Overall Effects of Consumer-Wearable Activity Tracker-Based Programs
The present systematic review and meta-analysis synthesizes the evidence to date about the effectiveness of consumer-wearable activity tracker-based programs on daily objectively measured PA and SB among apparently healthy school-aged children [20, 21, 43,44,45, 52,53,54,55, 60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94]. The overall results showed that the consumer-wearable activity tracker-based programs brought about significant moderate improvements in school-aged children’s daily total steps, and small but significant improvements in daily MVPA levels after the intervention program. However, regarding daily levels of total PA the effect was trivial. Therefore, the use of a consumer-wearable activity tracker as a motivation tool for young people is strongly recommended to reduce the high levels of physical inactivity among school-aged children [8]. However, the intervention programs seem not to reduce the school-aged children’s SB.
These results agree with similar previous meta-analyses carried out in adults [24,25,26]. Firstly, regarding daily total steps, all previous reviews found improvements, although they seem greater in the present systematic review (d = 0.612 vs. 0.240–0.449; D = 1,692.79 vs. 950.54). Regarding changes in actual units, adding 1,000 steps per day, which is equivalent to a 10% increase within the recommended 10,000 steps per day [19, 95], has been shown to be a significant and clinically meaningful change related to a substantial reduction of the risk of all-cause mortality in the adult population [96, 97]. Therefore, although there is no such evidence in studies carried out with school-aged children, the increase of 1,692.79 daily steps obtained in the present meta-analysis might be considered an important and meaningful change considering the existing evidence with adults. Moreover, with reference to MVPA, similar improvements in daily levels of MVPA were obtained (d = 0.220 vs. 0.270; D = 5.583 vs. 6.160). With reference to changes in actual units, to our knowledge, there is no clinical evidence of a link between changes in minutes involved in MVPA and health outcomes in any population. However, extrapolating the results obtained as a percentage of the international recommendations (i.e., 60 min of MVPA per day [4]), the obtained results (i.e., 5.6 min) represent almost a 10% increase. Therefore, considering the same reasoning explained above with daily total steps, these changes in daily levels of MVPA may also be considered clinically significant.
The greater improvement in school-aged children’s daily total steps than in their MVPA levels may be due to the kind of goal established in the program. Most of the studies included in the systematic review with a goal-setting strategy set only a step-based goal (30 of 37 studies), while only five studies established both step-based and minutes of total PA-based goals [62, 87] or SB-based goals [88, 91]. Therefore, all the motivational strategies included, such as reminders [44, 54, 66], counseling sessions [70], or rewards [65, 73] were carried out around this goal of increasing the number of steps. Moreover, the reason for relying mainly on the number of steps as the reference output for goal-setting may be due to steps having the advantage of being easier to understand and interpret by school-aged children compared to MVPA minutes [19]. In addition, in many studies, the consumer-wearable activity-tracker used were pedometers, which only show feedback about the number of steps. Therefore, similar to the evidence found with adults [24,25,26], consumer-wearable activity tracker-based programs seem to be effective for improving objectively measured daily PA, especially for daily total steps, although the effects seem to be greater for school-aged children. This greater effect may be due to the fact that school-aged children have a higher affinity with new technologies playing an important role in their daily life, and therefore these kinds of technology-based interventions may be more interesting for them [98, 99]. In addition, during the period of childhood and early adolescence, school-aged children are still forming their daily habits and they are more sensitive to changing their PA behavior which could explain these greater intervention effects, while in adulthood the stability of PA patterns is high and more difficult to be changed [100]
Regarding total PA, results showed a trivial favorable effect (d = 0.151, 95% CI 0.038–0.264). These results are curious because daily total steps is an indicator of total PA [101], and therefore, results were expected to be similar. However, it may be because very slow steps were included in daily total steps, but they do not reach the threshold to be considered light PA, or because fewer studies are assessing total PA compared to studies assessing daily total steps. Nevertheless, in the present meta-analysis, only the study by Corepal et al. [65] measured both variables in the same study, without any clear relationship between both outputs. Therefore, future studies including the evaluation of both variables (i.e., daily total steps and total PA) would be very interesting to establish a real cause–effect relationship between them.
