Correlates of the Physical Activity Decline during Childhood

Abstract

Purpose:

To describe longitudinal trends in children’s moderate-to-vigorous physical activity (MVPA), and to investigate associations with individual, familial and school characteristics.

Methods:

A sample of 341 5-10 years old Portuguese children (173 girls) from 6 age-cohorts was followed over three years using a mixed-longitudinal design. Physical activity, body mass index (BMI), gross motor coordination and musculoskeletal fitness were assessed annually. Information on socioeconomic status (SES) and school characteristics was collected and analyzed with mixed-models.

Results:

MVPA shows a similar declining trend in both sexes, but on average boys exceeded the WHO recommendations of 60 min·d⁻¹. The best model showed that boys spend, on average, more time in MVPA than girls. Children with lower BMI are less prone to the decline in MVPA, whereas higher levels of musculoskeletal fitness were associated with lower declines in MVPA. Of all school characteristics, only playground dimension was related to MVPA decreasing trajectories.

Conclusions:

MVPA systematically declines from 5 to 10 years of age in both boys and girls, but boys remained more active than girls across the age range. The best predictors of MVPA decline are sex, BMI, musculoskeletal fitness and school playground dimension.

INTRODUCTION

Physical activity has been shown to decline with increasing age during childhood (1), and there is evidence of sex differences in the negative rates of change (2, 3). However, what drives this decline is not well known yet.

Physical activity is a complex behaviour constrained by a myriad of factors, including but not limited to aspects of the individual and the environment (4). For example, there is evidence regarding the importance of gross motor coordination and physical fitness on children’s physical activity involvement, i.e., those with higher gross motor coordination and physical fitness levels being more likely to engage in more physical activities (5, 6). Although reviews have been written about children’s physical activity correlates (7, 8) few studies have used longitudinal data to examine, simultaneously, the relationship between sets of individual and environmental factors during the primary school years (9). This is an important aspect to be considered as cross-sectional studies are unable to tease out the independent effects of normal growth and development from the effects of sex, body mass index (BMI), gross motor coordination, physical fitness, socioeconomic status (SES) and school characteristics on physical activity.

In 2008, David Stodden and colleagues (10) suggested that motor competence and health-related physical fitness are putative causal determinants of physical (in)activity. Interestingly, there is scant longitudinal evidence supporting the causal effects of gross motor coordination and physical fitness on physical activity developmental trajectories during the primary school years (11–13). In addition, family and school characteristics also appear to play important roles on the expression of children’s physical activity across time. Moreover, differences in SES among children have been related to physical activity (14). Butte, Wong (15) suggested that individual characteristics (age, sex and BMI) were more important than family influences on children’s developmental trends in physical activity. Since children spend most of their awake time at school, this likely influences their physical activity engagement (16). For example, it was previously reported that schools explained ~18% of the total variance in moderate-to-vigorous physical activity (MVPA) in children aged 5 to 10 years (17). However, from the set of school-level predictors suggested by the Joint Consortium for School Health planner and adapted to meet the region specificities namely in terms of school characterization, policies and practises for PA, school physical infrastructure, and physical education class duration, only playground dimension was a significant one (17). Differently, Steenholt, Pisinger (18) using Danish children’s PA questionnaire data reported that school predictors explained from 4% (offers physical activity outside the school hours, good selection of sports equipment, good changing and shower facilities) to 8% (good access to outdoor football fields) of the total PA variance. Additionally, Fairclough, Ridgers (19) only included playground size in multilevel models investigating biological, sociodemographic and school on children MVPA but found positive association between playground size and vigorous physical activity during the weekdays. More recently, Pereira, Borges (20) in a study with Portuguese children aged 9-11 years old including more school variables (school size, number of intra-mural sports offered in PE programs, the number of distinct facilities (e.g., gymnasia, tracks, etc.) available during school hours, and policies and/or practices concerning PA) showed that only available facilities were related with children MVPA compliance. Additionally, Gomes, Katzmarzyk (21) extended to children from 12-countries and included: playground size, break-time, promotion of students active transportation, access to gymnasium, playground equipment, indoor facilities and sport equipment (in school hours and out of school hours); they reported that only access to playground equipment (during school hours) and sport equipment (out of school hour) are related with MVPA compliance. To our knowledge, no studies with a longitudinal or mixed-longitudinal design have investigated the joint facets of individual and environmental characteristics, using a multilevel approach, to probe into children’s objectively measured MVPA trajectories during their primary school years. Therefore, the aim of this study was to: (1) describe longitudinal trends in children’s MVPA and identify putative sex differences in the rates of change; (2) verify the influence of BMI and SES on MVPA trajectories; (3) investigate the associations of gross motor coordination with MVPA trajectories; (4) probe the putative positive effects of musculoskeletal fitness levels on MVPA trajectories; and (5) examine the role of the school characteristics on children’s MVPA trajectories.

