Determinants of change in accelerometer‐assessed sedentary behaviour in children 0 to 6 years of age: A systematic review

Summary Sedentary behaviour tracks from early to middle childhood, suggesting the need to intervene early. The aim of this systematic review was to identify determinants of change in accelerometer‐assessed sedentary behaviour in young children, with a view to informing interventions. Ten electronic databases were searched. Longitudinal and intervention studies were included if they (a) targeted sedentary behaviour in young children (less than of equal to 6 years), (b) assessed change in accelerometer‐assessed sedentary behaviour, and (c) reported on at least one determinant of change in sedentary behaviour. Intervention components were coded according to clusters of behaviour change technique (BCT) (ie, grouping similar BCTs components). Data synthesis was guided by the socioecological model. Sixteen studies (four longitudinal; 12 intervention) met the inclusion criteria. Two (out of five identified determinants) were associated with an increase in sedentary behaviour in longitudinal studies: the after childcare/school period and transition from childcare to school. Three (out of 21 identified determinants) were associated with a decrease in sedentary behaviour in intervention studies: “goals and planning” (ie, “behavioural contract”), “repetition and substitution” (ie, “graded tasks”), and “reward and treat” (ie, “incentives”). The environmental and interpersonal determinants identified in this review may help to inform behavioural strategies, timing, and settings for future interventions.

intervention) met the inclusion criteria. Two (out of five identified determinants) were associated with an increase in sedentary behaviour in longitudinal studies: the after childcare/school period and transition from childcare to school. Three (out of 21 identified determinants) were associated with a decrease in sedentary behaviour in intervention studies: "goals and planning" (ie, "behavioural contract"), "repetition and substitution" (ie, "graded tasks"), and "reward and treat" (ie, "incentives"). The environmental and interpersonal determinants identified in this review may help to inform behavioural strategies, timing, and settings for future interventions. KEYWORDS behavioural change technique, early years, objectively-measured, socioecological model There is growing evidence that the most efficient and cost-effective way to prevent health problems is to intervene in early life before behaviour and health patterns have been firmly established. 1 Although there is a general perception that young children are spontaneously active, a review examining levels of accelerometer-assessed sedentary behaviour in children 0 to 6 years old revealed high levels of sedentary time in this age group. 2 They found that children spent a median of 77% of the day (range across studies 34% to 94%) or approximately 10 hours sedentary.
Sedentary behaviour, defined as any waking behaviour characterized by an energy expenditure less than or equal to 1.5 metabolic equivalents (METs), while in a sitting, reclining, or lying posture, 3 has been associated with obesity in children and young people. [4][5][6] It is unclear whether this association is causal, 7 and interventions targeting sedentary behaviour in children and young people (0 to 17 years old) have only shown small and clinically irrelevant effects on BMI reduction. 8 This is however a complex field as sedentary behaviour is frequently targeted alongside with other behaviours (eg, diet and physical activity) in a multibehaviour approach to prevent and treat obesity. 7 In addition to overweight and obesity, sedentary behaviour in school-age children has been associated with a range of other negative health effects including a higher clustered cardiometabolic risk score, lower fitness, unfavourable behavioural conduct, and lower self-esteem. 9 Although there is little evidence about the role of sedentary behaviour on developmental outcomes in the early years, certain screen-based sedentary behaviours may have no benefit and potential to harm motor and cognitive development. 10 It is however important to recognize that sedentary behaviours such as reading has wellknown benefits for cognitive development 11 and parent-child interaction. 12 Sedentary behaviour appears to track at moderate to high levels from early to middle childhood years. 13,14 This suggests that benefits of early intervention to reduce sedentary behaviour may be carried over into school age, where evidence of the health benefit reducing sedentary behaviour for health is more robust. 9 Establishing the determinants of a behaviour in early life is therefore important in order to intervene effectively. 15 Determinants of sedentary behaviour have been investigated previously in a systematic review of children up to 18 years old. 16 However, evidence was limited for our population of interest (less than or equal to 6 years), including only one study with very young children (toddler and preschool age). 17 This study provided a proxy report of sedentary behaviour (parent self-reported) and accelerometer-based data. In this age group, proxy-reported questionnaires are commonly used to assess sedentary behaviour due to cognitive limitations of young children. However, the use of self-report for sedentary behaviour, usually restricted to screen time, has been criticized as it accounts for only a small proportion of the sedentary behaviour that children engage in. 18,19 Parents' proxy-reported sedentary behaviour might also be influenced by social desirability and recall bias especially due to the intermittent and incidental nature of children's sedentary behaviour. 20 This is particularly true in young children where sedentary behaviour includes being restrained in a car seat, high chair, or pushchair. This systematic review is part of a collection of reviews that aim to explore the determinants of obesity-related behaviours in young children (eg, diet, physical activity, and sedentary behaviour). [21][22][23] The aim of this review is to identify determinants of change in accelerometerassessed sedentary time in young children (0-6 years old), with a view to informing interventions. Determinants will be organized according to the social ecological model as done previously. 21,23 The socioecological model provides a useful framework for identifying potential determinants at individual (ie, age, weight status, and sedentary behaviour at baseline), interpersonal (ie, family, carers, and teachers) environmental (ie, playground density and equipment), and policy (ie, safe places to cross roads and longer lunch breaks) levels.
The socioecological model recognizes that individuals are embedded within a large interactive social system, which has a cumulative effect on health outcomes. 24,25 The use of this framework will allow us to identify the level-specific determinants of sedentary behaviour. Additionally, it will use the taxonomy of behaviour change techniques to code the content of behavioural interventions. 26 The use of BCT taxonomy is included with a view to gathering knowledge to guide future research and implementation by reporting the "active ingredients" of interventions with precision.

