A school-based health and mindfulness curriculum improves children’s objectively measured sleep: a prospective observational cohort study

Abstract

Study Objectives:

Poor sleep impedes children’s cognitive, emotional, and psychosocial development. Pediatric sleep dysregulation is common, and children who live in communities of low socioeconomic status experience additional risk factors for short sleep duration and poor sleep quality. School-based training in mindfulness and yoga-informed practices can improve children’s behavior and well-being, but effects on objectively measured sleep are unknown.

Methods:

Effects of a school-based health and mindfulness curriculum, which taught practices such as paced breathing, on sleep and stress were examined in 115 children (49 girls, ages 8 to 11 at baseline). Fifty-eight children in a community of low socioeconomic status received the curriculum twice weekly for 2 years. Fifty-seven children in a socioeconomic status–matched community engaged in their usual physical education class instead. In-home ambulatory polysomnography and perceived social stress were measured in all children at 3 time points: at baseline (ie, prior to curriculum exposure) and at 2 yearly follow-ups.

Results:

Children receiving the curriculum gained an average of 74 minutes of total sleep time, and 24 minutes of rapid eye movement sleep, per night over the 2-year study period. Children not receiving the curriculum experienced a decrease in total sleep time averaging 64 minutes per night, with no changes in rapid eye movement sleep. Sleep improved within the first 3 months of curriculum exposure, in a dose-dependent fashion. Higher curriculum engagement (eg, using the breathing exercises outside of class) was associated with larger gains in total and rapid eye movement sleep duration. Aggregate within-group changes in social stress were not significant. However, among children receiving the curriculum, those who experienced larger gains in total and rapid eye movement sleep duration also experienced larger increases in perceived social stress.

Conclusions:

A school-based health and mindfulness curriculum improved children’s objectively measured sleep over 2 years. Social stress did not mediate these effects; instead, mindfulness training may have increased awareness of environmental stressors, while developing tools to reduce stress vulnerability.

Citation:

Chick CF, Singh A, Anker LA, et al. A school-based health and mindfulness curriculum improves children’s objectively measured sleep: a prospective observational cohort study. J Clin Sleep Med. 2022;18(9):2261–2271.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Secular trends of decreasing sleep times among children are problematic given the crucial role of sleep in physical, cognitive, and emotional development. School-based training in mindfulness and yoga-informed practices can improve behavior and well-being, but effects on objectively measured sleep are unknown.

Study Impact: Children who received a school-based health and mindfulness curriculum, which taught skills such as paced breathing, experienced increased total and rapid eye movement sleep duration over the 2-year study period. In contrast, children from a neighboring school district, matched for sociodemographic variables, experienced decreased total sleep time, with no changes in rapid eye movement sleep. Skills taught in the curriculum may have supported improved sleep quality and quantity during a critical developmental period characterized by increasingly short sleep.

INTRODUCTION

Sleep is integral to physiological, psychological, and emotional development.^1–3 Yet a recent study reported that only 15.6% of middle school students meet the optimal 9 hours of sleep recommended by the National Sleep Foundation,⁴ and as many as 46.1% of students slept for less than 7 hours per night.⁵ Pediatric sleep is frequently dysregulated,^6,7 especially among children living in communities of low socioeconomic status (SES).⁸ Children in low-SES communities experience shorter sleep duration and poorer quality sleep, as well as more daytime sleepiness, compared with children of higher-SES communities.^9,10 In turn, poor sleep may exacerbate other negative outcomes associated with low SES.^11,12 Decreased sleep times in children have been associated with obesity,¹³ poor academic performance,^14,15 behavioral problems,¹⁶ and risky behaviors.⁵ These effects are compounded for children of immigrants and for Latino and other ethnic minorities.⁸ Thus, among the highly interrelated factors that contribute to poor health in low-SES communities, sleep appears to represent a nexus affecting physical, mental, and emotional health.