MATERIAL AND METHODS

Sample

Data comes from the Growth, Motor Development and Cognition Study (GMDC – Vouzela study) which took place in Vouzela county, located in the midlands of Portugal. A detailed description of the study can be found elsewhere (22). In brief, the purpose of the study, using a mixed-longitudinal cohort, was to investigate the dynamic links between children’s physical growth, motor performance (gross motor coordination and physical fitness), lifestyle characteristics (nutritional behaviors, physical activity, screen time and sleep habits), metabolic syndrome risk, cognition (reading skills and intelligence), and school characteristics.

Children were recruited into one of six age-cohorts and followed for three consecutive years (2013-2014; 2014-2015; 2015-2016). Cohorts had a two-year overlap such that in cohort 1 children were followed from 4 to 6 years, in cohort 2 from 5 to 7 years, and so on until cohort 6 followed children from 9 to 11 years. In the present paper, we will only consider children from 5 to 10 years of age because of different constraints. For example, when assessing gross motor coordination only children from 5 years onwards were considered because the battery used at the present study was the Körperkoordination Test für Kinder battery (KTK), that tests children from 5 years of age. Further, the number of children with complete physical activity data at 11 years was relatively low.

Children are from a convenient cluster sample (schools and children) from 19 public schools. The number of drop-outs was relatively small at 8.5%. In the 2nd year of the study, the drop-outs among girls included one 8 years old, three 9 year olds, and two 10 year olds, while the drop-outs among boys included two 5 year olds, one 6 year old, and one 10 year old. In the 3rd year of the study, the drop-outs among girls included one 6 years old, one 7 years old, one 10 years old, and one 12 years old, while the drop-outs among boys included one 5 years old, one 8 years old, three 9 year olds, and three 10 year olds. In total 23 children dropped out. Given the multilevel data structure, and the implied statistical model, there is no need for complete data at each measurement occasion from every child, assuming that the missing data are at random (23). In any case, we tested for putative differences in children with complete data versus those with missing information across a set of variables such as sex, gross motor coordination, birth weight, height, weight, laterality, physical fitness, physical activity and SES and only found differences in standing long jump and shuttle run test favoring children used in the present analysis. Children with physical disabilities (n=27), as reported by their educators, that would limit their ability to perform all tests were not included in the study. In summary, the sample for this article comprises 168 boys and 173 girls. Written informed consent was obtained from all parents or legal guardians; further, we gathered permission from all Vouzela school directors, political and health authorities to conduct the study, and the ethics committee of the Faculty of Sport, University of Porto approved the research project.

Physical activity

Physical activity was objectively assessed using the Actigraph GT3X+ accelerometer (Actigraph, Pensacola, USA). All children were instructed to wear the device for seven consecutive days (5 week-days and 2-week-end days). Further, they were also instructed to only remove the device during water activities (e.g., in their normal daily showers and swimming) as well as when sleeping, and we kept daily contacts with parents to ensure that their sons/daughters were wearing the device. The accelerometer was placed on the child’s iliac crest and was held in place by an elastic band with an adjustable clip.

Actilife® software (version 6.5.4.) was used to download the recorded data immediately upon retrieval of each accelerometer. For data to be valid, two conditions had to be met: children had to wear the accelerometer for at least 4 days (with one weekend day), with a minimum of 10 hours of daily wear time (girls=13.5±1.8 hours·d⁻¹; boys=13.6±1.8 hours·d⁻¹). All sequences of at least 20 consecutive minutes of zero activity counts were considered as non-wear time (24), and physical activity distinct expressions were derived from the cut-off points recomended by Evenson, Catellier (25). In the present study we only consider MVPA (min·day⁻¹) using the cut-point of 574 counts/15 s epoch.