| METHODS
This systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria. 27 The protocol for the overall systematic review process has been registered in the International Prospective Register for Systematic Reviews (PROSPERO), registration number CRD42012002881.
As stated earlier, this systematic review is part of a suite of reviews to explore the determinants of obesity-related behaviours. 21-23 A detailed protocol including study design, search, and quality assessments strategies has been published elsewhere. 28 This review deviates from the overall protocol with respect to the following inclusion criteria: (a) exclusion of cross-sectional studies; (b) exclusion of subjective measures of sedentary behaviour; and (c) exclusion of diet and physical activity search terms (an example of the search strategy 28 is presented in Data S1). One deviation from protocol on this present review was also present in another systematic review from this collection (ie, exclusion of cross-sectional studies). 21

| Study selection
For quality control, two batches of titles and abstracts (570 in total) were screened for inclusion by four reviewers. Disagreements were discussed until consensus was achieved. Since discrepancies between reviewers were low (less than 5%), the lead reviewer (L.A.) screened all remaining titles and abstracts. Full texts were subsequently obtained and read in full; eligibility for inclusion was assessed independently by two reviewers. Discrepancies were resolved by discussion or by consultation with a third reviewer until consensus was reached. studies that referred to "failure to meet a physical activity guideline" as a definition for sedentary behaviour, and (e) for intervention studies, had no control group.

| Quality assessment
Study quality was evaluated using assessment tools specific to each study design, published by the Evidence for Policy and Practice Information (EPPI) centre. 29 The quality assessment criteria are specified in or high: greater than or equal to 5. Quality assessment was performed independently by two reviewers and any disagreements resolved by a third reviewer.

| Data extraction
A standardized data extraction form was piloted, completed by one reviewer, and checked by a second reviewer. The following information was extracted by reviewers: study information (eg, author and year); baseline descriptive characteristics; study design; setting; sedentary behaviour measurement and outcomes; methods of analysis; follow-up (duration, sample, and results); and potential determinants and their association with the outcome. For all studies, the latest follow-up data available before the children were 6 years old (or as close to as possible afterwards) were included. If results were stratified by specific times of the day, data for the largest time periods were extracted. For intervention studies, all factors targeted in the intervention (eg, parental knowledge and parental modelling) were extracted as potential determinants of change in sedentary behaviour. To score these determinants, the difference in change in sedentary behaviour between control and intervention groups over time was assessed. This was deemed to provide evidence of factors targeted in interventions (ie, determinants), which were associated with change in the outcome.
Where possible, results of multivariable rather than univariable models were included.

| Data synthesis
Because of heterogeneity across studies (including design, setting, measures of determinants, and analysis type), a meta-analysis was not appropriate. A narrative synthesis and harvest plot analysis were therefore undertaken.  Hierarchically Clustered Techniques, 26 was also used to identify and cluster BCT applied. Information on protocols of included papers were also examined. The BCT coding was performed by one reviewer and verified by two others; in case of discrepancies, they were resolved through discussion.
Consistency regarding the association of each determinant from longitudinal and intervention studies with accelerometer-assessed sedentary behaviour was summarized according to Sallis et al. 30 The consistency of association was based on the percentage of reported findings that supported the hypothesized association as follows: "0" Finally, each study was presented as a bar chart and summarized using the harvest plot format. 31 The harvest plot emulates the visual representation of a forest plot providing evidence between the competing hypothesis (no change and positive or negative change), weighted by study quality and sample size.

| RESULTS
A total of 14 966 references were retrieved, of which 282 were read in full, and 16 studies (four longitudinal and 12 intervention studies) met the inclusion criteria ( Figure 1). Sixty-one studies were excluded because of proxy or self-assessed sedentary behaviour.

