Comparison of actigraphy‐measured and parent‐reported sleep in association with weight status among preschool children

This study compared weekday and weekend actigraphy‐measured and parent‐reported sleep in relation to weight status among preschool‐aged children. Participants were 3–6 years old preschoolers from the cross‐sectional DAGIS‐study with sleep data for ≥2 weekday and ≥2 weekend nights. Parents‐reported sleep onset and wake‐up times were gathered alongside 24 h hip‐worn actigraphy. An unsupervised Hidden‐Markov Model algorithm provided actigraphy‐measured night time sleep without the guidance of reported sleep times. Waist‐to‐height ratio and age‐and‐sex‐specific body mass index characterised weight status. Comparison of methods were assessed with consistency in quintile divisions and Spearman correlations. Associations between sleep and weight status were assessed with adjusted regression models. Participants included 638 children (49% girls) with a mean ± SD age of 4.76 ± 0.89. On weekdays, 98%–99% of actigraphy‐measured and parent‐reported sleep estimates were classified in the same or adjacent quintile and were strongly correlated (rs = 0.79–0.85, p < 0.001). On weekends, 84%–98% of actigraphy‐measured and parent‐reported sleep estimates were respectively classified and correlations were moderate to strong (rs = 0.62–0.86, p < 0.001). Compared with actigraphy‐measured sleep, parent‐reported sleep had consistently earlier onset, later wake‐up, and greater duration. Earlier actigraphy‐measured weekday sleep onset and midpoint were associated with a higher body mass index (respective β‐estimates: −0.63, p < 0.01 and −0.75, p < 0.01) and waist‐to‐height ratio (−0.004, p = 0.03 and −0.01, p = 0.02). Though the sleep estimation methods were consistent and correlated, actigraphy measures should be favoured as they are more objective and sensitive to identifying associations between sleep timing and weight status compared with parent reports.

characterised weight status.Comparison of methods were assessed with consistency in quintile divisions and Spearman correlations.Associations between sleep and weight status were assessed with adjusted regression models.Participants included 638 children (49% girls) with a mean ± SD age of 4.76 ± 0.89.On weekdays, 98%-99% of actigraphymeasured and parent-reported sleep estimates were classified in the same or adjacent quintile and were strongly correlated (r s = 0.79-0.85,p < 0.001).On weekends, 84%-98% of actigraphy-measured and parent-reported sleep estimates were respectively classified and correlations were moderate to strong (r s = 0.62-0.86,p < 0.001).Compared with actigraphy-measured sleep, parent-reported sleep had consistently earlier onset, later wake-up, and greater duration.Earlier actigraphy-measured weekday sleep onset and midpoint were associated with a higher body mass index (respective β-estimates: À0.63, p < 0.01 and À0.75, p < 0.01) and waist-to-height ratio (À0.004, p = 0.03 and À0.01, p = 0.02).Though the sleep estimation methods were consistent and correlated, actigraphy measures should be favoured as they are more objective and sensitive to identifying associations between sleep timing and weight status compared with parent reports.Over the past decades, an increasing amount of epidemiologic evidence suggests that insufficient sleep is associated with obesity (Morrissey et al., 2020;Wang et al., 2019).Sleep deprivation and a delayed sleepwake rhythm may disrupt the hunger-satiety hormonal balance (Taheri et al., 2004), influence physical activity levels (Foti et al., 2011;Merikanto et al., 2020), or simply provide more time for energy consumption (Nedeltcheva et al., 2009).Recent studies on preschoolers with parentreported sleep largely support the association between a short sleep duration and risk for obesity but lack data from objective sleep measures (Bolijn et al., 2016;Miller et al., 2018;Rangan et al., 2018;Wang et al., 2016).Actigraphy-measured shorter sleep and increased adiposity have been observed in school-aged children (Morrissey et al., 2020), but studies on preschool-aged children are scant with inconsistent findings (Krishnan et al., 2017;Wyszy nska et al., 2021;Xiu et al., 2020).
A comparison of parent-reported and actigraphy-measured sleep found that parents over-estimated sleep duration in preschool-aged children by close to an hour (Kushnir & Sadeh, 2013;Mazza et al., 2020), which is a significant amount for children's wellbeing (Matricciani et al., 2019).Therefore, it is as valuable to investigate whether associations between sleep and health outcomes also remain comparable when sleep estimation methods vary.This is especially important for early childhood studies that lack reliable sleep measurements and present many open questions regarding the associations between sleep and health.
Discrepancy in weekday and weekend night sleep is common for school-aged children and has been associated with obesity (Roenneberg et al., 2012;Stoner et al., 2018).Sleep debt accumulated during weekdays from early school times are compensated for on weekends, creating variability in weekday-weekend sleep.This discrepancy is already observed among young children, as preschoolers are at increased risk for greater weekend-weekday differences compared with children who do not attend preschool (Giannoumis et al., 2022).Furthermore, sleep comparability between studies accounting for weekday and weekend night differently remains unclear, and more studies comparing these differences by assessment method and in association with health outcomes are needed.
Accurate assessment tools are needed to better understand the relation between health outcomes and sleep timing and quality among children (Hjorth et al., 2012;Miller et al., 2018;Phillips et al., 2021).
Epidemiological studies have often relied on reported sleep or heuristic algorithms developed three decades ago that require guidance from subjective sleep logs to identify sleep onset and wake-up times (Cole et al., 1992;Sadeh et al., 1994).Furthermore, many commonly used sleep-identification algorithms have not been validated for hipplaced actigraphy, including the Sadeh algorithm (Sadeh et al., 1994), which is often favoured for physical activity research among children (Hjorth et al., 2012;Quante et al., 2018).Researcher access to offline raw actigraph data has improved (Sundararajan et al., 2021;Troiano et al., 2014), elevating the need for updated open-source data processing methods for increased interpretability and comparability both within and between study populations.
We applied a Hidden Markov Model (HMM)-based sleep/wake identification algorithm, which is an unsupervised, data-driven machine learning algorithm accounting for temporal sequencing that can be applied directly to actigraph data from different populations (Li et al., 2020).To the best of our knowledge, no previous study has compared the associations of parent-reported sleep with weight status to that of actigraphy-measured sleep among young children, especially for weekday and weekend nights, respectively.Therefore, the objectives of this study were to (1) compare weekday and weekend parent-reported sleep with actigraphy-measured sleep for preschoolaged children and (2) explore the associations of the two sleep assessment methods with weight status.

