Academic Performance and Lifestyle Behaviors in Australian School Children: A Cluster Analysis

Abstract

Poor academic performance has been linked with particular lifestyle behaviors, such as unhealthy diet, short sleep duration, high screen time, and low physical activity. However, little is known about how lifestyle behavior patterns (or combinations of behaviors) contribute to children’s academic performance. We aimed to compare academic performance across clusters of children with common lifestyle behavior patterns. We clustered participants (Australian children aged 9-11 years, n = 284) into four mutually exclusive groups of distinct lifestyle behavior patterns, using the following lifestyle behaviors as cluster inputs: light, moderate, and vigorous physical activity; sedentary behavior and sleep, derived from 24-hour accelerometry; self-reported screen time and diet. Differences in academic performance (measured by a nationally administered standardized test) were detected across the clusters, with scores being lowest in the Junk Food Screenies cluster (unhealthy diet/high screen time) and highest in the Sitters cluster (high nonscreen sedentary behavior/low physical activity). These findings suggest that reduction in screen time and an improved diet may contribute positively to academic performance. While children with high nonscreen sedentary time performed better academically in this study, they also accumulated low levels of physical activity. This warrants further investigation, given the known physical and mental benefits of physical activity.

Keywords

academic performance diet health behavior physical activity screen time sedentary behavior

Education plays a foundational role in the development and prosperity of today’s society (Parsons, 2014). Academic success provides children with higher education and employment opportunities and may lead to socioeconomic advantage, a sense of personal empowerment, and the ability to make healthy choices (Allensworth, 2015; Berg, Willis-Stewart, & Kendall, 2016). Traditionally, girls outperform boys in language-related areas, while boys generally achieve higher in numeracy (Voyer & Voyer, 2014). Nevertheless, the optimization of educational outcomes by modifying potential contributors to boys’ and girls’ academic performance remains a key goal for school systems and parents alike (Grolnick, Raftery-Helmer, & Flamm, 2013).

A considerable body of knowledge suggests that children’s academic performance is influenced by their lifestyle behaviors (Burrows, Goldman, Pursey, & Lim, 2016; Carson et al., 2016; Dewald, Meijer, Oort, Kerkhof, & Bögels, 2010; Trinh, Wong, & Faulkner, 2015). Previous research has examined the links between academic performance and particular lifestyle factors. For example, favorable academic performance is commonly related to longer sleep duration (Chaput et al., 2016), healthy diet (Haapala et al., 2014), and physical activity (PA; Poitras et al., 2016) in school-aged children. The contribution of sedentary behavior is less clear, and appears to depend on the type of activity conducted while sedentary. Academic performance has been negatively associated with screen-based activities such as TV viewing (Busch et al., 2014; Sigman, 2012) and positively associated with reading and doing homework outside of school (Carson et al., 2016).

While the associations between academic performance and individual lifestyle behaviors have been widely studied, the contribution of children’s lifestyles patterns (i.e., the combination of lifestyle behaviors) has rarely been examined. Better academic performance was associated with meeting three to four lifestyle behavior guidelines (for sleep, PA, television viewing, and fruit intake) than meeting none or one guideline in Spanish female adolescents aged 13 to 17 years (Martínez-Gómez et al., 2012). Higher academic achievement was related to the combination of healthy diet, PA, and lower body mass index (BMI) in a study of Icelandic adolescents (aged 14-15 years), with these lifestyle behaviors explaining up to 24% of the variance in academic performance (Sigfúsdóttir, Kristjánsson, & Allegrante, 2007). Another study of Canadian 9- to 12-year-old students found that academic performance was independently related to healthy diet and PA but not to these behaviors in combination (McIsaac, Kirk, & Kuhle, 2015). The relative lack of research on how children’s academic performance is linked to their lifestyles represents a gap in the literature, as lifestyle behaviors do not exist in isolation, but interact and combine to influence a range of health outcomes in children such as weight status (Gubbels, van Assema, & Kremers, 2013; Leech, McNaughton, & Timperio, 2014) and health-related quality of life (Hunt, McKay, Fitzgerald, & Perry, 2014).

The way in which children’s lifestyle behaviors combine has recently been explored through data-driven exploratory methods such as cluster analysis (Gubbels et al., 2013). Children’s lifestyle behaviors have been observed to aggregate into “healthy” (e.g., high PA/low sedentary), “unhealthy” (e.g., low PA/low sleep/poor diet or high screen/poor diet), and “mixed” (e.g., high screen/high PA) patterns (Leech et al., 2014). Few previous cluster analyses have explored children’s lifestyle behaviors comprehensively by including variables such as PA, sedentary behavior, screen time, sleep and diet quality (Leech et al., 2014), and none have considered the time spent in PA, sedentary behavior, and sleep as mutually exclusive and exhaustive components, which are subject to a constant sum restraint (i.e., the time spent in all behaviors must always sum to 24 hours/day). Recent methodological literature recommends the use of compositional data analysis which takes into account that if more time is spent in one behavior (e.g., PA), there must be a correspondingly lower amount of time spent in one or more of the other behaviors (e.g., sedentary behavior or sleep; Chastin, Palarea-Albaladejo, Dontje, & Skelton, 2015; Pedišić, 2014).

This study set out to address this gap in the literature to date. In order to examine the associations between academic performance and clusters of children’s lifestyle behaviors, we aimed to (a) explore how Australian school-aged children cluster in lifestyle behavior groups, using a comprehensive range of behaviors, and applying compositional data analysis techniques and (b) explore the associations between academic performance and membership of lifestyle behavior clusters, with academic performance indicators obtained through a nationwide standardized academic assessment.

Method

Participants

This study involved 9- to 11-year-old children from the Australian arm of the International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE), a large multinational cross-sectional study of approximately 7,000 children across 12 countries. Full methodological details are described in Katzmarzyk et al. (2013).