| Longitudinal studies
The main characteristics and findings of the longitudinal studies are summarized in Table 2. A total 1454 children took part in the four included studies. Follow-up duration varied between 1 and 3 years.
Overall, sedentary time increased significantly over time in two studies 45,46 and remained stable in the other two studies. 34,41

| Intervention studies
The main characteristics and findings of the intervention studies are summarized in Table 3

| Quality assessment
One longitudinal study scored high in the quality assessment, 45 while the remaining three were of intermediate quality. 34,41,46 Eight of the intervention studies were considered high quality, 32,33,35,37,38,40,44,47 including those demonstrating a significant change in sedentary behaviour, 37,44 and four intervention studies were of intermediate quality. 36,39,42,43 A description of the quality assessment score of each study is provided in Data S2. Table 4 shows a summary of all identified determinants and the direction and strength of the association combined with the harvest plot.

| Intervention studies
Only one intervention study targeted all the levels of the socioecological model 39 ( Only three BCT components were identified as determinants of decreases in sedentary behaviour. These included "behavioural contract" (cluster-"goals and planning"), "graded tasks" (cluster-"repetition and substitution"), and "incentive" (cluster-"reward and treat"). However, these determinants were extracted from a single high-quality study with a small sample size (N = 43 intervention and N = 33 control). 37 There was inconsistent evidence for "non-specific reward and material reward" (cluster-"reward and treat") as determinants of sedentary behaviour. The findings gathered from intervention studies suggest that "behavioural contracts" (BCT cluster-"goals and planning"), "graded tasks" (BCT cluster-"repetition and substitution"), and "incentives" (BCT cluster-"reward and treat") were associated with decreases in sedentary behaviour. However, these determinants were only identified in one study each. According to the Behaviour Change Taxonomy (v1), 26 "behavioural contracts" are when a targeted behaviour is specified, written and signed in a contract, agreed by one person, and witnessed by another. For "graded tasks" individuals are initially set easy to perform tasks and are then challenged to progress at achievable levels until the behaviour is performed. Finally, for "incentives," participants are informed that a reward will be delivered only if there has been an effort (or progress) in achieving a behaviour.
It is important to note that although only two intervention studies 37,44 showed statistically significant reductions in accelerometerassessed total sedentary behaviour, four others studies included here found a significant decrease in screen-viewing behaviour. 33,42,43,47 In two studies, 42,47 there was a reduction in electronic media use 47 or TV viewing, 42 while in the others, 33,43 although there was no effect on total TV viewing, there was a subgroup effect (ie, girls reduction in TV viewing on weekends) 33 and changes in parenting outcomes related to TV viewing (ie, TV snacks and dinner in front of TV). 43

| Findings in context of previous research
In this review, we found evidence, although limited, 45        Overall summary of findings for each outcome with the direction of association. 3 Number of studies which support the association of the grouping behaviour change technique 4 Overall summary of findings for the grouping behaviour change technique Notes: Bar charts were displayed as follow: 1. Position based on direction of association (decrease in sedentary behaviour (−), no change in sedentary behaviour, increase in sedentary behaviour (+); 2. Height of bar represented size of study (short <300 participants, medium 300-500 participants, high >300 participants); 3. Colour of bar representing quality: black, dark grey and white with darker bars representing higher quality studies; 4. Symbol on top show study for identification.    16 However, this may be because of the limited age range of participants included in studies conducted in early years, which restricts the opportunity to investigate this exposure as a determinant.
By focusing on the key ingredients of interventions, the BCTs identified in this review might help to inform future interventions to aid longer term behaviour change in young children. "Behavioural contracts" have been shown previously to positively impact physical activity for older adult populations and disease-specific conditions. 59,60 There is, however, limited evidence on younger and healthy populations. One example is an adolescent-targeted intervention that used behavioural contract in addition to other intervention features, which was successful in reducing screen-time in the intervention group, although no between-group differences were observed. 61 Likewise, the BCT "graded tasks" has predominantly been used in adults. A previous systematic review found that implementing "graded tasks" was associated with successful outcomes in longer term when promoting physical activity and healthy eating in adults with overweight and obesity. 59 Evidence in young children is however limited. Finally, while "incentives" appear to support change in behaviour in adults, [62][63][64] few studies have investigated the effect on behaviour change in children and those that have focussed primarily on diet in children at school age. 65,66 In the studies identified in this review, 37,42,43 different forms of incentives were delivered under the "reward and threat" cluster including "incentives (outcomes)," "material reward," "social reward," and "non-specific reward." Only "incentive (outcome)" (definition according to Michie et al 26 "inform that a reward will be delivered if and only if there has been effort and/or progress in achieving the behavioural outcome") was successful. 37 Previous review-level evidence 21 highlighted parental monitoring as a determinant of physical activity in young children. The same systematic review found that provider training was moderately associated with vigorous physical activity; however, child and parental knowledge was consistently not associated. In this study, we found that "shaping knowledge" (BCT component-"instruction on how to perform a behaviour") was not associated with changes in sedentary behaviour at all levels of the socioecological model. This reflects findings (ie, child and parental knowledge) of the previous physical activity review 21 ; however, shaping knowledge at childcare and preschool level is not associated with changes in sedentary behaviour.
Elements of the physical environment have been frequently investigated as a determinant of physical activity and sedentary behaviour. 16,21 Similar to our systematic review, a previous review that focused on determinants of physical activity found that restructuring the physical environment in preschool did not lead to changes of this behaviour in early years. 21 Likewise, another systematic review on the determinants of sedentary behaviour in youth 16 found that although environmental determinants were explored in a large number of studies, few found an association with sedentary behaviour.

