The population under scrutiny is far from arbitrary. First-year undergraduates face a perfect storm of academic pressure, irregular schedules, and newfound independence, and their movement behaviours, meaning sleep, physical activity, and screen time, are tightly interwoven. Insufficient physical activity contributes to obesity and cardiovascular disease over the long term, while poor sleep has been linked to impaired cognition, weaker academic performance, and elevated risks of mental health problems. Excessive screen time, meanwhile, displaces physical activity, disrupts sleep through blue-light exposure and psychological stimulation, and has been associated with poorer mental health in young people. Because these behaviours influence one another throughout the day, the researchers argue that any serious intervention must capture them simultaneously and in real time, rather than relying on retrospective questionnaires that are vulnerable to recall bias.

The technological rationale is compelling. In Singapore, smartphone penetration among people aged 15 to 24 reached 100 percent by 2021, and wearable device use in the general population is estimated at 35 to 55 percent. Modern iPhones and Apple Watches continuously record step counts, stair climbs, heart rate, and time in bed, all retrievable through Apple’s HealthKit framework, which has independently demonstrated validity for measuring gait parameters in adults. The MOVE@NUS team built a dedicated study app on the Cogniss no-code platform, integrating HealthKit data streams with two distinct research tools: ecological momentary assessments, or EMAs, which collect self-reported data in real time, and ecological momentary interventions, or EMIs, which deliver tailored health prompts at the moment they are most likely to matter. Every EMA and EMI is text-based and takes less than two minutes to complete, a deliberate design choice to minimise participant burden.

The study’s architecture is its most distinctive feature. Rather than testing one sprawling multi-component programme, the pilot embeds three sequential randomised controlled trials within a single continuous monitoring cohort. RCT-1 targets sleep, RCT-2 targets physical activity, and RCT-3 targets screen time, each running for two weeks and separated by at least two weeks of passive observation. For every trial, participants are re-randomised in a 1:1:1 ratio into two intervention arms and a control group, with allocation concealed from both the delivery system and the participants themselves. Because each participant serves as their own comparator across phases, and because prior allocation status is modelled statistically as a covariate, the design allows the team to isolate the effect of each intervention while balancing carryover effects across the study population.

The interventions themselves are grounded in behavioural science and tailored to the rhythms of student life. In the sleep trial, one group receives weekly personalised feedback on sleep duration aligned with the recommended seven to nine hours per night for young adults, while a second group selects weekly from evidence-based sleep hygiene strategies, such as avoiding screens before bed or skipping coffee after 3 p.m., with daily reminders tied to their chosen tactics. In the physical activity trial, one arm receives daily nudges to climb stairs instead of taking lifts, with feedback drawn from HealthKit data on flights climbed, while the other is prompted to choose short bursts of vigorous activity, like brisk walks between classes or squats while waiting. The screen time trial contrasts context-aware break reminders, which query current phone usage before suggesting stretches or walks, against a goal-setting arm in which participants set daily reduction targets supported by iOS screen time data.

Feasibility, not definitive effectiveness, is the primary endpoint, and the team measures it from every angle. Recruitment metrics include eligibility, enrolment, and non-response rates. Retention and engagement are tracked through questionnaire completion, EMA burst response rates, and the number of days of valid HealthKit data, with a day counted as valid only if the watch records at least ten hours of wear time. User experience is evaluated through a mixed-methods approach combining short surveys after each trial phase with optional semi-structured interviews, structured by a micro-macro framework that draws on the Technology Acceptance Model, the COM-B model of behaviour, and social and digital determinants of health. This dual lens captures both immediate app-level experiences and the broader contextual factors that shape whether students actually engage with the technology.

Quality assurance receives unusually rigorous treatment for a pilot. The app underwent internal alpha testing to catch functional errors, followed by beta testing with a small independent group of five users who assessed usability, navigation, and system reliability before launch. Intervention delivery fidelity is verified on two axes: notifications must arrive within ten minutes of their programmed schedule, and dynamically personalised content is tested to ensure it parses and displays correctly according to predefined triggers. The team is candid about one technical constraint: because some personalised messages depend on participants manually checking and entering their own Apple Health data, the system functions as an adaptive intervention rather than a fully automated just-in-time adaptive intervention. The main MOVE@NUS cohort is expected to close this gap by integrating real-time HealthKit data directly into the personalisation algorithms.

The analytical plan reflects the complexity of the design. Preliminary effectiveness will be explored separately for each embedded trial using linear mixed-effects models, which handle repeated measures by incorporating random effects for individual participants. Sleep duration is proxied by HealthKit time-in-bed records restricted to nights with valid nocturnal data, physical activity is captured through step counts and flights climbed, and screen time comes from weekly uploads of iOS Settings screenshots. Missingness patterns will be examined across groups, with sensitivity analyses planned if attrition appears linked to intervention allocation, since digital health studies often produce data that are missing not at random. Generalised estimating equations may be deployed for categorical outcomes, and all analyses will be conducted in R with two-sided tests at a five percent significance level.

