PhD Research Project: Breaking Free with Artificial Intelligence: The Data Science of Digital Behav

Using digital technology to improve health and care is now a priority focus for the UK National Health Service. New and maturing technological developments hold the potential to have a transformative effect on health service delivery. Digital health solutions can aid in the reduction of waiting times, more effective use of limited resources and improved treatment fidelity but, once such technologies are implemented, the dearth of data that can be acquired through their adoption itself presents a challenge. Data-driven insights, driven by machine learning and artificial intelligence techniques, can offer a glimpse of previously unknown patterns and identify opportunities for the development of new healthcare practices or policies. As digital health platforms become more sophisticated, the data that can be acquired often becomes more complex and heterogeneous, machine-based interpretation becomes a near-necessity.

The Breaking Free Group (http://www.breakingfreegroup.com/) is a leading digital health and behaviour science company. Founded in 2010, it is formed from a team of psychologists, substance misuse experts and technology specialists. With an organisational philosophy built around the principles of state-of-the-art behaviour science and the UK Medical Research Council evaluation framework, they have developed a bespoke digital behavioural intervention platform that provides 24/7 access to a secure, GDPR-compliant, online digital cognitive behaviour therapy (CBT) platform with a particular focus on substance misuse. Users of the platform complete pre- and post-therapy questionnaires and interactive exercises designed to teach strategies to overcome addictive behaviours and improve wellbeing. Scoring on questionnaire responses can signpost users to personalised information sources. Real-time quantitative data from the platform can be anonymously fed to health service providers and commissioners for planning and policy development purposes.

Breaking Free has seen its digital platform commissioned for use by the Ministry of Justice in UK prisons, Non-Government Charitable Organisations and NHS Trusts. Uniquely, Breaking Free operates under a research ethics protocol approved by the NHS Health Research Authority and endorsement by the National Institute of Clinical Excellence allowing the data they collect to be utilised for broader research applications. More recently, two new online platforms have been deployed more widely to the public targeting alcohol consumption reduction and smoking cessation.

In this PhD project, we propose to initiate a new multi-disciplinary collaboration between Turing, Birmingham and Breaking Free Group with the PhD student at the heart of the partnership. The student will develop novel data science methodologies to exploit real-time digital health monitoring and intervention data. Our working exemplars will be based on the unique behavioural insights captured by the Breaking Free Group but the research outputs would have generic applicability in other related settings including youth mental health conditions.

The project will address a number of key data problems including:

1. The identification of complex multi-faceted predictive factors from self-reported user data that might inform outcomes,
2. Stratification to identify previously uncharacterised sub-groups of users for novel therapeutic planning and design,
3. Dynamic models, including behavioural trajectory models and feedback mechanisms, to quantitatively describe the effect of behavioural interventions.

While analytical techniques exist for tackling such questions in the retrospective analysis of static data sets, extensions to the live, quasi-experimental settings provided by digital health providers, such as Breaking Free, opens new research challenges and opportunities. Examples include the integration of heterogeneous data types from highly structured questionnaire responses to usage event data and the use of formal approaches for causal and counterfactual modelling in order to understand intervention effects (e.g. what if the intervention hadn't been applied?). Scalable implementations of algorithms will also be required for any interactive applications using machine learning, Breaking Free offers its services across a variety of digital platform including smartphones, tablets and personal computers. In addition, with data collected on a UK-wide population scale and location data (postcode), there is the need to consider potential spatial relationships and integrating with publicly available socioeconomic data to understand geographical variation. Data-driven insights could lead to opportunities to develop robust evaluation studies (e.g. randomised control trials) which underpins the research evaluation philosophy of Breaking Free and which they have significant experience in developing.

Quantitative background acquired from a first or second degree in Mathematics, Computer Science, Physics or Engineering. Candidate should have a strong interest in scientific applications of machine learning. This project may also be suitable for quantitatively capable psychology or behaviour science graduates interested in cross-training in statistical machine learning.

Funding Notes

To support students the Turing offers a generous tax-free stipend of £20,500 per annum, a travel allowance and conference fund, and tuition fees for a period of 3.5 years.

References

• KR Campbell, C Yau (2018). Uncovering pseudotemporal trajectories with covariates from single cell and bulk expression data. Nature Communications 9 (1), 2442.
• T Rukat, C Yau (2018) Probabilistic Boolean Tensor Factorisation. International Conference on Machine Learning.
• S Brugger, M Broome (2018) Computational psychiatry. The Routledge Handbook of the Computational Mind, 468-484
• Elison S, Jones A, Ward J, Dugdale S, Davies G: Examining effectiveness of tailorable computer-assisted therapy programmes for substance misuse: Programme usage and clinical outcomes data from Breaking Free Online. Addictive Behaviors 2017, 74:140-147.
• Elison S, Ward J, Williams C, Espie C, Davies G, Dugdale S, Ragan K, Chisnall L, Lidbetter N, Smith K: Feasibility of a UK community-based, eTherapy mental health service in Greater Manchester: repeated-measures and between-groups study of ‘Living Life to the Full Interactive', ‘Sleepio' and ‘Breaking Free Online' at ‘Self Help Services'. BMJ Open 2017, 7:1-10.