All government, Catholic, and independent schools in metropolitan Adelaide, South Australia, were allocated a socioeconomic status score based on their “Index of Community Socio-Educational Advantage” (Australian Curriculum Assessment and Reporting Authority, 2013). The schools were then divided into socioeconomic tertiles. Schools were randomly selected from each tertile to ensure each socioeconomic level was equally represented in the sample. If a school declined to participate, another school from the same tertile was randomly selected to replace it. Participating schools did not differ significantly from nonparticipating schools in terms of socioeconomic status (p = .32). Twenty-six schools (43% of invited schools) consented to participate. All Grade 5 children within these schools were invited to participate in the study. The child participation rate was 57%, resulting in a sample of 528 children. Children without complete data for academic performance (N = 179), as well as lifestyle behaviors (N = 52) and sociodemographic covariates (N = 13), were excluded. Five school principals did not consent to the release of academic performance data; however, these schools did not differ socioeconomically from included schools (p = .34). Lifestyle behaviors (activity behaviors, screen time, diet) of children with complete and incomplete data did not differ significantly; therefore, it was decided to exclude participants with missing data rather than perform multiple imputations (Cheema, 2014). The final sample consisted of 284 (131 boys and 153 girls). Data collection occurred between October 2011 and December 2012.

Approval for the Australian protocol was provided by the University of South Australia Human Research Ethics Committee, the South Australian Department of Child Development, and the Catholic Education Department of South Australia. ISCOLE is registered on ClinicalTrials.gov, Identifier: NCT01722500. Parents provided signed informed consent and child assent was attained prior to inclusion in the study.

Materials

Academic Performance

Academic performance was not collected as part of the international ISCOLE protocol; however, at the Australian site, additional consent was requested from school principals and parents to allow access to measures of their child’s academic performance via the Department of Children’s Education and Development. Academic performance was gained from the results of a nationwide standardized, compulsory assessment: the National Assessment Program–Literacy and Numeracy (NAPLAN). Five academic components are evaluated: language (grammar and punctuation), reading, writing, spelling, and numeracy. Scores are standardized to a mean of 500, with a standard deviation of 100, with a higher score representing greater academic achievement. Overall academic performance was calculated as the arithmetic mean of the five component scores.

Use of Time

Daily use of time was measured by 7-day, 24-hour accelerometry. Participants wore an Actigraph GT3X+ accelerometer (ActiGraph, Ft. Walton Beach, FL) on their right hip, with instructions to only remove the device for bathing or water-based activities. An acceptable wear time for valid accelerometry was set at 4 days with at least 10 hours of waking-hours wear time per day, including at least 1 weekend day. Activity data were sampled at a rate of 80 Hz and downloaded in 1-second epochs using the low-frequency extension filter (Acti-Life software version 5.6; Tudor-Locke et al., 2015).

Nocturnal sleep was estimated from activity data processed in 60-second epochs using a fully automated algorithm specifically designed for ISCOLE (Tudor-Locke, Barreira, Schuna, Mire, & Katzmarzyk, 2013). The algorithm was refined to produce more precise estimates of nocturnal sleep duration by identifying and removing periods of nocturnal wakefulness (Barreira et al., 2015). Participants with less than 160 minutes of total sleep period for at least 3 nights were considered invalid. Time spent in daily activity domains was determined once total sleep and waking nonwear time (any sequence of ≤20 consecutive minutes of 0 activity counts) had been removed. Remaining data were processed in 15-second epochs to determine sedentary time (≤25 counts per 15 seconds): light (26-573 counts per 15 seconds), moderate (574-1,002 counts per 15 seconds), and vigorous PA (≥1,003 counts per 15 seconds) congruent with Evenson’s cut-points (Evenson, Catellier, Gill, Ondrak, & McMurray, 2008). Participants’ mean sedentary behavior; light, moderate, and vigorous PA; and sleep duration were calculated for all weekdays and weekend days, and combined to create a daily index for each component, weighted at 5:2 for weekdays–weekend days.

Screen Time

Child participants were asked to self-report time spent on an average weekday and weekend day watching TV and playing computer/video games, by selecting one of the following categories: none, <1 hour, 1 hour, 2 hours, 3 hours, 4 hours, and ≥5 hours. For analysis, a continuous variable was created for both TV time and video/computer time and transformed into a daily index by weighting weekday and weekend screen time at a ratio of 5:2.

Eating Pattern

A Food Frequency Questionnaire of 23 food categories was completed by child participants. Children reported the frequency of consumption of individual food items (categorized as never, less than once per week, once per week, 2-4 days per week, 5-6 days per week, once a day every day, and more than once a day). Two factors were identified by principal component analyses using the Food Frequency Questionnaire as input variables: (a) a healthy eating pattern (positive loadings for vegetables, fruit, whole grains, etc.) and (b) an unhealthy eating pattern (positive loadings for fast food, soft drinks, sweets, etc.; Saloheimo et al., 2015). The factor scores were standardized across the final sample to create “healthy eating pattern” and “unhealthy eating pattern” Z scores, where higher scores represent either a healthier or unhealthier eating pattern, respectively.

Covariates

BMI was derived from anthropometric measures collected at school visits through standardized protocols. Weight was measured using a portable Tanita SC-240 scale (Arlington Heights, IL; Barreira, Staiano, & Katzmarzyk, 2013). Participants’ standing height was measured without shoes, by a Seca 213 portable stadiometer (Hamburg, Germany). Following calculation of BMI for each participant, BMI = weight (kg) per height (m²), a Z score was created using the World Health Organization reference data (de Onis et al., 2007). Parental education was determined by the highest education level reported by either parent, which were collapsed to represent: 1 = less than high school and some high school; 2 = completed high school and some postsecondary (e.g., vocational diploma or certificate); 3 = bachelor degree and postgraduate. Participants’ sex, mother’s weekly employment hours (0, 0 > 15, 15-35, ≥36), the number of parents (≤1 or ≥2) and the number of siblings (0, 1, 2, 3, ≥4) residing in the home were determined from parent survey.

Procedure

The analysis consisted of (a) identification of clusters and (b) relating academic performance to membership of identified clusters.

