Effectiveness and moderators of a multicomponent school-based intervention on screen time devices: the Movimente cluster-randomized controlled trial

Sedentary behavior is a highly predominant risk behavior adolescents perform throughout the day [1, 2]. Among the different types of sedentary behavior, screen time (ST) is widely investigated and is defined as the time spent on screen-based behaviors performed predominantly in a stationary position, such as watching television (TV) and using computer, video game or smartphones [3]. Evidence shows that exposure of adolescents to excessive ST is associated with several health risks such as increased cardiovascular risk indicators [4], anxiety [5], depression [5] and low social connectedness [6]. International guidelines recommend that children and adolescents should spend less than 2 h per day in recreational ST [7], however, most do not meet this recommendation worldwide [1]. Thus interventions have been encouraged and developed focusing on the reduction of ST in this population group [8‐11], with strategies mainly delivered through school environments [12]. The schools are considered promising environments for changing adolescent behavior, it is the place where they spend a large proportion of their daily time and a context conducive to new learning [10, 13].

Although some interventions show significant effects on the reduction of ST [8, 10], systematic reviews highlight that the effect size of interventions are small (highest effect size reported = − 0.29, 95% CI: − 0.35; − 0.22)[14], the methodological quality low [15, 16], and the results are often inconsistent (studies that present conflicting results) [15]. Another aspect is that interventions frequently report the effect on total ST [11, 17]. According to a recent systematic review, most interventions performed in Latin America analysed the intervention effect on clustered computer/videogame and total screen time [18]. However, analysing each screen-based device separately is important to identify whether the intervention’s strategies have device specific effects [8, 9] as they may impact adolescents’ health differently [19]. Excessive TV viewing, for example, is associated with unhealthy diet behaviors and obesity [20], while playing computer/videogames and the use of smartphones is related with unfavourable social behaviors [6] and depression [5]. There is currently little evidence of the impact of interventions on more recent mobile devices such as smartphones, despite being widely used by adolescents [9, 21].

There is also a lack of evidence amongst low and middle-income countries [22]. Although a systematic review of 49 screen time interventions considered in the inclusion criteria studies from high and middle-income countries, all studies included were from high-income countries [21]. In another systematic review of sedentary behavior interventions developed in Latin America, a region composed mainly of low and middle-income countries, only nine studies were included [18]. Screen time behaviors and their correlates differ between high and middle-middle income countries highlighting the need for specific intervention strategies [23].

Furthermore, there is a gap in the literature about which subgroups respond best to interventions to improve adolescent’s health behaviors [24]. Despite this need, the analysis of subgroups has rarely been performed in interventions to reduce ST [9, 16]. Some findings indicate that sex [9], age [8], and socioeconomic status (SES) [25] may influence different responses in interventions. For instance, boys appear to show greater effects and significant reductions in ST compared to girls [9, 26], and younger adolescents (11–13 years old) respond better than older adolescents (14–17 years old) [8]. Considering SES, the higher the income the more time is spent in ST among adolescents in low-middle-income countries, thus SES may be a potential moderator of the intervention effect [23]. Therefore, the aim of this study was to determine the effects of the Movimente cluster-randomized controlled trial intervention on ST device use and identify possible moderators (i.e., sex, grade, SES) in adolescents from public schools in Florianopolis, Santa Catarina, Brazil.

Of the 1427 eligible students from the six schools, 921 students answered the questionnaire at baseline (n = 538 intervention group; n = 383 control group; response rate: 65%) and 788 participants provided complete baseline data (n = 472 intervention group, n = 316 control group). From those, 678 provided complete post-intervention data (n = 399 intervention group, n = 279 control group) (Fig. 1).

Table 2 shows the baseline characteristics of students and the time spent per day on different screen devices. There were no significant differences in sociodemographic variables and ST duration between participants and dropouts (Additional file 1: Table A1). No differences were found between the control and intervention groups at baseline for any demographic or ST variables, except for game time (p-value = 0.013) as shown in Table 2. Adolescents spent an average of 299 min per day in ST, and most adolescents did not meet the ST recommendation (97.2%).

