University of Southern California Dissertations and Theses

Sleep health and variability in youth: a real-time data capture study to examine influences on daily dietary intake patterns and longitudinal weight trajectories

(USC Thesis Other)

Sleep health and variability in youth: a real-time data capture study to examine influences on daily dietary intake patterns and longitudinal weight trajectories

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

SLEEP HEALTH AND VARIABILITY IN YOUTH: A REAL-TIME DATA CAPTURE
STUDY TO EXAMINE INFLUENCES ON DAILY DIETARY INTAKE PATTERNS AND
LONGITUDINAL WEIGHT TRAJECTORIES

by

Sydney G. O’Connor

A Dissertation Presented to the
Faculty of the USC Graduate School
University of Southern California

In Partial Fulfillment of the Requirements for the Degree
Doctor of Philosophy (Preventive Medicine, Health Behavior)

August 2019

ii
DEDICATION

To my family: Cheri, Ed, Meredith, Astrid, and Ivy.

iii
ACKNOWLEDGEMENTS

I would like to acknowledge and thank my mentor and committee chair, Genevieve
Dunton. You embody the definition of excellence in research, and your brilliance and grace will
never cease to inspire me. Thank you for the support and guidance that you have provided me
throughout the dissertation process and beyond. I am eternally grateful.
I would also like to acknowledge and thank my committee members: Jimi Huh, for your
statistical expertise and infinite patience in relaying it; Britni Belcher, for your impeccable
advice and willingness to share it; and Katie Page, for your critical perspective and cheerful
support. I am also immensely grateful for the mentorship of Susan Redline and Sue Schembre.
Thank you both for your generosity in sharing expertise and resources. Thank you to the stellar
faculty members of the Department of Preventive Medicine for the lively discussions, timely
pep-talks, and general good cheer that made my doctoral years both enjoyable and memorable.
Of note, I would like to acknowledge and thank Tom Valente, Jennifer Unger, Adam Leventhal,
Richard Watanabe, Jessica Barrington-Trimis, Kim Miller, and Rima Habre for their support.
I am grateful for my friends and colleagues in the REACH Lab, both past and present, of
which there are nearly too many to list: Frank Cedeno, Daniel Chu, Bridgette Do, Eldin Dzubur,
Danait Fessahai, Rosario Gutierrez, Malia Jones, Wangjing Ke, Yue Liao, Gigi Lopez, Jaclyn
Maher, Tyler Mason, Christine Naya, Amy Nguyen, Lissette Ramirez, Brian Redline, Christy
Rico, Eleanor Shonkoff, Shirlene Wang, Jason Yang, Li Yi, and Jennifer Zink. Our research
would not be possible without such a dedicated, detail-oriented, and passionate team. Thank you
also to the families who participated in the MATCH Study, who provided an incredible wealth of
data with the potential to yield insights for years to come.
Thank you to my doctoral colleagues and friends: Amanda Goodrich, Stephanie Pitts,
Christian Cerrada, Chris Warren, Brooke Bell, Sam Cwalina, Sheila Yu, Jessi Tobin, Karen Ra,

iv
Charlotte Deng, and many others. You made the journey fun. To the student organizations
(HBRSA and GSG) and their stellar leaders that connected me to the larger USC community. To
the indomitable Marny Barovich, my very first contact at USC. Your tireless efforts on behalf of
the students in your care do not go unnoticed.
I am grateful for my parents (Ed and Cheri O’Connor), grandparents (Joan Holman and
Joseph Pirani), aunt (Kristina Pirani), and sisters (Meredith, Astrid and Ivy) and their endless
support and love, despite the physical distance between us. Thank you also to Sophia and Wood
Nousome for instantly welcoming me into your family, and for all the home-cooked meals.
Finally, thank you to Darryl Nousome. You have been my rock and my motivation
throughout every step of the doctoral journey. Thank you for supporting and caring for me
through the good and the bad, and for making it all worthwhile.
This dissertation research was made possible through an NIH Ruth L. Kirschstein
National Research Service Award (F31HL13734626; PI: O’Connor) from the National Heart,
Lung, and Blood Institute. This dissertation was also supported by the University of Southern
California Graduate School’s Provost Fellowship and an Institutional Training grant from the
National Cancer Institute (T32CA009492, PI: Pentz). The MATCH Study, on which this
dissertation is based, was funded by the NHLBI (R01HL119255; PI: Dunton).
v
Table of Contents
DEDICATION ......................................................................................................................................................................... II
ACKNOWLEDGEMENTS .................................................................................................................................................. III
ABSTRACT .............................................................................................................................................................................. X
CHAPTER 1: INTRODUCTION ......................................................................................................................................... 1
BACKGROUND AND SIGNIFICANCE ..................................................................................................................................................... 1
Childhood Overweight and Obesity ................................................................................................................................................... 1
Sleep Health Among Youth ................................................................................................................................................................... 3
Relationship between Sleep Health and Obesity .......................................................................................................................... 9
Hypothesized Mechanisms Linking Sleep Health and Obesity ............................................................................................. 9
Challenges and Limitations of Previous Studies ...................................................................................................................... 13
Real-Time Data Capture Approaches for Sleep Assessment ............................................................................................... 17
GAPS IN THE LITERATURE .................................................................................................................................................................. 20
OVERVIEW OF COMPLETED STUDIES ............................................................................................................................................. 23
SPECIFIC AIMS ......................................................................................................................................................................................... 24
CHAPTER 2: AGREEMENT OF ECOLOGICAL MOMENTARY ASSESSMENT (EMA) WITH
ACTIGRAPHY FOR THE ASSESSMENT OF WITHIN-SUBJECT MEAN, VARIABILITY, AND DAY-
TO-DAY SLEEP HEALTH AMONG YOUTH ............................................................................................................. 26
ABSTRACT ................................................................................................................................................................................................. 26
INTRODUCTION ........................................................................................................................................................................................ 28
Real-Time Data Capture Approaches for Repeated Sleep Assessment ........................................................................... 29
Previous Studies Comparing Agreement of Actigraphy with Self-Report Sleep ........................................................ 34
Specific Aims and Hypotheses .......................................................................................................................................................... 36
METHODS .................................................................................................................................................................................................. 37
Participants ............................................................................................................................................................................................... 38
Procedures ................................................................................................................................................................................................. 38
Measures .................................................................................................................................................................................................... 40
Statistical Analysis ................................................................................................................................................................................. 42
RESULTS ..................................................................................................................................................................................................... 44
Data Availability ..................................................................................................................................................................................... 45
Descriptive Results ................................................................................................................................................................................. 46
Agreement of EMA with Actigraphy for Within-Subject Mean and Variability in Sleep ...................................... 46
Agreement of EMA with Actigraphy for Day-Level Sleep ................................................................................................... 47
DISCUSSION ............................................................................................................................................................................................... 48
Limitations ................................................................................................................................................................................................. 54
Implications .............................................................................................................................................................................................. 56
Future Directions ................................................................................................................................................................................... 58
CONCLUSIONS .......................................................................................................................................................................................... 59
CHAPTER 3: EXPLORING THE DAILY EFFECTS OF SLEEP DURATION AND BEDTIME ON 24HR
RECALL DIETARY QUALITY AND EATING BEHAVIOR IN YOUTH ............................................................ 67
ABSTRACT ................................................................................................................................................................................................. 67
INTRODUCTION ........................................................................................................................................................................................ 69
The Association of Usual Sleep Health with Dietary Quality and Eating Behavior ................................................ 69
Real-Time Data Capture Approaches for Repeated Sleep Assessment ........................................................................... 73
Specific Aims and Hypotheses .......................................................................................................................................................... 76
METHODS .................................................................................................................................................................................................. 77
Participants ............................................................................................................................................................................................... 77
Procedures ................................................................................................................................................................................................. 77

vi
Measures .................................................................................................................................................................................................... 78
Statistical Analysis ................................................................................................................................................................................. 83
RESULTS ..................................................................................................................................................................................................... 87
Data Availability ..................................................................................................................................................................................... 87
Descriptive Results ................................................................................................................................................................................. 88
Preliminary Analyses ............................................................................................................................................................................ 89
Association of Sleep Duration and Bedtime with Dietary Quality .................................................................................... 90
Association of Sleep Duration and Bedtime with Eating Behavior ................................................................................. 91
Combined Effects of Duration and Bedtime on Dietary Quality and Eating Behavior .......................................... 91
Interaction Effects of WS Duration and Bedtime on Dietary Quality and Eating Behavior ................................ 91
DISCUSSION ............................................................................................................................................................................................... 92
Role of Usual (BS) Sleep Health on Dietary Quality ............................................................................................................. 92
Role of Daily (WS) Sleep Health on Dietary Quality.............................................................................................................. 95
Role of Usual (BS) and Daily (WS) Sleep Health on Eating Behavior .......................................................................... 96
Relative and Interaction Effects of Sleep Health on Dietary Quality and Eating Behavior ................................. 98
Limitations ................................................................................................................................................................................................. 99
Implications ........................................................................................................................................................................................... 101
Future Directions ................................................................................................................................................................................ 103
CONCLUSIONS ........................................................................................................................................................................................ 104
CHAPTER 4: INDEPENDENT AND COMBINED EFFECTS OF WITHIN-SUBJECT MEAN AND
VARIABILITY IN SLEEP DURATION ON LONGITUDINAL WEIGHT TRAJECTORIES AND
ATTAINED BMI IN YOUTH ......................................................................................................................................... 111
ABSTRACT ............................................................................................................................................................................................... 111
INTRODUCTION ...................................................................................................................................................................................... 113
The Role of Mean Sleep Duration in Obesity ......................................................................................................................... 114
The Role of Within-Subject Variability in Sleep Duration on Obesity ........................................................................ 115
Potential Combined Effect of Within-Subject Mean and Variability in Sleep Duration on Obesity .............. 118
Specific Aims and Hypotheses ....................................................................................................................................................... 120
METHODS ................................................................................................................................................................................................ 121
Participants ............................................................................................................................................................................................ 121
Procedures .............................................................................................................................................................................................. 121
Measures ................................................................................................................................................................................................. 122
Statistical Analysis .............................................................................................................................................................................. 124
RESULTS ................................................................................................................................................................................................... 129
Data Availability .................................................................................................................................................................................. 129
Descriptive Results .............................................................................................................................................................................. 130
Preliminary Analyses ......................................................................................................................................................................... 130
Growth Curve Models ........................................................................................................................................................................ 131
DISCUSSION ............................................................................................................................................................................................. 133
Limitations .............................................................................................................................................................................................. 138
Implications ........................................................................................................................................................................................... 141
Future Directions ................................................................................................................................................................................ 142
CONCLUSIONS ........................................................................................................................................................................................ 142
CHAPTER 5: DISCUSSION AND CONCLUSIONS ................................................................................................ 152
SUMMARY OF FINDINGS ..................................................................................................................................................................... 152
OVERALL LIMITATIONS ..................................................................................................................................................................... 161
IMPLICATIONS ....................................................................................................................................................................................... 162
Methodological Implications .......................................................................................................................................................... 162
Theoretical Implications .................................................................................................................................................................. 164
Intervention and Policy Implications ......................................................................................................................................... 166
FUTURE RESEARCH DIRECTIONS ................................................................................................................................................... 170
CONCLUSIONS ........................................................................................................................................................................................ 173

vii
REFERENCES ................................................................................................................................................................... 175

viii
LIST OF TABLES

TABLE 1. DEMOGRAPHIC CHARACTERISTICS FOR THE MOTHERS' AND THEIR CHILDREN'S HEALTH STUDY
(MATCH) STUDY AND FOR THE MATCH SLEEP SUB-STUDY AT BASELINE AND AT SUB-STUDY
ENROLLMENT ............................................................................................................................................................................. 61
TABLE 2. DESCRIPTIVE STATISTICS FOR EMA AND ACTIGRAPHY SLEEP ASSESSED FOR 7-DAYS AMONG N=39
MATCH SLEEP SUB-STUDY PARTICIPANTS, OVERALL AND STRATIFIED BY DAY OF WEEK ........................... 62
TABLE 3. CORRELATIONS OF WITHIN-SUBJECT MEAN AND VARIABILITY FOR SLEEP VARIABLES OBTAINED FROM
7-DAYS OF EMA AND ACTIGRAPHY AMONG N=39 MATCH SLEEP SUB-STUDY PARTICIPANTS .................. 63
TABLE 4. RESULTS FROM PAIRED T-TEST COMPARING ESTIMATES OF WITHIN-SUBJECT MEAN AND VARIABILITY
FOR VARIABLES DERIVED FROM 7-DAY EMA AND ACTIGRAPHY AMONG N=39 MATCH SLEEP SUB-
STUDY PARTICIPANTS ............................................................................................................................................................. 64
TABLE 5. RESULTS OF MULTILEVEL LINEAR REGRESSION MODELS PREDICTING DAILY EMA REPORTED SLEEP AS
A FUNCTION OF WS DEVIATION IN ACTIGRAPHY SLEEP, AND MODERATION BY DEMOGRAPHIC AND
TEMPORAL COVARIATES ........................................................................................................................................................ 66
TABLE 6. BASELINE DEMOGRAPHIC CHARACTERISTICS FOR 159 DYADS ENROLLED IN THE MATCH STUDY .. 105
TABLE 7. DESCRIPTIVE CHARACTERISTICS OF SLEEP AND DIET COLLECTED FOR 159 CHILDREN ENROLLED IN
THE MATCH STUDY ........................................................................................................................................................... 106
TABLE 8. RESULTS OF MULTILEVEL LINEAR REGRESSION MODELS PREDICTING CHILDREN'S DIETARY QUALITY
AS A FUNCTION OF THEIR PREVIOUS NIGHT'S EMA-REPORTED SLEEP DURATION AND BEDTIME
AGGREGATED ACROSS MEASUREMENT BURSTS 2-6 ................................................................................................... 107
TABLE 9. RESULTS OF MULTILEVEL LOGISTIC REGRESSION MODELS PREDICTING CHILDREN'S EATING BEHAVIOR
AS A FUNCTION OF THEIR PREVIOUS NIGHT'S EMA-REPORTED SLEEP DURATION AND BEDTIME
AGGREGATED ACROSS MEASUREMENT BURSTS 2-6 ................................................................................................... 108
TABLE 10. RESULTS OF MULTILEVEL REGRESSION MODELS PREDICTING CHILDREN'S DIETARY QUALITY AND
EATING BEHAVIOR AS A COMBINED FUNCTION OF THEIR PREVIOUS NIGHT'S EMA-REPORTED SLEEP
DURATION AND BEDTIME AGGREGATED ACROSS MEASUREMENT BURSTS 2-6 ................................................. 109
TABLE 11. RESULTS OF MULTILEVEL REGRESSION MODELS PREDICTING CHILDREN'S DIETARY QUALITY AND
EATING BEHAVIOR AS A COMBINED FUNCTION OF THEIR PREVIOUS NIGHT'S EMA-REPORTED SLEEP
DURATION, BEDTIME, AND INTERACTION OF DURATION*BEDTIME, AGGREGATED ACROSS MEASUREMENT
BURSTS 2-6 .............................................................................................................................................................................. 110
TABLE 12. BASELINE AND FOLLOW-UP CHARACTERISTICS FOR 153 CHILDREN ENROLLED IN THE MATCH
STUDY ....................................................................................................................................................................................... 145
TABLE 13. BASELINE DESCRIPTIVE STATISTICS FOR 7-DAY EMA SLEEP AMONG N=153 MATCH CHILDREN 146
TABLE 14. RESULTS OF MULTILEVEL GROWTH CURVE MODEL SHOWING THE EFFECTS OF BASELINE SLEEP
DURATION ON BMI GROWTH (SLOPE) AND ATTAINED BMI (INTERCEPT)
A
AT THE END OF THE STUDY ... 149
TABLE 15. RESULTS OF MULTILEVEL GROWTH CURVE MODEL SHOWING THE EFFECTS OF BASELINE SLEEP
DURATION VARIABILITY ON BMI GROWTH (SLOPE) AND ATTAINED BMI (INTERCEPT)
A
AT THE END OF THE
STUDY ....................................................................................................................................................................................... 150
TABLE 16. RESULTS OF MULTILEVEL GROWTH CURVE MODEL SHOWING THE EFFECTS OF BASELINE SLEEP RISK
FACTOR SCORE ON BMI GROWTH (SLOPE) AND ATTAINED BMI (INTERCEPT)
A
AT THE END OF THE STUDY
.................................................................................................................................................................................................... 151

ix
LIST OF FIGURES

FIGURE 1. PLOT OF DAY-TO-DAY SLEEP DURATION ACROSS ONE WEEK FOR N=5 HYPOTHETICAL INDIVIDUALS 20
FIGURE 2. BASIC CONCEPTUAL MODEL OF THE DAILY, USUAL, AND LONGITUDINAL EFFECTS OF SLEEP HEALTH
ON DIETARY QUALITY, EATING BEHAVIOR, AND WEIGHT STATUS ........................................................................... 22
FIGURE 3. SCREENSHOTS OF THE MATCH EMA SLEEP ITEMS ........................................................................................... 60
FIGURE 4. BLAND-ALTMAN PLOTS FOR EMA VS. ACTIGRAPHY ........................................................................................ 65
FIGURE 5. HYPOTHESIZED EFFECTS OF WITHIN-SUBJECT MEAN AND VARIABILITY IN SLEEP DURATION ON
OVERALL SLEEP HEALTH ..................................................................................................................................................... 144
FIGURE 6. SPAGHETTI PLOT OF AVERAGE (RED) AND CHILD-SPECIFIC (BLACK) TRAJECTORIES OF BMI ACROSS
MEASUREMENT BURSTS 0 - 3, FOR CHILDREN WITH HIGH (LEFT) AND LOW (RIGHT) BASELINE WITHIN-
SUBJECT MEAN SLEEP DURATION ..................................................................................................................................... 147
FIGURE 7. SPAGHETTI PLOT OF AVERAGE (RED) AND CHILD-SPECIFIC (BLACK) TRAJECTORIES OF BMI ACROSS
MEASUREMENT BURSTS 0 - 3, FOR CHILDREN WITH HIGH (LEFT) AND LOW (RIGHT) WITHIN-SUBJECT
VARIABILITY IN BASELINE SLEEP DURATION. .............................................................................................................. 148
FIGURE 8. OVERALL CONCEPTUAL MODEL OF THE RELATIONSHIPS AMONG SLEEP HEALTH AND VARIABILITY,
ALTERATIONS TO BIOLOGICAL, COGNITIVE, PSYCHOSOCIAL, AND BEHAVIORAL PROCESSES, INCREASED
ADIPOSITY, OBESITY AND OTHER HEALTH CONSEQUENCES AMONG YOUTH. .................................................... 160

x
ABSTRACT

This dissertation consists of three unique, yet interrelated, studies that leveraged real-time
data capture methodologies to examine the influence of sleep health and variability on patterns of
dietary quality and eating behavior, as well as weight trajectories over time, in a sample of youth.
The aims of this project were to: (1) determine the agreement of ecological momentary
assessment (EMA) with actigraphy for assessment of within-subject mean, variability, and day-
to-day sleep health, as well as potential moderators of day-level agreement; (2) examine the daily
effects of sleep duration and timing on next-day dietary quality and eating behavior; and (3)
explore the longitudinal association of baseline within-subject mean and variability in sleep
duration on rate of change in body mass index (BMI) across 1.5 years and attained BMI at the
end of the study. Study 1 revealed moderate-to-high agreement of EMA-assessed sleep variables
(i.e., bedtime, wake time, duration) with actigraphy for estimates of within-subject mean (i.e., the
average value across days), variability (i.e., the standard deviation across days), and day-level
values. This study provides preliminary evidence that EMA is an acceptable and resource-
effective alternative to actigraphy for assessing naturalistic sleep health among free-living youth.
Study 2 revealed a between-subject effect of short usual sleep duration and late usual bedtime on
children’s 24hr recall-assessed Healthy Eating Index (HEI-2015) scores, an indicator of overall
dietary quality. Findings support an association of poorer overall sleep health with elevated
intake of ‘moderation’ foods (e.g., added sugars, refined grains), and decreased intake of
‘adequacy’ foods (e.g., fruits and vegetables). However, Study 2 did not support a within-subject
effect, suggesting that the effects of poor sleep on diet may be cumulative as opposed to acute,
such that a given night of poorer that one’s own usual sleep health does not alter diet the next
day. Contrary to hypotheses, Study 3 found that there was no significant effect of either within-
subject mean sleep duration, within-subject variability in sleep duration, or a combined effect of

xi
both short mean duration and high duration variability at baseline on the rate of BMI change over
1.5 years or attained BMI at the end of the study. The null findings may be attributed to a healthy
overall sleep profile, the relatively older age of the sample (whereas previous studies have found
strong effects in young children), or a follow-up duration that was too short to observe an effect.
Overall, results from this dissertation support the use of EMA to measure free-living sleep
among youth; results also support the incorporation of sleep improvement protocols into youth
dietary improvement and obesity prevention programs.

1
CHAPTER 1: INTRODUCTION

Background and Significance

Childhood Overweight and Obesity

Childhood overweight and obesity rates have steadily increased throughout the past
several decades, and current estimates indicate that approximately one in three children and
adolescents in the US are overweight or obese. (Ogden, Carroll, Kit, & Flegal, 2012) Body mass
index (BMI, kg/m2) is the most commonly used measure for determining weight status, and in
children BMI is highly correlated with underlying total body fat. (Katzmarzyk et al., 2015)
Among youth, overweight and obesity are defined as an age- and sex- specific BMI at or above
the 85
th
and the 95
th
percentile, respectively. (Ogden & Flegal, 2010) Obesity rates increase with
age, with 12.1% of children ages 2-5, and 18.0% of children ages 6-11, and 18.4% of adolescents
ages 12-19 classified as obese. (Ogden et al., 2012) Rates also differ by demographic
characteristics, with significantly higher rates in both black and Hispanic youth as compared to
white non-Hispanic youth. (Hales, Carroll, Fryar, & Ogden, 2017)
Childhood overweight and obesity can have both immediate and lifelong consequences.
Consequences of overweight and obesity include decreased quality of life, (Friedlander, Larkin,
Rosen, Palermo, & Redline, 2003; Schwimmer, Burwinkle, & Varni, 2003) social ostracism,
lower lifelong economic earnings, elevated risk of cardiovascular disease, diabetes, cancers, and
shortened life expectancy. (Dietz, 1998; Schwimmer et al., 2003; A. S. Singh, Mulder, Twisk,
Van Mechelen, & Chinapaw, 2008) Additionally, overweight and obesity in youth tends to
persist into adulthood, (Dietz, 1998; A. S. Singh et al., 2008) which highlights the importance of
obesity prevention and intervention efforts. A recent simulation study published in the New
England Journal of Medicine estimated that, based on current trends, a majority (57%) of
children will be obese at age 35, and of those, half will become obese during childhood. (Giles,

2
Cradock, Sc, Gortmaker, & Ph, 2017) Understanding the factors that contribute to the
development and perpetuation of obesity is essential for informing obesity prevention strategies
to halt this rapid increase.
Etiology of childhood obesity. Obesity is a complex, multi-factorial medical condition
with a multitude of contributing factors across several levels of influence. (Brown, Halvorson,
Cohen, Lazorick, & Skelton, 2015; Han, Lawlor, & Kimm, 2010) At the most basic level, weight
gain results from a caloric imbalance arising from excessive caloric intake (i.e., food intake) and
insufficient caloric expenditure (i.e., physical activity). Although genetic factors affect an
individual’s predisposition to obesity, the rapidly increasing rate of childhood obesity observed
over the past several decades within a genetically stable population points to the critical
contributions of environmental factors and lifestyle behaviors. (Ebbeling, Pawlak, & Ludwig,
2002; Han et al., 2010) The conveniences and demands of modern society have fueled a food
environment that promotes excessive intake of energy-dense foods and beverages, while the
changing home, work, and school physical environments have led to decreased opportunities for
physical activity through play, sports, and active transport. Although childhood obesity is
primarily a consequence of excessive energy intake and insufficient energy expenditure,
effective interventions require the consideration of the individual, interpersonal, environmental,
and societal influences that interplay to shape an individual’s risk of overweight and obesity.
(Ebbeling et al., 2002)
Effectiveness of childhood obesity prevention. Intervention and prevention programs
that attempt to halt the incidence and progression of childhood obesity have had modest to
minimal success in decreasing or preventing weight gain, and many initially successful programs
result in relapse and weight re-gain. (Ebbeling et al., 2002; Waters et al., 2011) This lack of
success is likely primarily due to the incredibly complex set of factors that contribute to

3
childhood obesity, and the difficulty of designing interventions that can effectively address and
encompass the various levels of influence. (Brown et al., 2016) One potential explanation for the
lack of success of existing obesity prevention and interventions programs is their focus on
behavioral targets that are only modestly associated with weight status, such decreasing intake of
dietary fat (vs. decreasing overall caloric intake), or increasing planned physical activity (vs.
increasing overall lifestyle activity). (Ebbeling et al., 2002; Kamath et al., 2008; Sims,
Scarborough, & Foster, 2015; Waters et al., 2011) This has led to a call for a broader range of
factors to be considered in the design of obesity prevention programs. (Yoong et al., 2016) One
potential factor is children’s sleep health. There is emerging evidence for an effect of sleep,
specifically sleep duration, on both weight-related behaviors (i.e., dietary intake) and obesity
risk, suggesting that sleep may represent an understudied ‘tertiary’ lifestyle contributing to
childhood obesity. (Arora, Choudhury, & Taheri, 2015; Arora & Taheri, 2017; Yoong et al.,
2016)
Sleep Health Among Youth

