Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Comparison of children’s physical activity profiles before and after COVID-19 lockdowns: A latent profile analysis

  • Ruth Salway ,

    Roles Conceptualization, Data curation, Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing

    ruth.salway@bristol.ac.uk

    Affiliations Centre for Exercise, Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, United Kingdom, Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom

  • Frank de Vocht,

    Roles Conceptualization, Funding acquisition, Writing – review & editing

    Affiliations Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom, The National Institute for Health Research, Applied Research Collaboration West (NIHR ARC West), University Hospitals Bristol and Weston NHS Foundation Trust, Bristol, United Kingdom

  • Lydia Emm-Collison,

    Roles Project administration, Writing – review & editing

    Affiliation Centre for Exercise, Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, United Kingdom

  • Kate Sansum,

    Roles Investigation, Writing – review & editing

    Affiliation Centre for Exercise, Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, United Kingdom

  • Danielle House,

    Roles Investigation, Project administration, Writing – review & editing

    Affiliation Centre for Exercise, Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, United Kingdom

  • Robert Walker,

    Roles Writing – review & editing

    Affiliation Centre for Exercise, Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, United Kingdom

  • Katie Breheny,

    Roles Writing – review & editing

    Affiliation Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom

  • Joanna G. Williams,

    Roles Conceptualization, Funding acquisition, Writing – review & editing

    Affiliations Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom, Communities and Public Health, Bristol City Council, Bristol, United Kingdom

  • William Hollingworth,

    Roles Conceptualization, Funding acquisition, Writing – review & editing

    Affiliations Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom, The National Institute for Health Research, Applied Research Collaboration West (NIHR ARC West), University Hospitals Bristol and Weston NHS Foundation Trust, Bristol, United Kingdom

  • Russell Jago

    Roles Conceptualization, Funding acquisition, Supervision, Writing – review & editing

    Affiliations Centre for Exercise, Nutrition & Health Sciences, School for Policy Studies, University of Bristol, Bristol, United Kingdom, Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom, The National Institute for Health Research, Applied Research Collaboration West (NIHR ARC West), University Hospitals Bristol and Weston NHS Foundation Trust, Bristol, United Kingdom, NIHR Bristol Biomedical Research Centre, University Hospitals Bristol and Weston NHS Foundation Trust and University of Bristol, Bristol, United Kingdom

Abstract

Physical activity is important for children’s health, but moderate to vigorous physical activity (MVPA) declines with age. COVID-19 lockdowns resulted in reduced MVPA and increased sedentary time among children. Characterising children’s activity patterns may help identify groups who are most likely to be inactive post-lockdown. Data were combined from a pre-COVID-19 cohort study on children aged 5–6 years (Year1: n = 1299), 8–9 years (Year4: n = 1223) and 10–11 years (Year6: n = 1296) and cross-sectional post-lockdown data from a natural experiment on 10-11-year-olds in 2021 (Year6-W1: n = 393) and 2022 (Year6-W2: n = 436). The proportions of time spent in MVPA, light physical activity (LPA) and sedentary time on weekdays and weekends were derived from accelerometer data. Latent class analysis was used to identify activity profiles pre and post-lockdown, and estimate pre-COVID-19 transitions between Year4 and Year6. We identified six pre-COVID-19 activity profiles in Year6, including a new profile characterised by very low MVPA and high sedentary time (19% of children). There was substantial movement between profiles at Year4 and Year6, with 45% moving to a profile with lower MVPA. Likelihood ratio tests suggested differences in Year6 activity profiles pre and post-lockdown, with a new post-lockdown profile emerging characterised by higher LPA. The percentage of children in the least active profiles (where under 20% meet UK physical activity guidelines), rose post-lockdown, from 34% pre-COVID-19 to 50% in 2021 and 40% in 2022. We also saw gender and socioeconomic gaps widen, and increased separation between high and low physical activity levels. Children’s physical activity has changed post-COVID-19, in terms of who is being active and how. The impact varies by activity profile, which is influenced by gender and socio-economic position. A greater understanding of these differences and targeting of low active groups is needed to increase both individual and population levels of physical activity.

Introduction

Increasing physical activity and decreasing sedentary time have both been found to be important for children’s health, with physical activity associated with emotional well-being, better cardiometabolic health and reduced risk of obesity [1, 2], while sedentary time is independently associated with increased adiposity [2, 3]. Moreover, low physical activity in childhood tracks into adulthood [4], leading to increased risk of heart disease, type 2 diabetes and some cancers [5, 6]. Despite this, children’s daily moderate to vigorous physical activity (MVPA) decreases throughout primary school by an average of 2 minutes per year. Over the same period daily sedentary time increases by an average of 13 min per year [7], and by age 11, the majority of children do not meet national UK guidelines of an average of 60 minutes of MVPA per day [7]. The COVID-19 pandemic has further impacted on children’s physical activity, with global reductions in physical activity and increases in sedentary time during periods of lockdown and restrictions [812]. In the UK, schools, leisure facilities and non-essential businesses were closed between 23rd March and 13th May 2020, with a second lockdown between 6th January and 3rd March 2021. After restrictions were lifted in 2021, MVPA initially remained lower and sedentary time higher compared to pre-pandemic [1215]. In 2022, children’s sedentary time remains higher than pre-pandemic by 13 mins [16]. Qualitative data have suggested the emergence of a new normal for child physical activity, with a stronger emphasis on structured activity, such as active clubs, rather than unstructured activity, such as free play, and highlight the possibility of some groups who are at higher risk of being inactive [17, 18]. As still only 41% of 10-11-year olds children meet UK physical activity guidelines [16], there is an urgent need to increase children’s activity levels to improve health. A key component of this is understanding whether patterns of children’s physical activity have changed after the lockdowns and if so, how.

The associations between MVPA, light intensity physical activity (LPA) and sedentary behaviour are complex [19], and studying average levels of these behaviours in isolation may not give a complete picture of activity patterns [20]. For example, some children who engage in very high levels of MVPA also have high sedentary time. Latent class analysis [21] is a data-driven technique that has been used to characterise groups of children who share similar patterns of physical activity and sedentary time [2225]. Identifying these activity profiles and exploring how they may change over time is important for designing behaviour change interventions as it facilitates the targeting of intervention approaches to those at greatest need.