Besides, regarding SB, similar to the trivial unfavorable effect obtained in the present systematic review (d = 0.172; 95% CI 0.039–0.305), no real differences were found in any previous review [24, 26]. Firstly, this may be due to the fact that most of the programs that evaluate SB only used strategies to encourage and support PA behavior change (i.e., goals, tips and challenges, behavioral incentives or reinforcement messages based only on PA practice) and were not specifically designed to reduce SB [21, 43, 62, 65, 77, 85]. Only Leinonen et al. [53] specifically included some SB-based strategies, such as feedback showing a thumb either up or down if the day included over two hours of sedentary (sitting) periods or not, and rewards regarding decrement in weekly school-aged children’s sedentary time. The lack of specific focus on reducing SB may have contributed to this trivial unfavorable effect. Moreover, these results may also present significant measurement bias due to all SB-studies (except Morris et al. [85]) analyzed raw time involved in SB per day instead of valid wear time-based standardized scores (e.g., percentage of time of each day engaged in SB of the total valid wear time; standardized mean SB in minutes) [102, 103]. In addition, it must be considered that for accelerometer-based measures only a valid minimum time per day is established (normally 600 min), but that standardized values are not taken into account [104]. This is even more accentuated for consumer-based wearables for which valid wear time cannot be known and, therefore, it cannot be controlled if school-aged children wore the wearable for a long enough time, or if they were more motivated to wear it for more time in the baseline or post-intervention measure. In this sense, it is easier for school-aged children with more valid time to have higher registered time involved in SB, so SB time outcomes could be especially affected by potential systematic valid wear-time variation between measurement moments (i.e., pre–post-intervention measures) or groups [102]. This could also affect daily total steps and daily MVPA levels although to a much lesser extent than for SB [102]. For instance, Gaudet et al. [43] showed that school-aged children in the experimental group had approximately 48 min more wear time per day in comparison with the control group, which may directly affect their differences regarding time involved in SB. However, most studies [21, 62, 65, 77] only reported the minimum time per day needed to be included in the study but did not report the actual mean valid wear time per day, and it is important that this be reported in order to compare compliance between groups. Finally, it should be noted that there are very few studies measuring school-aged children’s SB in comparison with studies measuring steps and MVPA, which implies a wider confidence interval and consequently greater uncertainty about the real value.
Influence of Participants’ Characteristics
According to the results of the present meta-analysis, a significant relationship has been observed for PA status from the within-study subgroups analysis, the intervention being much more effective for improving daily total steps in school-aged children who were physically inactive in the baseline measure than in those who were physically active. Furthermore, this positive influence was also found for improving objectively measured daily MVPA levels, although only from the between-study analysis (i.e., an observational relationship). A potential reason for this influence could be that school-aged children who already are physically active before the intervention are motivated enough for PA practice without the need for extra motivation with the proposed intervention [105]. Furthermore, it is difficult to further increase school-aged children’s PA levels when the baseline levels are high, just as has occurred with the improvement of physical fitness levels [106, 107]. Therefore, consumer-wearable activity tracker-based programs may be an especially appropriate strategy for less active school-aged children [82, 85]. In line with the recommendation by Love et al.'s [108] systematic review, future studies should analyze their results distinguishing by school-aged children’s baseline PA profiles to correctly identify the intervention impact because all participants do not react in the same way to the intervention. Regarding school-aged children who already are physically active before the intervention, it may be necessary to study which specific strategies should be implemented in these interventions to help them maintain their PA levels, or even to continue increasing them, therefore obtaining greater health-related benefits [4]. Finally, regarding school-aged children’s age, it does not seem to affect the intervention program effect.