Parallel with increases in childhood obesity rates, there have been significant declines in
children’s sleep health in recent decades. (Matricciani, Olds, & Petkov, 2012). Current reports
describe a shifting of sleep patterns and habits and an overall decline in sleep duration among
youth, amounting to a total overall reduction in nightly sleep duration of approximately of one
hour over the past century. (Matricciani, Olds, & Petkov, 2012) Despite the importance of
adequate sleep for a number of outcomes, a significant proportion of youth fail to meet
guidelines for sleep duration. A longitudinal national study of high school students conducted by
the Youth Risk Behavior Surveillance System in 2007 to 2013 found that fewer than 28.9% of
females and 34.5% of males reported obtaining 8 or more hours of sleep on a typical night.
(Basch, Basch, Ruggles, & Rajan, 2014) In addition to insufficient sleep duration, poor sleep

4
quality is also prevalent and may contribute to adverse health outcomes. For example, a clinic-
based survey of over one thousand youth found that 41% experienced insomnia, and 14%
reported excessive daytime sleepiness, (Archbold, Pituch, Panahi, & Chervin, 2002) suggesting a
high prevalence of sleep disturbances. Evidence of poor sleep health among youth underscores
the need to better understand the relationship of sleep with health and well-being.
The importance of sleep for overall health. The high prevalence of sleep issues among
youth is concerning, because sleep is essential for cognitive, psychological, physical and
emotional well-being. Adequate sleep is critical for the development of children’s executive
cognitive function, (Anderson, Storfer-Isser, Taylor, Rosen, & Redline, 2009; Jan et al., 2010)
working memory and memory consolidation, (Kopasz et al., 2010) stress response, (Meerlo,
Sgoifo, & Suchecki, 2008) and mood regulation. (Baum et al., 2014) Additionally, poor sleep
health has been shown to negatively affect numerous biological processes linked to health and
disease states. For example, sleep is related to metabolic homeostasis, (Spruyt, Molfese, &
Gozal, 2011) cardiovascular functioning, (Javaheri, Storfer-Isser, Rosen, & Redline, 2008)
inflammation, (Patel et al., 2009) insulin sensitivity, (Klingenberg et al., 2013) and glucose
metabolism. (Spiegel, Tasali, Leproult, & Van Cauter, 2009) Among adults, insufficient sleep
duration has been linked to elevated risk of disease; meta-analyses have demonstrated a link
between insufficient sleep and type 2 diabetes, (Cappuccio, D’Elia, Strazzullo, & Miller, 2010a)
cardiovascular disease, (Cappuccio, Cooper, Delia, Strazzullo, & Miller, 2011) and mortality
(Cappuccio, D’Elia, Strazzullo, & Miller, 2010b) among adults. Among both adults and children,
insufficient sleep is also associated with elevated obesity risk. (Cappuccio et al., 2008)
Physiology of sleep and circadian rhythms. Sleep can be defined as “a reversible
behavioral state of perceptual disengagement from and unresponsiveness to the environment”.
(Carskadon & Dement, 2011) Humans have an endogenous (i.e., inherent) biological drive to

5
sleep, which follows a circadian (circa: about; dia: a day) pattern. Beyond the endogenous drive
for sleep, humans also become entrained to the 24hr earth day through exposure to various
zeitgebers (i.e., external cues) that serve to align endogenous circadian rhythm to the external
day. Sleep is driven by the suprachiasmatic nucleus (SCN), a brain region located within the
hypothalamus, and often referred to as the ‘master clock’. In humans, the dynamic physiologic
and behavioral process of sleep is encapsulated by the two-process model of sleep regulation, a
framework of sleep posited by Borbély in 1982. (Borbély, Daan, Wirz-Justice, & Deboer, 2016)
The two-process model posits that there are two independent, yet highly interrelated, processes
governing sleep: Process S and Process C. Process S describes the homeostatic drive to sleep,
which builds with extended wakefulness and decreases during rest. Process C represents
circadian aspects of sleep regulation, such as core body temperature and melatonin, which
govern an individual’s susceptibility and ease for sleep throughout the 24hr day. The homeostatic
(Process S) and circadian (Process C) processes continually interact to influence the timing and
intensity of sleep across the day. (Borbély et al., 2016; Wright, Lowry, & LeBourgeois, 2012)
Developmental considerations in sleep. Normal human sleep consists of two distinct
states: rapid eye movement (REM) and non-rapid eye movement (NREM) sleep. Generally, a
sleep episode begins with NREM sleep, and progresses through successive stages, termed N2
and N3, during which there is low muscle tone and minimal psychological activity. These two
states ebb and flow, alternating during a sleep episode, with longer periods of REM sleep (during
which dreaming arises) occurring across the night. (Carskadon & Dement, 2011) There are
striking differences in sleep staging patterns across the lifespan. For example, in children, a large
proportion of sleep time is comprised of slow-wave sleep (SWS), which then decreases by 40%
during the second decade of life, even when total sleep time remains constant. Patterns of REM
sleep also change throughout the lifespan, and in children and preadolescents the first REM

6
episode is often skipped. However, by mid-adolescence, sleep is similar to that of young adults.
(Carskadon & Dement, 2011)
There is a well-described trajectory of sleep duration throughout the lifespan. Generally,
sleep duration needs decline with age, ranging from 14-17 hours per day in newborns (who have
not yet become entrained to the 24hr day or consolidated sleep to a single nocturnal period), to 7-
8 hours of sleep per night in adults. (Hirshkowitz et al., 2015) Recommendations for sleep
duration differ slightly across various medical and public health organizations, yet in general a
minimum of 8-9 hours is recommended for school-aged children and adolescents through age 18.
According to the National Sleep Foundation, school-aged children (i.e., ages 6-13 yrs.) require 9-
11 hours of sleep per night, while teenagers (i.e., ages 14-17 yrs.) require 8-10 hours per night.
(Hirshkowitz et al., 2015) Similarly, American Academy of Sleep Medicine recommends 9-12
hours for children aged 6-12 years, and 8-10 hours for adolescents aged 13-18 years, (Paruthi et
al., 2016) and the Centers for Disease Control and Prevention (CDC) recommends no less than 9
hours of sleep per night for adolescents. (Basch et al., 2014)
Adolescence and puberty. Adolescence is a stage of transition from childhood to
adulthood occurring during the second decade of life, which is characterized by biological, social
and cognitive maturation. (Colrain & Baker, 2011) Adolescence is also the developmental stage
during which pubertal development occurs. Puberty is initiated by the activation of the
hypothalamic-pituitary-gonadal axis, which releases gonadotropin releasing hormone (GnRH),
activating reproductive system and cueing the subsequent release of testosterone (males), and
estradiol (females), and growth hormones. (Colrain & Baker, 2011; Sisk & Foster, 2004) Along
with hormonal changes, puberty is a time of rapid physical development, including increases in
height and the development of secondary sex characteristics. (Colrain & Baker, 2011) The mean
age of puberty initiation and completion differs by sex, race, and country. In females, the average

7
age for puberty onset is between 10-11 years, while average age of completion is between 15-17
years, while in males the average age of pubertal onset is 11-12 years, terminating around ages
16-17 years. (Parent et al., 2003)
Sleep changes in adolescents undergoing puberty. In adolescence, a number of
alterations to the overall sleep process occur as part of or in conjunction with the onset of
puberty. (Carskadon, 1999; Hagenauer, Perryman, Lee, & Carskadon, 2009; Sadeh, Dahl,
Shahar, & Rosenblat-Stein, 2009) This relationship between pubertal development and sleep in
adolescents is complex and primarily biologically driven. (Dahl & Lewin, 2002; Sadeh et al.,
2009) In the context of the two-process model of sleep, there are two primary biological
mechanisms underlying the link between pubertal development and alterations to the sleep cycle
and sleep behavior: 1) a delay in the release of melatonin and subsequent shifting of circadian
(Process C) sleep timing, and 2) alterations to the homeostatic drive to sleep (Process S). The
hallmark changes to adolescent sleep during puberty include delayed phase preference, decreased
sleep duration, increased daytime sleepiness and tolerance for sleepiness, and greater irregularity
in sleep timing patterns. (Sadeh et al., 2009)
Delayed circadian sleep timing. During puberty, changes to ‘Process C’, or the circadian
regulation of sleep, lead to biologically rooted alterations in sleep timing. The primary driver of
this shifting of the circadian rhythm is a delay in the timing of the release of melatonin, a
hormone excreted by the pineal gland in response to darkness. (Moore & Meltzer, 2008) The
later shift of the circadian rhythm and melatonin release leads to naturally later bed and wake
times, difficulty with falling asleep at earlier bedtimes, and transition to delayed phase
preference observed in adolescents during puberty. (Carskadon, Vieira, & Acebo, 1993;
Hagenauer et al., 2009; Jenni, Achermann, & Carskadon, 2005) As a result of the interaction
between delayed phase preference, which causes adolescents to have later natural bedtimes, and

8
early school start times, which require adolescents to wake up early, adolescents tend to have
chronically insufficient overall sleep duration. (Hagenauer et al., 2009) To make up for lost
sleep, adolescents tend to extend sleep on free days (i.e., weekends). (Carskadon et al., 1993;
Dahl & Lewin, 2002) This shift in sleeping pattern from weekday nights to weekend nights is
referred to as ‘social jet-lag’, (Wittmann et al., 2006) as adolescents who dramatically shift their
sleep habits as a result of weekday demands are somewhat similar to individuals experiencing jet
lag (e.g., traveling from one’s typical time zone to one that is several hours earlier or later), and
are plagued by similar symptoms (e.g., insomnia, difficulty falling asleep or waking up, fatigue).
(Sadeh et al., 2009)
Alteration to the homeostatic regulation of sleep. Puberty is also accompanied by
changes to ‘Process S’, or the homeostatic regulation of sleep. These homeostatic alterations
underlie adolescent’s increased tolerance for sleepiness, caused by a slower build-up of sleep
pressure, or sleep debt. Homeostatic alterations also underlie adolescents’ delayed onset of
sleepiness, which also makes it difficult to adhere to early bedtimes. (Colrain & Baker, 2011;
Hagenauer et al., 2009) Adolescents also experience increases in daytime sleepiness, even when
total sleep duration obtained is held constant, at the same levels of sleep obtained in childhood.
(Carskadon et al., 1980; Sadeh et al., 2009)
External factors influencing sleep alterations in puberty. In addition to the two primary
biological mechanisms outlined above, various external influences may exacerbate the
biologically rooted alterations in adolescent sleep behavior. Increased social demands, greater
personal freedom, use of electronic devices, and poor sleep hygiene may also lead to delayed
bedtimes, while academic demands and school or work schedules may delay bedtimes or require
early awakenings. (Calamaro, Yang, Ratcliffe, & Chasens, 2012; Moore et al., 2009; Watson et
al., 2017) However, the preservation of altered ‘adolescent pattern’ of sleep behavior under

9
extended laboratory conditions free from external influences suggest a stronger role of biological
and physiological mechanisms than that of external influences. (Hagenauer et al., 2009)
Relationship between Sleep Health and Obesity
The dramatic rise in obesity rates and marked decline in sleep health may not be
independent phenomena. Numerous observational, experimental, and epidemiological studies
have provided supported for an association between poor sleep health and risk of overweight and
obesity, especially among children. (Börnhorst et al., 2012; Gonnissen et al., 2013; Nielsen,
Danielsen, & Sørensen, 2011) A recent meta-analysis of longitudinal studies found that children
with short mean sleep duration have twice the odds of overweight or obesity compared to
children with the longest sleep (OR: 2.15; 95% CI: 1.64–2.81). (Fatima, Doi, & Mamun, 2015)
There is also some evidence that this association may be causal: In an experimental study,
children randomized to increase their sleep duration had significantly lower weight gain
compared to children who decreased sleep duration. (Hart et al., 2013) In addition to a role of
short sleep duration in obesity risk, the literature has also illuminated an association between
chronotype, or a later bedtimes, and increased risk of obesity, independent of sleep duration.
(Arora & Taheri, 2015; Asarnow, McGlinchey, & Harvey, 2015) Though evidence for an effect
of short duration on obesity is fairly strong, the physiological and behavioral mechanisms
underlying this association among youth, are not well understood. (Börnhorst et al., 2012)
Hypothesized Mechanisms Linking Sleep Health and Obesity

There are a number of pathways or mechanisms that may underlie the relationship
between sleep health and risk of overweight and obesity among youth. Broadly, these include
biological or physiological, neurological or cognitive, psychological, and behavioral pathways.
Biological and physiological mechanisms. Poor sleep health may affect obesity risk
through biological and physiological alterations. For example, poor sleep has been shown to

10
increase activation of the hypothalamic pituitary adrenal (HPA) axis, (Kumari et al., 2009; Ly,
McGrath, & Gouin, 2015; Matricciani, Olds, & Petkov, 2012) producing stress hormones known
to increase appetite, and suppressing the expression of leptin, a hormone known to produce
feelings of satiety. (Pervanidou & Chrousos, 2012) Poor sleep has also been associated with
inflammation, including changes in IL-6, CRP, and TNFalpha. (Patel et al., 2009) Poor
subjectively-rated sleep quality was associated with negative cardiovascular effects, including
elevated blood pressure or prehypertension among adolescents. (Javaheri et al., 2008)
Additionally, a small pilot study found that children and adolescents with greater sleep-wake
behavior problems had decreased salivary cortisol response, (Capaldi, Handwerger, Richardson,
Laura, & Capaldi, 2010) suggesting that sleep-wake behavior may be linked to the biological
stress response.
Neurological and cognitive mechanisms. Sleep may also impact obesity risk through its
effects on brain structure and neurological and cognitive processes. For example, insufficient
sleep has been shown to promote brain hyperreactivity to food cues, (St-Onge, Wolfe, Sy,
Shechter, & Hirsch, 2014) and lead to elevated neuronal response to food cues in brain regions
implicated in reward. (Demos et al., 2017; Mcreynolds, Trivedi, Roberts, Sy, & Hirsch, 2012)
Previous studies have demonstrated an effect of sleep restriction on decreased food-related
inhibitory control in adolescents. (Duraccio, Zaugg, & Jensen, 2019) There is also increasing
evidence that long-term sleep insufficiency may cause structural brain changes, including
neuronal loss. (Jan et al., 2010) One longitudinal study found that sleep duration variability was
inversely associated with white matter integrity one year later in adolescents, (Telzer,
Goldenberg, Fuligni, Lieberman, & Gálvan, 2015) a change which can lead to deficits in
information processing and cognitive control, and difficulty in planning and executing healthy
behavior across the lifespan, which has clear implications for heightened obesity risk.

11
Psychosocial mechanisms. Poor sleep may also alter psychosocial mechanisms related
to downstream weight-related behaviors and weight gain. Insufficient sleep can decrease one’s
ability to cope with stressors and lead to increased emotional stress. (Vgontzas et al., 2008)
Greater variability in sleep across nights has been linked to elevated psychosocial stress (Mezick
et al., 2009) and greater negative affectivity. (Fuligni & Hardway, 2006) Insufficient sleep
increases the hedonic response to food, (McDonald, Wardle, Llewellyn, & Fisher, 2015) and
decreases dietary restraint. (Markwald et al., 2013) Insufficient sleep may also impact food
cravings; in a small experimental study, young adults who were randomized to extended sleep
duration had significantly lower overall subjective appetite and decreased desire for sweet and
salty foods. (Tasali, Chapotot, Wroblewski, & Schoeller, 2014)
Behavioral mechanisms. Poor sleep health may negatively impact other behavioral risk
factors for overweight and obesity. This may be a direct consequence (e.g., shorter sleep duration
allows for more sedentary screen time due to greater time spent awake), or as a consequence of
the biological, cognitive, or psychological changes that accompany insufficient sleep. For
example, there is evidence for an effect of sleep on children’s overall activity levels. Despite the
greater waking time that arises as a consequence of short sleep duration, children and adolescents
with short sleep do not have increased physical activity levels compared to those with longer
sleep. (Felső, Lohner, Hollódy, Erhardt, & Molnár, 2017; Markwald et al., 2013) In fact, a
systematic review found strong evidence for an inverse association between sleep duration and
sedentary screen time among school-aged children, (Hale & Guan, 2016) suggesting that more
time spent awake results in more time spent engaging in sedentary behavior. A study in
overweight and obese youth found that children with later bedtimes had significantly reduced
moderate-to-vigorous physical activity (MVPA) minutes. (Krietsch, Armstrong, McCrae, &
Janicke, 2016) Another study in children found that compared to long sleepers (≥10 hrs), short

12
sleepers (<9 hrs) spent less time in moderate-to-vigorous physical activity (MVPA), and more
time in sedentary behavior. (Stone, Stevens, & Faulkner, 2013)
Dietary Quality. The literature supports a consistent association between poor sleep
health, poorer overall dietary quality, and sub-optimal eating behaviors in youth. (Felső et al.,
2017) Short sleep duration is associated with increased feelings of hunger and increases in
energy intake that surpass the increased energy requirements of simply being awake for a longer
duration. (Hanlon et al., 2016) Several studies have documented a link between insufficient sleep
and excessive intake of high-fat, high-sugar (HF-HS) foods among youth. Specifically, studies
have found increased intake of snack foods, (Weiss, Xu, et al., 2010) fast food, (Kruger, Reither,
Peppard, Krueger, & Hale, 2014) energy-dense foods, (Westerlund, Ray, & Roos, 2009) and fat
(Weiss, Xu, et al., 2010) among youth with poorest sleep. Poor sleep has also linked to
insufficient intake of health-promoting foods, such as decreased consumption of nutrient-dense
fruits and vegetables among children with short sleep duration (Kruger et al., 2014; Westerlund
et al., 2009) and late bedtimes. (Arora & Taheri, 2015)
Limited results from experimental sleep studies have shown promising effects on dietary
quality. In a study by Asarnow et al. (2017), adolescents who were enrolled in a sleep
intervention to advance (i.e., make earlier) bedtimes ate fewer sweet and salty snacks in a ‘snack
task’, and consumed foods of a lower glycemic index at a standardized breakfast. (Asarnow,
Greer, Walker, & Harvey, 2017) Conversely, another study in adolescents found that those who
were randomized to a restricted sleep condition of 6.5 hours of sleep for five nights had elevated
overall energy (kcal) intake, and elevated intake of grain-based desserts and sweet foods
compared to those in the non-restricted 10 hours of sleep condition. (Beebe et al., 2013)
Eating Behavior. In addition to alterations in children’s dietary quality, poor sleep health
may also influence eating behavior, such as the timing, frequency, and amount of intake across

13
the 24hr day. This is important because certain patterns of eating are thought to have detrimental
effects on cardiometabolic heath, including impaired glucose tolerance, heightened inflammatory
response and weight gain. (St-Onge et al., 2017) Increasing evidence from population-based
studies has revealed altered patterns of eating among youth with poorer usual sleep health. For
example, children with short usual sleep duration have higher odds of skipping breakfast, (Gong
et al., 2017) while children with later usual bedtimes have greater frequency of (unhealthy) snack
consumption, (Arora & Taheri, 2015) and higher odds of excessive caloric intake from non-meal
snacks. (Weiss, Xu, et al., 2010) Poor sleep is also associated with increased total caloric intake
among children, especially in the evenings after dinner time. (Spaeth et al., 2019) In an
experimental study, children assigned to decrease their sleep duration consumed significantly
more (i.e., 134 kcals) average daily kilocalories as compared to children in the increased sleep
condition, and post-hoc analyses revealed that the elevated intake occurred in the evening. (Hart
et al., 2013) These studies suggest that poor sleep health (i.e., shorter duration or later timing)
may alter obesity-related patterns of eating behavior among youth.
Challenges and Limitations of Previous Studies

Although previous studies have demonstrated an association between sleep health and
obesity risk in children, with some evidence for an effect of sleep on diet and eating, previous
studies have been subject to several limitations. Major limitations in include a failure to examine
within-subject effects, reliance on cross-sectional design, lack of controlling for key covariates
and confounders, retrospective (vs. repeated) measurement of sleep and diet, and reliance on a
unidimensional conceptualization of sleep (i.e., usual duration).
Failure to examine within-subject effects. The literature reviewed above suggests a
compelling association between sleep, dietary quality and eating patterns among youth.
However, it is important to note that all of the studies reviewed in the previous sections have

14
been conducted at the person-level, examining the effect of children’s usual sleep health with
their average dietary quality and eating behaviors. Failure to capture and account for daily
fluctuations in both sleep health and eating behaviors may contribute to the ecological fallacy, in
which associations at the between-subjects level differ from associations observed at a lower
level (e.g., at the within-subject level). (Piantadosi, Byar, & Green, 1988) To date, no known
studies have examined the effects of within-subject (WS) day-to-day deviations from one’s own
usual sleep health may on next-day dietary quality or eating behaviors. For example, although
individuals with poor usual sleep health have elevated intake of high-fat, high-sugar (HF-HS)
foods, the consumption of these HF-HS foods may not necessarily follow nights in which that
individual experienced poorer than his or her own usual sleep. By disentangling the between-
subject (BS) effects from the WS effects of sleep on diet and eating, it is possible to determine
the unique effects of day-to-day deviations from one’s own sleep health on next-day diet and
eating behavior. (Schwartz & Stone, 1998) Understanding the time scale through which sleep
affects diet and eating is imperative for the successful design and implementation of prevention
and intervention programs that target sleep as a risk factor for obesity.
Cross-sectional vs. longitudinal design. The majority of studies examining the
relationship between sleep and obesity risk in children have been cross-sectional, which prevents
the ability to examine temporal trends or to infer causality. (Kraemer, Yesavage, Taylor, &
Kupfer, 2000) Although these studies have largely demonstrated an effect of short sleep duration
on risk of overweight or obesity, it is unknown whether poor sleep temporally proceeds weight
gain. Relatively few longitudinal studies have been conducted, and there is some evidence that
longitudinal studies provide weaker evidence for an effect compared to cross-sectional studies.
For example, in a systematic review of fifteen studies examining the relationship of sleep
duration and BMI among adolescents (ages 10-19), a majority (92%) of cross-sectional studies

15
found a positive association between short sleep duration and BMI, while none of the
prospective longitudinal studies found an association. (Guidolin & Gradisar, 2012b) There is also
evidence that participant age and study design may interact to produce differing results. In
contrast to the review by Guidolin & Gradisar, a systematic literature review by Magee & Hale
(2012) included n= 7 longitudinal studies in children examining the effect of childhood sleep
duration on subsequent weight gain. This review determined a consistent negative association
between childhood sleep duration and weight gain, with a median follow-up period of 5.5 years
(range: 3-27 years). (Magee & Hale, 2012) Findings from these two systematic reviews suggest
that study design and length of follow-up time may have important implications for the
significance of the association between sleep and adiposity outcomes.
Key confounders and covariates. Although previous studies have elucidated
demographic characteristics associated with both poorer sleep health and heightened obesity risk,
few studies adequately controlled for these potential confounders. For example, children of low
socioeconomic status (SES) have both poorer sleep health (Felden, Leite, Rebelatto, Andrade, &
Beltrame, 2015) and elevated life-long obesity risk. (Wells, Evans, Beavis, & Ong, 2010)
Similarly, Hispanic and black children have both group greater obesity prevalence (Hales et al.,
2017) and elevated rates of insufficient sleep. (Wong et al., 2013) These characteristics may act
as confounders in analyses seeking to understand the relationship between sleep with obesity, yet
these factors have not been adequately screened for in many previous studies. (Arora et al., 2015;
Nielsen et al., 2011) Failure to control for key confounders may weaken the ability to quantify
the relationship between sleep and weight outcomes. (Nielsen et al., 2011)
Uni-dimensional conceptualization of sleep health. A major limitation of previous
studies is a reliance on mean sleep duration as the sole indicator of children’s sleep health.
(Arora et al., 2015) Although mean sleep duration is an important contributor in obesity risk,

16
several recent studies have highlighted the role of other sleep health characteristics for overall
health and obesity risk. Beyond mean sleep duration, characteristics such as sleep timing (i.e.,
bedtime, wake time, sleep midpoint), sleep quality (i.e., efficiency, nocturnal awakenings, sleep
disturbances), and the overall degree of within-subject variability in sleep health may be
independently associated with obesity risk in youth, even when controlling for mean sleep
duration. (Arora & Taheri, 2015; Asarnow et al., 2015; Golley, Maher, Matricciani, & Olds,
2013; He, Bixler, Berg, et al., 2015; D. C. Jarrin, McGrath, & Drake, 2013; Kjeldsen et al., 2014;
A. L. Miller et al., 2014; Moore et al., 2011) Although mean sleep duration has strong links to
overall health and wellbeing, these other sleep health conceptualizations may be important to
consider, and may yield new insights into how sleep may affect weight related behaviors and
outcomes. (Bei, Manber, Allen, Trinder, & Wiley, 2017; Bei, Wiley, Trinder, & Manber, 2016;
He, Bixler, Liao, et al., 2015)
Assessment of diet and eating. The majority of previous studies examining the effects of
sleep health on children’s diet have used retrospective food frequency questionnaires (FFQs),
which ask participants to recall their average intake of various types of food over a specified
period of time (e.g., over the past month). (Kruger et al., 2014; Westerlund et al., 2009) Due to
their timescale (e.g., weekly or monthly average), this type of data is not well suited for drawing
conclusions regarding the within-subject effects of sleep (i.e., how does poorer than one’s own
usual sleep on a given night relate to dietary the following day?). In contrast, 24hr dietary recalls
collect detailed, time-stamped information on foods and beverage consumption for a given day.
24hr dietary recalls are appropriate for use in children and adolescents, and are considered the
gold standard for measuring dietary intake in population-based studies. (Burrows, Martin, &
Collins, 2010). In contrast to FFQs, the data obtained from 24hr dietary recalls can be linked
with day-level sleep data and to analyze the within-subject effects of sleep on dietary and eating.