Typical profiles include high physical activity/low sedentary time, high physical activity/high sedentary time and low physical activity/high sedentary time, with differences by gender and age [24, 25]. In a related study, we identified five activity profiles for children aged 5–6, and six profiles for the same children aged 8–9, ranging from very active to inactive [23], and with substantial movement between activity profiles between ages. We also saw strong gender differences in activity profile membership, with a higher proportion of boys in the more active profiles and girls in the more inactive profiles.

The aims of this paper are twofold. First, to extend our previous analyses to see if activity profiles change between ages 8–9 and age 10–11 pre-COVID-19. Second, to repeat the analysis in a post-lockdown sample to explore whether activity profiles at age 10–11 have changed post-pandemic, and if so in what way, either in terms of the profiles themselves or the proportions of children who fall into each. This will allow us characterise children’s activity patterns to identify groups who are most likely to be inactive post-lockdown, and thus at risk of poorer health and well-being. If physical activity patterns have changed, different strategies to increase children’s physical activity may be needed in future, and such groups may benefit from targeted strategies.

Methods

This is an exploratory study, which combines data from two studies. B-Proact1v [7, 26, 27] was a longitudinal cohort study of physical activity and sedentary behaviours of children aged 5–11 years and their parents. Data were collected in three phases between January 2012 and May 2018, when the children were aged 5–6 years (in Year 1 of primary school in the UK; Y1), 8–9 years (Y4) and 10–11 years (Y6). Fifty-seven schools in and around Bristol, UK, participated in Y1, with 47 and 50 of these participating in Y4 and Y6 respectively, with all children in the relevant year group eligible to take part. The Active-6 study [13, 16, 17, 28] was designed to investigate the impact of COVID-19 on children’s physical activity by comparing the Phase 3 B-Proact1v data on 10-11-year-olds to new data collected after the 2020/21 lockdowns. Data was collected in two waves: Wave 1 in June-December 2021 (Y6-W1) and Wave 2 in January-July 2022 (Y6-W2). Fig 1 shows the timings of the five samples. We invited the same 50 schools from BProact1v Y6 to participate, with 23 and 27 schools agreeing in Y6-W1 and Y6-W2 respectively (22 schools participated in both waves). Both studies collected accelerometer and questionnaire data from children in Year 6 (aged 10–11 years). Both studies received ethical approval from the School of Policy Studies Ethics Committee at the University of Bristol with parental informed consent obtained online or in writing for all participants. Only the Project Manager had access to participants’ identifiable information, which was stored separately from the main dataset.

Data

In total, data were collected from 1299 children at Y1, 1223 at Y4 and 1296 at Y6 from B-Proact1v, and 393 and 436 children in Y6-W1 and Y6-W2 from Active-6, respectively. Children wore a waist-mounted ActiGraph wGT3X-BT accelerometer (Actigraph LLC; Florida, US) during waking hours for five consecutive days including two weekend days in B-Proact1v, and for seven consecutive days in Active-6. An open-source R script [29] was used to process the accelerometer data. We excluded data between midnight and 6am and defined a valid day as at least 500 minutes of data, excluding intervals of ≥60 minutes of zero counts and up to two minutes of interruptions [30]. All participants who provided at least two valid weekdays and one valid weekend day of data were included. Evenson population-specific cut-points for children [31] were used to derive mean weekday and weekend minutes of MVPA, LPA and sedentary time from 10s epochs. These were converted to proportions of the total wear time spent in MVPA, LPA and sedentary time. We also calculated whether children met the UK physical activity guidelines of a daily average of 60 min per day.

Child gender and the highest education qualification in the household were reported in a parent/carer questionnaire. Children completed a questionnaire about the frequency with which they engaged in different forms of activity outside school hours: sport or exercise club at school, sport or exercise club elsewhere, playing outdoors in their neighbourhood, and playing outdoors at home (0 = ‘Never’, 1 = ‘1–2 days a week’, 2 = ‘3–4 days a week’, 3 = ‘5–7 days a week’). These were coded with the midpoint number of days for each category and averaged to form two variables for the number of days participating in structured (clubs) and unstructured (play) activities, on a scale of 0–6 days.

Statistical analysis

Descriptive summaries and missing data were reported for all variables. This analysis falls into two parts. Firstly, previous pre-COVID-19 longitudinal results on children’s activity profiles for Y1 and Y4 were extended to Y6. Secondly, the Y6 analysis was repeated cross-sectionally for children post-lockdown (Y6-W1 and Y6-W2) to investigate the potential impact of the COVID-19 lockdowns on activity profiles. For consistency with previous results, we used the same model structure as before [23]. Briefly, the latent profiles were identified based on separate weekday and weekend proportions of time spent in MVPA and sedentary time (with LPA implicit, as proportions sum to one). We assumed a non-zero covariance between MVPA and sedentary proportions, with weekday and weekend variances the same within a profile, but differing for MVPA and sedentary time, and between profiles. All latent variable analyses were performed in MPlus v8.8 [32] and rerun with multiple start values to ensure that the maximum log-likelihood value was replicated. We used full information maximum likelihood, which has been shown to produce unbiased parameter estimates and standard errors under a missing at random assumption [33].

Pre-COVID-19 extension to Y6.

We fitted cross-sectional latent profile models for 2–10 activity profiles and determined the appropriate number of profiles using a mixture of statistical criteria, interpretability and context (profiles that capture meaningful differences or specific populations of interest) and parsimony (fewer profiles and avoidance of very small profiles preferred), as recommended [34, 35]. Statistical criteria used to measure model fit were the Bayesian Information Criterion (BIC) and the sample-size adjusted BIC (saBIC) where lower values indicate better fit [36]. To compare models with different numbers of profiles, we reported the Lo-Mendell-Rubin (LMR) [37] and bootstrapped likelihood ratio (BLRT) tests [38]. These tests compare a k-1 versus a k-class model, with a low p-value rejecting the k-1 class model in favour of the k class model. We also reported the entropy, which measures the separation between classes, with higher values denoting better separation, although we do not use this to determine the optimum number of classes [39]. Descriptive labels were applied to each profile. For each profile, we also estimated the expected proportion who met the physical activity guidelines, by calculating the probability of achieving at least 60 mins of MVPA, based on the estimated distribution of proportion of time spent in MVPA and the overall average wear time. To examine change in profile membership between Y4 and Y6, we first tested for measurement invariance (that is, whether the same latent profiles were present at both timepoints) and, if there was evidence to support invariance, we estimated a new joint model. We then used a latent transition model [40, 41] for those children with data at both time points (n = 805), fixing the profiles as specified in either the joint or individual cross-sectional models (depending on measurement invariance), and estimated transition probabilities between Y4 and Y6 profiles.

