Correlates of Objectively Measured Physical Activity in Obese Children




The aim of this study was to identify potential correlates of objectively measured physical activity in a sample of obese children. A cross-sectional design was used to assess 137 5–9-year-old obese children (mean ± s.d. age = 8.3 ± 1.1 years; mean BMI z-score = 2.76 ± 0.70; 58% girls) from two regional cities in New South Wales, Australia, before commencement in a treatment trial. Correlates examined included age, BMI z, parental BMI, perceived competence, health-related quality of life, daily minutes spent in small screen recreation (SSR), and fundamental motor skill (FMS) proficiency. Physical activity was assessed using accelerometers and values were calculated for % of monitored time spent in moderate- (MPA) and vigorous (VPA)-intensity physical activity and mean counts per minute (CPM). Analyses were conducted separately for boys and girls. Motor skill proficiency was significantly correlated with a number of physical activity variables for boys and girls. For boys, regression analysis revealed object-control proficiency predicted CPM (R2 = 0.25) and age was a predictor of %MPA (R2 = 0.56). Age and object-control skill proficiency were salient predictors of %VPA (R2 = 0.34). For girls, age and daily minutes of SSR were the only significant predictors for CPM (R2 = 0.13). Age was the sole predictor of %MPA (R2 = 0.38) and %VPA (R2 = 0.15). The targeting of FMSs at an early age should be tested in experimental studies as potential strategies to increase physical activity among obese children, particularly for boys. Interventions aimed at reducing sedentary behaviors among obese girls should also be considered.


Over the past 30 years, childhood obesity has increased significantly and is now considered a serious public health issue with both short- and long-term health consequences (1). Increasing physical activity is recognized as a key strategy for the treatment of pediatric obesity (2), as studies have demonstrated a strong negative relationship between physical activity and obesity (3). Moreover, regular participation in physical activity is associated with numerous physiological and psychological health benefits for children (4). However, obese children are generally considered to be less active than their normal-weight peers (5).

A range of factors may influence the complex behavior of physical activity, but not all have been studied extensively with obese children. Competence Motivation Theory (6) posits that children's motivation to participate in physical activity is influenced by their fundamental motor skill (FMS) proficiency, perceived physical competence, social support, and enjoyment of physical activity. However, these relationships have not been explored in an obese sample despite studies finding that children may not be sufficiently active as they have poor levels of FMS proficiency and perceived competence (7,8). Our current knowledge of the influence of FMS proficiency and key psychosocial variables on physical activity in obese children is limited.

In addition, our understanding of potential predictors of physical activity behavior for obese children is quite weak, largely due to an over-reliance on self-report measures. Previous studies of healthy-weight children have measured physical activity using self-report that are limited by recall issues and have low levels of reliability and validity particularly among young children (9). This is seen as a major limitation of the vast majority of correlates studies (10) as the correlates of physical activity may vary depending on the type of measurement employed (11). The use of an objective measure of physical activity may solve some of these measurement problems (5).

There is a critical need to better understand the factors that are associated with physical activity among obese children to inform the design and delivery of intervention programs that promote physical activity. Intervention design must be based on an understanding and application of the correlates of physical activity among specific populations such as obese children (4). Research in this area is urgently needed as current treatment strategies for obese children are invariably unsuccessful (12).

The purpose of our study was to assess a range of potential correlates of accelerometer-assessed physical activity in a sample of obese children prior to participation in a treatment trial. We examined selected biological (sex, age, and parental BMI), psychological (perceived competence and health-related quality of life), and behavioral attributes (motor skill proficiency and sedentary behavior) that may influence physical activity. This is the first study to examine the relationship between these key psychological and behavioral attributes and objectively measured physical activity among obese children. A subsidiary purpose was to examine sex and physical activity-intensity differences in potential correlates. Knowledge of sex and physical activity intensity-specific correlates would provide insight into potential targets for experimental studies designed for obese children.