| METHODS
Cross-sectional data from the DAGIS (the Increased Health and Wellbeing in Preschools) study 2015-2016 was utilised.The DAGIS study investigated the socioeconomic differences in energy-balance related behaviours among preschool children in Finland and is described in detail elsewhere (Lehto et al., 2018).In total, 864 children aged 3-6 years and their families from 66 early childhood education and care centres from eight municipalities across southern and western Finland consented to participate in the study.The current study population consisted of 638 children, after exclusion of children with insufficient sleep data (less than four nights, of which two were weekend nights) from actigraph (n = 167) and additionally parent-reports (n = 25), as well as missing waist circumference, height, and weight measures (n = 34).There were no significant differences with respect to demographic or anthropometric measures, except that those children who were excluded from the analytical sample had significantly less highly educated parents compared with children who were included in the sample (p < 0.01).The study was approved in 2/2015 by the University of Helsinki Ethical Review Board in the Humanities and Social and Behavioural Sciences (#6/2015).

| Sleep data
Night time sleep onset and morning wake-up times were determined by parent-report and 24 h worn actigraph devices over a 7 day collection period.Adequate sleep data consisted of a minimum of four nights, with at least two week nights and two weekend nights.Weekend night was defined as the nights between Friday-Saturday and Saturday-Sunday.All children had sleep data for two weekend nights and only 0.01% (n = 6) children had weekday data for two nights, 0.01% (n = 4) for three nights, 8% (n = 49) for four nights, and 91% (n = 579) for five nights, respectively.