Cluster analysis, traditionally used in geographical contexts, has become increasingly used to detect groups in population data (Leech et al., 2014). Recent research has used cluster analysis to explore how children group together based on their lifestyle behaviors and subsequent cluster membership has been compared across health indicators (Leech et al., 2014). Clusters can be thought of as groups of individuals, closer together in multidimensional space, like galaxies in the universe (Hair, Black, Babin, Anderson, & Tatham, 2006). Continuous data can be used as cluster inputs, unlike other methods (e.g., latent class analysis), where inputs must be categorized into classes. However, caution must be exercised when data inputs are compositional (Aitchison, 1982). Compositional data are data where a domain (such as time use) is subdivided into a number of subdomains (such as sleep, sitting, and moving), which are exhaustive and mutually exclusive (Aitchison, 1982). These data exist in a constrained geometry (the Simplex) because collectively they must always sum to a constant (i.e., 24 hours; Aitchison, 1982). Accordingly, distance measures used within the clustering algorithm must be consistent with this constrained geometry (i.e., Aitchison distance; Aitchison, Barceló-Vidal, Martín-Fernández, & Pawlowsky-Glahn, 2000). Alternatively, compositional data can be isometrically transferred to the normal Euclidean geometry via a one-to-one log ratio transformation (Egozcue & Pawlowsky-Glahn, 2005). Once compositional data are expressed as isometric log ratio coordinates, they can be treated as regular data, and traditional clustering algorithms using Euclidean distance can be applied (Palarea-Albaladejo, Martín-Fernández, & Soto, 2012).

Cluster analyses were conducted using the Compositions Package (van den Boogaart & Tolosana-Delgado, 2008) for the open source R statistical programming language and software (R Development Core Team, Vienna, Austria). Children’s lifestyle behaviors were used as cluster inputs: Z scores of 24-hour time use (i.e., isometric log ratio coordinates); screen time; and healthy and unhealthy eating pattern scores. Cluster analysis is sensitive to outliers; however, outliers may provide a source of valuable information regarding subgroups in the population (Osei-Bryson & Samoilenko, 2014). To further explore potential outliers in context of the multivariate data structure, a principal components biplot was created (Kynčlová, Filzmoser, & Hron, 2016; see also Supplementary File 1, available online at heb.sagepub.com/supplemental). Ten observations were observed to be outside the 97.5% probability ellipse. To assess the sensitivity of the clustering procedure to the presence of the 10 outliers, a comparison was made between a k-means four-cluster solution (a) before and (b) after removal of the outliers. Comparison between the two solutions identified no significant difference (93% agreement, Cohen’s kappa = 0.91). Accordingly, outliers were retained for subsequent analyses.

A dendrogram was plotted though agglomerative hierarchical clustering using Ward’s method and Euclidean distance (Hair et al., 2006; see also Supplementary File 2, available online heb.sagepub.com/supplemental). Agglomerative hierarchical clustering begins with all individuals as separate entities, and then clusters closest individuals into pairs (Hair et al., 2006). Successively, clusters are combined together with their nearest cluster, until all clusters are incorporated into one group. The dendrogram depicts the agglomerative clustering process and can be used to visualize the number of clusters in a population. Inspection of the dendrogram revealed a four-cluster solution.

Subsequently, a k-means partitioning cluster analysis was used. Partitioning cluster analysis begins with all individuals in one group, which is subsequently divided into two clusters. Clusters are successively divided until a predetermined amount of clusters remain (Hair et al., 2006). A four-cluster solution was predetermined from agglomerative hierarchical clustering; therefore, the k-means analysis was set to identify four clusters. Descriptive statistics were used to characterize the four clusters: 24-hour time use by using compositional means to represent the center of the compositional data points (the geometric mean of each component, adjusted to sum to 1,440 minutes; Martín-Fernández, Daunis-i-Estadella, & Mateu-Figueras, 2015), and variation matrices (see Supplementary File 3, available online at heb.sagepub.com/supplemental) to describe the multivariate spread of the components, as traditionally used univariate standard deviations cannot represent the spread of interdependent components (John Aitchison, 1982). Screen time and diet scores were described using arithmetic means and standard deviations.

Children’s academic performance was compared across the four lifestyle behavior clusters using linear models (Stata/IC 14.0; StataCorp LP). The significance of associations was examined adjusting for covariates (including sex, due to sex differences in children’s academic performance; Voyer & Voyer, 2014) and using Bonferroni corrected p values for multiple comparisons. Linearized standard errors were used to account for the nested sampling design (participants within schools).

Results

Participant Characteristics

Included participants differed slightly from excluded by parental education level and number of siblings, and the academic performance of included participants was consistently higher than excluded participants (see Supplementary File 4, available online at heb.sagepub.com/supplemental).

Cluster Lifestyle Behavior Characteristics

Four lifestyle behavior clusters were identified. Descriptive statistics for each cluster’s sociodemographic characteristics are presented in Table 1. Each cluster’s 24-hour time use, screen time, and dietary behavior are presented in Table 2 and Supplementary File 3 (variation matrices for 24-hour time use components). The clusters can be defined as (a) “Junk Food Screenies”: unhealthiest diet, highest screen time, moderate PA; (b) “All-Rounders”: healthiest diet, lowest screen time, moderate PA; (c) “Actives”: highest PA, lowest sedentary behavior; and (d) “Sitters”: highest sedentary behavior, lowest PA. Sleep duration was poorly differentiated among the clusters, ranging from 571 minutes/day (Sitters), to 579 minutes/day (Actives). The All-Rounders cluster contained the largest proportion of children (30%), while the Junk Food Screenies was the smallest cluster overall (19%). Boys were most likely to be Junk Food Screenies and Actives, whereas girls were most likely to be All-Rounders and Sitters.

Table 1.

Sociodemographic Characteristics of Clusters.

