
reality
• Partnerships, networks and federated
arrangements are the only way to maximise
Scotland’s administrative data
• Single organisations or individuals cannot achieve
this in isolation
• But this raises new challenges around
coordination: researchers need,
– a single point of entry
– help to navigate the system
– to know what’s possible
NSS’s eData, Research and
Innovation Service (eDRIS)
eDRIS aims to:
• facilitate partnership working between the
NHS, academics and industry
• provide coordination and expert advice to
support researchers
• make research more efficient, easier, save
time and encourage better research
The eDRIS Service
Provide analyses,
interpretation and
intelligence about data
(where required)
Liaison with
technical
infrastructure
(safe havens)
7
8
1
A named
Person from
start to finish
Single point of
entry for health
research
2
Support projects
from start to
finish
Facilitate
completion of
required
permissions
6
Build relationship
between data
suppliers and
customers
5
Liaison with data
suppliers to
secure data
3
Help with
study
design
Provide expert advice
on coding, terminology,
meta data and study
feasibility
4
Agree deliverables
and timelines
The Latest News
• The Economic and Social Research Council have provided
additional funding for the analysis of public sector data
(£40m UK; £8m Scotland)
• Expertise and research across wider administrative
datasets
–
–
–
–
–
Housing
Education
Criminal Justice
Work and pensions
Vital events back to 19th Century
• This resource will also be at 9 BioQuarter
Farr institute
Health
eAdmin data research
centre
Cross sectorial data
eData Research and Innovation Service
Public communication & engagement
FOUNDATION STONE:
GUIDING PRINCIPLES
Clear governance processes
No9
Scottish Informatics & Linkage Centre
High
High performance computing
Data Linkage Framework
What are the challenges?
1. We must ensure the public support the use of their
data and contribute to debates about what is in the
‘public interest’
2. Individual privacy must be protected and the law
complied with
3. Permission processes must be efficient and not
overly bureaucratic
4. We link data efficiently and have high end
computing
5. We need multi-disciplinary working
A new creative environment/culture
Multiple skills - same location
– Informatics and computing experts
– Academics from multiple universities
– Data experts
– Policy experts
– Over time we hope industry will join us
Can Scotland become a global leader?
• We have an opportunity
– expertise
– data
– infrastructure
– computing resource
• We have challenges around
– culture change (working across institutions &
disciplines)
– developing efficient access and linkage
processes
Thank you for listening
s.pavis@nhs.net
0131 2756670