Actigraphic sleep variability is associated with lower positive mood in adolescents

Abstract

Purpose:

Poor sleep health is associated with lower positive mood in adolescents, and more variable sleep is associated with more negative mood. There is a lack of research on the associations between sleep variability and positive mood in adolescents. We investigated whether several types of sleep variability, measured with actigraphy, were associated with positive mood reported on a daily diary in adolescents.

Methods:

Data were collected from a sub-study of the Year 15 wave of the Future of Families and Child Wellbeing Study (n=580; 53% female, mean age±SD=15.4±.5 years, range 14.7–17.6). Adolescents wore an actigraphy device (M±SD=5.6±1.4 nights per adolescent, range: 3–10) and completed daily diaries (M±SD=5.5±1.4 days per adolescent, range: 3–9) for ~1 week, where they rated their levels of happiness and excitement during that day from 0 (not at all) to 4 (extremely). Happiness and excitement were averaged into "positive mood." Separate linear regression models assessed whether actigraphy-measured variability of sleep duration, onset, and offset (residual individual standard deviation, riSD), sleep regularity index (SRI), social jetlag, and free night catch-up sleep were associated with average positive mood per person. Analyses adjusted for age, birth sex, race/ethnicity, household income, and the primary caregiver's education level.

Results:

Greater variability in sleep duration (p=.011, β=−.11) and lower SRI (p=.034, β=.09) were associated with lower ratings of positive mood. There were no other significant associations (p>.10).

Conclusions:

Variable and irregular sleep are associated with lower levels of positive mood in adolescence, which may increase the risk of poor emotional health in adulthood.

Positive mood is an essential component of mental wellbeing that improves quality of life [1]. Positive mood refers to a pleasant state of being and includes aspects such as satisfaction, happiness, and excitement [2]. Nearly 10% of 15-year-olds in the United States report not being happy sometimes or always [3], and the United States ranks 32^nd out of the 36 richest countries in level of mental wellbeing among adolescents aged 15 [4]. Positive wellbeing in adolescence (including aspects such as happiness and enjoyment of life) predicts fewer health risk behaviors (including low physical activity, consuming fast food, smoking, binge drinking, and using illicit substances) and better general subjective health in adulthood [5]. It is therefore important to examine factors, particularly modifiable ones, that are associated with positive mood in adolescence.

Compared to both children and adults, adolescents may experience more variable sleep [6], which has been linked to poor emotional health. Adolescents experience a shift toward later preferred sleep timing (i.e., chronotype) [6] and short sleep during the school week due to early obligations such as school [7] that promote misalignment of sleep timing across the week, increasing social jetlag and free night catch-up sleep [6]. Social jetlag is the difference in sleep timing between non-school nights and school nights, and free night catch-up sleep is the difference in sleep duration. Studies have demonstrated that self-reported social jetlag is associated with anxious symptoms [8] and irritable mood [9], and self-reported free night catch-up sleep is associated with more depressive symptoms [10,11] and presence of mood and anxiety disorder [12] in adolescents. Greater variability in self-reported sleep duration (measured with standard deviation, SD) is associated with more internalizing emotions (e.g., anxious symptoms) [13], and more variable actigraphic time in bed is associated with more negative mood (depressive and anxious symptoms) [14] in adolescents. These studies demonstrated that increased variability in sleep (largely measured with self-report) is associated with more negative mood in adolescents.

Few studies have evaluated the association between sleep variability and positive mood in adolescents, especially measuring sleep with actigraphy. Some studies have demonstrated a link between sleep variability and negative mood in adolescents [8–14], but positive mood is independent from negative mood, and the two distinct constructs are not merely opposites of one another [2]. One study demonstrated that self-reported free night catch-up sleep was associated with lower life satisfaction, an aspect of positive wellbeing, in adolescents [15]. Given the lack of research on the association between sleep variability and positive mood in adolescents, particularly using objective measures of sleep, the present study examined whether several types of sleep variability measured with actigraphy were associated with self-reported positive mood ratings in adolescents. We hypothesized that greater sleep variability would be associated with lower ratings of positive mood.