Cluster	Total, %	Sex		Number of siblings, %					Number of parents, %		Highest parental education level,^a %			Mother’s work (hours/week), %				z-BMI, M (SD)
	Total, %	Male %	Female %	Number of siblings, %					Number of parents, %		Highest parental education level,^a %			Mother’s work (hours/week), %
	N = 284	n = 131	n = 153	0	1	2	3	≥4	≤1	≥2	1	2	3	0	0 > 15	15-35	≥36
Junk Food Screenies	19	21	16	4	51	25	15	6	23	77	25	45	30	26	9	42	23	0.67 (1.14)
All-Rounders	30	27	33	3	48	25	18	5	11	89	3	38	59	22	13	39	26	0.35 (1.07)
Actives	24	34	16	6	32	46	10	6	10	90	12	46	43	21	7	54	18	0.41 (0.99)
Sitters	27	18	35	7	49	24	13	8	21	79	20	46	34	28	9	43	20	0.87 (1.32)

Note. z-BMI = body mass index Z-score (World Health Organization).

Parent education levels are as follows: 1 = less than high school and some high school; 2 = completed high school and some postsecondary (e.g., vocational diploma or certificate); 3 = bachelor degree and postgraduate.

Table 2.

Lifestyle Behavior Characteristics of Clusters.

Cluster	24-hour time use (minutes/day)^a					Screen (hours/day), M (SD)	Diet, M (SD)
Cluster	Sleep	SB	LPA	MPA	VPA	Screen (hours/day), M (SD)	Healthy	Unhealthy
Junk Food Screenies	576	483	316	43	21	4.3 (1.6)	0.20 (1.14)	1.35 (1.20)
All-Rounders	577	492	303	44	23	1.6 (1.1)	0.58 (0.80)	−0.52 (0.48)
Actives	579	421	353	56	30	2.4 (1.5)	−0.40 (0.83)	−0.24 (0.55)
Sitters	571	545	286	28	10	2.7 (1.7)	−0.44 (0.88)	−0.13 (0.75)

Note. SB = sedentary behavior; LPA = light physical activity; MPA = moderate physical activity; VPA = vigorous physical activity.

Values presented as compositional mean for time use variables.

Academic Performance

Academic performance differed by cluster membership (Table 3). The highest academic performance was achieved by the All-Rounders in unadjusted models (overall mean = 505); however, following adjustment for covariates, Sitters achieved significantly higher academic performance (overall mean = 504) than Junk Food Screenies (overall mean = 477, effect size [ES] = 0.47; Table 3). The same disparity patterns in academic performance were observed across most NAPLAN components between Junk Food Screenies and Sitters (grammar [ES = 0.50], reading [ES = 0.40], spelling [ES = 0.52]).

Table 3.

Academic Performance by Cluster.

Cluster	Grammar	Reading	Writing	Spelling	Numeracy	Overall
Junk Food Screenies	472 ^a (82)	491 ^a (77)	471 (64)	479 ^a (64)	470^a (60)	477 ^a (57)
All-Rounders	514 (80)	516^b (76)	487 (52)	509 (64)	498 (63)	505 (55)
Actives	497^b (75)	501 ^c (76)	476 (54)	502^b (65)	496 (69)	494^b (54)
Sitters	515 ^{a b} (89)	522 ^{a b c} (78)	478 (51)	515 ^{a b} (72)	488^a (63)	504 ^{a b} (58)

Note. Values are presented as arithmetic mean (standard deviation). Superscript indicates significant difference between pairwise comparisons at alpha = .05 (values with the same superscript are significantly different from each other), and bold indicates maintained significance following Bonferroni correction, that is, alpha = .008. Models are adjusted for sex, parental highest education, mother’s employment hours, number of siblings and number of parents, child body mass index Z score (World Health Organization) and school attended. All analyses were adjusted for potential clustering at the school level.

Discussion

This study aimed to identify clusters of Australian children based on their lifestyle behaviors (24-hour time use, screen time, and diet pattern), and to explore differences in academic performance across these clusters. We found four clusters of distinct combinations of lifestyle behaviors (“Junk Food Screenies”: high screen, unhealthy diet; “All-Rounders”: low screen, healthy diet; “Actives”: high PA, low sedentary behavior; “Sitters”: high sedentary behavior, low PA). Following adjustment for sociodemographic covariates, Sitters achieved the highest academic performance and Junk Food Screenies the lowest.

Considering the univariate findings from previous research, one may have expected that lifestyles of children with poor academic performance would be characterized by a combination of unhealthy diet and high screen behaviors, and a 24-hour time use pattern of short sleep, low PA, and low sedentary time. No such lifestyle cluster was identified among this study’s sample; however, the Junk Food Screenies’ lifestyle is an approximation. Children in this cluster had the unhealthiest diet and longest screen time; however, their levels of sleep, PA, and sedentary behavior were moderate.

Unhealthy diet has consistently been linked with screen time in previous research, possibly due to the influence of advertisements and the distraction away from bodily cues of satiation (Sigman, 2012). Poor academic performance has been associated with both unhealthy diet and screen time in observational research, with a number of mechanisms proposed to explain the deleterious relationships. First, unhealthy diet (low fruit and vegetable intake, high saturated fat intake) may result in a deficiency of various nutrients, which have been associated with cognitive performance (Frisardi et al., 2010; Wu, Ying, & Gomez-Pinilla, 2004). Flavonoids are polyphenolic molecules involved in the production of brain-derived neurotrophic factor, known to be important for synaptic plasticity, neurogenesis, neural survival, learning, and memory (Frisardi et al., 2010; Gómez-Pinilla, 2008; Wu et al., 2004). Dietary intake of docosahexaenoic acid (DHA; one of the omega-3 fatty acids commonly found in fish oils) provides the brain with integral components for neuronal cell membranes (Gómez-Pinilla, 2008). Second, screen time is thought to impede academic performance through displacement of time available for learning opportunities afforded by educational tasks (e.g., reading; Sigman, 2012). Furthermore, screen time (specifically TV viewing) is a passive behavior with little intellectual engagement or mental effort (Shin, 2004). Superficial intellectual processing may stimulate impulsive behaviors and attention deficits, thus hindering academic performance (Shin, 2004). Since unhealthy diet and screen time are both regarded as detrimental to academic performance, it is not surprising that the combination of these two behaviors are linked to poorer academic performance in this study.