SES-associated stress has been proposed as a mediator of the impact of low SES on dysregulated sleep.¹⁷ Previous studies have emphasized the role of a stressful environment in the disturbed sleep of low-SES children, including social and family stress, food and housing insecurity, and crowded or noisy residential environments.^10,18,19 Indeed, SES-associated chronic stress in childhood has been linked to poor long-term outcomes, including exaggerated physiological responses to acute stress in adulthood.²⁰ Thus, the development and assessment of early interventions that teach children methods for coping with chronic stress may be particularly important for mitigating effects of SES-associated stress on trajectories of physical and mental health. The relationship between stress and dysregulated sleep in school-age children is such that multiple studies have examined the impact of stress-reducing curricula on overall well-being and self- and parent-reported sleep.²¹ Among these curricula, school-based practices that incorporate yoga-informed principles, including mindfulness, breathing, and movement posture education, have been successfully implemented in over 900 schools across the United States.²² A recent review concluded that children’s cognitive functioning, behavior, and overall well-being improved when their schools incorporate yoga-based education.²¹ Self-reported sleep improvements following a yoga curriculum have been reported in children with neurodevelopmental disorders.^23,24 Additionally, a recent review found that mindfulness meditation may benefit self-reported sleep in adults.²⁵ However, to date, no study has measured the effects of a yoga-derived health or mindfulness curriculum on objectively measured sleep in any age group.

We examined the impact of such a health and mindfulness curriculum, delivered during the school day, on objectively measured sleep in 115 typically developing children from low-SES communities. Both communities have historically high rates of violence and crime, which have been associated with biological markers of stress in children.²⁶ Previous research has linked chronic, social, and posttraumatic stress to decreased total sleep time (TST) and increased wakefulness, with particularly marked effects on rapid eye movement (REM) and deep non-REM sleep.^27–29 We hypothesized that curriculum exposure would improve children’s ability to manage stress, and that this would improve objectively measured sleep. Specifically, we predicted that curriculum exposure would decrease social stress and would increase the duration of total sleep and REM sleep. We further predicted that social stress would mediate improvements in sleep associated with curriculum exposure.

METHODS

Overview

A total of 1,175 children in the third and fifth grade were recruited from 2 Northern California school districts to participate in a prospective observational cohort study evaluating the feasibility and potential benefits of a yoga and mindfulness curriculum, which one of the districts planned to implement during the school day. The curriculum, which taught mindfulness, paced breathing, and movement/posture education, was mandatory for children in this district (ie, the curriculum cohort). The other school district (ie, the non-curriculum cohort) did not offer this curriculum; instead, children engaged in activities planned for their usual physical education (PE) hour. The districts were matched for SES and other sociodemographic variables (see “Demographics”). Curriculum assignment was not random, as it was determined by school district officials. Data collection began in December 2014 (ie, prior to the onset of the curriculum) and continued through January 2019.

Children participating in this larger study were sent home with a flyer offering the chance to participate in additional data collection (ie, objective sleep assessment via polysomnography [PSG]). Families who agreed to participate in the sleep study received 3 in-home PSG assessments, spaced at 1-year intervals (baseline, time [T] 2, and T3). One hundred fifteen children and their parents (58 from the curriculum cohort and 57 from the non-curriculum cohort) provided informed assent and consent to participate in the PSG assessments. Children in the curriculum cohort were scheduled to receive their baseline PSG prior to the start of the curriculum. All methods and procedures were approved by the Institutional Review Board of the Stanford University School of Medicine.

Demographics

Demographic characteristics are presented in Table 1. Children in the non-curriculum cohort were older than children in the curriculum cohort by an average of 8 months. The cohorts did not differ by sex, median household income, or number of adults or children living in the home. All families qualified as low-income according to 2016 county guidelines for housing assistance. Over half of families qualified as extremely low-income; nearly all participants (> 97%) qualified for free or reduced-cost school lunches. More than 90% of participants reported Hispanic ethnicity, and Spanish was the primary language spoken in over 90% of homes. Baseline sleep and psychosocial characteristics are presented in Table 2. Children in the non-curriculum cohort had more total sleep (54 minutes, 25 seconds) and more REM sleep (15 minutes, 55 seconds) than did children in the curriculum cohort at baseline. There were no significant differences in psychosocial characteristics at baseline.

Table 1.

Demographics by cohort.