| Actigraphy-measured sleep
Actigraph sleep estimates were obtained from a hip-worn ActiGraph wGT3X-BT activity monitor (Pensacola, FL, USA), which contains a triaxial accelerometer.Children were instructed to wear the actigraph for 1 week, 24 h per day, excluding only water-based activities.Nonwear hours were defined as published previously with the GGIR package version 2.5 for processing of actigraph data (Migueles et al., 2019;van Hees et al., 2015van Hees et al., , 2022)).Non-wear time under 2 h was excluded from sleep duration periods.A valid day was defined as <8 h of nonwear time from a noon-noon window.
The HMM algorithm was applied to identify the sleep and wake states and further infer the sleep onset and wake-up times (Li et al., 2020).The data-driven HMM algorithm distinguishes sleep and wake states by identifying different movement patterns, as activity in gravitational acceleration under the sleep state are mostly near-zero with small values accommodating movement during sleep, whereas activity under the wake states are large values describing the active movements.The HMM-algorithm was utilised because it is neither heuristic, which uses subjective thresholds for all populations that may not be accurate especially for younger populations, nor does it rely on labelled data from specific populations to train the model before use, which limits use in new populations since labelled data are not always available.Instead, the HMM algorithm is individual-based and datadriven that can infer sleep and wake patterns accurately without labelled data.HMM has been validated previously against polysomnography with 85.7% overall accuracy, better than the embedded Actiwatch software algorithm and the widely used supervised UCSD algorithm in Jean-Louis et al. (2001), and HMM also has a high sensitivity of 99.3% for detecting sleep epochs (Li et al., 2020).As different individuals can have different activity and sleep habits, the data-driven HMM algorithm accounts for individual variation to learn the unique sleep and wake patterns for each individual and further calculate the sleep onset and wakeup time based on changes in the movement pattern for every night.The algorithm is available as an R package "hmmacc".
Six variables were formed from actigraphy-measured sleep: (1) night time sleep onset, (2) morning wake-up, (3) sleep duration, (4) time spent asleep within sleep duration, (5) wake after sleep onset (6), sleep efficiency, (7) sleep midpoint, and (8) weekend-weekday differences in sleep midpoint.Sleep duration period was defined as the total time between sleep onset and wake-up.Sleep efficiency was calculated as the time spent asleep within sleep duration period divided by the total sleep duration period.The sleep midpoint was defined as the half of the time passed between sleep onset and wake-up.
Weekend-weekday difference in sleep midpoint was defined as weekday sleep midpoint subtracted from weekend sleep midpoint.

| Parent-reported sleep
Parents were asked to report night time sleep and morning wake-up time in a 7 day sedentary behaviour diary for the same week the child was instructed to wear the actigraph.Sleep duration period was calculated as the length between parent-reported sleep onset and wake-up time.Sleep midpoint and weekday-weekend differences in sleep midpoint were calculated in accordance to the methods described previously for actigraphy-measured sleep.

| Anthropometry data
Body weight was measured to the nearest 0.01 kg by a trained researcher with portable bench scales (CAS PB-100/200), height was measured to the nearest 0.1 cm with stadiometers (SECA 217), and waist circumference was measured to the nearest 0.1 cm with measuring tapes (SECA 201).The waist-to-height ratio was calculated as the waist circumference (cm) divided by height (cm).The body mass index (BMI) was calculated as body weight (kg)/height 2 (m) and converted to age-and-sex-adjusted BMI (ISO-BMI).Since BMI varies by ethnicity in paediatric populations (Daniels et al., 1997), BMI classifications were defined by Finnish growth reference based on ISO-BMI defining underweight <17 kg/m 2 , overweight as 25-30 kg/m 2 , and obese as >30 kg/m 2 (Saari et al., 2011).BMI was categorised as either "not overweight" or "overweight", which included children with overweight and obesity.

| Sociodemographic and energy-balance related data
Parents reported the child's date of birth and sex, as well as both parent's education level.Age was used as a continuous variable determined from the parent-reported date of birth.The highest reported parental education level was used as an indicator of socioeconomic status (SES) and categorised into three groups: high school or vocational school diploma or less, college degree (i.e.associates or bachelor's degree), or master's degree and higher (i.e.licentiate or doctoral degree).The reported number of household adults were separately categorised as one or over one.All additional children reported to be in the same household were considered to be siblings, and participants were further categorised as either having siblings or not.The research season was categorised as fall (September-October), early winter (November-December), and late winter/spring (January-April).
Physical activity as well as sedentary time were expressed as minutes per hour measured from actigraph using Butte et al. thresholds to distinguish activity intensities (Butte et al., 2014).Parents were instructed to report the children's daily TV, computer, or tablet and phone screen time in the 7 day sedentary behaviour diary (Määttä et al., 2017), which was expressed as a continuous variable (min/day).
Energy intake was calculated from 3-5 day food records, which included at least two weekdays and one weekend day, collected from parents and preschools (Korkalo et al., 2019).