Anthropometry

Standard techniques were used for measuring height and body mass with children in light clothing. Height was measured with a portable stadiometer (Holtain, UK) with the head positioned to the Frankfurt plane and the accuracy was 0.1 cm. Body mass was measured with a portable scale (TANITA BC-418 MA segmental body composition analyzer, Japan) with an accuracy of 0.1 kg. BMI was calculated with the standard formula: BMI= body mass (kg)/height (m)².

Gross motor coordination

Gross motor coordination was assessed using the Körperkoordination Test für Kinder battery (KTK) developed in Germany by Kiphard and Schilling (26) for children and adolescents aged 5 to 14.9 years. The battery has systematically been used in Europe (27–29) and comprises four tests marking different aspects of gross motor coordination: balance stability (test: walking backwards on balance beams); lower limb coordination and dynamic energy/strength (test: hopping for height on one leg); speed in alternating jumping (test: jumping sideways); laterality and space-time structuring (test: moving sideways). Using the results from each of the four tests, i.e., the number of points, and as advocated by Schilling (30), we summed them to produce an overall measure of gross motor coordination. Using this sum, within each age and sex grouping, we used the median to construct two gross motor coordination groups: low (below the median), and high (median and above).

Musculoskeletal fitness

Musculoskeletal fitness was assessed using three tests (IOM, 2011): explosive leg power with the standing long jump test (expressed in cm); agility with the shuttle-run test (expressed in seconds); static strength with handgrip strength test (expressed in kg using the Takei GRIP-D dynamometer, Japan). We then standardized all scores (z-scores), but the shuttle-run performance was multiplied by −1. Then, we summed individual z-scores to obtain a composite to express overall musculoskeletal fitness as advocated (31).

Socioeconomic status

Socieconomic status (SES) was obtained from the Portuguese school system social support which is based on a Portuguese Ministry of Education directive (32). In brief, a budget index classifies families into three distinct categories: low (up to 2,934 €·y⁻¹), middle (from 2,934 to 5,896 €·y⁻¹), and high (≥5,897 €·y⁻¹). We created a dummy coding scheme with low level as the reference category.

School characteristics

An inventory of the school environment was obtained via an objective audit developed by authors of this study based on previous information from the ISCOLE project (33) as well as in partnership with the Vouzela city-hall Education Department. Five domains were mapped: school characteristics (location; size and setting - rural, semi-urban and urban, as determined by the Portuguese National Statistics Institute); policies and practices to stimulate physical activity and nutrition; physical structure of the school (playground, multi-sports roofed complex, facilities and equipment for physical education); physical education (PE) classes; human resources (number of teachers in each school, academic degrees of teachers, and teaching time).

For this article we included the following items: school size based on the number of students; location was dummy defined, such that rural was our reference category; policies and practices was also dummy coded with no policies nor practices as our reference category; playground was also dummy defined with 10 to 39 m² as the reference category; the presence of a multi-sports roofed (0=no; 1=yes); Number of infrastructures for PE (0=one infrastructure; 1=two infrastructures); PE material, equipment for gymnastics, equipment for team or individual sports, equipment for traditional games, swimming pool extracurricular activity were binary coded (0=no; 1=yes) as well as PE time (0=45 minutes; 1=60 minutes).

