Less screen time and more physical activity is associated with more stable sleep patterns among Icelandic adolescents

Abstract

Objectives:

Emerging evidence suggests that inconsistent sleep may affect physical and psychological health. Thus, it is important to identify modifiable determinants of sleep variability. Screen time and physical activity are both thought to affect sleep, but studies of their relationship to sleep variability using objective measures are lacking. We examined cross-sectional associations between these variables in mid-teen adolescents using objectively measured sleep and activity.

Methods:

Wrist-worn accelerometers were used to measure one week of sleep and activity in 315 tenth grade students (mean age 15.8y) from six Reykjavik compulsory schools. Participants reported their daily hours of screen time. Regression analysis was used to explore associations of screen time and physical activity with variability in duration, quality, and timing of sleep, adjusting for DXA-measured body fat percentage, parental education, and physical activity or screen time.

Results:

Screen time, especially game playing, was associated with variability in duration, timing, and quality of sleep, most strongly with variation in bedtime. Physical activity was inversely associated with variability in duration, timing, and quality of sleep, most strongly with variation in the number of awakenings. Boys had less stable sleep patterns and higher screen time than girls, and sex-specific associations of screen time with sleep variability parameters were significant for boys only.

Conclusions:

Less screen time and more physical activity were independently associated with less sleep variability among mid-teen adolescents. Our results indicate that encouraging youngsters toward an active lifestyle with limited screen use may be important to achieve more consistent sleep.

Introduction

Adequate sleep is essential for the maturation, growth, daily functioning, and health of children and adolescents¹. Insufficient sleep duration is known to have adverse effects on the health and wellbeing of youngsters.^2–4 However, sleep quality and timing, and consistency of sleep patterns may also affect the restorative properties of sleep and overall health.^4,5 Thus, it is recommended that adolescents not only get an adequate amount of sleep but also maintain a consistent sleep routine.⁶ Two modifiable behavioral factors that are thought to affect sleep are screen time⁷ and physical activity.⁸

In adolescence, a circadian shift toward later hours^9,10 may lead to sleep deprivation, especially on schooldays.^11,12 To compensate, adolescents typically sleep longer on nonschool days, i.e. there is a tendency toward weekend shift in sleep duration.^12,13 Furthermore, research indicates that intra-individual variability in sleep is common amongst adolescents, resulting in unstable sleep patterns across the week.⁴ Although the causes of adolescent sleep variability are likely complex and multifaceted, potential contributors include biological changes due to puberty and brain maturation, increased academic, extracurricular, and social demands, and reduced parental supervision.¹⁴ Participation in screen-based activities⁷ and physical activity⁸ may also affect the sleep health of youngsters. In recent years, screen use has increased rapidly with advances in technology, most notably with the introduction of portable devices, such as tablets and smartphones, and only a small minority of youth limits daily screen time to the recommended maximum of 2 hours.^15–17 At the same time, physical activity levels have been declining,^18–20 and minority of adolescents meets the guidelines for activity.^20–22 Greater adolescent screen time has been linked to shorter sleep duration, delayed bedtime, and poorer sleep quality,^7,23,24 and greater variability in sleep duration and bedtime.^25,26 The relationship between physical activity and sleep in adolescents is more complex, but greater physical activity is generally found to have beneficial effects on sleep.⁹

Although research on the effects of sleep variability on health has been limited, emerging literature indicates that it may be just as important as the mean values of sleep parameters.^4,25 For instance, in youth, inconsistent sleep patterns have been found to correlate with poorer behavioral, cognitive, and psychological functioning, unhealthy diet, increased body weight, and less favorable body composition.⁴ Therefore, it is important to determine the impact of modifiable behaviors, such as screen time and physical activity, on adolescent sleep habits. However, such studies are scarce and have relied on subjective measures of sleep and activity.⁴

The main purpose of this study was to examine the relationship between variations in objectively measured sleep quantity, quality, and timing and measures of screen time and physical activity in Icelandic adolescents.

Methods

Sample and data collection

All four hundred and eleven tenth-grade students (age 15–16 years, 47% boys and 53% girls) in six compulsory schools in metropolitan Reykjavik, Iceland, were invited to participate in this cross-sectional study, 315 (79%) of which agreed to participate. Nonparticipation was mainly due to absence from school during measurement days and lack of interest in the study. Data collection was performed between mid-April and early June of 2015, with all measurements completed within a period of 7–10 days for each individual. Participants provided information regarding their socioeconomic background, health, and lifestyle by answering a tablet-based questionnaire administered at school under the supervision of research team members. Objective measurements of free-living physical activity and sleep, weight, height, and body composition were also performed. Written informed consent was obtained from all participants and their guardians. Strict procedures were followed to ensure confidentiality. The research project was approved by the Icelandic Data Protection Authority, the National Bioethics Committee (VSNb2015020013/13.07), and the Icelandic Radiation Safety Authority.

Measures

Primary measures

Screen time.

