Physical activity and screen time of children and adolescents before and during the COVID-19 lockdown in Germany: a natural experiment

The impact of COVID-19 on social life has been drastic and global. However, the different numbers of cases and different actions in different countries have been leading to various interesting yet unexplored effects on human behavior. In the present study, we compare the physical activity and recreational screen time of a representative sample of 1711 4- to 17-year-olds before and during the strictest time of the first COVID-19 lockdown in Germany. We found that sports activity declined whereas recreational screen time increased. However, a substantial increase in habitual physical activities leads to an overall increase in physical activity among children and adolescents in Germany. The effects differ in size but not in their direction between age groups and are stable for boys and girls. We conclude from this natural experiment that physical activity among children and adolescents is highly context-driven and mutual and does not act as a functional opposite to recreational screen time.

Sports activity. Organized sports at school and in sports clubs were shut down during the lockdown which leads to no measurable organized sports during the peri measurement. Table 2 shows the average daily minutes spent doing organized and nonorganized sports, and the total amount of sports pre and peri the lockdown. A significant main effect of the lockdown on nonorganized sports was found (F 1, 1601 = 626.3, p < 0.01, p. η 2 = 0.281). Participants reported an increase in nonorganized sports of 17.7 min per day. Additionally, an interaction between the lockdown and age was found for nonorganized sports (F 3, 1601 = 9.8, p < 0.01, p. η 2 = 0.018). 14-to 17-year-olds increased their nonorganized sports by 17.0 min per day whereas 4-to 5-year-olds increased it by 11.8 min. For the total amount of sports, we found a main effect of the lockdown (F 1, 1644 = 122.3, p < 0.01, p. η 2 = 0.069) in a negative direction. Participants reported a decrease in the total amount of sports of 10.8 min per day, and an interaction between the lockdown and age (F 3, 1644 = 8.6, p < 0.01, p. η 2 = 0.015) was significant: 14-to 17-year-olds reduced their total amount of sports by 15.6 min per day whereas 4-to 5-year-olds reduced it by 2.2 min.

Discussion
Due to consecutive tracking of the PA, fitness, and health of children and adolescents in Germany since 2003, the MoMo study had the opportunity to rely on longitudinal PA data from a representative national cohort directly before and during the COVID-19 pandemic.
Our results show a decline in SA among boys and girls of all age groups paralleled by an increase in HA and recreational ST. In summary, the negative effects of the lockdown were stronger pronounced among adolescence, which showed a larger decline of SA and a lower increase in HA compared to younger children. A decline in SA is in line with other studies that tracked PA during the COVID-19 lockdown by questionnaires in Canada 24,26, China 20 , Spain 27 , as well as by step counts in Italy 23 and Canada 22 . In contrast to SA, the adherence to the PA guidelines and total PA increased among children and adolescents in Germany, which is in line with a study from Belgium 28 but in contrast to studies from other countries [20][21][22][23][24]26,27 . The different behaviors of children and Table 3. Habitual physical activity before and during the COVID-19 lockdown in Germany (MoMo Study). m male, f female, Ø mean of males and females, M mean, s standard deviation, 95% CI 95% confidence intervals, d effect size Cohen's d. * / **Significant difference (*p < .05; **p < .01).   29 , whereas Belgium had only mild restrictions. On the other hand, we assume that differences may be caused by methodological issues of tracking PA. For example, the questionnaire we used to track PA differentiated between a multitude of settings and activities and therefore we were able to analyze and contrast shifts among different types of PA and different settings. In our sample, the increase in guideline adherence, which stands in contrast to the 28.5 min decline in organized SA, was not only initiated by an increase in nonorganized SA of 17.7 min per day but ultimately achieved by a 36.2 min increase in daily HA. Data from the US confirm that nonorganized PA and outdoor play were highly relevant during the COVID-19 pandemic, especially among younger children 30 . We found a change in total daily PA from 142.2 min pre to 167.8 min peri the lockdown in our study. In contrast, the study among 6-to 17-yearolds from Shanghai reported a change in total daily PA from 77.1 min pre to 15.0 min peri the lockdown 20 . PA behavior is a multi-dimensional construct with fluid settings. The different reactions to the pandemic all over the world showed that many children, adolescents, and adults can compensate for the elimination of settings within weeks. Nevertheless, being able to go outdoors safely with rules of social distancing and contact to at least one known person was the most striking difference in the execution of lockdowns between countries and seems crucial in maintaining sufficient levels of PA among children and adolescents.

