The Scottish Health Informatics Programme Health Statistics User

The Scottish Health Informatics Programme Health Statistics User Group Frank Sullivan FRSE, FRCP(Glas. ), FRCGP NHSTayside Prof of R&D in GP &1 y care Director Health Informatics Centre

Deterministic linkage Lab Data CHNo Investigations Social Services Hospital SMR Pharmacy Screening Dental Primary Care

Community Health Number 28 12 57 02 5 Date of Birth Sequence Se x 0 Checksum

Record-Linked Data Completing the Jigsaw Lab Data Pharmacy Investigations Dental CHNo Social Services GP Hospital Screening

Scottish data from cradle to grave … Dental SMR 13 Neonatal Record SMR 11 BIRTH Maternity Mental Health SMR 04 Out patients SMR 00 Community care SMR 50 Hospital Admissions SMR 01 GP consultations Immunisation Prescribing A& E Child health surveillance Screening Cancer registry Cancer registrations SMR 06 DEATH

Navigation • • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Navigation • • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Who we are Our funding comes from: Universities of Dundee, Edinburgh, Glasgow and St Andrews and the Information Services Division of NHS Scotland

Governance International Advisory Board Scientific Management Group SHIP Programme Manager Core Programmes Research Programmes

• • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

The present Linkage of a national diabetes dataset (SCI-DC) to a a datamart of Scottish Morbidity Register (ACa. DMe) • R&D approval from each board – 8 page form, covering letter, CV, proposal, sponsor letter, funder letter • Ethics approval 23 page form • PAC approval 11 pages • 14 Caldicott guardian approvals – Calidcott guardians difficult to identify – Took 4 months to get all replies – many needed chasing, 5 requested further information

We aim to: Create a research portal for EPRs already held by NHS Scotland that will provide rapid, secure, access to the type of data that clinical scientists require.

We aim to: Develop and evaluate systems that work across institutional boundaries to allow linkage between large, federated, third party research datasets and the NHS research portal. Health Education Environment Linked dataset

• • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Core Programmes C 1: Provisioning Datasets C 3: C 2: Engaging Governance Researchers C 4: Engaging The Public

C 1: Provisioning Datasets Aims • To create a research portal for EPRs already held by NHS Scotland, the Scottish Health Information Service for Research (SHIS-R). • To develop and evaluate innovative technical approaches that allow linkage between large, federated, third party research datasets between themselves and with SHIS-R. • To develop and evaluate systems that work across institutional boundaries with adequate data manipulation and statistical functionality that provide rapid, secure, access to the type of data that clinical scientists require.

C 2: Governance Aim To analyse the ethico-legal and cultural challenges associated with the secondary use of EPRs with a view to mapping the elements necessary to contribute to an optimal governance regime • 3 dimensions: o Scottish o International o Interdisciplinary

C 3: Engaging Researchers Aims • To host a biennial conference “Exploiting Existing Data for Health Research”. It has 5 themes: o o o The value of record linkage in health research Record linkage for health care improvement Longitudinal record linkage The methodological challenges of record linkage Confidentiality, disclosure and ethical issues • To develop and deliver training programmes and workshops for EPR research

C 4: Public Engagement Aims • To synthesise existing evidence on citizens attitudes towards sharing personal data for research. • To generate new evidence on the acceptability of different levels of data sharing under varying conditions. • To engage the wider public with the aims and outcomes of the SHIP through specific consultation exercises. • To link the public engagement activities into the development of governance frameworks (C 2). • To examine novel methods of making health data available to the public.

• • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Research Programmes RP 1: EPR Support of Clinical trials RP 2: National Epidemiology RP 3: Pharmacovigilance RP 4: Demographic, Socio-Economic & Environmental Data Linkage

RP 1: Support of Clinical Trials Aim To evaluate how EPRs can best support the conduct of a range of clinical trials • Can we identify eligible patients? • Can we extract follow-up information from GP records? • Can we get endpoints from routinely collected data?