Participants were asked to report on how many hours per day on average, separately for weekdays and weekend days, they played computer games, watched TV/DVD/internet material, used the internet for web-browsing/Facebook/e-mail and participated in “other” screen use. Each item was scored on a seven-point Likert scale, with the following response options: 1 = “none,” 2 = “about ½ h,” 3 = “1 up to 2 h,” 4 = “2 up to 3 h,” 5 = “3 up to 4 h,” 6 = “4 to 5 h,” and 7 = “more than 5 h.” Average daily hours for each type of screen-based activity were computed using the midpoint for each scoring category (a value of 5.5 h was used for the highest category) and weighted averages for weekdays and weekend days. All screen-based activities were then summed for total daily screen time (h/day).

Physical activity.

Free-living physical activity was objectively measured using small (3.8 cm x 3.7 cm x 1.8 cm) and lightweight (27 g) triaxial raw signal accelerometer-based Actigraph activity monitors (model GT3X+, Actigraph Inc., Pensacola Florida). Each participant was asked to continuously wear the monitor on his/her nondominant wrist for 7 consecutive days. The monitor recorded raw triaxial data (in milliG’s) sampled at 80 samples per second (Hz). Data were later reduced to the vector magnitude of activity counts over 60 s epoch (counts per min = cpm) using Actilife 6 software (Actigraph Inc., Pensacola, FL, USA; version 6.13.0). Wear-time was detected automatically with customized programs in Matlab (The Mathworks, Natick, MA, USA; version R2013a). Data were considered valid if wear-time was ≥ 14 h from 12 midnight to 12 midnight the following day. A minimum of 3 valid school days and 1 valid nonschool day of activity was set as an inclusion criterion.²⁷

Sleep.

Sleep was measured using Actigraph GT3X+ accelerometers, in the same manner as described above for physical activity. Data were considered valid if wear-time was ≥ 14 h from 12 noon to 12 noon the following day. A minimum of 3 valid school nights (Sunday through Thursday nights) and 1 valid nonschool night (Friday and Saturday night and nights prior to holidays) of sleep was set as an inclusion criterion.¹² Sleep parameters for the duration, quality and timing of sleep (see Table 1) were derived for school days, nonschool days, and all valid days of the week using the well-validated Sadeh automatic sleep detection algorithm for adolescents²⁸, implemented in the Actilife software. Detected sleep periods were verified and manually adjusted when necessary, with the aid of participant-recorded sleep logs of bedtimes and rise times (completed as part of the actigraphy study), using the Actilife software. The individual’s standard deviation of each sleep parameter was used to convey within-subject night-to-night variability. The difference in mean values between nonschool nights and school nights was calculated for each sleep parameter to obtain the within-subject weekend shift.

Table 1.

Sleep parameters for analysis

Sleep parameter	Abbreviation	Definition	Unit
Total rest time	TRT	time between bed-time and rise-time	Hours or Minutes
Total sleep time	TST	sleep time during rest period	Hours or Minutes
Bedtime	BT	time of going to bed	Clock time
Rise time	RT	time of getting out of bed	Clock time
Sleep onset latency	SOL	time it takes to go from wakefulness to sleep	Minutes
Number of awakenings	NOA	number of transitions from sleep to wakefulness, during the rest period	Counts
Sleep efficiency	SLE	(total sleep time divided by total rest time) x 100	%

Covariates

Body composition and parental education were selected as covariates, based on prior studies and our own correlation analysis. More unfavorable body composition has been linked to more screen time²⁹ and lower physical activity levels.³⁰ Parental education is an indicator of socioeconomic status, which may affect youngsters’ participation in recreational activities; higher parental education has been linked to less screen time³¹ and more physical activity among adolescents.³² Body composition and parental education have also been related to sleep parameters. Higher body fat has been associated with late bedtime and inadequate sleep duration² and abdominal obesity with nightly variability in sleep duration among adolescents.³³ Low parental education has been found to correlate with sleep problems in adolescence,³⁴ whereas in another study, adolescents having parents with higher educational level were found to have a shorter sleep duration.³⁵

Body composition.

Body mass index (BMI, kg/m²) was calculated from standing height (m) measured to the nearest mm with a stadiometer (Seca model 217, Seca Ltd., Birmingham, UK) and body weight (kg) measured to the nearest 0.1 kg on a calibrated scale (Seca model 813, Seca Ltd., Birmingham, UK). Measurements were recorded at individual schools, with participants wearing light clothing. Body fat percentage was derived from whole-body dual-energy X-ray absorptiometry (DXA) scans performed on a GE Lunar bone densitometer (GE Healthcare, Madison, Wisconsin USA) by a certified radiologist at the Icelandic Heart Association in Kopavogur, Iceland.

Parental education.

Participants reported the educational levels of their parents by choosing from the following options: 1 = “elementary degree,” 2 = “secondary degree,” 3 = “trade school degree,” 4 = “university degree,” 5 = “other,” 6 = “do not know,” 7 = “do not want to answer.” These options were used to form a new binary variable with the following categories: 1 = “at least one parent with a university degree” and 0 = “neither parent with a university degree.”