Methods

Participants

Data for the current analyses come from the Future of Families (named Fragile Families at the time of data collection) and Child Wellbeing Study (FFCWS; www.ffcws.princeton.edu), a longitudinal birth cohort oversampled for nonmarital births, which resulted in a greater proportion of racial/ethnic minority mothers and those of lower socioeconomic status and education level compared to the national population. More details regarding the sample and design may be found elsewhere [16]. This study was conducted according to the guidelines established in the Declaration of Helsinki of 1975 (revised 1983), and all procedures involving human participants were approved by the Princeton University and Stony Brook University (CORIHS B) (FWA #00000125) Institutional Review Boards. Written (for in-home interviews) or recorded verbal (for phone interviews) informed consent was obtained from primary caregivers, and assent was obtained from adolescents.

The research survey firm Westat^® communicated with the participants, obtained informed consent, distributed the actigraphy devices, and compensated the participants. Westat paid participants through pre-paid gift cards sent through postal mail. Parents received $100 USD for completion of the one-time parent survey. Adolescents received $50 USD for completion of the one-time youth survey, $50 for wearing the actigraphy device, and $20 for diary completion.

The original FFCWS birth cohort consists of 4,898 children born from 1998–2000 in 20 large U.S. cities [17]. Families were recruited from local hospitals at the time of the child’s birth. Study staff maintained records about the participants and their families for follow up at subsequent waves, when participants were approximately ages 1, 3, 5, 9, and 15 years of age. Families were eligible for inclusion in the Year 15 follow-up wave if the child was alive, not legally adopted, and participated in the year 9 wave. Data in the current analyses were collected from February 2014 to March 2016. Therefore, the results are based on pre-COVID-19 school schedules and may not be applicable to samples collected during the pandemic. During the Year 15 wave of the FFCWS (wave 6), 3,444 adolescents and their primary caregivers completed separate surveys querying household and demographic characteristics, administered either over the phone or in person at the participant’s place of residence. A randomly selected subsample (N = 1,090) were asked to participate in a FFCWS sub-study [18]. Adolescents who agreed to participate (N = 1,049) were asked to wear a wrist-worn accelerometer and answer a daily diary for seven consecutive days in the evening. Out of 1,049 assenting adolescents, n = 414 were excluded due to providing fewer than 3 valid nights of actigraphy recordings (see the "Wrist actigraphy" section; current sample M ± SD = 5.6 ± 1.4 nights per adolescent; range 3–10; interquartile range, IQR 5–7), and n = 55 were excluded due to providing fewer than 3 reports of their levels of happiness and excitement (current sample 5.5 ± 1.4 reports of each mood per adolescent; range 3–9; IQR 4–7), leaving a sample of N = 580 adolescents (55.3% of the subsample). Adolescents could provide any three actigraphy nights and mood reports; the actigraphy recording did not need to immediately precede each mood report for that adolescent. An additional n = 216 adolescents were excluded from social jetlag and free night catch-up sleep analyses due to not providing data from at least one school night and one free night, resulting in n = 364 adolescents included in the social jetlag and free night catch-up sleep analyses. Sex and income were not associated with missingness from the n = 364, but Hispanic and/or Latino adolescents had significantly lower odds of being missing from the 364 compared to White (OR = .45, p = .018) and Black (OR = .53, p = .030) adolescents. See Figure S1 for a participant flow chart.

Materials and measures

Wrist actigraphy

Sleep measures were collected with a wrist-worn accelerometer with off-wrist detection (Actiwatch Spectrum; Philips-Respironics, Murrysville, PA) and study participants were asked to wear the watch on their non-dominant wrist for one week. Accelerometer devices measure movements, from which patterns of sleep and wake may be estimated [19]. Data from the Spectrum device were downloaded with Philips Actiware software (Version 6.0.4, Philips Respironics, 2017). At least two independent scorers (blinded to each other) determined cut-point (i.e., start and end time that define a 24-hour day), validity of days, and sleep intervals with a duration of 30 minutes or more using a validated procedure [20]. A program was used to compare differences between the scorers in the determination of the number of valid days, cut-point, number of sleep intervals, and any > 15-minute differences in duration or wake after sleep onset (WASO) for each recording. The scorers adjudicated recordings, and a third scorer adjudicated any remaining discrepancies. The scorers determined sleep intervals using a decrease in activity levels and the aid of light levels for sleep onset and sleep offset [21], and a nighttime sleep interval was split into two intervals (main sleep and nap) if there was an awakening ≥ 1 hour during this interval. No sleep intervals were set if the duration was less than 30 minutes. A sleep actigraphy day was determined invalid and no sleep interval was set if there were ≥ 4 total hours of off-wrist time, with the exception of the first and last day (device should have been worn at least 2 hours on the first day). Other invalidation criteria were constant false activity due to battery failure, data unable to be retrieved or recovered, or an off-wrist period of ≥ 60 minutes within 10 minutes of the scored beginning or end of the main sleep period for that day. Nights were excluded from current analyses if the adolescent had an all-nighter (i.e., no sleep interval was set within that 24-hour day; n = 2 days).

