My major current interest is to understand how we can use novel digital footprint data to study human behaviour and real-life outcomes, such as health. Currently I am focusing on transaction data, specifically loyalty and banking cards, and working on realising the value of using these data to improve population health. My other research includes work on Value of Personal Data, Cooperation/Prosocial Behaviour and Individual Differences in Decision-Making.
I am leading a Transaction Data for Population Health programme funded through UKRI Future Leaders Fellowship (2021-2028) at Population Health Sciences, Bristol Medical School. This research programme is grouped around four main themes: Data Donation, Analytic Techniques for Large Transaction Datasets, Ground Truth for Patterns in Transaction Data, and Data Linkage.
I am also interested in applied health research with transaction data. Loyalty cards data can help to understand diet and alcohol consumption, while banking records can help to unpack factors influencing wellbeing and mental health. I am interested in a wide range of applications areas that can benefit from information derived from very large population level datasets including reproductive health, pain management, nutrition (Skatova et al, in preparation), gambling and others.
Data Donation (Skatova & Goulding, 2019; Blog). Data donation is a concept encapsulating an active decision by an individual to donate their personal data for public good (e.g., population health research). While data ownership is still a grey area, General Data Protection Regulation (GDPR, introduced in the UK in 2018) allows individuals to exercise “right to portability”. This right allows an individual to request a copy of personal data which is collected on them by an organisation. Individuals can also authorise/consent to a third party (e.g., an academic health researcher) to access their personal data. I work on unpicking motivations and barriers to donate personal data, and study how those can fit with the broader industry and regulatory landscape of personal data sharing.
Analytic Techniques for Large Transaction Datasets. Using very large transaction datasets opens up exciting opportunities to study human behaviour and health. However, there are many pitfalls and obstacles to overcome. By drawing on behavioural science, statistics and machine learning I develop new data analytic techniques that can be employed to understand the mechanisms of choices and behaviour using transaction data.
Ground truth for patterns in transaction data. A major difficulty with realising full value of transaction data is the lack of “ground truth” in standalone datasets such as loyalty cards data. The patterns in the data (e.g., escalating alcohol purchasing) need to be verified against other data to be sure that such patterns reflect individual behaviour (e.g., escalating alcohol consumption).
My work on linking transaction data into Longitudinal Population Studies (LPS) allows to use depth and breadth of LPS health and social variables as ground truth for patterns that we can detect from transaction data. By establishing patterns in linked datasets through LPS, we can set out to study those patterns in large standalone transaction datasets contributing to the knowledge about physical and mental health, wellbeing, everyday behaviours and life events.
Data Linkage (Skatova et al, 2019; Blog). I work with Avon Longitudinal Study of Parents and Children (ALSPAC) and other Longitudinal Population Studies (LPS) to create a methodology for linking transaction data into LPS. Using qualitative and quantitative methods, I study a range of issues including participants’ attitudes to sharing transaction data with academic researchers, privacy-preserving, ethical approaches for data linkage, assessing the quality of linked datasets in terms of sampling and various biases, creating appropriate infrastructure for the linkages, creating routes for the linked data to be available for academic research via secure protocols.