Sociodemographic moderators of longitudinal changes in active play between childhood and adolescence in Australia

ABSTRACT

Physical activity (PA) participation is prone to decline during childhood and adolescence. In Australia, this decline has been shown to particularly occur in active play. This study aimed to identify sociodemographic moderators of change in active play between 10-11y and 12-13y among Australian youth. The data were sourced from Waves 6–7 of the Longitudinal Study of Australian Children (n = 3567). Active play participation was measured using one-day time-use diaries (TUDs) completed by youth. Potential sociodemographic moderators were tested using multilevel mixed modelling, adjusted for pubertal development, body mass index z-score and TUD contextual variables (school attendance and season). Active play declined more among girls (β= −7.6 min/day, 95% CI = −13.3, −1.8), those who spoke English at home (β= −12.3 min/day, 95% CI = −22.0, −2.7) and marginally among those in regional/remote areas (β= −6.3 min/day, 95% CI = −12.8, +0.1). A widening gap in active play by sex was observed, while differences by language spoken at home and geographical remoteness weakened or became marginal over time. Interventions to promote active play could target girls in the transition to adolescence. Future studies could investigate whether active play declines earlier than 10-11y among youth who speak languages other than English at home and those living in urban areas.

KEYWORDS:

• ‘Exercise’
• ‘health promotion’
• ‘recreation’
• ‘youth’
• ‘physical activity’

Introduction

A range of health benefits have been linked with regular participation in physical activity (PA) (Poitras et al., 2016). However, PA participation is often insufficient among youth and tends to decline with age (Aubert et al., 2021). Studies in recent years have sought to explore whether this decline occurs in certain groupings of PA participation, known as domains of PA (Carson & Hunter, 2020; Kemp et al., 2019). For example, a previous Australian longitudinal study reported that almost the entire decline in PA participation between childhood and adolescence occurred in the domain of non-organised PA (Kemp et al., 2020). A subsequent study narrowed this down further, revealing that more than 80% of the decline in non-organised PA occurred in the subdomain of active play (Kemp et al., 2022). Active play includes types of PA that occur for their own sake in a “playful context”, such as playground games (e.g., “tag”, hide and seek) and using play equipment (e.g., trampoline, hula hoop) (Kemp et al., 2022; Pellegrini & Smith, 1998). As such, active play appears to be a particularly important component of the change in PA participation between childhood and adolescence in Australia.

Information about factors that are associated with changes in active play between childhood and adolescence could influence the design of targeted interventions to promote PA participation. However, limited evidence exists about the factors that are longitudinally associated with change in active play during this stage of life. Previous longitudinal studies of the predictors of active play have largely focussed on early- to mid-childhood (Huston et al., 1999; Remmers et al., 2014; Xu et al., 2016) with only one known study having explored changes during the transition from late-childhood to adolescence (Cairney et al., 2014). However, this study only tested three predictors of active play (sex, chronological age and biological age) among Canadian youth between 11y and 14y (Cairney et al., 2014). This study reported that active play declined more among girls and among those with more advanced pubertal development (after the time of peak height velocity) (Cairney et al., 2014). Although this study provides useful evidence, testing a broader range of sociodemographic variables is likely to yield additional information to support PA promotion.

An initial step in the design of targeted interventions is to use broad sociodemographic characteristics to identify which population group(s) may be a priority (Dietrich, 2017). After this, health promotion practitioners who have limited time or resources can target these groups on an a priori basis (Dolnicar & Grün, 2017). This might involve developing intervention strategies for specific sub-groups of the population, such as social marketing initiatives or after-school programmes. Alternatively, practitioners with more available resources could use data-driven techniques such as cluster analysis to segment priority groups further (Dolnicar & Grün, 2017). Therefore, this study aims to identify the sociodemographic moderators of change in active play participation between 10-11y and 12-13y among Australian youth. To maximize the potential for research translation, this study focusses on broad sociodemographic variables that health promotion practitioners could access and apply in their local geographic context (e.g., via publicly-available census data) (Australian Bureau of Statistics, 2022b).