Statistical analysis

Customized programs written with Matlab software (The Mathworks, Natick, MA, USA; version R2013a) were used to compile daily and weekly averages of objectively measured physical activity and sleep. Descriptive summaries are presented as means and standard deviations for continuous variables, and as frequencies and percentages for categorical variables, for all days, school days, and nonschool days. Sex differences were evaluated by unpaired t-tests for continuous variables and chi-square tests for categorical variables. Paired t-tests were used to compare measures on school days and nonschool days.

Linear models were used to explore potential associations of average daily screen time and physical activity with mean values of sleep parameters across the entire week. The main regression analysis of the study was then carried out to evaluate the association of average daily screen time and physical activity with nightly variations and weekend shifts in sleep parameters. We started with univariate linear regression analyses and then went on to multivariate linear models to explore a) the effects of screen time, both total screen time and time spent in subtypes of screen use, on nightly variations and weekend shifts in sleep parameters, adjusting for physical activity, body fat percentage, and parental education, and b) the effects of physical activity on nightly variations and weekend shifts in sleep parameters, adjusting for screen time, body fat percentage, and parental education. In additional analyses, models were further adjusted for total sleep time and daylength. Analyses were performed both for the total group and the sexes separately, as we observed a sex difference in behavioral variables (screen time and physical activity, see Table 2) and sleep parameters (see Table 3), and because sex-specific data is lacking from the literature. Significant differences or relations were accepted at α < 0.05. Statistical analyses were performed using SAS statistical software, version 9.4 (SAS Institute Inc., Cary, NC; www.sas.com).

Table 2.

Characteristics of the study participants-anthropometric and socioeconomic measures, screen time, and physical activity

	All (n=247)	Boys (n=103)	Girls (n=144)	p-valuec
Age, years, mean (SD)	15.8 (0.3)	15.8 (0.3)	15.9 (0.3)	0.07
BMI, kg/m2, mean (SD)	21.9 (3.0)	21.5 (2.8)	22.2 (3.2)	0.10
Body fat, %, mean (SD)	25.0 (8.8)	17.6 (6.5)	30.3 (6.0)	< 0.0001
Parents with university degree, N (%)	171 (69.2)	73 (70.9)	98 (68.1)	0.64
Screen timea, h/day, mean (SD)
All days
Total	5.6 (2.3)	6.0 (2.3)	5.3 (2.3)	0.03
Game playing	1.0 (1.3)	1.9 (1.3)	0.3 (0.7)	< 0.0001
Internet: Facebook, browsing, mail	2.2 (1.3)	1.7 (0.9)	2.6 (1.4)	< 0.0001
Watching TV/DVD/Internet	1.7 (0.9)	1.6 (0.9)	1.7 (0.9)	0.18
Other	0.7 (0.8)	0.8 (0.8)	0.7 (0.9)	0.49
School days	5.3 (2.3)	5.6 (2.2)	5.0 (2.4)	0.06
Nonschool days	6.4 (2.7)*	6.9 (2.8)*	6.0 (2.7)*	0.01
Physical activityb, cpm/day, mean (SD)
All days	2037 (471)	1993 (445)	2069 (488)	0.21
School days	2206 (510)	2205 (484)	2207 (529)	0.97
NonSchool days	1773 (540)*	1648 (535)*	1863 (528)*	0.002
Accelerometer compliance
Wear time, h/day, mean (SD)	23.81 (0.36)	23.74 (0.41)	23.86 (0.31)	0.01
Valid days, n, mean (SD)	6.61 (0.62)	6.50 (0.63)	6.69 (0.61)	0.02

BMI = body mass index; SD = standard deviation; cpm = counts per minute; n/N = number.

^aRange (h/day): 0.9-11.9 (all days), 0.5-11.5 (school days), 1.0-14.0 (nonschool days); 10th percentile (h/day, all days): 2.9 (all), 3.5 (boys), 2.5 (girls); 90th percentile (h/day, all days): 9.0 (all), 9.3 (boys), 8.5 (girls).

^bMean of total vector magnitude activity/wear time; Range (cpm/day): 969-3366 (all days), 1099-3677 (school days), 563-3828 (nonschool days); 10th percentile (cpm/day, all days): 1425 (all), 1409 (boys), 1516 (girls); 90th percentile (cpm/day, all days): 2671 (all), 2569 (boys), 2760 (girls).

^cp-values for tests of between sex differences, significant values are bolded.

^*Indicates difference between school days and nonschool days (p < 0.0001).

Table 3.

