Developing and testing the effectiveness of mindfulness intervention for reducing loneliness in Hong Kong older adults: a hybrid sequential multiple assignment randomized trial, micro randomized trial experimental design study protocol

Abstract Background Loneliness is a growing public health concern among older adults, likewise in Hong Kong. Mindfulness-based interventions have shown promise in addressing loneliness, but often yield modest and short-lived effects. Digital delivery makes such interventions scalable but sustaining user engagement remains challenging. Adaptive interventions that integrate human support, mobile technology, and machine learning show promise, yet rigorous evidence remains limited. This study aims to develop and evaluate a machine learning-enabled mindfulness intervention to reduce loneliness and enhance well-being among older adults in Hong Kong. Methods Using hybrid Sequential Multiple Assignment Randomized Trial–Micro Randomized Trial (hybrid SMART–MRT), 250 lonely older adults will be randomized into one of the five first-line intervention conditions over 4 weeks; single session mindfulness intervention (SSMI) with no coaching, app-based coaching, human coaching, a combination of app-based and human coaching, or an active control. Nonresponders, defined as those who do not exhibit a decrease of at least 2.86 points on the Revised UCLA Loneliness Scale from baseline (T0) to the post-first-line assessment (T1) will be re-randomized to either intensified human coaching or booster session. Concurrently, MRTs will evaluate the proximal effects of push notifications, recommended mindfulness activities, and personalized motivational messages on engagement. Data will be collected at baseline (T0), within 1-week after each stage of intervention (T1 & T2), and three-month (T3) and six-month (T4) after completion of second-line intervention. Primary outcome is assessed by the Revised UCLA Loneliness Scale. Secondary outcomes include De Jong Gierveld Loneliness Scale, symptoms of anxiety and depression, perceived stress, and sleep quality. Potential mediators include mindfulness, perceived social support, and social network. Linear mixed models and weighted least squares regression will be used for analysis. This study is currently in the finalization phrase for intervention materials, mobile application development and personnel training. Recruitment will begin in May 2026, with data collection expected to be concluded by December 2026. Discussion This study will be among the first to apply a hybrid SMART–MRT design to a mindfulness-based loneliness intervention for older adults. Findings will inform the development of scalable, adaptive, and personalized interventions by identifying optimal combinations of human support and machine learning–driven digital strategies to enhance engagement and reduce loneliness. Clinical Trial Registration ClinicalTrials.gov, NCT07308613