	Curriculum Cohort	Non-Curriculum Cohort	Comparison
n	57	58	—
Age at baseline, mean (SEM), y	10.89 (0.063)	10.22 (1.05)	t (df = 113) = 4.45, P < .001*
Sex	33 male (57.89%)	33 male (56.9%)	χ2 (df = 1) = .012, P = .914
Race	64.2% Hispanic, Latino, or Mexican American; 17.6% White/Caucasian; 13.7% American Indian or Alaska Native; 3.9% Black or African American (not Hispanic)	82.5% Hispanic, Latino, or Mexican American; 8.8% White/Caucasian; 5.3% Native Hawaiian or other Pacific Islander; 1.8% American Indian or Alaska Native; 1.8% Black or African American (not Hispanic)	χ2 (df = 4) = 8.61, P = .072
Ethnicity	98% Hispanic; 2% non-Hispanic	94.7% Hispanic; 5.3% non-Hispanic	χ2 (df = 1) = 0.824, P = .364
Primary language spoken at home	96.8% Spanish; 3.2% English	92.9% Spanish; 3.6% English; 3.6% other	χ2 (df = 3) = 1.16, P = .763
Number of adults living in the home (median)	2 (range: 1–7)	2 (range: 1–5)	χ2 (df = 5) = 3.959, P = .555
Number of children living in the home (median)	2 (range: 1–8)	2 (range: 1–5)	χ2 (df = 6) = 6.75, P = .345
Household income (median, all sources)	$20,000–$30,000	$20,000–$30,000	χ2 (df = 4) = 11.57, P = .021*
Low-income designation (per county guidelines)	100% low (of which 54% extremely low)	100% low (of which 73% extremely low)	χ2 (df = 1) = 2.610, P = .121
Receiving free or reduced-cost school lunch	97%	100%	χ2 (df = 1) = 0.149, P = .357

^*P < .05. Curriculum cohort statistics include dose-response participants. df = degrees of freedom, SEM = standard error of the mean.

Table 2.

Baseline sleep architecture and social stress by cohort.

Measure	Non-Curriculum Cohort, Mean (SEM)	Curriculum Cohort, Mean (SEM)	Comparison
Total sleep (min)	452.02 (12.87)	397.37 (19.21)	t (df = 56) = 2.36, P = .022*
N1 sleep (min)	17.87 (1.24)	14.85 (2.23)	t (df = 48) = 1.18, P = .224
N2 sleep (min)	201.98 (7.47)	181.27 (11.71)	t (df = 75) = 1.57, P = .122
N3 sleep (min)	147.57 (5.98)	133.19 (8.17)	t (df = 75) = 1.45, P = .151
REM sleep (min)	84.61 (4.49)	68.06 (5.44)	t (df = 75) = 2.34, P = .022*
Sleep-onset latency (min)	34.85 (7.78)	32.58 (9.44)	t (df = 75) = 0.19, P = .854
N1 sleep latency (min)	20.70 (6.08)	37.47 (13.35)	t (df = 60) = −1.30, P = .200
N2 sleep latency (min)	3.51 (1.45)	1.31 (0.86)	t (df = 60) = 0.87, P = .388
N3 sleep latency (min)	6.16 (0.65)	6.03 (1.45)	t (df = 60) = 0.09, P = .925
REM sleep latency (min)	132.43 (8.19)	143.18 (12.32)	t (df = 75) = −0.76, P = .452
WASO (min)	36.67 (5.57)	27.51 (5.06)	t (df = 75) = 0.83, P = .410
N1% of total sleep time	3.88% (0.25%)	3.36% (0.42%)	t (df = 50) = 1.09, P = .282
N2% of total sleep time	44.53% (0.09%)	45.25% (1.53%)	t (df = 75) = −0.43, P = .667
N3% of total sleep time	33.35% (1.33%)	34.68% (2.08%)	t (df = 75) = −0.57, P = .573
REM sleep percentage of total sleep time	18.24% (0.88%)	16.71% (0.99%)	t (df = 75) = 1.23, P = .224
BASC Social Problems (t value)	51.08 (1.41)	50.97 (1.61)	t (df = 69) = 0.50, P = .713

^*P < .05. Presented are raw means (ie, not adjusted for age or sex). Means for the curriculum cohort exclude participants included in dose-response analyses. For a comparison of the full curriculum cohort to the dose-response subset at baseline, see Table S1 ^{(368.6KB, pdf)} . For means by cohort at time points 2 and 3, see Table S3 ^{(368.6KB, pdf)} and Table S4 ^{(368.6KB, pdf)} in the supplemental material. BASC = Behavior Assessment System for Children II, REM = rapid eye movement, SEM = standard error of the mean, WASO = wake after sleep onset.