Pre-COVID-19 and post-lockdown cross-sectional comparison.

As the focus of these analyses was on the potential impact of the COVID-19 pandemic rather than purely descriptive for each wave, we adopted a slightly different approach in the comparison of pre- and post-lockdown data, by fixing the Y6 pre-COVID-19 activity profiles, and prioritising consistency between different time points where justified. This also avoided issues with smaller profiles being lost due to the smaller sample sizes in the two post-lockdown waves. We used multiple group latent class analysis [21], treating each post-lockdown wave (Y6, Y6-W1 and Y6-W2) as a separate group, as previous analysis has shown differences in overall average physical activity [13].

We first fit a model with Y6-W1 and Y6-W2 latent profiles fixed to be the same as the Y6 latent profiles above and compared it to a model that allowed the post-lockdown Y6-W1 and Y6-W2 profiles to differ. A likelihood ratio test was used to test whether the Y6, Y6-W1 and Y6-W2 latent profile definitions were the same (test for measurement invariance). If activity profile definitions differed, we further tested (a) whether Y6 and Y6-W2 were the same (reflecting a hypothesis of an initial change in profiles at Wave 1 followed by a return to previous profiles) and (b) whether Y6-W1 and Y6-W2 were the same (reflecting the hypothesis that pre and post lockdown activity profiles differ). The final model was used to produce estimates of the proportions of children in each activity profile at each time point. We also examined whether there were differences in gender, household education and structured/unstructured activities between profiles with a Wald test via the BCH method, which includes the classification error and is robust to assumption violations [42]. Accelerometer wear time may potentially affect estimates of MVPA/LPA/sedentary proportions and hence could influence profile membership probabilities. To check whether this was an issue we also tested for differences in wear time between profiles.

Results

Missing data (S1 File: Table 1) were similar across Y6, Y6-W1 and Y6-W2, with child characteristics missing 0–8% of data and accelerometer data missing 6–30%. Most missing data was due to a child having insufficient weekend accelerometer data. The gender balance was similar across all three samples, with a higher percentage of those from households with higher educational qualifications in the post-lockdown Y6-W1 and Y6-W2 samples. The overall proportion of time spent in MVPA, LPA and sedentary time did not differ markedly (Table 1), but the average number of days children reported engaging in structured and unstructured activity was lower post-lockdown by just under half a day for each. The percentage of children meeting UK physical activity guidelines dropped from 40% pre-COVID-19 to 37% in Wave 1 but returned to 41% in Wave 2.

thumbnail
Table 1. Proportions of time spent in MVPA, LPA and sedentary time for age 10–11 samples.

https://doi.org/10.1371/journal.pone.0289344.t001

Pre-COVID-19 cross-sectional profiles at Y6

Table 2 gives model fit indicators for the Y6 data for 2–10 profiles. The LMR supported a three-profile model, with BIC and adjusted BIC indicating five or activity profiles, respectively. The five- and six-profile models produced profiles that were similar to those seen in the previously reported Y1 and Y4 analyses. The main difference between five and six profile models was the addition of an extra sedentary and inactive activity profile, comprising 10% of the data. We chose the six-profile model on the basis of interpretability and context, as this additional profile captured a specific population of policy interest, was relatively large and was consistent with the post-COVID-19 lockdown analysis below. The weekday and weekend MVPA/LPA/sedentary proportions for each profile are shown in Fig 2 (left) with the estimated proportion of children in each profile shown in Fig 3 (left) and Table 3; full details of estimates are given in S1 File: Table 2. The six pre-COVID-19 activity profiles, ordered from most active to least (based on estimated proportion meeting physical activity guidelines) were as follows:

  • Highly Active (10%): High and very high proportions of MVPA, especially on weekends, and low proportions of sedentary time (95% meet guidelines)
  • Active (13%): Higher than average proportions of MVPA, and typical proportions of sedentary time (83% meet guidelines)
  • Moderate (20%): Average proportions of MVPA and weekday sedentary time; more LPA at weekends (62% meet guidelines)
  • Sedentary (23%): Average proportions of MVPA but high levels of sedentary time, with more sedentary time and less MVPA at weekends (28% meet guidelines)
  • Inactive (15%): Low proportions of MVPA and average levels of sedentary time. (10% meet guidelines)
  • Sedentary & Inactive (19%): Low to very low proportions of MVPA and high to very high levels of sedentary time (2% meet guidelines)
thumbnail
Fig 2. Pre and post-lockdown latent profile definitions for 10-11-year-old children.

Red = Sedentary time, yellow-light physical activity, green = moderate to vigorous physical activity.

https://doi.org/10.1371/journal.pone.0289344.g002

thumbnail
Fig 3. Distribution of Y6, Y6-W1 and Y6-W2 activity profiles pre- and post-lockdown for 10-11-year-old children.

Y6: March 2017-May 2018; Y6-W1: June—December 2021 Y6-W2: January–July 2022.

https://doi.org/10.1371/journal.pone.0289344.g003

thumbnail
Table 2. Model fit for models with 2–10 classes for Y6.

https://doi.org/10.1371/journal.pone.0289344.t002

thumbnail
Table 3. Percentage of time spent in sedentary (SED), light (LPA) and moderate to vigorous (MVPA) physical activity for pre-COVID-19 and post-lockdown profiles.

https://doi.org/10.1371/journal.pone.0289344.t003

Weekday and weekend patterns were similar within a profile, with weekends typically more extreme versions of weekday patterns. Active profiles (Highly Active and Active, in which over 80% of children were estimated to meet physical activity guidelines) comprised 23% of the population, while the least active profiles (Inactive and Sedentary & Inactive, in which under 20% of children were estimated to meet physical activity guidelines) comprised 34%.

