Study setting, design, and population
Within the Södermalm district of Stockholm Sweden, a convenience sample of 30 out of the total 51 municipal preschools were invited to participate in the current cross-sectional observational study. Out of 30 preschools invited, 27 preschools chose to participate, including a total of 1178 children. At the participating preschools, all children between 3-5 years of age were invited to participate. Out of the 1178 total children, 405 (23%) children and their parents consented to participation in the study. In Sweden, children aged below 6 attend preschools every weekday (i.e. 5/7 in a week). The study was approved by the Stockholm Ethical Review Board (Dnr: 2018/890-31/2), and informed consent was obtained from participating children’s parents and preschool teachers. The fieldwork measurements, including body measures of children, 7 days of accelerometer measures of PA in children and parental questionnaires, were carried out during September to November 2018.
Participation in organized sports (exposure)
Parents to participating children filled in a questionnaire with the question “Does your child participate in any kind of organized sports?”. The answer options were (i) no participation, (ii) 1-2 hours organized sports/week, (iii) 3 hours organized sports/week or (iv) 4 or more hours organized sports/week. Organized sports are generally accepted as structured leisure time activities in non-profit organization (24). Therefore, the participation in organized sports assessed here was explicitly about sports participation outside preschool hours. Due to the low participation in organized sports of more than 2 hours/week (10%), we classified participation in organized sports into a dichotomous variable: (i) No organized sports and (ii) at least one-hour participation in organized sports/week.
Physical activity and sedentary time outside preschool time, during preschool time and throughout the day (outcomes)
The outcome measures, PA and ST outside preschool time, during preschool time and throughout the day, were measured via the triaxial Actigraph GT3X+ accelerometer, which has been widely used to assess PA and ST in pediatric research (25). We consulted best practices for wear protocol and analysis (25) and as such decided our procedure to be the following: All children were instructed to wear accelerometer, at right hip for 7 consecutive days, during all waking hours children that had worn the accelerometer for at least 3 days with 10 or more wear-time hours per day were included in the analytical dataset (25). Non‐wear time was defined as 60 or more consecutive minutes with zero counts, allowing up to 2 minutes of interruptions with non-zero counts (26) after adaptation for a potentially less compliant sample (preschool children) (27). We analyzed vector magnitude (Vm) activity counts (Vm = √ (X2 + Y2 + Z2)) in 60s epochs following the calibration study by Butte et al that developed MVPA, LPA and ST cut-offs specifically for Vm activity counts in 4‐year old children (28): ST was calculated as any minute of less than 820 counts per minute (cpm), LPA as 820-3907 cpm and MVPA as ³3908 cpm. Steps were determined using the manufacturer’s step algorithm, using the normal filter setting.
During the 7 days of accelerometer measures in children, preschool staff recorded time arrival to and leaving the preschool on daily basis for each participating child. This data was thereafter matched, on daily level, to time-stamped accelerometer data which enabled us to calculate PA and ST before, after and during preschool. PA and ST before and after preschool time were then combined with PA and ST during the weekend to calculate PA and ST outside preschool time. PA and ST throughout the day was considered as all wear-time hours during the day. The mean daily PA and ST outside preschool time, during preschool time and throughout the day were calculated on the individual level and then matched with data on organized sports participation.
Covariates
Anthropometry and family characteristics were also documented. Weight and height of participating children were measured using validated scales and stadiometers, respectively (scale: VB2-200-EC, Vetek AB, Väddö, Sweden, stadiometer: Seca 213, Seca, Chino, CA, USA). We used an age and sex specific international body mass index (BMI) classification by Cole et al (29) to classify children as normal weight, overweight or obese. Parents filled out a questionnaire on demographical variables such as number of siblings within the family and highest parental education level, categorized into elementary school, upper secondary school and university education.
Statistical analysis
Appropriate measures of variability and central tendency, mean and standard deviation (SD) for normally distributed variables or median and interquartile range (IQR) for variables with skewed distribution, are presented for various background characteristics and PA outcomes of the total study population and stratified by participation in organized sports.
We used linear mixed models, nested on preschool level, to examine associations between participation in organized sports with children’s daily levels of MVPA, LPA, steps and ST outside preschool time, during preschool time and throughout the day. Separate models were fitted for each activity intensity. In addition, we performed robust multilevel mixed-effects Poisson regression, nested on preschool level, to examine the likelihood of meeting the current World Health Organization (WHO) PA guidelines for children under five (30), of at least 60 minutes of daily MVPA, in organized sports participation with no participation in organized sports as reference level. All models were adjusted for confounders, factors that potentially influences both exposure (participation in organized sports) and outcomes, selected based on causal reasoning (31). The selected confounders were age of the children, sex, overweight/obesity status (32), accelerometer wear-time (33), number of siblings (34) and parental education (35). We calculated all p-values for both Poisson and linear mixed models using Wald tests, testing the coefficient of interest being equal to zero. To test for sex-specific associations, we performed a Wald test of the coefficient of a product interaction between sex and organized sports participation and stratified our main analysis.
In sensitivity analyses, we compared the descriptive characteristics between the analytical dataset (n=290) and excluded observations (n=104).
Raw accelerometer data was processed in Actilife version 6.13.3 and all statistical analysis were conducted in Stata version 14.2.