| Statistical analysis
Actigraphy-measured and parent-reported mean sleep parameters were divided into quintiles and comparison of participants in categories were classified as same, same or adjacent, or opposite quintiles.Spearman's correlation coefficient was used to compare the sleep estimate methods and paired t-tests were used to estimate mean differences.Additionally, Bland-Altman plots comparing agreeability of actigraphy-measured and parent-reported sleep parameters during weekdays and weekends are illustrated in supplemental materials.Differences in descriptive variables between BMI categories and differences between weekday and weekend night sleep estimates were tested with analysis of variance for continuous variables and chi-squared tests for categorical variables.Multivariate regression was performed to assess associations with sleep and weight status as continuous variables.
Descriptive variables that significantly or nearly significantly differed by BMI category were included in the models as covariates.Model 1 included adjustment for age, sex, and number of adults in households (1 or >1 adult).Model 2 included model 1 variables and energy intake and screen time.To assess the impact of SES, parental education level was additionally assessed as a potential covariate but was excluded in the final models because the results did not change significantly.Associations between the time spent asleep within sleep duration period and weight status provided similar findings and therefore, were not separately reported.General additive model diagrams were utilised to further understand significant associations from adjusted models.The moderation effects of age and sex were additionally tested in adjusted regression models.Additional linear regression was performed for statistically significant results with underweight children removed from the reference groups (n = 21) and logistic regression was performed with BMI categorised to further elucidate the associations between sleep as a continuous variable and weight status.Supplementary analyses on the associations between sleep and weight status in children with five nights of weekday sleep (n = 579) is provided to better assess typical weekday sleep behaviours.Values of p < 0.05 were considered statistically significant.Analysis was performed with statistical software R version 4.1.2(R Core Team, 2020).

| Study population
Of the 638 participants, 16% (n = 102) were classified as overweight (13%) and obese (3%).Children with overweight and obesity were more likely older, boys, and had a greater energy intake (Table 1).Children categorised as overweight were more likely to have an earlier sleep midpoint and onset according to actigraphy estimates, as well as to have an earlier wake-up for both weekdays and weekends (Table 2).The sleep midpoint was later on weekends compared with weekdays for both parent reports (mean ± standard deviation: 29 ± 32 min) and actigraphy measures (26 ± 27 min).Statistically significant differences were found between weekday and weekend nights (p < 0.001) in both actigraphy-measured and parent-reported sleep for all sleep estimates.

| Comparison of actigraphy-measured and parent-reported sleep measures
Approximately 91% (n = 579) of the study participants had data for seven nights (five weekday and two weekend nights) and 99% (n = 632) had data for at least five nights (three weekdays and two weekends).Actigraphy-measured and parent-reported sleep correlated strongly on weekday variables (r s = 0.79-0.8,p < 0.0001), with weekday (r s = 0.79, p < 0.0001) and weekend (r s = 0.62, p < 0.0001) sleep duration periods displaying the lowest correlations (Table 3).

Comparison of weekend and weekday actigraphy-measured and
T A B L E 1 Background characteristics overall and according to categorised BMI from the DAGIS study (n = 638)   parent-reported sleep found that 98%-99% of children on weekdays and 84%-98% on weekends were classified into the same or adjacent quintile.The proportion of children classified into opposite quintiles ranged from 1.1%-2.2%on weekdays and 0.8%-16% on weekends.
The greatest misclassification (16%) was found for sleep duration period on weekends.
Overall parent-reported sleep duration period was estimated to be 37 min higher than actigraphy-measured sleep (Table 4).Parents reported earlier sleep onset by 23 min for weekday and 17 min for weekend, as well as later wake-up times, including 14 min for weekday and 22 min for weekend than observed by actigraph estimates.No differences were observed regarding weekend midpoints, but weekday midpoints were earlier with parent-reports than actigraphy-measured sleep.Bland-Altman plots display general agreement between actigraphy-measured and parent-reported sleep measures for weekdays and weekends and display the systematic error of underestimation of sleep onset and overestimation of wake-up times.Nonetheless, variation is moderate to fairly high among all sleep measures, with 3%-7% of data outside the set confidence intervals in the plots (Figure S1).

