Patterns and correlates of objectively measured physical activity in 3-year-old children

Early STOPP is a longitudinal clustered randomized controlled trial (RCT) with an obesity prevention intervention including 238 children with a 5-year consecutive follow-up period. The children have high (n = 181) or low (n = 57) risk of developing obesity based on parental BMI; where high-risk is defined as 1 parent with obesity (BMI ≥ 30) or 2 parents with overweight (BMI 25–29.9). Low-risk is defined as having both parents with BMI < 25 [21]. The families were recruited when the child was 1 year old from Child Health Care Centers (CHCC) in the Stockholm region between fall 2009 and spring 2013. The high-risk families were randomized to intervention (n = 66) or control group (n = 115), based on the CHCCs (cluster) and low-risk families serves as a reference group. More information about the project and the intervention can be found in earlier published papers [21‐23]. To reduce a potential effect of the intervention, the intervention group was excluded in the present study. Among families in the control and low-risk groups (n = 172), 100 families attended the annual visit at 3 years of age (second wave follow-up), taken place ±2 months from the child’s third birthday.

Early STOPP was approved by the Stockholm regional ethics committee in Stockholm in March 2009 (file no. 2009/217–31). The parents signed a written consent to be a part of the Early STOPP project and this sub-study falls under the project’s intent.

PA was assessed using a tri-axial motion sensor, the Actigraph GT3X+ accelerometer (Actigraph, Pensacola, FL). Children wore the accelerometer on their non-dominant wrist, wrist dominance estimated by the parents, for seven consecutive days. For a day to be considered valid it had to contain at least 10 h of PA measurements [24, 25] and all children with less than 4 valid days, or missing weekend data, were excluded [26]. The data was collected between June 2012 and June 2015.

Data was analyzed in the ActiLife program, version 6.11.9. (Actigraph, Pensacola, FL). Night-time sleep was excluded by removing the hours between 8.45 pm – 7.20 am [27] and any potential daytime sleep was considered as sedentary time. The outcome variable was average PA expressed in counts per minute (CPM), with a sampling rate of 30 Hz [28]. We used CPM from the vector magnitude (VM) a variable that combines the 3 axes in 1 outcome defined as √(x² + y² + z²). The outcome was calculated for every hour and for weekdays and weekends respectively.

Body weight was measured to the nearest 0.1 kg with a portable scale Tanita HD-316 (Tanita corp, Tokyo, Japan) and height to the nearest 0.1 cm with a fixed stadiometer (Ulmer; Buss Design Engineering, Elchinge, Germany). BMI (kg/m²) was calculated for children at age 3, second wave follow-up, classifying children as normal weight, overweight or obese using international cut-off values corresponding to adult BMI 25 and 30 [29]. Parental BMI was calculated at baseline and at second wave follow-up using international standards to classify normal weight (BMI 18.5–24.9), overweight (BMI 25–29.9) and obesity (BMI ≥ 30).

Parental educational level was used to determine SES [30]. Mothers and fathers reported their highest level of education as; Elementary school, High school or Academic education. Parental education level was considered high if at least 1 parent had an academic education, otherwise as low [22, 23].

Additional demographic and family-related information was collected in order to adjust for possible confounders [31]. Child daycare was reported by the parents as either staying at home with a parent and/or guardian or preschool care. Preschool was then categorized as full time (≥ 30 h/week) or part time (< 30 h/week). Country of origin was categorized as either Nordic or non-Nordic, where non-Nordic was defined as at least 1 parent born in a non-Nordic country. Number of siblings were collected as either 0, 1 or ≥ 2. Season of measurement was categorized as spring (March–May), summer (June–August), fall (September–November) and winter (December–February). All information was derived from questionnaires filled out by the parents in connection to the visit.

Factorial repeated analysis of variance (ANOVA) was performed in order to evaluate differences between days of week and time of day for the main outcome CPM. A paired t-test was performed to compare differences in PA between weekdays and weekends. To test if PA differed (values through all days/weekdays/weekend days) by child weight status and sex and parental weight status and education, independent ANOVA, with a post hoc equation Bonferroni, was performed. Significant factors (p < 0.05) were further included in a model using simple linear regression, adjusting for: sex, age, family group, daycare, country of origin, number of siblings and season [31]. Cohen’s d were used to calculate effect size for group comparisons, with cutoffs 0.2, 0.5 and 0.8 indicating a small, medium or large effect size respectively [32]. Post hoc power equation for multiple regression was used to calculate power. All analyses were performed with IBM SPSS Statistics version 23.0 (IBM, Armonk, New York, USA).

This cross-sectional study contributes to the knowledge regarding young children’s PA patterns and its correlates. Our findings herein are an extension of our previous findings on 2-year-old children [22], that young children vary their activity pattern over the days as well as between weekdays and weekends. Furthermore, child PA was inversely associated to parental education with a large effect size. No differences in PA between boys and girls was found.