Curriculum

The curriculum (PurePower, developed by Pure Edge, Inc.; pureedgeinc.org) was administered by trained instructors. Curriculum sessions were conducted twice a week, for 2 academic years, during the students’ scheduled PE class. Sessions lasted 15 to 50 minutes and consisted of active breathing exercises, mindfulness activities, and movement/posture education. Children in the non-curriculum cohort received their standard PE classes.

Polysomnographic assessment of sleep

Overnight PSG was recorded in participants’ homes using the Compumedics Siesta Ambulatory PSG System. This small ambulatory device records breathing using movements of the rib cage and abdomen (piezoelectric bands), a nasal pressure transducer, and an oral airflow monitor (thermistor). Other sensors included a body position sensor, a finger pulse oximeter, an electrocardiogram, and chin and leg (bilateral anterior tibialis) electromyograms. A microphone recorded snoring. Electroencephalography was recorded from the frontal, central, and occipital regions of the scalp (according to the International 10–20 System of Electrode Placement), in addition to standard electro-oculography to track eye movements during sleep. The sleep equipment was placed on the child before bedtime, and parents were asked to have their child fall asleep and wake up at their usual times.

Each PSG was visually scored by a registered PSG technologist, according to guidelines of the American Academy of Sleep Medicine.³⁰ The technologist was blind to experimental group. The following sleep stages were scored: N1, N2, N3, and REM sleep, and wake after sleep onset (WASO). TST, time in each sleep stage, latency to each sleep stage, percentage of TST spent in each stage, and WASO were recorded by the registered PSG technologist.

Other measures

Stress and psychosocial functioning

Psychosocial functioning was assessed via children’s self-report on the Behavior Assessment System for Children II Child (ages 8–11) and Adolescent (ages 12–21) versions (BASC).³¹ Raw scores were converted to age-normed t scores, with higher scores indicating poorer psychosocial functioning. The social problems subscale was used as a measure of social stress.

Engagement with the curriculum

Each year, children in the curriculum cohort answered 3 questions about their perception of, and level of engagement with, the curriculum. The questions were as follows: (1) “I like health and wellness (yoga) at school,” (2) “I feel like health and wellness (yoga) helps me feel better,” and (3) “I use the breathing tools from health and wellness (yoga) when I am not in health and wellness class.” Children answered on a 3-point scale: not at all, somewhat, or very much so.

Statistical analyses

All statistical analyses were performed in SPSS version 26 (IBM Corporation, Armonk, NY).

Baseline comparisons

Cohorts were compared on demographic characteristics (Table 1), sleep architecture, and social stress (Table 2) at baseline by conducting independent-samples t tests for continuous variables and Pearson chi-square tests for nominal and ordinal variables.

Longitudinal analyses of sleep

We conducted a multivariate multiple regression using maximum-likelihood estimation, including the following independent variables: sex (male, female), age (years), cohort (curriculum, non-curriculum), time point (1, 2, 3), and cohort × time point. We conducted 2 primary analyses, with total sleep duration and REM sleep duration, respectively, as the dependent variables. To account for multiple testing, we set an α of .05/2 = .025. Additionally, we conducted exploratory analyses on the following dependent variables: sleep duration in minutes (N1, N2, N3), sleep latency (latency to sleep onset), percentage of TST (N1, N2, N3, and REM sleep), and WASO.

Longitudinal analyses of social stress

We conducted a mixed-model analysis on the BASC social problems subscale. Independent variables were identical to those described for models of sleep.

Dose-response analyses of sleep (exploratory)

As described below (see “Scheduling delays”), 16 participants in the curriculum cohort completed their baseline PSG after initial exposure to the curriculum. We used these 16 participants’ data to conduct exploratory analyses examining short-term effects of the curriculum on sleep. To do so, we conducted multiple regressions using age, sex, and time elapsed (in days) since first exposure to the curriculum at the time of the first PSG as an independent variable. As in the longitudinal analyses, the primary outcomes of interest were TST and REM sleep. Additional exploratory analyses were run on N1, N2, and N3 sleep duration. A natural-log transformation was applied to duration of curriculum exposure, which was right-skewed.