| Associations with weight status
Neither parent-reported nor actigraphy-measured sleep duration period estimates were significantly associated with BMI or waist-to-height ratio ( for BMI after further adjustment.Similarly, later actigraphy-measured weekday sleep midpoint was associated with lower BMI and waistto-height ratio in both models, but not with respective parent-reported estimates.For every 1 h increase in actigraphy-measured sleep midpoint during weekdays, there was a decrease in both BMI (β-estimate: À0.75, 95% CI: À1.30 to À0.21, p < 0.01) and waist-to-height ratio (β-estimate: À0.01, 95% CI: À0.01 to À0.001, p = 0.02).Similar associations remain when only including children with all five weekday nights (Table S1).
The modelled associations between weekday sleep onset and midpoint are illustrated in Figure 1.For both weight status measures, optimal sleep onset falls between 21:00-22:30 and sleep midpoint between 2:00-3:30.
Age did not significantly moderate any other associations of BMI or waist-to-height ratio with sleep midpoint.Sex did not have any moderation effects.

| DISCUSSION
This study found that purely data-derived actigraphy-measured sleep and parent-reported sleep were generally comparable, though parentreported estimates consistently observed earlier sleep onset and later wake-up times.The greatest difference was in the sleep duration period, which averaged 37 min higher in parent-reported than actigraphymeasured estimates for both weekday and weekend nights.Regardless of the sleep estimation method, children tended to sleep longer and have later sleep timing on weekends compared with weekdays.This study also investigated the associations between actigraphy-measured and parentreported sleep measures and weight status and found the sleep estimation methods were comparable.Still, actigraphy-measured sleep parameters were more sensitive to detecting significant associations with weight status.Findings observed that earlier weekday sleep onset and earlier weekday midpoint were associated with higher risk of being overweight.
Proxy reporters, e.g.parents, are prone to overestimation of sleep duration, as they may report the time their child was put to bed or simply not been able to accurately judge sleep onset and wake-up times (Kushnir & Sadeh, 2013;Martinez et al., 2014;Mazza et al., 2020;Short et al., 2013).Though parents consistently reported earlier sleep onset and later wake-up times compared with actigraphy-measured estimates, parent-reported sleep onset were Regardless of the sleep estimation method, the associations with sleep duration period and weight status were non-significant among the study participants.A longitudinal study on 3857 children aged 0-7 years found that the total daily sleep duration was not predictive of later BMI z-score and that longer reported sleep duration was only associated with lower BMI between the ages of 6 to 7, indicating that associations with sleep duration and adiposity may develop in older children (Hiscock et al., 2011).As age increases and caregiver dependency naturally decreases, children's autonomy in food choices also increases.This greater control over food choices may be a considerable factor behind the sleep-obesity nexus as children grow.Supporting this theory, longitudinal analysis from a randomised controlled intervention on 2-6 year old children already predisposed to obesity, found inverse associations between reported night time sleep duration and change in BMI to be mediated by energy intake (Rangan et al., 2018).
Somewhat surprisingly, our study observed that later sleep onset and later midpoint on weekdays were associated with lower BMI and waist-to-height ratio, especially among children under 5 years of age.
Studies assessing associations with sleep midpoint and weight status are limited in children.One cross-sectional study on 3-6 year old children found no association with sleep midpoints from parent reports, despite children having notably earlier sleep midpoints compared with the current study population (Pattinson et al., 2018).Similar to sleep duration, the associations between later sleep timing and increased adiposity may begin to form in later childhood.For example, a study on 115 primary school-aged children found a later weekday sleep midpoint to associate with higher BMI (Martoni et al., 2016).However, similar to our findings, overweight children were reported to have earlier mean sleep midpoints compared with normal weight children and, therefore, the associations previously reported in the primary-school children may have been driven by the children with obesity (6%) who had a later mean sleep midpoint (Martoni et al., 2016).
In the current study, the association between later sleep timing and lower weight status observed on weekdays seem to be driven by because naptime is offered daily during preschool.Napping is common for children under 5 (Kaar et al., 2020), and can be an important contributor to overall sleep duration as well as influence sleep timing.
In the current study, there was little variation in BMI with a low number of children classified as overweight and obese.Other studies on children have reported associations solely between sleep and obesity (Ji et al., 2018;Miller et al., 2018), which may produce more pronounced associations with sleep measures.However, this study population of preschoolers from western and southern Finland had a 3% prevalence of obesity, which is slightly lower than the national averages (4% of girls and 8% of boys) in Finland (Mäki et al., 2018), and a more representative study population may be needed to examine the detailed associations between sleep and obesity.
One strength of this study is the relatively large sample size with simultaneously collected data on at least two weekdays and weekend nights, with 1 week of sleep data for over 90% of participants.Additionally, the purely data-driven, HMM-algorithm is a feasible and objective method to measure sleep which captures individual variations in large populations without specificity to study demographic.
Since actigraphy devices measure sleep by defining sleep and wake states through acceleration, they can be prone to overestimation (e.g.lying still in bed) or underestimation (e.g.movement during sleep) of sleep.Similarly, although the HMM-algorithm can attempt to capture daytime sleep, the validity of the sleep episodes should be confirmed or compared against reported sleep, as daytime estimations can be heavily confounded by sedentary time.However, it is not uncommon to only study associations between night time sleep and weight status among preschool-aged children (Pattinson et al., 2018;Rangan et al., 2018;Wang et al., 2016).Furthermore, this study did not account for heritability, which has been found to influence actigraph-measured sleep duration among school-aged children (Breitenstein et al., 2021).The direction of associations or causality cannot be concluded from the findings because of the cross-sectional design.We also note that two nights of weekend sleep data may not be enough to assess habitual sleep, and future studies should consider data collection to include a longer time period when feasible.Lastly, our findings should be cautiously extrapolated, as our study popula- sleep, chronotype, hidden Markov model, machine learning, overweight and obesity, sleep rhythm, sleep-obesity nexus 1 | INTRODUCTION

