Hybrid digital intervention cohort in university students: feasibility pilot study using smartphone- and smartwatch-based monitoring and ecological momentary interventions

This feasibility pilot study demonstrates that a hybrid digital intervention design combining passive smartwatch/smartphone monitoring with sequential ecological momentary interventions is feasible and acceptable for promoting healthy sleep, physical activity, and screen time behaviors among university students, despite observed declines in engagement over time.

The Gist

Imagine a university campus as a busy, chaotic train station. The students are the travelers, often running late, tired, and glued to their phones. They need help getting back on track with their sleep, exercise, and screen time, but traditional advice (like a pamphlet in a waiting room) often gets ignored.

This paper describes a pilot study called MOVE@NUS, which tried a different approach: instead of a pamphlet, the researchers gave the students a digital "co-pilot" living inside their own Apple Watches and iPhones.

Here is the story of what they did, how it went, and what they learned, explained simply.

The Big Idea: A "Test Drive" for Digital Health

Think of this study as a test drive for a new car before it goes into mass production. The researchers wanted to see if a specific type of digital car (a hybrid study design) could work for university students.

This "car" had two main engines running at the same time:

1. The Passive Engine (The Silent Observer): The Apple Watch and iPhone acted like a silent, invisible camera. They automatically recorded how much the students walked, climbed stairs, and slept, without the students having to do anything.
2. The Active Engine (The Nudge Coach): Every two weeks, the students played a mini-game called a "Randomized Controlled Trial" (RCT). They were secretly split into three groups:
   • Group A: Got specific tips to sleep better.
   • Group B: Got specific tips to move more.
   • Group C: Got specific tips to use their phone less.
   • (Note: There was also a "Control" group that just used the app normally, like a baseline).

The goal wasn't to prove which tip was the best yet, but to see if the whole system could run without breaking down.

The Journey: What Happened?

The study ran for five months with 65 first-year students. Here is how the journey unfolded:

1. The Sign-Up (Recruitment)
It was like trying to find people who own a very specific, high-end model of car. The researchers only accepted students who had both an iPhone and a newer Apple Watch.

• Result: They found 65 students who fit the bill. It was a bit hard to find enough people because the requirements were strict, but they got a solid group to start with.

2. The Silent Observer (Passive Data)
The Apple Watches were great at tracking movement. It was like having a faithful dog that never forgets to walk.

• Result: The watches successfully recorded data for 95% of the students. However, there was a catch: many students took their watches off at night (like taking off a watch to sleep), so the "sleep" data was a bit spotty. The "movement" data (steps and stairs) was very complete.

3. The Nudge Coach (Interventions)
The app sent little "nudges" (reminders) to the students.

• Sleep: "You slept 6 hours; try for 7!"
• Movement: "Take the stairs instead of the elevator!"
• Screen Time: "Put the phone down for a break!"
• Result: The students generally liked the sleep and movement nudges. They felt helpful. However, the screen time nudges were a bit tricky. Some students admitted that checking the notification to see if they were using their phone actually made them pick up the phone more, accidentally increasing their screen time!

4. The Fatigue Factor (Engagement)
Imagine a marathon where you have to stop and fill out a survey every few miles.

• Result: At the start, almost everyone filled out the surveys. But as the months went on, the "survey fatigue" set in. By the end, fewer than half of the students were completing the daily check-ins. The longer the study went on, the more people dropped out of the active parts, even though they kept wearing the watch.

The Verdict: Did the Car Start?

Yes, but it needs a tune-up.

The study proved that this "hybrid" approach (combining automatic tracking with active coaching) is feasible. It works. The technology didn't crash, the students didn't quit entirely, and the data came through.

However, the researchers found three main "potholes" on the road:

1. The "Apple Only" Barrier: By only accepting iPhone/Watch users, they missed out on many students. Future versions might need to work on Android phones too.
2. The "Nighttime Gap": Students didn't want to sleep with their watches on, making sleep data less reliable.
3. The "Nudge Backfire": For screen time, the reminders sometimes made the problem worse by drawing attention to the phone.

The Takeaway

The researchers concluded that this digital "co-pilot" is a promising vehicle for helping students get healthy. It's like a prototype that successfully drove off the lot, but the engineers now know they need to:

• Make the car compatible with more brands (Android).
• Make the "sleep mode" more comfortable so people don't take the watch off.
• Be smarter about how they remind people to put their phones down, so the reminder itself doesn't become a distraction.

In short: The idea works, the tech works, but the human habits need a little more fine-tuning for the next version.

Technical Summary

1. Problem Statement

University students face significant challenges in maintaining healthy movement behaviors (sleep, physical activity, and screen time) due to academic pressures, irregular schedules, and high reliance on digital devices. While digital health interventions using Ecological Momentary Assessments (EMAs) and Interventions (EMIs) offer real-time, context-aware solutions, their implementation faces critical feasibility hurdles:

• Engagement Decay: Response rates for self-reported EMAs often decline over time.
• Data Fragmentation: Passive sensor data (from wearables) can be incomplete due to device non-wear or technical synchronization issues.
• Lack of Integrated Frameworks: There is a scarcity of rigorous evaluations combining continuous passive monitoring with sequential randomized controlled trials (RCTs) to test multiple behavioral interventions simultaneously in a naturalistic setting.

2. Methodology

The study, titled MOVE@NUS, was a five-month pilot study (September 2024 – January 2025) conducted at the National University of Singapore (NUS).