Statistical analysis

Descriptive statistics are presented as means, standard deviations and percentages. Given that the data have a hierarchical structure – repeated observations (level-1) nested within children (level-2) which are themselves nested within schools (level-3), we relied on the mixed model (23) also known as the multilevel hierarchical linear model (34). In this three-level model we followed the suggestions of Hedeker and Gibbons (35) in terms of model testing strategy, but applied to the study aims. First, we anchored the time metric at 5 years, so that 0, 1, 2, 3, 4 and 5 corresponds to 5, 6, 7, 8, 9 and 10 years of age. Second, we tested a series of nested models with increasing complexity in terms of number of covariates: model 1 is a “growth” model that defines the best trend for MVPA changes (linear or curvilinear), and once found we then tested for a significant interaction with sex, i.e., testing if boys and girls show distinct MVPA changes across time; then, in model 2 BMI (time-varying predictor) and SES were added; gross motor coordination (time-varying predictor) was added in model 3, and in model 4 musculoskeletal fitness (time-varying predictor) was added. Our final “growth” model (model 5) included school characteristics (number of students, school setting, policies and practices for physical activity, playground dimension, multi-sports roofed, PE material, equipment for gymnastics, equipment for team or individual sports, equipment for traditional games, swimming extracurricular activity and PE class time). Third, whenever required, covariates were centered at the grand mean as advocated (Hedeker & Gibbons, 2006). Fourth, competing models (from 1 to 4) were tested using the Likelihood ratio chi-square test (LRT) (Hedeker & Gibbons, 2006). Fifth, all model parameters were simultaneously estimated using maximum liklelihood procedures with the Mixed module implemented in the statistical software STATA 16. The significance level was set at 5%.

RESULTS

Table 1 shows descriptive information from the sample grouped by sex and age. As expected, there are increments in almost all variables with age (BMI, gross motor coordination, and musculoskeletal fitness) in both sexes. Although boys show higher mean values of MVPA than girls, both sexes show a steady MVPA decline with age (Figure 1).

Table 1:

Descriptive statistics (means±standard deviations or percentages) for boys and girls across chronological age groups.

	5 years G¥, n=23; B¥, n=39	6 years G, n=57; B, n=59	7 years G, n=70; B, n=64	8 years G, n=73; B, n=65	9 years G, n=83; B, n=63	10 years G, n=47; B, n=41
MVPA§ (min·d−1)
Girls	65.3±20.8	62.3±19.5	60.2±17.4	59.7±19.8	55.0±15.3	57.8±18.3
Boys	83.6±23.8	82.8±21.8	77.3±23.5	74.0±23.4	71.3±20.8	68.2±22.9
BMI (kg·m−2)
Girls	16.7±2.0	16.8±2.0	17.8±2.9	18.1±3.2	18.8±3.2	19.3±3.4
Boys	16.9±2.0	17.0±2.1	17.1±2.3	17.9±3.6	18.4±3.8	19.2±4.4
SLJump (cm)
Girls	78.9±17.1	94.1±17.7	102.1±16.5	114.7±16.9	123.5±17.0	132.2±17.3
Boys	82.8±19.8	99.2±19.6	114.1±18.0	121.0±19.0	130.6±17.1	139.0±16.5
Shuttle Run (s)
Girls	16.1±2.2	14.6±1.5	13.9±1.3	13.3±1.1	12.8±1.1	12.2±1.1
Boys	15.6±2.2	14.5±1.9	13.3±1.3	13.1±1.1	12.3±0.9	11.8±0.7
Handgrip (kg)
Girls	6.3±1.0	7.9±2.0	10.0±2.5	12.7±2.9	14.3±3.0	16.5±3.2
Boys	6.6±1.8	9.1±2.4	11.2±2.5	13.5±3.4	15.4±3.3	18.2±4.4
PF (Σ z-scores)‡
Girls	−4.3±1.3	−2.4±1.4	−1.2±1,4	0.3±1.4	1.3±1.5	2.6±1.5
Boys	−3.8±1.7	−1.9±2.0	−0.1±1.5	0.9±1.5	2.1±1.3	3.5±1.2
GMC categories†
Girls (low)	42.9%	50.0%	44.6%	50.0%	47.2%	43.2%
Boys (low)	36.8%	45.7%	54.4%	50.8%	43.6%	47.5%
Girls (high)	57.1%	50.0%	55.4%	50.0%	52.8%	56.8%
Boys (high)	63.2%	54.3%	45.6%	49.2%	56.4%	52.5%
SES categories℄
Girls_low	26.1%	17.9%	20.0%	20.3%	24.1%	21.3%
Boys_low	17.9%	13.6%	14.1%	10.8%	14.1%	7.1%
Girls_middle	39.1%	37.5%	37.1%	28.4%	27.7%	21.3%
Boys_middle	20.5%	25.4%	23.4%	24.6%	26.6%	35.7%
Girls_high	34.8%	44.6%	42.9%	51.4%	48.2%	57.4%
Boys_high	61.5%	61.0%	62.5%	64.6%	59.4%	57.1%

^¥, G=Girls^, B=Boys;

^§, MVPA= Moderate-to-vigorous physical activity;

^‡, sum of z-scores for Physical Fitness tests;

^℄socioeconomic status categories (low up to 2.934 €·y⁻¹; middle from 2.934 to 5.896 €·y-1; high ≥5897 €·y-1).