Furthermore, programs seem to be more effective in females than in males for improving daily total steps, which could be related to PA status since females tend to be more physically inactive than males throughout childhood and adolescence [109, 110]. Therefore, sex-specific interventions could be considered in future research like Böhm et al. [29] suggested, although the conclusions in relation to the effectiveness of sex-specific interventions should be taken with caution because the success of those programs was not obtained by the within-study subgroups analysis (i.e., cause–effect relationships), but rather by the between-study analysis, which only establishes observational relationships [20, 44, 69, 73, 74, 83]. Furthermore, the between-study subgroups analysis results still showed moderate-to-high heterogeneity in each subgroup (i.e., males and females), so it is likely that there were differences in other strategies of the intervention or participants’ characteristics that could influence its effect, even more than sex itself. Additionally, these differences are even more accentuated because the analysis includes twice as many publications carried out with females than with males. However, the number of studies including this within-study sex comparison and which allow establishing cause–effect relationships is very limited (k = 7), and for this reason, the observational relationships of the between-study comparisons have been considered in the present meta-analysis.
Influence of Intervention Programs’ Characteristics
Firstly, overall meta-regression analysis results showed a higher effect for daily total steps when a greater number of strategies were included in the programs. Therefore, multi-dimensional interventions that include most of these strategies seem to be preferable for mediating PA behavior. That agrees with some psychological theories which include most of those strategies as positive mediators influencing PA behavior. For instance, the Social Cognitive Theory [111], highlighted PA-related knowledge included in counseling sessions, positive reinforcement such as reminders with encouraging messages about PA practice (e.g., “You’re in charge! Make the choice to meet your step goal today!”), and the importance of setting achievable goals as determining factors in the design of PA promotion interventions that could lead to behavior change in school-aged children. In addition, the Self-Determination Theory [112], emphasizes the need for relatedness included in most motivational strategies (e.g., teamwork or the use of social networks), and the perceived competence reflected in the reinforcement reminders praising their efforts (e.g., “You can meet your step goal; just keep stepping!”), or the evolution that school-aged children could observe in the diary they filled in during the program, as necessary to increase their intrinsic motivation. Furthermore, recent systematic reviews highlight that using multi-strategy approaches as behavior change techniques show better PA outcomes than singular change approaches [113, 114]. For instance, the PA-related knowledge provided in the counseling sessions, the autonomy support environment by the consumer-wearable activity tracker feedback, the inclusion of additional motivational strategies like social networks, setting goals of moderate difficulty, or sending reminders with encouraging messages about PA practice have been shown to positively change school-aged children’s PA behavior [115, 116]. However, it should be noted that the explained variance was low (R2 = 0.02), as well as the results being highly heterogeneous (I2 = 89.02).
Furthermore, there was considerable heterogeneity in the strategies included in the reviewed studies. Most of the programs used a goal-setting strategy, participants’ logbooks, educational counseling sessions, and/or some kind of motivational strategy. Nevertheless, the inclusion of reminders to persuade participants to move or exercise more was a less frequently included strategy. The influence of these intervention program characteristics in school-aged children’s daily total steps and MVPA has been analyzed in the present meta-analysis. Firstly, interventions including some kind of counseling [54, 60, 71] and/or goal-setting techniques [73, 86, 90], in addition to consumer-wearable activity trackers, were highlighted as more effective than those without them for improving school-aged children’s daily total steps. These results are in accordance with previous studies’ recommendations about the inclusion of these explicit strategies (e.g., advice about PA benefits, strategies to reduce SB and increase PA, resolution of barriers to PA practice, or goal-setting strategies based on the international guidelines) which make students feel that they are making an informed decision about their health in any kind of program for PA promotion [11, 13] and specifically in wearable-based programs [27, 115]. Nevertheless, apparently contradictory results showed that consumer-wearable activity trackers-based programs which did not include any exercise routine seem to be more effective than those that included it for improving school-aged children’s daily MVPA levels. However, analyzing the kind of exercise routine included, had