17
Real-Time Data Capture Approaches for Sleep Assessment

As discussed in previous sections, many of the existing studies in this body of literature
have been limited by (1) retrospective recall of key variables, (2) singular focus on usual sleep
duration as the indicator of sleep health, and (3) a failure to examine the potential within-subject
(WS) effects in addition to the between-subject (BS) effects. Many of these limitations can be
resolved by the use of real-time data capture (RTDC) methods. RTDC approaches repeatedly
assess behaviors in real time as they occur, within an individual’s naturalistic setting. Two
common examples of RTDC modalities include wearable sensors (i.e., actigraphy) and
ecological momentary assessment (EMA).
RTDC methods pose many advantages over traditional survey measures and laboratory-
based studies. First, in contrast to traditional retrospective surveys, which ask participants to
estimate their usual behavior over a given period of time (e.g., usual sleep duration over the past
month), RTDC methods allow for the assessment of behaviors that may vary from day-to-day
(e.g., actual sleep duration on a given night). This allows for the disaggregation of the BS from
WS effects of a predictor of interest on various outcomes. Additionally, by assessing behaviors
immediately or soon after they occur, recall bias is decreased. RTDC techniques are easily
incorporated into an individual’s daily life and naturalistic setting, which increases ecological
validity and representative design. (Tate et al., 2013) Furthermore, RTDC methods can
incorporate the measurement of contextual factors (e.g., social, environmental) that proximally
affect processes of interest. (Schwartz & Stone, 1998) Insights gleaned from RTDC approaches
have led to insights on human behavior that could not otherwise have been obtained using
traditional methods. (Shiffman, Stone, & Hufford, 2008)
Variability in sleep health. Importantly, RTDC methods can be used to understand the
role of variability in sleep health in relation to diet and obesity. This is important because beyond

18
mean (i.e., average) sleep duration, several recent studies have highlighted the role of high
within-subject variability in duration (i.e., to the extent that an individual’s sleep duration varies
across repeatedly assessed days) as a novel risk factor for obesity. Some degree of variability in
sleep across repeatedly assessed days is common. (Dillon et al., 2015) However, a high degree of
variability in sleep has been associated with less healthy lifestyle choices, such as poorer dietary
quality and elevated sedentary behavior. (Duncan, Kline, Rebar, Vandelanotte, & Short, 2016)
Growing evidence suggests that, even after adjusting for mean sleep duration, greater within-
subject variability in sleep duration may be independently associated with children’s obesity risk.
(He, Bixler, Berg, et al., 2015; He, Bixler, Liao, et al., 2015; Moore et al., 2011; Spruyt et al.,
2011) However, due to the limited number of studies have assessed the role of within-subject
variability in sleep duration on children’s obesity, and overall mixed findings, there is a need for
additional investigation. (Bei et al., 2016; D. Jarrin, Mcgrath, & Drake, 2013)
Conceptualization of within-subject sleep using RTDC methods. By repeatedly
assessing sleep health across multiple days, three primary sleep conceptualizations can be
determined for any given sleep health variable (e.g., bedtime, duration): (1) within-subject mean,
(2) within-subject variability, and (3) day-level. Within-subject mean sleep health reflects an
individual’s average (i.e., usual) sleep health, and is calculated by taking the average of repeated
measures across days. It is important to note that although an individual’s mean sleep value is the
most commonly used conceptualization of sleep health, the value obtained by averaging
repeatedly assessed sleep may differ from the value obtained through retrospective recall.
Within-subject variability represents the extent to which an individual exhibits variation (i.e.,
lack of consistency) in sleep across days. Within-subject variability takes into account lower-
level (i.e., day-level) observations to calculate an overall measure of variability, calculated by
taking the standard deviation of the mean of repeatedly measured sleep. Additionally, and in

19
contrast to the previously described subject-level traits, day-level, or day-to-day sleep describes
the actual levels of sleep for a given day. Day-level sleep values can be used to calculate the
within-subject deviation of a given night of sleep from an individual’s own usual level (i.e., by
subtracting the individual’s mean level from the actual observed value on a given night to
produce a deviation score). RTDC approaches allow for novel examination of the effects of these
various sleep health conceptualizations on diet and weight outcomes among youth.
Figure 1 displays hypothetical data for sleep duration across one week for n=5
individuals. Based on the actual day-to-day sleep values, it is clear that these individuals differ in
their within-subject mean duration (i.e., average value of sleep duration across days), and their
level of within-subject variability (i.e., the degree of erraticism in sleep duration across nights,
represented by the standard deviation). Further, this data can be used to calculate how the BS and
WS values, which describe how a given subject’s mean sleep differs from the group mean level
(i.e., the between-subjects [BS] mean), as well as the degree of deviation of a given night’s sleep
from that individual’s own mean level (i.e., the within-subject [WS] mean). (Schwartz & Stone,
1998) Thus, each of these three conceptualizations of sleep health (i.e., within-subject mean,
within-subject variability, and day-to-day levels) may be important to consider when exploring
the role of sleep health on weight-related behaviors (i.e., diet) and obesity risk.

20

Figure 1. Plot of day-to-day sleep duration across one week for n=5 hypothetical individuals

Gaps in the Literature

The scientific literature demonstrates a significantly increased risk of obesity for children
with the shortest usual sleep duration. However, there are questions that remain unanswered, and
these research gaps form the core of this dissertation.
An overarching question concerns the measurement of sleep health: while many previous
studies have measured children’s sleep using a single retrospective survey item regarding usual
sleep duration, RTDC methods are becoming more commonly used to repeatedly asses sleep
across multiple days and to calculate subject-level sleep characteristics (i.e., within-subject mean
and variability). However, it is unknown how this repeated sleep assessment may lead to similar
or differing associations or new insights into the role of sleep on children’s obesity risk. For
example, while previous studies have found that retrospective, self-reported ‘usual’ sleep
duration is cross-sectionally related to BMI, this relationship may not remain when sleep is
repeatedly assessed across nights. Additionally, while previous studies have assessed the
agreement of self-report sleep logs compared to actigraphy, (Tremaine, Dorrian, & Blunden,
2010; Werner, Molinari, Guyer, & Jenni, 2008) a commonly used and validated sleep assessment

21
tool, no study to date has examined the agreement of EMA sleep assessment with actigraphy for
estimates of children’s within-subject mean sleep. Nor have any known studies sought to
examine the agreement of self-report compared to objective actigraphy for determining day-to-
day levels of sleep, or the overall degree of within-subject variability across days. It is important
to understand the agreement of EMA with actigraphy for these various sleep health
conceptualizations, to inform future studies using EMA to assess sleep health.
While several cross-sectional studies have suggested that sleep may be related to diet and
eating, (Felső et al., 2017) none have examined the within-subject, daily effects of deviations
from children’s own mean levels of sleep health (e.g., mean duration) on their next-day dietary
intake, such as overall quality and patterns of eating behavior. Understanding the time scale
through which insufficient sleep (e.g., short duration) may affect dietary intake is important, as it
can inform theory and interventions relating to the interplay between sleep and weight-related
behaviors and outcomes.
Finally, few studies to date have examined the unique role of within-subject variability
(i.e., standard deviation of an individual’s repeatedly measured levels) in sleep duration on
children’s obesity risk, and the vast majority have been cross-sectional in nature. Due to the
emerging evidence for a cross-sectional effect of within-subject sleep variability on obesity
status, it is essential to understand how within-subject variability in sleep duration may also
affect the change in children’s weight status over time.
In order to address these research gaps, this dissertation examined the interrelationships
between sleep health and variability, dietary quality and eating behavior, and weight status in
children, using longitudinal design and a RTDC approach. The source of data for this project was
the Mothers’ and Their Children’s Health (MATCH) Study, a longitudinal study of mother-child
dyads, which consisted of six bi-annual assessments (‘measurement bursts’) across three years.

22
(Dunton et al., 2015) The use of repeated EMA sleep assessments across multiple years in the
MATCH Study provides a unique framework in which to examine sleep health and its
relationship to weight-related behaviors and outcomes.
The basic conceptual model of the hypothesized relationships between sleep, diet and
eating, and weight outcomes is presented in Figure 2. In this model, distinct aspects of children’s
sleep health may have downstream effects on dietary intake, such as dietary quality or patterns of
eating behavior. These effects may unfold at the day-level (i.e., sleep on a given night affects
diet/eating the next day), or at the usual level (i.e., overall sleep affects overall diet and eating).
This model posits that over time, both sleep health and diet/eating alterations may directly or
indirectly lead to weight gain and elevated obesity risk.

Figure 2. Basic conceptual model of the daily, usual, and longitudinal effects of sleep health on
dietary quality, eating behavior, and weight status

23
Overview of Completed Studies

This dissertation examined the relationships between sleep health, diet and eating
behavior, and obesity outcomes in children, using real-time data capture techniques and
longitudinal study design. Despite the strong evidence for an association between short usual
sleep duration and elevated risk of overweight and obesity among youth, few studies have
considered the role of day-to-day variability in sleep duration as it relates to weight gain. Nor
have studies examined the within-subject effects of sleep health on next-day weight-related
behaviors, such as dietary intake quality and patterns of eating behavior. Further, although EMA
is increasingly used to repeatedly assess children’s free-living behaviors and psychological
states, no studies to date have examined the agreement of EMA sleep assessment against
objective actigraphy for determining children’s day-to-day and overall sleep health levels. To
address these gaps in the literature, this dissertation employed repeated day-do-day sleep
assessment using both EMA and actigraphy (in a sub-sample), multiple 24hr dietary recall
assessments to capture day-level dietary quality and patterns of eating behavior, and bi-annual
measurements of BMI within a longitudinal study of youth.
The three completed studies address: (1) the agreement of EMA-assessed within-subject
mean, variability, and day-to-day sleep health in children as compared to actigraphy as well as
moderators of the day-level agreement; (2) the daily effects of sleep duration and timing on next-
day dietary quality and eating behavior; and (3) the longitudinal relationship of baseline within-
subject mean and variability in sleep duration on rate of change in body mass index (BMI) across
1.5 years and attained BMI at follow-up.

24
Specific Aims

Study 1: Agreement of Ecological Momentary Assessment (EMA) with actigraphy for the
assessment of within-subject mean, variability, and day-to-day sleep health among youth
1. To test the correlation and agreement between EMA and sleep-log informed actigraphy
for measuring within-subject mean and variability (i.e., standard deviation) in sleep (i.e.,
bedtime, wake time, duration, and WASO).
2. To test the day-to-day (i.e., day-level) agreement of EMA with sleep-log informed
actigraphy sleep across seven days.
3. To determine the moderating effects of children’s sex, overweight/obesity status, pubertal
development, as well as day of the week on the day-to-day (i.e., day-level) agreement of
EMA with sleep-log informed actigraphy across seven days.
Study 2: Exploring the daily effects of sleep duration and bedtime on 24hr recall dietary
quality and eating behavior in youth
4. To test whether, within individuals, a night with poorer than usual sleep (i.e., shorter
duration, later bedtime) affects next-day dietary quality (i.e., HEI-2015 total score, HEI-
2015 Moderation and Adequacy Sub-Scores).
5. To test whether, within individuals, a night with poorer than usual sleep (i.e., shorter
duration, later bedtime) affects next-day eating behavior (i.e., breakfast consumption,
frequency of eating events, proportion of kilocalories consumed in the evening).
6. To test the independent (i.e., the effect of one sleep health predictor controlling for the
other sleep health predictor) and moderation (i.e. interaction effect) effects of nights with
poorer than usual sleep duration and bedtime on next-day dietary quality and eating
behavior.

25
Study 3: Independent and combined effects of within-subject mean and variability in sleep
duration on longitudinal weight trajectories and attained BMI in youth
7. To test the association of baseline within-subject mean sleep duration on the rate of
change of BMI across 1.5 years and intercept of BMI at 1.5 years.
8. To test the association of baseline within-subject variability in sleep duration on the rate
of change of BMI across 1.5 years and BMI intercept at 1.5 years.
9. To test the combined effect of baseline shorter within-subject mean sleep duration and
higher sleep duration variability (i.e., using a sleep risk factor score, with possible scores
of 0 [adequate duration/low variability], 2 [either adequate duration/high variability; or
short duration/low variability], and 4 [short duration/high variability]) on the rate of
change of BMI across 1.5 years and attained BMI at 1.5 years.

26
CHAPTER 2: AGREEMENT OF ECOLOGICAL MOMENTARY ASSESSMENT
(EMA) WITH ACTIGRAPHY FOR THE ASSESSMENT OF WITHIN-SUBJECT
MEAN, VARIABILITY, AND DAY-TO-DAY SLEEP HEALTH AMONG YOUTH

Abstract

Introduction: Because sleep health is important for myriad developmental, cognitive, and health
outcomes in children and adolescents, research seeking to measure sleep within this population is
increasingly common. There are a variety of methods available for assessing free-living sleep
among youth, though all are subject to various limitations. Ecological momentary assessment
(EMA) is an increasingly utilized research method, which provides a novel tool for repeatedly
assessing youth sleep health. However, no study to date has examined the agreement of EMA as
compared to actigraphy, an objective and well-accepted sleep assessment method, for capturing
within-subject mean and variability, as well as day-to-day sleep health variables among youth.
Methods: Participants included 40 youth (mean age: 11.9 ± 1.1 years; 58% female) participating
in a longitudinal study. Participants completed daily EMA sleep surveys and wore a wrist-worn
Actigraph GT3-x accelerometer for one week at home. Correlation and paired t-tests compared
the agreement of within-subject mean and variability (SD) of four sleep health variables (i.e.,
bedtime, wake time, duration, wake-after sleep onset [WASO]) in EMA compared to actigraphy,
and Bland-Altman plots were used to visually represent the agreement between methods.
Multilevel linear regression was used to examine the agreement of within-subject (WS) day-level
deviations from a participant’s mean actigraphy-assessed sleep level of a given sleep variable
(i.e., duration) with EMA report of that sleep variable for the same day, and to test for temporal
(e.g., weekend) and person-level (e.g., sex) moderators of these associations.
Results: EMA revealed the following within-subject mean levels for bedtime (22:04 ± 00:59),
waketime (06:56 ± 00:34), duration (532 ± 65 min), and WASO (41.9 ± 18.1 min). The EMA
sleep measure had overall moderate to high correlation (r= 0.63 – 0.86; p<0.0001) with

27
actigraphy for measuring within-subject mean duration, bedtime, and waketime but had poor
correlation for WASO. Similarly, EMA had good correlation with Actigraphy for within-subject
variability of bedtime (r= 0.70, p<0.0001) and duration (r= 0.44, p<0.01), but not waketime or
WASO. Paired t-tests revealed that EMA estimates of within-subject mean wake time and
duration were not significantly different from actigraphy, but EMA WASO and bedtime were
significantly lower and earlier than actigraphy, respectively. In day-level analyses using
multilevel models and controlling for between-subject levels of a given sleep health variable,
there was a significant positive association of within-subject deviations in actigraphy sleep with
EMA-reported sleep for the same day, indicating significant day-level agreement between the
methods for all four variables. Analyses revealed a significant moderation effects of sex and
pubertal stage for WASO, indicating that the agreement of actigraphy and EMA WASO for a
given night was stronger for boys (vs. girls), and for those in earlier stages of pubertal
development (vs. later stages).
Conclusions: EMA sleep assessment demonstrates good correlation and agreement with
actigraphy for measuring levels of within-subject mean, variability and day-to-day bedtime,
wake time and sleep duration (but not WASO). EMA may be used as a cost- and resource-
effective alternative to actigraphy in studies seeking to examine the within-subject antecedents
and consequences of sleep health among youth.

28
Introduction

Adequate sleep among youth is essential for cognitive functioning and memory
consolidation, (Anderson et al., 2009; Kopasz et al., 2010) while insufficient sleep is linked to
poorer mood regulation, (Baum et al., 2014) behavioral difficulties, (Biggs, Lushington, van den
Heuvel, Martin, & Kennedy, 2011) and physical health consequences including elevated obesity
risk (Cappuccio et al., 2008). Despite the high cost associated with insufficient sleep, youth
today obtain less sleep than in the past, and sleep disturbances and clinical sleep problems are
increasing in prevalence. (Matricciani, Olds, & Petkov, 2012) Due to its importance for overall
health, sleep research among youth populations is increasingly common, and Healthy People
2020 for the first time included sleep health as an major objective. (US Department of Health
and Human Services, 2010)
Sleep is a reversible state, characterized by minimal movement, supine posture, reduced
responsiveness to stimuli, an increased arousal threshold, and a set of physiological changes
including neurological brain wave activity alterations. The gold standard for sleep assessment is
polysomnography (PSG), (Douglas, Thomas, & Jan, 1992) which uses electroencephalogram
(EEG) to measure brain dynamics in conjunction with other biophysiological measurements
(e.g., heart rate, eye movement). However, full PSG is limited by to laboratory (non-naturalistic)
settings and a high administration cost that makes its use impractical for home and/or repeated
assessment. Alternatively, many population-based studies have used retrospective sleep
questionnaires, often a single item, which ask individuals to estimate their usual sleep health
characteristics over a designated period of time (e.g., ‘Over the past month’). Retrospective
reports of usual behavior are problematic for a number of reasons, primarily due to the
propensity for recall bias, and low agreement when compared to repeatedly assessed objective
sleep in the same individual. (Girschik, Fritschi, Heyworth, & Waters, 2013) It is important to

29
also note that retrospective approaches are unable to fully capture sleep health within individuals
from day-to-day or over time. This is a major limitation, as studies increasingly seek to measure
and understand the day-to-day effects of sleep on subsequent biological, psychological, and
behavioral outcomes. (Cox, Sterba, Cole, Upender, & Olatunji, 2018; Schmid, Hallschmid, &
Jauch-, 2008) Thus, it is important to develop and test assessment methods that can repeatedly
assess youth sleep in a naturalistic setting.
Real-Time Data Capture Approaches for Repeated Sleep Assessment

To date, there are several primary sleep assessment methodologies that allow for repeated
day-to-day assessment of sleep in a naturalistic setting. These real-time data capture (RTDC)
methods pose several advantages. First, by assessing sleep behavior in real- or near-real time,
RTDC approaches greatly reduce the potential for recall bias. (Shiffman et al., 2008) Because
RTDC approaches measure sleep within the naturalistic setting, they typically result in greater
ecological validity. By measuring sleep over multiple days, RTDC methods are well suited to
capture the natural variability in sleep across nights (Dillon et al., 2015; Gradisar, Gardner, &
Dohnt, 2011), which can help to illuminate the within-subject antecedents and consequences of
sleep levels across days. Finally, repeated sleep assessment allows one to calculate the overall
degree to which an individual’s sleep health fluctuates across repeatedly assessed days (i.e.,
within-subject variability). This is important because within-subject variability in sleep health
has recently emerged as potential independent risk factor for numerous adverse health outcomes.
(Bei et al., 2016; Buysse, 2013; Gooley, 2016)
While there are many emerging RTDC methodologies for day-to-day sleep assessment,
the two most commonly used approaches are sleep logs and actigraphy, which have both been
validated for use in youth populations. (Meltzer, Montgomery-Downs, Insana, & Walsh, 2012)
Despite their utility, both approaches are subject to limitations. Aside from sleep logs and

30
actigraphy, a subset of RTDC approaches called Ecological Momentary Assessment (EMA)
offer a novel approach for sleep assessment that may pose advantages over existing RTDC
methods. EMA is a set of tools used to repeatedly assess individuals in the context of their daily
lives, which is typically characterized by the frequent delivery of short surveys through a
smartphone application (‘app’). (Shiffman et al., 2008)
While EMA is most often characterized by repeated assessment throughout the day, with
the aim of capturing a representative sample or full coverage or the phenomena of interest, EMA
may also be used for capturing less frequent (i.e., day-level) phenomena, such as sleep health.
However, in contrast to traditional sleep logs, in which participants are instructed to complete a
paper or electronic survey at the end of each day to report on their previous night’s sleep, EMA
methods are distinct in that they: (1) prompt the participant to complete entries via a signal or
tone, instead of relying on participants to remember to complete an entry; (2) may be completed
more quickly or with less effort than traditional methods; (3) can be limited in nature, such that
surveys time out after a pre-specified window (e.g., 10 minutes after the initial signal), which
standardizes and certifies (via a time-stamp) and prevents lengthy response delays known to
decrease accuracy; and (4) can be uploaded to the cloud in real-time and used to inform real-time
interventions that tailor approach based on a participant’s day-level sleep health. Although EMA
methodologies present a useful yet underutilized approach to measuring and understanding sleep
among free-living individuals, no studies to date have systematically examined the agreement of
EMA sleep assessment to existing methods, such as actigraphy. A brief discussion of the
advantages and disadvantages of sleep logs, actigraphy, and EMA for repeated sleep assessment
are summarized next.
Sleep logs. Sleep logs are commonly used in population-based studies. Sleep logs
generally involve providing participants with a set of paper or electronic questions to be

31
completed daily. Typically, individuals are asked to complete these logs upon awakening
(reporting on their previous night’s sleep) and prior to going to bed (reporting on other factors
that might be relevant to sleep, including daytime naps and medication use). The primary
advantage of sleep logs is their ability to capture patterns of sleep behavior across multiple days,
with minimal cost and fairly high reliability, even among youth. (Wolfson & Carskadon, 2003)
Beyond the lack of objective data on sleep patterns, sleep logs are also subject to several other
limitations. First, sleep logs lack a completion timestamp, and although participants are
instructed to complete an entry each day, they may forget and back-fill pages at a later date. The
potential lag between actual sleep and report of sleep is unknown and thus reports may be
differentially subject to recall biases. Additionally, because they require participants to
proactively initiate filling out the log each day, sleep logs may be perceived as burdensome.
Finally, as discussed in the previous section, sleep logs are not appropriate for studies that seek
to incorporate real-time feedback on daily sleep health into adaptive interventions. Despite their
limitations, sleep logs are often assessed alongside actigraphy in order to guide interpretation of
the objective sleep data.
Actigraphy. Actigraphs are small wearable sensors that record body motion. While there
a number of commercially available devices and proprietary sleep scoring algorithms, the current
paper focuses on the Actigraphy GT3-x accelerometer (Actigraph, Pensacola, FL), and the Cole-
Kripke sleep algorithm. The Actigraph GT3-x device uses an acceleration sensor to continuously
collect 3-dimensional movement data (in Hz) within pre-specified time windows (i.e., epochs)
such as 10-sec, 15-sec, 30-sec. Using the ActiLife software, a proprietary algorithm is then
applied to the raw data to convert each epoch to activity count units. (ActiLife, 2012) The
activity count data is then processed using a validated sleep algorithm (e.g., Sadeh, Cole-Kripke)
to score each epoch as asleep or awake, using a moving average approach that takes into account

32
the activity count levels immediately preceding and following a given epoch to inform the
classification of that given epoch as asleep or awake. (ActiLife, 2012) The ActiLife software
produces several sleep variables at the day-level and person-average level. A minimum of 5 days
of monitoring is recommended to obtain reliable estimates of usual sleep in children and
adolescents. (Acebo et al., 1999) Scoring and interpretation of the actigraphy data is further
improved when informed by a concurrently completed sleep log, specifically for determining
sleep start and end time; in such cases this may be considered as sleep-log informed actigraphy.
Compared to self-report methods, actigraphy collects objective movement data to
estimate sleep patterns. A primary advantage of actigraphy is its passive sensing of movement,
which keeps participant burden at a minimum. Actigraphy is recommended by the American
Academy of Sleep Medicine for estimating sleep patterns in free-living individuals, (Jr et al.,
2007) and actigraphy has been validated against PSG in both youth and adults. (Meltzer,
Montgomery-Downs, et al., 2012; Paavonen, Fjällberg, Steenari, & Aronen, 2002; Slater et al.,
2015) Despite its utility, there are some disadvantages to the use of actigraphy. Although
actigraphy has been validated against PSG, there are known systematic measurement biases,
especially in estimating aspects of nocturnal sleep quality (e.g., night awakenings, WASO).
(Sadeh, 2011) Actigraphy can be cost and resource intensive, as it requires specialized devices
and software. Further, as with all wearable sensors, there is potential for missing data due to
removed or lost devices, or technical errors. Additionally, actigraphy cannot capture an
individual’s subjective appraisal of their own sleep quality, which may be distinct from objective
sleep yet and uniquely associated with psychobiological consequences of poor sleep health. (El-
Sheikh, Buckhalt, Keller, & Granger, 2008) Finally, the breadth of actigraphy devices, protocols,
and scoring algorithms can yield slightly different estimates of sleep and complicate comparison
and interpretation across studies. (Ancoli-Israel et al., 2003; Sadeh & Acebo, 2002)