Table 3 shows the intervention effect on ST devices. According to the group-by-time interaction analysis there was no significant effect of the intervention condition after adjusting for sex, grade and SES in TV time (β = − 6.4, 95% CI: − 6.1;13.4), game time (β = − 8.2, 95% CI: − 7.2;10.8), computer time (β = 1.1, 95% CI: − 6.3;18.5), smartphone time (β = − 10.2, 95% CI: − 32.5;12.1), ST (β = − 12.8, 95% CI: − 50.5;24.8), ST guidelines (OR: 1.29, 95% CI: 0.65,2.57) and ST guidelines with smartphone (OR:1.66, 95% CI: 0.37,7.40). Thus, no differences in changes in ST device use over time between control and intervention conditions were observed.

Possible differences in intervention effects between subgroups were analysed by examining the interactions effects for sex, grade and SES. The three-way interaction terms (group*time*moderator) showed no differences in the effects of the intervention between sex, grade and SES for game time, computer time, smartphone time, ST, ST guidelines and ST guidelines with smartphone (Additional file 1: Tables A2, A3 and A4). However, an interaction effect of grade on time by group term for TV (p-value = 0.089) was observed. There was an intervention effect on reducing TV time (β = − 37.1, 95% CI: − 73.0, − 1.3) only among 8th grade students (Table 4).

This study examined whether a multicomponent intervention was effective in reducing overall ST and device specific ST among adolescents. There was no effect of the Movimente intervention on each screen device among the overall sample. However, strategies were effective in reducing TV viewing of adolescents in 8th grade by almost 40 min per week. Despite evidence that TV viewing time is being replaced by other new screen devices [33], TV still represents one of the most accessed devices in this population group, particularly in low- and middle-income countries [34] and among people of low SES [33]. According to the latest nationwide survey conducted in Brazil, 60% of 9th grade students watch more than 2 h of TV per day [35] and a systematic review showed that 58.8% of Brazilian adolescents (10–19 years) had excessive TV time. Given consistent associations between TV viewing and unhealthy outcomes [33, 34], this reduction in TV viewing may have a beneficial impact on these adolescents’ health.

The effect found only for the 8th grade may be explained by the age of the students, the characteristics of elementary schools in Brazil and strategies of the Movimente study. Perhaps the ST behavior of younger students (7th grade) is more dependent and influenced by parents and the Movimente intervention had no specific strategies either for the parents or relatives, or for the family environment. Interventions with family involvement are often more effective, particularly for reducing recreational ST [14, 36]. In comparison, 9th grade students have many different school activities, the school calendar ends earlier and they are preparing for high school, which may have made participation and access to Movimente program strategies even more difficult. Thus, the results suggest that it may be interesting to consider the different characteristics of the grades in the formulation of intervention strategies. In addition, participants’ perception of changes to their screen time throughout the intervention was not evaluated but may be interesting in further investigations.

Although the effect of the Movimente intervention on ST was not gender-moderated, other studies have shown that gender can moderate intervention effects [8, 37]. A survey study amongst French students showed that recreational ST is strongly sex dependent, with boys participating more in games (play video games with consoles, tablets, smartphones, or other electronic media) and girls in social media interactions [38]. Even in intervention studies with all participants undergoing the same strategies, there is a reduction in TV time among girls but not among boys [8] and a reduction in computer time only in boys [39]. To develop policies and actions that are more inclusive, future interventions need to document gender differences and similarities in strategies and results, since information reported in the literature is not enough yet [37].

Most studies that have investigated the effect of interventions on reducing ST among adolescents, have found distinct results for different screen components. For example, while one study observed a reduction for TV viewing and total ST [40], another study found it for TV viewing and computer/video game time [8], and another show it for computer time [9]. It is possible that some ST behaviors are more difficult to reduce or need specific intervention strategies. For example, the main strategies of the ACTIVITAL intervention focused on TV, which resulted in a greater reduction in the time spent watching TV but to the detriment of other ST behaviors [40]. Therefore, it seems that the intervention strategies need to be better elaborated to cover all screen-based devices as the excessive recreational ST provides harmful effects on health. Strategies need to be tailored towards specific behaviors (i.e. TV viewing and/or computer time), as trying to change multiple screen behaviors without specific strategies, i.e. the “one fits for all” approach, does not seem to be effective.

In addition, many interventions neglect new technologies such as smartphones, which are widely used by adolescents [3, 21]. Our intervention was not effective at reducing smartphone time. Only two other interventions with adolescents have examined smartphone use [9, 11], however, both studies included smartphone use in their sum of overall ST measure. This makes understanding and comparability difficult where other studies consider ST as TV, computer and video game time [3]. One of these interventions did also consider the smartphone separately and observed significant reduction in the prevalence of adolescents that do not meet recommendations (≥2 h per day) after intervention [9]. The study suggests that curricular and extracurricular actions, and greater involvement of students, the school community and family members are effective strategies in reducing smartphone use in adolescents [9].

