A Personalized Technology-Based Intervention for Physical Activity

Although the positive effects of physical activity on well-being are well documented, many individuals still failed to exercise sufficiently. This study leveraged information and communications technology (ICT) based interventions to promote exercise behavior among healthy NUS students in Singapore. The study introduced a personalized technology-based intervention that used a mobile app paired with a wireless activity tracker, with app features personalized based on each participant's current motivations and motives to exercise.

The goal was to evaluate changes in individual exercise behavior and motivation following use of the intervention. Both behavioral and psychological data were collected from participants. Behavioral data were collected from participants' assigned activity trackers at the beginning and end of the study, consisting of daily information such as exercise duration and step count. Psychological data on participants' exercise motivations and motives were collected via survey questionnaires. Behavioral and psychological data were triangulated to study the relationship between the psychological and behavioral aspects of participants' physical activity.

Drawing on self-determination theory and organismic integration theory, which conceptualize exercise motivation along a continuum from amotivation to intrinsic regulation, this study tested two hypotheses:

Salient nudging - dynamically increasing the visual prominence of app features that align with a user's current motivational profile, without altering feature availability - would be more effective than a non-tailored feature layout at promoting physical activity behavior. Individuals receiving salient nudging were expected to show greater increases in objective physical activity (daily step count, MVPA minutes) than individuals in the non-tailored control condition.

The effect of salient nudging on physical activity would be partially mediated by exercise motivation. Individuals receiving salient nudging were expected to progress toward more autonomous forms of motivation, and this shift in motivation was expected to account, in part, for the increase in physical activity attributable to salient nudging.

Recruitment

Potential participants were recruited via an in-campus recruitment advertisement and posters placed at strategic locations in the National University of Singapore (NUS). The advertisement informed potential participants that they would be using a smartphone and a wireless activity tracker throughout the study, and that they could choose not to join if they did not want to use either device.

Interested participants signed up via a URL link or QR code on the advertisement, which led to a Google Form. The form informed them that they would be invited to attend an information session, briefed together with other interested participants as a group. Information sessions were conducted over multiple timeslots to accommodate different schedules. The form collected each participant's NUS email address and preferred timeslot; a member of the research team then sent a confirmation email inviting them to attend at their selected timeslot.

At the start of each information session, participants were briefed on the study overview, the main features of the mobile app, and what to expect during the study. The Participant Information Sheet (PIS) and Consent Form (CF) were then handed out and explained. Interested participants who declined to sign the CF after the briefing could leave without penalty. Participants who wished to take part indicated their interest by signing and submitting the CF, and were permitted to keep their PIS.

The personalization aspect of the app was not explicitly disclosed in the PIS and CF, and was not mentioned when briefing participants in the first session, to avoid compromising the internal validity of the study through experimenter or participant expectancy effects. The tailored-versus-nontailored comparison was not disclosed to the participants throughout the trial.

Description of the Technology-Based Intervention Platform

The intervention platform comprised a mobile app, named "iMovin'," and a wireless activity tracker. Developed for iOS and Android, iMovin' was designed to promote physical activity and was free to download. The ICT features in iMovin' included:

• A dashboard showing daily physical activity history
• An in-app questionnaire
• An achievement dashboard
• Notifications when users were close to achieving exercise milestones
• A social media feed for users to post pictures and text about their exercise activities
• An online library of exercise-based research articles
• A goal planning module for setting exercise goals
• A dashboard for monitoring exercise goals
• An online forum for posting questions, commenting on posts, and replying to comments Synchronization (pairing) with an assigned wireless activity tracker to process physical activity data iMovin' personalized its features for each predefined user group, with each group consisting of participants who shared a common set of motivations and motives to exercise. The system automatically selected a predefined set of ICT features for each group. For example, a user who intended to self-regulate their exercise behavior in anticipation of a reward was automatically assigned features that enabled them to experience that reward. Through repeated use, iMovin' aimed to encourage each user toward a more autonomous self-regulation of exercise behavior. The app did not track or collect participants' physical location data. A non-personalized version of the app, in which all features were available to every user, was also used in the study.

The paired wireless activity trackers provided behavioral data on exercise behavior, used both for the app's physical activity history dashboard and as a study outcome measure. Each tracker contained a 3-axis accelerometer that tracked movement in every direction; raw accelerometer data were processed by the tracker into steps and activity data via a series of algorithms.

