All students in the two Grade 8 (13-14 years old) classes in one school were invited to participate in this study. To participate, students had to obtain written informed consent from parents or legal guardians and provide assent as approved by the Centre Hospitalier de l’Université de Sherbrooke ethics committee. This study employed a quasi-experimental crossover design with a 7-week physical activity tracker-based physical activity intervention and a control period. Participants in one class (group A) were randomly assigned to the immediate intervention group and participants in the other class (group B) were assigned to the delayed intervention group (Figure 1). The crossover design was used to allow participants from both groups to have a chance to experiment with the physical activity tracker. This design also allowed controlling for individual-level covariates as participants served as their own control. Baseline measurements were obtained before the first week of intervention. Physical activity measurements for all participants were also obtained at the end of both intervention periods: weeks 7 and 14. This study took place between February and June, 2016.

The intervention focused on increasing physical activity via self-monitoring and self-regulation. During the intervention period, each participant was provided a wrist-worn physical activity tracker (Fitbit, model Charge HR; FitBit Inc. San Francisco, USA) that was equipped with a small screen displaying real-time summary data for steps, heart rate, distance, calories, and stairs climbed. Physical activity intensity minutes and temporal patterns were also available to participants through the accompanying Web-based Fitbit user account. Individual Fitbit user accounts were prepared by the study coordinator for each participant before the distribution of the devices. The study coordinator was also responsible for demonstrating how to use the Fitbit device and Web-based Fitbit user account. The intervention was similar to that used in previous studies [10,16] and was based on Behavior Change Technique research [24]. Empirical research shows that self-monitoring, in combination with at least one other self-regulation skill, can lead to positive physical activity related behavior change [24]. An individualized goal that can be tracked using measures provided by the physical activity tracker was set by each participant, following a brief researcher-led presentation of the SMART (specific, measurable, attainable, realistic, and timely) principles of goal-setting [25]. Beyond receiving physical activity trackers during the intervention period and being asked to select a goal, participants did not receive other interventions. During the control period, participants received no intervention.

Actical accelerometers (Philips - Respironics, Oregon, USA) were worn on the hip for seven days at baseline, week 7, and week 14. This lightweight omnidirectional accelerometer has been validated as an objective measure of physical activity in youth aged 10 to 15 years [26]. Accelerometer data were recorded in 15-second intervals, then cleaned and managed using procedures recommended by Statistics Canada [27] through a series of publicly available SAS codes adapted for this type of study [28]. Time spent in different physical activity intensities were determined by using cut-points established in previous research involving Actical accelerometers in children [29]. Sedentary activity corresponded to count values below 100, light physical activity to counts between 100 and 1500, moderate physical activity to counts between 1500 and 6500, and vigorous physical activity to counts greater than 6500. Only data for valid days, defined as 10 hours or more of wear time, were retained for analyses. Daily averages for MVPA were calculated from valid days.

Using the Fitabase analytics system (Small Steps Labs, San Diego, CA, USA), data from all physical activity trackers were remotely collected and aggregated whenever data were transmitted to users’ personal Fitbit dashboards. Data captured included heart rate, daily steps, and minutes of intensity-specific physical activity. Wear time was calculated by subtracting non-wear time from 24 hours and non-wear time was defined as any interval with at least 60 consecutive seconds of zero recording of heart rate. As heart rate was recorded at variable time periods by the physical activity tracker, allowing for 60 consecutive seconds of zero recording of heart rate was sufficient to distinguish non-wear from wear time. Indeed, identification of valid days remained stable across use of higher thresholds, for example 5, 15, 30, and 60 minutes, whereas significant information was lost with a threshold set under 60 seconds.