Pre (M ± s) Peri (M ± s) Pre (M ± s) Peri (M ± s) Pre (M ± s) Peri (M ± s) Pre (M ± s) Peri (M ± s) Pre (M ± s) Peri (M ± s)
Interestingly, an increase in nonorganized sports was not limited to those who were already doing sports before the lockdown in our study. 30.2% of those who had reported no SA before the lockdown engaged in nonorganized sports during the lockdown whereas only 60.3% of those who had engaged in SA in sports clubs before the lockdown transferred their organized to nonorganized SA. Within-person analyses pre to peri the lockdown showed that the mean correlation coefficients were substantially lower for nonorganized sports (r = 0.13, p < 0.01) compared to HA (r = 0.43; p < 0.01), days active (r = 0.41; p < 0.01), or recreational ST (r = 0.69; p < 0.01). These results are supported by a cross-sectional study with 13,515 adults from Belgium 28 . The data from Belgium shows a general increase in exercise frequencies, but exercise response to the lockdown differed between active and nonactive people before the lockdown: Of those who had exercised at least once a week before the lockdown, 23% reported less and 36% reported more exercising during the lockdown. In contrast, 58% of those who had exercised less than once a week before the lockdown reported more exercising during the lockdown 28 . Most likely, a more pronounced focus on a healthy lifestyle during the pandemic triggered this behavior in countries where policy restrictions allowed exercising. We also analyzed the distribution of PA and ST among a multitude of socioeconomic variables and found a slightly smaller impact of the socioeconomic status on PA and ST peri compared to pre lockdown. This may be explained by the fact that children and adolescents from families with lower socioeconomic status engage less often in organized and more often in nonorganized SA compared to those from households with a higher socioeconomic status 11 . The Gini coefficient 31 , which has recently been used to describe inequalities among the PA distribution between countries 32 , decreased from 0.80 to 0.70 for SA and from 0.47 to 0.39 for recreational ST but increased from 0.35 to 0.40 for HA among children and adolescents during the lockdown in Germany. A decrease in the Gini coefficient for SA and ST can be interpreted as a step towards a more equally distributed amount of SA and ST among children and adolescents as entry barriers to organized SA played no role and the demand for a healthy lifestyle was high during the lockdown. We can derive from this that countries and policies need to enable access to nonorganized PA to all children and adolescents during potential future lockdowns. Especially for adolescents, internet interventions from sports clubs are one www.nature.com/scientificreports/ potential option but scientists should also focus on upcoming affordable virtual and augmented-reality solutions to enable comprehensive PA opportunities for children and adolescents from all social classes, especially in urban areas. In reaction to the drastic decrease in PA among children and adolescents in China 20 , authors from the Shanghai University of Sport recommended exercising at home but also mentioned the danger of prolonged stays at home which can increase behaviors that lead to inactivity and contribute to anxiety and depression 29 . In the context of the studies from other countries mentioned before, results speak for the fact that the current technical possibilities are not adequate to comprehensively maintain sufficient levels of PA at home. Until there are virtual options for PA that satisfy the needs of users to an extent that they are comparably motivational and continual to analog forms of PA, policies should focus on providing safe natural space for nonorganized outdoor activities in addition to organized offers. The finding that children and adolescents in Germany spent more active time peri than pre lockdown may be explained by the simple fact that they had more recreational time to do so, but also theoretically by self-determination theory 33 and a more pronounced focus on health. Studies show that boys get most of their moderateto-vigorous PA at school during recess and non-structured times 34 . During non-structured days, children and adolescents can choose PA behaviors that allow them to develop mastery, skills, and competence and can satisfy the need that they are in control of their behavior which leads to more motivation for PA 34 .
