Impact of Gamification in Behaviour Change Intervention: A Randomised Controlled Trial with YuLife's Health and Wellbeing App

Abstract

Background: Companies in the Health and Life Insurance space are increasingly turning to digital tools to promote healthier behaviours among their user base and reduce future health risks. This approach shifts insurers' role from passive underwriters to partners in health management. These tools, often smartphone or wearable-tracker-based, enable real-time monitoring of behaviours (such as physical activity or meditation), providing fruitful targets for behavioural change interventions. Gamification, a Behavioural Change Technique with rich theoretical backing, is increasingly used in this context; however, despite its theoretical promise, current evidence remains mixed, and makes it hard to disambiguate its effect compared to more isolated financial incentives, the extent to which initial effects may be sustained over time, and how such changes in behaviour potentially translate to downstream health risk reductions. Objective: This 9-month parallel-group, open-label Randomised Controlled Trial was designed to assess the causal impact of gamification in promoting health behaviours, independent of financial incentivisation. This was conducted in a real-world workplace setting, involving a cohort of participants using the YuLife Health and Wellbeing app, provided within an employer-sponsored group cover setting. Methods: For the purposes of the RCT, the app was adapted such that gamification features could be turned on or off in a controlled manner, and in-app rewards in the form of "YuCoin" were adjusted between treatment groups to account for the effect of financial incentives. Following a baseline phase involving acquisition of baseline step estimates and questionnaire data, 1,288 participants -- recruited from a number of companies partnered with YuLife, spanning various sectors -- were randomised to gamified versus non-gamified versions of the app using stratified block-randomisation, and evaluated at specific milestones over a 9-month period, to enable comparison of short-term to long-term outcomes. The primary outcomes assessed were absolute differences in mean daily step count and engagement with the YuLife app. The data were analysed using Linear Mixed-Effects Models (LMMs). Additionally, a Cox Proportional Hazards model fitted to UK Biobank data was used to map step differences directly onto downstream health risks, and reductions were evaluated using an LMM. Further secondary outcomes (such as smoking and alcohol consumption) were also evaluated using non-parametric statistics. Results: Compared with control, the gamified intervention was associated with greater mean daily steps throughout the study, with month / intervention interaction effects reaching one-sided 5% significance at months 3 (β=473.84, p=0.027), 5 (β=626.54, p=0.006), and 9 (β=480.91, p=0.033). Additionally, strong seasonal effects were identified, with fewer steps in Autumn (β≈−943.50, p<0.001) and Winter (β≈−1,145.45, p<0.001) versus Summer; higher baseline activity was a strong predictor of later activity (β≈0.85, p<0.001) and higher BMI was negatively associated with steps (β≈−60.84 per unit, p<0.001). For app engagement, month / intervention interactions were positive and significant from Month 3 onwards (Month 3 β=0.205, Month 5 β=0.182, Month 7 β=0.170, Month 9 β=0.175, all p<0.001), effectively showing sustained engagement while main milestone terms indicated declines in the control arm. Sensitivity analyses demonstrated the potential for baseline step inflation due to novelty effects, motivating repeating the step count analyses under an alternative baseline definition; this showed similar results, but with interaction effects achieving one-sided significance over all study milestones. Predicted partial‑hazard analyses showed progressively larger month / intervention reductions in hazard, reaching one-sided significance at months 5 (coef=−0.018, p=0.016) and 9 (coef=−0.026, p=0.002). No significant intervention effects were observed for other secondary outcomes (e.g. smoking, alcohol) following Bonferroni-Holm correction. Conclusions: Gamification elements can be an effective component in the context of digital interventions aiming to promote positive health behaviours, leading to improved engagement with the intervention and positive behavioural outcomes. Through progressive risk-reduction, even small but sustained improvements can be shown to meaningfully improve long-term health outcomes. Gamification is likely to add value to workplace health promotion initiatives, particularly for targeted short- to medium-term behavioural change interventions operating within a larger risk-management framework. Trial Pre-registration: https://osf.io/926pd