^†GMC categories (low < median; high ≥ median)

Figure 1:

Linear trend in mean MVPA (with 95% confidence intervals) in boys and girls. The dashed-dot line corresponds to the WHO recommendation of 60 min·d⁻¹ of MVPA for children and adolescents.

Table 2 presents the school characteristics for the 19 participating schools. There is variability in school size with a range of 5 to 87 children. The majority are located in rural areas (63.2%), have policies and practices for PA (52.6%), and have playground areas with obstacles. Most playground areas (68.4%) are relatively large (>70 m²), but most schools (73.7%) do not have roofed multi-sports arenas. The type of infrastructures varied between schools but the most common are unpaved outdoor space (52.6%), cement floor (36.8%), and multi-sports roofed (26.3%). One school (5.3%) reported using a classroom for PE. The majority of schools (78.9%) have equipment for PE and distributed as follows: 52.6% have equipment for gymnastics, 63.2% for team and individual sports, and only 26.3% have equipment for traditional games; yet, 73.7% have swimming extracurricular activity. Finally, 68.4% of the schools have PE classes with a duration of 60 minutes.

Table 2:

School-level characteristics.

School size	Mean±SD (min-max)
Number of children	23±22 (5-87)
Number of teachers	4±3 (1-8)
School setting	n (%)
Rural	12 (63.2)
Semi-urban	7 (36.8)
Policies and practices for physical activity	n (%)
Policies and practices	10 (52.6)
Only policies	5 (26.3)
Only practices	4 (21.1)
Playground dimension	n (%)
Small (10 m2 to 39 m2)	2 (10.5)
Medium (40 m2 to 69 m2)	4 (21.1)
Large (≥70 m2)	13 (68.4)
Multi-sports roofed facility dimension	n (%)
No multi-sports roofed facility available	14 (73.7)
Small to medium (≤49 m2)	2 (10.5)
Large (≥50 m2)	3 (15.8)
Number of infrastructures for physical education	n (%)
One infrastructure	15 (78.9)
Two infrastructures	4 (21.1)
Type of the infrastructures for physical education	n (%)
Unpaved outdoor space	7 (36.8)
Multi-sports roofed	1 (5.3)
Cement floor	6 (31.6)
Classroom or similar	1 (5.3)
Unpaved outdoor space and multi-sports roofed	3 (15.8)
Cement floor and multi-sports roofed	1 (5.3)
Equipment for physical education	n (%)
Yes	15 (78.9)
No	4 (21.1)
Equipment for gymnastics	n (%)
Yes	9 (47.4)
No	10 (52.6)
Equipment for Team Sports	n (%)
Yes	12 (63.2)
No	7 (36.8)
Equipment for Individual Sports	n (%)
Yes	12 (63.2)
No	7 (36.8)
Equipment for traditional games	n (%)
Yes	5 (26.3)
No	14 (73.7)
Swimming extracurricular activity	n (%)
Yes	14 (73.7)
No	5 (26.3)
Physical education class duration	n (%)
45 min	6 (31.6)
60 min	13 (68.4)

In accordance to our aims and the modelling strategy, our first step was to determine the best MVPA trend (linear vs curvilinear), and a linear “growth” model fitted the data better (LRT=1.11, p=0.291). Then we tested if boys and girls had different changing rates (an interaction of sex-by-age). Since no significant sex-by-age interaction was found (β=−1.492±1.046, p=0.154), we dropped it from the model. We plotted MVPA trends for boys and girls across age (Figure 1). Boys and girls experience declines in MVPA across age from 5 to 10 years. However, in boys, their means across age (age trend) were always higher than the WHO daily recommendations; yet, girls’ mean values were lower than the recommended values from 8 years old onwards.