The increase in recreational ST among children and adolescents in Germany is in line with data about leisure screen time in Spain 27 and China 20 . The study from Spain found an increase in ST between 132 (3-5-year-olds) and 198 (13-16-year-olds) minutes per day and a significant increase in overall sleep time among the 13-16-yearolds of 36 minutes per day during strict COVID-19 confinement. The Chinese study shows an increase in daily recreational ST among 6-to 17-year-olds from 24.3 to 64.3 min and an increase in daily total ST from 87.1 to 334.3 min. Adding nonrecreational screen media usage to our data, we find an increase from 146.1 to 237.5 min in total ST pre to peri the lockdown and an increase from 215.1 to 338.7 min in daily total media usage (including reading and listening to music). Home confinement restricted opportunities for children in China more than in Germany, and the fact that, as of reports by free journalists 35 , homeschooling started much earlier in China (March 2, 2020) than in Germany (May, 4, 2020) explains the pronounced increase in total ST in China. The numerically larger increase in ST with increasing age we found is supported by a German study that interviewed more than 8000 parents of children aged 3-15 years during the lockdown and found an increase of 38% (kindergarten) to 72% (secondary education) among recreational ST 36 . An increase in ST paralleled by an increase in PA among children and adolescents in Germany during the COVID-19 lockdown adds a piece to the puzzle of the relationship between PA and ST. Although it may be against common understanding at first glance that PA and ST are mutually exclusive in youth's recreational time, a meta-analysis of 163 studies about the associations between sedentary behavior and physical activity in children and adolescents found a global correlation of only r = 0.11 between PA and ST 37 . The authors stated that the activities do not behave as functional opposites and substitute each other only marginally. Cluster analyses of the MoMo baseline sample confirmed that there are a multitude of different behavioral patterns concerning PA and recreational ST including children and adolescents with very large amounts of PA and ST and some that do not report any PA but also report low amounts of ST at the same time 38,39 . Other studies support the idea of ST and physical inactivity being different risk factors for an unhealthy lifestyle [40][41][42] . In the cluster-randomized, controlled trial "Switch-Off 4 Healthy Minds", the authors found that a successful reduction of ST did not lead to a meaningful increase in PA among adolescents 43 .
Despite the comprehensive approach of tracking PA in Germany within the MoMo study, there are some limitations to this study. First, all our results are based on self-reports. Our questionnaire has been tailored to the different PA settings in Germany with a higher focus on recall bias compared to the relatively short and unspecific, yet economical, GPAQ 44 , which has been used for example in the Chinese study 20 . We also use device-based measures (Actigraph GT3X) to validate and augment our PA data in MoMo, but decided that it is not ethically justifiable to use them on a large sample during the lockdown. Device-based studies of PA during the lockdown are scarce and the little data that have been published so far rely on nonrepresentative ad-hoc samples, for example, children with congenital heart disease 22 or already active people 45 . The largest sources of published device-based measured PA peri the lockdown are country-stratified data from Garmin wearables 45 and a smartphone app 46 . The Garmin data show that different activities such as walking or outdoor and indoor cycling changed differently among countries, with most of them increasing 45 . The study which analyzed step counts from a smartphone app confirmed substantial differences between countries but reported decreases in step counts within 30 days of the pandemic from 6.9% in Sweden to 48.7% in Italy with a mean decrease of 27.3% 46 . Second, as it is a natural experiment, there is no control group and we can only assume that the lockdown was causal for the changes in PA and ST behavior. Since a full representative MoMo study takes 2 years to be completed, the mean age peri exceeded the mean age pre by almost exactly 1 year (Table 5). Since studies show that especially nonorganized PA declines during maturation in humans 9,16,47,48 as well as in most nonhuman species 49 , this may have led to a small underestimation of the observed increase in PA in our study.
The weather was untypically warm in April 2020 in Germany. A mean temperature of 10.4 °C (9.6 °C in 2019) and on average 292.4 sunshine hours (227.9 in 2019) were measured across Germany in April, 2020 50,51 . Warmer temperatures are associated with more PA and less ST in mild climates 52 . We also analyzed our data for seasonal effects. When considering only those MoMo participants who were tested in April 2019, the effects on PA and recreational ST were stable. Nonorganized sports increased by 15.9 min per day, HA by 46.2 min per day, recreational ST by 49.5 min per day, and WHO guideline compliance increased from 17.6 to 33.7%.
