Using behavioural insights to improve the effectiveness of digital weight loss interventions

Excess weight is one of the leading preventable causes of morbidity and mortality in high income countries. Behavioural weight loss programmes are effective for weight loss but often resource-intensive and costly, limiting their scalability. Digitally delivered interventions offer a promising alternative, yet their effectiveness is limited by challenges in maintaining engagement and retention. Therefore, this thesis investigated possible strategies to optimise the effectiveness of digital weight loss programmes, in part through enhancing engagement. The work was conducted in partnership with Second Nature, a digital weight management service provider to the NHS. I initially conducted a prospective observational study using existing Second Nature data to explore the association of goal setting with weight loss and programme dropout. I found that individuals who set higher weight loss goals or reported health- or fitness-related motivations achieved greater weight loss and were less likely to drop out. Alongside this I helped to develop a fully automated, self-management app, grounded in self-regulation theory (ARTEMIS: Adults Regulating Their weight Everyday with Mobile Internet Support). The app was evaluated in a large-scale randomised controlled trial (n = 1,607), compared to simple advice to lose weight. At six months, the app led to 1.85 kg greater weight loss, doubled the odds of losing ≥ 5% body weight, halved the odds of symptoms of disordered eating, and improved body image. The modest engagement levels and effect sizes from a self-managed intervention like ARTEMIS prompted me to explore whether adding support components could improve outcomes. I applied the Multiphase Optimisation Strategy (MOST) framework to develop and test four candidate components: an introductory video call with a health coach, coaching drop-in webchat sessions, goal-setting statements, and food diary reviews plus feedback. A 2⁴ factorial optimisation trial was conducted to evaluate their individual and combined effects. I found that the health coach intro call led to 1kg greater weight loss at 24 weeks, while the food diary was associated with poorer outcomes and engagement. Engagement and retention were a recurring issue throughout the thesis. Thus, the final study explored ‘planned pauses’ in dieting as a novel strategy to sustain adherence. A systematic review and meta-analysis compared planned pauses with continuous energy restriction. Planned pauses produced weight loss outcomes comparable to continuous energy restriction, but with no clear evidence that they improved retention. Overall, this thesis illustrates how behavioural theory can inform multiple stages of intervention development, from observational studies to optimisation trials, to enhance engagement and outcomes. Both fully automated and supported interventions were effective, though not universally so, and sustaining engagement remained a consistent challenge. Further work is needed to identify what works best for whom and when, while enabling more responsive collaboration between academia and industry. This highlights the need for more flexible, adaptive evaluation frameworks, and in the discussion, I proposed a set of design principles to support a more fit-for-purpose framework for digital health interventions.