Sleep variability measures

Variability in sleep duration, onset, and offset across the monitoring period were each represented by the residual individual SD (riSD) per adolescent. The riSD adjusts for any potential systematic changes of the variable over time, providing an advantage over the traditionally used measure of intraindividual variability, the SD [22,23]. The riSD is calculated by fitting a model per person where the predictor is time (in this case, sleep period number for each individual, skipping for invalid days) and the outcome is the variable of interest (in this case, sleep duration, onset, or offset). The residual from this model is the riSD for that person, for that variable.

Other types of sleep variability were the sleep regularity index (SRI)[24], social jetlag [6], and free night catch-up sleep. SRI was calculated based on the formula from Phillips et al. [24] and Lunsford-Avery [25] and ranges from 0 (low regularity) to 100 (high regularity). The score encapsulates within-person sleep patterns and represents the percentage probability that an individual is in the same state (sleeping or awake) at any two time points 24 hours apart, averaged across all days. A score of 0 indicates the individual is sleeping and waking totally at random, whereas a score of 100 indicates the individual is sleeping and waking at exactly the same times each day. Social jetlag, a misalignment of sleep midpoint between school and free days, was calculated in hours through the following formula: | average sleep midpoint on free nights − average sleep midpoint on school nights | [6]. Free night catch-up sleep was calculated as average sleep duration on free nights − average sleep duration on school nights and included negative values. Only adolescents with at least one school night and one free night (of their minimum three total nights of actigraphy) were included in the social jetlag and free night catch-up sleep measures; therefore, some adolescents did not have social jetlag or free night catch-up sleep values (N = 364 of the 580 total adolescents had values).

Daily diary

Participants were asked to complete a diary each evening after 7:00 PM (19:00) and before going to sleep each day. The questions in the diary asked about last night’s sleep through the current day, and adolescents were not to complete the diary after they went to bed that night (i.e., the next day). Sleep and daily diary data were merged by participant identification number and date/time. A “day” was defined as spanning from sleep onset to sleep onset (typically about 24 hours). Therefore, a previous night’s sleep and the following day’s diary answers were aligned on the same observation record. If there was not a main sleep onset at the end of a day due to an invalid sleep period (e.g., device removed), the day ended when it encountered more than 1 hour off-wrist time or an apparent sleep interval scored by a validated algorithm [20], whichever occurred first.

Positive mood

The mood scale was created by the FFCWS team. Adolescents were presented with the prompt, “This list describes feelings or experiences. Mark the selection that best describes how you felt during that day” on the daily diary. They were presented with the items “Happy” and “Excited” and were given the options “Very slightly or not at all” (1), “A little” (2),” “Moderately” (3), “Quite a bit” (4), and “Extremely” (5). The happiness and excitement ratings were both recoded such that 0 became low and 4 became high. The mean of happiness and excitement across days were computed separately. The two variables were highly correlated (r = .78); therefore, the measures were averaged into one composite “positive mood” rating by computing the mean of the cross-day means of happiness and excitement.

School attendance

Adolescents were asked, " Did you go to school?" and chose "Yes" (1) or "No" (0). School attendance was used to calculate social jetlag and free night catch-up sleep.

In-person assessments

Sex information was collected at birth, and age was reported at the in-person year 15 interview.