33
EMA. EMA is measurement tool which may pose some practical and methodological
advantages over existing RTDC sleep assessment methods (e.g., actigraphy). In brief, studies
may use EMA to solicit short daily or within-day surveys pertaining to sleep health, including
bed and wake times, ratings of sleep quality, or other related constructs. Studies utilizing EMA
generally encourage participants to use their own smartphones, which can be resource-effective,
and can increase compliance and decrease participant burden by preventing the need to carry an
additional study device. (Shiffman, 2007) Because EMA is characterized by short (e.g., less than
3 minute) surveys, and because delivery of prompts can be customized to individuals, daily
routines may be minimally disrupted. Additionally, because survey windows are typically
programmed to be limited in nature (e.g., to ‘time-out’ within several minutes) the elapsed time
between the sleep event and sleep report are kept at a minimum, thereby reducing recall bias.
(Shiffman et al., 2008)
EMA is appealing for sleep assessment within the context of interventions that require
real-time data: EMA sleep reports can be uploaded to the cloud in real-time and used to inform
interventions that seek to incorporate within-subject sleep changes. (Nahum-Shani et al., 2016)
Most compellingly, day-to-day EMA sleep assessment can be paired with within-day EMA
assessment of relevant processes that may act as antecedents to or consequences of sleep health.
For example, correlates of sleep such as affective states, perceived stress, social, and
environmental context can be easily queried through EMA and linked to day-level sleep.
(Shiffman et al., 2008) Despite its numerous advantages, EMA methodologies are primarily
limited by subjective (vs. objective) sleep report. And importantly, while EMA methods are
acceptable for use in youth ages 7 and above, (Heron, Everhart, McHale, & Smyth, 2017) and
show moderate (78%) compliance (e.g., the percentage of prompts to which participant
responded on average), (Wen, Schneider, Stone, & Spruijt-Metz, 2017) the acceptability of EMA

34
for repeated assessment of sleep health variables in youth is yet unexamined.
Previous Studies Comparing Agreement of Actigraphy with Self-Report Sleep

Agreement of within-subject mean. Although no studies to date have examined the
agreement of EMA with actigraphy for youth sleep assessment, there have been numerous
studies comparing the agreement of actigraphy with paper sleep logs among youth; this literature
provides insight into the expected agreement for actigraphy with EMA assessment of children’s
usual (i.e., within-subject mean) sleep health. Overall, these studies have found good agreement
of paper sleep logs with actigraphy for estimates of within-subject mean sleep timing (i.e., bed
and wake times), (Tremaine et al., 2010; Werner et al., 2008) but lower agreement for sleep
quality measures (e.g., sleep efficiency or night awakenings). (Sadeh, 2011) Additionally,
several studies have observed a tendency for sleep logs to overestimate overall sleep duration
compared to actigraphy. (Short, Gradisar, Lack, Wright, & Carskadon, 2012; Tremaine et al.,
2010; Werner et al., 2008) Due to the general similarity between EMA and sleep logs (i.e., both
self-report daily measures), it is likely that comparison of EMA with actigraphy for sleep
assessment would yield similar overall trends. However, features of EMA such as the prompting
scheme (e.g., reports occur closer in time to actual sleep events compared to sleep logs), limited
reporting window (e.g., participants are not able to back-fill entries in EMA, which is possible
with sleep logs), and lower burden (e.g., participants respond to brief daily survey vs. complete
paper logs) may yield estimates of sleep health that are more in agreement with actigraphy.
Agreement of within-subject variability and day-to-day levels. Furthermore, beyond
agreement of within-subject mean levels, there are few existing studies that assess the agreement
of within-subject variability (i.e., standard deviation of the average across all days) or the day-
level agreement (i.e., how sleep levels on any given night agree between two methods) of
repeatedly assessed sleep across methods. Despite the importance of repeated day-to-day sleep

35
assessment for determining an individuals’ degree of within-subject variability in sleep, which is
a unique contributor to several health outcomes, (Bei et al., 2016) the vast majority of studies
aggregate repeated sleep assessments reports in favor of calculating the within-subject mean (i.e.,
average across all days) value of sleep variables across observations. (Weiss, Johnson, Berger, &
Redline, 2010; Werner et al., 2008) Although five days of measurement is recommended to
obtain an accurate estimate of an individual’s usual sleep (Acebo et al., 1999) in a variety of
research contexts, the actual levels of sleep on a given day, or the overall degree of within-
subject variability across days may be the parameter of interest. (Becker, Sidol, Van Dyk,
Epstein, & Beebe, 2017; Bei et al., 2017, 2016) To our knowledge, no study has examined the
agreement of within-subject sleep variability estimates across methods. Only one study has
examined day-to-day agreement, however, this study used daily paper sleep logs and actigraphy,
measured only two consecutive nights among in older adults with insomnia and chronic
musculoskeletal pain. (Wilson, Watson, & Currie, 1998) To date, no studies have examined the
within-subject mean or day-to-day agreement of EMA with actigraphy among youth.
Characteristics associated with agreement. Previous studies in youth that compared the
agreement of paper sleep logs with actigraphy for within-subject mean levels have identified
several participant and temporal characteristics that may affect the degree of agreement across
methods. For example, previous studies have found lower agreement between sleep log and
actigraphy for estimates of WASO and duration in males (vs. females), which is attributed to
greater nighttime movement and misclassification of these movements as ‘awake’ by the
actigraphy scoring algorithm. (Short et al., 2012) Additionally, pubertal development is
accompanied by a number of physiological and social changes known to impact sleep health,
leading to delayed sleep onset and wake times, shorter sleep duration, and more irregular sleep
patterns. (Sadeh et al., 2009) Likely due to the rapid changes occurring during this time, and the

36
potentially diminished correspondence between perceived vs. objective sleep health, later stage
of pubertal development has been associated with lower agreement of self-report sleep log with
actigraphy sleep. (Short et al., 2012) A child’s body mass index (BMI, kg/m
2
) may also affect the
agreement of sleep logs with actigraphy, due to a higher likelihood of sleep disordered breathing,
and poorer objective sleep quality within this population, (Vgontzas, Bixler, & Basta, 2010)
which may not be salient via self-report. Finally, children’s sleep patterns tend to differ on
weekend days as compared to weekdays, and previous studies have found higher agreement for
between sleep logs and actigraphy for weekend days (vs. weekdays). (Tremaine et al., 2010)
However, as all of these previous studies have examined the agreement of within-subject mean
sleep levels, it is unexplored how these characteristics may moderate the day-level agreement of
EMA with actigraphy across different sleep variables.
To summarize, in light of the growing use of EMA methods in studies of health behavior
and the importance of sleep for myriad health behaviors and outcomes, (Cappuccio et al., 2010b;
Luyster, Strollo, Zee, & Walsh, 2012) EMA represents a novel yet underexamined tool for
repeatedly assessing sleep in youth. However, it is unknown how EMA estimates of sleep health
agree with estimates from actigraphy, a commonly used and validated tool for youth sleep
assessment. To address the gaps in the literature, the present study measured children’s day-to-
day sleep health across one week with EMA and sleep-log informed actigraphy in order to
determine 1) the agreement of within-subject mean and variability in sleep health variables
obtained from EMA and actigraphy; 2) the day-to-day (i.e., day-level) agreement of EMA with
actigraphy; and 3) the temporal (i.e., weekend) and demographic (i.e., sex, BMI, pubertal
development) factors that moderate day-to-day agreement.
Specific Aims and Hypotheses

1. To test the correlation and agreement between EMA and sleep-log informed actigraphy

37
for measuring within-subject mean and variability (i.e., standard deviation) in sleep (i.e.,
bedtime, wake time, duration, and WASO).
Hypothesis: It was hypothesized that there would be good agreement and a large (r > 0.5)
correlation between EMA and actigraphy for bedtime and wake time and duration, and
poorer agreement for WASO due to underestimation of nocturnal awakenings through
EMA. Agreement of within-subject variability across methods was an exploratory aim.
2. To test the day-to-day (i.e., day-level) agreement of EMA with sleep-log informed
actigraphy sleep across seven days.
Hypothesis: It was hypothesized that the strength of association between EMA and
actigraphy and EMA for day-level sleep would differ by variable, with the strongest
associations for bedtime, wake time, and duration, weaker associations for WASO due to
underestimation of nocturnal awakenings through EMA.
3. To determine the moderating effects of children’s sex, overweight/obesity status, pubertal
development, as well as day of the week on the day-to-day (i.e., day-level) agreement of
EMA with sleep-log informed actigraphy across seven days.
Hypothesis: It was hypothesized that boys, children in later stages of pubertal
development, and children who are overweight or obese would have weaker day-level
agreement between EMA and actigraphy, due to evidence from previous studies
suggesting that these groups have lower agreement of subjective paper sleep log report
with actigraphy. It was also hypothesized that associations will be stronger for weekend
days as compared to weekdays, due to previous literature finding this trend for paper
sleep logs compared to actigraphy.
Methods

A subset of 40 youth participating in the longitudinal Mothers’ and Their Children’s

38
Health (MATCH) Study completed one week of at-home assessment, during which they wore a
wrist-worn actigraphy device and completed daily paper sleep logs (which were used to inform
interpretation of actigraphy) and EMA sleep reports.
Participants

Dyads were recruited to participate in the MATCH Study from schools and community
centers in greater Los Angeles County. Recruitment occurred on a rolling basis, with the first
group of dyads enrolling in the fall of 2014, and the final batch of dyads enrolling in the spring
of 2016. A total of 202 dyads enrolled in the study and completed the first measurement burst.
MATCH Study inclusion criteria included: a) in 3
rd
– 6
th
grade at the time of enrollment (child);
b) reside together at least 50% of time (mother and child); and c) ability to speak and read in
English or Spanish (mother and child). Study exclusion criteria were: a) use of medication for
thyroid or psychological condition (mother); b) a health condition limiting ability to be
physically active (mother or child); c) enrolled in a special education program (child); d)
currently using oral or inhalant corticosteroids for asthma (mother or child); e) pregnancy
(mother); f) underweight (BMI < 5
th
% for age and sex) (child); and g) working more than two
evenings (between 5-9pm) during the week or more than one 8-hour weekend shift (mother). All
MATCH children were eligible to enroll in the Sleep Sub-study.
Procedures

MATCH dyads completed measurement bursts approximately every six months for three
years, for a total of six semi-annual assessments (‘bursts’). At each burst, mothers and children
attended an in-person visit, during which they completed paper and pencil surveys and received
written and verbal instructions on the proper use of all study devices. Children were trained on
use of the study EMA smartphone application (‘app’), each on their own device. Participants
who owned their own Android compatible phone are asked to download the study app on their

39
own phone, and participants without a phone or with a non-Android phone borrowed a MotoG
phone (Motorola Mobility, USA) for use during the study period. Upon leaving the visit,
children completed eight consecutive days of ecological momentary assessment (EMA) surveys,
including daily sleep questions.
In addition to the main MATCH Study procedures, a subset (n= 40) of children enrolled
in the Sleep Sub-Study, which could coincide with any MATCH Study measurement burst. Sub-
study participants were recruited in the fall of 2017, until the target of n=40 children was
reached. Sub-study participation did not coincide with major school holidays, vacations, or clock
changes (e.g., Daylight Savings time). Dyads were notified of the sub-study during the
scheduling phone call or email for an upcoming in-person study visit. At the in-person visit,
interested dyads provided parental consent (mothers) and written assent (youth). Following
informed consent, the study coordinator fit the participant with an Actigraph accelerometer, on
an adjustable real-time sensor Velcro wristband. Participant were instructed to wear the device
on their non-dominant wrist.
Participants were provided with a sleep log booklet with eight sheets and instructed to
complete a daily entry each day for one week; the first entry was completed during the in-person
visit with the study coordinator, providing the opportunity for the participant to ask for
clarification as needed. Participants were instructed to wear the wrist monitor at all times during
the following week, except when bathing, showering, swimming, or engaging in hard contact
sports (e.g., football), and participants were provided with written instructions and answers to
common questions. Upon drop-off at the end of the week, the accelerometer and the sleep log
were collected. Participants with ≥ 80% compliance with the wrist-worn accelerometer received
$40, and those with < 80% compliance received $20.

40
Measures

EMA sleep. Participants completed up to 7 days of EMA sleep reports, from Day 2 - Day
8 of the study. Figure 3 displays screenshots of the EMA sleep items. Sleep questions were
included in the first answered prompt of each day; on weekend days, the first survey occurred
between 7:30-8:00am, and on weekdays the first survey of the day occurred between 3:30-
4:00pm (after school). If the sleep items were skipped or incomplete, they were repeated at
subsequent prompting windows, giving participants multiple opportunities to respond. The EMA
items asked children to report on their previous night’s sleep each day, including “What time did
you fall asleep last night?”, “What time did you wake up this morning?”, and “How many times
did you wake up during the night? (i.e., 0, 1, 2, 3, 4, 5-8, 9+).” Responses were used to denote 1)
bedtime (i.e. hh:mm); 2) wake time (i.e. hh:mm); 3) sleep duration (i.e., the elapsed time
between reported bed and wake time); and 4) estimated wake after sleep onset (i.e., WASO). To
estimate WASO, the question regarding nighttime awakenings was used. This item was re-coded
so that a report of ‘5-8 times’ was equal to 6.5 (the mean) while a response of ‘9+ times’ was
equal to 9 (the other options were not re-coded). Next, the value for night-time awakenings was
multiplied by 30 minutes to estimate total nightly WASO. This indirect approach for estimating
WASO in the absence of a direct EMA WASO item was selected based on findings from a study
employing concurrent sleep log and actigraphy that found that mean sleep diary-reported
nighttime awakening was 2 minutes, while mean actigraphy WASO was 60 minutes,
approximately 30 times higher. (Tremaine et al., 2010)
Sleep logs. Participants also completed 8 days of standard sleep logs, which were used to
guide the analysis of actigraphy data. Participants were instructed to respond to the following
questions each morning upon awakening: “What time did you try to fall asleep last night? The
time you closed your eyes and tried to fall asleep”, and “What time did you first wake up this

41
morning?” These responses were used to guide the determination of ‘In Bed’ and ‘Out of Bed’
markers within the ActiLife software.
Actigraphy. Children completed up to 7 nights of actigraphy sleep assessment, on Days
1 - 8 of the study, using an Actigraph GT3x accelerometer (Firmware v1.5.0). Devices were
initialized to begin collecting data at 3:00pm on Day 1 (pick-up appointment), and to stop data
collection at 3:00pm on Day 8 (drop-off appointment). Devices were initialized to record data at
100Hz, at 15-second epochs, with the ‘Idle Sleep Mode’ setting enabled. Actigraphy data were
processed using the Cole-Kripke (CK) algorithm (Cole, Kripke, Mullaney, & Gillin, 1992) in the
ActiLife 6 Software (Actigraph, Pensacola, FL). A proprietary algorithm was applied to the raw
data (in Hz) to convert each epoch to activity count units, then a shifting window approach was
used to classify each 60-second epoch as either awake or asleep based on activity using a shifting
window approach. Although the CK algorithm was originally validated in an adult population,
the CK algorithm was selected for use in this study due to growing empirical evidence that the
CK algorithm has higher accuracy in comparison to PSG than other available algorithms, for
total sleep time (Meltzer, Walsh, Traylor, & Westin, 2012) and WASO (Quante, Kaplan, Cailler,
Rueschman, Wang, Weng, et al., 2018) for both youth and adults. Sleep log data were used to
assist in assigning the ‘in bed’ and ‘out of bed’ time markers when possible, thus the Actigraphy
sleep data is considered to be sleep-log informed actigraphy. Actigraphy sleep were screened for
nights of non-wear, and invalid nights (i.e., nights with complete non-wear or with only partial
wear) were visually identified by examining the activity counts for consistent 0 activity counts
and excluded. To correspond to the EMA derived sleep variables, the following actigraphy sleep
variables were used to denote: 1) bedtime (i.e., the ActiLife ‘in bed time’ variable); 2) wake time
(i.e., the ActiLife ‘out bed time’ variable); 3) duration (i.e., the ActiLife ‘total minutes in bed’
variable); and 4) WASO (i.e., the ActiLife ‘WASO’ variable, indicating the total number of

42
minutes scored as awake after sleep onset occurred). (ActiLife, 2012)
Anthropometric measures. Children’s weight and height were assessed using a digital
scale (Tanita WB-110A) and stadiometer. Anthropometric measurements were taken in duplicate
and averaged when discrepant. Body mass index (BMI; kg/m
2
) and weight classification were
calculated using EpiInfo 2005, Version 3.2 (CDC, Atlanta, GA). A dichotomous variable
overweight/obese (=1) vs. normal weight (=0) was used in analyses.
Demographic measures. Children completed surveys reporting on their sex, age and
pubertal development. Pubertal development was measured using the Pubertal Development
Scale (PDS), a 5-item, sex-specific self-report scale, which assessed the presence of physical
changes associated with puberty in various domains. (Petersen, Crockett, Richards, & Boxer,
1988) Scores across items were summed to assign the PDS category, which ranged from 1 (pre-
pubertal) – 5 (post-pubertal). At each measurement burst, participants’ mothers completed
surveys reporting on their marital status (married vs. non-married), number of individuals living
in the household (continuous), annual household income (<$35,000, $35,001-$74,999, $75,000-
$104,999, and $105,000+), maternal education level (college vs. no college) as well as their
child’s race (white vs. non-white) and ethnicity (Hispanic vs. non-Hispanic).
Statistical Analysis

The analytical sample included all sub-study observation days. Prior to analysis, EMA
and actigraphy data were merged at the day-level. In order to accommodate observations in
which bedtime occurred after midnight, data were merged by wake-up date (i.e., by linking the
actigraphy ‘out-of-bed’ date with the EMA survey completion date). Anthropometric and
demographic covariates were then merged by participant ID. All clock times were converted so
that bedtimes occurring at midnight or later were re-coded with an extra 24 hours; thus, later
bedtimes were increasingly larger values (e.g., 1:00 am becomes 25:00). Additionally, bedtime

43
and waketime minutes were re-coded to represent the proportion of time out of 60 minutes (e.g.,
7:30am becomes 7.5, and 23:05 becomes 23.12). Within-subject mean and variability (i.e.,
standard deviation of the mean) were calculated for each of the four sleep variables for EMA and
actigraphy. All analyses were conducted using SAS v. 9.4 (SAS Institute, Cary, NC). Statistical
significance was determined at p <0.05.
Research Question 1. To test the correlation and agreement between EMA and sleep-log
informed actigraphy for measuring within-subject mean and variability (i.e., standard deviation)
in sleep (i.e., bedtime, wake time, duration, and WASO).
Analyses tested the agreement across measurement methods for estimates of 1) within-
subject mean; and 2) within-subject variability in a) bedtime; b) wake time; c) duration; and d)
WASO. First, descriptive analyses were conducted to examine the group-level mean and
standard deviation of each sleep health variable for EMA and actigraphy. Descriptive statistics
were also calculated separately by weekend vs. weekday in order to determine differences by day
of the week. Paired t-tests were used to test for differences in EMA vs. actigraphy for estimates
of both (1) within-subject mean and (2) within-subject variability for each sleep variable.
Pearson’s correlation was calculated to determine the linear associations (r) between EMA and
actigraphy for within-subject mean and variability for each sleep variable. Correlations were
interpreted as high (r= 0.70 – 1.00), moderate (r= 0.50 – 0.70), low (r= 0.30 – 0.50) and
negligible (r= 0.00 – 0.30). (Mukaka, 2012) Bland-Altman plots were constructed to visually
represent the level of agreement of within-subject mean values for EMA compared to actigraphy
for each variable. (Bland & Altman, 1999)
Research Question 2. To test the day-to-day (i.e., day-level) agreement of EMA with
sleep-log informed actigraphy sleep across seven days.
To test the agreement of day-to-day sleep values reported in EMA as compared to

44
actigraphy, a series of multilevel linear regression models were tested using SAS PROC
MIXED. The purpose of these analyses was to determine the effect of day-level deviations from
an individual’s usual actigraphy-assessed sleep (controlling for the group mean centered value
for the actigraphy sleep variable) on that individual’s EMA-reported sleep for that same day. The
analytical sample included all days for which a participant had an available value of a given
sleep variable for both EMA and actigraphy. Separate models were run for each of the sleep
variables (i.e., bedtime, wake time, duration, WASO). Models tested the day-level associations
between WS effects (i.e., the deviation from each subject’s mean level of a given sleep variable)
measured via actigraphy with EMA report of the same sleep variable for the same day. Models
controlled for the BS effect (i.e., the group mean centered value for each actigraphy sleep
variable) and adjusted for the clustering of days (Level 1) within individuals (Level 2).
Research Question 3. To determine the moderating effects of children’s sex,
overweight/obesity status, pubertal development status, and day of the week on the day-to-day
(i.e., day-level) agreement of EMA with sleep-log informed actigraphy across seven days.
To test the effect of demographic (i.e., male vs. female sex, pubertal development
category, overweight/obese vs. normal weight) and temporal characteristics (i.e., weekend vs.
weekday) in the association between WS effects (i.e., deviation from each subject’s mean level
of a given sleep variable) measured via actigraphy with EMA report of the same sleep variable
for the same day, models from RQ2 were re-run with the consecutive addition of each covariate
as well as an interaction term (e.g., product) of each covariate with the WS term for each sleep
health variable. Each covariate (and its interaction with the WS sleep term) was tested in a
separate model.
Results

Participants consisted of 40 youth (57% female, 68% Hispanic), who were on average

45
11.9 ± 1.1 years at sub-study enrollment. The majority (N=31, 78%) of participants completed
the sub-study during the 6
th
measurement burst of the MATCH Study, while the remainder (N=9,
22%) completed the sub-study during the 5
th
burst. Table 1 shows characteristics for the
MATCH sample, and the sub-study sample both at baseline and at sub-study enrollment.
Data Availability

Of the 40 participants enrolled in the sub-study, one participant did not complete the
study or return the study devices, resulting a total of 39 participants (L2) in the analytical sample.
EMA sleep. All 39 participants provided at least one day of EMA sleep report. Of the 39
participants with any EMA sleep data, the mean number of observation days was 6.0 (SD: 0.89,
range: 4-7) for a mean subject-level compliance rate of 85.71%. Twelve participants (30.8%) had
7 days of data, n=15 (38.5%) had 6 days of data, n=10 participants (25.64%) had 5 days of data,
and n=2 participants (5.13%) had 4 days of data. Overall, of the total possible n=273 possible
observation days (i.e., 7 days x 39 participants), there were n=232 complete, n=9 incomplete,
and n=30 missing observations, for an overall compliance rate of 84.98%.
Actigraphy. Among the 39 participants who completed the sleep sub-study, n=2
participants lost the accelerometer device in the field, and n=1 participant’s device did not record
any data due to a device error (i.e., the device was initialized with the wrong start and end date).
Of the 36 participants with any actigraphy data, the mean number of observation days was 6.2
(SD: 1.4, range: 2-7, for a mean subject-level compliance rate of 88.57%). Twenty-five
participants (69.4%) had 7 days of valid data, n=3 (8.3%) had 6 days, n=3 (8.3%) had 5 days,
n=2 (5.6%) had 4 days, n=2 (5.6%) had 3 days, and n=1 participant (2.8%) had 2 days. Overall,
of the total possible n=273 days (i.e., 7 days x 39 participants), there were n=224 available
observations (no incomplete observations) for an overall compliance rate of 82.05%.