A baseline questionnaire was used to collect information regarding age, sex, stage of change for physical activity behavior, and goal commitment. Specifically, participants indicated whether they participated in at least 60 minutes of physical activity per day, using an item corresponding to the five stages of behavior change (precontemplation, “No, I do not participate in physical activity and I do not intend to in the next 6 months;” contemplation, “No, I do not participate in physical activity regularly but I intend to in the next 6 months;” preparation, “No, I do not participate in physical activity regularly but I intend to in the next 30 days;” action, “Yes, I have been participating in physical activity regularly, but for less than 6 months;” maintenance, “Yes, I have been participating in physical activity regularly for more than 6 months”) [30,31]. For analyses, these five stages of behavior change were grouped into two categories representing adoption (ie, action and maintenance) and preadoption (ie, precontemplation, contemplation, and preparation) as done by De Bourdeaudhuij et al [32]. Goal commitment, defined as determination to attain an objective, was assessed using a five-item scale refined and validated by Klein et al [33]. In this scale, participants indicated their level of commitment to the personal goal they had set with the following items using a 5-point Likert scale: (1) It’s hard to take this goal seriously; (2) Quite frankly, I don’t care if I achieve this goal or not; (3) I am strongly committed to pursuing this goal; (4) It wouldn’t take much to make me abandon this goal; (5) I think this is a good goal to shoot for. Items 1, 2, and 4 were reverse-scored before calculating a mean of the five items meant to represent the construct of goal commitment [33].

Based on previous research, which showed that a similar Fitbit-based physical activity tracker intervention induced a 36% increase in MVPA (pre-post change mean=172, SD=83 to mean=234, SD=119), albeit in a sample of adult women [16], we estimated that 22 participants per group would provide 80% power with 95% CIs.

Wilcoxon rank-sum tests were used to compare between group differences at baseline and following both intervention periods. Wilcoxon signed-rank tests were used to compare within group difference at different time points. A multiple linear regression model was used to assess pre-post change in physical activity while controlling for the effects of time, goal commitment, and stage of behavior change. Physical activity data, from both physical activity trackers and accelerometers, were adjusted for valid wear days. A valid wear day was defined as at least 10 hours of wear time. Repeated measures analysis and Tukey post-hoc tests were used to assess changes in physical activity tracker-measured physical activity throughout the intervention period. All analyses were conducted using SAS 9.4 (SAS Institute, Inc., Cary, North Carolina, USA).

With little research to date on the effects of using direct-to-consumer mHealth trackers as behavior change tools, the current study sought to examine the effectiveness of using a minimalist physical activity tracker-based intervention as a means of encouraging increased physical activity in adolescents. Although the main results suggested no change in MVPA as a result of having been exposed to a physical activity tracker, secondary analyses suggest that the interventions may have had beneficial effects for some sub-groups. For example, an intervention response occurred in a subset of participants who reported being in the action and maintenance stages of behavior change (adopters) in relation to participation in physical activity. Whereas these participants increased their daily average of MVPA by more than 15 minutes during the intervention period, those in the precontemplation, contemplation, and preparation stages (preadopters) had no change in their level of MVPA. This distinction manifested itself without the intervention being intentionally tailored to any specific stage of behavior change. Nevertheless, previous studies show that psychosocial determinants positively associated with physical activity generally increase across the stages of behavior change [32,34]. Adolescents in the adoption stages typically perceive fewer barriers, more benefits and have a better attitude and more self-efficacy in relation to physical activity participation [32]. Such underlying conditions likely predisposed these participants to be more receptive and to respond favorably to the exposure to a physical activity tracker. Although minimalist in nature, the introduction of a physical activity tracker may therefore represent a sufficient trigger for youth in the adoption stages to increase their level of physical activity. However, the intervention was likely too simple to induce a behavior change among youth in the preadoption stages. Stage-specific intervention research suggests that in order to successfully motivate individuals in the preadoption stages, it is necessary to consider cognitive aspects of behavior change such as raising consciousness, social liberation, self-re-evaluation, self-liberation and counter-conditioning, helping relationships, and reward management [35,36].

The increase in physical activity among adopters may also be attributable to the fact that there was room for growth. Despite perceiving themselves as being active, objective measures suggested that participants in the adoption stages were not more active than those in the preadoption stages at baseline. This is similar to results from another study, which found little to no difference among levels of objectively measured MVPA of adolescents at different stages of behavior change [30]. Our results therefore point to the potential usefulness of assessing readiness to change before intervening. Although preadoption and adoption adolescents presenting similar objectively measured physical activity at baseline, the simple one-item questionnaire used to assess stages of change behavior change in this study correctly pre-identified participants who would best respond to the introduction of a minimalist physical activity tracker-based intervention.

