Descriptive analysis of preschool physical activity and sedentary behaviors – a cross sectional study of 3-year-olds nested in the SKOT cohort

Nested cross-sectional data used in this study are from the 3-year examination in the longitudinal observational SKOT cohort study, which has been described in detail previously [14, 15]. In short, based on random selection from the National Danish Civil Registry of infants born in the greater Copenhagen area, 330 children were enrolled in the study between April 2007 and May 2008 at age 8.5 months and examined at 9, 18, and 36 months. Inclusion criteria were being a singleton infant born at term (≥37 weeks of gestation), with no diseases expected to affect growth or nutritional intake. Anthropometric measures and questionnaire data were collected at all examinations. Twenty children were lost to follow-up before age 3 years, and one child was excluded due to severe chronic disorder with late manifestation. The remaining 309 children still eligible for participation were invited to the 3-year examination. All 3-year examinations (± 3 months) took place at the Department of Nutrition, Exercise and Sports, Frederiksberg, Denmark during the period from October 2009 to October 2010. Two-point-6 % of the children attended daycare in private homes, 12.6% attended day nursery, and 84.8% attended kindergarten (preschool) – together they are referred to as children attending daycare.

Body weight and height were measured according to standard procedures, as described elsewhere [14, 15]. Means of available measures were used in all analyses. Age- and sex-specific Z-scores for body mass index (BMI-Z) were calculated by the software WHO Anthro (Department of Nutrition, World Health Organization, Geneva, Switzerland). The proportions of children who were overweight or obese were calculated according to sex- and age specific body mass index (BMI) cut-points [16].

A proxy of socio-economic status was obtained through information on mothers’ level of education in a parent-completed questionnaire. We defined educational categories based on length of completed academic education, completion of vocational education or training, and gymnasium/high school or lower secondary school education. In this Danish context, academic education refers to higher educations at universities or university colleges, vocational education or training refers to education offered at special state-funded vocational schools, gymnasium/high school refers to a 3-year academic-oriented upper secondary educational programme, and lower secondary school refers to education provided in state schools or private schools from Year 0 to Year 9 or 10. Information was also gathered regarding night sleep (“At what time does your child usually fall asleep in the evening?”, “At what time does your child usually wake up in the morning?”), and day naps (“How many days per week does the child sleep during the day?”, “For how long does the child sleep on days where the child sleeps during the day?”).

Information regarding time and date of dropping-off/picking-up children from daycare and date of any sick days was obtained from a parent-completed log.

In this study, continuous 24-h accelerometer data were recorded, resulting in a need to separate PA during waking hours from sporadic movements made during night-sleeping. We used Individual Manual Inspection (IMI) to identify specific times of waking up in the morning and falling asleep in the evenings for each child for each day of monitoring. This method involved manual visual inspection of activity graphs produced by the sleep analysis module in the Actilife software. Falling asleep was defined as the time during which more continuous patterns of PA were followed by at least 5 min of zero counts and an obvious change in behavior pattern to a few single volatile sporadic movements, thereby allowing for tossing and turning in bed. Likewise, wakeup time was defined as the time where more steady patterns of PA followed at least 5 min of zero counts, with allowance given for single sporadic movements during sleep prior to the 5 min zero counts. The visual inspection was performed twice by two trained research assistants and the results were compared. Scoring differences of more than 10 min were double-checked and reevaluated by the second research assistant and LBC. Only PA occurring during waking-hours, as identified by morning wake-up and times of falling asleep at night, was included in the analyses. We used information obtained in the parent-completed questionnaire to compare morning wake-up time and sleep time in evenings as identified by the IMI and parents’ reports, respectively. Additionally, we used the filtering of 20 min or more of consecutive zero counts to exclude sleep time in children who were napping during the daytime.