Table 3 shows the results of the multilevel models for minutes spent in MVPA conditional on children’s age changes. Model 1 shows that, on average, the mean MVPA of a 5-year-old girl was 65.1±2.5 minutes, and that boys had, on average, 17.0±1.9 minutes (their MVPA mean was 82.1±2.4 minutes) more than girls of the same age. Moreover, the slope of the relationship between MVPA and age (trend) was statistically significant (p=0.05), meaning that with increasing age MVPA systematically declines similarly in both boys and girls (β=−1.85±0.55 minutes·year⁻¹). Model 2, which fitted the data better than model 1, (LRT=11.61, p<0.05) showed that, on average, children with high BMI had lower MVPA (β=−0.88±0.31 minutes). Yet, no MVPA differences were found across SES categories. Model 3 with gross motor coordination fitted the data significantly better than Model 2 (LRT=8.02, p<0.05). Here, children with high levels of gross motor coordination tend to have higher MVPA (β=4.44±1.54 minutes). However, when musculoskeletal fitness was added in model 4 a better fit was obtained relative to model 3 (LRT=15.55, p<0.05) but gross motor coordination levels (low or high) are no longer statistically significant; further, musculoskeletal fitness is significant showing that, on average, children with higher levels (greater z-scores) also tend to show greater MVPA values across age (β=2.31±0.58 minutes). Finally, our “growth” model 5 with school-level variables fitted the data better than the previous one (LRT=93.93, p<0.05). In this model all previous predictors remain similar to Model 4. From our set of school-level characteristics (model 5), only playground dimension was statistically significant (p<0.05), indicating that children from schools with a playground between >40m² tend to have lower MVPA levels than those from schools with playground with less than 40 m².

Table 3:

Parameter estimates (±standard errors) for the five tested models

	Model 1 Estimate±se	Model 2 Estimate±se	Model 3 Estimate±se	Model 4 Estimate±se	Model 5 Estimate±se
Fixed effects
Children
At 5 years (min.day−1)	65.05±2.48	65.69±3.00	64.00±3.04	74.49±3.99	119.08±25.60
Age (min·year−1)	−1.85±0.55	−1.41±0.57	−1.40±0.57	−4.64±0.98	−4.72±1.00
Sex (boys)	17.03±1.88	17.47±1.86	17.47±1.83	15.61±1.84	15.29±1.87
BMI (kg·m−2)		−0.88±0.31	−0.72±0.31	−0.65±0.30	−0.59±0.31
Socioeconomic status (middle)		−0.16±2.76ns	−0.76±2.71ns	−0.62±2.65ns	−0.55±2.70ns
Socioeconomic status (high)		−4.10±2.52ns	−4.61±2.48ns	−4.30±2.42ns	−4.31±2.47ns
KTK (high level)			4.44±1.54	2.29±1.63ns	2.24±1.64ns
Physical fitness (z-scores)				2.31±0.58	2.46±0.59
Schools
Size (n° students)					0.07±0.44ns
Location (rural-urban)					5.79±13.99ns
PP (only practises)					−6.44±23.45ns
PP (polices + practises)					−17.12±15.67ns
Playground (40-69 m2)					−45.19±18.47
Playground (>70 m2)					−47.67±22.02
Multi-sports roofed (no)					9.26±8.11ns
Number of infrastructures for PE (two					8.28±6.45ns
PE material (yes)					−7.63±7.62ns
Equipment for gymnastic (yes)					2.45±15.55ns
Equipment for team or individual sports (yes)					40.00±44.38ns
Equipment for traditional games (yes)					−4.16±22.76ns
Swimming pool extracurricular activity (yes)					3.01±9.30ns
PE class time (60 min)					−35.17±20.60ns
Random effects
Schools
Constant	35.92±20.18	35.28±19.77	37.4±20.15	40.02±21.22	25.22±30.60
Children
Constant	172.68±24.94	162.88±24.15	150.73±23.63	137.47±22.41	143.10±22.93
Residual	219.57±16.88	219.50±16.87	222.16±17.18	222.23±17.10	221.89±17.03
LogLikelihood	−2980.97	−2975.17	−2971.00	−2963.39	−2908.255
Likelihood ratio test		11.61 (2)	8.02 (1)	15.55(1)	93.93 (14)

