JMIR Publications

ABSTRACT

Background:

Low levels of physical activity (PA) can be attributed to present lifestyles and mHealth approaches are gaining popularity to address this issue.

Objective:

This pilot study aimed to: (1) develop a multi-component mHealth technology suite (i.e. PA wearables and an interactive smartphone application (App) supported by a web-based data management system), (2) determine the validity of the PA wearables in measuring steps and floor counts, and (3) pilot-test a real-time team-challenge intervention delivered through the mHealth suite.

Methods:

Staff and students from National University of Singapore were recruited between 2015 and 2016. This pilot study involved a validation (Phase 1) and an evaluation phase (Phase 2). In Phase 1, every participant was requested to climb 4 bouts of floors in an indoor stairwell, and simultaneously wear a Fitbit® tracker and an ActiGraph™ for 7 days under free-living condition. In Phase 2, participants received trackers and were allocated to control or intervention group. The intervention participants were randomized to 4 teams and were registered to use the mHealth suite. Intervention was 6 weekly team-challenges (Monday - Friday) integrated with behavioural change techniques delivered in real-time through the technology suite. In Phase 1, Fitbit-measured steps and floors were compared against ActiGraph and direct observation, respectively. In Phase 2, the difference between groups in change in average daily steps or floors from baseline to end of intervention was analysed using multiple linear regression analysis.

Results:

Of 40 participants eligible for Phase 1, 32/40 and 40/40 participants provided valid steps and floors data, respectively. The Fitbit tracker over-estimated step-count in free-living condition (median absolute error of 17%) but accurately estimated floors-count in stairs-climbing. High positive correlation and good agreement were observed in the validation of steps/day (Fitbit vs ActiGraph: Spearman Rho=0.89, P < .001) while very high positive correlation and excellent agreement were observed in the validation of floors-count (Fitbit vs observation: Spearman Rho=0.98, P < .001). In Phase 2, 40 participants were allocated to intervention (n=20) and control (n=20). 24/40 completers (i.e. provided complete covariates and valid PA data) were included in the analyses. At the end of intervention, the intervention group achieved higher PA levels compared to the controls: (1) average steps/day=2,990, 95%CI=(89, 5892) and (2) 90% increase in geometric mean of floors/day (95%CI=(1.2, 3.2)), after adjusting for age, gender and respective baseline PA values.

Conclusions:

This pilot study suggested that the consumer wearables provide an acceptable validity in estimating step count and stair climbing, despite some overestimation of absolute step counts. The mHealth suite developed for real-time interventions was feasible to increase step counts and especially stair use in the target population. More generalizable studies with long-term follow-up are warranted to strengthen the evidence for the presented mHealth strategy.