Based on accelerometer data defined by the use of the Individual Manual Inspection approach and information on daycare attendance provided by parents’ logs, time-stamped data were analyzed separately for predefined typical settings of the day and week (i.e., all days overall, sick days, daycare-days (DC-days), time before daycare, time in daycare, time after daycare, and non-day-care-days (non-DC-days). All children except one were enrolled in childcare, but some of the children reported having some weekdays without daycare attendance, and 40 children did not report any days in daycare during the monitoring period. These children may have worn the accelerometer during days off or during non-typical weeks like vacation time. Weekdays where parents reported their child did not attend daycare had more similarities with weekend days than with weekdays where parents reported that children where in daycare (data not shown). Accordingly, days were recoded as either DC-days or non-DC-days.

All accelerometer outcomes were processed separately across the following domains: total time all days, sick days, DC-days, time before daycare, time in daycare, time after daycare, and non-DC-days. Crude hourly mean total PA intensity was processed to illustrate more general PA levels throughout the day. To assess if PA or SB differed across settings – (e.g., when children were in daycare and not in daycare), linear regressions were performed with the relevant accelerometer output included as the dependent variable. “Settings” were treated as a dummy variable and included as independent variable in the model. “Cluster option” in Stata IC 14.0 was used on child ID to obtain robust standard errors thereby accounting for repeated observations. The statistical significance of the overall influence of the different settings was evaluated using a Wald test. In case of overall significant influence due to settings, we post hoc used the “lincom” command in Stata to examine if PA or SB outcomes differed across the following combinations of settings: 1) time before daycare vs. time in daycare, 2) time before daycare vs. time after daycare, 3) time in daycare vs. time after daycare. Furthermore, potential differences were examined between DC-days and non-DC-days.

Differences in proportions of children meeting PA recommendations across settings and sex were examined based on the chi-square test. Quartiles based on mean total PA (CPM) were generated post hoc, and subgroups of low active (1st quartile of CPM) and high active (4th quartile of CPM) children were compared for MVPA and SB. Analyses were restricted to individuals with complete data and statistical significance was based on α = 0.05.

Participant characteristics are presented in Table 1. Two-hundred-sixty-four children completed the 3-year examination, whereas 7 withdrew from the study before the 3-year examination, 12 children did not respond to the invitation, and 26 children were unable to participate. Four accelerometers were lost in the mail or not returned by the families, and 29 children provided less than 4 valid days of monitoring. No instrument malfunction occurred.

Of the children’s mothers, 40.7% had long academic education (>4 years), 35.5% had medium length academic education (3–4 years), 11.3% had short academic education (<3 years), 8.7% had vocational education or training, and 3.9% had no education above gymnasium/high school or lower secondary school. The 231 children (87.5%) who provided valid PA data were slightly lighter (14.5 ± 1.5 kg vs. 15.0 ± 1.5 kg, p = 0.0018), shorter (95.7 ± 3.4 cm vs. 97 ± 3.5 cm, p = 0.04), and had a lower BMI (15.8 ± 1.14 vs. 16.5 ± 0.96, p = 0.006) compared to the children who did not provide valid PA data, but they did not differ significantly by age, sex, overweight/obesity status, or maternal educational level (Additional file 1: Table S1).

Overall, children had a mean of 7.0 ± 1.1 valid monitoring days and on average 12.6 ± 0.8 h of valid wear time per day. On average, children attended daycare 4.1 ± 1.2 days, and daily valid wear time while in daycare was 6.4 ± 1.2 h (Table 2). Children were dropped off in daycare at 8:32 AM ±42 min and picked up at 3:31 PM ± 43 min, and thus spent an average of 6.98 ± 1.08 h per day in daycare. There were no significant differences in valid wear time by sex, reporting of any days in daycare, or napping during the day (Additional file 2: Table S2).

Overall, boys and girls spent an average of 1.4 ± 0.3 h/day (11.2 ± 2.6% of daily time) and 1.2 ± 0.4 h/day (9.7 ± 2.9% of daily time) in MVPA, respectively. Likewise, boys and girls accumulated 6.7 ± 0.8 h (53.5 ± 5.1% of daily time) and 6.8 ± 0.9 h (54.7 ± 5.9% of daily time) of daily SB time, whereas 5.9 ± 0.7 h (46.5 ± 5.1% of daily time) and 5.7 ± 0.8 h (45.3 ± 5.9% of daily time) of non-SB time were accumulated in boys and girls (Table 3).