Correlations (exploratory)

To examine sleep changes in relation to curriculum engagement and changes in perceived social stress, we first computed a mean change score for each outcome measure by subtracting the score at baseline from the score at T2, subtracting the score at T2 from the score at T3, and (for participants missing data at T2) subtracting the score at baseline from the score at T3 and dividing by 2. We then computed the average yearly change rate by taking the mean of these measures. Spearman correlations between average yearly change scores in sleep and psychosocial measures were computed, with a 2-tailed α of P < .05. Correlations were run separately for each cohort.

Treatment of missing data

We used an intent-to-treat approach.³² To address missing data, we conducted reduced maximum-likelihood analyses on all available observations, without imputation.³³ In secondary analyses, we performed 20 imputations with fully conditional specification, using pooled statistics.³⁴ Results were identical to analyses of unimputed data.

Enrollment and missing data by time point are presented in Table 3. Data were missing due to attrition and scheduling delays.

Table 3.

Data collected, data included in analyses, and attrition by time point.

Time Point	PSGs Collected	PSGs Included in Analyses	Attrition
Non-curriculum cohort
1	57	46 (81%)	—
2	33	29 (88%)	24 (42%)
3	23	15 (65%)	10 (30%)
Curriculum cohort
1	58	49 [32i] (85%)	–
2	43	38 (88%)	15 (26%)
3	30	26 (87%)	13 (30%)

“PSGs Collected” describes the number of participants contributing data at each time point. “PSGs Included in Analyses” describes the subset of PSGs collected, excluding equipment malfunction or removal. ⁱIndicates number of participants included in analyses of the curriculum cohort (participants not included were analyzed in separate dose-response analyses). PSG = polysomnography.

Attrition:

Attrition was defined as declining to contribute a night of in-home ambulatory PSG at 1 or more time points after baseline. Attrition from baseline to T2 was 26% in the curriculum cohort and 42% in the non-curriculum cohort. Attrition from T2 to T3 was 30% in the curriculum cohort and 30% in the non-curriculum cohort (Table 3). Attrition was not predicted by age, sex, cohort, baseline sleep characteristics, or interactions among these variables (all P > .1).

Despite our best attempts to acclimate participants to the electroencephalography cap and position it in a comfortable manner, some participants may have removed it during the night. Equipment malfunction or removal was highest in the non-curriculum cohort at T3 (Table 3). This reduced the number of non-curriculum participants included in analyses to 15 at T3, although PSGs were actually collected from 23 non-curriculum participants at that time point.

Scheduling delays:

Sixteen participants in the curriculum cohort (28%) completed their baseline PSG after initial exposure to the curriculum (range = 9 to 247 days, median = 43 days, after first curriculum session). They did not differ in age, sex, stress, or sleep measures compared with other participants in the curriculum cohort, with the exception that the 16 delayed participants had longer latency to N2 sleep ( Table S1 ^{(368.6KB, pdf)} in the supplemental material). To ensure baseline data reflected sleep prior to curriculum exposure, we excluded these participants’ baseline PSG data from longitudinal analyses. (Secondary analyses including these participants’ baseline data yielded the same significant effects; see Supplemental Results in the supplemental material). We did include these participants’ T2 and T3 data in longitudinal analyses. Additionally, we used these 16 participants’ data to conduct exploratory analyses of short-term effects of the curriculum on sleep—in effect, a dose-response analysis (described above).

RESULTS

Longitudinal analyses of sleep

TST

We observed a cohort × time interaction: F(2, 95) = 7.81, P = .001 (Figure 1, and Table S2 ^{(368.6KB, pdf)} in the supplemental material). For children receiving the curriculum, TST increased by an average of 74.02 minutes (standard error of the mean [SEM] = 27.97) from T1 to T3 (P = .009). The increase from T1 to T2 (mean = 42.76 minutes, SEM = 23.83) missed significance (P = .075). T2 did not differ significantly from T3 (P = .176). In contrast, for children not receiving the curriculum, TST decreased by an average of 63.58 minutes (SEM = 26.76) from T2 to T3 (P = .020). The decrease from T1 to T3 (mean = 55.07 minutes, SEM = 28.57) missed significance (P = .057). T1 did not differ significantly from T2 (P = .717). No main effects were significant (all P > .2).

