Pediatric obesity continues to be a serious public health issue with trends indicating that the prevalence of childhood obesity will reach 24% in the USA by 2015 [1]. Reversing the pediatric obesity epidemic has been established as a critical priority [2]. Schools are the most frequent target for prevention and intervention programs because they reach the largest segments of the youth population and provide an available infrastructure and support [3–6]. Many school-based interventions aimed at modifying physical activity (PA) and diet have been conducted but the overall effectiveness of these programs has been limited [7, 8]. One limitation is that many have not included families or communities.

Family components are critical for youth obesity prevention programs because parents directly and indirectly influence children's PA and nutrition behaviors at home and also dictate the physical and social environments that are available to their children [9]. Several studies have confirmed the importance of parenting behaviors and home environments in influencing children's nutrition and PA behaviors [10–12]. Despite the clear importance of families, family-based interventions (even those implemented through schools) have not been shown to be effective [13, 14]. A possible reason for this is that these studies were conducted under isolated conditions and have not been well integrated into the schools and communities where families live, work and play. Several studies have demonstrated that home, school, and community environments interact over time to influence children's health and development [15, 16], so it may prove more effective to integrate school and family interventions into broad community-based initiatives.

Social ecological models that target multiple levels of influence have been recommended to address the obesity epidemic [2, 17]. Although a number of social ecological models exist to facilitate planning of multilevel interventions [18–21], there are few, if any, examples of studies that have implemented a multilevel intervention aimed at changing behavioral risk factors for obesity. This paper presents the main outcomes of Switch™, a family-, school-, and community-based intervention aimed at changing key behaviors (PA, television viewing/screen time, and nutrition) related to childhood obesity. Third through fifth graders were targeted to potentially avert or reduce the documented declines in fruit and vegetable consumption (FV) [22], PA [23], and increases in screen time (ST) [24, 25] that occur during the late elementary and middle school years. Multiple behaviors were targeted in order to produce larger overall changes in energy balance. The American Dietetics Association (ADA) recently completed a comprehensive evidence analysis of behavioral factors that have been shown to be associated with childhood overweight [26]. Adequate evidence (Grade II) was found for five key behaviors (promoting PA, discouraging TV watching, decreasing dietary fat, decreasing sweetened beverages, and increasing FV). Consistent with these findings, the primary objectives of Switch were to: (1) increase the amount of PA; (2) reduce the amount of ST (television and video game time); and (3) increase FV consumption among children from the third grade to fifth grade.

Characteristics of the sample at baseline are shown in Table 1. The average BMI was 18.4 kg/m² (SD = 3.4, range 12.4 to 35.4 kg/m²). Overall, 19% of children were classified as overweight and 8% were obese. There were significant differences between boys and girls on child-reported and parent-reported weekly ST (t (1,144) = 6.35, P < 0.001, and t (976) = 5.69, P < 0.001, respectively). This was due primarily to video game play. By child report, boys and girls were not significantly different on the number of fruits and/or vegetables they had eaten the previous day. Based on national guidelines (five servings of fruits and vegetables per day) approximately 44% of the children met the recommended guidelines. By parent report, girls ate almost two more servings of fruits and vegetables per week compared with boys (M = 21.7 and 23.4 for boys and girls, respectively; t (957) = 2.22, P < 0.05). Only 17% of children met the recommended guidelines of at least five servings a day. As expected, boys tended to have greater step counts than girls (t (823) = 9.01, P < 0.001). Depending on the recommendation used, between 28% and 43% of children achieved the recommended level of PA. Boys and girls were not significantly different on meeting the Vincent and Pangrazi recommendation (41% and 45%, respectively), but were significantly different on meeting the stricter Tudor-Locke recommendation (23% and 31%, respectively; χ² = 7.7, df = 1, P < 0.01).

The goal of the present research was to determine whether a multilevel intervention conducted in two states could modify key behaviors related to overweight or obesity. The novel aspect of the study was the community-, school-, and family-based approach to modify nutritional and activity behaviors. In general, baseline data indicated that these children were similar to national averages for this age group. The intervention yielded encouraging results despite the relatively short-term length of the intervention, the relatively small number of schools, and the naturalistic design. The experimental group showed significant differences from the control group on both ST and FV consumption at both the immediate post-intervention assessment and at the 6-month follow-up assessment, controlling for baseline values. The results for selected variables were moderated by child sex, family involvement, and child BMI. The intervention was designed to influence three behaviors that together can have an impact on BMI over the long term and was not expected to have a short-term direct effect on BMI. It is likely that longer-term studies would be needed to document changes in BMI resulting from changes in nutritional and activity behaviors.

The effect sizes reported here range from small to large and are similar in magnitude to other published intervention studies [39–41]. Although modest, these results are not trivial. Obesity is a multifactorial health condition, and the target behaviors of the Switch program are only three out of many risk factors [39, 42]. If even small changes can be enacted and maintained, the changes may have large effects in the long term [43–45]. It is also encouraging that parents and children perceived that substantial changes had been made during the intervention, and that several positive changes appear to have been maintained 6 months after the program was ended.