Pre-COVID-19 transition between Y1, Y4 and Y6

Fig 4 shows the previously reported transitions between profiles at Y1 and Y4 [23] extended to the newly created profiles at Y6, with transition probabilities provided in S1 File: Table 3. Note that activity profile labels from the previous analysis [23] have been relabelled and reordered slightly (see S1 File: Table 4 for further details) so that profiles reported in this paper are ordered from most active (top) to least active (bottom), based on estimated proportion meeting physical activity guidelines.

thumbnail
Fig 4. Transitions between activity profiles at Y1, Y4 and Y6, and cross-sectional comparison with Y6-W1 and Y6-W2.

Y1: January 2012—July 2013 (age 5–6); Y4: March 2015—July 2016 (age 8–9); Y6: March 2017- May 2018 (age 10–11) Y6-W1: June—December 2021 (age 10–11); Y6-W2: January—July 2022 (age 10–11).

https://doi.org/10.1371/journal.pone.0289344.g004

There was substantial movement between profiles between Y4 and Y6 (Fig 4; S1 File: Table 3). The Active/Light profile, characterised by higher-than-average MVPA and high levels of LPA, reduced in size from 29% to 6% between Y1 and Y4, and disappeared completely from the Y6 profiles. The Sedentary & Inactive profile, characterised by very low MVPA and very high sedentary time, emerged as a new activity profile at Y6, with those in the Sedentary and Inactive profiles at Y4 most likely to move to this profile. Overall, the most common patterns of movement were to a profile with a lower MVPA (45%) or no change (36%), with only 19% moving to a more active profile.

Pre-COVID-19 and post-lockdown cross-sectional comparison

We estimated six latent profiles for Y6, Y6-W1 and Y6-W2 with the pre-COVID-19 Y6 profiles fixed to be those estimated above. Likelihood ratio tests suggested that activity profiles differed between waves, with additional tests concluding that the differences were between pre (Y6) and post (Y6-W1 & Y6-W2) lockdown, with profiles in the post-lockdown waves the same, although with different proportions (Table 4). The post-lockdown activity profiles (Table 3; Figs 2 and 3; S1 File: Table 5) are described below, with the Y6-W1 and Y6-W2 percentages given in brackets, and highlighting similarities and differences to the pre-lockdown profiles:

  • Highly Active (Y6-W1:5% Y6-W2: 12%): High and very high proportions of MVPA, especially on weekends, and low proportions of sedentary time. Similar MVPA to Highly Active pre-COVID-19 profile, but more sedentary. (96% meet guidelines)
  • Active/Light (Y6-W1: 7% Y6-W2: 7%): higher than average proportions of MVPA, and low proportions of sedentary time, both more extreme on weekends. This profile differs from the pre-COVID-19 Active profile, with similar MVPA, but more LPA/less sedentary time. (82% meet guidelines)
  • Moderate (Y6-W1: 20% Y6-W2: 20%): Average proportions of MVPA and weekday sedentary time; more LPA at weekends. Similar MVPA to Moderate pre-COVID-19 profile, but more sedentary, and fewer meet the guidelines. (58% meet guidelines)
  • Sedentary (Y6-W1: 16% Y6-W2: 20%): Average proportions of MVPA but high levels of sedentary time, and more sedentary at weekends. Slightly more weekend MVPA and weekday sedentary compared to Sedentary pre-COVID-19 profile, and more likely to meet the guidelines. (42% meet guidelines)
  • Inactive (Y6-W1: 14% Y6-W2: 13%): Low proportions of MVPA and average levels of sedentary time. Very similar to Inactive pre-COVID-19 profile. (8% meet guidelines)
  • Sedentary & Inactive (Y6-W1: 38% Y6-W2: 27%): Low to very low proportions of MVPA and high to very high levels of sedentary time. Similar MVPA to Sedentary & Inactive pre-COVID-19 profile, but more sedentary. (3% meet guidelines)
thumbnail
Table 4. Likelihood ratio tests for pre- and post-lockdown profile comparison.

https://doi.org/10.1371/journal.pone.0289344.t004

The largest difference between pre- and post-lockdown activity profiles was the replacement of the pre-COVID-19 Active profile (characterised by higher MVPA and average sedentary time) with an Active/Light profile, which had similar MVPA but less sedentary time and more LPA. This profile was similar to the Active/Light profile seen in Y4 (although more sedentary, in common with all Y6 activity profiles). In general, post-lockdown activity profiles were more sedentary than pre-COVID-19, particularly the Highly Active and Moderate profiles, although the proportions of MVPA remained similar for most. There were also differences in the percentage meeting guidelines, with those in the Moderate profile slightly less likely to meet guidelines post-lockdown (59% compared to 62%) but those in the post-lockdown Sedentary profile more likely (42% compared to 28%).

Although the activity profiles were the same in Y6-W1 and Y6-W2 post-lockdown, there were substantial differences in the proportions within each profile. In Y6-W1 1, only 5% were in the Highly Active profile, whereas over a third (38%) were in the Sedentary & Inactive profile. In Wave 2, proportions were closer to pre-pandemic, with the exception of the Sedentary & Inactive which was higher (27% compared to 19% pre-pandemic). In Y6-W1, over half of children (52%) were in the least active profiles (less than 20% meet guidelines), decreasing to 40% in Y6-W2, compared to 34% pre-pandemic (S1 File: Table 5).

Differences by gender, household education and structured/unstructured activities.

The Inactive and Inactive & Sedentary profiles contained the highest proportion of girls at 85% and 69% respectively pre-COVID-19 (Y6), while post-lockdown in Y6-W2 the highest proportions were for the new Active/Light profile (86%), with proportions in the Inactive and Inactive & Sedentary profiles remaining high at 73% and 74% (Table 5). In contrast, over two-thirds (70%) of the pre-COVID-19 Y6 Highly Active profile were boys, rising to 82% post-lockdown in Y6-W2. The new post-lockdown Active/Light profile was predominately boys in Y6-W1, similar to the pre-COVID-19 Y6 Active profile, but switching to a majority of girls by Wave 2. Pre-COVID-19 Y6 differences by household education were in the less active activity profiles, with a higher proportion of those from households with lower education in the Inactive profile, compared to more from higher-education households in the Sedentary profile included. These differences disappeared in Y6-W1, but returned in Y6-W2, along with differences in the Highly Active and Moderate profiles, with the Y6-W2 post-lockdown profiles more likely to be from higher-education households. The new Active/Light profile included a higher proportion of children from lower-education households, compared to other post-lockdown activity profiles, and compared to the pre-COVID-19 Y6 Active profile. The average number of days engaging in structured activities was highest for the Highly Active and Active profiles, but lower for Active/Light in Y6-W2 with all post-lockdown activity profile averages lower than those in their pre-COVID-19 counterparts (Table 6). Pre-COVID-19 Y6 differences in unstructured activities (highest in Active and Moderate profiles) were not evident post-lockdown. There were no differences in accelerometer wear time between profiles (S1 File: Table 6), indicating that the distribution of activity profiles was not dependent on wear time.