This study was initiated during the middle of winter and extended to the end of spring. Thus, the start and end points of the study coincide with the typical periods of lowest and highest annual levels of physical activity in this age group, respectively [37-39]. This is noteworthy as the physical activity tracker-based intervention elicited an increase in physical activity during the study phase that corresponded to winter. Although it needs to be corroborated by other studies, it is possible that the introduction of physical activity trackers during this season could help some adolescents increase physical activity during the colder winter months. This is in support of findings from Dean et al [19] who observed, among a sample of adolescents with congenital heart disease, that wearing a physical activity tracker, as compared with not wearing one, was associated with a less abrupt decline of physical activity during winter months.

Continuous objective measurements obtained from physical activity trackers provided information suggesting that there was an acute effect of receiving the physical activity tracker. Specifically, compliance to wearing the physical activity tracker was at its highest during the first weeks of intervention. This would suggest that the device had a novelty effect, as demonstrated by others [23,40,41]. For instance, Shih et al [40] measured 50% attrition rates after the 2 week mark in a 6-week study in undergraduate students, while Schaefer et al [23], had only 2 participants (8%) use their physical activity trackers for a 4 month follow-up study. Beyond a decrease in compliance, it was noted that the average number of daily steps and minutes of physical activity were also at their peaks early in the intervention period. During the first three weeks, participants averaged between 9800 and 12,000 steps per day, which is close to the 10,000 to 12,000 steps per day recommended for this age group [42-44]. After the third week, however, this number declined to less than 9000 steps per day. Whereas normative data indicate that the majority of adolescents do not meet the step count recommendation [44], our findings suggest that there may be potential for physical activity trackers to encourage adolescents to perform near recommended levels of physical activity, at least over a short period of time.

It is possible that accompanying the distribution of physical activity trackers with a more intensive intervention would have led to greater compliance in wearing the device and greater increases in physical activity. The participants in our study, nevertheless, wore physical activity trackers to a greater extent than adolescents from under-resourced communities in another study [23], but also considerably less than post-menopausal women in another study [16]. Direct comparison to wear time during intervention in other studies involving adolescents is not possible as this kind of information tends not to be reported [41,45].

Limitations of this study need to be considered when interpreting the results. First, even though none of the participants was lost during the study, there was an unanticipated decline in compliance in wearing accelerometers at both post-intervention periods. This contributed to a loss of power to detect meaningful differences in physical activity, despite having initially recruited enough participants for adequate power. This drop in compliance may also have contributed to selection bias, wherein participants least likely to become more active did not wear the device at follow-up periods. Research is warranted to better understand adolescent engagement toward physical activity trackers to develop tailored interventions aimed at increasing compliance in this population [41,46]. Comparison between participants who completed all three evaluation periods and those who did not revealed no significant differences in physical activity level at baseline or distribution in stages of behavior change. Second, it needs to be considered that some activities were not measured because the accelerometer or physical activity tracker could not be worn (eg, swimming). Third, caution must be taken in interpreting physical activity tracker measured physical activity data as the proprietary algorithms used to calculate minutes of physical activity at different intensities are not publicly available. Fourth, although our results revealed a difference in intervention response between adolescents in the preadoption and adoptions stages, the small sample did not allow for in-depth analyses between each stage. Given the theoretical and empirical evidence of psychosocial and processes of change differences between each of the stages, future research with a larger sample is warranted to help elucidate which stages benefit the most from mHealth devices such as physical activity trackers. Future research should also assess whether similar physical activity tracker-based interventions lead to progressions in the stages of behavior change even among individuals who do not change their level of MVPA. Finally, although this study benefited from the strengths of randomization and crossover, having more randomization units in future studies would help attain group similarities even among unmeasured potentially confounding variables.

In summary, although no overall effect was found, the secondary results of this study suggest that there is potential value in using physical activity trackers to increase physical activity among adolescents in adoption stages of behavior change related to participation in physical activity. Adolescents in the adoption stages of behavior change may benefit from simply gaining access to a direct-to-consumer mHealth device designed to track physical activity. Future research is needed to better understand what additional strategies could be paired with physical activity trackers to lead to improvements in physical activity levels of adolescents in all stages of behavior change. This study also has implications for research as it demonstrates the feasibility of continuously and objectively measuring physical activity during an intervention involving adolescents.