Lastly, the representativeness of our longitudinal pre-peri sample is limited because of the unforeseen COVID-19 outbreak during the collection of the representative pre-study sample. However, 120 out of 167 initial sample points across Germany were reached during the pre-study. Since the survey design includes a circle concept which ensures that sample points with different environmental contexts are tested during different months of the year, the reduction in representativeness because of missing sample points can be considered as small. The Scientific Reports | (2020) 10:21780 | https://doi.org/10.1038/s41598-020-78438-4 www.nature.com/scientificreports/ main effects of the lockdown remained stable after stratifying the sample by potential influencing socioeconomic factors such as housing situation, region within Germany, socioeconomic status, and the number of siblings. What we can conclude from this natural experiment is that in some individuals the insufficient levels of PA are context-driven (e.g. too much forced sitting during school, homework, and other organized sedentary activities) and will vanish, once they regain autonomy over their everyday life. It is striking that in countries with suitable circumstances like in Belgium 28 , people found their way to PA during the lockdown. Phases with high alertness and focus on health may be a chance to positively interact with people's preventive behavior.
The fact that some children and adolescents did not engage in PA peri the lockdown and that the decline in SA is more pronounced in older age groups is alarming. Among European countries, there is a weak agreement that total PA among children and adolescents declined during the 1980/1990s, however comparable data with high standards regarding the methodology is missing among studies from that time 47 . This trend is followed by a slight increase in PA and a substantial increase in ST during the 2000s in Germany 16,53,54 and most European countries 48,55 , leading to the stagnation of total PA with a shift from nonorganized to organized settings during the 2010s 11 . The fact that the recommended 60 min of daily moderate to vigorous PA nearly doubled among children and adolescents during the COVID-19 lockdown after having been stable for two decades is striking and needs to be analyzed further from a scientific yet political point of view.
The special circumstances during the COVID-19 lockdown also told us that PA and recreational ST behavior do not act as functional opposites and may decrease or increase concurrently 43 . Although there are cases where youth show very high levels of recreational ST, sometimes even diagnosed as gaming disorder, which can lead to an almost complete denial of PA 56,57 , the majority of studies among healthy or representative samples report a weak relationship between PA and ST or other forms of media usage among children and adolescents [37][38][39][40][41][42] and speak for the fact that issues among both behaviors need to be addressed separately.

Methods
We obtained self-reported data from 1717 participants at two measurement points before (pre) and during (peri) the COVID-19 lockdown in Germany to examine the influence of the lockdown on PA and ST of children and adolescents aged 4-17 years pre. The STROBE statement 58 guided the reporting of this study.
Procedures. The data were obtained within the framework of the MoMo study 59 . The study follows a cohort sequence design, which means that in addition to a longitudinal observation of the MoMo participants, a new representative sample of children and adolescents living in Germany is recruited at each follow-up. The baseline data was collected from 2003 to 2006, Wave 1 was conducted between 2009 and 2012, and Wave 2 between 2015 and 2017 59 . Wave 3 started in August 2018 and was planned to be finished in June 2020, but had to be interrupted in March 2020 because of the COVID-19 lockdown in Germany. During the strictest time of the lockdown (20/04/2020-01/05/2020), we asked every participant of Wave 3 (pre-study) to answer our questionnaire again (peri-study) to track the direct changes in behavior (c.f. Fig. 2).
The study was conducted in accordance with the Declaration of Helsinki. For MoMo, ethics approval was obtained by the Charité Universitätsmedizin Berlin ethics committee, by the University of Konstanz, and the ethics committee of the Karlsruhe Institute of Technology. The Federal Commissioner for Data Protection and Freedom of Information was informed about the study and approved it.
For the pre-study, parents and children were invited to examination rooms at central locations in the proximity of their homes in the 167 cities and municipalities in which the study was conducted. Parents gave their written consent for minors and the presence of a legal guardian was mandatory under the age of 15. Questionnaires were filled out on-site by the participants on laptops (pre) or online (peri). Participation in the study was voluntary and every participant received a gift worth 20 €. The participants or their custodians were informed about the contents of the study and about data protection and gave their written consent. During the COVID lockdown in Germany, the participants were contacted via email and were asked to answer the peri-questions online.