Study Design

• Hybrid Approach: The study combined continuous passive monitoring with three sequentially embedded 3-arm Randomized Controlled Trials (RCTs).
  • RCT-1: Sleep (2 weeks)
  • RCT-2: Physical Activity (2 weeks)
  • RCT-3: Screen Time (2 weeks)
• Randomization: Participants were randomized 1:1:1 into Control, Intervention 1, or Intervention 2 for each RCT phase.
• Participants: 65 first-year undergraduates (mean age 20.4 years; 53.8% female).
  • Inclusion Criteria: Aged 18–25, owned an iPhone and Apple Watch (Series 6/SE or newer), and agreed to continuous wear.
  • Exclusion: Pregnancy or overseas travel >6 weeks.

Data Collection Architecture

1. Passive Monitoring (Primary):
   • Devices: Apple Watch and iPhone.
   • Platform: Integrated via Apple HealthKit through a custom app built on the Cogniss platform.
   • Metrics: Sleep duration (time in bed), step counts, flights climbed, and vigorous physical activity (VPA).
   • Screen Time: Collected via weekly manual uploads of iOS Settings screenshots (as HealthKit does not provide raw screen time data).
2. Active Reporting (Secondary):
   • EMAs: Administered in 8 bursts (3 days each) every two weeks. Participants reported on behaviors and context.
   • Surveys: Baseline, midway (2.5 months), and endpoint (5 months) questionnaires via REDCap.

Intervention Strategies (Ecological Momentary Interventions - EMIs)

Interventions were text-based, personalized, and limited to ≤7 notifications/day to prevent fatigue.

• Sleep:
  • Nudge: Weekly feedback on sleep duration vs. guidelines (7–9 hrs).
  • Choice: Participants selected weekly sleep hygiene strategies (e.g., no coffee after 3 PM).
• Physical Activity:
  • Nudge: Contextual reminders to take stairs, based on real-time flight data.
  • Choice: Daily selection of brief vigorous activities (e.g., squats between classes).
• Screen Time:
  • Nudge: Break reminders with alternative activity suggestions triggered by current usage status.
  • Goal-Setting: Personalized daily reduction goals with progress feedback.

Analysis

• Feasibility Metrics: Recruitment rates, retention, data completeness (HealthKit sync rates), and EMA response rates.
• Preliminary Effectiveness: Compared pre-intervention vs. during-intervention means for objective metrics (HealthKit) and self-reports, calculating mean differences with 95% Confidence Intervals (CIs).
• Framework: Used a micro-macro framework integrating TAM, COM-B, and MARS models to assess user experience.

3. Key Contributions

• Novel Hybrid Design: Validated a scalable framework embedding multiple sequential RCTs within a single continuous digital cohort, allowing for the testing of diverse intervention types without requiring separate recruitment cycles.
• Technical Integration: Demonstrated the feasibility of a unified app (Cogniss) integrating HealthKit for passive tracking, manual screenshot uploads for screen time, and EMA delivery on consumer-grade Apple devices.
• Hierarchy of Data Burden: Provided empirical evidence on the "hierarchy of compliance," showing that passive physical activity data had the highest completeness, while sleep data suffered from non-wear, and screen time data (requiring manual uploads) had the highest attrition.
• Behavioral Insights: Identified that intervention acceptability varies by domain (Sleep/PA > Screen Time) and that notification frequency significantly impacts user satisfaction.

4. Results

Recruitment and Retention

• Enrollment: 65 participants enrolled from 229 screened (28.4% enrollment rate).
• Retention:
  • Questionnaire completion was high: 100% (baseline), 89.2% (midway), 86.2% (endpoint).
  • EMA Engagement: Declined significantly over time, from 88.7% in the first burst to 49.2% in the final burst. Only 30.8% of participants completed all 7 EMA bursts.
  • Passive Data: 95.4% (62/65) provided HealthKit data. However, only 48.0% of the initial cohort maintained data provision through the final phase.

Data Completeness

• Physical Activity: Highest completeness (flights climbed and steps).
• Sleep: Moderate completeness; limited by participants not wearing watches overnight.
• Screen Time: Lowest completeness due to the manual nature of screenshot uploads.

Preliminary Effectiveness

• Sleep: No significant changes in duration across groups (stable ~7 hours).
• Physical Activity: Intervention groups showed small, non-significant declines in steps and flights climbed compared to controls.
• Screen Time: The "Goal-Setting" group showed a slight reduction (-1.0 hours), while "Break Nudge" and control groups showed minor increases.
• Note: The study was underpowered for definitive efficacy claims; the primary goal was feasibility.

User Experience

• Satisfaction: Highest for Sleep (78.9%) and Physical Activity (70.6%), lowest for Screen Time (60.0%).
• Feedback: Participants found sleep/PA nudges useful and relevant. Screen time interventions were perceived as repetitive and sometimes counterproductive (checking notifications increased screen time).
• Technical Issues: Occasional app crashes and app uninstallation after intervention phases disrupted data continuity.

5. Significance and Future Directions

• Proof of Concept: The study confirms that a hybrid digital cohort with embedded micro-trials is feasible for university students, offering a model for future large-scale behavioral research.
• Design Implications:
  • Platform Diversification: Future studies must include Android users to improve generalizability and recruitment.
  • Technical Robustness: Apps require seamless background synchronization and persistence across reinstalls to prevent data loss.
  • Intervention Duration: Engagement drops significantly after 2–3 months; future interventions should consider shorter, intensive bursts or milestone-based incentives.
  • Domain Specificity: Interventions for screen time require different strategies (e.g., automated tracking vs. manual reporting) compared to sleep or PA.
• Methodological Shift: The authors suggest moving from traditional sequential RCTs to micro-randomized trials (MRTs) or factorial designs in future iterations to optimize intervention components more efficiently.

Conclusion: The MOVE@NUS pilot successfully demonstrated the technical and behavioral feasibility of a complex, multi-behavior digital intervention framework. While challenges in long-term engagement and data completeness (particularly for sleep and screen time) remain, the study provides a critical roadmap for refining scalable digital health solutions for emerging adults.