Methods

This study utilised data from the Birth (B) cohort of the Longitudinal Study of Australian Children (LSAC), a project managed by the Australian Department of Social Services (Mohal et al., 2021). Data are available to researchers in Australia upon application (Department of Social Services; Australian Institute of Family Studies; Australian Bureau of Statistics, 2021). LSAC data collection procedures were approved by the Australian Institute of Family Studies ethics committee and parents/guardians provided informed consent for children’s participation at baseline (Johnstone, 2022). Approval to use data in the present study was granted by the University of Wollongong (UOW) Human Research Ethics Committee (HREC 2022/046).

LSAC B cohort participants were recruited in 2004 at 0–12 months old (Mohal et al., 2021). A two-stage clustered design was used to recruit participants, involving the random selection of postcodes and then families based on the Australian Medicare database (Mohal et al., 2021). Participant selection was also stratified to ensure national representation in each Australian state and territory and across urban, regional and remote areas (Soloff et al., 2022). A total of 5107 families were recruited at baseline (57% of those invited to participate) (Soloff et al., 2022). Data collection waves have occurred every 2 years since baseline (Mohal et al., 2021). This study uses B cohort data from Wave 6 (2014, 10-11y) and Wave 7 (2016, 12-13y). These waves coincide with the transition from primary to secondary school in Australia, which often aligns with the onset of puberty and marks the transition from childhood to adolescence.

The main outcome in this study was active play participation (min/day), measured at 10-11y and 12-13y using a one-day time-use diary (TUD). Self-report instruments such as TUDs are often used to collect information about PA context, as this cannot easily be determined using device-based measures (e.g., accelerometry) (Carson & Hunter, 2020). Only two waves of data could be included in this study because reduced “light” diaries were used before 10-11y and free-text descriptions of activities were not available after 12-13y. Participants were provided with a structured paper diary along with instructions to record, in their own words, the activities they were engaged in during the waking period on the day preceding their LSAC interview at each time point (Corey et al., 2017). Participants who attended school on this day were instructed to record activities occurring both before and after school, as well as during school breaks (including recess and lunch breaks) but not activities that happened during school lessons (including physical education (PE) classes) (Corey et al., 2017). TUD entries were coded by LSAC interviewers using a predetermined framework during the home visit, and interviewers were trained to prompt the child for information during this process (e.g., to fill gaps in the diary) (Corey et al., 2017).

In the present study, the duration of active play participation in both waves was extracted from LSAC datasets in a similar manner as in the previous research (Kemp et al., 2020). In addition, a quality improvement process was undertaken to refine TUD codes based on free-text descriptions of activities in the datasets (Kemp et al., 2022). Active play was derived from TUD entries that were deemed likely to: (1) consist of moderate-to-vigorous PA (≥3.0 metabolic equivalents, evaluated in a process described previously) (Kemp et al., 2022); (2) satisfy Pellegrini and Smith’s criteria for physically-active play (occurring for its own sake in a “playful context”) (Pellegrini & Smith, 1998); and (3) align with at least one example of physical play in the Tool for Observing Play Outdoors (Loebach & Cox, 2020). Specific activities that were coded as active play are: flying disc games/boomerang throwing, going to the park, flying kites, hide and seek, hula hoop, kicking ball/ball games, playground/play on play equipment, rope skipping/skipping, slip n slide, throwing ball against wall, tips/chasing/running around, trampolining, water fight, unstructured active play (not further defined).

Ten sociodemographic variables were tested as potential moderators of active play participation: sex (male/female), the child’s age at the time of the interview (in months), whether a language other than English was spoken at home (yes/no), Indigenous status (whether the child was of Aboriginal or Torres Strait Islander origin: yes/no), whether the child had two parents living in the home at the time of the interview (yes/no), the number of siblings living in the home, the type of school that the child attended (public versus private/independent), whether all carers living in the home were working full-time (defined by LSAC as working at least 30 hours a week: yes/no), geographical remoteness (major city/regional-remote) and the socioeconomic position of the family (z-score). Socioeconomic position was an LSAC-derived scale based on household income, parental educational attainment and parental occupation (Blakemore et al., 2009). This scale has been validated against experiences of economic hardship and receiving income support payments (Blakemore et al., 2009). Geographical remoteness was based on the Remoteness Structure in the Australian Statistical Geography Standard (Australian Bureau of Statistics, 2022a).