There are several strengths of the Switch intervention, including targeting multiple ecological levels and multiple behaviors. A number of school-based intervention studies have been conducted to promote healthy eating and PA [40, 41, 46, 47], but few have also targeted reductions in ST [48, 49]. Although often thought of as synonymous constructs, it has been demonstrated that PA and ST have independent effects on risk for overweight [50, 51], so both should be included in interventions. Most school-based studies have focused only on curricular or policy changes in physical education and school food service [52–54]. Comprehensive obesity prevention necessitates cooperation from families but this has often been neglected. In their review, Flynn et al. [40] highlighted the need for studies that include home and community settings, and the present study found significant changes using measures that focused on behavior in both home and school settings. Another strength of this study was the participatory approach, allowing teachers and parents to adapt the program to their needs. Most previous school-based interventions have utilized tight controls to ensure uniform implementation, but these require frequent staff training and ongoing support [52, 55]. That approach is costly and limits sustainability. Our approach was to standardize recommendations to communities, teachers and parents but to allow flexibility in how the materials were used. This approach provides sufficient standardization for statistical analyses while maintaining the external validity of the study. Other studies have demonstrated that this participatory approach has advantages for long-term sustainability [56, 57]. A final strength was the employment of multiple informants. Although the effects on FV consumption and ST were in the same direction for both child and parent respondents, the effects of the intervention appeared stronger based on parental report. It is unclear why this was found. It is possible that parents are more accurate reporters of children's habits. Another possibility is that parents are more susceptible to demand characteristics. This does not seem particularly likely, since we would expect this bias in both the experimental and control groups. Furthermore, parents were not uniformly more positive than children: parents were more positive about ST, similar about FV consumption, and less positive about PA (Table 4). Although we cannot determine why the effects appear somewhat different based on the informant, the fact that the effects were in the same direction demonstrates a certain degree of robustness. Furthermore, by obtaining data from multiple informants we were able to minimize the impact of potential measurement problems.

A key challenge in the Switch project was to document and track the extent to which children, parents, and teachers utilized the specific strategies that were provided. Other studies have also reported challenges in getting families to maintain participation in school and/or community projects. Most large-scale, prevention-based interventions targeting families have generally been ineffective because of this limitation [13, 14]. In contrast, obesity treatment programs involving families have been shown to be highly effective, if not essential [58]. In fact, an evidence analysis supported by the ADA on youth obesity interventions [26], rated the strength of evidence supporting family involvement with the strongest possible grade (Grade I). It is not clear why family involvement matters but it is likely that parents assume greater responsibility and are more willing to be involved in a clinically based weight loss program than a project aimed at prevention. A major challenge for prevention intervention programs is to determine how to increase the proportion of participants who fully engage in the intervention activities [59, 60]. Our results demonstrate that participants who were more involved in the program saw larger improvements in ST and FV than participants that were not as involved. This suggests that the effectiveness of programming would be greatly enhanced if it was possible to get more families to engage fully in the intervention activities.

Community messages were distributed by newspapers, radio advertisements, and billboards, but it is difficult to document the impact that these social-change messages independently had on the outcomes as experimental families were receiving parallel messages from several ecological sources (that is, direct, school, and community). Furthermore, these community-based messages were available to both experimental and control families and, therefore, may have minimized differences between the groups.

Although over 1,300 families participated in the study, a major limitation of this study was the sample size. Statistical analysis of treatment effects was at the school level due to families being nested within schools and randomization at the school level, resulting in low power to detect differences. Future studies should randomize at the classroom or family level to have greater degrees of freedom (although this could lead to greater problems with cross-contamination between experimental and control groups). Another limitation is the potential for social desirability to influence the results. It seems unlikely, however, that only the experimental group would be subject to such desirability biases. To reduce the potential effects of bias and method variance, we used multiple informants. An additional limitation was reliance on pedometers to measure PA. Although it seems preferable to have an objective measure, compliance with the requirements was lower than anticipated. Furthermore, the pedometer captures the total amount of PA and does not provide information on the intensity of PA. It is unknown if the amount of time spent in moderate-to-vigorous PA was altered by the intervention. IHere, it is possible that the total number of steps/day was unchanged yet the time spent in moderate-to-vigorous PA increased. Future studies should use PA monitoring tools that capture both the volume and intensity of PA.

In conclusion, the results indicate that the Switch program produced modest treatment effects for children's FV and ST. It is noteworthy that the effects remained significant in the 6-month follow-up evaluation as this indicates maintenance of these differences over time. Although levels of BMI were not statistically different between groups, the maintenance of behaviors over time may contribute to reduced risks for overweight in the future. The Switch program which targets multiple behaviors through multiple channels offers promise for use in youth obesity prevention.