As the poorest academic performance was achieved by the Junk Food Screenies’ lifestyle, it would be expected that best academic performance would be achieved by the juxtaposed All-Rounders’ lifestyle (healthiest diet, lowest screen, moderate sleep, PA, and sedentary behavior). Although this was the case, Table 2 reveals that the All-Rounder lifestyle was clearly associated with high socioeconomic status, and after adjustment for sociodemographic factors, the Sitters’ lifestyle was linked with the best academic performance in this study. The difference in average academic performance between Sitters and the Junk Food Screenies was approximately 30 points (ES ~ 0.5). The difference can also be conceptualized using a metric based on year-level equivalence recently developed for the 2013 Australian NAPLAN data (numeracy and reading; Goss & Sonnenmann, 2016). Using this metric, the difference between the Sitters and the Junk Food Screenies cluster was estimated to be 0.56 years for numeracy, and 1.06 years for reading. The association between academic performance and objectively measured sedentary behavior has rarely been studied. One study found no association (Syväoja et al., 2013), and another found a positive association between academic performance and a high sedentary behavior/high PA cohort (Maher et al., 2016). In the present study, no lifestyle clusters characterized by high sedentary behavior and high PA emerged. Rather, high sedentary behavior tended to occur in combination with low PA, and this lifestyle pattern was associated with the best academic performance. This finding suggests that out of the common lifestyle patterns seen in children, spending more time in sedentary—particularly nonscreen pursuits (e.g., reading or doing homework) and less in PA—is associated with the best academic outcomes. Furthermore, the lifestyle cluster characterized by high PA (Actives) was associated with mediocre academic performance. These findings warrant further consideration.

Previous research has found deleterious relationships between sedentary behavior and academic performance (Chinapaw, Proper, Brug, Van Mechelen, & Singh, 2011; Cliff et al., 2016; Tremblay et al., 2011). However, there is some evidence that particular forms of sedentary behavior are associated with better academic performance (e.g., reading books, doing homework; Carson et al., 2016); therefore, the current study’s findings are not altogether surprising. The current study’s findings regarding PA are more difficult to reconcile in light of existing literature linking high PA with better cognition. For example, a meta-analysis reported a positive relationship between PA and cognition (categorized as perceptual skills, intelligence quotient, achievement, verbal tests, mathematics tests, memory, developmental level/academic readiness, and other) in school-aged children (Sibley & Etnier, 2003). A positive relationship was found for each category of cognition, except memory, which was unrelated (Sibley & Etnier, 2003). Neuroanatomical research has shown that PA, and especially fitness, activate the frontoparietal cortex (Colcombe et al., 2004; Colcombe et al., 2006; Marks et al., 2007), a region of the brain also related to reading comprehension (Maguire, Frith, & Morris, 1999) and mathematical operations (Ansari & Dhital, 2006; Göbel, Johansen-Berg, Behrens, & Rushworth, 2004; Rivera, Reiss, Eckert, & Menon, 2005) in children. Furthermore, rodent-based models have demonstrated PA to selectively enhance angiogenesis, synaptogenesis, and neurogenesis, and to stimulate the production of neurotropic factors within the brain (Van Praag, Christie, Sejnowski, & Gage, 1999). In the present study, it could be hypothesized that, rather than high PA jeopardizing the academic performance of the Actives cluster, insufficient nonscreen sedentary behavior (Actives had the lowest sedentary behavior, yet still accrued over 2 hours of screen time/day) may be limiting their academic performance.

Sleep duration did not emerge as a defining feature of any lifestyle identified by the clustering algorithm applied in the present study. Similarly, sleep duration did not feature as an important characteristic in our recent cluster analysis using the full international ISCOLE data set of >7,000 children (Dumuid et al., 2016). Sleep was relatively similar among the clusters, with only 8 minutes/day difference between the clusters with highest and lowest sleep duration, possibly due to parental influence over bedtimes in this age group. A number of previous cluster analyses of children’s lifestyle behaviors have identified high or low sleep clusters (Fernández-Alvira et al., 2013; Magee, Caputi, & Iverson, 2013); however, these studies did not use compositional data analysis methods. Because sleep duration was relatively unimportant in the present study’s cluster solution, our results cannot provide clear evidence regarding the contribution of sleep duration to the relationship between academic performance and children’s lifestyles.

This study had a number of strengths. In particular, children’s lifestyles were studied comprehensively, taking into account objectively measured 24-hour time use: sedentary behavior, all intensities of PA, sleep duration, as well as their dietary intake and screen time. Academic performance was collected through a standardized national assessment, minimizing the risk of bias. Statistical analyses incorporated compositional data analysis techniques to accommodate the closed nature of 24-hour time use. A relatively large sample was included, potential sociodemographic confounders were adjusted for in the regression models and potential clustering by schools was accounted for.

The limitations of this study must also be acknowledged. Causation cannot be inferred due to the study’s cross-sectional design. The screen time measure relied on child self-report and did not specifically capture screen time accrued on increasingly popular mobile devices such as tablets and cell phones. Activity was measured using waist-worn accelerometers, which may not perfectly distinguish between sitting and standing, and have been reported to cause the overestimation of sleep duration (Hjorth et al., 2012). It is possible that different lifestyle behavior clusters may have been identified if wrist-worn accelerometers were used. Analyses were not sex-specific, with the combination of boys’ and girls’ overall academic performance potentially introducing bias due to sex differences in specific academic areas (e.g., girls have been reported to achieve higher scores than boys in language-based assessments; Voyer & Voyer, 2014). Furthermore, it is possible that lifestyle clusters, as well as the relationships between academic performance and these clusters, may have differed on the basis of sex. It is worth noting, however, that our recent analyses regarding clustering of lifestyle patterns using the full international ISCOLE data set found that lifestyle clusters were remarkably consistent between boys and girls and also between countries (Dumuid et al., 2017). Finally, the generalizability of the findings is unclear; the study had moderate participation rates from schools (43%) and children (57%), potentially creating bias. Further bias may have been introduced by the listwise deletion of subjects with missing data, particularly as excluded children consistently had poorer academic performance than included children. It is also important to acknowledge that the study participants were from a single Australian city; children from different cities or rural settings may have different lifestyle behaviors.