Sleep parameters mean value and night-to-night variability

	All (n=247)	Boys (n=103)	Girls (n=144)	p-valueb
Sleep parameters
Sleep/rest duration
Total sleep time, h/night, mean (SD)
All days	6.6 (0.7)	6.6 (0.7)	6.6 (0.6)	0.43
School days	6.2 (0.7)	6.2 (0.8)	6.2 (0.7)	0.92
Nonschool days	7.3 (1.1)**	7.2 (1.3)**	7.4 (1.0)**	0.19
Total rest time, h/night, mean (SD)
All days	7.5 (0.7)	7.5 (0.7)	7.6 (0.7)	0.60
School days	7.1 (0.8)	7.1 (0.9)	7.1 (0.8)	0.95
Nonschool days	8.4 (1.3)**	8.3 (1.5)**	8.4 (1.1)**	0.43
Sleep quality
Sleep efficiency, %, mean (SD)
All days	87.8 (4.2)	87.6 (3.9)	87.9 (4.5)	0.58
School days	88.0 (4.5)	88.0 (4.6)	88.0 (4.5)	0.92
Nonschool days	87.5 (4.8)	86.9 (4.2)*	87.9 (5.2)	0.09
Sleep onset latency, min, mean (SD)
All days	1.6 (0.8)	1.6 (0.8)	1.6 (0.7)	0.83
School days	1.6 (1.0)	1.6 (1.0)	1.6 (1.0)	0.91
Nonschool days	1.6 (1.2)	1.6 (1.3)	1.6 (1.2)	0.98
Number of awakenings, N, mean (SD)
All days	20.0 (5.6)	20.0 (5.5)	19.9 (5.7)	0.92
School days	18.5 (5.6)	18.2 (5.7)	18.6 (5.4)	0.56
Nonschool days	22.8 (8.0)**	23.6 (8.3)**	22.3 (7.8)**	0.21
Sleep timing
Bedtime, o’clock, mean (SD (min))
All days	00:46 (53)	00:55 (51)	00:40 (53)	0.047
School days	00:20 (53)	00:24 (55)	00:17 (51)	0.29
Nonschool days	01:40 (75)**	02:00 (84)**	01:26 (64)**	0.0012
Rise time, o’clock, mean (SD (min))
All days	08:21 (43)	08:30 (47)	08:15 (40)	0.008
School days	07:26 (36)	07:31 (41)	07:22 (31)	0.08
Nonschool days	10:10 (82)**	10:31 (85)**	09:55 (76)**	0.0005
Night-to-night variability in sleep parametersa
Total sleep time, min, mean (SD)	75.8 (37.2)	81.8 (44.5)	71.6 (30.5)	0.04
Total rest time, min, mean (SD)	85.6 (42.0)	94.5 (48.6)	79.2 (35.5)	0.008
Sleep efficiency, %, mean (SD)	4.0 (1.6)	4.3 (1.7)	3.8 (1.6)	0.02
Sleep onset latency, min, mean (SD)	1.7 (0.7)	1.6 (0.6)	1.7 (0.7)	0.27
Number of awakenings, N, mean (SD)	6.2 (2.6)	6.8 (2.8)	5.6 (2.3)	0.001
Bedtime, min, mean (SD)	67.7 (39.5)	77.9 (52.0)	60.4 (25.5)	0.002
Rise time, min, mean (SD)	99.3 (40.4)	107.4 (45.0)	93.5 (35.8)	0.01

Difference between school days and nonschool days is marked *p < 0.05 and **p < 0.0001.

SD = standard deviation; h = hours; min = minutes; n/N = number.

^aAcross the entire week.

^bp-values for tests of between sex differences, significant values are bolded.

Results

Characteristics of participants

A total of 247 participants, 103 boys and 144 girls, had complete data according to the inclusion criteria. The characteristics of the participants are shown in Table 2. The average total screen time over the entire week was 5.6 ± 2.3 h/day, around 40 min longer for boys than girls (p = 0.03). Screen time was higher on nonschool days than school days (p < 0.0001); the increase was, on average, 60 min for girls and 78 min for boys, and boys reported almost 1 h higher screen time on nonschool days than girls (p = 0.01). Girls spent most of their screen time on the internet, whereas boys preferred game playing. As we have previously reported,²⁰ boys and girls had similar physical activity over the entire week and on school days, but girls averaged about 13% higher activity than boys on nonschool days (p = 0.002). Participants of both sexes recorded higher activity on school days (+33.8% for boys and +18.5% for girls, both p < 0.0001).

Mean values and night-to-night variability in sleep parameters are summarized in Table 3. Mean values of sleep quality and duration did not differ between the sexes. The average sleep and rest durations across the week (6.6 ± 0.7 h/night and 7.5 ± 0.7 h/night, respectively) were below the recommended minimum of 8 h/night, and the number of awakenings was on average 20/night. Boys had, on average, 15 min later bed and rise times than girls across the week, and this difference was around 35 min on nonschool days (all p < 0.05). Average night-to-night variations in sleep time, rest time, bedtime, and rise time all exceeded 1 h, with rise time having the highest variability around 100 min. Variability in the number of awakenings was about 6/night. Variations in all sleep parameters, except sleep onset latency, were significantly greater for boys than girls (p-values from 0.001 to 0.04). As we have previously reported,¹² there was around 1 h weekend shift toward longer sleep and rest durations, but a higher number of awakenings indicated lower sleep quality on weekend nights (all p < 0.0001). Similarly, bed and rise times shifted to later hours on weekends (both p < 0.0001), more so for boys, who went to bed 96 min later and rose 3 h later on nonschool nights, while girls went to bed 69 min later and rose about 2.5 h later (differences between sexes both p < 0.005). Participants with incomplete data on screen time and physical activity did not differ in sleep parameters from those with complete data.