Participants. The MoMo participants for Wave 3, respectively the pre-study, were selected based on a nationwide multi-stage sampling approach with two evaluation levels 60 to maximize representativeness. First, a systematic sample of 167 primary sampling units was selected from an inventory of German communities stratified according to the classification system that measures the level of urbanization and geographic distribution 61 . The probability of any community being picked was proportional to the number of inhabitants younger than www.nature.com/scientificreports/ 18 years of age in that community. Second, based on the official registers of local residents, an age-stratified sample of randomly selected children and adolescents was drawn. 2722 MoMo Wave 3 participants (preliminary response 25.2%) were contacted for the peri-study and data from 1711 participants were gathered. Twenty-three e-mails could not be delivered because of incorrect addresses, and a total longitudinal response of 63.6% was achieved. Table 5 shows the characteristics of the longitudinal sample and the ages pre and peri the lockdown.
Measures. The pre-lockdown data were derived via questionnaires (PA and ST) and interviews (socioeconomic data) on the pursuant screening site. For children under the age of 11 years, parents filled in the questionnaires together with the child. During the peri-study, the same questions about PA and ST were asked via an online version of the questionnaire except for the questions about organized PA which was shut down and forbidden by governmental law. We also changed each instance of time mentioned in the description of the questions during the peri-study to "during the Corona lockdown".
Physical activity. The MoMo PA Questionnaire (MoMo-PAQ) was used to assess PA via self-reported SA and HA in different settings (sports clubs, leisure time, and school) 62 . The MoMo-PAQ consists of 28 items and measures frequency, duration, type, intensity, and setting of SA and HA. The data obtained with the MoMo-PAQ are sufficiently reliable and valid (test-retest reliability: ICC ≥ 0.68) 62 . SA at school was assessed by two items about the frequency (times per week) of 45-min classes in curricular and extracurricular sports activities which are multiplied by a factor of 8.5/12 to correct for vacations. SA in sports clubs was assessed by three items which could be answered up to four times: type of sports club activity, duration (minutes per week), and time throughout the year (months per year). Minutes were then multiplied by months per year divided by twelve to control for periodic effects of SA. Nonorganized leisure-time SA was assessed in the same way by three items which could be answered for up to four types of sports: type, duration, and time throughout the year. These items were combined in an index that reflects the daily minutes with organized (school and sports clubs) and nonorganized SA. Types of sports that do not lead to an increase in energy expenditure using large skeletal muscles (for example esports) were not defined as SA according to the definition of PA 1 , and minutes spent were multiplied with zero. Nonorganized playing outside, gardening, and housework were each assessed separately by two 8-scaled items about days per week and minutes a day in which the participants pursued the activity ("On how many days do you normally play outside/garden/work in the household during a week?", "How long do you on average work/ play on one of those days?"). Walking and cycling were assessed by one item each about the daily distance participants travel by walking or cycling 62 .
We also asked the participants on how many days they are active for more than 60 min with moderate to vigorous intensity to reproduce the WHO PA guideline: "Over a typical or usual week (peri: "during the Corona lockdown"), on how many days are you physically active for a total of at least 60 min per day?" 63 .
Recreational screen time. ST was measured via self-reported screen-time behaviors which are commonly used to report ST in youth and which have similar directions with health outcomes as direct measures of sedentary behavior 3 . Participants were asked to report the time spent watching television (TV), playing games on any device (Gaming), and using the internet for recreational use (Internet) separately for weekdays and weekends using an 7-point scale including (almost) never, 15 min per day, 30 min per day, 1 h per day, 2 h per day, 3 h per day, and 4 h per day 64 . An index reflecting weekdays and weekend days at a 5:2 ratio was then calculated for each activity. www.nature.com/scientificreports/ Sociodemographic variables. Sociodemographic variables were assessed within the pre-study, including age, sex, and a multitude of socioenvironmental and socioeconomic variables. A detailed description of the MoMo survey instruments can be found elsewhere 65 . Statistics. All statistical tests were conducted using IBM SPSS 25 (IBM Corporation, Armonk, NY). Statistically significant (two-sided) differences pre and peri were determined via ANOVA for repeated measurements where the effect of time represents the lockdown effect and 95% confidence intervals for differences of units and proportions 66 , as well as effect sizes Cohen's d for pre-to-peri differences in the form of d = |Diff| σ(Diff) were reported. Means and standard deviations were reported for PA in different settings. Means of difference and 95% CIs were reported for overall SA, HA, and ST.

Data availability
The datasets generated and analyzed during the current study are not publicly available due to the strict ethical standards required by the Federal Office for the Protection of Data with which study investigators are obliged to comply but are available from the corresponding author on reasonable request.