Year 15 surveys

The year 15 surveys were administered once to adolescents and their primary caregivers in person and queried demographic and household characteristics. Race/ethnicity was reported on the adolescent survey and was grouped into exclusive categories of White/Caucasian (not Hispanic or Latino), Black/African (not Hispanic or Latino), Hispanic and/or Latino (any race), or a category with other (including Asian, Central American/Caribbean, Native American/Alaska Native, and/or Native Hawaiian/Pacific Islander), mixed, and no reported race/ethnicity (all not Hispanic or Latino). Annual household income (in USD) and the primary caregiver's highest education level (did not complete high school, completed high school, completed some college, or college graduate) were reported on the primary caregiver survey. Missing income values were imputed by the FFCWS staff at Princeton University through Stata statistical software regression-based impute command with the following covariates: original sample city, total adults in the household, and primary caregiver age, years of education, race/ethnicity, earnings, immigrant status, employed last year, hours worked, welfare receipt, and marital status.

Statistical analyses

Analyses were conducted in SAS 9.4 software (SAS Institute, Cary, North Carolina). Most variables met standards for normality (skew < |3| and kurtosis < |10|) [26]. The riSDs of sleep onset and offset and social jetlag were positively skewed (skew ≥ 3) and/or leptokurtotic (kurtosis ≥ 10) and were winsorized (i.e., values 3 SDs above or below the mean were replaced with the nearest value that was within 3 SDs of the mean). Of 580 adolescents, 9 sleep onset riSD values, 11 sleep offset riSD values, and 1 social jetlag value were winsorized. After winsorization, these variables also met criteria for normality.

Analyses among types of sleep variability

We used bivariate Pearson correlations to test between-person associations among the types of sleep variability. Correlations split by sex are in Table S1.

Main analyses

Linear regression models were used to test whether types of sleep variability (riSD of sleep duration, onset, and offset, SRI, and social jetlag) predicted average positive mood rating. Sex, race/ethnicity, adolescent age, household income, and primary caregiver education level were included as covariates in all analyses. Categorical covariates were dummy coded before being entered into analyses (sex, race/ethnicity, and primary caregiver education level). Alpha < .05 (two-sided) was deemed statistically significant.

Sensitivity and supplementary analyses

We conducted four sets of sensitivity analyses: (1) Using traditional SD as a predictor of positive mood (instead of riSD); (2) adjusting for average sleep duration; (3) using only the sample where social jetlag and free night catch-up sleep values were available (n = 364), and (4) examining happiness and excitement ratings as separate outcomes (Table S2). In supplementary analyses, we examined associations between the types of sleep variability with ratings of "lonely" and "angry," aspects of negative affect (see Table S3 for results). We additionally explored interactions between the types of sleep variability and sex, race/ethnicity, and household income on positive mood (sex results are in Table S4). Results of all sensitivity and supplementary analyses are in Supplementary Material.

A separate set of unrelated analyses examined within- and between-person associations of sleep duration and maintenance efficiency with self-reported happiness, anger, and loneliness ratings on the daily level (Master et al., under review).

Results

Participant characteristics

In all analyses except where social jetlag or free night catch-up sleep predicted mood, 580 adolescents provided at least 3 nights of actigraphy, reports of mood, and complete covariate data (53% female, n = 307; mean age ± SD = 15.4 ± .5 years, range 14.7–17.6), with an average of 5.6 ± 1.4 actigraphy nights (range 3–10 days; IQR 5–7) and 5.5 ± 1.4 diary mood reports (range 3–9; IQR 4–7) per adolescent (n = 364 for social jetlag and free night catch-up sleep analyses). Ethnoracial composition of the full sample was as follows: 41% Black/African American (n = 240), 25% Hispanic or Latino (n = 144), 19% White/Caucasian (n = 111), and 15% other, mixed, or none (n = 85). The mean positive mood rating was 1.95 ± .92 (range 0–4). Other sample information (n = 580), including descriptive statistics for sleep variables and covariates, is in Tables 1 and 2. See Table 3 for correlations among types of sleep variability (n = 580).

Table 1.

Average demographic and household statistics for analytical sample (N = 580).

Variable	Mean or %	(SD or n)
Demographic and household characteristics
Age	15.39	(.51)
Sex (female)a	53%	(307
Race/ethnicity
Black/African American	41%	(240)
Hispanic and/or Latino	25%	(144)
White/Caucasian	19%	(111)
Other,b mixed, or none	15%	(85)
Household incomec	$66,103	($61,903)
Primary caregiver's education level Did not graduate high school	14%	(79)
Completed high school	18%	(104)
Completed some college	47%	(271)
Graduated from college	22%	(126)

^aData collected at birth.