Methods and Procedures


A total of 165 families with overweight or obese children aged between 5 and 9 years were recruited, as part of the Hunter Illawarra Kids Challenge Using Parent Support (HIKCUPS) trial (13). HIKCUPS is an ongoing multisite randomized controlled trial in overweight/obese children comparing the efficacy of three interventions: (i) a parent-centered dietary modification program, (ii) a child-centered physical activity skill-development program, and (iii) a program combining both 1 and 2 above (13). Children aged 5–9 years were selected as they had not yet experienced a pubertal growth spurt (which may deflate their BMI score), were postadiposity rebound, and at an appropriate age to develop their motor skills (14). Participants were recruited through a range of strategies including notices in school newsletters, posters placed in general practitioner surgeries, and paid advertisements in local newspapers. The HIKCUPS study and recruitment strategies have been described in detail elsewhere (13,15). Baseline data were used in the present study for those children who had sufficient physical activity data (n = 137). The sample included 58% girls (79/137) and 77% (105/137) were categorized as obese for their age and sex. The sample has a mean ± s.d. age of 8.3 ± 1.1 years and a mean BMI z-score of 2.76 ± 0.70.

Eligibility was determined using a standardized telephone-screening script, which addressed specific inclusion/exclusion criteria. Participants were classified as overweight or obese as defined by BMI according to International Obesity Task Force cutoff points for age and sex (16). Children with extreme obesity (BMI z-score >4.0), known syndromal causes of obesity, medications known to be associated with weight gain, chronic illness, or children who had started puberty (parent reported—Tanner staging) or subject to significant dietary restriction were excluded. Prior to participation in the study, written informed consent was obtained from parents. The study was approved by both the University of Newcastle and University of Wollongong Human Ethics Committees.

Measurement of physical activity

Each child's physical activity was objectively measured for eight consecutive days using Manufacturing Technology Industries (MTI Health Services, Fort Walton Beach, FL) Actigraph 7164 accelerometers. Children were recruited as part of a treatment trial and four different cohorts were required. Cohort 1 and 4 were measured in April (Autumn) in consecutive years while Cohort 2 (July-Winter) and Cohort 3 (September-Spring) were measured at different times of the year. The Actigraph 7164 accelerometer is considered to be a reliable and valid measure of physical activity in children (17). Accelerometers were attached to adjustable elastic belts and worn over the right hip. Physical activity was recorded in 1 min epochs. Strings of “0” counts in bouts ≥20 min were considered times of nonmonitoring and were subtracted from the total minutes monitored during data reduction (18). The average number of minutes that children wore the accelerometers and the number of activity counts per minute (CPM) were calculated. Mean CPM as a summary measure of total physical activity in children has been validated against doubly labeled water (19). Minute-by-minute activity counts were uploaded to determine the amount of time spent in moderate (MPA) (3–5.9 METS) and vigorous (VPA) (≥6 METS) activity during each 60-min segment of the monitoring period. Age-specific count ranges relating to the above intensity levels were based on prediction equations for energy expenditure (20).

Values were calculated for percentage of monitored time spent in MPA (%MPA) and VPA (%VPA) to account for variation in time spent wearing monitors. Participant data were included in analyses if accelerometers were worn for ≥600 min on ≥4 days (21). Physical activity measurements were able to be calculated for 137 of the 165 participants after exclusion of participants with insufficient data. There were no significant differences between results for participants with sufficient physical activity data (n = 137) and those without (n = 28) for any of the study variables. As recommended by Trost et al. (20), we estimated the contribution of times when monitors were not worn (e.g., swimming and cycling) to total physical activity through examination of self-reported log sheets completed by families. However, no significant changes in any of the physical activity variables were found after adjustment and so were excluded from further analyses.

Assessment of physical activity correlates

Adiposity. Weight was measured to 0.1 kg with the children barefoot and wearing light clothing using Tanita HD646 scales (Tanita Corporation of America, Arlington Heights, IL). Height was measured to 0.1 cm using the stretch-stature method and PE87 portable stadiometers (Mentone Educational Centre, Victoria, Australia). Assessors were trained by the same experienced researcher and used the protocol prescribed by the International Society for the Advancement of Kinanthropometry (22). Weight and height were used to calculate BMI (kg/m2) and age- and sex-adjusted standardized scores (z-scores) based upon the UK 1990 reference charts (23).