46
Descriptive Results

Means and standard deviations for the four sleep variables (i.e., bedtime, wake time,
duration, and WASO) measured by EMA and actigraphy are presented in Table 2. Compared to
actigraphy, mean EMA bedtime was 12 minutes earlier (22:04 vs. 22:16), wake time was 10
minutes earlier (6:56 vs. 7:06), duration was 15 minutes shorter (532 vs. 547 min), and estimates
of WASO were 42 minutes lower (42 vs. 84).
Table 2 also displays descriptive statistics stratified by day of the week. On average,
participants had later bedtimes and wake times, and longer duration on weekends as compared to
week days, measured through both EMA and actigraphy. Compared to weekdays, on weekend
days (i.e., Friday and Saturday nights), mean EMA bedtime was 36 minutes later (22:27 vs.
21:51), wake time was 73 minutes later (7:43 vs. 6:30), duration was 36 minutes longer (556 vs.
520). This weekend vs. weekday trend was not observed for WASO, which was similar in
weekdays and weekend days. A similar trend of later bed and wake times, and longer duration
was also observed in actigraphy sleep. Additionally, both EMA and actigraphy-assessed within-
subject variability appears to be greater on weekend days as compared to week days.
Agreement of EMA with Actigraphy for Within-Subject Mean and Variability in Sleep

Correlation. Correlations within-subject mean and variability for each sleep variable
from EMA and actigraphy for the are shown in Table 3. Correlations for within-subject mean
values were moderate-to-high and significant (Mukaka, 2012) for bedtime, waketime, and
duration (r= 0.63 – 0.96, p’s <0.001), but negligible for estimates of WASO (r= 0.21).
Correlations for within-subject variability in sleep between EMA and actigraphy ranged from
negligible to high, with highest correlation for bedtime variability (r= 0.70, p< 0.01).
Paired t-tests. Results for paired t-tests are shown in Table 4. There were no significant
differences between EMA and actigraphy for estimates of within-subject mean wake time or

47
sleep duration (p’s > 0.05). Compared to actigraphy, within-subject mean EMA bedtime was 13
minutes earlier (p= 0.015), while mean EMA WASO was 44.2 minutes lower (p< 0.0001).
Paired t-tests for estimates of within-subject variability revealed no significant differences
between EMA and actigraphy for bedtime, wake time or duration, however mean within-subject
variability estimate for WASO was 24 minutes lower in EMA (p< 0.001).
Bland-Altman plots. Bland-Altman plots were constructed to visualize the agreement of
EMA with actigraphy, as displayed in Figure 4. The mean of EMA and actigraphy assessed
values for each sleep variable is plotted on the X-axis, while the mean difference between EMA
and actigraphy for each sleep variable is plotted on the Y-axis. The line at Y= 0 represents the
line of equality, on which all points would rest if there were no differences in values obtained
from the two assessment methods, and the lines above and below the Y= 0 line indicate the
(upper and lower) 95% limits of agreement. (Bland & Altman, 1999) While these plots show a
wide range of mean differences, the majority of points lie within the 95% limits of agreement.
For WASO, there is a clear negative bias for EMA vs. actigraphy (i.e., EMA systematically
produces lower estimates of WASO compared to actigraphy, which increases at higher values).
Agreement of EMA with Actigraphy for Day-Level Sleep

Table 5 displays the results of multilevel linear regression models predicting day-level
EMA sleep as a function of WS deviation in actigraphy measured sleep. There was a significant
positive association of the WS deviation in actigraphy sleep with the EMA reported sleep level
for that same variable (e.g., actigraphy WS bedtime à EMA bedtime) across all four sleep
variables (p’s< 0.001). This finding indicates that, controlling for usual actigraphy-assessed sleep
levels, on days when actigraphy detected higher (or later, in the case of bed and wake time) level
than a participant’s own mean level for a given sleep variable, their EMA report of the same that
variable was also significantly higher. To illustrate, controlling for the group-mean centered

48
wake time, on nights when participant’s actigraphy-assessed wake time was one unit (one hour)
later than their own usual, their EMA-reported wake time for that same day was 0.81 units (0.81
x 60 min = 49 minutes) later.
Moderators of the day-level agreement of EMA with actigraphy. Table 5 also
displays the moderation effect of demographic, anthropometric, and temporal covariates. Results
revealed a significant interaction effect of both pubertal development category (bint= - 0.12 (SE:
0.04), p< 0.01) and male sex (bint= 0.28 (SE: 0.08), p< 0.01) with actigraphy-assessed WS
WASO on EMA-assessed WASO. These findings indicate that the day-level agreement of WS
actigraphy WASO is weaker for participants who are later in their pubertal development (vs.
those in earlier stages), but stronger for males (vs. females).
Discussion

With growing recognition of the importance of sleep for cognitive, mental, and physical
health, it’s increasingly important to develop and validate novel sleep assessment tools for
repeated day-to-day sleep assessment that both maximizes resources and minimizes burden. This
study was the first to compare EMA with actigraphy sleep assessment in a sample of youth
across one week. Using several statistical approaches, this study examined the agreement of
EMA with actigraphy for four key indices of sleep health – bedtime, wake time, duration, and
WASO, for estimating three distinct sleep conceptualizations – within-subject mean (i.e., usual)
sleep, within-subject variability (i.e., the degree of variability within an individual across nights)
in sleep, and day-to-day (i.e., day-level) sleep levels. Results revealed that EMA sleep has
generally moderate-to-high agreement with actigraphy for estimates of within-subject mean and
variability for all variables but WASO, as well as significant day-to-day agreement between
methods that generally did not differ by day of week or other participant characteristics. Results
suggest that EMA can be used to capture children’s day-level and usual levels of sleep health

49
within their natural environments.
Participants in our study had an EMA reported sleep duration of 532 ± 65 minutes, which
is equivalent to 8 hours and 52 minutes. This suggests that children were just short of meeting
age-specific sleep recommendations on average (i.e., at least 9hrs for youth through age 13, and
at least 8hrs for youth ages 14 and above). (Hirshkowitz et al., 2015) Within-subject mean values
from both EMA and actigraphy reveal that children displayed different sleep patterns on
weekend days as compared to weekdays. On weekends, children went to sleep about 30 minutes
later, woke up over an hour later and obtained nearly 45 minutes extra sleep on average. As has
been observed in other studies, (Arora, Broglia, Pushpakumar, Lodhi, & Taheri, 2013) children
also displayed greater variability in their sleep patterns on weekend days as compared to week
days, which may reflect more relaxed household rules regarding sleep compared to weekdays.
Overall, this shifting pattern of sleep timing is characteristic of ‘social jetlag’. (Wittmann et al.,
2006) Social jetlag refers to the chronic weekly shifts in sleep patterns observed among
adolescents between school nights (when youth must wake up early) and weekends nights (when
youth can ‘catch up’ on sleep by delaying wake time), which has been found to be associated
with metabolic dysregulation and elevated obesity risk in youth. (Malone et al., 2016; Stoner et
al., 2018)
Agreement of within-subject mean. As hypothesized, the present study revealed
moderate-to-high correlation and good agreement of EMA with actigraphy for assessing within-
subject mean sleep levels in three of the sleep variables (with the exception of WASO). Overall,
our EMA measure showed much higher correlation with actigraphy than what has been observed
in previous studies, even among similar populations. A study by Arora et al. compared
agreement of sleep logs to actigraphy across one week in n=225 youth ages 11-13, and found
low correlation (r= 0.30) of within-subject mean duration across methods. (Arora et al., 2013) In

50
comparison, the current EMA measure was highly correlated with actigraphy for sleep duration
(r=0.71). The high degree of correlation observed for bedtime, waketime, and duration within the
present study suggests that EMA is capturing the same underlying information that is detected
through actigraphy. Additionally, the agreement of within-subject mean levels for sleep duration,
wake time, and bedtime were acceptable (not significantly different). Other studies comparing
agreement of sleep assessments among youth have generally found a tendency for self-report to
overestimate total sleep duration as compared to the objective actigraphy; for example, one study
found overestimation of sleep duration by nearly one hour in sleep logs as compared to
actigraphy. (Arora et al., 2013; Short et al., 2012) In contrast, the current EMA measure was not
significantly different from actigraphy, and in fact underestimated duration by approximately 14
minutes compared to actigraphy. However, although the differences between EMA and
actigraphy for within-subject mean estimation were not significantly different, the observed
mean difference may be clinically or practically meaningful. For example, for a child hovering
around 9 hours of sleep, a 15-minute difference in sleep duration estimate could result in being
classified as not meeting recommendations by EMA estimation, while actigraphy duration would
suggest adequate sleep. Furthermore, EMA sleep assessment may not be suitable for studies that
seek to experimentally manipulate (e.g., extend or curtail sleep) within a home setting, which
require precise estimates of sleep timing.
This study found low correlation (r= 0.21) and large mean difference (- 44.2 min) for
EMA as compared to actigraphy within-subject mean WASO estimates. Because EMA and
actigraphy are assumed be capturing the same underlying process, it is expected that estimates
between sources to be strongly related. For example, if a child reports three nighttime
awakenings on a given night, compared to one nighttime awakening on another night, actigraphy
estimates of WASO should also be higher on the night with three awakenings. Yet, the low

51
correlation suggests that self-report and actigraphy may be capturing different underlying
processes. (Mukaka, 2012) This is likely partially due to the WASO estimation method used in
the current study, in which EMA nighttime awakenings were multiplied by a factor of 30
minutes to estimate total EMA WASO. However, several other studies have found low
correlation of self-report with actigraphy for WASO, including a study in adolescents with a
correlation of r= 0.02 between sleep logs and actigraphy. (Short et al., 2012) Additionally, EMA
produced significantly lower and earlier reports for WASO and bedtime, respectively. That EMA
underestimated WASO as compared to actigraphy is not surprising; it is well known that
actigraphy (as compared to PSG) significantly overestimates WASO, (Quante, Kaplan, Cailler,
Rueschman, Wang, & Redline, 2018) especially among youth. Similarly, other studies in youth
and adolescents have found significantly higher estimates of WASO in actigraphy vs. self-report,
from 67 minutes in adolescents, (Short et al., 2012) to 72 minutes in kindergartners (Werner et
al., 2008) One potential reason for the low overall agreement of EMA with actigraphy for
WASO is that actigraphy detects nocturnal movement as wake time; this may be due to elevated
nocturnal limb movement in this population. Overall, results suggest that WASO is an especially
problematic sleep variable to estimate in a naturalistic setting due to large measurement error
present outside of PSG methods.
Agreement of within-subject variability. A unique strength of the current study is its
focus not only on the agreement of within-subject mean sleep, but also on the agreement of
within-subject variability of sleep variables across methodologies. The ability to capture within-
subject variability in sleep is important, due to emerging evidence for its role in overall health
and well-being. (Becker et al., 2017; Bei et al., 2016) Our analyses revealed a moderate
correlation of within-subject variability for bedtime and duration; however, there were low and
non-significant correlation for wake time and WASO. Additionally, although estimates of

52
within-subject variability for bedtime, waketime, and duration were not significantly different
between EMA and actigraphy, for WASO, EMA produced significantly lower estimates of
within-subject variability compared to actigraphy. The difference between methods for within-
subject variability in WASO is likely due to both (1) the construction of the EMA WASO
variable, and (2) the known tendency for actigraphy to overestimate WASO as compared to the
ground truth (i.e., PSG). Interestingly, compared to actigraphy, EMA produced (non-
significantly) lower within-subject variability estimates for bedtime, and higher estimates of
within-subject variability for wake time. This pattern of findings may be attributed to children’s
tendency to sleep later on weekends, when they are able to delay waketime. While EMA report
would capture the self-reported wake-up time (the time the child gets up for the day with the
intention of remaining out of bed), actigraphy might be more sensitive in capturing increased
morning movement, such as moving around in bed, which would be classified as time awake.
This hypothesized discrepancy would be expected to produce higher estimates of within-subject
variability in waketime for EMA compared to actigraphy. Overall, the observed moderate
correlation and good agreement suggest that EMA has similar psychometric properties with
actigraphy and is able to capture the magnitude of day-to-day variability in sleep levels.
Day-to-day agreement. In multilevel models, there was a positive association between
within-subject deviations in actigraphy sleep with EMA sleep for the same day. However, the
effect sizes varied by sleep variable (e.g., effect estimate for waketime was double than that of
bedtime (b= 0.81 vs. b= 0.40), which suggests differing strengths of agreement for day-to-day
actigraphy and EMA sleep assessment by variable. The strongest effect (highest estimate) was
for wake time. This may be due to the shorter delay between actual wake time and EMA report
of wake time; especially on weekend days, children were prompted to report on their wake time
as early as 8:00am, at which time their ability to accurately recall and report on their recent wake

53
time is at its highest due to the recency of awakening. Overall, results from the day-level analysis
show that EMA is in agreement with actigraphy for capturing within-subject, day-to-day change
in sleep. This is useful for studies which seek to tailor interventions based on daily sleep health
(e.g., Just-in-time adaptive interventions [JITA]), (Nahum-Shani et al., 2016) which may seek to
intervene in participant’s health behaviors on days following nights with poorer than usual (e.g.,
later bedtime, shorter duration) sleep.
Moderators of day-to-day agreement. Consistent with hypotheses, the day-level
agreement of WS actigraphy with EMA for WASO was weaker for participants who are later in
their pubertal development (vs. those in earlier stages). Contrary to hypothesis, the agreement for
WASO was stronger for males (vs. females). However, none of the tested factors (i.e., sex,
pubertal development, weight status, weekend) moderated the agreement of EMA with
actigraphy for duration, bedtime, or wake time. Nor did we find any moderation effects for
obesity status or day of the week for agreement of any sleep variable. This lack of moderation
effect suggests that day-to-day actigraphy and EMA report are in agreement regardless of
participant or temporal characteristics. This finding diverges from what has been found in
previous studies comparing sleep log with actigraphy, which determined that females, later
pubertal development, and weekdays had lower agreement rates. However, the present study is
distinct in that it examined the agreement of day-level (vs. within-subject mean) sleep variables.
This lower level of analysis may weaken any broader trends that might be detected if differences
by group were examined in aggregate. Alternatively, it is possible that EMA enabled more
accurate day-to-day sleep report as compared to paper sleep logs, leading to more consistent
overall agreement than what has been observed in previous studies. However, due to the small
sample size for detecting differences at the L2 unit of analysis, future studies should continue to
examine potential moderators of agreement across sleep assessment methods, including whether

54
EMA may be more or less in agreement with actigraphy for certain youth or on certain days.
Limitations

The present study has several strengths. This study successfully assessed youth sleep
across one week with concurrent EMA and actigraphy, with high subject-level compliance (85%
for EMA; 82% for actigraphy). To our knowledge, this study is the first to compare EMA to
actigraphy sleep assessment, and to compare estimates of not only within-subject mean levels,
but also within-subject variability and day-to-day levels of sleep variables. However, there are
also important limitations to note.
Although actigraphy has high validity as compared to PSG (the gold standard sleep
assessment), actigraphy is not considered to be a ‘gold standard’ for sleep assessment. (Ancoli-
Israel et al., 2003) Thus, although this study found overall acceptable agreement of EMA with
actigraphy, it could not determine the overall ‘accuracy’ of EMA, or the overall sensitivity and
specificity of EMA for measuring sleep. Despite this limitation, EMA is strengthened by its
ability to capture naturalistic sleep, in contrast to PSG which may be limited by laboratory
settings. Thus, although real-time data capture approaches such as EMA are not intended to
‘replace’ PSG for sleep assessment, a major advantage of EMA is its ability to capture free-
living sleep habits across repeated nights.
Although the present study revealed generally high correlation for within-subject mean
sleep across methods, this does not preclude a lack of systematic over- or under- estimation in
EMA as compared to actigraphy. (Ancoli-Israel et al., 2003) Indeed, the present analysis
revealed that EMA tended to produce earlier estimates of bed and wake times, and lower
estimates of overall duration and WASO compared to actigraphy. Though not statistically
significant, this may be a clinical or practical limitation depending on the context.
An additional limitation of the present study is the variations across the sleep assessment

55
methods for capturing each sleep variable. For example, the EMA item asked: “What time did
you fall asleep last night?”, the sleep log asked: “What time did you try to fall asleep last night?”,
while the actigraphy ‘in bed time’ variable was used to represent bedtime. Children’s differing
interpretations or heuristics for responding to these items may produce varying degrees of
agreement across methods. Despite the potential differences in items across methods, the overall
agreement was high, which indicates that stronger consistency of wording might have resulted in
even greater agreement in the present analysis. This study used EMA-reported nocturnal
awakenings to estimate WASO, by multiplying the number of night-time awakenings by 30
minutes. The lack of direct question regarding the amount of time participants were awake
during the night is a limitation of the present study that likely contributed to low agreement and
correlation for actigraphy and EMA estimates of WASO at all levels of conceptualization.
An additional limitation in the current study is that asking participants to record their
sleep on paper sleep logs may have artificially improved the agreement of their EMA report with
the sleep-log informed actigraphy. When available, sleep logs were used to guide assignment of
actigraphy ‘in bed’ and ‘out of bed’ times. This a common practice in studies using actigraphy
because it can be difficult to determine the beginning and end of a sleep event due to high
sedentary behavior (low movement) immediately prior to sleeping and after awakening, which
can be difficult to differentiate from sleep. Therefore, it is a limitation that actigraphy values are
influenced by self-report sleep logs, which may be similar to EMA report, leading to higher
agreement than if actigraphy had been automatically processed without any subjective input from
sleep logs. However, because the majority of existing actigraphy literature among youth uses
sleep logs to guide the interpretation of actigraphy, when comparing a new tool such as EMA it
is important to consider it in light of existing methods (i.e., sleep-log informed actigraphy).
The EMA prompting schedule may also be a limitation, in that the time participants first

56
gain access to the survey questions depends based on the day of the week (i.e., 7:30-8:00am on
weekend days, and 3:30-4:00pm on weekdays). While this prompting schedule was designed in
order to capture day-to-day sleep while minimizing participant burden and avoiding prompting
during school time, the difference in recall length on weekdays vs. weekends may result in a
difference in recall bias, with more accurate recall on weekend days due to the earlier prompting
time (shorter delay). However, as the day-level agreement analysis did not reveal a significant
moderating effect of weekend, this does not appear to be significantly impacting the findings.
This study used the Actigraph GT3-x accelerometer device, and the Cole-Kripke
approach for measuring and scoring sleep data. The current results may not be generalizable to
studies using other devices or scoring approaches. Additionally, although the sub-study
population was representative of the larger MATCH Study cohort in terms of key covariates,
participants were not randomly selected to participate and may differ in key ways from the larger
sample. Additionally, findings from the present study may not be generalizable to younger
children or older adolescents, or to children from lower income families (e.g., those below the
poverty line) or children of single mother households or of mothers with lower education levels;
future studies should replicate this study in more representative samples.
Implications

Several large population-based studies (e.g., NHANES, the UK Biobank, Nurses’ Health
Study) have added actigraphy sleep assessment to their protocols, demonstrating increased
interest among the research community to repeatedly asses day-to-day sleep within the natural
environment. The development of reliable, cost-effective sleep assessment methods that pose
low participant burden is of the utmost importance. While actigraphy is considered to be the best
option for measuring free-living sleep, (Jr et al., 2007) actigraphy has important limitations (e.g.,
cost). The present study suggests that EMA is an acceptable alternative for measuring day-to-day

57
sleep health among youth, with comparable estimates of within-subject mean and day-level sleep
characteristics as compared to actigraphy. EMA methods provide an alternative, low-cost, and
low burden methodology for free-living sleep assessment among youth.
There are some methodological implications to consider. The present study has
demonstrated good agreement of EMA with actigraphy for sleep assessment in children.
However, while EMA was not significantly different from actigraphy for measuring within-
subject mean and variability across sleep variables (with the exception of WASO and within-
subject mean bedtime), in some cases the mean differences across methods were relatively large
(e.g., EMA estimates of mean wake time were nearly 10 minutes earlier than actigraphy).
Similarly, while the day-to-day agreement of EMA with actigraphy was positive and significant,
estimates across sources are not identical. Studies that require more precise estimates of sleep
characteristics should consider the relative benefits of EMA as compared to actigraphy. For
example, studies that plan to utilize EMA sleep reports to inform other aspects of their procedure
(e.g., for prompting the collection of a biological sample, such as salivary cortisol) might prefer
to select a method that systematically produces earlier mean estimates of wake time (i.e., EMA),
so as to not miss an important window (e.g., the first 30 minutes upon awakening to capture the
morning cortisol rise and peak). (Adam & Kumari, 2009) When interpreting these results, it is
important to remember that actigraphy is not the gold standard for sleep assessment; thus, while
EMA produces different mean estimates from actigraphy, this does not necessarily imply that
these estimates are more or less ‘accurate’ in capturing the ground truth. In the future, comparing
actigraphy and EMA against PSG would allow researchers to directly compare the relative
accuracy of EMA vs. actigraphy against the gold standard. (Douglas et al., 1992)
The present study also has implications for theory. In light of the growing use of EMA
methods in behavioral health research and the increasing recognition of the negative health

58
effects of high within-subject sleep variability, repeated sleep assessment (as compared to
retrospective recall of usual sleep health) is becoming more and more important. Because sleep
is a high-frequency behavior (i.e., occurs daily), static measurements are unable to fully capture
its complexity. (Dunton, 2018) For example, a single-item survey item querying a child’s usual
sleep duration cannot capture the complexity and variability that is likely present. The current
study demonstrates that EMA is a useful assessment tool for capturing estimates of sleep health
within individuals across time. With increasing interest in the antecedents and consequences of
sleep as it unfolds across days, EMA also provides an ideal setting for assessing other time-
varying constructs related to sleep health (e.g., stress, energy-balance behaviors).
Future Directions

This small study showed that EMA sleep reports display moderate-to-high agreement
with actigraphy across 7 days. An additional priority for future studies is to test the predictive
ability of the various conceptualizations of EMA reported sleep in relation to other psychosocial,
behavioral, and health outcomes. Future studies can assess day-to-day sleep health among youth
and determine whether the relationship of within-subject mean and/or within-subject variability
in EMA sleep health with health outcomes (e.g., obesity status). Most importantly, future studies
can leverage EMA sleep assessment in conjunction with other EMA-assessed temporal factors to
disentangle the within- and between-subject effects of poor sleep health. For example, using
EMA, studies can determine the temporal predictors (e.g., perceived stress) and consequences
(e.g., elevated intake of sweets) of a night with poorer than one’s own usual sleep health (e.g.,
short duration).
Additionally, although it was not done in the current study due to practical and resource
limitations, it may be important to solicit qualitative feedback from participants on their
interpretations of the EMA sleep questions, which may assist with fine-tuning items to better

59
reflect and capture youth sleep health behaviors. (Driscoll, Salib, & Rupert, 2007) For example,
qualitative feedback may reveal that participants interpret the question ‘What time did you go to
sleep last night’ to mean ‘get into bed’, while others interpret the question to mean ‘first shut
your eyes’, while still other participants interpret the question as ‘finally fall asleep’.
Understanding the frame of reference participants use when responding to the EMA sleep items
might lead us to adapt the questions, such as separately assessing the time youth enter their
bedroom for the night, the time they physically get into bed, the time they close their eyes with
the intention of falling to sleep, and the approximate time they actually fell asleep.
Conclusions

The current study found that EMA sleep assessment demonstrates moderate-to-high
agreement with actigraphy for determining within-subject mean and variability for all variables
but WASO, as well as good agreement for day-to-day sleep level that generally did not differ by
day of week or other participant characteristics. Results suggest that EMA can be used to capture
children’s day-level and usual levels of sleep health within their natural environments. Given the
growing recognition of the importance of repeated, daily behavior in overall health status, the
development of new sleep assessment tools, such as EMA, is essential. Future studies should
leverage day-to-day EMA assessment of sleep in conjunction with other temporal factors to
illuminate the within-subject antecedents and consequences of sleep health.

60

Figure 3. Screenshots of the MATCH EMA Sleep Items

61
Table 1. Demographic characteristics for the Mothers' and Their Children's Health Study
(MATCH) Study and for the MATCH Sleep Sub-Study at baseline and at Sub-Study enrollment

MATCH Study Sleep Sub-Study Sleep Sub-Study
At Baseline At Baseline At Study Sub-
Enrollment

(N=202) (N=40) (N=40)
M SD M/N SD/% M/N SD/%
Age (years) 9.6 0.9 9.6 0.9 11.9 1.1
Male 99 49.0 17 42.5 27 42.5
Weight category

Normal weight 125 34.2 18 45.0 18 45.0

Overweight 40 19.9 13 32.5 8 20.0

Obese 34 16.9 9 22.5 14 35.0
Hispanic
a
114 57.0 26 68.4 26 68.4
White race
a
90 45.2 17 44.7 17 44.7
Pubertal development category
a

Pre Pubertal
64 31.8 7 18.0 2 5.7
Early Pubertal
61 30.4 17 43.6 6 17.1
Mid Pubertal
65 32.3 13 33.3 17 48.6
Late Pubertal
7 3.5 2 5.1 3 8.6
Post Pubertal
4 2.0 0 0.0 7 20.0
Mother married
a
136 67.7 23
59.0
23
59.0
Mother college graduate
a
118 59.0 26 66.7 26 66.7
Mother works full-time
a
114 57.6 25 65.8 25 65.8
Household size 4.5
1.5 4.4
1.3
4.2
1.3
Household income
a

≤ $35,000 54 26.9 13 33.4 10 27.0

35,001 - 75,000 60 29.9 10 25.6 8 21.6

75,001 - 105,000 39 19.4 8 20.5 9 24.4

> $105,000 48 23.9 8 20.5 10 27.0
a
Data missing on variable

62
Table 2. Descriptive statistics for EMA and actigraphy sleep assessed for 7-days among N=39
MATCH Sleep Sub-Study participants, overall and stratified by day of week

EMA Actigraphy
M ± SD M ± SD

Overall

Bedtime 22:04 ± 59 22:16 ± 51
Wake time 06:56 ± 34 07:06 ± 39
Duration 532 ± 65 547 ± 48
WASO
a
42 ± 18 84 ± 35

Weekdays
b

Bedtime 21:51 ± 72 22:10 ± 63
Wake time 6:30 ± 61 6:43 ± 54
Duration 520 ± 92 529 ± 68
WASO
a
41 ± 26 76 ± 43

Weekend Days
c

Bedtime 22:27 ± 80 22:35 ± 95
Wake time 7:43 ± 102 8:00 ± 86
Duration 556 ± 117 582 ± 96
WASO
a
43 ± 28 94 ± 51

Note: Bedtime and wake time are in hh:mm. Duration and WASO are in minutes.
a
WASO = wake after sleep onset
b
Weekdays include Sunday - Thursday nights (reporting on Monday - Friday)
c
Weekend days include Friday and Saturday (reporting on Saturday and Sunday)

63
Table 3. Correlations of within-subject mean and variability for sleep variables obtained from 7-
days of EMA and actigraphy among N=39 MATCH Sleep Sub-Study participants

EMA vs. Actigraphy
r
Mean
Bedtime 0.86 ***
Wake time 0.63 ***
Duration 0.71 ***
WASO 0.21

Variability
a

Bedtime 0.70***
Wake time 0.25
Duration 0.44**
WASO 0.08
Note: Duration and WASO are in minutes.
a
Variability refers to the standard deviation of the mean, representing
the degree of within-subject variation in each sleep variable across
nights.
*p<0.05; **p<0.01, ***p<0.001

64
Table 4. Results from paired t-test comparing estimates of within-subject mean and variability
for variables derived from 7-day EMA and actigraphy among N=39 MATCH Sleep Sub-Study
participants

EMA vs. Actigraphy
Mean difference df p
Mean
Bedtime - 00:13 35 0.015
Wake time - 00:09 35 0.082
Duration - 13.8 35 0.090
WASO
a
- 44.2 35 <0.0001

Variability
b

Bedtime - 8.0 35 0.440
Wake time 16.0 35 0.061
Duration 8.5 35 0.190
WASO
a
- 23.6 35 <0.0001
Note: Bedtime and wake time are in hh:mm. Duration and WASO are in minutes.
a
WASO = wake after sleep onset
b
Variability refers to the standard deviation of the mean, representing the degree of
within-subject variation in each sleep variable across nights.

