ENABLING DATA LINKAGE TO MAXIMISE THE VALUE OF PUBLIC HEALTH RESEARCH DATA Presentation of findings to the Public Health Research Data Forum University.

ENABLING DATA LINKAGETO MAXIMISE THE VALUE

OF PUBLIC HEALTH RESEARCH DATA

Presentation of findings to the Public Health Research Data Forum

University of the West of England, Bristol

DataFirst, University of Cape Town

CIPRB, Dhaka

1

• Aims and methods of the project• Key findings• The HIC experience• The LMIC experience• Recommendations

Outline of presentation

Introduction

• how data linkage could boost public health research

• the barriers to useful data linkage

Aim: to investigate

Objectives and methods

• “to produce a synthesis fully grounded in both theory and empirical evidence to generate recommendations and practical guidelines for short- and long-term public health data strategies”

• practical and useful rather than exhaustive

Objective

Objectives and methods

• Faculties of business and health, UWE• DataFirst, University of Cape Town• Centre for Injury Prevention Research,

Bangladesh

– Mix of expertise in data access, socioeconomic data, and public health and clinical data

Project team

Objectives and methods

• non-systematic literature review– including conference presentations

• formal and informal interviews• case study examples• internal team perspective

Methods

Objectives and methods

Key findings

Key findings

1. Change the tone of the debate

1. Data should not be used for research or linked unless it can be done safely and securely

2. Data should be available for research and linking unless it cannot be done safely and securely

Key findings

Key findings

1. Change the tone of the debate

• default closed → default open

Key findings

Key findings

2. Policy decisions need to be more evidence-based

• research data use is safe

Key findings

Key findings

2. Policy decisions need to be more evidence-based

• ‘intruder’ model → ‘idiot’ model

Key findings

Key findings

3. Narrow informed consent is not enough for good epidemiological research

• broad consent supported by public/researchers• where broad consent not feasible, we know how to

manage the social contract

Key findings

Key findings

4. Maintaining good relationships is the key

• relationships with everyone: data depositors, ethics committees, general public, researchers

• early planning with stakeholders vital– especially for strategic projects

Key findings

Key findings

5. Incentives to manage and share data are weak

• funding bodies have some responsibility• the research community needs to consider its role

Key findings

Key findings

6. Different things matter in difference places

• A hierarchy of problems?– data– organisation– institutions

Key findings

Key findings

• Data issues exist• Dominated by institutional issues

– relationships with data depositors/ethics committees

– public acceptability– unrealistic risk-assessment, worst-case

scenario planning

The HIC experience

The HIC experience

• What works: stakeholder management– early planning– education

The HIC experience

The HIC experience

• Dominated by operational and quality issues

The LMIC experience

The LMIC experience

• Operational issues: access to health data– Publicly funded health data held by state research institutes,

universities only available to research collaborators– No data sharing requirement from national funding bodies– Data sharing requirements of international funders not enforced

• No critical mass of researchers engaged in quantitative research – rather “pools of expertise”

The LMIC experience

The LMIC experience

• The base situation– We have useful, linkable data– ADHSS, other household survey, hospital

information systems, civil registration, laboratory data, drug dispensation, encounters, episodic data, social grants and schools

The LMIC experience in SA

• What data linkage has there been?– ADHSS to civil registration systems, clinical

data (PHCU, HIV/AIDS, hypertension clinics)– Data harmonisation project– HIV cohort data to national population

registers

The LMIC experience in SA

• Operational Barriers– High level data skills and database

management skills rare– Outsourcing of complex information system

management– Pay scale issues and incentives, public vs

private

The LMIC experience in SA

The LMIC experience

• Statistical Barriers– ID numbers not always available– ID number penetration correlated with

individual characteristics– Probabilistic matching issues: date of birth,

names, twins

The LMIC experience in SA

• Ethical Concerns– Protection of personal information perceived

as more important if data used for research purposes (vs clinical)

– WCDoH trying to operationalise due diligence by setting up preapproved database procedures, anonymize data effectively

The LMIC experience in SA

• Two types– Changing the conceptual framework– Practical guidelines and measures

Recommendations

Recommendations

• Much evidence of what works, but– in the wrong place– not used in decision-making

• Many wheels being re-inventedÞ need for clear, strong, evidence-based

guidance to address fear and ignorance

Recommendations: changing the conceptual framework

Recommendations

• Everything has been solved somewhere• Make sure this information is known

– Technical information• managing access; collecting good ID data

– Institutional tips• getting ethics/data depositors on your side

Recommendations: practical guidance

Recommendations

• Establish Research Data Infrastructure to support health data usage and linkagese.g. DataFirst’s Secure Data Service

• Build quantitative skills

Recommendations: practical guidance for LMICs

Recommendations

• Data management is a problem:– shortage of ‘data science’ skills– need to encourage data sharing– data collection and research timetables don’t fit

Þ some funding tailored towards good data collection and curation

Recommendations: planning for and funding data collection

Recommendations

http://www.wellcome.ac.uk/About-us/Policy/Spotlight-issues/Data-sharing/Public-health-and-epidemiology/WTP056860.htm

Thank you

Questions?

Next steps