thumbnail
Table 5. Model-based estimates of gender and household education proportions in each activity profile, and test for differences across profiles.

https://doi.org/10.1371/journal.pone.0289344.t005

thumbnail
Table 6. Model-based estimates of mean number of days in structured and unstructured activities in each activity profile, and test for differences across profiles.

https://doi.org/10.1371/journal.pone.0289344.t006

Discussion

This paper has explored pre-COVID-19 longitudinal change in children’s activity profiles between ages 5–6 and 10–11, and compared cross-sectional profiles at ages 10–11 at two time points before and after COVID-19 lockdowns. Pre-COVID-19, we found that activity profiles changed between ages 8–9 and 10–11, with a move away from profiles with high proportions of LPA (Active/Light) and the emergence of a new activity pattern characterised by a combination of very low proportions of MVPA and high proportions of sedentary time (Sedentary & Inactive), in which only 2% were estimated to meet UK physical activity guidelines. Overall, 45% of children moved to a less active profile at age 11, and a considerable proportion of children became less active and more sedentary, in particular those who were less active to begin with. COVID-19 lockdowns have disrupted children’s activity [13], and we saw substantial differences between pre-COVID-19 in 2018, Y6-W1 in 2021 and Y6-W2 in 2022. These differences could be due to other factors that have changed over this time, rather than a direct consequence of the pandemic, for example, a secular trend, changes to physical activity provision or other events such as increased cost of living, although our results suggest a period of disruption in 2021–22, especially among the most and least active profiles. Post-lockdown, we found: 1) a change in the activity profiles themselves, with an activity profile characterised by high proportions of MVPA (Active) replaced by one characterised by a mix of MVPA and light activity (Active/Light), and all activity profiles more sedentary than pre-COVID-19; 2) a change in the proportions of children in each activity profile with a higher proportion of children in the least active profiles, especially in 2021; and 3) changes in the gender and socioeconomic composition of profiles. These all point to important changes in children’s physical activity by 2022, in terms of who is being active and how, that need to be understood before developing new strategies to increase children’s physical activity.

The profiles identified here are similar to those observed in other studies, ranging from very active to inactive [2325, 43]. In the pre-COVID-19 longitudinal analysis, the increase in sedentary time between Y4 and Y6 and the new Sedentary and Inactive profile, consisting of about a fifth of 10-11-year-olds, reflect a general pattern of lower physical activity and higher sedentary time observed with age [7, 30]. Unstructured activities, such as active play, have also been found to decline with age [23, 44] and may explain the shift away from the Active/Light profile (characterised by unstructured activity) towards structured activity (more common in Highly Active and Active profiles). A third of children stayed in the same profile, while only a fifth moved to a more active profile, similar to the transition patterns seen between Y1 and Y4 [23]. It may therefore be more productive to encourage children to retain physical activity levels at an earlier age, rather than to become active again once they have stopped, especially as children’s physical activity tracks into adulthood [4].

Post-lockdown, the pre-COVID-19 Active profile, a more moderate version of the Highly Active profile with slightly less MVPA and more sedentary time, was replaced by an Active/Light profile, characterised by high levels of LPA, similar to that seen pre-COVID-19 at Y1 and Y4 but which previously disappeared in Y6. Differences in the activity levels and socio-demographic composition suggest that these two profiles capture different activity patterns, reflecting underlying differences in the types of physical activity and who is engaging in them. This difference in physical activity patterns is supported by our qualitative evidence [18] which has found a ‘new normal’ for children’s physical activity in 2022, a year following the COVID-19 lockdowns, with a greater dependence on organised activities. On average, those in the Active/Light profile attended fewer active clubs than other activity profiles in which the majority met physical activity guidelines, and so this small group of atypical children may struggle to maintain activity levels in the ‘new normal’. Given the large differences between Y6-W1 and Y6-W2, it is not clear if this profile is a permanent change, or if activity profiles are still changing, and that the Active/Light profile may disappear in the future as it has in the past. Pre-COVID-19, children in the Active/Light profile in Y4 were most likely to move to the less active Moderate (30%) or Inactive profiles (39%). Thus, post-lockdown it is important to make sure that children with this activity profile are encouraged to maintain their physical activity. Future research should explore the reasons for their lower participation in structured activities, to ensure accessibility for all and a range of both structured and unstructured active options.

Profiles were generally more sedentary post-lockdown compared to pre-COVID-19, especially in the Highly Active profile, where the average proportion of MVPA was similar to pre-COVID-19 but the average proportion of time spent sedentary increased from 56% to 62% on weekdays and 48% to 56% on weekends. While evidence on children’s physical activity after lockdowns is still limited, this is consistent with our previous findings that overall average sedentary time remained lower by 13 minutes on average a year after UK lockdowns were lifted [16], with some lockdown habits, such as increased screen-viewing, continuing [18, 45]. Findings are also conceptually consistent with a Texas-based longitudinal study [15] that found larger increases in sedentary time of 50 min, with only 15% of 8-11-year-olds decreasing their sedentary time between 2019/20 and 2021/22 (although some of this increase may be attributable to age-related rather than COVID-19 changes). Moreover, we found that the proportion of children in the least active profiles (Inactive and Sedentary & Inactive; those in which fewer than 20% of children were estimated to meet physical activity guidelines) rose from 34% pre-COVID-19 to 52% when lockdowns were first lifted and remained higher than pre-pandemic at 40% a year later. Most notably, the Sedentary & Inactive profile doubled in size between pre-COVID-19 and 2021 to 38%. Although smaller by 2022, it remains the largest activity profile, with over a quarter of children (27%), which is concerning given that children in this profile are much less likely to meet physical activity guidelines than those in other profiles. The negative health impacts of sedentary behaviours may be independent of inactivity [2, 46], and so these children are potentially at risk of a double negative health impact. Thus, we urgently need strategies to target this increasingly inactive and sedentary population.