The tracker was worn by the user, typically on the wrist, throughout the study, and was required to be worn daily except when charging. Tracker data were automatically synchronized with iMovin' each time the user launched the app. Because the tracker was in contact with skin, prolonged wear could lead to slight skin discomfort or irritation for some individuals. To reduce this risk, users were briefed to keep their trackers clean and dry and not to wear them too tightly. Participants experiencing prolonged discomfort could stop wearing the tracker and notify the research team; in this case, they returned the tracker and were withdrawn from the study.

Experimental Design

A double-blind randomized field experiment study was conducted to develop and evaluate salient nudging, an adaptive intervention that adjusts the visual prominence of mHealth app features according to each user's motivational profile. It was conducted at NUS and spanning 112 days, divided into four consecutive 28-day periods (T1: Days 1-28; T2: Days 29-56; T3: Days 57-84; T4: Days 85-112).

Participants were randomly assigned in a 1:1 ratio to either the salient-nudging condition or the control condition. The allocation sequence was generated by computer, stratified by gender. Allocation was concealed using sequentially numbered, opaque, sealed envelopes prepared by a research assistant not involved in participant recruitment. Both participants and researchers involved in data collection were blinded to treatment assignment throughout the study.

Measurements were taken at baseline and at each subsequent 28-day interval. Survey questionnaires collected data on participants' motivations and motives to exercise at each interval, classifying participants into one of six motivational profiles along the self-determination continuum: amotivation, external regulation, introjected regulation, identified regulation, integrated regulation, and intrinsic motivation. Objective physical activity data were continuously recorded via wearable activity trackers throughout the study.

Research Procedure

The study was conducted as a randomized controlled trial (September-October 2019 to April-May 2020).

The procedure began once a participant signed and submitted a Consent Form (CF) at the first of two sessions. Each participant was issued a wireless activity tracker on loan, and, with assistance from research staff, registered a new app account by providing an email address (as username), a chosen password, and the loaned tracker's identifier via a web-based administrator platform. The platform stored account credentials in a local database and preloaded default values, to be updated when the participant next used the app.

The session concluded with instructions that participants would use the app and tracker for 16 weeks; that they were required to wear the tracker daily and to keep it charged, except when showering, sleeping, or charging it, starting from a baseline date to be communicated to them; that the tracker, if loaned, was the property of the Department of Information Systems and Analytics and was to be cared for accordingly; that they would receive a follow-up email with the baseline date; and that they should download and install the app via the Apple App Store or Google Play.

After downloading the app, participants logged in with their created account and completed the first of four "profiling" survey questionnaires, covering current exercise behavior, motivations, and motives; the first questionnaire also collected demographic information. A new questionnaire was administered every 28 days via the app, at the start of each period (T1-T4). Participants could save progress within a questionnaire. If a questionnaire was not completed within three days of being sent, a first in-app reminder was sent the following day, and a second warning notification one day after that.

On submission of the first profiling questionnaire, each participant's profile was updated and they were automatically assigned according to the experimental design. Participants in the salient-nudging condition saw their app's landing-page feature prominence configured to match their motivational profile, with targeted push notifications reinforcing the salient features; participants in the control condition saw the same set of features displayed in a randomized, non-tailored layout. All participants retained access to the same full set of app features throughout the study. A follow-up email was sent shortly after the first session indicating the baseline date and reminding participants to install the app and set up the tracker beforehand.

From the baseline date, participants were responsible for keeping the tracker charged and taking due care of any loaned tracker until it was returned at the end of the study. At the start of each subsequent 28-day period (three further intervals), participants were prompted on app login to complete the next profiling questionnaire, after which the salient-nudging condition's landing-page configuration was updated to reflect their current motivational profile.

Tracker data were monitored daily. Damaged or misplaced trackers were handled, with participants issued a replacement; repeat incidents resulted in withdrawal from the study. Participants withdrawn from the study were contacted by email to arrange retrieval of their loaned tracker.

At four 28-day periods after the baseline date, Participants were asked to uninstall iMovin' and emailed a link to an online webpage to select a preferred timeslot for a second, walk-in session. Uninstallation needs to be done before this defriefing session because not all participants ended at the same time. At this session, participants were screened for completion of the final profiling questionnaire and correct attendance; returned their loaned tracker, if applicable; were debriefed on the study's objective, data collected, and use of that data, and future directions for the study; had the opportunity to ask questions. They were finally thanked for their participation.

Participants attended two sessions with the research team, held at multiple locations around NUS, including the School of Computing, Faculty of Science, Faculty of Engineering, and student hostels such as Prince George's Park Residence. The first session lasted about 10-20 minutes; the second, about 15-20 minutes.