In summary, four distinct lifestyle behavior clusters were identified (Junk Food Screenies, All-Rounders, Actives, and Sitters). Academic performance was poorest in the Junk Food Screenies cluster and best in the Sitters cluster. Taken together, these findings suggest the minimization of screen time and promotion of healthy diet behaviors should be encouraged to promote academic performance. Furthermore, it appears that sitting (high nonscreen sedentary time) may be of benefit to academic performance. It is of concern that high sedentary behavior coexisted with low PA, given the substantial body of previous research linking each of these behaviors to a range of deleterious health measures. While Sitters performed the best academically in this study, they also had the highest z-BMI of all clusters. Interventions encouraging physically active study breaks may be warranted to prevent future health complications for high achieving Sitters. Future studies should investigate the change and stability of children’s lifestyle behavior clusters and associations with academic performance over time, through longitudinal studies.

Supplemental Material

sj-docx-1-heb-10.1177_1090198117699508 – Supplemental material for Academic Performance and Lifestyle Behaviors in Australian School Children: A Cluster Analysis

Supplemental material, sj-docx-1-heb-10.1177_1090198117699508 for Academic Performance and Lifestyle Behaviors in Australian School Children: A Cluster Analysis by Dorothea Dumuid, Timothy Olds, Josep-Antoni Martín-Fernández, Lucy K. Lewis, Leah Cassidy and Carol Maher in Health Education & Behavior

Footnotes

Acknowledgements

We wish to thank the Australian participants of the International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE) and their families who made this study possible. In addition, we would like to acknowledge Peter T. Katzmarzyk and Timothy S. Church, who conceptualized and designed ISCOLE.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Dorothea Dumuid is supported by an Australian Government Research Training Program Scholarship. Martín-Fernández is partially supported by the Spanish Ministry of Economic Competitiveness under the project CODA-RETOS (MTM2015-65016-C2-1(2)-R). Carol Maher is the recipient of a postdoctoral fellowship award from the Australian National Heart Foundation (100188). ISCOLE was funded by The Coca-Cola Company. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of this article.

Supplementary Material

Additional supporting information is available at .

References

Aitchison

(1982). The statistical analysis of compositional data. Journal of the Royal Statistical Society. Series B (Methodological), 44, 139-177.

Aitchison

Barceló-Vidal

Martín-Fernández

Pawlowsky-Glahn

(2000). Logratio analysis and compositional distance. Mathematical Geology, 32, 271-275.

Allensworth

D. D.

(2015). Health services and health education. In Bosworth

(Ed.), Prevention science in school settings (pp. 105-121). New York, NY: Springer.

Ansari

Dhital

(2006). Age-related changes in the activation of the intraparietal sulcus during nonsymbolic magnitude processing: An event-related functional magnetic resonance imaging study. Journal of Cognitive Neuroscience, 18, 1820-1828.

Australian Curriculum Assessment and Reporting Authority. (2013). My school: Guide to understanding 2012 Index of Community Socio-Educational Advantage (ICSEA) values. Retrieved from http://www.acara.edu.au/verve/_resources/About_icsea_2014.pdf

Barreira

T. V.

Schuna

J. M.

Jr. Mire

E. F.

Katzmarzyk

P. T.

Chaput

J.-P.

Leduc

Tudor-Locke

(2015). Identifying children’s nocturnal sleep using 24-h waist accelerometry. Medicine & Science in Sports & Exercise, 47, 937-943.

Barreira

T. V.

Staiano

A. E.

Katzmarzyk

P. T.

(2013). Validity assessment of a portable bioimpedance scale to estimate body fat percentage in White and African–American children and adolescents. Pediatric Obesity, 8(2), e29-e32.

Berg

Willis-Stewart

Kendall

(2016). Creating healthy schools and student engagement: The Got Health? Initiative. Revue phénEPS/PHEnex Journal, 8(1). Retrieved from http://ojs.acadiau.ca/index.php/phenex/article/view/1618

Burrows

Goldman

Pursey

Lim

(2016). Is there an association between dietary intake and academic achievement: a systematic review. Journal of Human Nutrition and Dietetics. Advance online publication. doi:10.1111/jhn.12407

10.

Busch

Loyen

Lodder

Schrijvers

A. J.

van Yperen

T. A.

de Leeuw

J. R.

(2014). The effects of adolescent health-related behavior on academic performance: A systematic review of the longitudinal evidence. Review of Educational Research, 84, 245-274.

11.

Carson

Hunter

Kuzik

Gray

C. E.

Poitras

V. J.

Chaput

J.-P.

. . . Connor Gorber

(2016). Systematic review of sedentary behaviour and health indicators in school-aged children and youth: An update. Applied Physiology, Nutrition, and Metabolism, 41, S240-S265.

12.

Chaput

J.-P.

Gray

C. E.

Poitras

V. J.

Carson

Gruber

Olds

. . . Sampson

(2016). Systematic review of the relationships between sleep duration and health indicators in school-aged children and youth. Applied Physiology, Nutrition, and Metabolism, 41, S266-S282.

13.

Chastin

S. F.

Palarea-Albaladejo

Dontje

M. L.

Skelton

D. A.

(2015). Combined effects of time spent in physical activity, sedentary behaviors and sleep on obesity and cardio-metabolic health markers: A novel compositional data analysis approach. PLoS One, 10(10), e0139984.

14.

Cheema

J. R.

(2014). Some general guidelines for choosing missing data handling methods in educational research. Journal of Modern Applied Statistical Methods, 13(2), 3.

15.

Chinapaw

Proper

Brug

Van Mechelen

Singh

(2011). Relationship between young peoples’ sedentary behaviour and biomedical health indicators: A systematic review of prospective studies. Obesity Reviews, 12, e621-e632.

16.

Cliff

Hesketh

Vella

Hinkley

Tsiros

Ridgers

. . . Hardy

(2016). Objectively measured sedentary behaviour and health and development in children and adolescents: Systematic review and meta-analysis. Obesity Reviews, 17, 330-344.

17.