^bAsian, Central American/Caribbean, Native American/Alaska Native, and/or Native Hawaiian/Pacific Islander.

^cMissing values imputed by the Fragile Families and Child Wellbeing Study staff at Princeton University through Stata statistical software regression-based impute command with covariates original sample city, total adults in the household, and primary caregiver age, years of education, race/ethnicity, earnings, immigrant status, employed last year, hours worked, welfare receipt, and marital status.

Table 2.

Average sleep and mood descriptive statistics for analytical sample (N = 580).

Variable	Mean	(SD)
Sleep duration, timing, and quality
Sleep duration (hrs)	7.81	(1.07)
Sleep onset (midnight-centered hrs)	0:25	(1:43)
Sleep offset (midnight-centered hrs)	8:17	(1:45)
Sleep maintenance efficiency (%)	90.76	(3.38)
Sleep variability
Sleep duration riSD (hrs)a	1.32	(.77)
Sleep onset riSD (hrs)a	1.04	(.65)
Sleep offset riSD (hrs)a	1.25	(.84)
SRIb	48.75	(13.23)
Social jetlag (hrs)c	1.80	(1.16)
Free night catch-up sleep (hrs)d	1.75	(1.36)
Emotional health
Happiness ratinge	2.21	(.88)
Excitement ratinge	1.69	(1.06)
Positive mood ratingf	1.95	(.92)

^aResidual standard deviation; higher values denote more variability.

^bCalculated based on formula from Phillips et al. [24]; ranges from 0 (low) - 100 (high).

^cCalculated as |average midpoint on non-school nights − average sleep midpoint on school nights| [6]. N = 364.

^dCalculated as average sleep duration on non-school nights − average sleep duration on school nights. N = 364.

^eRanges from 0 (very slightly or not at all) - 4 (extremely).

^fCalculated as average of happiness and excitement ratings; ranges from 0 (very slightly or not at all) - 4 (extremely).

Table 3.

Between-person correlations among types of sleep variability (N = 364–580).

Variable	1.	2.	3.	4.	5.	6.
1. Sleep duration (riSDa, hrs)	---	.49***	.63***	−.43***	.17**	.58***
2. Sleep onset (riSDa, hrs)	---	---	.37***	−.41***	.38***	−.04
3. Sleep offset (riSDa, hrs)	---	---	---	−.37***	.51***	.36***
4. SRIb	---	---	---	---	−.22***	−.17**
5. Social jetlag (hrs)c	---	---	---	---	---	.19***
6. Free night catch-up sleep (hrs)d	---	---	---	---	---	---

Notes. Values are bivariate Pearson correlation coefficients (r) with sleep variability measures calculated per adolescent. Sleep timing measures (onset, midpoint, and offset) were centered around midnight (0:00) before riSD was calculated. The mean number of valid actigraphy nights per youth was 5.6 ± 1.4 (range: 3–10 nights).

^aResidual standard deviation; higher values denote more variability.

^bCalculated based on formula from Phillips et al. [24]; ranges from 0 (low) - 100 (high).

^cCalculated as |average midpoint on non-school nights − average sleep midpoint on school nights| [6]. N = 364.

^dCalculated as average sleep duration on non-school nights − average sleep duration on school nights. N = 364.

hrs, hours; riSD, residual standard deviation; SRI, sleep regularity index.

^**p < .01,

^***p < .001, two-tailed.

Associations of sleep variability with positive mood

Greater variability (riSD) in sleep duration was negatively associated with average positive mood rating (p = .011, β = −.11), such that for every additional riSD-hour in variability, positive mood rating decreased by .13 units (see Table 4 and Figure 1). SRI was positively associated with average positive mood rating (p = .034, β =.09), such that for every additional unit increase in SRI, positive mood rating increased by .01 units. Sleep onset riSD (p = .204), sleep offset riSD (p = .146), social jetlag (p = .321), and free night catch-up sleep (p = .235) were not associated with positive mood rating.

Table 4.

Sleep variability predicting adolescent positive mood ratings (N = 364–580).