65
Panel 1 Panel 2

Panel 3 Panel 4

Figure 4. Bland-Altman Plots for EMA vs. Actigraphy
Note: Panel 1 displays results for bedtime; Panel 2 displays results for waketime; Panel 3
displays results for duration (min); Panel 4 displays results for WASO (wake after sleep onset;
min).

66
Table 5. Results of multilevel linear regression models predicting daily EMA reported sleep as a
function of WS deviation in actigraphy sleep, and moderation by demographic and temporal
covariates

EMA Bedtime EMA Waketime EMA Duration EMA WASO
a

Level-1 (day) n = 195
Level-2 (individual) N = 36

! SE ! SE ! SE ! SE
Model 1
WS Actigraphy 0.40*** 0.06 0.81*** 0.06 0.62*** 0.07 0.14*** 0.04
Model 2
WS Actigraphy
0.56** 0.21 0.68*** 0.18 0.42 0.25 0.53*** 0.15
PCS
b

0.04 0.09 -0.01 0.06 -4.68 7.14 -2.24 2.23
WS Actigraphy *
PCS
b

-0.03 0.05 0.05 0.05 0.05 0.06 -0.12** 0.04
Model 3

WS Actigraphy
0.38*** 0.07 0.81*** 0.08 0.65*** 0.08 0.05 0.04
Sex
c

0.17 0.19 0.01 0.13 -17.00 17.32 4.92 5.23
WS Actigraphy *
Sex
c

0.08 0.12 0.01 0.13 -0.09 0.15 0.28** 0.08
Model 4

WS Actigraphy
0.42*** 0.08 0.88*** 0.09 0.70*** 0.09 0.16** 0.06
BMI
d

-0.33 0.17 -0.14 0.13 -1.15 16.9 -0.54 5.02
WS Actigraphy *
BMI
d

-0.02 0.11 -0.12 0.12 -0.21 0.15 -0.03 0.08
Model 5

WS Actigraphy
0.44*** 0.08 0.70*** 0.10 0.55*** 0.12 0.13* 0.06
Weekend
e

0.39*** 0.11 0.29 0.17 12.9 11.0 -1.82 3.09
WS Actigraphy *
Weekend
e

-0.17 0.13 0.07 0.15 0.08 0.17 0.02 0.09

Note: Bedtime and wake time are in hh:mm. Duration and WASO are in minutes.
a
WASO = wake after sleep onset
b
PCS = Pubertal Development Score Category
c
Sex is coded as Male=1, Female = 0
d
BMI is coded as Overweight/Obese=1 and normal weight =0
e
Weekend is coded as weekend day=1, weekday= 0
*p<0.05; **p<0.01, ***p<0.001

67
CHAPTER 3: EXPLORING THE DAILY EFFECTS OF SLEEP DURATION
AND BEDTIME ON 24HR RECALL DIETARY QUALITY AND EATING
BEHAVIOR IN YOUTH

Abstract

Introduction: Sleep and obesity are strongly linked in youth. Poor sleep health may increase
obesity risk through effects on weight-related behaviors, including dietary intake. Though
growing research has reported poorer diet in children with insufficient sleep, it is unknown
whether poorer than one’s own usual sleep on a given night affects diet quality and eating
behaviors the next day. This study used Ecological Momentary Assessment (EMA) and 24hr
dietary recalls to examine usual and daily effects of sleep health on children’s dietary intake.
Methods: Children (N= 159, 46% male, age: 9.6 ± 0.9 yrs.) completed up to 6, 8-day
measurement bursts of over three years, using an EMA mobile app to report on sleep health at
the first prompt of each day. At each burst, children also completed up to two days of 24hr
dietary recall, used to calculate daily dietary quality (e.g., Healthy Eating Index [HEI-2015]) and
several indices of eating behavior (e.g., breakfast consumption). A series of multilevel models
examined the between-subject (BS) and within-subject (WS) effects of sleep (i.e., duration,
bedtime), as well as their interactions, on next-day diet and eating outcomes, controlling for sex,
age, burst number, day of the week, and several other covariates.
Results: The analytical sample consisted of n= 772 days (average of 4.9 ± 2.1 days per child).
Mean sleep duration was 9.2 hrs ± 52 min, and mean bedtime was 22:07 ± 53 min. At the WS
level, on nights when children slept longer than their own average, they were nearly one and a
half times more likely to consume breakfast the following day (OR: 1.44 [95% CI: 1.02, 2.02],
p< 0.05). At the BS level, children with longer usual sleep duration than their peers had higher
HEI-2015 Total Scores (b= 2.14 [SE: 0.83], p< 0.01), and children with later usual bedtimes had

68
lower HEI-2015 Total Scores (b= -3.35 [SE: 0.78], p< 0.001), which remained significant after
adjusting for multiple comparisons. In exploratory models combining duration and bedtime,
bedtime (not duration) emerged as a significant BS predictor of HEI-2015 Total Score (b= -3.25
[SE: 0.95], p< 0.001). There was no significant interaction effect of WS bedtime and WS
duration on next-day diet and eating.
Conclusions: This analysis revealed that children with longer sleep duration and earlier bedtimes
had higher dietary quality, with preliminary evidence for a stronger relative effect of bedtime.
While analyses revealed that children are more likely to consume breakfast on days following
longer than usual sleep duration, more research is needed in order to clarify the daily effects of
sleep on diet and eating patterns.

69
Introduction

Poor sleep health and elevated obesity rates are interrelated health issues that have
reached epidemic levels in modern society. Approximately one in three US youth are classified
as overweight or obese (i.e., body mass index [BMI] ≥85
th
and 95
th
percentile), with rates
increasing during the transition from childhood to adolescence. (Ogden et al., 2012) The
psychological and physical health consequences of obesity are severe and, as obesity in youth
tends to persist into adulthood, often lifelong. (Dietz, 1998; A. S. Singh et al., 2008) Similarly,
insufficient sleep among youth is common, and increasingly prevalent. (Matricciani, Olds, &
Petkov, 2012) Despite current guidelines recommending 9-11 hours of sleep for school-age
youth up to 13 years, and 8-10 hours of sleep for adolescents 14-17 years, a substantial
proportion of youth fail to obtain minimum recommended duration, with rates of insufficient
sleep increasing with age. (National Sleep Foundation, 2006)
There is substantial work documenting the link between sleep health and weight status
among youth, which has determined double the of obesity among children who obtain the least
sleep. (Cappuccio et al., 2008; Fatima et al., 2015) Similar effects have emerged for the
detrimental effects of later timing of sleep (i.e., later bedtimes) even when controlling for the
effect of overall duration, suggesting that the timing of sleep within the 24 hour day may also
influence obesity risk. (Arora & Taheri, 2015; Asarnow et al., 2015) In light of the limited to
modest effectiveness of childhood obesity prevention and intervention programs to date,
(Kamath et al., 2008; Sims et al., 2015; Waters et al., 2011) it is important to understand the
potential for sleep health to influence weight gain and obesity outcomes among youth.
The Association of Usual Sleep Health with Dietary Quality and Eating Behavior

A primary pathway through which sleep health may alter obesity risk is alterations in
dietary intake and eating behaviors. Eating behavior is one of the primary contributors to an

70
individual’s weight status. As demonstrated by numerous previous studies, eating is a
multifactorial behavior, influenced by numerous individual (e.g., biological, psychosocial,
behavioral), interpersonal (e.g., family dynamics), environmental (e.g., school and neighborhood
setting) and societal (e.g., norms, advertisements) factors. (Story, Neumark-Sztainer, & French,
2002) These levels of influence interact to impact individual eating behavior and dietary choices,
which may in turn modify or be modified by other behaviors. (Baranowski et al., 2003) For
example, sleep behavior may play a role in subsequent dietary quality and eating behavior, and
indeed, previous studies have found children with poor sleep have worse dietary quality and
eating behavior. (Cespedes et al., 2016; Spaeth et al., 2019)
There are several posited cognitive, physiological, and biological mechanisms which may
explain the link between poor, mis-timed, or insufficient sleep with altered diet quality and
eating behavior, (Chaput, 2014) although the exact mechanisms are a matter of debate. (Felső et
al., 2017) It is important to first note that short sleep duration leads to longer time awake, along
with elevated caloric expenditure and increased caloric needs as compared to a sleeping state (in
which energy expenditure is comparatively lower). (Hanlon et al., 2016) However, despite the
relatively higher caloric need, it has been repeatedly observed that the actual increased energy
intake among individuals experiencing prolonged wakefulness far exceeds what is necessary to
compensate for the extra time awake, perhaps partially due to increased opportunity for eating to
occur. (St-Onge et al., 2011)
Insufficient sleep may be conceptualized as a stressor, in that it contributes to elevated
psychological distress and biological stress response including elevated cortisol. (Vgontzas et al.,
2008) Poor sleep increases the hedonic responsive to food, (McDonald et al., 2015) leads to
greater neuronal sensitivity and response to food cues, (St-Onge et al., 2014) as well as decreased
food-related inhibitory control, (Duraccio et al., 2019) and dietary restraint. (Markwald et al.,

71
2013) Poor sleep also affects the production of appetitive hormones, leading to increased levels
of the appetite-stimulating hormone ghrelin and decreases in the satiety hormone leptin.
(Morselli, Leproult, Balbo, & Spiegel, 2010; Spiegel et al., 2009)
These cognitive, physiological, and biological responses to poor sleep may play a role in
an individual’s diet and eating patterns, leading to positive energy balance, weight gain, and
elevated obesity risk. To date, dozens of studies have provided emerging evidence for an
association between poor sleep health (e.g., shorter sleep duration, later bedtimes) and both
poorer dietary quality and altered eating patterns in children; findings are summarized here.
Dietary quality. Previous observational studies have found that poorer usual sleep health
is associated with indicators of poorer usual dietary quality in children. The poorer overall
dietary quality observed among children with insufficient sleep health is concerning, because a
diet high in fruits, vegetables, whole grains, and limited in added fats, sugars, and empty calories
is essential for optimal health and the prevention of chronic diseases, including cardiovascular
disease, type 2 diabetes, a variety of cancers, (Spring, King, Pagoto, Horn, & Fisher, 2015) and
obesity. (Guo, Warden, Paeratakul, & Bray, 2004; Tande, Magel, & Strand, 2010)
Dietary quality can be conceptualized as (1) the adequate intake of essential, health-
promoting nutrients and foods, and (2) the limiting foods and beverages that are detrimental to
health when consumed in excess. Several studies have documented a link between poor usual
sleep with failure to limit (i.e., excessive intake) high-fat, high-sugar foods among youth.
Specifically, studies have found increased odds of consuming snack foods, (Weiss, Xu, et al.,
2010) fast food, (Kruger et al., 2014) energy-dense foods, (Westerlund et al., 2009) overall fat
intake (Weiss, Xu, et al., 2010) among youth with poorest sleep. Poor sleep has also linked to
insufficient intake of health-promoting foods. For example, several studies in youth have
documented decreased overall consumption of nutrient-dense fruits and vegetables among those

72
with short duration (Kruger et al., 2014; Westerlund et al., 2009) as well as those with late
bedtimes. (Arora & Taheri, 2015)
Limited results from experimental sleep manipulation protocols have shown promising
effects on improving dietary quality. Beebe et al (2013) found that adolescents (ages 14-16 yrs)
who were randomized to a restricted sleep condition of 6.5 hours of sleep for five nights had
elevated overall energy (kcal) intake, and elevated intake of grain-based dessert and sweet foods
as compared to adolescents randomized to the non-restricted 10 hours of sleep condition. In
another study, adolescents who enrolled in a sleep intervention to advance (i.e., make earlier)
bedtimes ate fewer sweet and salty snacks in a ‘snack task’, and consumed foods of a lower
glycemic index at a standardized breakfast. (Asarnow et al., 2017) While the majority of studies
have isolated specific dietary nutrients or food types, some studies have studied the relationship
of poor usual sleep with overall pattern of dietary quality among youth. One cross-sectional
study of n=1,522 European adolescents found that those with ≥ 9 hours of sleep scored
significantly higher on the Diet Quality Index for Adolescents with Meal index (DQI-AM) (a
measure of overall dietary quality), as compared to adolescents with < 8 hours of sleep (i.e.,
DQI-AM of 65.6 vs. 62.05). (Bel et al., 2013)
Eating behavior. While dietary quality refers to the contents of consumed food and
beverages, eating behavior refers to the context and patterns of consumption, including the
overall amount, frequency, and timing of intake across the day. In addition to alterations in
children’s dietary quality, poor sleep health may also influence eating behavior, such as the
timing, frequency, amount, and of intake across the 24hr day. This is important because certain
patterns of eating have been linked to detrimental impacts on cardiometabolic heath, including
impaired glucose tolerance, heightened inflammation (e.g., C-reactive protein), and weight gain.
(St-Onge et al., 2017) According to circadian science, mis-timed eating may be especially

73
detrimental to weight outcomes because eating during an ‘inactive’ phase (i.e., a time of day in
which the body is not prepared to nutrient intake, processing, and nutrient uptake) can result in
desynchrony of one’s internal biological processes from the external environment, leading to
inflammation and oxidative stress. (Foster & Kreitzman, 2017; Kahleova, Lloren, Mashchak,
Hill, & Fraser, 2017) For example, breakfast consumption, longer nocturnal fasting duration (i.e.,
the length of time without food or drink between the last meal of one day and the first meal of
the following day), and consuming more calorically dense meals in the morning as compared to
the evening have all been shown to be beneficial for biological processes and weight outcomes.
(Kahleova et al., 2017; Longo & Panda, 2016)
Increasing evidence from population-based studies has revealed altered patterns of eating
among youth with poorer usual sleep health. For example, children with short usual sleep
duration have higher odds of skipping breakfast, (Gong et al., 2017) while children with later
usual bedtimes have greater frequency of unhealthy snack consumption, (Arora & Taheri, 2015)
and higher odds of excessive caloric intake from snacks. (Weiss, Xu, et al., 2010) Poor sleep is
also associated with increased total caloric intake among children, especially in the evenings
after dinner time (Spaeth et al., 2019). In an experimental study by Hart et al. (2013) children
assigned to decrease their sleep duration consumed significantly more (i.e., 134 kcals) average
daily kilocalories as compared to children in the increased sleep condition; post-hoc analyses
revealed that this elevated intake occurred in the evening, (Hart et al., 2013) which adds further
credence to the notion that poor sleep health (i.e., shorter duration or later timing) may alter
patterns of eating behavior among youth.
Real-Time Data Capture Approaches for Repeated Sleep Assessment

The literature reviewed herein demonstrates a compelling link between poorer sleep
health, lower dietary quality and worse eating patterns among youth. However, it is important to

74
note that all of the studies reviewed in the previous sections have been conducted at the person-
level, examining the effect of children’s usual sleep health with their average dietary quality and
eating behaviors. To date, no known studies have examined the effects of within-subject (WS)
day-to-day deviations from one’s own usual sleep health may on next-day dietary quality or
eating behaviors. This is an important methodological limitation; understanding the varying
levels of association may yield new insights on the role of sleep on children’s diet and eating
behavior.
There is large inter-individual (i.e., between subject) differences in sleep need and
behaviors. Additionally, while an individual’s usual (i.e., average) sleep health across the
lifespan is largely governed by stable biological needs and genetic predispositions, there are
many day-level factors that may influence a given night’s sleep, which can contribute to
significant night-to-night variability. (Dillon et al., 2015; Moore et al., 2011) This variability
across nights cannot be captured by simply asking individuals to retrospectively report on their
usual sleep health over a certain period of time (e.g., the past month). (Bei et al., 2016) Failure to
capture and account for daily fluctuations in both sleep health and eating behaviors may
contribute to the ecological fallacy, in which associations at the between-subjects level differ
from associations observed at a lower level (e.g., at the within-subject level). (Piantadosi et al.,
1988) For example, although individuals with poor usual sleep health have elevated intake of
high-fat, high-sugar (HF-HS) foods, the consumption of these HF-HS foods may not necessarily
follow nights in which that individual experienced poorer than his or her own usual sleep.
Understanding the time scale in which poor sleep may affects dietary intake and eating behavior
is imperative for the successful design and implementation of prevention and intervention
programs that target sleep as a risk factor for obesity.
Ecological momentary assessment, or EMA, is a real-time data capture approach that

75
seeks to repeatedly assess the cognitions, behaviors, and/or contexts of individuals within their
naturalistic environments. (Shiffman et al., 2008) EMA presents a useful tool for the repeated
daily assessment of children’s sleep health characteristics, capturing sleep close to the time that it
occurs, and resulting in less response or recall bias than other methods. Using EMA
methodologies, one can determine both the average, or usual levels of an individual’s sleep
health across repeatedly assessed nights and how it differs from the group mean level (i.e., the
between-subjects [BS] mean), as well as the degree of deviation of a given night’s sleep from
that individual’s own usual level (i.e., the within-subject [WS] mean). (Schwartz & Stone, 1998)
EMA methods are thus tremendously useful for disentangling the BS effects from the WS effects
of sleep on diet and eating, allowing us to determine not only the BS effects of usual sleep health
on usual diet and eating, but also the WS effects of daily deviations from one’s own sleep health
on their next-day diet and eating behavior.
To summarize, although poor sleep health has been linked to poorer overall dietary
quality and eating behavior among youth, the majority of previous studies have been limited by
retrospective assessments or aggregate measures of sleep and eating, which are subject to biases,
and preclude the ability to examine the WS effect of poorer than one’s own usual sleep on next-
day diet and eating behaviors. (Westerlund et al., 2009) Given that one night of altered sleep may
have significant impacts on biological processes linked to obesity including insulin resistance
(Donga et al., 2010) and increased ghrelin production, (Schmid, Hallschmid, & Jauch, 2008)
understanding how poorer than one’s usual sleep may relate to next-day dietary intake and eating
behaviors is of upmost importance, and may provide new insight to obesity prevention and
intervention programs. The present study builds upon the BS literature to explore the daily WS
effects of sleep health on next-day dietary quality and eating behavior in youth.

76
Specific Aims and Hypotheses

4. To test whether, within individuals, a night with poorer than usual sleep (i.e., shorter
duration, later bedtime) affects next-day dietary quality (i.e., HEI-2015 total score, HEI-
2015 Moderation and Adequacy Sub-Scores).
Hypothesis: It was hypothesized that nights with poorer than usual sleep would be
followed by days with lower HEI-2015 total score, as well as lower HEI-2015
Moderation and Adequacy Sub-scores. At the between subject level, it was hypothesized
that individuals with poorer usual sleep (i.e., shorter usual duration, later usual bedtime)
as compared to others would have poorer overall dietary quality.
5. To test whether, within individuals, a night with poorer than usual sleep (i.e., shorter
duration, later bedtime) affects next-day eating behavior (i.e., breakfast consumption,
frequency of eating events, proportion of kilocalories consumed in the evening).
Hypothesis: It was hypothesized that nights with poorer than usual sleep would be
followed by days with lower odds of eating breakfast, greater likelihood of >4 eating
events, and greater likelihood of consuming the largest proportion of kilocalorie intake in
the evening (as compared to morning or mid-day). At the between subject level, it was
hypothesized that individuals with poorer usual sleep as compared to others would have
worse overall eating behavior.
6. To test the independent (i.e., the effect of one sleep health predictor controlling for the
other sleep health predictor) and moderation (i.e. interaction effect) effects of nights with
poorer than usual sleep duration and bedtime on next-day dietary quality and eating
behavior.
Hypothesis: As there are no studies to date examining the within-subject independent or
moderation effects of sleep duration and bedtime on next-day diet and eating, this aim

77
does not have an a priori hypothesis.
Methods

This study used EMA to repeatedly asses sleep and multiple 24hr dietary recalls to
determine dietary quality and eating behavior patterns across multiple measurement bursts
(‘bursts’) of data collection among 8-12-year-old children. Analyses tested the WS and BS
effects of sleep health on next-day dietary quality and eating behavior.
Participants

Participants were drawn from the Mothers’ and their Children’s Health (MATCH) Study,
a longitudinal investigation of maternal and child factors influencing child obesity risk. Mother-
child dyads were recruited from schools and community centers in greater Los Angeles County.
Recruitment occurred on a rolling basis, with the first dyads enrolling in the study in the fall of
2014, and the final dyads enrolling in the spring of 2016. A total of 202 dyads enrolled in the
study and completed the first burst. At the time of enrollment, children were in 3
rd
– 6
th
grade,
spoke and wrote in English or Spanish, and had no heath conditions limiting physical activity or
requiring inhaler. Full inclusion and exclusion criteria as well as detailed information on the
MATCH study design and sample is published elsewhere. (Dunton et al., 2015)
Procedures

At the time of enrollment, children provided assent and mothers provided consent for self
and child to participate. Dyads then completed an 8-day measurement burst approximately every
six months for three years, for a total of six semi-annual assessments (occurring in the spring and
fall academic sessions, with no summer data collection). Although both mothers and children
completed the study procedures, children are the focus of the current analysis. Children were
provided with a study smartphone at each burst and trained on the use of a study phone with the
custom MATCH Study application (‘app’) installed. At each burst, children reported on their

78
previous night’s sleep health characteristics at the first answered EMA prompt of each day.
Additionally, at each burst, children completed up to two 24hr interviewer-assessed dietary
recalls, one for a weekday and one for weekend day. The Institutional Review Boards at the
University of Southern California and Northeastern University approved all aspects of this study.
Measures

EMA sleep. Beginning in the fall of 2015, a set of questions regarding previous night’s
sleep were added to the daily EMA surveys. As the EMA sleep questions were added after the
study had begun, the majority of participants first received these questions in the 2
nd
or 3
rd
burst.
Sleep questions were included in the first answered survey prompt of each day, across study days
2 – 8, for a total 7 assessments per burst across all available bursts. Each day, children reported
on their previous night’s sleep, including 1) bedtime (i.e. hh:mm); and 2) wake time (i.e.
hh:mm). Surveys were prompted in a stratified random sampling scheme; during weekdays, the
first survey of the day occurred between 3:30-4:00pm, and on weekend days, the first survey
occurred between 7:30-8:00am, or later if children pre-programmed a later wake-up time.
Children received an audible notification when it was time to complete a survey and were asked
to pause their current activity in order to respond. Surveys took approximately 2-3 minutes to
complete, and children were unable to access the survey 10 minutes after the original
notification. In order to maximize their ability to report on previous night’s sleep, the sleep
questions were programmed so that if a child ignored the prompt or did not completely respond
to the set of sleep questions at the first prompt of the day, the sleep questions were repeated at
subsequent EMA prompting windows, until they were completed. Among a subset of the study
population, EMA sleep was significantly correlated with concurrently assessed actigraphy.
(O’Connor, in preparation)
Day-level sleep variables for the present study included sleep duration (hh.mm) and

79
bedtime (hh.mm). Sleep duration was calculated as the number of elapsed minutes between self-
reported bedtime and wake-time, and bedtime was the self-reported bedtime. Prior to analyses,
bedtimes occurring after midnight were re-coded by adding 24 to the hours (e.g., 1:30am
becomes 25:30), so that later bedtimes were successively larger numbers Additionally, bedtime
minutes were re-coded to represent the proportion of time out of 60 minutes (e.g., 1:30am
becomes 25.30, and 22:05 becomes 22.12). Data were examined for plausibility and cleaning
rules were applied to remedy common user entry issues (e.g., swapping ‘am’ for ‘pm’ when
reporting waketime). On days when participants received multiple opportunities to report on
their sleep, the first complete set of responses were retained. Day-level EMA sleep data were
screened for implausible or incomplete data, and observations were excluded if they were
implausible or outliers (i.e., within the upper 99% or lowest 1% of observations).
24hr dietary recall. Children completed up to two 24hr dietary recalls at each bust.
Recalls were scheduled so that one was completed for a weekday (Mon – Fri) and one was
completed for a weekend day (Sat, Sun) whenever possible; this was done in order to obtain a
more accurate and representative estimate of children’s usual intake at each measurement burst.
Dietary intake data were collected by trained staff from the Northeastern University Dietary
Assessment Center (DAC) and analyzed using Nutrition Data System for Research software
developed by the Nutrition Coordinating Center (NCC), University of Minnesota, Minneapolis,
MN. (Harnack, 2013) The NDSR software provides detailed nutrient and food group serving
count data, as well as a timestamp for each reported eating event. Children were equipped with
reference guides for portion size estimation, and mothers were asked to be available in order to
assist with the call when needed. (Burrows et al., 2010) In the multiple-pass method, children
were first asked to freely recall what they ate during the course of the previous calendar day, and
in successive reviews of the previous-day’s food list, the interviewer probed for additional

80
details. Dietary recall days were excluded due to technical errors (e.g., mis-labeled ID) or
extreme observation (e.g., daily kilocalorie intake <500 or >4000). (Willett, 1998)
Dietary quality. Using the 24hr dietary recall data, children’s daily dietary quality was
assessed using the Healthy Eating Index 2015 (HEI-2015). The HEI-2015 is a diet quality index
that reflects the extent to which an individual’s diet adheres to the 2015 federal Dietary
Guidelines for Americans (DGA). (Guenther et al., 2014; U.S. Department of Health and Human
Services and US Department of Agriculture, 2015) The HEI-2015 contains 13 dietary component
scores. The range of possible scores for each individual component varies, ranging from 0 – 5
(total fruits, whole fruits, total vegetables, greens and beans, total protein foods, seafood and
plant proteins), or 0-10 (whole grains, dairy, fatty acids, refined grains, sodium, added sugars,
and saturated fats). Each component is categorized into one of two groups: nutritional adequacy
(i.e., dietary components to increase), or moderation (i.e., dietary components to decrease). For
adequacy components, increasing intake results in higher score, to reflect better alignment with
dietary guidelines, whereas for moderation components, decreasing intake reflects higher scores
(i.e., reverse coded). Maximum scores are assigned for meeting the minimum requirement.
The HEI-2015 Total Score was calculated by taking the sum or each of the 13 component
scores. Scores for the HEI-2015 Total Score could range from 0-100. A score of 100 is consistent
with an optimal diet in complete compliance with the dietary guidelines, while a total score of 0
reflects the lowest possible dietary quality in relation to guidelines.
The HEI-2015 Adequacy Sub-Score was calculated by summing the nine adequacy
component scores, including: total vegetables, greens and beans, total fruit, whole fruit, whole
grain, total dairy, total protein, seafood and plant protein, and fatty acid. The Adequacy Sub-
Score could range from 0-60, with higher scores indicating greater adherence to guidelines (i.e.,
sufficient intake of ‘healthy’ foods).