There were no differences between children’s wake up time according to the IMI approach and parents’ reports, respectively (06:47:31 AM ±31.7 min vs. 06:49:06 AM ±40.1 min, p = 0.55). However, children fell asleep later based on the IMI approach compared to parent provided information (asleep: 20:35:28 PM ± 46.6 vs.19:58:31 PM ± 32.8 min, p < 0.0001).

Crude hourly mean total PA level (CPM) revealed patterns of low PA levels around 11:30 AM, 02:30 PM, and 06:00 PM on DC-days. Furthermore, on DC-days, differences in PA patterns were observed between boys taking a nap and boys not taking a nap during early (12:30 PM-02:00 PM) and late (04:00 PM-07:00 PM) afternoon. In girls, this phenomenon could only be observed during the early afternoon (12:00 PM-02:30 PM) (Fig. 1). Post hoc sub-analyses supported these findings; overall, on DC-days, on non-DC-days, and in daycare time, the girls who took a nap registered lower mean total PA compared to those girls not taking a nap. Boys taking a nap were more physically active after daycare on DC-days compared to boys who did not take a nap (Additional file 2: Table S2).

Significant differences among settings were observed for all PA and SB variables in both girls and boys (all p < 0.0001). Generally, post hoc analyses revealed that the daycare setting contributed to higher PA levels and lower SB than non-daycare. In girls, higher total PA (CPM) was obtained during daycare compared to time before and after daycare (β = 228, p < 0.001 & β = 39, p = 0.08, respectively). Furthermore, girls accumulated more time in MVPA in daycare compared to time before daycare (β = 4.8%, p < 0.001). Similarly, compared to time before and after daycare, lower amounts of SB time were accumulated during daycare (β = −12.2%, p < 0.001 & β = −3.8%, p < 0.001, respectively). Opposite results were observed for non-SB time. On days during which they attended daycare, girls engaged in significantly less SB but more MVPA compared to days without daycare attendance (β = −1.6%, p = 0.002 & β = 0.57%, p = 0.015). Compared to all other settings with the exception of sick days, girls were least active and accumulated most SB time in the morning before arriving to daycare.

Boys engaged in the highest level of total PA in daycare compared to time before and after daycare (β = 277, p < 0.001 & β = 82, p = 0.006, respectively). Furthermore, boys accumulated more time in MVPA during time in daycare compared to time before and after daycare (β = 6.2%, p < 0.001 & β = 2.0%, p < 0.001, respectively). Also, during time in daycare boys accumulated less SB time compared with time before and after daycare (β = −15.4%, p < 0.001 & β = −7.6%, p < 0.001, respectively). DC-days in boys consisted of more time spent in PA and less SB than non-DC-days (CPM: β =29, p = 0.046, %MVPA: β = 0.83, p = 0.007, %SED: β = −2.3, p < 0.001, respectively). As observed in girls, boys were least active and accumulated most SB time in the morning hours before arriving to daycare – this was true for all variables and settings with the exception of sick days.

Boys exhibited higher levels of total PA (CPM) on all days overall (p < 0,001), on DC-days (p = 0.01), and when in daycare (p = 0.007) when compared to girls. In daycare, boys also accumulated less SB time and more non-SB time compared to girls (p = 0.009). Furthermore, significant sex differences in MVPA were observed across all settings with the exception of sick days and time after daycare (all p-values < 0.03) (Table 3).

Overall, all boys and girls fulfilled the recommendation of at least 3 h of activity per day. For sick days specifically, these numbers decreased to 82% for girls and 90% for boys, respectively (Table 4). Overall, 89% of the boys met the recommendation for 5-year-olds of at least 1 daily hour of MVPA. A significantly (p = 0.001) smaller proportion (72%) of girls met this recommendation (Table 4). Furthermore, fewer girls than boys (p = 0.015) met the MVPA recommendations on DC-days. Generally, more children met MVPA recommendations on DC-days than on non-DC-days although these differences only were significant in boys (boys: p = 0.02, girls: p = 0.24). Finally, as expected, the lowest proportions of girls (18%) and boys (40%) were observed to fulfill MVPA recommendations on sick days (Table 4).