Figure 1. Total and REM sleep duration by cohort and time.

Error bars denote SEM. Plotted are estimated marginal means. REM = rapid eye movement, SEM = standard error of the mean, T1, T2, T3 = time points 1, 2, and 3 (with T1 corresponding to baseline).

When data from the dose-response participants at T1 were included in this analysis, the cohort × time interaction remained significant, and no additional effects emerged as significant.

REM sleep

We observed a cohort × time interaction: F(2, 80) = 4.43, P = .015 (Figure 1, and Table S2 ^{(368.6KB, pdf)} ). For children receiving the curriculum, REM sleep increased by an average of 24.53 minutes (SEM = 10.49) from T1 to T3 (P = .021). The increase from T1 to T2 (mean = 14.89 minutes, SEM = 8.01) missed significance (P = .065). T2 did not differ from T3 (P = .298). For children receiving the curriculum, REM sleep did not change over time (all P > .2). No main effects were significant (all P > .3).

When data from the dose-response participants at T1 were included in this analysis, the cohort × time interaction remained significant, and no additional effects emerged as significant.

Additional features of objective sleep (exploratory)

Children receiving the curriculum experienced an increase in N1 sleep from T1 to T3 (mean = 11.24 minutes, SEM = 3.23, P = .001), and from T2 to T3 (mean = 7.83 minutes, SEM = 2.82, P = .007), with no changes in sleep efficiency, sleep latency, WASO, N2, N3, or percentage of time spent in any sleep stage. In contrast, children not receiving the curriculum experienced a decrease in sleep efficiency from T1 to T3, with no change in sleep latency, WASO, N1, N2, N3, or percentage of time spent in any sleep stage. For detailed results, see the Supplemental Results.

Longitudinal analyses of social stress

There were no significant changes in social stress (all P > .05). For means by cohort and time, see Table S2 ^{(368.6KB, pdf)} .

Dose-response analyses of sleep (exploratory)

TST

The model explained more variance than would be expected by chance: F(3, 34) = 5.09, P = .006. Longer exposure to the curriculum was associated with more TST (standardized β = 3.89, P = .001) (Figure 2). We also observed a significant effect of age, such that younger age was associated with more TST (standardized β = −2.53, P = .017). The effect of sex was not significant (P = .825).

Figure 2. Dose-response effect of curriculum exposure on total and REM sleep.

PSG = polysomnography, REM = rapid eye movement.

REM sleep

The model explained more variance than would be expected by chance: F(3, 34) = 3.85, P = .019. Longer exposure to the curriculum was associated with more REM sleep (standardized β = 3.29, P = .002) (Figure 2). Effects of age and sex were not significant (all P > .1).

Additional features of objective sleep

Longer exposure to the curriculum was associated with more N1 sleep and more WASO; no association was detected with N2, N3, sleep latency, or sleep efficiency. For detailed results, see the Supplemental Results.

Correlations (exploratory)

Curriculum cohort

Use of the breathing exercises outside of the classroom (feedback question 3) was associated with increases in total sleep (ρ [n = 35] = .341, P = .045), N1 sleep (ρ [n = 35] = .490, P = .003), N2 sleep (ρ [n = 35] = .360, P = .033), and WASO (ρ [n = 35] = .448, P = .007). Increases in perceived social stress (BASC social problems subscale, t score) were associated with increases in total sleep (ρ [n = 36] = .443, P = .007), REM sleep (ρ [n = 36] = .480, P = .003), and N3 sleep (ρ [n = 36] = .387, P = .020).

Non-curriculum cohort

No correlations were significant (all P > .1).