Parental BMI. BMI was calculated from parents' self-reported height and weight. To improve the quality of the self-report data, parents were sent a letter requesting their height and weight. As recommended by Flood et al. (24), validity of self-report BMI data is improved if participants are given notice and information is not collected over the telephone or in person. There were no differences in physical activity participation for participants who had complete parent BMI data and those without.

Perceived competence. Perceived competence was assessed among 8–9 year olds using the Self-Perception Profile for Children (25) and among 5–7 year olds using the Pictorial Scale of Perceived Competence and Social Acceptance for Young Children (26), which both have established validity and reliability (26,27). The older version assesses six domains (athletic, scholastic, physical, social, behavioral, and global self-worth) while four domains are measured in the younger version (cognitive, peer acceptance, physical, and maternal acceptance), which is considered to be a downward extension of the older version. Only the common domains from both versions were used to allow analysis of all children together and maximize statistical power. These included physical competence (which comprised the domain of athletic competence for older children and physical competence for younger children), social acceptance (social acceptance for older children and peer acceptance for younger children), and scholastic competence (scholastic competence for older children and cognitive competence for younger children). Internal consistency reliability coefficients for the common domains for both older and younger children were considered acceptable: Older version Cronbach's alpha scores were scholastic (0.76), athletic (0.71), and social (0.74). For younger children, alphas were cognitive (0.81), physical competence (0.62), and peer (0.86). The appropriate questionnaire was administered to each child individually by trained assessors.

Health-related quality of life. Health-related quality of life was assessed by child and parent proxy using the Pediatric Quality of Life Inventory (PedsQL) version 4.0. Children were read each item by a research assistant. Parents completed the parent-proxy version. The questionnaire comprises four domains including physical functioning, social functioning, school functioning, and emotional functioning. Recently published data indicate high levels of internal consistency for both the self-report and parent-proxy report in this age group (28). However, reliability scores (Cronbach's alpha) for the four domains were generally low in our sample for children—physical functioning (0.61), emotional functioning (0.54), social functioning (0.67), school functioning (0.49), and of moderate strength for parents—physical functioning (0.85), emotional functioning (0.77), social functioning (0.80), school functioning (0.77). All domains, with the exception of physical functioning, were combined to create a psychological health scale.

FMS proficiency. FMS proficiency was video assessed by a single assessor using the Test of Gross Motor Development, 2nd edn. (TGMD-2) (29). This battery involves two subtests, a locomotor subtest (run, gallop, hop, leap, horizontal jump, and slide) and an object-control subtest (strike, dribble, catch, kick, overhand throw, and underhand roll). Children were given a standardized visual demonstration of the correct technique for performing the skill before each test, but were not told what components were being assessed. Children were asked to perform two trials of the skill. A score out of two was recorded for each skill component and summed to give a total skill score. Skill scores were then combined to form the two subtest total raw scores. Finally, the two subtest scores were converted to standard scores based on each child's sex and age, and these standard scores were summed and then standardized to give a gross motor quotient (GMQ) (29). In this study, inter- and intra-observer reliability tests were conducted with 38 randomly selected children. Assessor evaluations were compared to an independent criterion assessor's evaluation of children's motor skill proficiency. Reliability coefficients were: GMQ (Inter ICC = 0.91; Intra ICC = 0.89), locomotor (Inter ICC = 0.92; Intra ICC = 0.90), and object-control (Inter ICC = 0.81; Intra ICC = 0.80).

Sedentary behavior. Parents proxy-reported their child's sedentary behavior using the Children's Leisure Activities Study Survey which has been validated in Australian children (30). The Children's Leisure Activities Study Survey instructs parents to think of a “typical” week and report the time their child spends in a range of sedentary activities. For the purpose of this study, only time spent in small screen recreation (SSR) activities (watching television, DVDs, playing computer games, and using the Internet for fun) were used. Parents were asked to indicate the total hours/minutes their child spent engaged in SSR over a typical week (including weekends).