Large differences in MVPA and SB were consistently observed across all settings between the most (4th PA quartile) and the least active (1st PA quartile) children. In contrast to the total sample significant sex differences were also observed in time after daycare (p = 0.04) and on non-DC-days (p = 0.0006) in the least active children. No significant sex differences were observed in any setting in the most active quartiles of boys and girls (all P > 0.11) (Table 5).

Overall significant influences of settings on MVPA and SB were observed for the most and least active girls and boys (all p < 0.03). Post hoc analyses revealed that both girls and boys in the fourth PA quartile accumulated significantly less SB time in daycare compared to time before daycare (girls: β = −13.6%, p < 0.001; boys: β = −17.9%, p < 0.001) and time after daycare (girls: β = −4.36%, p = 0.07; boys: β = −7.8%, p = 0.003). Compared to all other settings, both high- and low-active girls and boys participated in the least MVPA and most SB time during morning hours before daycare (Table 5).

Overall, only 7% of the least active girls and 59% of the least active boys fulfilled the recommendation for 5-year-olds of at least 1 h of MVPA per day. In high active girls and boys these numbers were 97% and 100%, respectively (data not shown).

We observed clear patterns of low PA on DC-days around 11:30 AM, 2:30 PM and 6:00 PM, possibly reflecting consistent meal times across DC-days and the daycare institutions. The forenoon (10:00 AM-11:00 AM) and early afternoon (12:00 PM-02:00 PM) were characterized by periods of high PA levels. This corresponds well with commonly scheduled playtime in Danish preschools. Although day-time-specific routines to some extent are expected to affect children’s PA level across different daycare centers (e.g., around lunch time), it is noteworthy how different daycares seem to follow the same hourly routines throughout the day. Such patterns did not appear on non-DC-days, indicating more flexibility across individual families on those days.

The fact that results revealed higher PA levels in daycare than out of daycare supports previous findings in preschoolers observed in a Danish study [19]. Olesen et al. reported lower PA levels in children during leisure time than during their preschool day. Furthermore, the authors noted substantially lower total PA levels on weekend days compared to weekdays. Interestingly, these differences were reported to change throughout the day (i.e., children were more active on weekdays compared to weekend days from 8 AM to 4 PM but more active in weekends from 4 PM to 8 PM). These findings correspond well to our findings of children being more active on weekdays during daycare hours. Generally, previous studies from other countries have reported some similar and some contradictory results when examining preschool vs. non-preschool PA by comparing weekdays and weekends [20‐22].

High PA levels in the daycare setting are in contrast to findings observed in other countries when describing PA levels within the childcare setting [12]. Unlike our study, these observations, though, did not take into account the potential individual differential impact of time and place on children’s PA, although the authors noted that other data available suggest that the problem of low PA is not unique to the child care setting. Furthermore, these differences are in part likely due to heterogeneity between samples, differences in monitor type, measurement protocol (e.g., minimum criteria for valid day and time of start/stop in morning/evening). Hesketh et al. previously examined preschoolers’ PA in the UK by specifically exploring the potential differential impact of time and place including childcare. In accordance with our observations, they reported that young children accumulated more MVPA in childcare compared to when at home [23].

According to Danish daycare law, it is compulsory for all kindergartens and day nurseries in Denmark to create a written pedagogical curriculum plan describing targets for working with children within specific themes, including social competences and relations, nature, and the body and body movements. Typically, within the Danish context it is also mandatory for preschool children to spend time outdoors on a daily basis and outdoor environments, such as playground and portable equipment, are usually available. A recent review conducted by Tonge et al. identified size, use, and presence of outdoor environment as consistent PA and SB correlates in preschoolers [24]. Furthermore, outdoor childcare hours have been suggested to comprise more PA than indoor hours [25]. Other factors, such as presence of peers and peer prompts, in the daycare setting may contribute to our findings of elevated PA levels during daycare hours [24].