ns – non-significant

DISCUSSION

In this study, longitudinal trajectories of MVPA in primary school-children, as well as their individual and environmental correlates, were investigated. We showed a systematic linear decline in MVPA with age in both boys and girls, which is not entirely consistent with what has been previously reported. For example, Farooq, Martin (1) recently examined year-to-year changes in MVPA among children and adolescents across 52 studies, and reported that the percentage decline at 9 years is different in boys (−7.8%) than in girls (−10.2%). Further, Schwarzfischer, Gruszfeld (36) reported MVPA changes in children from five European countries (German, Italy, Belgium, Poland and Spain) measured at 6, 8 and 11 years of age, and indicated that MVPA remains stable between 6 to 8 years of age, but steeply dropped off at 11 years of age. Alternatively, Basterfield, Adamson (37), using 2-year MVPA changes in 7-year old English children followed-up at 9 years, found a decline of 3 min in girls but not in boys. Additionally, a study with Swiss children aged 8-12 years (38) showed a linear decrease in time spent in MVPA with age but only during PE classes. This discrepancy between studies may be related to several factors, including: (i) differences in study design, (ii) duration of follow-up, (iii) statistical techniques used, and (iv) the use of different cut-points to define MVPA.

Despite the systematic decline in MVPA observed in both Portuguese boys and girls, boys tend to be more active than girls, on average by as additional 17 minutes, and comply with the WHO daily recommendations across age. However, girls, on average, do not comply starting from 8 years old. Previous studies have attempted to explain why girls are less active than boys; girls tend to be less engaged in organized sports (39), in extracurricular sport activities and also show a lower perceived competence during their PE classes (40). This is a very important message, reinforcing previous evidence (41), that the primary school years are crucial to developing more assertive intervention programs to increase opportunities to promote physical activities of different sorts and expressions, in systematic play and or non-organized forms aiming to attenuate the decline in MVPA across age, especially in girls.

In our model which includes all individual and school characteristics (model 5), the rate of decline was 4.61 minutes of MVPA per year. This sheds light on the importance of the joint influence of these characteristics on MVPA trajectories. However, when we only looked at the associations of sex, BMI, SES and gross motor coordination on MVPA trajectories (model 3) we found that more coordinated children also tended to be more active. Previous studies (11, 37) also revealed similar associations i.e., that higher gross motor coordination levels positively influenced MVPA (11, 37). Of importance in our results, when we added musculoskeletal fitness into the equation (model 4) the significant association between gross motor coordination and MVPA was no longer significant, and this result remained in the full model (model 5). It is worth mentioning that previous longitudinal studies did not consider the joint influence of these characteristics, i.e., gross motor coordination and musculoskeletal fitness, on MVPA trajectories during primary school. As such, comparisons with our study are problematic. Nonetheless, we were able to identify a study with children conducted in Finland (5), using cross-lagged associations with stability, locomotor and manipulative skills, cardiorespiratory fitness, muscular fitness and MVPA during two consecutive years (baseline with 11 years old’s). In these Finn children, the results showed that cardiorespiratory fitness was the only significant predictor of MVPA and this association appeared only in boys. We think that this “trend” may be related to the fact that more physically fit children are perhaps more motivated to participate in physical activities but to test the influence of musculoskeletal fitness on the MVPA trajectories we need to look carefully at the type of physical activity children are engaged in. Given that musculoskeletal fitness is more related to MVPA than gross motor coordination, this may reinforce the suggestion that it is a good marker of a more active and healthier life during childhood (42). Furthermore, these results also highlight the importance of a careful planning of type of physical activities in PE classes as well as sports’ coaches training sessions with children aiming to increase their physical fitness levels.