The researchers are equally transparent about the study’s limitations. Requiring participants to own an iPhone and a recent Apple Watch introduces selection bias that may limit generalisability across socioeconomic groups, and the impossibility of blinding participants to whether they receive daily nudges creates a risk of performance bias and the Hawthorne effect, the well-documented tendency for people to change behaviour when they know they are being observed. Objective HealthKit metrics are less susceptible to this distortion than self-reports, but subjective findings will be interpreted with caution. No formal data monitoring committee or interim stopping rules are in place, reflecting the low-risk, non-invasive nature of the interventions, though the team reviews operational issues every two weeks.

What makes MOVE@NUS more than a campus experiment is its scalability thesis. Smartphones can reach vast populations in naturalistic settings at low marginal cost, and students who learn healthier routines during the transition to independent living will soon carry those habits into the workforce. By demonstrating whether continuous monitoring can be combined with rapid-cycle, embedded trials of personalised nudges, the pilot aims to establish the methodological foundation for a full-scale digital intervention cohort. If it succeeds, the payoff could extend well beyond Singapore’s universities, offering a template for real-time, behaviour-aware health promotion that meets people where they already are: on their phones.

Subject of Research: A pilot digital intervention study using smartwatch monitoring and embedded randomised trials to promote healthy movement behaviours in university students

Article Title: MOVE@NUS digital intervention cohort: protocol of a pilot study to promote healthy movement in university students

Article References: Chen, M., Movia, M., Chua, X. H., Tan, S. Y. X., Zheng, S., Jin, K., Topothai, T., Padmapriya, N., Müller-Riemenschneider, F., & Edney, S. (2026). MOVE@NUS digital intervention cohort: protocol of a pilot study to promote healthy movement in university students. Journal of Activity, Sedentary and Sleep Behaviors, 5(1), Article 5. https://doi.org/10.1186/s44167-026-00097-z

Image Credits: AI Generated

DOI: 10.1186/s44167-026-00097-z

Keywords: digital health, mHealth, ecological momentary intervention, wearable technology, physical activity, sleep, screen time, university students, randomised controlled trial, behaviour change, Apple HealthKit, Singapore

The stakes are enormous. The second Lancet Commission on Adolescent Health and Wellbeing has estimated that by 2030 more than one billion adolescents aged 10 to 24 — roughly half of all adolescents globally — will live in countries where preventable health problems threaten their wellbeing. During early and middle adolescence, six intertwined risk behaviours tend to emerge: poor nutrition, inadequate physical activity, poor sleep, excessive recreational screen time, alcohol use, and smoking or vaping. Researchers call these the “Big 6,” and they rarely appear in isolation. They cluster together, reinforce one another, and track stubbornly into adulthood, laying the foundations for chronic disease decades later. Families experiencing socio-economic disadvantage are hit hardest, facing structural and systemic barriers that make every one of these behaviours more likely. In Australia, adolescents in the most disadvantaged areas are more likely to drink at risky levels and to smoke, and less likely to meet fruit consumption guidelines, than their more advantaged peers.

Parents, the research team argues, are the most underused lever in prevention. Caregivers shape adolescent health through role modelling, rule-setting, monitoring and communication, and prior trials of parent-based programs have shown real promise, particularly when paired with youth-focused components. Yet engaging parents is notoriously difficult. Time scarcity and competing demands consistently emerge as barriers, and these are magnified for parents juggling financial stress, household pressure and limited social support. Digital interventions promise a way around these obstacles — flexible, remote, and free of travel and childcare costs — but a nagging question has remained: how much human support do parents actually need? Guided programs tend to boost engagement and adherence, while unguided ones are cheaper and scale effortlessly. Nobody knew where the sweet spot lay, especially for disadvantaged families, and a systematic review by the same group found only one parent intervention in the entire literature designed specifically for them.

Enter the Multiphase Optimisation Strategy, or MOST, a framework borrowed from engineering that treats an intervention not as a monolithic package but as an assembly of independently testable parts. Rather than running a traditional randomised controlled trial that asks “does the whole program work?”, MOST asks “which parts work, which parts are dead weight, and which combinations deliver the most benefit for the least burden?” The framework unfolds in three phases — preparation, optimisation and evaluation — and the new study represents the optimisation phase, the first time MOST has ever been applied to a digital intervention helping parents improve multiple adolescent health behaviours. In the preparation phase, the team built a conceptual model grounded in Social Cognitive Theory and Family Systems Theory, then co-designed every component with parents and adolescents from disadvantaged communities through workshops, school focus groups and a standing parent–adolescent advisory group.

The trial itself was a feat of experimental design. Between April and October 2025, the researchers recruited 303 parents of 11-to-15-year-olds living in the 40 per cent most disadvantaged areas of New South Wales, Australia, identified using the Australian Bureau of Statistics’ Index of Relative Socio-economic Disadvantage. Every parent received a core package of six online learning modules — one each for screen time, smoking and vaping, alcohol, sleep, healthy eating and physical activity — each roughly 20 minutes long, built around podcast-style videos with health professionals, interactive strategy summaries and goal-setting activities. Parents were then randomised across 16 experimental conditions in a 2×2×2×2 factorial design, with four additional components switched on or off: automated text messages, tailored feedback on their parenting, a stress management module, and weekly telehealth coaching calls from trained non-specialist health coaches. After data quality checks removed fraudulent responses, 292 parents formed the final baseline sample, with 79 per cent retained at the three-month follow-up.