Data analysis was undertaken using SPSS version 14.0. Prior to analysis, normality and equal variance of the data were assessed using a Kolmogrov-Smirnov test (with Lillefors' correction) and Leven median test, respectively. Means and standard deviations were calculated for all normally distributed variables. Medians and interquartile ranges were calculated for variables that were not normally distributed. Percent time spent in VPA was transformed (square root) to satisfy normality criteria for both boys and girls. Crude associations between %MPA, %VPA, and mean CPM were assessed using Pearson's Product-Moment correlations. Due to widely recognized sex differences in physical activity levels and correlates, separate analyses were performed for boys and girls.

To establish whether potential correlates were independently associated with the physical activity variables, a series of sex-specific multiple regression analyses were conducted using the backward elimination method. As previous studies have shown correlates of physical activity are different depending on intensity (31), variables with significant zero-order correlations (P < 0.05) were entered into separate models for %MPA, %VPA, and CPM. Full models were run to predict the strongest correlates of physical activity after controlling for age and BMI z-score. Unlike the other variables, age and BMI z-score were entered regardless of evidence of bivariate association. All two-way interactions between significant main-effect independent variables were also tested for significance in the multiple linear regression models. In all analyses, statistical significance was set at an alpha level of 0.05.


The summary data for the physical activity and correlates variables are presented in Table 1. The median duration and median number of days of activity monitoring were 757 min (720–794) and 7 (7, 8) days, respectively. Participating children spent on average 194 ± 54 min/day in MVPA, equating to ∼26% of daily monitoring time. Analyses were conducted for both minutes in MPA and VPA and %MPA and %VPA, and it made no difference to our findings (data not shown).

Table 1.  Descriptives and sex differences for physical activity variables and correlates.
inline image

No sex differences were apparent for %MPA, %VPA, or CPM, or for any of the perceived competence or quality of life variables. Compared with girls, boys had higher BMI z-scores (P < 0.001) and spent more time in SSR (P < 0.05). When sex differences were examined for raw motor skill scores, boys were found to be better performers of object-control skills (P < 0.001) while girls were more proficient at locomotor skills (P < 0.001).

Zero-order correlation coefficients between each of the correlates and %MPA, %VPA, and CPM are shown in Table 2. For boys, age (P < 0.001) and father's BMI (P < 0.01) were inversely correlated to %MPA and %VPA, while locomotor skill proficiency (P < 0.05), object-control skill proficiency (P < 0.001) and GMQ (P < 0.001) were positively correlated with %MPA and %VPA. Scholastic competence (P < 0.05) was an additional correlate for %MPA. Age (inverse) (P < 0.001), father's BMI (inverse) (P < 0.01), object-control skill proficiency (P < 0.001), locomotor skill proficiency (P < 0.05), and GMQ (P < 0.001) were correlated with CPM among boys.

Table 2.  Bivariate correlations between potential correlates and physical activity behavior
inline image

For girls, age (inverse; P < 0.001), locomotor skill proficiency (P < 0.01) and GMQ (P < 0.01) were correlated with %MPA. Similarly, age (P < 0.001) was inversely correlated with %VPA but object-control skill proficiency (P < 0.01) and GMQ (P < 0.01) were positively correlated. Minutes spent in SSR (P < 0.05) was the only variable correlated (inversely) with CPM among girls. There were no significant correlations found between any of the perceived competence or quality of life measures and physical activity in boys or girls.

Results of the stepwise regression analyses for higher intensity (%MPA and %VPA) and total physical activity (CPM) for boys and girls are displayed in Tables 3 and 4, respectively. For boys, age was the sole predictor of %MPA (P < 0.001) (R2 = 0.56). Age and object-control skill proficiency were predictors of %VPA (P < 0.001), explaining 34% of the variance. Object-control skill proficiency was the sole predictor for CPM (P < 0.001) and the model explained 25% of the variance. Paternal BMI (n = 24) was not entered in the regression models for boys because of substantial missing data.