Maternal and paternal behaviors have previously been observed to associate with children’s home- and neighborhood-based SB and MVPA [26], and mothers’ SB during the morning period has related negatively to children’s MVPA [27]. We speculate that early morning time periods, especially during less flexible weekdays, are characterized by essential daily living activities (e.g., bathing, dressing, and eating) with little options for parents and children to be active before daycare attendance. Although the present study was not designed to identify specific PA- and SB-correlates but more should be seen as a descriptive hypothesis-generating study, our findings indicate that morning and afternoon during weekdays are obvious intervention opportunities in preschoolers. Promotion of PA and reduction of SB during early morning-hours seems difficult because of a need to perform practical tasks which might leave fewer opportunities to be physically active. Generally, there may be more potential for promoting parents’ and children’s home-based PA on weekdays in the afternoon after daycare or on weekends. It is possible that factors such as pedagogical curriculum and institutional policy, outdoor environment, and the potential importance of peers might influence attempts to increase childcare PA levels.

It is important to note that we observed considerable differences between low- and high-active children across all settings and PA variables examined. This is interesting and indicates that low-active children are substantially and consistently less active compared to highly active children, irrespective of the context. Most studies find that boys are more active than girls [28]; however, sex-based differences were not observed among the most active children in the current study. These findings support that targeting of PA and non-SB seems conceivable and should be favored across all settings in the least active children.

Previous PA reviews conducted in preschoolers generally have confirmed low PA levels, high levels of SB [12], and low prevalence of children adhering to recommendations [29], or the authors have been unable to make conclusions regarding the PA levels due to methodological inconsistencies [30]. Cardon et al. [20] previously reported that only 7% of Dutch preschoolers engaged in 60 min of MPVA per day, but they used a cutoff point of ≤615 counts/15 s to define MVPA as suggested by Sirard et al. [31]. In our study, we reported that 72% of girls and 89% of boys accumulated at least 60 min of daily MVPA which is recommended for children when they reach the age of 5 years. However, when we tried to apply the Sirard cut-points [31] results became quite comparable to the findings observed by Cardon et al., as proportion of children complying with MVPA guidelines decreased substantially to 7% and 12%, respectively (results not shown). Similarly, we found that all children met the recommendation of 3 h of daily engagement in PA with any activity. However, when applying high SB cut-points, as suggested by Sirard et al. [31], only 1% of children in the present study fulfilled recommendations of at least 3 h PA with any intensity per day (results not shown). This illustrates the impact that the selection of cut-points will have on overall conclusions. We would argue that a fairly “low” cut-point to define SB seems preferable in young children. A good example of why this might be preferable is a case such as sitting on the floor instead of on a chair, which often involves small movements to shift position. These small movements may be regarded as small breaks in SB time. Considering the intermittent nature of young children’s PA [32], free play activities in young children could be hypothesized to involve a high frequency of such breaks even though the overall activity (lying or sitting) in itself may qualify as SB according to some observational protocols (e.g., the Children’s Activity Rating Scale, CARS [33]). The use of ≤25 counts/15 s and ≤420 counts/15 s cut-points to define SB and MVPA, respectively, as applied in the present study, has previously been suggested by Trost et al. as the best cut-off thresholds to use in young children [34].

In the present study, both boys and girls were observed to accumulate most of their time in MPVA when in daycare (boys, 13% of daycare time and girls 10.8%). These numbers correspond well to previous results observed in US preschoolers indicating that 3–5-year-olds engage in 7.7 min of MVPA per hour of preschool attendance, corresponding to approximately 13% of the time [35].

We found that only 7% of the least active girls and 59% of the least active boys fulfilled the recommendation of at least 1 h of MVPA per day. Although MPVA recommendations do not apply for children younger than 5 years, these are discouraging results from a public health perspective. This, especially since early PA levels during childhood have been reported to be predictive for later PA levels [5, 6], and since risk factors for lifestyle-related diseases tend to cluster in the least fit and least active children [36]. Therefore, the least active children even in daycare or kindergarten are most likely in most need of early and targeted interventions.