Screen time and physical activity versus sleep parameters

Screen time was significantly associated with later bedtime (p = 0.03) and less rest (p = 0.02) and sleep (p = 0.047) durations, but not with rise time or sleep quality markers, in linear models adjusted for physical activity, body fat percentage and parental education. With each additional hour of screen time, rest and sleep durations were reduced by 2.8 min and 2.2 min, respectively, and there was a 0.12 min delay in bedtime.

Physical activity was inversely associated with the number of awakenings (p = 0.0002), rise time (p = 0.0003), and sleep and rest durations (both p < 0.0001) in linear models, adjusted for screen time, body fat percentage, and parental education. With each additional 100 cpm/day of physical activity (ca. 5% of the average daily activity), there was a reduction in rest and sleep durations of 3.1 min and 2.5 min, respectively, as well as 0.3 fewer awakenings and 0.15 min earlier rise time.

Screen time and night-to-night variation in sleep parameters

The results of the linear regression analyses between screen time and night-to-night variations in sleep parameters are shown in Table 4. Total screen time was significantly related to night-to-night variations in total sleep time, total rest time, bedtime, and rise time (all p < 0.005). Significance persisted even after adjusting for body fat percentage, parental education, and physical activity (p-values from 0.0001 to 0.03). Standardized betas (see Additional File 1: Table S1) demonstrated that screen time was most strongly associated with variations in bedtime (0.254), followed by variability in total sleep (0.216) and rest times (0.215). Further adjustment for total sleep time and daylength did not meaningfully change the results of these analyses or any other regression analyses in the study. When regressions were performed separately by sex, the associations only remained significant for boys. With each additional hour in screen time, the nightly variation in rest and sleep duration of boys increased by 8.2 min and 7.4 min, respectively, and their nightly variation in bed and rise times increased by 7.8 min and 5.6 min, respectively. Boys in the 90th percentile of screen time, as compared to boys in the 10th percentile, had 43 min higher nightly variability in sleep duration, and their variability in bed and rise times was 46 min and 32 min higher, respectively. We did not observe significant associations between total screen time and variations in sleep quality parameters.

Table 4.

Linear relationships between screen time and night-to-night variations in sleep parameters^a

	All (N=247)			Boys (N=103)			Girls (N=144)
	β	SE	pb	β	SE	pb	β	SE	pb
Sleep duration
SD_Total sleep time (min)
Unadjusted	3.813	1.010	<0.001	6.300	1.865	0.001	1.683	1.109	0.13
Adjustedc	3.566	1.074	0.001	7.416	2.041	<0.001	0.795	1.207	0.51
SD_Total rest time (min)
Unadjusted	4.425	1.138	<0.001	7.395	2.019	<0.001	1.719	1.294	0.19
Adjustedc	4.012	1.208	0.001	8.203	2.216	<0.001	0.522	1.395	0.71
Sleep quality
SD_Sleep efficiency (%)
Unadjusted	−9.1x10−3	0.045	0.84	−0.024	0.074	0.75	−0.024	0.057	0.68
Adjustedc	0.012	0.048	0.79	−0.018	0.081	0.82	0.003	0.061	0.97
SD_Sleep onset latency (min)
Unadjusted	0.009	0.019	0.65	0.030	0.027	0.27	−6.6x10−4	0.027	0.98
Adjustedc	0.008	0.021	0.72	0.045	0.030	0.13	−0.010	0.030	0.74
SD_Number of awakenings (n)
Unadjusted	0.129	0.071	0.07	0.254	0.120	0.04	−0.010	0.084	0.90
Adjustedc	0.069	0.073	0.34	0.196	0.128	0.13	−0.073	0.088	0.41
Sleep timing
SD_Bedtime (min)
Unadjusted	4.020	1.070	<0.001	6.160	2.210	0.007	1.760	0.913	0.06
Adjustedc	4.469	1.135	<0.001	7.795	2.400	0.002	1.478	1.011	0.15
SD_Rise time (min)
Unadjusted	3.270	1.110	0.003	4.690	1.930	0.02	1.670	1.310	0.20
Adjustedc	2.453	1.149	0.03	5.573	2.086	0.009	0.151	1.403	0.91

SD = standard deviation; β = unstandardized regression coefficient; SE = standard error; min = minutes; n/N = number.

^aScreen time (h/day) and night-to-night variations (as SDs) in sleep parameters for the entire week.

^bp-value < 0.05 (bolded) was considered to show statistically significant relationship between screen time and outcome variables.

^cAdjusted for physical activity, body fat percentage, and parental education.

