Early effects of the COVID-19 pandemic on physical activity and sedentary behavior in children living in the U.S.

A prospective survey design assessed the effects of the COVID-19 pandemic on PA among U.S. children by using online parent-reported surveys. A baseline online survey was completed between April 25 – May 16, 2020, and a second online survey is scheduled to occur within 6–12 months. The current analyses focus on data collected from the baseline online survey.

A convenience sampling strategy, focused on recruiting a general population of parents living in the U.S. during the COVID-19 pandemic, was utilized. To avoid in-person interactions, potential respondents were electronically invited through various social media platforms (e.g., Facebook, Twitter) and university-based email list servs of students, faculty, and staff. Inclusion criteria were as follows: 18 years or older, able to speak and read English, live in the U.S., is a parent or legal guardian of a child between the ages of 5–13, greater than or equal to 50% of child’s custody resides with the parent/legal guardian, and planned parental custody for the next 12 months. Individuals were directed to an online screening form directly from an email or social media post. Once eligibility was determined, individuals agreed to participate through an anonymous online information sheet describing the study procedures, risks, and benefits. The Institutional Review Board of the University of Southern California determined the study procedures presented no more than minimal risk and approved it as exempt from full review.

Participants completed the online screening form, information sheet, and baseline survey in English through an online survey platform, Research Electronic Data Capture, which conforms to HIPAA security requirements for the protection of data [24, 25]. Baseline data collection took place over 22 days (April 25 – May 16, 2020). The baseline survey took approximately 20 min to complete. Participants had the option to complete it either on their mobile phone, tablet, or desktop device. Each parent reported on one child of theirs who was between ages 5–13. If a parent had more than one child between ages 5–13, they were asked to only report on the one child whose birthday occurred next chronologically after the date the survey was administered. Upon completion of the baseline survey, participants were eligible to be entered into a lottery to win one of ten $50 gift cards.

The current analyses utilized data on child’s past day PA and SB, parents’ perceived changes in level of PA and sitting time, locations of child’s PA, child’s use of remote/streaming services, and demographics.

Parents completed a measure of their child’s previous day PA created for the purposes of capturing non-school based activities frequently occurring during COVID-19, modeled upon the structure and format of previous day PA measures used in youth [26]. Parents also completed a measure of their child’s previous day SB with an instrument used in the “Active Where” survey [27]. The instructions asked parents to indicate how much time on the previous day their child performed each from a list of 11 types of common non-school-based PA and 12 common types of non-school-based SB for children. They were asked to think about the time that their child was awake on the previous day. The SB instrument has demonstrated acceptable intraclass correlations and test-retest reliability in previous research [27]. The specific types of PA included: sports practice or training; activity classes/lessons; free play or unstructured PA; jogging or running; biking; scootering, skateboarding, or roller skating; swimming; going for a walk; circuit training or conditioning; weightlifting, or other (write-in). Types of SB included: watching television, videos, or movies; playing computer or video games; using the internet, emailing, or other electronic media for leisure; doing school-related video calls; doing video calls; doing school-related work; sitting while listening to music; sitting while talking on the telephone or texting; sitting while hanging out or talking with friends or family in person; reading a book or magazine NOT for school; doing inactive hobbies; and riding or driving in a car. Duration in each activity was reported. After completing the measures, parents were asked to indicate how similar their child’s level of PA or sitting YESTERDAY, respectively, was compared to the past 7 days.

The Youth Compendium of Physical Activities (Youth Compendium) was utilized as a guide to calculate metabolic equivalents (METs) for each of the 11 types of PA. METy values for the age group 6–9 were applied for children who were 5 years old. If the type of PA in the baseline survey was not explicitly stated in the Youth Compendium, the METy values for the activity that most closely resembled the type of PA were used (e.g., the METy values for “jumping jacks” were utilized for the survey item “circuit training/conditioning). Five items on the survey were not explicitly stated in the Youth Compendium. If a participant wrote in an activity for ‘other’, METy values were inputted for each individual case. To calculate MET-minutes (MET-mins) for each of the types of PA, each age-group specific mean METy value was multiplied by the number of minutes parents reported their child performed in the previous day. Daily total MET-mins of PA were calculated for each child by taking the sum of MET-mins across the 11 specific types of PA.