some limitations. Firstly, Jago et al. [62] and Smith et al. [91] included a low frequency of supervised PA sessions (one 20-min PA session per week, and only six 20-min lessons in 20 weeks, respectively) with which it is very difficult to positively affect the school-aged children’s daily PA levels. Secondly, most of the activities included by Jago et al. [62] did not have a direct relationship with increasing the school-aged children’s number of steps or minutes involved in MVPA (i.e., stretching, technical drills, or strengthening tasks), and also it should be noted that despite including this exercise routine, they did not include other strategies that may be even more important than this one (e.g., reminders, diary, or motivational strategies). Finally, it is also important to denote that the analysis included 17 units of analysis without an exercise routine vs. three with an exercise routine [62, 91], so given this marked difference in the sample, results should be interpreted with caution. Moreover, it must be considered that only observational relationships have been obtained from the between-study subgroups analyses due to the low number of studies compared in the within-study subgroups analyses. Furthermore, the results regarding school-aged children’s daily total steps separated by each subgroup still showed a high level of heterogeneity which implies differences in other intervention characteristics. Therefore, future studies should include different intervention groups that compared some intervention characteristics (e.g., one experimental group including counseling and another without counseling) to establish causal-effect relationships between intervention characteristics and the effect of the programs.
Regarding the kind of consumer-wearable activity tracker, waist-worn trackers such as pedometers are more common than wrist-worn trackers. This could mainly be due to the fact that waist-worn trackers such as traditional step counters with digital displays, have featured in scientific research since approximately 1996, being an accepted method for assessing PA and a tool for walking interventions [117]. On the contrary, wrist-worn trackers have burst onto the market in the last decade and, therefore, their scientific evidence base is still scarce [14, 18]. Moreover, the meta-analysis results showed that programs carried out with a wrist-worn activity tracker seem to be more effective than those carried out with waist-worn trackers for improving school-aged children’s daily MVPA levels. This may be due to wrist-worn trackers having several advantages compared to waist-worn trackers, such as reporting real-time feedback that can be easily checked on their wrist or touch screens [14]. Moreover, unlike more traditional waist-based trackers, which only monitor and display simple feedback about PA levels, wrist-based trackers are much more interactive since the user is able to set reminders, notifications, or congratulatory messages upon reaching the proposed goal [118]. Finally, wrist devices have shown greater wear time compliance which could mean that if they wear it for a longer time school-aged children could interact more with its features [119].
Risk of Bias and Certainty of the Evidence
Firstly, based on the methodological risk of bias assessment, most studies were classified as “high risk” or “some concerns,” leaving only two studies classified as overall “low risk” [21, 81]. Therefore, this may have resulted in a biased assessment of the intervention effect, underestimating or overestimating the true intervention effect, which meant downgrading the GRADE certainty rating by one level for all outcomes (i.e., daily total steps, MVPA, total PA, and SB) regarding the risk of bias domain [34, 46]. In reference to the study designs, it is interesting to highlight that only 51.11% of the included studies are true or cluster-randomized controlled trials, which are markedly far stronger interventions to demonstrate effect significance [120]. However, sensitivity analysis showed no differences for school-aged children’s daily MVPA, total PA, and SB levels between study designs, but much greater effects were found in randomized controlled trials for school-aged children’s daily total steps than non-randomized trials.
Moreover, regarding daily total steps outcomes, a substantial level of heterogeneity was found, even in the follow-up subgroups analyses (except when separating by accomplishment of PA recommendations) and it meant downgrading the GRADE certainty rating by another level regarding inconsistency domain for daily total steps. This is most likely because it is the PA outcome that includes the largest number of studies and, consequently, the greatest variety in the types of intervention applied when compared with other outcome measures. Finally, regarding publication bias, although the funnel plots and Egger’s test suggested publication bias for daily total steps and daily MVPA levels, its impact seems to be very low given the unlikely number of “lost” studies suggested by the fail-safe N analyses. Furthermore, the Trim and Fill method did not trim any study for daily steps and only two studies for MVPA, resulting in an adjusted value similar to the observed values (d = 0.220 vs. 0.213).