RP 2: National Epidemiology Aim To perform epidemiological studies on a national scale and use the information to estimate current and future health costs using diabetes as an exemplar. o Develop risk engines for CVD in diabetes to inform statin prescribing o Use retinopathy data to look at screening intervals o Explore glycaemic control and CVD o Explore novel associations between diabetes and other diseases o Modelling current and future diabetes care costs

RP 3: Pharmacovigilance Aim • To link community prescribing data to EPRs to demonstrate the feasibility of national pharmacovigilance o using longitudinal datasets that link drug exposure and morbidity to discover previously unknown drug effects o developing novel methods of analysis and comparing them with traditional approaches Variational Bayes Algorithms

Year 2 of RP 3 • Development of Specific analytic tools and software for supervised learning (classification) • Development of Specific tools for unsupervised learning (clustering) • Definition of specific ADR hypotheses to test • Definition of the specific drugs to include in unsupervised inference of likely ADRs and pleiotropic effects • These tools will exploit several different kinds of contrast including survival analyses, case cross over analyses case control etc.

RP 4: Demographic, Socio-Economic and Environmental Data Linkage Aims • to link information given by respondents in genetic studies back through time using the records of births, marriages and deaths since 1855. o Complex genealogies are being built up which will be used to look at genetic effects in epidemiological studies • to estimate the effects of exposure to various environmental agents on health by linking environmental pollution data to hospital admissions and the Scottish Longitudinal Study (SLS)

• • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Formulating the plans • Review of existing practice (completed August 2010) – Information governance of use of patient data in medical research in Scotland: current and future scenarios • Public engagement – 8 focus groups by March 2011 • Expert working groups (final report 11/11/10) – On Governance, IT and Administration • Consultation with key stakeholders – Secure data linkage & access workshop 09/02/10 – Governance & public engagement workshop 13/12/10

Improving Governance • Increasing transparency & reducing uncertainty • Setting standards: Principles & Best Practices • Clarifying Responsibilities: Data Flows & Data Controllers • Seeking buy-in from stakeholders

Good Governance Framework • Principles: foundational starting points for deliberation and action • Best practice: instances of implementation of principles to a high standard • Content: • Public interest and the importance of research • Privacy/Anonymisation/Consent/Data Protection • Authorising/advisory bodies • Governance/Access • Trusted Third Parties (where appropriate) • Clinical Trials • Cross-sector sharing and sharing agreements • Public engagement and benefit sharing

Public Interest / Personal Privacy Scientifically sound and ethically robust research is in the interest of protecting the health of the public. Every effort should be made to consider and minimise risks of identification to data subjects

Consent yes no Anonymisation and / or Authorisation

The Vision • Aims to improve accessibility and provide metadata • Streamlined approvals process for health data • Work with SAHSC to provide local access for researchers at SAHSC nodes • A website for researchers with details of the process and available data • A national indexing service

Proposed Infrastructure • National Indexing Service – located in NHS NSS • National Safe Haven – Separately located in NSS • Linkage Agent – Within the national safe haven • Model to be mirrored at SAHSC nodes

• • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Safe Havens Thomas and Walport Data Sharing Review (2008): “environments for population based research and statistical analysis in which the risk of identifying individuals is minimised”

Safe Projects A committee decides whether the access request is for a valid statistical purpose

Safe People Researchers may have to be: • Attached to a known institution • Asked to attend a course • Bound by a strict code • Subject to sanctions

Safe Data • Global recoding – Combining categories • Top and bottom recoding – For normal distributions • Record swapping – Interchanging sensitive values between records • Post Randomisation Method – Misclassifies categorical data according to a predetermined probability mechanism eg 20% chance that a male is female • (Over)imputation – Values randomly deleted and likely values from similar donor records are substitued

Safe Settings Dumb terminal prevents removal of data: • No memory stick • No CD or DVD • No internet Can be expanded to the user’s own computer using eg Citrix or Terminal Services.

Safe Outputs Statistical Disclosure Control • Results produced may be checked by officers of the safe haven to make sure they do not contain any disclosive results.

• • Who we are Aims of SHIP Core programmes Research programmes The plan Safe havens The detail

Privacy Advisory Committee (approval of linkage requests) Data Source 1 Indexing Service ISD Data Source 2 SCI DC Data Source 3 University dataset Aggregate nondisclosive data Key coded data only I’d like to link some data Linkage Agent Project Governance Disclosure work space Safe Guards Control Secure Access Facility Data Archive SAFE HAVEN

Linkage part 1 Datasource 2 Datasource 1 Local IDs Study numbers Local IDs Names Addresses Dates of birth Indexing Service Local IDs Study numbers

Linkage part 2 Datasource 1 Study numbers Payload data Safe Haven Study numbers Payloaddata. Pay loaddata Study numbers Payload data

Exploiting Existing Data for Health Research International Conference 9 -11 th Sept '11 University of St Andrews www. scot-ship. ac. uk