There is no doubt that schools are important in children’s growth and development, and results from our model 5 suggest that it is also important in terms of MVPA changes across age. However, from the set of school predictors, only playground dimension was statistically significant and the direction was negative i.e., children from schools with medium to large playgrounds (≥40 m²) tend to have lower MVPA levels than those from schools with smaller playgrounds (<40 m2). Although we have previously found a similar tendency (17) this is an unexpected result, even though we were unable to locate studies relating playground dimensions with MVPA changes in primary school children. Playground dimensions have been a focus in different intervention programs aiming to increase physical activity levels also focused on playground dimensions as well as other related aspects like structured recess, playground markings, game equipment and physical structures (43). However, a recent meta-analysis (44) focusing on the effectiveness of playground interventions revealed distinct results ranging from negative small effects to positive small effects. Additionally, given the time children spend at school (on average ~8 hours), its influence on their physical activity levels is expected. However, most of their time is spent sitting in classrooms. One could hypothesize that children only have time and space to be physically active during recess (two breaks of approximately 20 minutes each) and in PE classes (twice a week for 45 or 60 minutes). However, Vouzela county is a very good example in providing equal opportunities for all children to participate in PE as well as in sports activities free of cost. Over the last few years, the city-hall managed to provide infrastructure and equipment to all schools. In fact, MVPA levels in Vouzela children are satisfactory (above the WHO recommendations), except in girls from 8 years old and these results reinforce the importance of creating intervention programs in girls especially in these ages. These results also reinforce the idea that policymakers need to provide and create suitable conditions to foster and develop healthy lifestyle trajectories. Additionally, no MVPA differences were found across SES categories. Previous studies showed inconsistencies between SES and MVPA in primary school children (45). For example, Matsudo, Ferrari (46) in a study with Brazilian children reported that families with higher SES had lower odds of complying with MVPA guidelines. However, Kristensen, Korsholm (47) in a study with European children showed that SES was unrelated to physical activity in 8-10-year-old children. The discrepancies between studies may arise from putative mediational effects of other variables on SES (48). Therefore, further research is needed to better comprehend of the precise link between SES and MVPA.

Notwithstanding the importance of these results, this study has some limitations: (a) the sample comes from a specific region of Portugal and generalization requires caution; (b) we did not consider other children’s characteristics that might influence their MVPA levels, such as perceived motor competence as well as extracurricular activities; (c) even though the number of children within schools varies, we relied on the full maximum likelihood methods to estimate all model parameters in our three-level models. In results not shown, we also re-ran all models using restricted maximum likelihood and difference in results was trivial and did not change the essence of what was reported. In any case, we caution readers regarding the generalization of the school-level covariates because of the available children within each school; (d) Moreover, we did not include variables related to teachers’ practices such as the didactics of teaching and learning as well as the effectiveness of school programming that may influence moderate-to-vigorous physical activity. However, in Portugal, there is a consolidated physical education program developed by the Ministry of Education for the primary school years that all teachers should follow. This study also has several strengths: (a) the use of objective and highly reliable measurements of MVPA, gross motor coordination, and musculoskeletal fitness; (b) sampling from a very important developmental time window (5 to 10 years of age); (c) the consecutive follow-up, and (d) the use of a multilevel model to consider the complexity of the data structure (repeated observations nested within students, which are nested within schools).

CONCLUSIONs

In conclusion, this study showed a linear declining trend in MVPA from 5 to 10 years of age. Boys tend to remain more physically active than girls over time. More fit children and those with lower BMI tend to have less steep MVPA declines. Finally, from the school characteristics only playground dimensions predicted MVPA declining trajectories. This is relevant information to be taken into consideration by policymakers when designing and implementing diversified strategies to attenuate children’s MVPA decline. In sum, this body of results indicate that one of the main challenges policymakers face is to also think in terms of tailor-made interventions instead of relying on generalized programs, i.e., the need is to focus on children whose characteristics are negatively related to the observed declining trends in MVPA and to provide ample opportunities for MVPA. Remedial programs should provide new opportunities for girls and children with high BMI. We suggest Physical Education teachers, together with school directors and staff from City-Hall education departments consider a broad and very inclusive assessment approach to identify children which, from the outset, may be more prone to show declining trends in MVPA and create opportunities to systematically and reliably change this condition. There is no doubt that the primary school years are also important windows of opportunity to understand what is (de)motivating these children, as well as the specificities of their environmental contexts, so that they may be systematically engaged in their physical activities with their peers. There is also an urgent need to identify which activities children perceive as more enjoyable not only at their school but also in their leisure times, as well as considering individual differences in order to provide suitable educational strategies where different children can be more active, have fun and perceive the importance of being active and healthy.