In the present study, children were less active during weekends than weekdays which is in line with previous findings [4‐7, 22, 33]. During weekends children are primarily in care of their parents/guardian and one hypothesis may be that parents do not have the same time to activate their children or that they consider PA as something the preschools should provide [33]. It has also been suggested that higher PA during weekdays could be due to the planned time for outdoor activities in preschools [24, 34]. The hourly PA-pattern differed between weekdays and weekends indicating that the variability could be caused by exogenous factors and that it might then be possible to influence the PA. The differences at 7 am and 8 pm might be explained by the fact that children sleep in during weekends and at the same time also stay up later at night [35]. The hourly pattern most likely confirms that preschools, in greater extent than parents, engage children in activities that are more physically active and have a fixed schedule for meals and naptimes. To further examine the parent-child PA relationship a comparison between their PA levels and patterns should be investigated in future studies.

We did not find any effect of weight status on PA. Studies on school-aged children, adolescents and adults have found normal weight individuals to be more active and less sedentary than individuals with overweight and/or obesity [9‐14]. In preschoolers, there is a positive association between PA and a reduced risk for excessive increase in body weight and adiposity [3]. However, to our knowledge, weight status has, at 3 years of age, not been established as either a determinant of PA nor an outcome of too little PA [8].

Boys and girls were found to be equally active. In a review by Sallis et al. [9], 80% of the studies reported boys to be more physically active than girls among children age 4–17. Wijtzes et al. [31] reported sex-differences only on sedentary behavior but not on low PA or high PA in the same age group. Johansson et al. [22] reported no differences between boys and girls aged 2 within the Early STOPP population. Analyzing and exploring PA within this cohort should for this reason continue, in order to establish if, and in that case when, differences in PA by sex starts. Why boys are being more active than girls has yet to be established. Many factors have been suggested; motor skills, biological age, physical maturity, participation in sports but also social and cultural norm behaviors [36‐42]. Most of these proposed factors applies primarily to older children however, motor skills could be a factor of interest for children younger than 4 years of age and should be considered to include in future studies.

Finding and targeting groups of individuals with risk for low PA might help improve effects of PA interventions [1]. As the predominant risk for childhood obesity [19, 20] parental weight could therefore be a target of interest. Meaning, if parental weight status correlates to children’s PA, children to parents with obesity might benefit from PA-interventions and the added knowledge may make interventions more efficient. However, in this study we could not find differences in PA by parental weight status and to the best of our knowledge no other study has investigated this in 3-year-old children. Previously Johansson et al. [22] found no differences in PA among 2-year-old children with respect to parental weight.

We found that children to parents with a lower educational level were more active on weekdays than their counterparts, with a large effect size. Previous results have been inconsistent and contradictory. In children from 6 years of age SES has in some studies been associated with PA [7, 8, 15‐18]. A systematic review showed that 58% of the included studies reported a positive relationship between SES and PA and the remaining reported the opposite or no relationship [17]. The authors argue that studying this relationship is problematic due to the different types of measurements on both variables [17, 30, 43] that is, subjective or objective PA [17, 43] and various proxy for SES, for instance education, occupation and income [30, 44]. This could lead to the contradictory results of the studies and could therefore make it somewhat hard to compare findings [15, 17]. Even so, Beckvid Henrikson et al. [15] also found that six-year-old children from families with low SES were more active and less sedentary, using objective PA and education as a proxy for SES [15]. They discuss the possible implication of age on the differences in results, arguing that at an older age spontaneous PA is replaced with more structured sports [15, 16]. Resulting in the strengthening of a positive relationship to SES, since the coast for activity increases [15, 16]. Our results on 3-year-old children might cohere with their hypothesis that the correlation between SES and PA might differ depending on age.

We used an accelerometer (Actigraph GT3X+) to objectively measure PA with a mean of 6.7 valid days, providing robust data [25, 26, 45, 46] making us able to provide a picture of the children’s PA pattern hour by hour. Accelerometry is the suggested and most preferable method to be used when measuring younger children [45]. Although wrist placement is not the most common wear placement it has been shown to increase the wear compliance among children [46]. Using a uniform sleep-time for all children on both weekends and weekdays may be a source of uncertainty in the interpretation of the results. Nevertheless, children at this age often have regular hours independent of day of week [35].

Due to the relatively small sample from the Stockholm area, the higher proportions of highly educated parents with mostly Nordic country background compared to the Swedish population [47], the results should be interpreted with caution to other populations. Moreover, in this study we did not have information on motor skills or parental PA that might be of value for the total PA in this age group. Education as proxy for SES is the most commonly used, but using only 1 variable for a person’s SES is questionable and creating an index might give a more equitable estimation [30]. However, there are no agreed upon definition on what variables to then include [30].