June 18, 2024

Healthy About Liver

Masters of Health

COVID-19 induced changes in physical activity patterns, screen time and sleep among Swedish adolescents – a cohort study | BMC Public Health

COVID-19 induced changes in physical activity patterns, screen time and sleep among Swedish adolescents – a cohort study | BMC Public Health

Data for this study comes from the project “Physical activity for healthy brain functions in school youth”, a cohort study with measures from September 26th to December 6th 2019 (baseline) and follow-up from April 12th to June 9th 2021. The study was performed in Stockholm, which is the capital of Sweden. The school system in Sweden is deregulated and allows the parents to freely choose between independent and public schools for their children, without paying any fees. The aim for the baseline study was to include at least 1000 participants to be able to explore associations between mental health and physical activity. A convenience sample was used. Schools within a 2–3 hour driving distance from Stockholm, Sweden (ranging from 2 to 200 km away) were invited to participate (n = 558). Recruitment was performed from March to June of 2019. Schools with a sports profile, with less than 15 students in grade 7 (13–14-year-olds) and schools with a non-Swedish speaking student population were excluded. Of the 84 schools that showed interest in participating in the study, 40 were chosen to maximize variation in the type of municipality and socioeconomic background of the schools. Six schools dropped out before participating, leaving a total of 34 schools with 1–4 classes of 7th graders participating in each school. Of these 11 (32{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161}) were independent schools and 23 (68{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161}) were public schools and a total of 22 municipalities are represented in the sample, with 44{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161} being located in large cities and municipalities near large cities, 47{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161} medium-sized towns and municipalities near medium-sized towns and 9{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161} in included smaller towns/urban areas and rural municipalities. All students from participating classes were invited to participate (n = 1556) and a total of 1139 participated (participation rate 73{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161}). Teachers and principals at each school handled the recruitment of the students in the fall of 2019 and gathered consent forms. The schools were contacted in March 2020 and asked to assist in distributing the questionnaires and accelerometers; 6 schools did not have the time or resources, so questionnaires and accelerometers were sent to the students in these schools directly by e-mail and post. A total of 585 participated at follow-up i.e. either answered the questionnaire or had at least one valid day of accelerometer measurements (51{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161} participation rate), see Fig. 1. The study was pre-registered (ISRCTN15689873) on 17/06/2021.

Fig. 1
figure 1

Flow chart of the data collection

Data collection

At baseline, the participants came in groups of at most 60 students per day to the Swedish School of Sport and Health Sciences. The participants arrived in the morning, were informed about the data collection procedures. They answered an electronic questionnaire, had their body composition, height and weight measured, performed a sub-maximal fitness test and were given an accelerometer for recording physical activity to be worn for 1 week. The data collection finished at noon. School schedules of all participating classes were collected.

At follow-up, questionnaires were sent electronically to the participants. In case of non-response after electronic reminders, paper questionnaires were sent out to the home address. Accelerometers were sent out to the schools that agreed to assist with the data collection and given to those willing to participate again. For non-participating schools, the accelerometers were sent directly to the participants after they had answered the electronic questionnaire. The accelerometers were again worn for 1 week. New school schedules for the spring of 2021 were collected. On average 18.4 months (sd 1 month) passed between baseline and follow-up.

Exposure variables

Background variables

In the baseline questionnaire, participants indicated their gender and their parents’ country of birth, classified as “both born in Sweden”, “one born outside and one in Sweden” and “both born outside Sweden”. Register data on parental education for each participant was collected from Statistics Sweden and classified as “12 years or less” and “more than 12 years”, chosen as primary and secondary school combined lasts 12 years in Sweden.

Anthropometric measures

Height and weight were measured at baseline using a stadiometer (SECA 5123, SECA Weighing and Measuring Systems) to the nearest mm and a calibrated scale (Tanita BC-418, Tanita Corporation, Tokyo, Japan) to the nearest 0.1 kg. Body mass index (BMI) was calculated (kg/m2) and categorised according to sex- and age-specific values from the International Obesity Task Force (IOTF) 2012 [16]. BMI was dichotomised as normal/underweight and overweight/obese. Furthermore, BMI standard deviation score (sds) was calculated, which adjusts for gender and age [17].

Body composition, reported as body fat percentage, was measured with bioelectric impedance analyses (BIA) using a Tanita body composition analysis scale (Tanita BC-418, Tanita Corporation, Tokyo, Japan).