Predictor	n	b	95% CI b	β	95% CI β	P
Unadjusted analyses
Sleep duration riSD (hrs)a	580	−.14	[−.24 −.05]	−.12	[−.20 −.04]	.004**
Sleep onset riSD (hrs)a	580	−.09	[−.21 .02]	−.07	[−.15 .01]	.105
Sleep offset riSD (hrs)a	580	−.09	[−.17 <.01]	−.08	[−.16 <.01]	.059 †
SRIb	580	.01	[<.01 .01]	.12	[ .04 .20]	.003 **
Social jetlag (hrs)c	364	−.05	[−.13 .03]	−.06	[−.17 .04]	.247
Free night catch-up sleep (hrs)d	364	−.04	[−.11 .03]	−.06	[−.16 .05]	.283
Adjusted analyses
Sleep duration riSD (hrs)a	580	−.13	[−.22, −.11]	−.11	[−.19, −.02]	.011 *
Sleep onset riSD (hrs)a	580	−.07	[−.19, −.05]	−.05	[−.14, .03]	.204
Sleep offset riSD (hrs)a	580	−.07	[−.16, −.06]	−.06	[−.14, .02]	.146
SRIb	580	.01	[<.01, .09]	.09	[ .01, .17]	.034 *
Social jetlag (hrs)c	364	−.04	[−.13, −.05]	−.05	[−.16, .05]	.321
Free night catch-up sleep(hrs)a	364	−.04	[−.12, −.06]	−.06	[−.17, .04]	.235

Note. Each row represents a separate linear regression model. Analyses in bottom portion adjust for age, birth sex, race/ethnicity, household income, and the primary caregiver's education level. Sleep variability was measured from nightly actigraphy. Higher values mean more variability, except SRI. Mood ratings were collected from daily diary. Positive mood was calculated as the average of happiness and excitement mood ratings, which each range from 0 (not at all) to 4 (extremely). The mean number of valid actigraphy nights per youth was 5.6 ± 1.4 (range 3–10) and the mean number of happiness and excitement daily reports was 5.5 ± 1.4 (range 3–9).

^aResidual standard deviation; higher values denote more variability.

^bCalculated based on formula from Phillips et al. [24]; ranges from 0 (low) - 100 (high).

^cCalculated as |average midpoint on non-school nights − average sleep midpoint on school nights| [6]. N = 364.

^dCalculated as average sleep duration on non-school nights − average sleep duration on school nights. N = 364.

b, unstandardized beta; β, standardized beta; CI, confidence interval; n, number; p, significance level; riSD, residual individual standard deviation; SRI, sleep regularity index.

^†p < .10,

^*p < .05,

^**p < .01, two-tailed.

Bold italic, p < .10; bold, p < .05.

Figure 1.

Associations of actigraphic sleep duration riSD (A) and SRI (B) with average positive mood reported on daily diaries in two separate linear regression analyses (predicted values displayed). For sleep duration riSD, higher values mean more variability; for SRI, higher values mean less variability. Positive mood was calculated as the average of happiness and excitement mood ratings, which each range from 0 (not at all) to 4 (extremely). The mean number of valid actigraphy nights per youth was 5.6 ± 1.4 (range 3–10) and the mean number of happiness and excitement daily reports was 5.5 ± 1.4 (range 3–9). Both models adjust for age, birth sex, race/ethnicity, household income, and the primary caregiver's education level. Shaded bands depict 95% confidence interval of the mean. b, unstandardized beta; β, standardized beta; p, significance level; riSD, residual individual standard deviation; SRI, sleep regularity index [24].

Discussion

The present study examined whether several types of sleep variability (riSD of sleep duration, onset, and offset, SRI, social jetlag, and free night catch-up sleep) measured with actigraphy were associated with positive mood ratings in adolescents. We demonstrated that variability (riSD) in sleep duration was negatively associated with positive mood rating, whereas SRI was positively associated. The findings show that more variable sleep duration and irregular sleep across the week are indicators of lower positive mood in adolescents. Strategies should be adopted to help adolescents stabilize sleep schedules across the week, which may potentially increase positive mood.

We found that more variability in sleep duration was associated with lower positive mood. A previous study demonstrated that variability (SD) in sleep duration was associated with more internalizing symptoms (a negative emotional trait) in adolescents [13]. The current study is among the first to examine the link between sleep variability and positive mood in adolescents. There are several potential explanations for the link between sleep duration variability and lower positive mood. Greater variability in sleep may lead to minor circadian misalignment [27], which may increase emotional dysregulation and reduce positive mood. Alternatively, adolescents with more positive mood may be better adjusted and more likely to maintain consistent sleep/wake schedules. Another possibility is that the relationship between sleep variability and mental health in adolescents is bidirectional, as between sleep duration and mental health [28,29]. There may also be other factors that contribute to both sleep variability and mental health not measured in the current study. For example, psychological stress [30] and parental sleep variability [31] may increase the risk for both increased adolescent sleep variability and lower positive mood. Future experimental and longitudinal research into the link between sleep variability and positive mood would elucidate the direction of the association.