Other variables, including TUD contextual variables (school attendance and season of measurement), body mass index (BMI) z-score and pubertal development, were deemed likely to be associated with children’s active play participation (Kemp et al., 2022; Lee et al., 2021) and were therefore included as potential confounding variables in analyses. School attendance on the day of TUD completion was included as a variable in LSAC datasets and missing data was imputed using the “school lessons” code (it was assumed that children attended school if this code was used). Season of measurement was based on the month of the LSAC interview. BMI z-score was calculated by LSAC based on the child’s height and weight (as measured by LSAC interviewers). Pubertal status was based on the Petersen pubertal development scale which asked parents to rate the extent of the child’s body hair development, growth spurt, skin changes, breast changes (females), menarche (females), facial hair growth (males) and voice changes (males) (Petersen et al., 1988). This scale has been validated against physicians’ assessments, and it has adequate reliability (α = 0.77) (Petersen et al., 1988).

Data processing was conducted using SPSS version 25 (IBM Corporation, Armonk, NY, USA), and analyses were performed using Stata 17 (StataCorp, College Station, TX, USA). As with previous research (Kemp et al., 2020), visual inspection of frequency histograms revealed that respondents tended to round their TUD entries to the closest 5 minutes. Therefore, participation in active play was consistently rounded to the nearest 5 minutes for all cases. Additionally, the LSAC Wave 1 and 6 population weight was used in analyses to help improve the representativeness of the data. Effect sizes and confidence intervals (instead of p-values) were used to report and interpret the results, as they provide a better understanding of both the magnitude and clinical significance of the observed difference, which are important for informing population health intervention (Matthay et al., 2021).

Longitudinal interactions between potential moderators and active play were tested using separate multilevel mixed-effects models. Autocorrelation in the outcome variable was tested and deemed negligible, although robust standard errors were used in models due to some heteroscedasticity and non-normality of residuals. Mixed models assessed the effect of wave of measurement (W6/W7) on the duration of active play (level 1), nested within individuals (level 2). Preliminary unadjusted models tested interactions between the wave of measurement and each potential moderator separately. These analyses were then repeated in preliminary adjusted models which included potential confounding variables. Following this, a final adjusted model was performed which included variables that were associated with active play in preliminary adjusted models (i.e., where the 95% confidence interval for the regression coefficient did not intersect with zero) as well as the potential confounders. All available data were included in models, as multilevel modelling does not require complete cases for every wave (Rabe-Hesketh & Skrondal, 2008).

Results

A total of 3567 participants were included in the analytic sample of this study. Of these, 3174 participants (89%) provided data for both waves (average length between waves = 24.7 months [SD 3.3]). Table 1 outlines the characteristics of the analytic sample. The sex distribution of participants in the analytic sample at W7 was similar to that of 12–13 year-olds in the 2016 Australian Census (48.7% were girls compared with 48.6% in Census data) (Australian Bureau of Statistics, 2022b). However, the analytic sample somewhat overrepresented participants who spoke English at home (91.7% at W7 versus 82.7% nationally) and those who lived in regional or remote areas (36.8% at W7 versus 30.4% nationally). The analytic sample also underrepresented Aboriginal and Torres Strait Islander participants (2.4% at W7 versus 5.4% nationally) and those who attended public schools (49.6% at W7 versus 58.3% nationally) (Australian Bureau of Statistics, 2022b). However, these discrepancies all somewhat reduced when data weights were applied (see Table 1, footnote d). BMI z-scores (0.3 [SD 1.3] versus 0.3 [SD 1.1]) and active play participation (39.0 min/day [SD 63.3] versus 38.4 min/day [SD 61.3]) were similar between boys and girls at W6; however, girls exhibited more advanced pubertal development than boys (1.9 [SD 0.5] versus 1.5 [SD 0.4]).

Table 1. Characteristics of the analytic sample ^a, unweighted LSAC data.