DISCUSSION

Using a prospective observational cohort design, we observed whether changes in objective sleep architecture and perceived social stress were associated with exposure to a school-based yoga and mindfulness curriculum. Across 3 time points over 2 years, we compared sleep and stress in children from a low-SES school district before, during, and after exposure to the curriculum with the same measures collected from a cohort of children from an SES-matched school district who did not receive the curriculum. Findings suggest that curriculum exposure was associated with an increase not only in total sleep, but also in REM sleep, which is particularly important to emotional well-being and resilience.³⁵ Remarkably, sleep improved within the first 3 months of exposure to the curriculum, in a dose-dependent fashion. Additionally, children who reported being more engaged with the curriculum (eg, using the breathing exercises at home) experienced larger changes in sleep architecture and perceived social stress, suggesting skills learned in the curriculum (such as mindful breathing) may have supported improved sleep.

In contrast, over the 2-year course of the study, participants in the non-curriculum cohort experienced decreased duration of total sleep and REM sleep, consistent with age-related decreases in total and REM sleep,^8,35 as well as with a secular trend of decreasing sleep time.³⁶ Although the mean sleep at T3 (6.5 hours per night in the non-curriculum cohort, 7.7 hours per night in the curriculum cohort) is far below the optimal 9 hours of sleep per night recommended by the National Sleep Foundation,⁴ these data are consistent with a recent study, which reported that as many as 46.1% of students had less than 7 hours of sleep per night.⁵

The cohorts were well matched based on the risk factors that determined selection criteria for the study. These included exposure to environmental and social stressors associated with living in a community of low SES. As shown in Table 1, the groups were nearly identical on key indicators such as household income, language spoken at home, ethnicity, and number of children and adults living in the home. Although there were no baseline differences in SES, sex, or social stress, children in the non-curriculum cohort were older than children in the curriculum cohort and had more total and REM sleep at baseline. We controlled for these baseline differences by conducting statistical analyses using a mixed model, which examines changes both between and within groups. In a mixed-model analysis, changes within each group are measured with respect to that group’s own baseline, thus controlling for baseline differences. Although we did observe significant between-group differences, our central finding was that the cohort exposed to the curriculum experienced increased TST from baseline to T2, which they maintained at T3. In contrast, the cohort not exposed to the curriculum experienced decreased sleep time from baseline to T3. Therefore, we controlled for baseline differences in sleep. We also controlled for age in all analyses. Over the course of adolescence, total and REM sleep, as well as N3 as a percentage of total sleep, typically decrease.³⁵ Given that the older cohort had more total and REM sleep at baseline, and did not differ in N3, it appears that this age difference did not strongly affect sleep architecture. It is important to keep these baseline differences in mind when interpreting results; however, they do not appear to account for the core finding of within-cohort changes over time. An additional between-group difference is that the non-curriculum cohort experienced a large decrease in TST between T2 and T3. The transition between T2 and T3 may have coincided with the onset of puberty. Previous research has shown that the changes in sleep that occur during adolescence occur in proportion to biological age (ie, pubertal development).^37,38 It will be important for future studies to replicate these effects and to examine a potential relationship to pubertal development.

Contrary to our hypothesis that the curriculum would improve sleep by reducing social stress, neither cohort showed significant changes in social stress across time. However, among children who received the curriculum, higher curriculum engagement was associated with increased self-reported social stress. Although these results run counter to our hypothesis, they are consistent with a broader literature showing that practicing mindfulness can increase awareness of stress.³⁹ Thus, an increase in mindfulness may account for the increase in perceived social stress. If the increase in self-reported stress reflected an increase in environmental stressors, rather than increased awareness of stress, then it is even more remarkable that sleep improved during this time period. However, the finding that sleep improved most in the children who used mindful breathing exercises outside of the classroom suggests these children might have learned to calm their nervous system in the presence of stressors, which could account for improvements in sleep.⁴⁰ This bottom-up mechanism (as opposed to top-down effects on perceived stress) would be consistent with literature documenting the physiological effects of the practices taught in the curriculum. Mindfulness practice, as well as yoga-derived breathing, posture, and movement practices, affects sympathetic and parasympathetic activity^41–44 and reduces oxygen intake, breath volume, breathing rate,⁴⁵ and mind wandering,^46–48 which may facilitate a calm transition into sleep.^46–48 Interestingly, N1 sleep is characterized by many of these components, including reduced body temperature, reduced breathing rate, and increased relaxation. It is therefore noteworthy that more frequent use of the mindful breathing exercises outside of the classroom was associated with larger increases in N1 sleep.