In this study we saw differences in activity patterns by gender and household education, with disparities widening between 2018 and 2022. Pre-COVID-19, the Highly Active profile was male-dominated, with only a quarter being girls, but post-lockdown girls were much less likely to have this activity profile, with only 8% female in 2022. So, while overall the proportion of children in the Highly Active profile was similar, this was driven mainly by an increase in boys in this group. The opposite pattern occurred in Sedentary & Inactive profile, with a higher proportion of girls post-lockdown. The Active Lives Children and Young People Survey 2021–22 reported that the gender gap in physical activity closed during the pandemic, mainly as boys became less active [47], and re-emerged post-lockdown [48] similar to before COVID-19. Our results show the gap is wider for the most and least active profiles than pre-pandemic, and suggest that averaging over all children, as in our previous analysis, may mask the extent of the gender gap in specific subgroups. Similarly, we saw increased differentiation in activity profiles post-lockdown by socioeconomic position, with a lower proportion of children from households with lower education attainment in the Highly Active profile, and a higher proportion in the Inactive profile. Previous research has suggested that reasons why girls may be less active than boys include lack of role models, desire for more choice and issues around body image [49], and it is possible that these have become more pronounced after lockdowns. Similarly, reasons for socioeconomic gaps include the physical environment, parental education, financial resources and accessibility [5052]. While the first two factors are unlikely to have changed over the time scale of this study, both cost and accessibility, for example impacted by the COVID-19 pandemic and/or rising cost of living, could potentially explain the widening gaps, especially with the observed stronger reliance on structured activities. Further research is needed to focus specifically on girls and those from lower socio-economic backgrounds to understand why these groups in particular are now less likely to be in the highly active profiles, especially in relation to accessibility and cost of provision, in order to develop interventions that target these groups to prevent these gaps widening still further.

Post-lockdown, the changes in activity profiles suggest that physical activity patterns have become more polarised. While overall averages suggest a recovery in children’s physical activity to pre-pandemic levels [16, 48], this may be masking an important change in the loss of an active ‘middle ground’ with children separating into more or less active groups. This may be further exacerbated if the Active/Light profile disappears in future, especially since it consists predominantly of girls (86%) and children from lower socio-economic background (65%), two groups who are typically less likely to meet physical activity guidelines [7]. If, as suggested by our previous qualitative work [18], physical activity is increasingly characterised by structured activities such as active clubs, the danger is that these become the preserve of those who are already active, dominated by boys and those from higher socioeconomic backgrounds, with others becoming more sedentary and inactive. Reduced levels of physical activity among those less likely to be active could thus lead to poorer health and well-being and the widening socio-economic inequalities in health. It is therefore imperative that we ensure active options are available for all abilities and interests rather than just catering to those who are already active.

The key implications of this study are therefore:

  • changes in who is being active and how, in particular the increase in the most sedentary profile, may require different strategies in future to increase physical activity;
  • those children who currently get their MVPA from less structured sources may need additional support as post-lockdown physical activity becomes more focused around structured activities;
  • the widening gender and socio-economic gaps need to be better understood and addressed;
  • opportunities to be active need to be accessible to all not just those who are already active.

Strengths and limitations

Latent class/profile analysis allows us to explore and characterise the complex relationship between physical activity and sedentary time beyond population averages. A key strength of our analysis is that we collected data using the same protocols at all timepoints. This is particularly important as latent profile analysis is a data-driven approach, and so we were able to use the same model approach and structure to estimate the full pre-COVID-19 longitudinal transition between ages 5/6 and ages 10/11, and comparable pre and post-lockdown models. Unlike cluster analysis, a similar approach for identifying groups, latent profile analysis is probabilistic. As a result, our conclusions are more statistically robust as we were able to formally test hypotheses about differences in profiles between waves rather than relying on subjective comparisons. In addition, pre- and post-lockdown samples were recruited from the same schools, reducing the impact of school differences. However, we cannot say that the differences observed between the two time points are directly due to the COVID-19 pandemic or associated lockdowns, as any age-specific activity profiles may change over time naturally, or differences may be due to other factors, for example, changes physical activity provision in both the school and community, changes in school priorities or the increase in cost of living. We were also unable to look at transitions between post-lockdown activity profiles, due to insufficient longitudinal data, or clustering within schools. Children and schools taking part in both B-Proact1v and Active 6 were relatively active, and the Active-6 sample were more likely to come from degree-educated households, and so our analysis may underestimate the proportion of children in the least active profiles, and our findings may not be generalisable to other schools or areas.

Conclusions

Children’s physical activity patterns have changed after the COVID-19 lockdowns, in terms of who is being active and how. The key changes are activity patterns and proportions which reflect underlying differences in the types of physical activity children engage in, and an increase in the percentage of children in the least active profiles, where under 20% meet UK physical activity guidelines. We also saw a widening of gender and socioeconomic gaps in physical activity, and increased separation between high and low physical activity levels. The key implication of this work is that age and lockdown-related impacts on children’s physical activity are not the same for all but vary according to their physical activity profile, which in turn is influenced by gender and socio-economic position. Thus, a greater understanding of these differences and how to help the different groups of children to be more active is needed to increase both individual and population levels of physical activity.

Supporting information

S1 Checklist. STROBE statement—checklist of items that should be included in reports of observational studies.

https://doi.org/10.1371/journal.pone.0289344.s001

(DOCX)

Acknowledgments

We would like to thank all the families and schools that have taken part in the B-Proact1v and Active-6 projects.