	Wave 6 (n = 3567)^b	Wave 7 (n = 3174)^b
Age in years, mean (SD)	10.4 (0.5)	12.5 (0.5)
Girls, n (%)	1749 (49.0%)	1547 (48.7%)^c
Speaks English at home, n (%)	3257 (91.3%)	2912 (91.7%)^c
Lives in regional or remote area, n (%)	1306 (36.6%)	1169 (36.8%)^c
Aboriginal or Torres Strait Islander, n (%)	95 (2.7%)	77 (2.4%)^c
Socioeconomic position z-score, mean (SD)	0.0 (1.0)	0.0 (1.0)
Attends public school, n (%)	2195 (61.6%)	1573 (49.6%)^c
Has two parents living in the home, n (%)	3044 (85.4%)	2651 (84.6%)
Number of siblings living in the home, mean (SD)	1.6 (1.0)	1.6 (1.0)
All carers in the home working full-time (≥30 hours/week), n (%)	1297 (36.5%)	1276 (41.1%)
Body mass index z-score, mean (SD)	0.3 (1.2)	0.3 (1.0)
Pubertal development score, mean (SD)^d	1.7 (0.5)	2.3 (0.7)
Attended school on day of TUD, n (%)	1664 (48.7%)	1535 (52.1%)
Season of measurement, n (%) Summer Autumn Winter Spring	71 (2.1%) 836 (24.4%) 1367 (39.9%) 1149 (33.6%)	253 (8.6%) 437 (14.8%) 1611 (54.7%) 646 (21.9%)
Active play participation (min/day), mean (SD)^e	38.7 (62.3)	17.1 (44.7)

LSAC = Longitudinal Study of Australian Children; n = number of participants; % = proportion of analytic sample; SD = standard deviation; TUD = time use diary.

^aThe sample shown here was included in main models which required valid data for active play on at least one occasion and a valid data weight.

^bVariable-specific missing data: two parents living in home (W6 n = 1, W7 n = 40), number of siblings (W6 n = 1, W7 n = 40), attends public school (W6 n = 1, W7 n = 5), all carers in the home working full-time (W6 n = 14, W7 n = 68), socioeconomic position (W6 n = 31, W7 n = 53), BMI z-score (W6 n = 84, W7 n = 145), pubertal development (W6 n = 18, W7 n = 19), attended school on day of TUD (W6 n = 151, W7 n = 228), season of measurement (W6 n = 144, W7 n = 227), active play participation (W6 n = 151, W7 n = 228).

^cWeighted proportions for Wave 7 (comparison with 12–13 year-olds in the 2016 Australian Census, excluding “not stated” responses): 48.9% were girls (Census: 48.6%), 87.8% spoke English at home (Census: 82.7%), 2.8% were of Aboriginal or Torres Strait Islander origin (Census: 5.4%), 33.4% lived in a regional or remote area (Census: 30.4%), 53.3% attended public school (Census: 58.3%) (Australian Bureau of Statistics, 2022b).

^dThe pubertal development scale ranged from 1 (least developed) to 4 (most developed).

^eActivities that were coded as active play: flying disc games/boomerang throwing, going to the park, flying kites, hide and seek, hula hoop, kicking ball/ball games, playground/play on play equipment, rope skipping/skipping, slip n slide, throwing ball against wall, tips/chasing/running around, trampolining, water fight, unstructured active play (not further defined).

Table 2 presents the models used to test potential moderators of change in active play participation between 10-11y and 12-13y. Adjusted preliminary models revealed that active play declined more among girls (β = −7.6 min/day, 95% CI = −13.3, −1.8), those who spoke English at home (β = −14.3 min/day, 95% CI = −23.9, −4.7) and those who lived in regional or remote areas (β = −8.3 min/day, 95% CI = −14.7, −1.9). When these variables were included together in the final model, the moderating effect of sex remained constant. However, the moderation effects of the other two variables weakened, with the effect of geographic remoteness in particular becoming marginal. These trends are depicted in Figure 1. The full results of all models (including results for confounding variables) are provided in Supplementary Files A and B.

Figure 1. Changes in mean active play duration (min/day) between 10-11y and 12-13y, stratified by sex, language spoken at home and geographical remoteness, with 95% confidence intervals, unweighted LSAC data.

Table 2. Results of preliminary models testing moderators of change in active play participation (min/day) between 10-11y and 12-13y, weighted LSAC data^a.