For all the above-mentioned reasons, it is important to highlight the “low” certainty of evidence found for daily total steps, which means that the confidence in the effect estimate is limited and the true effect may be different from the estimated effect [121]. Regarding MVPA, total PA and SB outcomes, “moderate” certainty of evidence was found, so the true effect is likely close to the estimated effect, but there is a possibility that it is substantially different [121]. Therefore, the findings of the present meta-analysis should be considered with caution and firmer conclusions should await the accumulation of a larger high-quality number of primary studies.
Strengths and Limitations
Regarding the strengths of the present systematic review, numerous measures to avoid, or at least to reduce, publication bias were followed (e.g., the inclusion of a great range of bibliographic databases from different disciplines and complementary search strategies, or not restricting the search by the language, type or date of publication). Then, several exploratory analyses were conducted to identify and assess the impact of any potential publication bias (e.g., funnel plots, or Orwin’s fail-safe N analyses), as well as sensitivity analyses (e.g., Hedges’ g with a random-effects model or Cohen’s d with a random-effects model separately for randomized controlled trial design or not) to verify the robustness of the results. Furthermore, the present review was focused only on objective measurements, which have shown high validity to measure PA and SB levels in comparisons with self-reported measures [35, 36]. Lastly, to our knowledge, to date this is the first systematic review and meta-analysis about the effects of consumer-based activity tracker-based programs on objectively measured PA and SB levels within apparently healthy school-aged children, including analyzing the influence of the intervention programs’ characteristics and school-aged children’s characteristics on the effects. This meta-analysis summarizes the effectiveness of those interventions in an overall statistical synthesis, improving the precision of the results by the estimation of the effect size and direction, and clarifying whether or not the effect size is consistent across studies.
However, the present systematic review and meta-analysis is not without limitations. First, although randomized controlled trials have higher methodological quality, the present systematic review includes several study designs. As expected, there were not many consumer-based activity tracker-based studies with a high level of quality design for improving the different PA-related behaviors. Therefore, a reason for including several designs is to provide evidence of the effects of interventions for which only a small number of randomized controlled trials are available, drawing on the “best available evidence” rather than the “highest tier” of evidence [34]. Nevertheless, sensitivity analyses were also performed comparing randomized controlled trials and non-randomized trials, showing no differences for school-aged children’s daily MVPA, total PA, and SB levels. Furthermore, even greater effects in school-aged children’s daily total steps were found in randomized controlled trials. Second, although the inclusion of a wide range of intervention types, populations, sample size, and study designs had some advantages regarding the generalizability of conclusions, it also means a high level of heterogeneity. For instance within intervention types, multiple behavior change strategies such as goal-setting (even including different kinds like static or adaptive goals) or extra motivational strategies (e.g., social networks or social support) were usually combined in the same study. Therefore, it makes the independent contribution of any intervention features and, therefore, making strong conclusions from the intervention difficult to establish. However, in addition to the overall effect size, subgroups analyses and meta-regression of the a priori hypothesized moderators were also performed. Therefore, not only general effect results are provided, but also results for each specific group based on the characteristics of the programs and school-aged children. Third, the present systematic review investigated effectiveness at the end of the program (i.e., short-term), but future studies should investigate long-term effectiveness to assess actual behavioral changes some months after the program. However, due to the very limited evidence, this was not performed in the present systematic review. Finally, coding some study outcomes was problematic due to authors not reporting them. Although authors were contacted, many of them did not reply and the particular study outcome had to be omitted. However, this is a common problem in most systematic reviews [34], and a great effort was made in contacting authors, recalculating data, or estimating values from figures. Finally, in some cases, consumer-based activity trackers were used both as a motivational instrument during the intervention and to objectively measure PA, which could affect results by increasing their actual PA levels in the control group or during baseline assessments.