Cardiovascular fitness

The Ekblom-Bak sub-maximal cycle ergometer test was used to estimate cardiovascular fitness (CVF) [18, 19]. To adjust for maturation, the values for prepubertal boys were calculated using the equation for girls [18]. Assessment of puberty was done using the Tanner scale, [20, 21] and boys at stages 1–2 were defined as prepubertal; see Kjellenberg and colleagues [22]. Change in heart rate response to a change in sub-maximal work rate was used to estimate maximal oxygen consumption and expressed as relative values (millilitres of oxygen per kilogramme of body weight per minute, ml/kg/min).


Physical activity and sedentary time

Triaxial accelerometers (Actigraph model GT3X+, LCC, Pensacola, FL, USA) were used to collect information on time spent physically active and sedentary. Participants wore the accelerometer on their right hip for seven consecutive days during all waking hours, except during water-based activities, e.g., showering. At baseline, the first of these 7 days was the day after the participants visited the School of Sport and Health Sciences, while at follow-up either the teacher or the participant indicated the start day. Data were collected at a sampling rate of 30 Hz and were processed in 5-s bouts for analyses using the programme ActiLife version 6.13.3. Non-wear time was classified as 60 or more consecutive minutes of 0 counts, with zero spike tolerance using triaxial data. A filter based on each participant’s reported bedtime and wake time in the questionnaire was used to separate sleep time from wake time. A valid day was classified as having at least 500 minutes of data after removal of non-wear time and application of the filter.

Additionally, a filter based on school schedules was applied to separate schooltime from leisure time during weekdays. Results are reported for 1) weekdays 2) schooltime, 3) leisure time during weekdays and 4) weekends. Each participant needed to have two valid weekdays of measurement at both baseline and follow-up to be included in the results for the first three domains, while one weekend day was required to be included in the results for weekends. The data were categorised into sedentary time (0–100 cpm), LiPA (101–2295 cpm) and MVPA (≥2296 cpm) [23].

Change in PA and sedentary time was calculated by subtracting baseline values from follow-up values. In total. 12 variables were created based on the accelerometer data by using the three intensity levels (sedentary, LiPA, MVPA) across the four time domains (weekdays, school time, leisure time during weekdays, and weekends). All variables are reported as average minutes per day.


The participants were asked to report their wake and sleep times during weekdays and weekends. The response options were in 30-minute intervals and were used to calculate sleep duration. Change in sleep duration was calculated by subtracting sleep duration at baseline from that at follow-up and is reported in minutes, separately for weekdays and weekends.

Screen time

Screen time was determined using the questions “During a normal weekday/weekend day, outside of school, how much time do you spend looking at a screen (not including schoolwork)?” The response alternatives were “no time at all”, “less than 1 hour”, “1–2 hours”, “3–4 hours”, “5–6 hours”, and “7 hours or more”. For ease of comparison with the other outcome variables, the response options were simplified to reflect the midpoint number of minutes i.e. 1–2 hours became 90 minutes. No time at all was counted as 0 minutes, less than 1 hour as 30 minutes and 7 hours or more as 420 minutes. Change in screen time was calculated by subtracting screen time at baseline from that at follow-up and is reported in minutes, separately for weekdays and weekends.

Statistical analyses

Descriptive data were presented as percentages for categorical variables and means with standard deviations for continuous variables. Differences between participants and drop-outs at baseline were tested using Chi2 tests and independent sample t-tests. As the clusters (schools) explained a much of the variation in outcome variables, multilevel linear regression analyses were performed. The models were applied on two levels: the school and the individual, and a random intercept was modelled for each school. Unstandardised betas (β) are reported with 95{6f90f2fe98827f97fd05e0011472e53c8890931f9d0d5714295052b72b9b5161} confidence intervals (CI). Residuals were plotted and checked for normality. A bootstrapping estimation was performed to produce sensible estimates for standard errors in the multi-level model, as suboptimal fitness of model or heteroscedasticity was detected. Each exposure was explored separately for association with each outcome. The models for anthropometrics and fitness variables were adjusted for gender and parental education. Crude models are reported in supplementary tables 1–2. No imputations for missing data or loss to follow-up were performed. The level of significance was set to p < 0.05. The IBM SPSS Statistics for Windows software, version 26 was used for the statistical analyses.