In analyses for subtypes of screen use (see Additional File 2: Table S2), game playing was associated with variability in all of the sleep parameters studied (p-values from <0.0001 to 0.04), except for that in sleep onset latency. Watching TV/DVD/internet material was associated with variability in rest and sleep durations (both p < 0.005), but spending time on the internet for web-browsing/Facebook/e-mail was associated with less variability in sleep efficiency (p = 0.04). When subtype regressions were stratified by sex, associations remained significant for boys but not girls (although the association for variability in sleep efficiency became nonsignificant for both sexes). In addition, we observed significant associations between internet use and variability in rest and sleep durations for boys (both p = 0.03), but not for girls or the total group.

Physical activity and night-to-night variation in sleep parameters

Results of the linear regression analyses between physical activity and night-to-night variations in sleep parameters are shown in Table 5. Physical activity was inversely related to night-to-night variations in the number of awakenings (p < 0.0001) and rise time (p = 0.002), even after adjusting for body fat percentage, parental education, and screen time (p = 0.0001 and 0.013, respectively). The association was stronger for variability in the number of awakenings than in rise time according to standardized beta values (−0.251 vs. −0.161, see Additional File 1: Table S1). When the analyses were performed for each sex separately, variation in the number of awakenings remained significantly associated with physical activity for both sexes, but variability in rise time was only significant for girls. With each additional 100 cpm/day in physical activity, the variability in the number of awakenings decreased by 0.14 and the variability in rise time decreased by 1.4 min (total group). Participants in the 90th percentile of physical activity, as compared to participants in the 10th percentile, had reduced variability in the number of awakenings of 1.7 and decreased variability in rise time of 17 min. We did not observe significant associations between physical activity and night-to-night variations in total sleep or rest times, sleep efficiency, sleep onset latency, or bedtime.

Table 5.

Linear relationships between physical activity and night-to-night variations in sleep parameters^a

	All (N=247)			Boys (N=103)			Girls (N=144)
	β	SE	pb	B	SE	pb	β	SE	pb
Sleep duration
SD_Total sleep time (min)
Unadjusted	−0.008	0.005	0.13	−0.006	0.010	0.58	−0.008	0.005	0.15
Adjustedc	−0.003	0.005	0.56	0.0038	0.010	0.71	−0.005	0.005	0.31
SD_Total rest time (min)
Unadjusted	−0.012	0.006	0.04	−0.012	0.011	0.27	−0.010	0.006	0.10
Adjustedc	−0.007	0.006	0.24	−0.0015	0.011	0.89	−0.008	0.006	0.22
Sleep quality
SD_Sleep efficiency (%)
Unadjusted	2.2x10−4	2.2x10−4	0.32	2.3x10−4	3.7x10−4	0.54	5.5x10−4	2.6x10−4	0.04
Adjustedc	1.8x10−4	2.2x10−4	0.42	−2.5x10−4	4.0x10−4	0.53	5.1x10−4	2.7x10−4	0.06
SD_Sleep onset latency (min)
Unadjusted	1.6x10−5	9.5x10−5	0.87	1.7x10−4	1.4x10−4	0.22	9.0x10−5	1.3x10−4	0.48
Adjustedc	3.9x10−5	9.8x10−5	0.69	2.8x10−4	1.5x10−4	0.06	−8.5x10−5	1.3x10−4	0.53
SD_Number of awakenings (n)
Unadjusted	−1.4x10−3	3.3x10−4	<0.001	−0.002	5.9x10−4	0.003	−0.001	3.8x10−4	0.006
Adjustedc	−1.4x10−3	3.5x10−4	<0.001	−0.002	6.3x10−4	0.015	−0.001	3.9x10−4	0.006
Sleep timing
SD_Bedtime (min)
Unadjusted	4.3x10−4	0.005	0.94	0.007	0.012	0.55	−0.001	0.004	0.80
Adjustedc	0.006	0.005	0.30	0.019	0.012	0.11	0.001	0.004	0.82
SD_Rise time (min)
Unadjusted	−0.017	0.005	0.002	−0.015	0.010	0.14	−0.016	0.006	0.008
Adjustedc	−0.014	0.005	0.013	−0.010	0.010	0.32	−0.015	0.006	0.016

SD = standard deviation; β = unstandardized regression coefficient; SE = standard error; min = minutes; n/N = number.

^aPhysical activity (counts/min/day) and night-to-night variations (as SDs) in sleep parameters for the entire week.

^bp-value < 0.05 (bolded) was considered to show statistically significant relationship between physical activity and outcome variables.

^cAdjusted for screen time, body fat percentage and parental education.

Screen time and physical activity versus a weekend shift in sleep parameters

Average total screen time was associated with a weekend shift in bed and rise times after adjusting for physical activity, body fat percentage and parental education (Table 6; p = 0.007 and 0.03, respectively), with the shift in bedtime having the stronger association according to standardized beta values (0.174 vs. 0.140). Game playing was the only subtype of screen use significantly associated with a weekend shift in bed and rise times (Additional File 3: Table S3; p = 0.0002 and p = 0.002, respectively). When performed separately by sex, these associations were only significant for boys (all p < 0.05). With each additional hour in screen time, the weekend shift in bed and rise times of boys increased by 8.8 min and 8.6 min, respectively. Both these weekend shifts were around 50 min greater for boys in the 90th versus the 10th percentile of screen time.