The primary outcome was change in parental encouragement of adolescent health habits, measured with the validated Health Habits sub-scale of the Parenting to Reduce Adolescent Depression Anxiety Scale, a 12-item measure scored from 0 to 12 with very high reliability in this sample. The statistical machinery was correspondingly sophisticated: pre-specified linear mixed models with effect coding estimated each component’s main effect over time, while every possible two-, three- and four-way interaction was tested for synergy or antagonism. Missing data were handled by maximum likelihood, and analyses ran on an intention-to-treat basis. Notably, the team used a more liberal significance threshold of p < 0.10, consistent with MOST guidance, to avoid prematurely discarding potentially useful components during this exploratory screening stage.

The results delivered surprises in both directions. Telehealth coaching significantly improved parental encouragement (Cohen’s d = 0.38), as did tailored feedback (d = 0.26), and there was no statistical difference between the two. Crucially, combining them produced no significant synergistic effect — a direct challenge to the assumption that more support is always better. Text messages, meanwhile, backfired spectacularly: the component significantly reduced parental encouragement over time (d = −0.41), despite being co-designed with parents and intended as gentle encouragement. The authors speculate that the semi-automated messages may have added pressure for time-poor parents, felt too generic, or simply drowned in the noise of daily notifications. Stress management training showed no effect at all, possibly because a single module delivered late in the sequence could not shift entrenched household stress within the trial’s timeframe.

The optimisation decision then hinged on burden, not just efficacy. Parents who received both effective components rated telehealth coaching as significantly more effortful than tailored feedback (mean 2.4 versus 2.1 on a five-point effort scale). Coaching also demands real-world resources — trained staff, scheduling, fidelity monitoring — that cap scalability, whereas automated feedback costs almost nothing once built. Weighing effectiveness against effort, the team selected the package combining the online modules with automated tailored feedback alone as the optimised intervention. Feedback surveys suggested parents embraced it: 97.7 per cent of respondents reported a good or very good experience, 90.6 per cent found the length about right, and 80 per cent found it easy to fit into their lives. Coaching fidelity, for the record, averaged 91.9 per cent, so the component’s screening-out reflected burden, not poor delivery.

The study is not without limitations. The sample was 87 per cent mothers and drawn from a single Australian state, and because eligibility was based on area-level rather than individual-level disadvantage, many participants held post-secondary qualifications and moderate incomes, meaning the findings may not extend to families facing the most severe deprivation. The primary outcome relied on parent self-report, leaving room for social desirability bias, and internet access was required to participate. Yet the strengths are considerable: strong retention, rigorous fraud screening, prospectively registered methods, and a design that finally puts disadvantaged families at the centre of prevention science rather than its margins.

The bigger story is methodological. By dismantling a multicomponent intervention and rebuilding it from empirically validated parts, the researchers demonstrated that intervention science can work like engineering — identifying what earns its place and what does not. The optimised Health4Life Parents & Teens program now moves into the evaluation phase, a cluster randomised controlled trial with economic analysis that will test whether pairing it with the existing school-based Health4Life program improves the Big 6 behaviours in parent–adolescent dyads. If it succeeds, the payoff extends far beyond Australia: the same optimisation logic could stretch scarce prevention dollars further in low- and middle-income settings, and future work may explore whether artificial intelligence can deliver coaching-like support without the human cost. For a field long guilty of bolting components together and hoping for the best, this trial offers a sharper, more honest blueprint — and a reminder that sometimes the most powerful thing a program can do is know what to leave out.

Subject of Research: Optimisation of a digital parent-based intervention for adolescent health risk behaviours using the MOST framework

Article Title: Optimising a digital parent-based intervention to improve adolescent health risk behaviours in families living in socio-economically disadvantaged communities: findings from the Health4Life Parents & Teens factorial randomised trial

Article References: Champion, K. E., Davidson, L., Sunderland, M., Hunter, E., Spring, B., Thornton, L., Haidinger, A., Finn, T., Osman, B., Chapman, C., Burrows, T., Slade, T., Partridge, S. R., Gardner, L. A., Parmenter, B. J., Baur, L. A., Teesson, M., Mihalopoulos, C., Johnson, G., … Newton, N. C. (2026). Optimising a digital parent-based intervention to improve adolescent health risk behaviours in families living in socio-economically disadvantaged communities: findings from the Health4Life Parents &amp; Teens factorial randomised trial. The Lancet Regional Health – Western Pacific, Article 101987. https://doi.org/10.1016/j.lanwpc.2026.101987

Image Credits: AI Generated

DOI: Not provided

Keywords: adolescent health, digital intervention, parenting, MOST framework, factorial randomised trial, health risk behaviours, socio-economic disadvantage, telehealth coaching, tailored feedback, prevention, Australia, behaviour change