References

  1. 1. Skrede T, Steene‐Johannessen J, Anderssen SA, Resaland GK, Ekelund U: The prospective association between objectively measured sedentary time, moderate‐to‐vigorous physical activity and cardiometabolic risk factors in youth: a systematic review and meta‐analysis. Obes Rev 2018. pmid:30270500
  2. 2. Chaput JP, Willumsen J, Bull F, Chou R, Ekelund U, Firth J, et al: 2020 WHO guidelines on physical activity and sedentary behaviour for children and adolescents aged 5–17 years: summary of the evidence. Int J Behav Nutr Phys Act 2020, 17(1):141.
  3. 3. Carson V, Hunter S, Kuzik N, Gray CE, Poitras VJ, Chaput J-P, et al: Systematic review of sedentary behaviour and health indicators in school-aged children and youth: an update. Appl Physiol Nutr Metab 2016, 41(6 Suppl 3):S240–265. pmid:27306432
  4. 4. Telama R, Yang X, Leskinen E, Kankaanpaa A, Hirvensalo M, Tammelin T, et al: Tracking of physical activity from early childhood through youth into adulthood. Medicine and science in sports and exercise 2014, 46(5):955–962. pmid:24121247
  5. 5. World Health Organization: Global Health Risks: Mortality and burden of disease attributable to selected major risks. In. Geneva: World Health Organization; 2009.
  6. 6. Bull FC, Al-Ansari SS, Biddle S, Borodulin K, Buman MP, Cardon G, et al: World Health Organization 2020 guidelines on physical activity and sedentary behaviour. Br J Sports Med 2020, 54(24):1451–1462. pmid:33239350
  7. 7. Jago R, Salway R, Emm-Collison L, Sebire SJ, Thompson JL, Lawlor DA: Association of BMI category with change in children’s physical activity between ages 6 and 11 years: a longitudinal study. Int J Obes 2020, 44:104–113. pmid:31712707
  8. 8. Stockwell S, Trott M, Tully M, Shin J, Barnett Y, Butler L, et al: Changes in physical activity and sedentary behaviours from before to during the COVID-19 pandemic lockdown: a systematic review. BMJ Open Sport Exerc Med 2021, 7(1):e000960. pmid:34192010
  9. 9. Rossi L, Behme N, Breuer C: Physical Activity of Children and Adolescents during the COVID-19 Pandemic-A Scoping Review. Int J Environ Res Public Health 2021, 18(21):11440. pmid:34769956
  10. 10. Neville RD, Lakes KD, Hopkins WG, Tarantino G, Draper CE, Beck R, et al: Global Changes in Child and Adolescent Physical Activity During the COVID-19 Pandemic: A Systematic Review and Meta-analysis. JAMA Pediatr 2022, 176(9):886–894. pmid:35816330
  11. 11. Burkart S, Parker H, Weaver RG, Beets MW, Jones A, Adams EL, et al: Impact of the COVID-19 pandemic on elementary schoolers’ physical activity, sleep, screen time and diet: A quasi-experimental interrupted time series study. Pediatr Obes 2022, 17(1):e12846. pmid:34409754
  12. 12. Scales J, Chavda J, Ikeda E, Tsocheva I, Dove RE, Wood HE, et al: Device-Measured Change in Physical Activity in Primary School Children During the UK COVID-19 Pandemic Lockdown: A Longitudinal Study. Journal of Physical Activity and Health 2023:1–9. pmid:37142406
  13. 13. Salway R, Foster C, de Vocht F, Tibbitts B, Emm-Collison L, House D, et al: Accelerometer-measured physical activity and sedentary time among children and their parents in the UK before and after COVID-19 lockdowns: a natural experiment. Int J Behav Nutr Phys Act 2022, 19(1):51. pmid:35570265
  14. 14. Hurter L, McNarry M, Stratton G, Mackintosh K: Back to school after lockdown: The effect of COVID-19 restrictions on children’s device-based physical activity metrics. J Sport Health Sci 2022, 11(4):530–536. pmid:35092856
  15. 15. Ganzar LA, Salvo D, Burford K, Zhang Y, Kohl HW, Hoelscher DM: Longitudinal changes in objectively-measured physical activity and sedentary time among school-age children in Central Texas, US during the COVID-19 pandemic. Int J Behav Nutr Phys Act 2022, 19(1):56. pmid:35590329
  16. 16. Jago R, Salway R, House D, Walker R, Emm-Collison L, Sansum K, et al: Short and medium-term effects of the COVID-19 lockdowns on child and parent accelerometer-measured physical activity and sedentary time: a natural experiment. Int J Behav Nutr Phys Act 2023, 20(1):42. pmid:37101270
  17. 17. Walker R, House D, Emm-Collison L, Salway R, Tibbitts B, Sansum K, et al: A multi-perspective qualitative exploration of the reasons for changes in the physical activity among 10-11-year-old children following the easing of the COVID-19 lockdown in the UK in 2021. Int J Behav Nutr Phys Act 2022, 19(1):114. pmid:36064548
  18. 18. Walker R, House D, Salway R, Emm-Collison L, Hollander LE, Sansum K, et al: The new normal for children’s physical activity and screen viewing: A multi-perspective qualitative analysis of behaviours a year after the COVID-19 lockdowns in the UK. BMC Public Health 2023:Forthcoming. pmid:37495976
  19. 19. Biddle S, Gorely T, Marshall SJ, Murdey I, Cameron N: Physical activity and sedentary behaviours in youth: issues and controversies. J R Soc Promot Health 2004, 124(1):2–48. pmid:14971190
  20. 20. Thompson D, Peacock O, Western M, Batterham AM: Multidimensional physical activity: an opportunity, not a problem. Exerc Sport Sci Rev 2015, 43(2):67–74. pmid:25607280
  21. 21. Collins LM, Lanza ST: Latent class and latent transition analysis: with applications in the social, behavioral, and health sciences. Hoboken, N.J.: Wiley; 2010.
  22. 22. Patnode CD, Lytle LA, Erickson DJ, Sirard JR, Barr-Anderson DJ, Story M: Physical Activity and Sedentary Activity Patterns Among Children and Adolescents: A Latent Class Analysis Approach. Journal of Physical Activity and Health 2011, 8:457–467. pmid:21597117
  23. 23. Jago R, Salway R, Lawlor DA, Emm-Collison L, Heron J, Thompson JL, et al: Profiles of children’s physical activity and sedentary behaviour between age 6 and 9: a latent profile and transition analysis. Int J Behav Nutr Phys Act 2018, 15:103. pmid:30352597