In contrast to variability in sleep duration, variability in sleep timing was not associated with positive mood. The effects of variability in sleep onset and sleep offset may have been too small to detect statistical significance; each effect was about half the size of the effect for variability in sleep duration. Stabilizing both sleep onset and sleep offset would reduce variability in sleep duration, which the current study suggests may allow for more positive mood in adolescents.

SRI was associated with greater positive mood in adolescents. The current study is the among the first to examine the link between SRI and emotional health in adolescents. In contrast to the riSD measures and social jetlag, SRI is not specific to duration or sleep timing. Rather, SRI represents the general consistency of sleep-wake cycles across days, epoch-by-epoch, and in addition to duration and timing, takes into account aspects of sleep such as napping and WASO [24] (a measure of sleep discontinuity). Higher SRI could reflect adolescents who have consistent sleep/wake schedules, take fewer naps, and have consistently high sleep efficiency. Increasing SRI could be accomplished by minimizing naps or maintaining a consistent nap time each day, maintaining a consistent sleep/wake schedule across the week, and increasing sleep efficiency.

Neither social jetlag nor free night catch-up sleep were associated with positive mood in the current study. Previous studies have demonstrated self-reported social jetlag was associated with anxious symptoms [8] and irritable mood [9], and self-reported free night catch-up sleep was associated with more depressive symptoms [10,11] and existence of mood or anxiety disorder [12]. However, there are a lack of studies that examine the link between social jetlag or free night catch-up sleep with positive emotional wellbeing in adolescents. One study demonstrated that free night catch-up sleep, calculated by querying the adolescen's typical sleep habits, was associated with lower life satisfaction in adolescents [15]. It is possible that social debt and social jetlag as calculated from at least one school night and one free night may not represent the individual's typical values. Future research may examine whether social jetlag and free night catch-up sleep as measured with two or more weeks of actigraphy, which may capture better averages, are associated with positive mood in adolescents.

The findings of this study inform recommendations for pediatricians, parents, and school administrators about the potential impact of sleep variability on positive mood. Pediatricians could recommend that parents set consistent bedtimes and wake times for their adolescent children, which may reduce sleep variability and may boost positive mood. School administrators may consider later high school start times, which may reduce sleep duration variability due to more consistent sleep timing across the week [6]. Stabilizing sleep schedules in adolescents may be an important tool to maintain optimal wellbeing.

The current study has some limitations and notable strengths. We did not use a validated scale such as the Positive Affect and Negative Affect Schedule-Child Form (PANAS-C) [32] that captures several aspects of positive emotional wellbeing to assess positive mood in the current study, including low-arousal emotions such as "calm." Future studies may assess the association between sleep variability and both low- and high-arousal states of positive emotional wellbeing. We were unable to determine temporal precedence in this cross-sectional study, and future research should consider employing longitudinal or experimental designs to determine the direction of the association between actigraphic sleep variability and positive mood. A strength of the current study is the objective measurement of multiple types of sleep variability with actigraphy, which provide novel contributions to the literature. Another strength is the large, diverse sample of adolescents across the United States.

The present study demonstrated that adolescents with more variable sleep duration and lower SRI as measured with actigraphy reported lower ratings of positive mood. It is possible that maintaining consistent sleep schedules may help boost positive mood, and increasing positive mood may assist in the maintenance of consistent sleep schedules in adolescents.

Implications and Contributions.

The current study found that greater sleep variability measured with wrist actigraphy predicted lower positive mood in a diverse sample of 580 adolescents. The findings may help in the development of adolescent mental health interventions that focus on obtaining consistently sufficient sleep duration across the week.

List of abbreviations:

FFCWS

Fragile Families and Child Wellbeing Study

IQR

interquartile range

riSD

residual standard deviation

SD

standard deviation

SRI

sleep regularity index

WASO

wake after sleep onset