Colcombe

S. J.

Erickson

K. I.

Scalf

P. E.

Kim

J. S.

Prakash

McAuley

. . . Kramer

A. F.

(2006). Aerobic exercise training increases brain volume in aging humans. Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 61, 1166-1170.

18.

Colcombe

S. J.

Kramer

A. F.

Erickson

K. I.

Scalf

McAuley

Cohen

N. J.

. . . Elavsky

(2004). Cardiovascular fitness, cortical plasticity, and aging. Proceedings of the National Academy of Sciences of the United States of America, 101, 3316-3321.

19.

de Onis

Onyango

Borghi

Siyam

Nishida

Siekmann

. (2007). Development of a WHO growth reference for school-aged children and adolescents. Bulletin of the World Health Organization, 85, 660-667.

20.

Dewald

J. F.

Meijer

A. M.

Oort

F. J.

Kerkhof

G. A.

Bögels

S. M.

(2010). The influence of sleep quality, sleep duration and sleepiness on school performance in children and adolescents: A meta-analytic review. Sleep Medicine Reviews, 14, 179-189.

21.

Dumuid

Olds

Lewis

Martín-Fernández

Katzmarzyk

P. T.

Barreira

T. V.

. . . Maher

(2016). The adiposity of children is associated with their lifestyle behaviours: A cluster analysis of school-aged children from 12 nations. Pediatric Obesity. Advance online publication. doi:10.1111/ijpo.12196

22.

Dumuid

Olds

Lewis

L. K.

Martin-Fernandez

Katzmarzyk

P. T.

Barreira

T. V.

. . . Maher

(2017). Health-related quality of life and lifestyle behavior clusters in school-aged children from 12 countries. The Journal of Pediatrics, 183, 78–183.e2

23.

Egozcue

J. J.

Pawlowsky-Glahn

(2005). Groups of parts and their balances in compositional data analysis. Mathematical Geology, 37, 795-828.

24.

Evenson

K. R.

Catellier

D. J.

Gill

Ondrak

K. S.

McMurray

R. G.

(2008). Calibration of two objective measures of physical activity for children. Journal of Sports Sciences, 26, 1557-1565.

25.

Fernández-Alvira

J. M.

De Bourdeaudhuij

Singh

A. S.

Vik

F. N.

Manios

Kovacs

. . . Moreno

L. A.

(2013). Clustering of energy balance-related behaviors and parental education in European children: The ENERGY-project. International Journal of Behavioral Nutrition and Physical Activity, 10, 5. doi:10.1186/1479-5868-10-5

26.

Frisardi

Panza

Seripa

Imbimbo

B. P.

Vendemiale

Pilotto

Solfrizzi

(2010). Nutraceutical properties of Mediterranean diet and cognitive decline: Possible underlying mechanisms. Journal of Alzheimer’s Disease, 22, 715-740.

27.

Göbel

S. M.

Johansen-Berg

Behrens

Rushworth

M. F.

(2004). Response-selection-related parietal activation during number comparison. Journal of Cognitive Neuroscience, 16, 1536-1551.

28.

Gómez-Pinilla

(2008). Brain foods: The effects of nutrients on brain function. Nature Reviews Neuroscience, 9, 568-578.

29.

Goss

Sonnemann

(2016). Widening gaps: What NAPLAN tells us about student progress. Retrieved from https://grattan.edu.au/wp-content/uploads/2016/03/937-Widening-gaps.pdf

30.

Grolnick

W. S.

Raftery-Helmer

J. N.

Flamm

E. S.

(2013). Parent involvement in learning. In Hattie

Anderman

(Eds.), International guide to student achievement (pp. 101-103). New York, NY: Routledge.

31.

Gubbels

J. S.

van Assema

Kremers

S. P.

(2013). Physical activity, sedentary behavior, and dietary patterns among children. Current Nutrition Reports, 2, 105-112.

32.

Haapala

E. A.

Poikkeus

A.-M.

Kukkonen-Harjula

Tompuri

Lintu

Väistö

. . . Lakka

T. A.

(2014). Associations of physical activity and sedentary behavior with academic skills: A follow-up study among primary school children. PLoS One, 9(9), e107031.

33.

Hair

J. F.

Black

W. C.

Babin

B. J.

Anderson

R. E.

Tatham

R. L.

(2006). Multivariate data analysis (6th ed.). Upper Saddle River, NJ: Pearson/Prentice Hall.

34.

Hjorth

M. F.

Chaput

J. P.

Damsgaard

C. T.

Dalskov

S. M.

Michaelsen

K. F.

Tetens

Sjödin

(2012). Measure of sleep and physical activity by a single accelerometer: Can a waist-worn Actigraph adequately measure sleep in children? Sleep and Biological Rhythms, 10, 328-335.

35.

Hunt

McKay

Fitzgerald

Perry

(2014). Time use and daily activities of late adolescents in contemporary Ireland. Journal of Occupational Science, 21, 42-64.

36.

Katzmarzyk

P. T.

Barreira

T. V.

Broyles

S. T.

Champagne

C. M.

Chaput

J.-P.

Fogelholm

. . . Kurpad

(2013). The International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE): Design and methods. BMC Public Health, 13(1), 900.

37.

Kynčlová

Filzmoser

Hron

(2016). Compositional biplots including external non-compositional variables. Statistics, 50, 1132-1148.

38.

Leech

R. M.

McNaughton

S. A.

Timperio

(2014). The clustering of diet, physical activity and sedentary behavior in children and adolescents: A review. International Journal of Behavioral Nutrition and Physical Activity, 11, 4. doi:10.1186/1479-5868-11-4

39.

Magee

C. A.

Caputi

Iverson

D. C.

(2013). Patterns of health behaviours predict obesity in Australian children. Journal of Paediatrics and Child Health, 49, 291-296.

40.

Maguire

E. A.

Frith

C. D.

Morris

(1999). The functional neuroanatomy of comprehension and memory: The importance of prior knowledge. Brain, 122, 1839-1850.

41.