Children’s previous day sitting/SB was examined by calculating the minutes spent in each of the 12 types of SB. Daily total minutes of SB were calculated for each child by taking the sum of minutes across the 12 types of SB. If the daily total minutes exceeded 1080 min, the value was truncated to 1080 min to address outliers. Daily total minutes of SB was also separated into two categories: minutes of school-related sitting (i.e., doing school-related video calls and doing school-related work) and minutes of sitting for leisure (i.e., all other types listed in the baseline survey).

Parents were asked to compare their child’s current levels of PA (i.e., past 7 days) and SB to the pre-COVID-19 period (i.e., February 2020). Specifically, they were asked, “Compared to Feb 2020, how physically active has your child been/how much sitting has your child been doing in the PAST 7 DAYS?” Response options were reported on a 5-point likert scale ranging from a “much more physically active in past 7 days as compared to February 2020,” to “much less physically active in past 7 days as compared to February 2020.”

Parents were asked to report the types of locations in which their child did PA in February 2020 and over the past 7 days (i.e., “where did your child do physical activity?”) with instructions to choose all that apply from the following options: inside my home or garage, in my yard or driveway, on the sidewalks and roads in my neighborhood, on the sidewalks and roads outside my neighborhood, gym or fitness center, at a park or trail, at an indoor sports facility (e.g., basketball/tennis court, ice rink), or at an outdoor sports facility (e.g., basketball/tennis court, baseball diamond). These settings were based upon locations assessed in the “Active Where” survey [28, 29] and correspond to where children frequently engage in physical activity based upon Global Positioning Systems (GPS) [30].

Parents reported whether their child used remote or streaming services to participate in PA during the during the early-COVID-19 period. Specifically, they were asked, on how many days of the past 7 days did their child participate in any team sports training sessions or practices, activity classes or lessons classes or sessions provided by a health club or gym through remote services, such as streaming classes via the internet or mobile applications.

Parents reported on their child’s biological sex at birth (male vs. female), birthdate, grade in school, ethnicity (Hispanic vs. non-Hispanic), and race (coded as American Indian or Alaska Native, Asian, Black, Native Hawaiian or Pacific Islander, White, Mixed race, Other). Parents also reported on their own birthdate, gender, marital status (coded as married vs. not married), employment status (coded as works full-time vs. does not work full-time), and annual household income (categorized as less than $24,999, $25,000–$54,999, $55,000–$94,999, $95,000 or more). All survey questions allowed participants to choose the option ‘do not know/prefer not to answer’.

Prior to data analyses, variables were screened for violations of statistical assumptions (e.g., normality, linearity). Variables representing the duration of participation in specific types of physical activities were highly skewed due to the substantial number of children who did not perform any type of activity. Therefore, these variables were coded as some vs. none for subsequent analyses. The total MET-min PA variable, and all the SB variables (i.e., specific types, total sitting minutes, minutes of school-related sitting, minutes of sitting for leisure) were also positively skewed and thus subjected to square root transformations. To test the first objective, chi-square and independent samples t-tests compared rates and means for participation in the specific types of PA and SB during the early-COVID-19 period (April – May 2020), respectively, by child sex (male vs. female) and child age group (5–8 years vs. 9–13 years). Multiple linear regression analyses further examined whether child sex and age predicted total MET-min, total sitting minutes, minutes of school-related sitting, minutes of sitting for leisure) after controlling for child ethnicity (Hispanic vs. non-Hispanic), parent employment status (works full-time vs. does not work full-time), parent marital status (married vs. not married), and annual household income. To test the second objective, ordinal logistic regression models examined whether child sex and age predicted the likelihood of parents perceiving changes (i.e., much more, somewhat more, about the same, somewhat less, much less) in PA and SB between the pre-COVID-19 period (February 2020) and early-COVID-19 period (April–May 2020) after adjusting for the same demographic covariates listed in the first objective. The third objective was tested by using generalized estimating equations (GEE) to examine within-subject changes in the likelihood of children engaging in PA at various locations differed by child sex and age group. The within-subject factor was Time (i.e., pre-COVID-19 vs. early-COVID-19), the between-subject factors were child sex and age group, and covariates included child ethnicity (Hispanic vs. non-Hispanic), parent employment status (works full-time vs. does not work full-time), parent marital status (married vs. not married), and annual household income. Interactions were tested for Time × Child sex and Time × Child age group. To address the ancillary goal of examining whether the likelihood of children engaging in some vs. none for each of three types of PA remote and streaming services (i.e., team sports, activity classes and lessons, classes offered by gym) also differed by child sex and age group, logistic regressions were conducted controlling for the covariates used in the above models.

