| Abstract | Background:
Healthy lifestyle behaviours, including healthy diet and regular physical activity, are important behavioural determinants of the quality of life in midlife women. Midlife for women is also a key life stage that is associated with increased health risks of developing long-term health conditions, including heart disease, dementia, and osteoporosis. With over half of the global population being women and over a third of the entire UK female population currently experiencing menopause, promoting greater options for women to manage their healthy lifestyle behaviours is a public health priority. Moreover, the accessibility, convenience, and wide reach of digital health media, including mobile applications and wearable devices, makes it possible to deliver digital health behaviour change interventions (DHBCIs) to midlife women that are personalised, timely, and at scale. However, theory-and-evidence-based intervention designs are needed to improve our understanding of how to explain, influence, and predict health behaviours. Although there is a lot of interest to develop accessible and scalable DHBCIs that empower people to self-manage modifiable risk factors, there are currently no DHBCIs that provide personalised long-term lifestyle health management to midlife women in the UK.
Aim:
The aim of this PhD thesis was to design and optimise theory and evidence-based personalised DHBCI to enhance lifestyle health behaviours among midlife women in the UK.
Methods:
The thesis consists of four empirical studies and a co-production approach with a design phase followed by an optimisation phase of a DHBCI. The multimethod design consists of three design workstreams including: 1) a systematic review of behaviour change techniques (BCTs) used in DHIs for midlife women, 2) a mixed-method study to understand UK-residing midlife women’s lived experiences with lifestyle health behaviours using focus groups, and 3) a co-production approach to co-design key components of the DHBCI. Underpinned by the Behaviour Change Wheel (BCW) framework, the design is theory and evidence-based, with a group-level personalisation facilitated by a Person-based approach (PBA) in the design phase. Subsequently, the multimethod intervention design was tested in a 4) a single-arm feasibility and acceptability study (intervention) utilising multiple data collection modalities (e.g., ecological momentary assessment (EMA) app, fitness tracker, survey), and optimised in 5) a study exploring machine learning (ML) algorithms to identify predictors and groups of BCTs to the target behaviours using the intervention dataset, and to identify strategies for dynamic, continuous and individualised DHBCI for women in midlife, in the UK. Results:
Many DHBCIs lack theoretical grounding and identification of their key intervention components (e.g., BCTs, mechanism of action, intervention content) and are therefore too generic to be effective and are difficult to replicate. Personalisation of interventions requires input from the target population strengthened by co-production to ensure the intervention meets the needs of the target population. Applying the BCW framework to identify intervention components in the combined multi-study design is feasible and provides a systematic and replicable design. Testing the design in the intervention reached feasibility with a high level of participant engagement with no dropouts, and high level of acceptability. Finally, using supervised ML feature selection models to identify sets of intervention components (with intervention features linked to groups of BCTs) with the greatest impact on predicting target behaviours (using the intervention dataset) was feasible. Prediction accuracy of the ML model and predictive power was found acceptable in social science research.
Conclusion:
Identifying behavioural determinants personalised for the target population and providing coaching support is essential to achieve high level of engagement and retention in DHIs. Considering multiple complementary design inputs provides the necessary foundation for DHBCIs designs that are multi-faceted with wide-ranging considerations. Additionally, co-designing DHBCIs with the public facilitates greater relevance of the content and acceptability of the intervention. The intervention’s multiple data collection modalities allow for capturing a wide range of behavioural data that can be further exploited by ML algorithms to identify predictors that are the most relevant in behaviour change. Utilising ML has the potential to improve personalisation, engagement, and effectiveness of adaptive and continuous DHBCIs on a long-term basis. Therefore, a multi-behavioural, multi-design, multimethod, and multimodal intervention aimed to improve healthy eating and regular physical activity in UK-residing midlife women is feasible and acceptable. |
|---|