Maher

Lewis

Katzmarzyk

P. T.

Dumuid

Cassidy

Olds

(2016). The associations between physical activity, sedentary behaviour and academic performance. Journal of Science and Medicine in Sport, 19, 1004-1009.

42.

Marks

B. L.

Madden

D. J.

Bucur

Provenzale

J. M.

White

L. E.

Cabeza

Huettel

S. A.

(2007). Role of aerobic fitness and aging on cerebral white matter integrity. Annals of the New York Academy of Sciences, 1097, 171-174.

43.

Martín-Fernández

J.-A.

Daunis-i-Estadella

Mateu-Figueras

(2015). On the interpretation of differences between groups for compositional data. SORT—Statistics and Operations Research Transactions, 39, 231-252.

44.

Martínez-Gómez

Veiga

Gómez-Martínez

Zapatera

Martínez-Hernández

Calle

M. E.

Marcos

(2012). Gender-specific influence of health behaviors on academic performance in Spanish adolescents: The AFINOS study. Nutricion Hospitalaria, 27, 724-730.

45.

McIsaac

J.-L. D.

Kirk

S. F.

Kuhle

(2015). The association between health behaviours and academic performance in Canadian elementary school students: A cross-sectional study. International Journal of Environmental Research and Public Health, 12, 14857-14871.

46.

Osei-Bryson

K.-M.

Samoilenko

(2014). Overview on cluster analysis. In Osei-Bryson

K.-M.

Samoilenko

(Eds.), Advances in research methods for information systems research (pp. 127-138). New York, NY: Springer.

47.

Palarea-Albaladejo

Martín-Fernández

J. A.

Soto

J. A.

(2012). Dealing with distances and transformations for fuzzy C-means clustering of compositional data. Journal of Classification, 29, 144-169.

48.

Parsons

(2014). The school class as a social system. In Ballantine

Spade

(Eds.), Schools and society: A sociological approach to education. (pp. 35-40). Thousand Oaks, CA: SAGE Publications.

49.

Pedišić

. (2014). Measurement issues and poor adjustments for physical activity and sleep undermine sedentary behaviour research—The focus should shift to the balance between sleep, sedentary behaviour, standing and activity. Kinesiology, 46, 135-146.

50.

Poitras

V. J.

Gray

C. E.

Borghese

M. M.

Carson

Chaput

J.-P.

Janssen

. . . Kho

M. E.

(2016). Systematic review of the relationships between objectively measured physical activity and health indicators in school-aged children and youth. Applied Physiology, Nutrition, and Metabolism, 41, S197-S239.

51.

Rivera

S. M.

Reiss

Eckert

M. A.

Menon

(2005). Developmental changes in mental arithmetic: Evidence for increased functional specialization in the left inferior parietal cortex. Cerebral Cortex, 15, 1779-1790.

52.

Saloheimo

González

Erkkola

Milauskas

Meisel

Champagne

. . . Fogelholm

(2015). The reliability and validity of a short food frequency questionnaire among 9–11-year olds: A multinational study on three middle-income and high-income countries. International Journal of Obesity Supplements, 5, S22-S28.

53.

Shin

(2004). Exploring pathways from television viewing to academic achievement in school age children. Journal of Genetic Psychology, 165, 367-382.

54.

Sibley

B. A.

Etnier

J. L.

(2003). The relationship between physical activity and cognition in children: A meta-analysis. Pediatric Exercise Science, 15, 243-256.

55.

Sigfúsdóttir

I. D.

Kristjánsson

A. L.

Allegrante

J. P.

(2007). Health behaviour and academic achievement in Icelandic school children. Health Education Research, 22, 70-80.

56.

Sigman

(2012). Time for a view on screen time. Archives of Disease in Childhood, 97, 935-942.

57.

Syväoja

Kantomaa

M. T.

Ahonen

Hakonen

Kankaanpää

Tammelin

T. H.

(2013). Physical activity, sedentary behavior, and academic performance in Finnish children. Medicine & Science in Sports & Exercise, 45, 2098-2104.

58.

Tremblay

M. S.

LeBlanc

A. G.

Kho

M. E.

Saunders

T. J.

Larouche

Colley

R. C.

. . . Gorber

S. C.

(2011). Systematic review of sedentary behaviour and health indicators in school-aged children and youth. International Journal of Behavioral Nutrition and Physical Activity, 8(1), 98.

59.

Trinh

Wong

Faulkner

G. E.

(2015). The independent and interactive associations of screen time and physical activity on mental health, school connectedness and academic achievement among a population-based sample of youth. Journal of the Canadian Academy of Child and Adolescent Psychiatry, 24, 17-24.

60.

Tudor-Locke

Barreira

T. V.

Schuna

J. M.

Mire

E. F.

Chaput

J.-P.

Fogelholm

. . . Lambert

E. V.

(2015). Improving wear time compliance with a 24-hour waist-worn accelerometer protocol in the International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE). International Journal of Behavioral Nutrition and Physical Activity, 12(1), 11.

61.

Tudor-Locke

Barreira

T. V.

Schuna

J. M.

Jr. Mire

E. F.

Katzmarzyk

P. T.

(2013). Fully automated waist-worn accelerometer algorithm for detecting children’s sleep-period time separate from 24-h physical activity or sedentary behaviors. Applied Physiology, Nutrition, and Metabolism, 39, 53-57.

62.

van den Boogaart

K. G.

Tolosana-Delgado

. (2008). “Compositions”: A unified R package to analyze compositional data. Computers & Geosciences, 34, 320-338.

63.

Van Praag

Christie

B. R.

Sejnowski

T. J.

Gage

F. H

. (1999). Running enhances neurogenesis, learning, and long-term potentiation in mice. Proceedings of the National Academy of Sciences, 96, 13427-13431.

64.

Voyer

S. D.

(2014). Gender differences in scholastic achievement: A meta-analysis. Psychological Bulletin, 140, 1174-1204.

65.

Ying

Gomez-Pinilla

(2004). The interplay between oxidative stress and brain-derived neurotrophic factor modulates the outcome of a saturated fat diet on synaptic plasticity and cognition. European Journal of Neuroscience, 19, 1699-1707.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.40 MB