T
A B L E 2 Weekday and weekend sleep and wakeful activity measures according to overall and categorised BMI groups (not overweight and overweight) from the DAGIS study (n General additive model diagrams of changes in BMI score (with 95% confidence intervals) with weekday sleep onset and sleep midpoint from the DAGIS study participants (n = 638) closer and wake-up times farther from actigraphy-measured estimates on weekends compared with weekdays.These results indicate that the parent's perception of children's sleep timing is likely influenced by daily routines.For instance, parents may be more vigilant in regards to time on mornings with early preschool and work schedules (i.e.weekdays versus weekends), enabling them to report more accurate wake-up times.Moreover, parent-reported sleep onset may be more accurate on weekends compared with weekdays because children may fall asleep quicker due to the shift in later sleep schedule on weekends.Future studies should take into consideration the possible variations between weekdays and weekends, especially when sleep is estimated with proxy reports.
the children that fall asleep before 21:00 and have a sleep midpoint before 02:00.The findings indicate that for some children with early night time sleep timing there is an elevated risk for adiposity.It is possible that either the sleep timing among these children is suboptimal in relation to their sleep need or the combination of timing and quality of food in relation to their sleep timing might increase the risk for adiposity among these children (Ward et al., 2021).Further studies on the interplay between preschool children's nutrition and sleep on health is required.All in all, additional attention to individual variation and needs is important, especially among younger children.Unmeasured daytime sleep may have been an important confounding factor behind the unexpected findings for weekday sleep, particularly tions sleep measures may not be representative of all preschool-aged children in Finland.Compared with national averages, families in the current study population were more educated.Parents in the analytical sample were also more educated than those excluded (37% vs. 30% in the highest education group, respectively), and therefore, the data might not be nationally representative.In conclusion, though sleep estimates from different methods may be comparable, associations may still vary between sleep and health outcomes, such as weight status.Researchers must interpret findings with caution when comparing results across studies with different sleep estimation methods or combined results for weekday and weekend sleep estimates.Most commonly used actigraphic sleep estimation methods are often heuristic, commercialised, and require the input of reported sleep times, which may make sleep research cumbersome and expensive in large study populations.This study highlights the potential of an openly accessible data-driven, unsupervised algorithm used to estimate night time sleep from actigraph data.Further studies are needed to better understand the associations with sleep timing and weight status.

Table 5
Comparison of mean sleep onset, wake-up, and duration from actigraphy-measured and parent-reported sleep estimates with participants (%) divided into quintiles from the DAGIS study (n = 638) Mean differences between actigraphy-measured and parent-reported sleep from the DAGIS study participants (n = 638) a Correlations assessed with continuous variables.*p-Value < 0.0001.T A B L E 4 *p value < 0.0001.T A B L E 5 Linear regression analysis of actigraphy-measured and parent-reported sleep estimates with BMI a and waist-to-height ratio from the DAGIS study participants (n a Age-and-sex-adjusted BMI (ISO-BMI).*p value < 0.05.**pvalue < 0.01.