Given that the curriculum was delivered in only 1 of the 2 school districts, it remains possible that factors other than curriculum exposure caused the observed effects. We consider this unlikely, since children from both cohorts experienced similar SES-related risk factors for poor sleep. Additionally, curriculum engagement predicted changes in sleep architecture, consistent with a role for the curriculum in improving sleep. It remains possible that children who had better social support, lower social stress, or other protective factors were better equipped to engage with the curriculum; however, we did not observe significant correlations between engagement and stress or other psychosocial measures. Since participants were exposed to the curriculum continuously throughout the assessment period, it is unknown how long sleep improvements were sustained following curriculum cessation. Additionally, children in this study had multiple risk factors for poor sleep, including low SES, ethnic minority, and first-generation immigrant status; it is unknown whether the curriculum was specifically well suited to addressing the risk factors in this sample, or whether it would also benefit children in other sociodemographic groups. It is also unclear which aspects of the curriculum—mindful movement, paced breathing, or all elements in combination—most contributed to these effects.

A limitation of this study is that some data were missing due to attrition. Missing data were highest at T3, although posthoc analyses revealed no association between study participation at T3 and participant characteristics or sleep architecture at T1 or T2. Baseline demographic and sleep characteristics did not differentiate study completers from those who would drop out. Considering the number of families who participated in a PSG assessment at each time point, our rate of retention is consistent with previous longitudinal curriculum or intervention studies.⁴⁹ Low SES is a predictor of attrition.^50–52 Families of low SES may be more likely to encounter logistical barriers to continued participation, such as transportation, which are associated with attrition.^53–55 Longitudinal studies requiring participation of both parent and child have reported attrition rates between 20% and 65%,^56,57 with continued attrition over additional follow-up periods.^50,58,59 In the current study, attrition varied between 26% and 41%. Unfortunately, attrition may be an inherent trade-off of obtaining externally valid data from this sensitive population.

The use of in-home ambulatory PSG confers both benefits and limitations. Ambulatory PSG can be recorded while the child sleeps in his or her own bed, which better reflects typical sleeping patterns than would PSG recorded in a laboratory.^60,61 Ambulatory PSG also reduced logistical barriers to participation, as study personnel set up the PSG equipment in participants’ homes, rather than requiring parents to transport their children to the laboratory and stay overnight. Despite these benefits, ambulatory PSG also has limitations, including that it cannot be monitored in real time.

PSG parameters may differ between the first and subsequent nights of recording, the so-called first-night effect.⁶² We cannot rule out that children’s sleep architecture might have differed during additional recorded nights. Mitigating this concern, first-night considerations may be more important when establishing normative parameters of sleep architecture. In contrast, our goal was to compare changes within and across cohorts. There is no a priori reason to expect first-night effects would differ between cohorts, although we acknowledge results might have differed if we had recorded more than 1 night of sleep.

In summary, we conducted the first study to assess the association between exposure to a school-based health and mindfulness curriculum and changes in children’s objectively measured sleep. Exposure to the curriculum, which taught stress-reduction skills such as mindfulness and paced breathing, was associated with increased duration of total and REM sleep over a 2-year period. Further, engagement with the curriculum, such as using the breathing techniques outside of class, was associated with larger improvements in objectively measured sleep. These results warrant additional investigation to assess whether benefits persist after discontinuation of the curriculum, and whether the benefits generalize to other populations.

DISCLOSURE STATEMENT

All authors have seen and approved this manuscript. Work for this study was performed at Stanford University School of Medicine. This study was funded by the Lucile Packard Foundation for Children’s Health and National Institutes of Health award UL1 TR001085. The authors report no conflicts of interest.

ABBREVIATIONS

BASC,

Behavior Assessment System for Children II

PE,

physical education

PSG,

polysomnography

REM,

rapid eye movement

SEM,

standard error of the mean

SES,

socioeconomic status

T1, T2, T3,

time points 1, 2, and 3 (with T1 corresponding to baseline)

TST,

total sleep time

WASO,

wake after sleep onset