Potential moderator	Preliminary unadjusted models^b		Preliminary adjusted models^b		Final model (n = 3567)^c
	n	β (95% CI)	n	β (95% CI)	β (95% CI)
Sex (girls)	3614	−8.3 (−14.1, −2.4)	3568	−7.6 (−13.3, −1.8)	−7.6 (−13.3, −1.8)
Speaks English at home	3614	−12.3 (−22.5, −2.1)	3567	−14.3 (−23.9, −4.7)	−12.3 (−22.0, −2.7)
Lives in regional or remote area	3614	−6.8 (−13.2, −0.4)	3568	−8.3 (−14.7, −1.9)	−6.3 (−12.8, +0.1)
Age in months	3614	+0.1 (−0.7, +0.8)	3568	+0.0 (−0.7, +0.8)	–
Aboriginal or Torres Strait Islander	3614	−11.5 (−31.2, +8.3)	3568	−10.0 (−29.9, +9.9)	–
Socioeconomic position z-score	3601	+1.2 (−1.5, +3.8)	3556	+1.7 (−1.0, +4.4)	–
Attends public school	3612	+1.5 (−4.5, +7.6)	3568	+1.6 (−4.5, +7.6)	–
Two parents living in the home	3613	+0.6 (−8.0, +9.3)	3568	+2.8 (−6.0, +11.6)	–
Number of siblings living in the home	3613	−0.3 (−3.0, +2.4)	3568	+0.1 (−2.5, +2.8)	–
All carers in the home working full-time^d	3611	−0.9 (−7.1, +5.3)	3566	−0.8 (−6.9, +5.3)	–

LSAC = Longitudinal Study of Australian Children; n = number of participants; β = unstandardised regression coefficient; 95% CI = 95% confidence interval for the unstandardised regression coefficient.

^aSupplemental Files A and B provide additional data, including results for potential confounding variables: TUD contextual variables (school attendance and season of measurement), body mass index z-score and pubertal development.

^bMixed models assessed the effect of wave of measurement (W6/W7) on the duration of active play, nested within individuals. Preliminary unadjusted models tested interactions between wave of measurement and each potential moderator separately. Preliminary adjusted models tested the same interactions separately but with potential confounding variables included in models.

^cThe final model included variables that were associated with active play in preliminary adjusted models (i.e. where the 95% CI did not intersect with zero), as well as potential confounding variables.

^d‘Full-time’ was defined by LSAC as being at least 30 hours per week.

Discussion

This study identified sociodemographic moderators of change in active play between 10-11y and 12-13y among Australian youth. Active play declined more among girls, those who spoke English at home and those who lived in regional or remote areas. However, the moderating effect of geographical remoteness became marginal in the final model. A widening gap in active play participation was observed by sex between 10-11y and 12-13y, while differences in participation by language spoken at home and geographical remoteness attenuated over time.

Active play participation declined more among girls than boys in the present study. This is consistent with a previous Canadian longitudinal study that also reported a larger decline in active play among girls between 11y and 14y (Cairney et al., 2014). In addition, a previous study based on LSAC data reported a larger decline in non-organised PA among girls between childhood and adolescence (Kemp et al., 2020) and a previous umbrella review reported that girls are likely to have lower PA overall (Aleksovska et al., 2019). The larger decline in active play observed among girls in the present study may be related to a combination of puberty and peer influence (Kemp et al., 2022). Although the present study controlled for the pubertal status of participants, it is possible that the pubertal development of peers may also influence active play. For example, declines in active play participation among earlier-developing girls may have quickly led to similar declines among other girls due to a desire to fit in with peer norms (Kemp et al., 2022). Such a change may not have influenced boys in the same way because peer groups are often stratified by sex during this stage of life and, as previously mentioned, boys often experience puberty later (Petersen et al., 1988).

Participants who spoke English at home also had a larger decline in active play participation between 10-11y and 12-13y in the present study. However, active play participation at 10-11y was higher among youth who spoke English at home, with differences attenuating over time. This attenuation may relate to processes of socialisation that may have been commonly experienced across groups. For example, the transition to high school often results in youth becoming more concerned about fitting in with peers and aware of the presence of older teenagers, which may have led to similar amounts of active play participation in both groups after this transition (Kemp et al., 2022). Despite this, a previous study based on LSAC data revealed contrasting findings, where participation in non-organised PA was shown to decline more among youth from non-English speaking backgrounds (Kemp et al., 2020). However, this study was conducted with a different cohort of youth who were born 4 years earlier than the present cohort (Kemp et al., 2020). It is possible that the specific cultural makeup of youth from non-English speaking backgrounds may have differed between cohorts. A previous systematic review reported a range of additional factors that may affect PA participation among ethnic minority groups, such as the level of English proficiency and the degree of acculturation to participants’ new settings (Langøien et al., 2017). Future studies could investigate these additional factors as potential mediators of the longitudinal relationship between language spoken at home and active play participation.