A total of 325 parents expressed interest in the study and completed the screening questions. Of this number, n = 41 individuals were not eligible. Twenty-seven individuals were no longer interested in the study after completing the online screening form, and 257 individuals agreed to participate in the study. Thirty-one cases were removed that had not yet completed the baseline survey at the time of data analysis, and 14 participants were missing data on child PA and SB, leaving an analytic sample size of 211. Additional participants were missing data on demographic variables, but they were retained in the analytic sample and treated with pairwise deletion. Table 1 shows the descriptive statistics for the demographic characteristics of the analytic sample. Individuals resided across 35 U.S. states and the District of Columbia. Parents ranged in age from 20 to 61 years, with an average age of 42.05 years (SD = 5.34). Children’s ages ranged from 5 to 13 years, with an average age of 8.71 years (SD = 2.58). Most parents who participated were mothers, identified as non-Hispanic, and had graduated from college. About half of the children reported on in the study were female and the majority were non-Hispanic. Over half of parents reported working full-time. Over half of the sample reported an annual household income of $95,000 or more.

Table 2 shows descriptive statistics for the frequencies of children who performed various types of PA on the previous day during the early-COVID-19 period (April–May 2020). The most frequently reported physical activities were free play/unstructured PA (e.g., running around, tag, other active games) and going for a walk. Child sex and age differences were observed in the frequencies of reported physical activities (see Table 2). Boys were more likely to participate in sports practice/training than girls. Additionally, younger children (ages 5–8) were more likely to participate in free play/unstructured physical activity, biking, and scootering/skateboarding/roller skating than older children (ages 9–13). However, older children were more likely to participate in circuit training/conditioning than younger children. Descriptive statistics for the duration (in minutes) of various types of sedentary behaviors performed by children on the previous day during the early-COVID-19 period (April–May 2020) are shown in Table 3. Children spent the most time watching television/videos/ movies, sitting while hanging out with friends and family in person, doing school-related work, and playing computer or video games. Child sex and age differences were found in the duration of reported sedentary behaviors (see Table 3). Boys spent more time than girls playing computer or video games, whereas girls spent more time than boys using the internet/emailing/electronic media for leisure, doing video calls w/friends or family, sitting while listening to music, sitting talking on the phone/texting, sitting while hanging out with friends/family in person, and doing inactive hobbies. As compared to younger children, older children spent more time playing computer or video games, using the Internet/emailing/ electronic media for leisure, sitting while listening to music, and sitting talking on the phone/texting.

Results of the multiple regression analyses for demographic variables predicting square-root transformed total MET-min of PA, total minutes of sitting, minutes of school-related sitting, and minutes of sitting for leisure during the early-COVID-19 period are shown in Table 4. On average, children expended a total of 892.0 (SD = 902.07) MET-min on the previous day through the specific types of physical activities assessed. Children engaged in an average of 91.1 (SD = 109.2) min of sitting for school-related activities, 398.5 (SD = 184.6) min of sitting for leisure activity, and 489.4 (SD = 211.5) min of total sitting on the previous day. After controlling for all the variables in the model, child age was significantly negatively associated with total MET-min of PA. Also, child age was positively associated with minutes of school-related sitting and total minutes of sitting. Lastly, child age and being a girl were positively associated with minutes of sitting for leisure during the early-COVID-19 period after controlling for the child ethnicity (Hispanic vs. non-Hispanic), parent employment status (works full-time vs. does not work full-time), parent marital status (married vs. not married), and annual household income.