The lack of effectiveness for the whole sample may be due to the strategies not being focused exclusively on ST since the Movimente intervention also focused on physical activity and other health behaviors [15]. Similarly, an intervention performed with students from New South Wales (Australia) that focused on several health behaviors (BMI, physical activity, emotional problems, behavioral problems and psychological well-being) and applied strategies similar to the current RCT but with more family involvement, health messaging and parental newsletters, also found no effect in reducing ST (− 21.3 min/day; p = 0.255, 11]. There was also no effect of a multicomponent intervention focused on promoting physical activity and healthy eating habits performed with Brazilian adolescents [41].

In contrast, a cluster RCT performed with adolescents in Cuenca (Ecuador) found significant effects for TV viewing (− 15.7 min/day; p = 0.003) and total ST (− 25.9 min/day; p = 0.03) after 18-months [40] with a multicomponent intervention. However, when the intervention strategies focusing on reducing ST stopped and the intervention focus changed to diet and physical activity, the effects were not maintained [40]. According to a recent systematic review, some interventions with high methodological quality that focused exclusively on ST reduction showed intervention effectiveness [15]. An umbrella systematic review found that interventions with strategies focused exclusively on sedentary behavior were more effective in reducing ST and sitting time than those with strategies towards both physical activity and sedentary behavior combined [22]. Though there is still no consistent evidence [15, 16], the results of these studies suggest that the interventions may be more effective if focused exclusively on reducing ST [15]. It is also important to highlight that interventions with a multicomponent approach have been widely used [14] and present an important design to promote behavioral change due to focusing on interrelated determinants of one or codependent outcomes. However, the interpretation of effects is limited to the whole intervention as the effectiveness of isolated strategies cannot be precisely estimated. Thus, qualitative process evaluation should be designed within school-based interventions to expand the understanding of the implementation’s reach and magnitude.

The lack of significant effects in the current study may also relate to intervention implementation. Poor implementation is one of main reasons that ST interventions have been unsuccessful [42]. For instance, our intervention reached only 19 of the 63 teachers (30%) who could have participated in the teacher training. After the intervention, teachers were asked which topics covered during teacher training (i.e., physical activity, diet, physical activity related to academic performance, and ST) were used in their class and ST was the least addressed. This was also identified by the students who, when asked about the topics covered during the classes, recalled more topics about physical activity and diet, with ST being rarely addressed. Thus, low teacher participation and students’ recall of the topics covered during the intervention may have contributed to the lack of effect [43].

Further, many of the strategies developed in the Movimente intervention were based on interventions performed in high-income countries [17]. The context of these countries is completely different from low- and middle-income countries, which may also affect implementation and effectiveness. For example, Australia [17], United States [10] and Finland [44] that develop good intervention models are highly ranked in regard to the educational system, while Brazil is in the last positions, with a school context where teachers are devalued and government incentives for public education are lacking. Another important difference is that most Brazilian schools are predominantly part-time, with students spending 4 h daily at school in a morning or afternoon ‘shift’. This means they are likely to be less exposed to the actions of the intervention than students in countries that require them to spend six or more hours at school at least5 days per week. There is a need for further studies to evaluate the implementation of intervention strategies in low- and middle-income countries.

This study has some limitations such as the ST variable measured by self-report, however, the used instrument has the strength of comprising multiple screen devices and was previously validated with this population [31]. The retention and response rates (65 and 80%, respectively) were lower than expected. Thus, sensitivity analyses of subgroups were conducted and no evidence of selection bias was observed. We understand that the effects on physical activity and screen time cannot be separated and are expected to be dependent due to time displacement. As an example, if strategies towards reducing overall screen time were successful, students would need to fill their extra time with another behavior. Physical activity may be the behavior of choice as stimulated by the intervention strategies focused on promoting this behavior. In such as case we cannot assume that changes of screen time were secondary or residual to our intervention. Another limitation was the low participation of teachers in teacher training, and the students’ lack of recall of the topics covered during the intervention.

The major strengths of this study are the randomized controlled trial design and the use of multilevel analysis, an approach that allowed to consider the nested nature of the data which are important to reduce inference bias and to avoid misleading conclusions on population estimates. It is also important to highlight the development of this intervention in a middle-income country that needs more evidence to support the formulation and implementation of future interventions more effective and better implemented.