Table 3.  Results of the multiple regression analyses to explain physical activity in boys (N = 58)
inline image
Table 4.  Results of the multiple regression analyses to explain physical activity in girls (N = 79)
inline image

For girls, age was a predictor of both %MPA (P < 0.001) and %VPA (P < 0.001) explaining 38% and 15% of the variance, respectively. Age and SSR were predictors of CPM (P = 0.01) explaining 13% of the variance. The unique contribution of SSR was 6%.

All two-way interactions between motor skill variables, age, and BMI z-score were examined as covariates for each physical activity outcome variable for boys and girls. However, no interactions were significant.


This study investigated the ability of potential correlates to explain the physical activity behavior of obese children prior to participation in a treatment trial. Our findings have revealed key predictors of physical activity which vary to some extent depending on sex and the physical activity variable examined. Motor skill proficiency was found to be an important correlate of physical activity, especially for boys. Time spent in SSR was inversely related to physical activity for girls. Our findings also indicated that age was inversely associated with physical activity among boys and girls. Other hypothesized correlates including quality of life and perceived competence measures were not associated with physical activity.

A key finding of our study was the significant associations established between motor skill proficiency and physical activity, which were evident across sexes and physical activity intensities. No previous study has explored the relationship between motor skill proficiency and objectively measured physical activity in obese children. Although motor skills were not a significant predictor of %MPA for boys, the amount of variance explained uniquely by object-control motor skills (10% for %VPA and 25% for CPM) is far greater than has been previously demonstrated in studies of preschool (32), elementary (33), and high school students (7). Okely et al. (7) used a self-report measure of physical activity and found that motor skills only explained a small amount (3%) of the variation in physical activity among adolescents. Of the limited studies that have examined motor skill proficiency and used an objective measure of physical activity in children, the variance explained in physical activity has ranged from 3 to 9%. Fisher et al. (32) found a significant but weak association between physical activity and motor skills in preschool children. Similarly, Wrotniak et al. (33) found children's motor skills were positively related to CPM and %MPA and %MVPA and explained 8.7% of the variance in CPM. Despite recent contention in the literature regarding the strength of the motor skill and physical activity relationship (32,34), previous studies have been limited by use of self-reported physical activity (7,34), product-oriented skills assessments (32,33), and a low number of skills assessed (35). Our positive findings may be attributable to the use of a process-oriented battery of 12 motor skills and an objective physical activity measure.

Obese children who have greater motor skill competence may be more likely to be physically active through improved self-esteem and enjoyment than those with poorer motor skill competence. The results of the current study suggest that the teaching of motor skills to obese children could be an important component of interventions delivered in home, school, and community settings. This takes on increasing significance for elementary school children who are at an optimal age in terms of motor skill learning (14) and for obese children who have poorer motor skill proficiency than healthy-weight children (33,36).

We also found that the minutes spent in SSR were a predictor of CPM but not higher-intensity physical activity in girls. Strauss et al. (31) demonstrated that sedentary behaviors replace low-intensity (playing, walking) rather than VPA (running, sports), which may help to explain why CPM and not %VPA was related to SSR. SSR, of which TV watching is a major part, is a commonly studied and contentious correlate of physical activity. While it is possible for some children to spend large periods of time in SSR and still be relatively active (37), strategies to reduce sedentary behaviors must be developed given the evidence linking TV watching to obesity (38). It is possible that, for obese girls, SSR may be a competitor to physical activity opportunities. Epstein et al. (39) has previously demonstrated the value in targeting reduced sedentary behaviors with obese children to increase physical activity levels. This strategy may need further investigation, particularly with obese girls.

We found that the correlates of physical activity were similar between sexes with more variance explained in boys than girls. For both boys and girls, age was strongly associated with physical activity, with younger children more active than older children. However, it is important to note that our data may have been confounded by the use of age-specific thresholds for MPA and VPA. The use of age-specific accelerometer definitions for MVPA in children is a contentious issue—some definitions used in differing age ranges are age specific (20) and others are not (17). However, limited empirical evidence is available to clarify the relevance of age specificity in definitions, as well as the appropriate magnitude of change in cut points with age. Preliminary data suggest that age-specificity might be less important than is commonly believed (40), although further longitudinal data is required to investigate this more thoroughly.