81
The HEI-2015 moderation sub-score was calculated by summing the four moderation
component scores, which included sodium, refined grains, added sugars, and saturated fats. The
HEI-2015 Moderation Sub-Score could range from 0-40, with higher scores indicating better
adherence to guidelines (i.e., lower intake of ‘unhealthy’ foods).
Eating behavior. Primary day-level eating behavior outcome variables include breakfast
consumption, frequency of eating events, and evening overconsumption. (Kahleova et al., 2017)
The eating behavior variables were created using the NDSR nutrients (File 03) and food group
serving counts (File 08) at the meal-level/eating occasion level datasets, which provide time-
stamped details on each eating occasion reported by the child including participant-denoted meal
type, nutrient and food group contents. All eating behavior variables are at the day-level.
Breakfast consumption. Breakfast consumption was defined using the NDSR child-
reported meal type variable (imname), which was self-reported for each reported eating occasion.
Breakfast consumption was equal to 1 on days when participants reported consuming breakfast
(imname=1) and equal to 0 on days when participants did not report consuming breakfast.
Frequency of eating events. The frequency of eating events (intended to capture the
amount or snacking or caloric beverage intake outside of meals) was calculated by summing the
total number of distinct eating events that a participant reported in a given day. Prior to creating
this variable, non-caloric eating events (i.e., events in which only water or non-caloric beverages
were reported) were excluded. A binary variable was created to represent whether or not
participant reported >4 or ≤ 4 eating events that day. This cut-off is based on evidence from a
recent representative study, which found that children consume on average 2-3 snacks per day,
on top of three main meals. (Piernas & Popkin, 2010) Given the ubiquity of snacking among this
population, four or fewer eating events was selected as an indicator of low (i.e., a healthy number
of) eating events, allowing a participant to have reported one snack in addition to their three

82
main meals per day. The binary events variable was equal to 1 on days when children reported
>4 eating events and equal to 0 on days when ≤4 eating events were reported.
Evening overconsumption. Evening overconsumption represents the relative distribution
of energy intake throughout the day. Within the meal-level dataset, each eating event was
categorized as occurring in one of three distinct time windows (Kahleova et al., 2017): early (i.e.,
4:00 – 10:59) midday (i.e., 11:00 – 16:59), or evening (17:00 – 23:59). These windows were
created in order to capture the windows in which breakfast, lunch, and dinner are typically
consumed, along with any snack and beverage intake occurring near primary meals. Although
the vast majority of eating occasions were expected to be consumed between 5:00 – 21:59 (a
span of 17 hours), the ‘early’ and ‘evening’ windows were expanded in order to capture any very
early (4:00 – 4:59) or very late (i.e., 22:00 – 23:59) dietary intake, without inadvertently
including the previous evening’s very late (i.e., 00:00 – 3:59) intake along with the next day’s
‘early’ intake calculation (e.g., to avoid including a late night’s snack consumed at 1:00 and prior
to the nocturnal sleeping period as part of the estimate for that day’s early intake total). Due to
these considerations, the three time windows have slightly differing lengths. After dividing each
day into the three time windows, total day-level kilocalorie intake within each windows was
summed (i.e., early kcal, mid-day kcal, evening kcal). Based on these summed values, an
evening overconsumption variable was created; this variable was equal to 1 if the greatest
proportion of daily kilocalorie intake occurred in the ‘evening’ intake window, and equal to 0 if
the greatest total kilocalorie intake of the day occurred in the ‘morning’ or ‘midday’ windows.
Anthropometric measures. At burst, children’s and mothers’ weight and height were
assessed using a digital scale (Tanita WB-110A) and stadiometer. Measurements were taken in
duplicate and averaged when discrepant. Body mass index (BMI; kg/m
2
) was calculated using
EpiInfo 2005, Version 3.2 (CDC, Atlanta, GA), and BMI categories (normal weight, overweight,

83
obese) were used for data analysis. (Kuczmarski, Ogden, & Guo, 2002)
Demographic measures. At each burst, mothers completed surveys reporting on their
marital status (married vs. non-married), household size (continuous), annual household income
(≤ $35,000; $35,001 to $75,000; $75,0001 to $105,000; >$105,000), employment status (full-
time vs. not full-time), mother’s education level (college vs. no college) as well as their child’s
race (white vs. non-white) and ethnicity (Hispanic vs. non-Hispanic). Children completed
surveys reporting on their sex, age and pubertal development. Pubertal development was
measured using the Pubertal Development Scale (PDS), a 5-item, sex-specific self-report scale,
which assessed the presence of physical changes associated with puberty in various domains.
(Petersen et al., 1988) Responses were used to calculate each child’s Pubertal Development
Category on a 5-point scale from pre-pubertal (1), to mid pubertal (3) to post-pubertal (5).
Statistical Analysis

All analyses were conducted using SAS v. 9.4 (SAS Institute, Cary, NC). Day-level EMA
sleep data and 24hr dietary recall were merged so that the analytical dataset contained all days
for which a child completed both (a) an EMA sleep assessment for the previous night; and (b) a
24hr dietary recall assessment, with each line of data representing a single observation day.
Baseline demographic, anthropometric, and other characteristics were defined as the earliest
available data point per child or dyad (in order to retain as many observations in the final model
as possible) and were merged into the day-level dataset.
Data were analyzed using a multilevel modeling framework, in order to accommodate the
nested structure of the dataset with observation days (Level 1) nested within children (Level 2).
The within-subject (WS; Level 1, day-level) and a between-subject (BS; Level 2, individual-
level) versions of each sleep predictor variable (i.e., duration, bedtime) were entered into each
model in order to partition the variance allowing for the interpretation of WS results while

84
controlling for BS effects. The WS terms were centered on the person-mean, or the participant’s
mean value of each sleep predictor across all bursts and represent the deviation in a given night’s
sleep as compared to that individual’s mean value. The BS terms were grand mean centered,
representing the difference in an individual’s mean sleep duration as compared to the grand (i.e.,
overall group) mean duration. (Curran & Bauer, 2011)
Continuous outcome variables were screened for normality. The following covariates
were included a priori in all models: burst number, weekend (vs. weekday), child age at baseline,
and child sex (male=1). In addition, the following covariates were screened and controlled for
when significantly (p <0.10) associated with the outcome of interest: child ethnicity (Hispanic
vs. non-Hispanic), child race (white vs. non-white), child and mother body mass index category
(underweight, normal weight, overweight, obese) pubertal status (Pubertal Development Scale
category), maternal work status (full-time vs. not full-time) maternal marital status (married vs.
not married), maternal education (college graduate vs. not a college graduate) and household
income (quartiles).
Because each research question tests several independent associations (e.g., multiple
comparisons), there is an increased possibility for Type 1 error, or false positives. In order to
adjust for erroneously significant results, the False Discovery Rate (FDR) was adjusted for using
the Benjamini-Hochberg approach. (Benjamini & Hochberg, 1995) The Benjamini-Hochberg
FDR approach adjusts the p-values for a series of comparisons by calculating a q-value
threshold; the q-value is used to determine the percent of significant results that are false
positives and is used to re-evaluate findings that were significant at the original alpha value in
light of the total number of comparisons tested. P-values are rank-ordered used to determine the
new significance threshold (q-value) using the following equation: "
( $)
≤
$
(
). In this equation,
"
( $)
represents the p-value for the i
th
rank ordered p-value, m represents the total number of

85
comparisons, and ) represents the original significance threshold () =0. 05) . This approach was
used post-hoc within each set of models to control for false positives by applying a more
stringent significance threshold. FDR adjusted significance levels are indicated in tables.
For each research question, a series of models were fit. The unconditional means model
allows for the calculation of the intra‐class correlation (ICC). The ICC, which ranges from 0-1,
represents the percent of total variance in an outcome (Y) that is due to mean differences between
subject (L2). An ICC of 0 indicates that subjects do not differ from each other in terms of mean
values of Y, whereas an ICC of 1 indicates that all of the variability in Y is between subjects.
Generally, in studies of within-subject processes, it is common for ICCs to be moderate (i.e., 0.2
– 0.4) which indicates non-negligible clustering of variance within upper-level (L2) units. To
determine the ICC for a given outcome, the unconditional means model was tested for each
outcome, using SAS PROC MIXED (PROX GLIMMIX for binary outcomes). To illustrate the
unconditional means model with a continuous outcome, .
$/
represents the outcome variable, 0
11

represents the grand mean of the outcome, 2
1/
represents each participant’s deviation from the
grand mean and 3
$/
denotes the error associated with participant j at time i.
.
$/
=0
11
+2
1/
+ 3
$/

Research Question 4. To test whether, within individuals, a night with poorer than usual
sleep (i.e., shorter duration, later bedtime) affects next-day dietary quality (i.e., HEI-2015 total
score, HEI-2015 Moderation and Adequacy Sub-Scores).
PROC MIXED in SAS v. 9.4 (SAS Institute, Cary, NC) was used to test the multilevel
linear associations between sleep and dietary quality. Models tested two predictors (a) sleep
duration, (b) bedtime, and three outcomes: (a) HEI-2015 Total Score; (b) HEI-2015 Adequacy
Sub-Score; and (c) HEI-2015 Moderation Sub-Score. Each combination of predictor and
outcome, with both the WS and BS terms entered simultaneously, were tested individually,

86
resulting in six separate models.
The full model includes the BS and WS terms, a priori covariates, and additional
covariates that are significant at p <0.10. In the below full model, 0
51
represents the BS effect of
the sleep term, and 0
15
represents the WS effect. Additional covariates were screened and added
when significantly associated with the outcome of interest. Post-hoc FDR was used to apply
more stringent criteria for significance and minimize type 1 error.
.
$/
=0
11
+0
15
67338
$/
+0
51
67338
9999999
1/
+0
:1
;<=3 +0
1:
>?@ + 2
1/
+ A
$/

Research Question 5. To test whether, within individuals, a night with poorer than usual
sleep (i.e., shorter duration, later bedtime) affects next-day eating behavior (i.e., breakfast
consumption, frequency of eating events, proportion of kilocalories consumed in the evening).
PROC GLIMMIX in SAS v. 9.4 (SAS Institute, Cary, NC) was used to test the multilevel
associations between sleep health predictor and each binary eating behavior outcome. Models
tested two predictors: (a) sleep duration; and (b) bedtime, with three outcomes: (a) breakfast
consumption (y/n); (b) frequency of eating events (> 4 vs. ≤ 4 eating events, y/n); and (c) and
evening overconsumption (y/n). Each combination of predictor and outcome, with both the WS
and BS terms entered simultaneously, was tested individually, resulting in six separate models.
Post-hoc FDR was used to apply more stringent criteria for significance and minimize type 1
error. Analyses paralleled those described for RQ4, with a logistic outcome, predicting the
likelihood of each eating behavior given the WS, BS, and covariates in the model.
log
8
$/
1− 8
$/
= 0
11
+0
15
67338
$/
+0
51
67338
9999999
1/
+0
:1
;<=3 +0
1:
>?@ + 2
1/
+ A
$/

Research Question 6. To test the independent (i.e., the effect of one sleep health
predictor controlling for the other sleep health predictor) and moderation (i.e. interaction effect)
effects of nights with poorer than usual sleep duration and bedtime on next-day dietary quality
and eating behavior.

87
A series of multilevel linear and logistic regression models were tested, with similar
approach to those described in R4 and RQ5. These combined models entered (1) the BS and WS
terms for both duration and bedtime simultaneously, and (2) an interaction term for the effect of
WS duration*WS bedtime, controlling for the main effects of duration and bedtime. These sets
of predictors were used to predict each diet and eating outcome while controlling for the other
sleep variable and all previously included covariates were tested as an exploratory aim. Post-hoc
FDR was used to apply more stringent criteria for significance and minimize type 1 error.
Results

Data Availability

A total of 202 dyads enrolled in the MATCH Study; retention remained at approximately
80% across the six study bursts.
EMA sleep. Aggregated across all bursts, there were n=7,840 total study observation
days. Of these, n=2,376 (30.61%) observation days occurred prior to the addition of the EMA
sleep questions, resulting in no available sleep data. Of the n=5,464 total observations days that
occurred after the introduction of the EMA sleep items to the protocol n=4,100 days (75.04%)
had at least some available EMA sleep data, while the remaining n=1,364 days (24.96%) were
missing due to non-compliance, meaning that a child received but did not complete any EMA
sleep items for that day. Prior to analysis, an additional n=182 of the n=4,100 available days
were excluded due to missing data on bedtime and duration needed for the present analysis (e.g.,
incomplete responses), and n=78 days were excluded due to extreme sleep duration values of <6
hours (n=38 days) or >13 hours (n=40 days), which represents the upper and lower 5% of the
distribution of all observations. Finally, n=3 observation days that occurred at measurement burst
1 were excluded in order to simplify the models and prevent biased estimates of the intercept due
to non-representative data for this measurement burst. The final level-1 (day-level) sample size

88
was n=3,837 days.
There were no differences in compliance rate by child age, sex, or BMI category (p’s
>0.05). However, the odds of compliance to the daily sleep items increased with each subsequent
measurement burst (OR: 1.29 [95% CI: 1.23 – 1.36], p < 0.001). Additionally, participants were
more likely to comply to EMA sleep report on weekends as compared to weekdays (OR: 2.30
[95% CI: 1.56 – 3.38], p < 0.001).
24hr dietary recalls. In total, children completed 1,529 days of 24hr dietary recall
assessments across the study. Of these, 22 days were excluded due to technical error and 17 days
were excluded due to extreme kilocalorie values, resulting in a total of 1,490 dietary recalls.
Analytic dataset. After merging the EMA sleep dataset with the 24hr dietary dataset, the
final analytical sample consisted of 772 observation days (L1), representing a total of 159
children (L2). The mean number of observations days per child was 4.85 (SD: 2.10), and the L1
sample size ranged from 1-10 days. The number of days at each burst is indicated in Table 7.
Descriptive Results

Demographics. At baseline, children (45.9% male) were 9.6 years (SD: 0.9), with a
range of 8-12 years. The majority of children were Hispanic (59.2%). The majority of children
had mothers who were married (66.5%), college graduates (59.5%), and full-time workers
(61.3%). Full demographic characteristics are presented in Table 6.
Sleep. Taking the mean of each child’s average value across observations (e.g., mean of
means), the average sleep duration was 9.2 hours (SD: 52 min), and the mean bedtime was 22:07
(SD: 53 min). Group-level mean sleep duration is in line with sleep recommendations for
children through age 13 (the age range represented in >95% of all participant measurement
bursts). (Hirshkowitz et al., 2015) Further investigation revealed that children’s mean sleep
duration met age-specific sleep duration recommendations (i.e., at least 9 hrs/night up to 13yrs;

89
at least of 8 hrs/night for 14yrs and up) at 59.87% of all measurement bursts, and that children’s
mean sleep duration at each burst was below recommended duration in the remaining 40.13% of
bursts. There was larger inter-individual (BS) variability in sleep health characteristics, with the
range of average sleep duration spanning from a minimum of 7 hours to a maximum of 12 hours.
Similarly, the range of average bedtimes spanned from an early average bedtime of 19:49 to a
late average bedtime of 01:10. Full sleep characteristics are displayed in Table 7.
Dietary quality and eating behavior. When taking the mean of each child’s average
value across observations (e.g., mean of means), children’s HEI-2015 Total Score was 50.44
(SD: 9.04), which is significantly below the minimum score recommended for health benefits,
which is a score of 80. (Tek et al., 2011; United States Department of Agriculture, 2000)
Additionally, on average across children, the mean kilocalorie intake was 1,733.18 kcal per day,
with the greatest mean kcal consumption (717.86 kcal) occurring in the evening (i.e., 17:00 –
23:59) as compared to morning or mid-day. On average across individuals, children reported
4.28 (SD: 0.98) distinct eating events per day. The majority of children (n=128, 80.5%) reported
consuming breakfast each day, while one child (n=1, 0.63%) reported never consuming
breakfast, and the remaining children (n=30, 18.87%) reported skipping breakfast on at least one
day. In terms of mean number of eating occasions, n=47 (29.56%) had zero days with five or
more eating occasions, while n=13 (8.18%) always reported five or more eating events, and the
remaining children (n=99, 62.26%) displayed variability across days in terms of consuming
greater or less than five meals. Full dietary quality and eating behavior characteristics are
displayed in Table 7.
Preliminary Analyses

When examining the overall correlation of children’s average sleep and diet/eating
values, there was a significant negative correlation between mean sleep duration and mean

90
bedtime (r= -0.48, p < 0.0001), indicating that children with later usual bedtimes tend to have
shorter usual sleep duration. Mean sleep duration was positively correlated with mean HEI-2015
Total Score (r= 0.17, p = .028), and HEI-2015 Moderation Score (r= 0.22, p=0.004), indicating
that children with longer usual sleep duration have higher mean dietary quality, which appears to
be driven by lower intake (lower intake = higher score) of ‘moderation’ foods. Mean bedtime
was significantly negatively correlated with mean HEI-2015 Total Score (r= -0.34, p<0.0001), as
well as both mean HEI-2015 Moderation Score (r= -0.23, p=0.003) and mean HEI-2015
Adequacy Score (r= -0.34, p<0.0001), suggesting that children with later usual bedtimes have
poorer overall dietary quality, driven by both elevated intake of ‘moderation’ foods, and
insufficient intake of ‘adequacy’ foods. There was not a significant correlation of mean sleep
duration or bedtime with average eating behaviors (i.e., breakfast consumption, meal frequency,
evening overconsumption).
Association of Sleep Duration and Bedtime with Dietary Quality

Table 8 shows the results from multilevel linear regression models predicting children’s
dietary quality as a function of their usual and previous night’s sleep duration and bedtime.
Results revealed a BS effect such that children with longer usual sleep duration had higher
overall HEI-2015 Total Scores (b= 2.14 [SE: 0.83], p< 0.01), as well as a higher HEI-2015
Moderation Sub-Scores (b= 1.05, [SE: 0.34], p< 0.01). There was no significant effect of sleep
duration on HEI-2015 Adequacy Sub-Scores, nor were there any significant WS effects of sleep
duration on next-day dietary quality. Findings for sleep timing revealed that children with later
usual bedtimes as compared to other children had lower overall HEI-2015 Total Score (b= -3.325
[SE: 0.78], p< 0.001) and lower HEI-2015 Moderation (b= -1.21 [SE: 0.35], p< 0.001) and HEI-
2015 Adequacy Sub-Scores (b= -2.24, [SE: 0.53], p< 0.001). There were no significant WS
effects of bedtime on next-day dietary quality. All five significant results remained significant

91
after adjustment for multiple comparisons (FDR q= 0.021 for 12 comparisons).
Association of Sleep Duration and Bedtime with Eating Behavior

Table 9 shows the results of multilevel logistic regression models predicting children’s
eating behavior as a function of their usual and previous night’s sleep duration and bedtime. At
the WS level, on nights when children reported sleeping longer than their own average, they
were more likely to consume breakfast the following day (OR: 1.44 [95% CI: 1.02, 2.02], p<
0.05). There were no other significant WS or BS effects observed for sleep duration and number
of eating events or evening overconsumption, nor were there any significant effects of bedtime
on any of the eating behavior outcomes. After adjusting for multiple comparisons, the effect of
WS duration on breakfast was no longer significant (FDR q= 0.004).
Combined Effects of Duration and Bedtime on Dietary Quality and Eating Behavior

Because sleep duration and bedtime are highly interrelated, a combined model was tested
in which both sleep duration and bedtime predictors were simultaneously entered, in order to
examine the relative effects of sleep duration and bedtime on each of the dietary quality and
eating behavior outcomes, controlling for the other sleep health predictor and other covariates.
Findings from these models, shown in Table 10, reveal a BS effect of bedtime on HEI-2015
Total Score (b= -3.25), HEI-2015 Moderation Sub-Score (b = -0.95), and HEI-2015 Adequacy
Sub-Score (b= -2.43), (p’s all <0.001). The WS effect of duration on breakfast consumption also
remained significant. However, when controlling for usual bedtime, there was no longer a
significant effect of usual duration on any of the diet or eating behavior outcomes. Controlling
for multiple comparisons, only the BS effects of bedtime on HEI-2015 Total Score (FDR q=
0.004) and on HEI-2015 Adequacy Score (FDR q= 0.008) remained significant.
Interaction Effects of WS Duration and Bedtime on Dietary Quality and Eating Behavior

In order to determine whether there was a significant interaction effect of WS bedtime

92
and WS duration on diet and eating outcomes, the combined models were tested with the
addition of an interaction term, representing WS duration*WS bedtime. Results (displayed in
Table 11) revealed no significant interaction effect for WS duration and bedtime on outcomes,
controlling for the main effects of the WS and BS terms, as well as covariates (p’s all > 0.05).
Discussion

This study is among the first to examine the WS and BS effects of children’s sleep
duration and timing on dietary quality and eating behaviors, using a longitudinal study design,
with daily EMA assessment of sleep duration and bedtime and multiple 24hr dietary recalls. This
study found strong evidence for an effect of usual sleep duration and bedtime on overall dietary
quality, such that children with longer usual sleep duration and earlier usual bedtimes had
significantly higher dietary quality than other children. Controlling for temporal factors including
day of the week, this study also found evidence for a daily effect of sleep duration on next-day
eating behavior, such that on days when children slept longer than their own usual, they were
more likely to consume breakfast the following day. This is the first known study to examine the
daily and usual effects of sleep on children’s HEI-2015 scores. It is also the first known study to
examine the WS effects of sleep duration and bedtime on the dietary quality and eating behaviors
of healthy children assessed within their naturalistic environments
Role of Usual (BS) Sleep Health on Dietary Quality

HEI-2015 total score. Overall, findings indicate a consistent positive association
between longer sleep duration and earlier usual bedtime with higher dietary quality. The mean
HEI-2015 total score (i.e., 50.68), though similar to nationally representative samples, (Banfield,
Liu, Davis, Chang, & Frazier-Wood, 2016) falls far below the minimum recommended levels of
80, and would in fact be classified as ‘needs improvement’ (i.e., an HEI-2015 between 50-79).
(Tek et al., 2011; United States Department of Agriculture, 2000). Though several factors have