For some participants, we were somewhat challenged in defining the precise time of wake-up. However, based on subjective information provided in the parental questionnaire, we found high agreement between times of wake-up based on IMI and information provided by parents. Sleep time, however, occurred later when recorded by IMI compared to parents’ reports. We hypothesize that differences could be due to parents reporting snuggle time instead of time where children actually fell asleep. We made great effort in trying to include all waking hours in the present study, and we believe that only limited time during morning hours and evening hours, which typically would be characterized by low PA levels and abundant SB time with the potential to downgrade out-of-daycare PA levels, falsely could have been eliminated from our analyses. Inspections of PA graphs for each day of monitoring (not shown) indicated that the children’s sleep was characterized by some small-scale “activity”, most likely from tossing and turning in bed. Thus, standard non-wear filters, as provided by various accelerometer software programs, may be insufficient to cut out time spent sleeping without simultaneously excluding low PA during waking hours.

No time-matched daytime napping data were available; however, we noted that on-and-off body movements and tossing and turning in the bed were observed in data files at time points where daytime napping roughly was expected to occur in subjects who reported daytime napping. This made it impossible for us to define valid times of falling asleep and wake-up when taking a nap based on the IMI approach. Therefore, higher PA levels as observed during early afternoon in children napping around noon compared to children not napping at noon most likely reflect that children either participated in high intensity playtime after lunch or were sleeping, and perhaps some of the nap time was misclassified as SB time. Short sleep duration has been found to be associated with increased risk of childhood obesity [37]. Since daytime napping is natural during preschool years [38], naps may likewise be hypothesized to be beneficial for child health. On the other hand, it is far from clear how much of daytime napping children actually spend sleeping. Thus, we believe it is a discussion point whether or not napping during the day should be excluded from final analyses to avoid misclassification as SB time. We suggest that future studies carefully consider how to handle daytime napping when analyzing PA data.

A major strength in the present study is that PA was assessed objectively with high wear compliance and that PA analyses encompassed all waking hours as defined by the IMI approach based on 24 h registration. Use of sophisticated time-stamped data processing based on the detailed information provided by parental logs, including institution check-ins and check-outs for each child, made it possible for us to isolate PA and SB outcomes in a number of relevant everyday settings, including time in and out of daycare.

The individual daycare has previously been identified as an important predictor of PA [35, 39, 40]. We did not take into account the specific daycare center in our analyses since the unit of recruiting in our study was the single child and not clusters of institutions. In the area from which the children were recruited there are 400–500 daycare centers. Thus, children who were enrolled in our study attended numerous different daycare institutions, which have contributed to more robust findings.

We did not distinguish between non-DC-days during weekdays and weekend days due to statistical power considerations. This could potentially have masked difference in children’s behaviors across these day types. However, weekdays for which parents reported that children did not attend daycare had more similarities with weekend days than with weekdays where parents reported that children were in daycare (data not shown). It was a limitation that it was not possible for us to discriminate between children’s time spent indoors and outdoors during daycare as time spent outdoors has been described as a predictor of preschool PA [28].

Accelerometer-determined intensity thresholds are a major issue when trying to quantify the minutes spent in specific PA intensities, and basically no uniform consensus exists regarding which cut-points are best to use in order to estimate valid amounts of time that children of different ages spend in different PA intensities during free living. However, relative differences in PA levels and SB across different everyday settings can still be meaningfully described if one recognizes the limitation of the cut-point approach.

Specific activities undertaken during time in SB and MVPA were not considered in the present study. However, supplementing accelerometer data with more detailed contextual information could provide further valuable insights into which contexts and settings children typically engage in SB and low/high PA intensities. As in any other study where PA is assessed using hip-worn accelerometers, the inability to capture cycling, swimming, and loadbearing activities correctly is a limitation to our study. We speculate, however, that generally these types of behaviors only constitute relatively small parts of preschoolers’ total habitual PA.

Finally, the children in the SKOT cohort are primarily from well-educated, high-income families. Danish preschoolers’ PA levels have previously been found not to be associated with either household income [40] or parental educational level [39]. The association between socio-economic status and PA may not emerge until school age, but we cannot rule out the possibility that generalizability of the results observed may not extend beyond the type of population sampled in the present study (i.e., majority of children from families with a high socio-economic status).