This study also found that geographical remoteness was a moderator of active play between 10-11y and 12-13y, although this effect was only marginal in the final model. Active play participation was initially higher among youth living in regional or remote areas at 10-11y. This is consistent with a previous Australian cross-sectional study which reported higher participation in free-play PA among youth living in regional and remote areas (9-16y) (Dollman et al., 2012). Potential enablers of active play in regional and remote areas may include the availability of free space and greater perceptions of neighbourhood social cohesion (Dollman et al., 2012). However, preliminary models in the present study showed greater declines in active play participation among youth living in regional and remote areas between 10-11y and 12-13y. Potential barriers to active play that might arise for youth living in these areas during the transition to adolescence may include new responsibilities (e.g., working in a family business or farm) or commuting further to high school (Kemp et al., 2022).

This study could inform the development of intervention strategies to promote active play among certain subgroups of the population. For example, social marketing campaigns or after-school active play programmes could potentially target girls in the transition to adolescence due to the larger decline in active play participation among this group. Data-driven techniques such as cluster analysis could also be used to identify more niche segments of girls to target based on other variables (e.g., attitudinal or behavioural characteristics) (Dietrich, 2017; Dolnicar & Grün, 2017). Interventions could also target other population groups based on the findings of this study, such as youth who speak English at home and those living in regional or remote areas. However, future studies could also explore whether declines in active play participation occurred prior to 10-11y among youth who spoke languages other than English at home and those living in urban areas and, if so, targeted interventions may be needed for such children.

This was the first known study to explore potential moderators of longitudinal changes in active play between childhood and adolescence. It was also the first known study to test longitudinal associations between a number of variables and active play in this age group, including language spoken at home, geographical remoteness and socioeconomic position. However, this study also has some limitations. Active play was measured using self-reported TUDs completed over one day. Self-reported measures have been shown to overestimate PA (Adamo et al., 2009) and the single day of measurement may not account for variations in daily routines. However, promising validity and reliability have been demonstrated by 24-h TUDs (Gershuny et al., 2018; Ridley et al., 2006) and the present analyses controlled for TUD contextual variables (school attendance and season). Additionally, although the population weight used in analyses was considered the best available option, it did not account for non-response beyond Wave 6. LSAC is also a “closed” longitudinal study and no new participants have been included since Wave 1. All participants were either born in Australia or arrived in the country as infants, which may explain the under-representation of youth who spoke languages other than English at home. Finally, this study was delimited by focussing on sociodemographic moderators – other factors may also be related to active play (e.g., neighbourhood characteristics) (Remmers et al., 2014).

Conclusions

This study identified sociodemographic moderators of active play between 10-11y and 12-13y among Australian youth. A widening gap in active play participation was observed by sex between 10-11y and 12-13y, while differences in participation by language spoken at home and geographical remoteness weakened or became marginal over time. Interventions to promote active play could potentially target girls in the transition to adolescence, and future studies might seek to investigate whether active play declines earlier than 10-11y among youth who speak languages other than English at home and those living in urban areas.

Acknowledgments

The authors wish to thank Prof Marijka Batterham at the UOW Statistical Consulting Centre for providing analytic advice. This paper uses unit record data from Growing Up in Australia: the Longitudinal Study of Australian Children (LSAC), conducted by the Australian Government Department of Social Services (DSS). The findings and views reported in this paper, however, are those of the author[s] and should not be attributed to the Australian Government, DSS, or any of DSS’ contractors or partners. DOI: http://dx.doi.org/10.26193/BAA3N6.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed online https://doi.org/10.1080/02640414.2023.2278932.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

Author 1 is the recipient of a Prioritising Emerging Research Leaders (PERL) Fellowship, provided by the University of Wollongong (UOW). This funding body did not influence the conduct or reporting of the research.