Study design
This study used a cross-sectional survey design to investigate the association of physical activity level and sport participation of adolescents with the personal, social, and ecological factors. After obtaining the necessary approvals from local governments and the ethical approval from Ethics Committee of Marmara University Faculty of Health Sciences (No: 239 Date: 19.12.2019), the study was conducted between February 2020 and March 2020. There were 39 secondary schools in Üsküdar district of Istanbul, and the Education Ministry provided a random selection of schools for data collection. To select the sample, the secondary schools in the district were listed via an electronic medium and 8 of these schools were randomly chosen. However, due to COVID-19 pandemic, data collection was only completed in 3 schools (including 996 participants) before the national shutdown.
Participants
Nine hundred ninety-six students aged 11–14 years were invited to take part in the study. Those who returned the informed consent form signed by their parents participated in the study. Students with orthopedic problems that prevent them from participating in physical activity, and those with any systemic, neurological, chronic diseases, or mental problems were not included in the study. The flow chart of the study was summarized in Fig. 1.
Procedure
Data was collected in the classroom by making face-to-face interviews with each student, under the supervision of the teachers. Information Form (Additional file 1: Appendix 1) and Child Physical Activity Form were used as outcomes. Information was obtained from the administration and teachers on the screening day about students with special conditions (who have a disability, inclusive student, etc.), their answers were obtained, but they were excluded from the study.
Outcomes
PAL and sport participation were measured by using Child Physical Activity Questionnaire (PAQ) and questioning sport-related habits of adolescents.
Child physical activity questionnaire
The questionnaire was developed to evaluate the physical activity level of primary schoolchildren aged 8–14, from the fourth grade to the eighth grade. The reliability and validity of the questionnaire have been well documented [15, 16]. The validity of the questionnaire in Turkish population was also reported in 2012 [17]. Each item of the questionnaire, except for the tenth question, which questions the disease status, is evaluated on a 5-point Likert scale and has an activity score between 1 and 5. “1” indicates low physical activity, “5” indicates high physical activity. The total score of the survey is 1–9. It is calculated by summing up the scores of the answers given to the question and dividing it by the number of questions. A cut-off point of 2.75 was used to identify adolescents who are active (a score of 2.75 or more) or inactive (a score of less than 2.75) [18].
Sport participation
The type of sport activity, duration, and frequency were questioned to decide regular sport participation. Adolescents were classified as regular sport participants if they involved in one of the previously identified sport activity at least once in a month [19] or not regular sport participants (i.e., those who did not involve in one of the previously identified sport activity at least once in a month).
Predictors
Personal factors
Personal factors included age (continuous), gender (categorical: male or female), sleep time (continuous: calculated as hours spent sleeping on average per a day), and screen time (continuous: calculated as hours spent in front of the TV, computer, tablet or phone). BMI z score (BMIz) (continuous) was measured by adjusting weight for participants’ age and sex [20]. Siblings (categorical) was classified as “Yes” if the adolescent had at least one sibling; if they did not, it was classified as “No”.
Ecological factors
Playground (categorical) was classified as “Yes” or “No” depending on presence of park or playground in the adolescent’s neighborhood. Type of school transportation (categorical) were classified based on the type of transportation they use to get to school as “Physically active (e.g., walking, cycling)” or “Physically inactive (e.g., using bus).”
Social factors
Family income (categorical) was grouped into three categories as “Lower”/”Middle”/”Higher.” Family activity time (categorical) was classified as “Yes” or “No” depending on if adolescent spends time with family for physical activity. Adolescent preference for school breaks (categorical) was classified considering the type of activity that adolescents prefer at their break time as “active (e.g., playing tag)” or “inactive (e.g., sitting).”
Data analysis
Descriptive statistics were used to summarize the demographic information of the participants, and all performance scores. The normality of data was visually evaluated by histograms, and Quantile–Quantile plots; and tested using the Shapiro–Wilk test. The observed outliers were removed from the data to improve the normality of the data. In the condition where data was not normally distributed after outlier removal, the log transformation was done for continuous variables.
Before the main analysis, the collinearity among independent variables were checked through variance inflation factor (VIF). Collinearity was determined to be present when the variance inflation factor was over 5 [21].
The two outcomes (physical activity level and sport participation) were regressed against 6 personal (age, gender, BMIz, sleep time, screen time, and siblings), 2 ecological (playground and school transportation choice), and 3 social (adolescent preference, family income, and family activity time) independent variables using logistic regression analysis. The binary variables (gender, adolescent preference, playground access, school transportation, sibling, family activity time) and ordinal variable (family income) were included into the regression analysis. The data was transformed into dummy variables with being female, being inactive, absence of a playground, using inactive mean of transportation, lower income, absence of a sibling, and lack of active time with family as the reference values.
The model was inspected visually for linearity, heteroscedasticity, and normality of the residuals, and goodness of the fit was evaluated using Hosmer-Lemeshow goodness of fit test, which is a Chi-square test conducted by dividing the sorted set into g=10 equal-sized groups [22]. Our previous study is consistent with the previous studies in the literature which showed that the inactivity rate in adolescents was around 80% in Turkish population [1, 2]. Based on this rate, a sample of at least 748 was needed to obtain 99% power with a confidence level of 95% and 5% Type 1 error, which is lesser than the current sample of 996 adolescents. All statistical analysis was done using R statistical software (Version 3.6.0, St. Louis, Missouri, USA), the package “ResourceSelection” [23]. The alpha level was .05.