Table 6.

Linear relationships between weekend shifts in sleep parameters and screen time/physical activity

	β	SE	pc
Screen time (h/day)a
TST NSchD – SchD (min)	−0.024	0.037	0.51
TRT NSchD – SchD (min)	−1.92	2.60	0.46
BT NSchD – SchD (min)	4.90 (boys: 8.78; girls: 2.16)	1.87 (boys: 3.74; girls: 1.87)	0.007 (boys: 0.02; girls: 0.24)
RT NSchD – SchD (min)	4.90 (boys: 8.64; girls: 1.37)	2.30 (boys: 3.89; girls: 2.88)	0.03 (boys: 0.03; girls: 0.64)
NOA NSchD – SchD (N)	−0.09	0.20	0.64
SLE NSchD – SchD (%)	−3.7x10−4	0.12	1.00
SOL NSchD – SchD (min)	0.07	0.04	0.12
Physical activity (cpm/day)b
TST NSchD – SchD (min)	−0.023 (boys: −7.2x10−4; girls: −2.5 x10−4)	0.011 (boys: 3.5x10−4; girls: 1.9 x10−4)	0.03 (boys: 0.04; girls: 0.18)
TRT NSchD – SchD (min)	−0.027 (boys: −0.045; girls: −0.019)	0.012 (boys: 0.025; girls: 0.013)	0.03 (boys: 0.08; girls: 0.15)
BT NSchD – SchD (min)	6.9x10−3	8.6x10−3	0.42
RT NSchD – SchD (min)	−0.019	0.011	0.09
NOA NSchD – SchD (N)	−2.0x10−3 (boys: −1.1x10−3; girls: −2.1x10−3)	9.4x10−4 (boys: 1.8x10−3; girls: 1.0x10−3)	0.04 (boys: 0.53; girls: 0.05)
SLE NSchD – SchD (%)	1.5x10−4	5.6x10−4	0.79
SOL NSchD – SchD (min)	−6.2x10−5	2.1x10−4	0.77

TST = total sleep time; TRT = total rest time; SLE = sleep efficiency; SOL = Sleep onset latency; NOA = number of awakenings; BT = bedtime; RT = rise time; NSchD = nonschool days; SchD = school days; h = hours; min = minutes; cpm = counts per minute; N = number; β = unstandardized regression coefficient; SE = standard error.

^aAdjusted for physical activity, body fat percentage and parental education.

^bAdjusted for screen time, body fat percentage and parental education.

^cp-value < 0.05 (bolded) was considered to show statistically significant relationship between screen time/physical activity and outcome variables.

Physical activity was negatively associated with a weekend shift in sleep time, rest time, and the number of awakenings (all p < 0.05) after adjusting for screen time, body fat percentage, and parental education, with all relationships showing a similar strength of association (standardized beta = −0.14). When performed separately by sex, physical activity was only significantly associated with a weekend shift in sleep time for boys (p = 0.04) and in the number of awakenings for girls (p = 0.05). With each additional 100 cpm/day in physical activity, there was a reduction in the weekend shift in the number of awakenings by 0.2 and in the weekend shift in rest and sleep duration by 2.7 min and 2.3 min, respectively (total group). Participants in the 90th percentile of physical activity, as compared to participants in the 10th percentile, had around 30 min less weekend shifts in rest and sleep duration, and less shift in the number of awakenings of 2.4.

Discussion

We found that more irregular sleep patterns were associated with greater screen time and less physical activity among mid-teen adolescents, especially among boys. Screen time was most strongly associated with variability in bedtime, and physical activity with variation in the number of awakenings. To the best of our knowledge, this is the first study to examine the relationship between intra-individual variations in sleep parameters and screen time and physical activity with objective measures in adolescents.

Greater screen time, especially game playing, was associated with greater night-to-night variations in bedtime, rise time, rest duration and sleep duration, and greater weekend shifts in bed and rise times (game playing was furthermore associated with higher variability in the number of awakenings and sleep efficiency). Screen time was also associated with later bedtime and less rest and sleep durations. Additionally, we found that the increase in reported screen time on nonschool days was correlated with the shift toward later bedtime on nonschool days (Beta = 0.195; p = 0.002). Collectively, these findings support a growing body of research that suggests that screen time disrupts bedtime routine and sleep duration,^23,24 and also previous findings based on self-report that adolescents with more screen/computer use experience greater nightly variation in sleep duration,²⁵ and greater weekend shift in bedtime.²⁶ Although mechanisms for sleep disruption by screen usage remain uncertain, factors, such as direct displacement of sleep by screen time, screen-light altering peak melatonin release, and arousal from screen usage are all thought to contribute.^23,24,36

