Schoolchildren (7–12 years-old) from 2nd to 5th-grade in part-time public schools in Feira de Santana (Bahia) participated in this cross-sectional study. Feira de Santana is in the Northeast region of Brazil (inhabitants: 624,107; Human Development Index: 0.712). Data collection covered weekdays (Tuesday to Friday), from March to October of the year 2019 and included a probability sample of students from 2nd to 5th-grade, from public schools in the urban area, with broadband Internet. The sample size was defined based on the following parameters: a population of 15,920 students enrolled in the education system, according to data from the Municipal Department of Education; expected prevalence of outcomes of 50%; confidence limit of three percentage points; design effect (deff) of 2.0; and 95% confidence interval (95%CI). Based on these parameters, the sample size was calculated at 2,000 students. A further 20% was added to make up for presumed losses, resulting in a sample of 2,400 students (Fig. 1).
The cluster sampling process was carried out in three stages: I) all schools in the municipal network were stratified according to the 11 geographic and administrative centers of the Department of Education (clusters); II) one school from each center was randomly drawn; III) all classrooms from 2nd to 5th grade within each school were selected (159 classrooms), and all subjects within the selected classrooms were invited to participate in the study. All methods were carried out in accordance with relevant guidelines and regulations of ethical standards set out in Resolution No. 466/2012 of Brazil’s National Research Ethics Council. Informed consent was obtained from all participants involved in the study and their parents/guardians provided authorization in writing. The study protocol was approved by the Research Ethics Council of the State University of Feira de Santana (Approval No. 02307918.5.0000.0053, Opinion No.: 3.116.495). The Municipal Department of Education provided information regarding the sex, age, and school shift of participants.
Measurement of sedentary behaviors and physical activities
The participants self-reported the SB and physical activity on the Food Intake and Physical Activity of Schoolchildren (Web-CAAFE) questionnaire. The Web-CAAFE is a previously validated self-report questionnaire , completed on the internet and based on the previous-day recall. The instrument collects information on weight status, food consumption, physical activity, and SB and includes screens to evaluate physical education classes and to investigate modes of commuting to school.
Participants choose up to 32 items, out of a total of 50 stored in the system, which they had performed the day before across three periods (morning, afternoon, evening). The list contains five SB icons (one for academic tasks, e.g. reading, writing, drawing, painting; four electronic devices, e.g. TV, video game, computer, and cell phone), and 27 physical activity icons classified into: Active play (Play with a ball, Play catch, Soccer, Dance, Marbles, Jump rope, Gymnastics, Elastics, Play in the park, Play in the water/Swim, Ride a bicycle, Rollerblade/Skateboard/Ride a scooter, Fly a kite, Dodgeball, Hide and seek, Play with a dog, Hopscotch), Non-active play (Board games, Playing with dolls/action figures, Playing with toy cars, Spinning top/Bayblade, Listen to music, Play musical instrument), Structured physical activity (Ballet, Fight Sports), and Household chores (Wash the dishes, Sweep). Information on the weekly frequency of participation in physical education classes is assessed through the question “How many times a week do you take part in physical education classes?” (none, 1, 2 3, 4, every day of the week). The closed list of leisure activities, sports, home chores, and sedentary activities was compiled based on results from focal groups, previous instruments for this age range, and the 7-day recall completed by 180 schoolchildren .
Participants completed the Web-CAAFE at the school, after receiving verbal explanations about how the software works and how to complete the questionnaire. Students were instructed not to interact during the task and the research team helped when requested, without inducing responses.
The study included weight and height measurements to calculate the Body Mass Index (BMI), measured by trained researchers, following recommended standardization . Weight was measured using an AVAnutri® digital scale with graduation every 100 g and a maximum capacity of 200 kg. Height was measured using a portable stadiometer, detachable, with a square platform, Seca® brand, with a 205 cm maximum height and graduation every 1 mm. The students were barefoot, wearing school uniform, and with no headwear during measurements. Age-and sex-specific BMI z-scores were calculated according to the International Obesity Task Force (IOTF) . The weight status was categorized into non-overweight (underweight and normal weight), overweight, and obesity according to IOTF reference values .
Classification of economic level
Socioeconomic status was investigated based on the analysis of possession of items, education level of the head of the household, and access to public services, according to the Brazilian Economic Classification Criteria . The socioeconomic status was classified into classes, related to the average household income in Reais (R$): A (R$25,554.33), B-C (R$1,748.59 to R$11,279.14), and D-E (R$719.81). Based on the average dollar exchange rate between March and October 2019, income ranges in these classes were: A (US$ 6,485.87), B-C (US$ 443.80 to 2,862.72), and D-E (US$ 182.69).
Data processing and analysis
The weekly attendance in PE was the main exposure analyzed (0/week; 1/week; ≥ 2/week). Daily frequencies of active play, non-active play, and structured physical activity were the main outcomes (count outcomes). These frequencies were obtained by summing all reports in the morning, afternoon, and night. For example, if a participant reported riding a bike in the morning period, playing with a ball in the afternoon, and playing with a dog in the evening, then their sum was 3 counts of active play. SB frequency was obtained by summing the daily reports of academic tasks and screen use. DPA frequency was obtained by summing the daily reports of all physical activities.
Students with intellectual disabilities and ages outside the age group of seven to 12 years participated in the study but were excluded from the statistical analyses. Descriptive statistics are used to present the study variables. Variables without normal distribution after verification of the histograms and the Shapiro–Wilk test are described by median and interquartile range values. Differences in non-normally distributed continuous variables were evaluated using the non-parametric Mann–Whitney test (U). Categorical variables are described as absolute and relative values and compared using Pearson's chi-square test (Χ2).
The associations between weekly attendance in PE and frequencies of active play, non-active play, and structured physical activity were analyzed using the values of prevalence ratios (PR) and respective 95%CI estimated via multiple Negative Binomial Regression, with adjustment for age (7–9 years; ≥ 10 years), school shift (morning; afternoon), and BMI z-scores, adopting a robust variance estimation method. Negative Binomial models analyzing the association between weekly attendance in PE and DPA and SB were also adjusted by the daily frequency of household chores. The group of household chores was not included in the present analysis as an outcome because there is no evidence of an association with attendance in PE.
The Negative Binomial distribution is suitable for fitting count data susceptible to overdispersion. In addition, it showed higher linearity in the comparison between observed and predicted values of the outcome. The zero-inflation between the factors was assumed to be constant. Although the negative binomial regression models provide a measure of association such as Incidence-Rate Ratios (IRR), we adopted the prevalence ratio (PR) as the most appropriate way to present our results, considering the cross-sectional design of the study. Statistical significance was assessed using p value < 0.05. Effect modification was tested using interaction terms between weekly attendance in PE and sex, age, school shift, and BMI z-scores. Interactions that showed statistical significance at the critical value of p < 0.05 were described.