Overall, parents perceived children’s PA had decreased whereas children’s SB had increased between the pre-COVID-19 period (February 2020) and the early-COVID-19 period (April – May 2020). About 36% of parents reported their child had done much less PA in the past 7 days as compared to February 2020, whereas only about 11% of parents reported their child had done much more PA in the past 7 days as compared to February 2020. In contrast, 41% of parents reported their child had done much more sitting in the past 7 days as compared to February 2020, whereas only about 6% of parents reported their child had done much less sitting in the past 7 days as compared to February 2020. Results of the ordinal logistic regression analyses predicting perceived changes in PA and SB from the pre-COVID-19 period to the early-COVID-19 period by child sex and age group controlling for child ethnicity (Hispanic vs. non-Hispanic), annual household income, and parental marital and work status are shown in Table 5. The predictor variable, child age group, was found to contribute to the model for perceived changes in PA (ordered log-odds estimate = −.838, SE = .277, Wald = 9.120, p = .003) and the model for perceived changes in SB (ordered log-odds estimate = .623, SE = .280, Wald = 4.947, p = .026) controlling for all of the other variables in the model. Parents of older children (ages 9–13) vs. younger children (ages 5–8) were over twice as likely (OR = 2.31, 95% CI [1.34, 3.98]) to have a one-unit change in the perception their children had done less PA in past 7 days as compared to February 2020. The unadjusted proportions of change in each PA category by child age group are shown in Fig. 1. Parents of older children (ages 9–13) vs. younger children (ages 5–8) were half as likely (OR = 0.54, 95% CI [0.31, 0.93]) to have a one-unit change in the perception their children had done less SB in past 7 days as compared to February 2020. The unadjusted proportions of change in each sitting category by child age group are shown in Fig. 2. The odds of perceiving changes in children’s PA and SB were not associated with child sex.

Generalized estimating equations (GEE) tested changes in the locations of children’s PA from the pre-COVID-19 period (February 2020) to the early-COVID-19 period (April – May 2020). Results found the likelihood of performing PA at home or in the garage (OR = 2.49, 95% CI[1.35, 4.60], Wald = 8.593, p = .003) and on sidewalks and roads in their neighborhood (OR = 1.92, 95% CI [1.04,4.60], Wald = 4.28, p = .038) increased from the pre-COVID-19 period to the early-COVID-19 period after controlling for child sex, child age group, child ethnicity (Hispanic vs. non-Hispanic), parent employment status (works full-time vs. does not work full-time), parent marital status (married vs. not married), and annual household income. In contrast, the likelihood of performing PA at a park or trail (OR = 0.47, 95% CI [0.23, 0.97], Wald = 4.22, p = .040) decreased from the pre-COVID-period to the early-COVID-19 period. The likelihood of children performing PA in their yard or driveway (OR = 1.32, 95% CI [0.76, 2.31], Wald = 0.95, p = .329) or on sidewalks and roads outside the neighborhood (OR = 0.76, 95% CI [0.25,2.33], Wald = .288, p = .633) did not change. Changes in the locations of children’s PA did not differ by child sex or age group (i.e., interactions between Time × Child sex and Time × Child age group were not significant). GEE models predicting changes in the likelihood of children performing PA in the common areas of their apartments or condos, a gym or fitness center, an indoor sports facility, or an outdoor sports complex did not adequately converge due to low frequencies in certain cells. The unadjusted proportions of children whose parents reported they performed PA in each location during the pre-COVID-period and early-COVID-19 period are shown in Fig. 3.