Transaction data for population health

Lead Research Organisation: University of Bristol
Department Name: Bristol Medical School

Abstract

Digital technology opens up a new era in the understanding of human behaviour and lifestyle choices, with people's daily activities and habits leaving 'footprints' in their digital records. For example, when we buy goods in supermarkets and use loyalty cards to obtain benefits (e.g., future discounts), the supermarket records our purchases and creates a representation of our habits and preferences. Until now the use of 'digital footprint' data has mostly been limited to private companies. Companies have been using aggregates of these data to track sales of their products, to understand the factors that impact sales levels, and to target marketing and promotions. Changes in Data Protection law in the UK, i.e. General Data Protection Regulation, mean the public can now access and donate their data for academic research. Shopping history data, recorded through loyalty cards by retailers, are an extremely useful source of information for population health research as it can provide granular, objective data on real world choices and behaviours (e.g. painkillers, food) and other behaviours (e.g., pain and weight, wellbeing management). This information is often hard to obtain in the health research domain. Links between lifestyle choices and health outcomes are commonly studied through self-report questionnaires that ask people to remember their everyday choices and behaviours, which can bias results: responses about behaviours do not always reflect the reality of what people actually do. If and when shopping history data are used in a privacy preserving and ethical manner, these data can be utilised for public good, benefiting health research (e.g., helping to understand how everyday behaviours and lifestyle choices impact health and social outcomes). For example, what are the exact levels of alcohol consumption that lead to irreversible health damage for unborn babies accounting for moderating factors (e.g., age, gender, genetic makeup, etc.)? Under which conditions do different types of ready meals contribute to obesity? Do chemicals in household products lead to higher risks of cancer and other adverse health outcomes in children?
The Transaction Data for Population Health research programme utilises commercially collected datasets for privacy-preserving, ethical research to benefit the public good. This program questions whether shopping history data can be used in a positive way to support health research and the development of new interventions. The fellowship will establish the feasibility of novel ways of assessing both health outcomes and associated lifestyle choices through objective measures of real world behaviours reflected in retail shopping history data recorded through loyalty cards. At the same time it will build a framework that can be used by future researchers. My research programme in Yrs 1-4 will unfold in three stages. First, it will use commercially collected datasets to identify and study reproductive health outcomes through patterns in the shopping data. Second, it will validate patterns in the data which are associated with health outcomes using established Longitudinal Population Studies such as the Avon Longitudinal Study of Parents And Children (aka Children of the 90s). Third, I will use the linked datasets to research questions of population health importance in the domain of reproductive health, such as what are the true rates of miscarriages, how do women manage postpartum health and wellbeing, whether breastfeeding is better in the long run for children's mental health, and others. This will be done through studies with Children of the 90s participants and the general public helping to validate the results. The impact of the project will realised in Yrs 5-7 and include a conceptual change in techniques for studying population health, making it possible to identify lifestyle causes of diseases, assess the impact of national policies, and provide recommendations for health interventions.

Planned Impact

Creating the data sharing mechanisms described in Transaction Data for Population Health programme provides a gateway to unlock the potential of large datasets to study individual lifestyle choices for population health research. The ultimate long-term goal of this program is to put large commercial datasets - such as shopping history data - at the service of the public healthcare through contributing to early detection of diseases, developing and testing targeted interventions, and contributing to the evidence-based healthcare and health research. If this research is successful, we will benefit healthcare through improved allocative efficiency in public spending in a variety of domains as well as through novel ways to assess the success of health interventions. The programme will provide significant long-term benefits to the wider public through better provision and access to healthcare. The development and implementation of frameworks allowing companies to share consumer data with academic researchers for public benefit will create long-term impact for retailers and other industries that rely on consumer data through reputation benefits and lead to long-term social impact through improving lives of people. All outputs of the programme will be open access and publicised through relevant routes (e.g., policy reports, journal publications, reports available through project website).
In addition, the outputs of the fellowship will provide multiple short-term benefits to different groups of users, including public and third sector (i.e., policymakers, healthcare services, health charities), retail industry and the general public. These have been identified through my previous research and impact work, through monitoring policy developments, responding to government consultations (e.g., on Online Targeting and Personalisation, Department for Culture, Media and Sport), membership on policy-makers' Advisory Boards (e.g., for Centre for Data Ethics and Innovation) and extensive work with retail industry and the general public.
Public service departments, such as the National Health Service and the Office for National Statistics, will learn about relevant work on data sharing and data linkage, as well as novel data analytic techniques. This will impact their strategies of using new streams of data in their practices and create new opportunities for data driven research. Policy-makers and regulators, such as Department of Culture Media and Sport and Information Commission Officer, short-term benefit will constitute learning about novel data sharing practices that can help to improve policies and regulations, and create more efficient routes of sharing data between academia, industry and third sector.
Retail industry project partners (e.g., Walgreen Boots Alliance) will benefit from skills development and knowledge transfer. They will learn novel techniques of data analysis helping to understand their consumers better and to provide more targeted advertising to their consumers. This will ultimately increase their profits, build stronger relationships with consumers, achieve better visibility in a crowded marketplace and help to realise the value of their data for public services. Retailers will benefit from a newly created blueprint for satisfying data sharing requests from consumers and will have reputational benefits by helping to create a more fair and sustainable society through efficient use of already collected health data for public good.
The members of the general public and population studies participants who took part in the research will benefit in the short-term though learning about data sharing and possibilities of using loyalty cards data for research and will be empowered in their decisions of sharing their personal data for public good.

Publications

10 25 50