Higher levels of physical activity have been associated with better sleep quality in adolescents,^8,26,37 but studies of its association with adolescent sleep duration have yielded mixed results.^26,38,39 In this study, we observed that physical activity was associated with less sleep and rest durations and earlier rise time, possibly due to direct displacement of sleep by physical activity. Prior studies exploring the relationship between physical activity and sleep variability in adolescents and children have also produced inconsistent results.^26,40 However, these studies were limited to self- or parent-report and focused only on weekend shifts in sleep. Our results, obtained using objective measures, suggest that physical activity contributes to both better and more consistent sleep quality since physical activity was inversely associated with the mean value, night-to-night variation, and weekend shift in awakenings. We also found that greater physical activity was associated with less nightly variation in rise time and smaller weekend shifts in sleep and rest durations, indicating that it may also promote more consistent sleep patterns. These results, taken together, suggest that physically active students have more consistent sleep quality and routine across the week and less need for catch-up sleep on weekends than their more sedentary counterparts. These findings are consistent with the idea that some of the physiological effects of physical activity, including enhanced body temperature control and melatonin production, may be favorable to sleep regulation.⁴¹

Our sex-specific linear regression analyses showed that associations between screen time and variability in sleep were significant for boys but not girls. Since boys had higher levels of screen time, and more profound nightly variation in sleep and weekend shifts in sleep timing, stronger associations between screen time and sleep variations for boys were not surprising. Among the subtypes of screen use, game playing was most often associated with variability and weekend shift in sleep parameters. A higher percentage of boys than girls reported this type of screen use in our study, which may further explain the sex differences in our results. Conversely, we found that the relationship between physical activity and night-to-night variation in rise time was significant for girls but not boys. This discrepancy may be due to insufficient power in our sex-specific analyses, which should be interpreted with caution due to our limited sample size.

Although the observed increases in the variability in sleep parameters among boys with each additional hour of screen time were modest (ranging from 5.6 to 8.8 min), the cumulative effect of screen use on the stability of sleep parameters may have public health relevance. The same applies to decreases in the variability of sleep parameters with increasing physical activity. Boys in the 90th percentile of screen time, as compared to boys in the 10th percentile, had considerably higher nightly variability in sleep duration, bed and rise times, as well as a substantially greater weekend shift in bed and rise times. This higher variability (28–59% of the average variability/shift in the corresponding sleep parameters among boys) may negatively affect the health, functionality, and well-being of individuals with the highest screen time.⁴ Participants (total group) in the 90th percentile of physical activity, as compared to participants in the 10th percentile, had considerably less variability and weekend shift in the number of awakenings, decreased variability in rise time, and less weekend shift in rest and sleep durations. Again, these are substantial differences (17–56% of the average variability/shift in the corresponding sleep parameters), supporting the notion that by promoting stable sleep patterns, being physically active may have definite and positive effects on health and well-being.

The results of the current study contribute to the limited data on the relationship of screen time and physical activity with the regularity of sleep among adolescents. A strength of our study was the use of objective rather than self-reported measures for sleep, physical activity, and body composition. The participation rate (79%) was also quite high, and the study sample represents a relatively large portion of the total number (4,254 individuals) of 15-year-old Icelandic adolescents,⁴² and the results are, therefore, likely to be representative for the Icelandic youth population. The generalizability to other adolescent populations is less clear as objective measurements of sleep patterns are scarce. The northern latitude and homogeneous racial and ethnic composition of Iceland may limit the generalizability of the findings.

The cross-sectional study design does not allow us to determine causal relationships between the study variables. Reverse causality cannot be ruled out; participants with irregular sleep patterns may spend more time in screen-based activities and less time in physical activity than their peers. Longitudinal and laboratory-based studies are needed to further clarify the causality of screen time and physical activity with sleep patterns. Another limitation of the present study is the use of self-report for screen time, which may be subject to recall and reporting biases. Employing time-use diaries and/or using discrete time periods to report on may help in minimizing such biases, but it also increases subject burden, which may reduce compliance.⁴³ Our questionnaire included separate questions for time spent on individual screen-based activities, which were combined for the total screen time used in our analyses. This may have resulted in an overestimation of the total screen time, as multitasking on different screens, such as watching TV and using a smart-phone at the same time, is quite prevalent in youth.¹⁵ We adjusted for several potential confounders in our analyses, but residual confounding may be present. Finally, although we have no evidence to suggest that the non-participants (N = 96) differed from the general student population in terms of socioeconomic status, lifestyle habits (including physical activity and screen time) and/or sleep habits, we cannot rule out the possibility of selection bias.

Conclusions

We found that less screen time and more physical activity were both associated with less variable sleep patterns among mid-teen Icelandic adolescents, especially among boys. The cumulative effect of screen time and physical activity was substantial and may impact the health, functionality, and well-being of participants. Our results suggest that encouraging youngsters to lead an active lifestyle and limit their screen use may be important to achieve recommendations for a more consistent sleep schedule.