Previous reviews of correlates of physical activity in childhood have reported inconsistent associations with age and physical activity with children (4,41), largely due to the narrow age range of children in most studies. However, our findings indicate that early physical activity interventions may be needed for obese children. In the current study, boys were more vigorously active than girls, spent more time in SSR, and were more proficient at performing object-control skills. Girls were more proficient at performing locomotor skills. Previous studies have also identified differences in motor skill proficiency between boys and girls, with boys being more proficient at activities requiring strength such as throwing and striking (33), while girls are superior at demonstrating locomotor skills (42). Differences in motor skills between boys and girls prior to puberty are believed to be a result of environmental (e.g., opportunity to practice, parent expectations) rather than biological factors (43). Our results suggest that interventions designed for obese children should focus on the development of FMSs. Programs for obese boys could suitably focus on object-control skills such as throwing, kicking, and striking. Interventions for girls could emphasize locomotor skills and/or a reduction in SSR.

Interesting relationships emerged out of bivariate analysis of paternal BMI and physical activity. Our findings suggest that obese boys with leaner fathers were more involved in %MPA, %VPA, and CPM. This relationship was not present for girls. Whether or not fathers with higher BMI scores were less likely to promote or role model physical activity is not able to be discerned from our study. It is likely that a father's BMI interacts with other variables to influence a child's physical activity behavior. However, future studies should look to explore this relationship longitudinally with obese children and with directly measured parental height and weight. No psychological health variables were associated with any of the physical activity variables for boys or girls. The lack of significant relationships may be partly explained by the reduced sample size resulting from subgroup analysis, low reliability of quality of life domains, and the small variation in scores for these variables, which may be due to studying a sample of obese children.

There are a number of limitations in the present study. Measurement of some physical activities is problematic with accelerometers including water activities, contact sports and cycling, and children's physical activity has substantial intraindividual variability throughout the year (44). The cross-sectional design of the study does not allow us to infer the causal directions of the observed relationships. Sex-specific analysis reduced our statistical power and ability to detect significant findings and our sample size limited our capacity to conduct analyses separately for age. The use of self-reported paternal/maternal BMI is also a limitation and the relationship to physical activity needs to be explored with a direct measure of parental height and weight.

Finally, the study population represented obese children who had enrolled in a treatment trial. As such, the sample may not be representative of the general population of obese children and hence generalizability of the results may be limited. Nevertheless, the findings of the study are likely to be useful to educators and health professionals who are managing childhood obesity.

There were several strengths of our study including the use of an objective measure of physical activity, a comprehensive and process-evaluated motor skill battery, a good sample size of obese children, and exploration of several new correlates. Our findings establish important and unique information regarding the correlates of physical activity among obese children. Compared to other studies using objective measures of physical activity, the correlates accounted for a comparable amount of variance for girls and a substantially greater amount of variance for boys. Notably, this is the first study to report a significant association between motor skill proficiency and objectively measured physical activity among obese children.

In summary, an improved understanding of the factors associated with physical activity may assist in the design of appropriate interventions to promote physical activity. The findings of this study have demonstrated that the targeting of motor skills may be a promising strategy to increase physical activity among obese children, as they are considered to be prerequisites to both short- and long-term participation in many physical activities. Longitudinal studies would provide further insight into the direction of the relationship between motor skills and physical activity. The study contributes to the literature on associations with the physical activity behavior of obese children and may be of assistance to researchers designing experimental trials for home, school, and community settings to increase physical activity in obese children.


HIKCUPS is funded by the Australian National Health and Medical Research Council of Australia (354101). We thank the participating children and parents, local schools in the Hunter and Illawarra regions of New South Wales, and the Universities of Wollongong and Newcastle. We also thank The Health Food Company Sanitarium for supplying the breakfast cereal for the assessments and Daniel Riethmuller for assisting with the motor skill assessments.


The authors declared no conflict of interest.