  24. 24. Parker KE, Salmon J, Costigan SA, Villanueva K, Brown HL, Timperio A: Activity-related behavior typologies in youth: a systematic review. Int J Behav Nutr Phys Act 2019, 16(1):44. pmid:31097036
  25. 25. Leech RM, McNaughton SA, Timperio A: The clustering of diet, physical activity and sedentary behavior in children and adolescents: a review. Int J Behav Nutr Phys Act 2014, 11:4. pmid:24450617
  26. 26. Jago R, Sebire SJ, Wood L, Pool L, Zahra J, Thompson JL, et al: Associations between objectively assessed child and parental physical activity: a cross-sectional study of families with 5–6 year old children. BMC Public Health 2014, 14:655. pmid:24970045
  27. 27. Jago R, Solomon-Moore E, Macdonald-Wallis C, Sebire SJ, Thompson JL, Lawlor DA: Change in children’s physical activity and sedentary time between Year 1 and Year 4 of primary school in the B-PROACT1V cohort. Int J Behav Nutr Phys Act 2017, 14:33. pmid:28449679
  28. 28. Jago R, Foster C, Williams J, de Vocht F, Hollingworth W: Protocol for Assessing the impact of COVID-19 on the physical activity of Year 6 children and their parents: Identifying scalable actions to mitigate adverse impacts & provide rapid evidence to policy makers (ACTIVE-6) 2021, [cited 2023 Mar 27] Available from: https://fundingawards.nihr.ac.uk/award/NIHR131847
  29. 29. Salway R: Accelerometer Processing Code, 2022, , Available from https://osf.io/y8mwu/
  30. 30. Cooper AR, Goodman A, Page AS, Sherar LB, Esliger DW, van Sluijs EM, et al: Objectively measured physical activity and sedentary time in youth: the International children’s accelerometry database (ICAD). Int J Behav Nutr Phys Act 2015, 12:113. pmid:26377803
  31. 31. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG: Calibration of two objective measures of physical activity for children. J Sports Sci 2008, 26(14):1557–1565. pmid:18949660
  32. 32. Muthén LK, Muthén BO: Mplus User’s Guide. Eighth Edition. Los Angeles, CA: Muthén & Muthén; 1998–2017.
  33. 33. Enders CK, Bandalos DL: The Relative Performance of Full Information Maximum Likelihood Estimation for Missing Data in Structural Equation Models. Structural Equation Modeling 2001, 8(3):430–457.
  34. 34. Nylund K, Asparouhov T, Muthén B: Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study. Structural Equation Modeling 2007, 14(4):535–569.
  35. 35. Yang C: Evaluating latent class analyses in qualitative phenotype identification. Computational Statistics and Data Analysis 2006, 50:1090–1104.
  36. 36. Sen S, Bradshaw L: Comparison of Relative Fit Indices for Diagnostic Model Selection. Appl Psychol Meas 2017, 41(6):422–438. pmid:29881100
  37. 37. Lo Y, Mendell N, Rubin D: Testing the number of components in a normal mixture. Biometrika 2001, 88(3):767–778.
  38. 38. McLachlan G, Peel D: Finite mixture models. New York: Wiley; 2000.
  39. 39. Sinha P, Calfee CS, Delucchi KL: Practitioner’s Guide to Latent Class Analysis: Methodological Considerations and Common Pitfalls. Crit Care Med 2021, 49(1):e63-e79. pmid:33165028
  40. 40. Chung H, Lanza ST, Loken E: Latent transition analysis: Inference and estimation. Stat Med 2008, 27:1834–1854.
  41. 41. Nylund-Gibson K, Grimm R, Quirk M, Furlong M: A Latent Transition Mixture Model Using the Three-Step Specification. Structural Equation Modeling A Multidisciplinary Journal 2014, 21:1–16.
  42. 42. Bakk Z, Vermunt JK: Robustness of stepwise latent class modeling with continuous distal outcomes. Structural Equation Modeling 2015, 23:20–31.
  43. 43. Evenson K R., Wen FHales D, Herring AH: National youth sedentary behavior and physical activity daily patterns using latent class analysis applied to accelerometry. International Journal of Behavioral Nutrition and Physical Activity 2016, 13:55. pmid:27142304
  44. 44. Kemp BJ, Parrish AM, Batterham M, Cliff DP: Changes in subdomains of non-organized physical activity between childhood and adolescence in Australia: a longitudinal study. Int J Behav Nutr Phys Act 2022, 19(1):73. pmid:35752838
  45. 45. Salway R, Walker R, Sansum K, House D, Emm-Collison L, Reid T, et al: Screen-viewing behaviours of children before and after the 2020–21 COVID-19 lockdowns in the UK: a mixed methods study. BMC Public Health 2023, 23(1):116.
  46. 46. Saunders TJ, Chaput JP, Tremblay MS: Sedentary behaviour as an emerging risk factor for cardiometabolic diseases in children and youth. Can J Diabetes 2014, 38(1):53–61. pmid:24485214
  47. 47. Sport England: Active Lives Children and Young People Survey, Academic Year 2020–21. UK: Sport England; 2021.
  48. 48. Sport England: Active Lives Children and Young People Survey, Academic Year 2021–22. UK: Sport England; 2022.
  49. 49. Women’s Sport and Fitness Foundation: Changing the Game for Girls. London. [cited 2023 May 30] Available from: www.wsff.org.uk/system/1/assets/files/000/000/285/285/f4894dccf/original/Changing_The_Game_For_Girls_Final.pdf
  50. 50. Andersen OK, Gebremariam MK, Kolle E, Tarp J: Socioeconomic position, built environment and physical activity among children and adolescents: a systematic review of mediating and moderating effects. Int J Behav Nutr Phys Act 2022, 19(1):149. pmid:36510203
  51. 51. Higgerson J, Halliday E, Ortiz-Nunez A, Brown R, Barr B: Impact of free access to leisure facilities and community outreach on inequalities in physical activity: a quasi-experimental study. Journal of epidemiology and community health 2018, 72(3):252–258. pmid:29330166
  52. 52. Inchley J, Dorothy Currie D, Young T, Samdal O, Torsheim T, Augustson L, et al (eds.): Growing up unequal: gender and socioeconomic differences in young people’s health and well-being. Health Behaviour in School-aged Children (HBSC) study: international report from the 2013/2014 survey. Copenhagen: WHO Regional Office for Europe; 2016.