| Strengths and limitations
To our knowledge, this is the first systematic review to assess the determinants of change in sedentary behaviour in young children.
The use of accelerometer-assessed sedentary time is a key strength, as self-report measures tend to focus on TV or screen viewing, which has been shown to have low validity to measure total sedentary time. 18,20 However, it can also be seen as a limitation as self/proxy report measures provide contextual information (ie, setting and type of activity) about sedentary behaviour that provide valuable information about sedentary activities undertaken by young children. 67 No time or language restrictions were applied, ensuring high sensitivity in identifying the literature. However, it is possible that all relevant publications were not included, and publication bias cannot be ruled out. Moreover, although all effort was made to extract information of intervention features (and therefore determinants) from relevant documents (ie, protocols, trial registers, supplementary files, and additional papers), it was not always possible to detail the exact intervention elements for all studies. 68 Furthermore, it is possible that the intervention strategies embedded in the included studies were not captured by the coding of BCT taxonomy if these were not clear or sufficiently precise. The use of template for intervention description and replication (TIDieR) checklist to specify essential elements of the Parents/carers target module consisting of educational materials (posters and flyers) School community group: implement modules at school level. Educational materials were distributed through the school and the community.
i. Children ii. Parents/care giver iii. Schools, community intervention and the use of the BCT taxonomy coding in future studies might help better identifying elements of interventions in future 68 and facilitate evidence synthesis that could guide implementation. 69 Despite substantial heterogeneity in the included studies, exposure and outcome measures, the combined used of summary tables to assess consistency of associations across studies, 30 and the use of Harvest plot 31 enabled us to provide a detailed summary of findings.
Although we included a limited the number of studies, they were of intermediate (n = 7) and high (n = 9) quality, strengthening the findings reported here. As the majority were intervention studies, this highlights a lack of high-quality longitudinal observational research in this age group. Moreover, all studies were conducted in high-income countries, and findings cannot therefore be generalized to low-and middleincome countries.

| Recommendations for policy and practice
Although several interventions have been developed to target sedentary behaviour in childcare/school setting, 70,71 it appears that the after school is a period of high prevalence in sedentary behaviour. 72 This suggests that more needs be done to prevent sedentary behaviour in the home environment. 73 Similarly, childcare settings may be more supportive than the formal school setting for reducing sedentary behaviour. The more structured curriculum in primary schools may reinforce sedentary behaviour; therefore, initiatives to reduce sitting time such as classroom-based physical activity 74 or standing desks 75 might be good strategies to be implemented at schools.
Also, from this review, we found strong evidence that shaping knowledge (instruction on how to perform a behaviour) at individual, parents/carers, and at childcare/preschool environment is not sufficient to change sedentary behaviour of young children. Therefore, we recommend that in practice, instruction on how to perform a behaviour should not be delivered in isolation, as it might not bring the expected benefits on the reduction of sedentary behaviour.
Similar findings we observed on restructuring the environment, as it seems that when this BCT component is implemented by its own 32 or in some cases in combination with other BCTs, 33,40,43,47 it does not promote the expected reductions in sedentary behaviour. However, more evidence is needed.
Interpersonal determinants such as having "behavioural contract" (cluster "goals and planning"), promoting "graded tasks" (cluster "repetition and substitution"), and receiving "incentives" (cluster "reward and treat") might be appropriate behavioural strategies to be incorporated into sedentary behaviour interventions in young children.
Although more evidence is needed, interventions may benefit from incorporating other BCT components in the cluster of "goals and planning," "repetition and substitution," and "reward and treat."

| CONCLUSION
We identified limited evidence on the determinants of change in accelerometer-assessed sedentary time in children 0 to 6 years. The available evidence suggests that the after childcare/school period and transition from childcare to formal school are potential determinants. Furthermore, the following determinants at the interpersonal level were associated with a decrease in sedentary behaviour: goals and planning (ie, behavioural contract), repetition and substitution (ie, graded tasks), and reward and treat (ie, incentives). More longitudinal and intervention research is needed to provide more robust evidence on the determinants of sedentary behaviour in children, to in turn inform the development of effective interventions.