93
been found to influence children’s dietary quality, including demographic factors, and (within
the current sample) maternal stress, (O’Connor, Huh, Schembre, Lopez, & Dunton, 2019) the
present analysis suggests that sleep duration and timing may be represent unique influences for
children’s dietary quality. This is the first known study to examine the relationship between
children’s sleep and their HEI-2015 score. In the current study, children who on average sleep
one hour longer than the group mean have mean dietary quality scores that are 2.14 points higher
than average, while children whose usual bedtime is one hour later than the group mean have
HEI-2015 scores that are 3.35 points lower on average, controlling for covariates. This difference
is meaningful, especially for youth whose HEI-2015 scores are near the threshold between low
and ‘needs improvement’, or between ‘needs improvement’, for whom a 3-point increase could
push their diet from one classification to another. The current study findings are similar to that of
a cross-sectional study of n=1,522 European adolescents, which found higher dietary quality
(DQI-AM) among youth with the longest sleep duration as compared to those with the shortest
sleep duration. (Bel et al., 2013)
HEI-2015 moderation and adequacy sub-score. Whereas bedtime and duration were
both associated with HEI-2015 Total Score in the hypothesized direction, the relative effects of
sleep and duration on the HEI-2015 Sub-Scores revealed a complex pattern of finding. While
usual bedtime was negatively associated with both Moderation and Adequacy Sub-Scores, usual
duration was positively associated with the Moderation Sub-Score only. This finding indicates
that, while both aspects of sleep health are related to overall diet quality, the effect of duration on
quality may be driven primarily through limited (i.e., lower) intake of ‘moderation’ foods-
including added sugars, fats, and sodium, while late bedtimes may impact dietary quality through
elevated intake of ‘moderation’ foods and lower intake of ‘adequacy’ foods. A more consistent
finding for sleep on intake of unhealthy, as compared to healthy foods has been found in several

94
studies. (Spaeth et al., 2019; Tasali et al., 2014)
Moderation. The finding that late bedtimes and short duration are associated with
elevated intake of ‘moderation’ foods suggests that habitual late bedtimes and short duration may
be tied to a pattern of consuming foods and beverages characterized by high sodium, refined
grains, added sugars, and saturated fat. This pattern of eating is in fact common among youth,
whose empty calorie consumption (e.g., grain desserts, sugar sweetened beverages), comprises
40% of their total energy consumption. (Reedy & Krebs-Smith, 2010) Potential explanations
underlying the relationship between poor sleep and elevated intake of high-fat, high-sugar foods
are several: poor usual sleep may activate the hedonic drive to eat, (McDonald et al., 2015)
increase psychological and biological stress, (Vgontzas et al., 2008) and decrease inhibitory
control. (Duraccio et al., 2019) These are all factors which may increase the desire for, and
inability to resist, highly palatable foods and drinks. As has been found in several previous
studies, late bedtimes tend to be associated with elevated caloric intake specifically in the
evening; children with habitually late bedtimes may be more likely to consume high-fat, high-
sugar snack and dessert foods after dinner. (Hart et al., 2013; Spaeth et al., 2019)
Adequacy. Beyond the effect of sleep on elevated intake of ‘moderation’ foods, the
results from the present study also suggest that late usual bedtimes may lead to decreased intake
of ‘adequacy’ foods, including fruits, whole grains, and protein. Previous studies have also
founds that youth with habitually late bedtimes may be less likely to desire healthy foods; one
study enrolled 42 adolescents with late bedtimes into a sleep extension intervention and found
that adolescents who achieved earlier bedtimes at post-treatment had significantly increased
intake of fruit and dairy foods as compared to pre-treatment. (Asarnow et al., 2017) This is
consistent with another study which found early usual bedtimes to be linked to greater intake of
health-promoting foods. (Baron, Reid, Kern, & Zee, 2009) Overall, maintaining sufficient

95
duration and avoiding late bedtimes may be protective for overall dietary quality among youth.
One important consideration for dietary quality is that, although the HEI-2015 is a score
of diet quality intendent of total energy intake, an individual’s overall dietary quality is also
influenced by their total energy intake (i.e., kilocalories). (Guenther et al., 2013) For example,
with increasing energy intake, the more likely a diet is to be proficient in terms of adequacy (e.g.,
sufficient fruits, whole grain intake), while the opposite is true for moderation components, for
which the greater overall caloric intake, the more likely to surpass the limits for moderate intake
(e.g., too much sodium, added sugars). Thus, the effect of overall energy (kcal) intake is
expected to differentially impact the HEI-2015 Adequacy and Moderation Sub-Scores, as higher
Adequacy scores may also lead to higher kilocalorie intake. This is a limitation, in that the
present study did not control for, or independently examine, the effects of sleep on kilocalorie
intake, which is another important dietary characteristic which plays an important role in overall
obesity risk.
Role of Daily (WS) Sleep Health on Dietary Quality

Interestingly, there was a lack of significant associations for a WS effect of sleep duration
or bedtime on dietary quality. The lack of WS findings indicates that while children with poorer
than average sleep have unhealthier diets overall, but that elevated consumption of unhealthy
‘moderation’ foods (and lower consumption of healthy ‘adequacy’ foods) may not necessarily
follow nights in with the individual experiences poorer than usual sleep. For example, while this
study found that youth with later usual bedtimes than their peers had lower HEI-2015 Total
Scores, it was not the case that HEI-2015 Total Scores were lower on days following nights with
shorter than one’s usual duration or later than one’s own usual bedtime.
One potential explanation for the lack of WS effects of sleep on dietary quality is that
short duration and late bedtimes may affect overall dietary quality through cumulative, as

96
opposed to acute, effects. Over repeated, chronic, exposure, poor sleep may contribute to an
overall pattern of unhealthy eating that has been observed in previous studies. Previous
laboratory studies of sleep restriction have found that chronic, sub-sufficient sleep duration has a
cumulative, dose-response effect on overall functioning similar to that of complete sleep
restriction on the immediately following day. (Dongen, Maislin, Mullington, & Dinges, 2003)
This suggests that, while a single night of poor or absent sleep has clear detrimental effects on
next-day processes, the cumulative effect of chronically short or insufficient sleep has a dose-
response relationship on neurobiological (and behavioral) function in healthy individuals.
Though the present study has important differences, including healthy mean levels of sleep at the
group level (e.g., no extreme restriction) results suggest that short usual sleep duration may play
a role in overall dietary quality and eating behavior (breakfast) through cumulative impacts on
behaviors. This idea is also supported by the many existing studies of sleep and diet among
youth to date, which collapse repeatedly assessed sleep to the average level and find that overall
poorer sleep is linked to aspects of diet; these studies may reflect the cumulative effects of poor
sleep across day and their overall effect on diet.
Alternatively, the lack of WS findings for sleep on dietary quality may reflect the
relatively healthy sleep within this population of youth. It is possible that small variations in
sleep health which are still within healthy ranges (i.e., meeting recommendations) may not
significantly impact biological or behavioral processes. For example, one study increased adults’
nightly sleep from 8hr to 9hr and did not detect significant improvements in cognitive
performance. (Belenky et al., 2003) This suggests that there may be a ceiling effect for ability of
sleep health to impact other processes.
Role of Usual (BS) and Daily (WS) Sleep Health on Eating Behavior

Contrary to hypotheses, this study did not reveal a BS effect of poor usual sleep on any of

97
the eating behavior outcomes, including breakfast consumption, frequency of meals, or evening
kilocalorie consumption. Additionally, neither meal frequency nor proportion of calories
consumed in the evening were associated with previous night’s sleep duration or bedtime. This is
in contrast to other studies, which have found a cross-sectional relationship between short usual
sleep duration with increased snacking as well as elevated caloric consumption in the evening
among youth at the person level.
As hypothesized, on nights when children sleep longer duration than their own usual,
they were more likely to consume breakfast the following morning. The finding for an effect of
longer than one’s own usual sleep on higher likelihood of next-day breakfast consumption
suggests that longer sleep duration than one’s own usual may affect morning meal intake.
Children with longer than usual sleep duration may feel more rested and have less sleep debt;
feeling rested can lead to improved decision making, (St-Onge et al., 2014) increasing their
likelihood of engaging in breakfast consumption. Additionally, obtaining longer sleep than usual
may be indicative of greater alignment of internal biological rhythms with the external
environment, leading to greater synchrony of biological hunger signals with external meal times.
(Foster & Kreitzman, 2017) While numerous cross-sectional studies have determined a link
between sufficient sleep duration and regular breakfast consumption, this is the first known study
to WS daily effects of sleep duration on breakfast consumption among youth.
The finding for a role of longer than one’s own sleep duration on breakfast consumption
is important, because both cross-sectional and prospective observational studies have found that
children and adolescents who skip breakfast are at elevated risk of developing obesity.
(Szajewska & Ruszczyński, 2010) The protective effect of breakfast on weight may be partially
attributed to the relationship of breakfast with dietary quality; previous studies have found that
on days that adolescents eat breakfast, they have higher overall dietary quality and intake of fiber

98
and fruits. (Medin, Myhre, Diep, & Andersen, 2019) The time of day may also play an important
role in biological processed underlying obesity and other chronic diseases. The same meal eaten
in the morning as compared to evening leads to more favorable glucose metabolism,
(Jakubowicz, Barnea, Wainstein, & Froy, 2013) due to alignment of eating with the 24hr
organization of the endocrine system. (Zarrinpar, Chaix, & Panda, 2016) Taken together, these
studies suggest that breakfast is protective against obesity due to (1) the tendency for breakfast to
contain ‘adequacy’ food items, such as fruits and whole grains, which are often lacking in
individuals overall diet, and (2) the biological alignment of early morning food consumption
with better metabolic profile and biological response to the food.
Relative and Interaction Effects of Sleep Health on Dietary Quality and Eating Behavior

It is important to consider the interrelated nature of bedtime and duration; in preliminary
descriptive analyses, sleep duration and bedtime were shown to be significantly negatively
correlated (r= -0.60), indicating that nights with later bedtimes have shorter durations. The
negative association of bedtime with duration is likely a product of the generally stable
waketimes among youth, for whom inflexible school start times require early awakenings on
most days. Adolescents commonly extend their sleep (i.e., ‘sleeping in’) on weekends, when
schedules are less constrained by academic demands, in order to compensate for the sleep debt
that accumulated across the week, resulting in ‘social jet-lag’. (Carskadon et al., 1993; Malone et
al., 2016) On non-school days, children generally have greater ability to self-select both bed and
wake times, which often leads to a pattern of delay in both bed and wake time, resulting in a
similar or longer overall sleep duration. Thus, the negative correlation observed between usual
bedtime and usual duration is likely reflecting the curtailed sleep duration that occurs on
weekday nights with late bedtimes, or on weekend nights when children are allowed to stay up
later but are still required to awaken early for sports or other activities. Results from models that

99
included both BS bedtime and BS duration revealed that only BS bedtime remained significantly
associated with dietary quality outcomes. This suggest that while bedtime and duration both are
related to HEI-2015, bedtime may play a relatively stronger role. However, due to the high
correlation between these two indices of sleep health, these findings should be interpreted with
caution. Ultimately, interventions to improve sleep health among youth should seek to extend
sleep duration through encouraging earlier bedtimes.
Limitations

This study has a number of strengths, including a longitudinal design, with daily EMA
sleep assessment and 24hr dietary recalls. However, there are some important limitations to note.
While EMA has poses several advantages over traditional survey approaches, the present study is
still limited by its observational design, and causality cannot be inferred. Additionally, as with
other observational designs, there is a possibility for confounding due to uncontrolled variables.
Although the present study screened for several important covariates related to the association of
interest, residual confounding cannot be ruled out.
A primary limitation of this study is missing data, which may lead to biased effect
estimates if the data are not missing at random. At each measurement burst, children completed
up to two 24hr dietary recalls, one on a weekend day and one on a weekday, in order to better
capture their ‘usual’ intake. Due to scheduling issues, missed calls, and/or other unknown
reasons, not all children completed both recalls at each burst. Additionally, while overall EMA
compliance rate was high (i.e., 75%), children were more likely to report on their sleep on
weekends (vs. weekdays) and during later measurement bursts as compared to earlier bursts.
Thus, results may not be representative of the average week, or the entire study period. The
higher weekend compliance was likely due to the sampling scheme, in which children received
up to 7 prompts/day on weekend days, but only up to 3 prompts/day on weekdays (participants

100
had multiple opportunities to respond to the sleep items). Additionally, it is likely that the
increasing compliance across the study period was due to the late introduction of the EMA sleep
items to the sampling protocol, in which several participants only began receiving the prompts in
the 3
rd
burst or later.
The EMA sleep items are also limited by timing of assessment. The EMA application
assesses sleep in the first answered prompt of each day, however the time of the first prompt
differs on weekdays vs. weekend days. On weekend days, participants reported on their sleep via
EMA as early as 7:30-8:00am, unless a later wake-up time was selected. On weekdays,
participants reported on the previous night’s sleep health beginning at 3:30-4:00pm; this is
because the EMA sampling schedule was designed to avoid time that children are in school.
There are two potential implications for this difference in reporting; first, the difference in recall
length on weekdays vs. weekends may result in a difference in recall bias. Secondly, although on
weekend days the reporting of the previous night’s sleep health is in the early morning, and
therefore unlikely to have occurred before the consumption of any food or beverages, weekday
reporting of sleep health occurred in the afternoon, after a substantial portion of the day’s intake
has already been consumed. It is possible that the foods a participant consumes in the morning
and mid-day may subsequently affect the child’s report of previous night’s sleep (i.e., high
caffeine intake that morning may bias ratings of the previous night’s sleep).
The study design was such that 24hr dietary recalls were collected for one weekend day
and one weekday per measurement burst. Because of this, the included days are biased toward
representing weekend days (52% of all L1 observations). By controlling for day of week
(weekend day vs. week day) in analyses, this study statistically controlled for the potential
confounding effect of weekend. This is important because bedtimes tend to be later while dietary
quality also tends to be lower on weekends compared to weekends. However, it is still possible

101
that the finding for an effect of late bedtimes on poorer dietary quality may be partially due to the
study design. Furthermore, it is possible that participants who provided more weekend
observations might have later overall estimates of usual bedtimes as well as poorer overall
dietary quality. However, it is important to note that there was not a WS effect of shorter than
usual duration or later usual bedtimes on next-day dietary quality; this suggests that, even with
the tendency for weekend days to have both later bedtimes and poorer diet, the between-subject
finding was not simply a result of this weekend effect. However, future studies should assess
sleep and diet across a greater number of days in order to obtain a more representative sample.
Several studies have pointed to the potential bidirectional effects between sleep and diet.
A healthy diet has been shown to promote sufficient sleep duration and quality. (Chaput, 2014)
Particularly, foods promoting the production of serotonin and melatonin and the availability of
tryptophan may promote sleep. (Peuhkuri, Sihvola, & Korpela, 2012) The present study selected
a priori to examine the effect of a given night’s sleep on next-day diet and did not test the effect
of daily dietary quality and eating behavior on the following night’s sleep. However, it is
possible that bi-directional effects may be at play, which may be investigated in future studies.
Implications

Although numerous studies have attempted to intervene on diet as a pathway to improve
obesity outcomes in children, the majority have been ineffective in producing sustained, long-
term improvements in dietary quality and subsequent improvements in weight outcomes.
(Gibson, Peto, Warren, & dos Santos Silva, 2006) This study provides further evidence for an
important role of adequate sleep duration and early bedtimes on overall dietary quality. Findings
suggest that sleep manipulation may have positive impacts on overall diet, potentially including
increased intake of healthy foods, and limiting of high-fat, high-sugar foods.
Even among studies that have repeatedly measured children’s sleep and diet across

102
several days, analyses have aggregated the repeated daily measures to obtain average values of
sleep and diet (i.e., the mean across all measured days). (Hart et al., 2013; Kjeldsen et al., 2014)
In contrast, the current study used repeatedly assessed sleep to calculate both the BS and WS
values and utilized multilevel models to statistically disentangle the within-subject from the
between-subject effects. This is an important methodological strength of the present study, which
led to new insights on the time-scale through which sleep may affect diet. This approach allowed
us to discover that, although usual sleep duration was not related to overall breakfast intake,
breakfast intake was more likely on days following nights with longer than one’s own usual
sleep duration. This suggests a unique window of opportunity for intervening upon children’s
breakfast consumption, which may in turn have positive effects on overall dietary quality.
(Medin et al., 2019)
Findings from this study suggest that sleep health should be considered for inclusion in
family-based obesity prevention and intervention programs. For example, given the strong effect
observed for short duration and late usual bedtimes on overall dietary quality, this study provides
additional support to the growing evidence that insufficient sleep greatly impacts youth well-
being. Families should be made aware of the association of poorer usual sleep health and lower
dietary quality; this may allow families to put measures in place to curb the negative effects of
poor sleep on weight-related behaviors (e.g., limiting snacking and ensuring availability of fresh
fruits and vegetables) in youth. (American Academy of Pediatrics, 2014) Furthermore, the
preliminary evidence for daily concordance of sleep health among parents and their adolescent
children (Fuligni, Tsai, Krull, & Gonzales, 2015) highlights the importance of addressing poor
sleep health within the context of the whole family.
Results are also relevant to public health and educational policy. Schools can play an
important role in children and adolescent’s overall sleep health by delaying school times to no

103
earlier than 8:30 am; districts who have successfully enacted this change have documented
improvements in sleep duration, as well as improvements in academic success and overall well-
being such as alertness, tardiness, truism, graduation rates, psychological health, and safe
driving. (Troxel & Wolfson, 2017; Watson et al., 2017) Thus, in addition to playing a role in
children’s weight-related behavior and obesity risk, improvements in sleep health can impact
numerous other important indices of quality of life and well-being.
Future Directions

The finding of shorter and later timing of sleep with lower overall dietary quality leads to
the question of whether dietary quality may mediate the relationship between sleep health and
obesity risk. To date, there are no known studies examining the either the longitudinal
association between the HEI-2015 and BMI or the mediating role of diet as it relates to HEI-
2015 and BMI in youth, thus understanding the interplay between children’s sleep, dietary
quality, and longitudinal obesity risk is an important area for future research.
Additionally, while this study did not examine potential mechanisms linking sleep with
diet and eating, this is an important area for future studies. Potential mechanisms linking sleep
with diet and eating can be categorized into homeostatic and non-homeostatic drives to eat.
(Chaput, 2014) Short sleep duration is thought to lead to increased total intake (kcal) through
non-homeostatic drive to eat, such as through greater potential time during the day for eating
opportunities, and greater sensitivity to food rewards, which have been well-supported in the
literature. (Chaput, 2014) Additionally, homeostatic mechanisms, which include to hunger
hormones (i.e. leptin, ghrelin, cortisol) may also play an important role in the relationship
between sleep, diet/eating, and obesity. (Broussard et al., 2016; Hart et al., 2013)
Future studies should also consider different time scales of the potential WS effect of
sleep on diet. In the present analysis, analysis was limited to the concurrent day (i.e., effect of

104
sleep on immediately following day of diet), due to evidence that a single night of sleep
restriction alters biological processes the following day. (Donga et al., 2010; Schmid,
Hallschmid, & Jauch, 2008) However, it is possible that the effects of short duration or later
bedtime affect dietary intake and eating patterns at a longer lag, such as two or even several days
later. Future study could test varying time scales beyond one day, or test the cumulative effects
of a given night of poorer than usual sleep on diet and eating behaviors over the span of several
days following poor sleep, to determine whether there is an effect.
Finally, although daily self-report of sleep health via EMA has been found to correlate
with objective monitoring of sleep with wrist-worn actigraphy in children, (O’Connor,
unpublished) future studies should attempt to validate the WS and BS effects of nightly sleep on
next-day diet and eating with objective monitoring. This may reveal important differences in
association, as children’s self-report sleep may be affected by their subjective appraisal of their
own sleep, while objective monitoring takes into account more than simply time in bed,
including sleep latency and awakenings.
Conclusions

This is the first known study to examine the relationship between children’s sleep and
their HEI-2015 score, and the first know study to differentiate the within- and between-subject
effects of sleep duration and bedtime on next-day diet and eating behavior. By linking daily
EMA sleep reports to 24hr dietary recall-assessed dietary quality and eating behavior outcomes,
the present study revealed that children with longer usual sleep duration and earlier usual
bedtimes have higher overall dietary quality, and that nights with an hour longer than one’s usual
sleep duration is associated with nearly double the odds of breakfast consumption. This study
provides novel insight to the potential role of diet and eating as primary behavioral pathway
linking poor usual sleep health with elevated obesity risk in youth.

105
Table 6. Baseline demographic characteristics for 159 dyads enrolled in the MATCH Study
Children
M SD N %
Age (years) 9.59 0.94
Male 73 45.91
Weight category
Normal weight 103 64.78
Overweight 31 19.50
Obese 25 15.72
Hispanic a 93 59.24
White race a 69 44.23
Pubertal development category a
Pre Pubertal
48 30.38
Early Pubertal
50 31.65
Mid Pubertal
51 32.28
Late Pubertal
6 3.80
Post Pubertal
3 1.90
Mothers
M SD N %
Age (years) 42.95 6.21
Weight category
Normal weight 55 34.59
Overweight 54 33.96
Obese 50 31.45
Hispanic a 79 59.32
Married a 105 66.46
College graduate a 94 59.49
Works full-time a 92 61.33
Family
M SD N %
Household size 4.52 1.52
Household income a
≤ $35,000 41 25.95
35,001 - 75,000 49 31.01
75,001 - 105,000 34 21.52
> $105,000 34 21.52
a Data missing on variable

106
Table 7. Descriptive characteristics of sleep and diet collected for 159 children enrolled in the
MATCH Study

Variable M SD N %
Sleep health variables
Duration (hrs) 9.2 52 min
Bed time (hh:mm) 22:07 53 min
Wake time (hh:mm) 7:19 54 min
Dietary quality variables
HEI-2015 total score
a
50.44 9.04
HEI-2015 moderation sub-score
b
22.19 3.92
HEI-2015 adequacy sub-score
c
28.24 6.21
Eating behavior variables
Total kilocalories 1733.18 405.35
Total kilocalories- early 459.52 177.95
Total kilocalories - midday 645.27 220.87
Total kilocalories - evening 717.86 263.5
Evening overconsumption 331 42.88
Breakfast (1=yes) 729 94.43
N eating events 4.28 0.98
>4 eating events 321 41.58
Temporal characteristics
N observations/participants 4.85 2.10
Weekend day (vs. weekday) 405 52.46
Total L1 observations 772 100.00
Burst 2 37 4.79
Burst 3 143 18.52
Burst 4 171 22.15
Burst 5 197 25.52
Burst 6 224 29.02
Note: Mean (SD) values reflect the mean of each child's average value across observations; N (%)
reflects the number and percent across observation days
a
HEI-2015 Total Score ranges from 0 - 100 (higher=better)
b
Moderation Sub-Score is the sum of all moderation components; scores range from 0 - 40
(higher=better)
c
Adequacy Sub-Score is the sum of all adequacy components; scores range from 0 – 60
(higher=better)

107
Table 8. Results of multilevel linear regression models predicting children's dietary quality as a
function of their previous night's EMA-reported sleep duration and bedtime aggregated across
measurement bursts 2-6

HEI-2015 HEI-2015 HEI-2015

Total Score
a
Moderation Adequacy

Sub-Score
b
Sub-Score
c

Level-1 (day) n = 769 n=754 n=769
Level-2 (individual) N = 158 N=156 N=158
BS WS BS WS BS WS
! SE ! SE ! SE ! SE ! SE ! SE
Duration 2.14** 0.83 -0.04 0.46 1.05** 0.34 -0.02 0.23 1.09 0.57 0.01 0.30
Bedtime -3.35*** 0.78 -0.53 0.52 -1.21*** 0.35 -0.38 0.27 -2.24*** 0.53 -0.21 0.34
Note: Each combination of sleep health predictor and dietary outcome tested as a separate model
All models control for L1 covariates: burst #, and day of week (weekend vs. week day), and L2 covariates child sex and
child age at baseline
Models also control for any of the following baseline covariates that were significantly associated with the outcome of
interest at p <0.10: BMI category (mother or child), maternal education (college graduate vs. not college graduate),
maternal marital status (married vs. not married), maternal work status (full time vs. not full time), household income
quartile, household size, child race (white vs. non-white), child ethnicity (Hispanic vs. non-Hispanic), and Pubertal
Development Category
BS= Between-subjects (centered on the group mean); WS= Within-subjects (centered on the person mean)
The sample size for each model differs due to missing covariates
a
HEI-2015 Total Score ranges from 0 - 100 (higher=better)
b
Moderation Sub-Score is the sum of all moderation components; scores range from 0 - 40 (higher=better)
c
Adequacy Sub-Score is the sum of all adequacy components; scores range from 0 - 60 (higher=better)
Bold font indicates that the finding remained significant after adjusting for multiple comparisons using Benjamini-
Hochberg FDR
*p<0.05; **p<0.01, ***p<0.001

108

Table 9. Results of multilevel logistic regression models predicting children's eating behavior as a function of their previous night's
EMA-reported sleep duration and bedtime aggregated across measurement bursts 2-6

Breakfast > 4 Eating Evening

Consumption
a
Events
b
Overconsumption
c

Level-1 (day) n=769 n=754 n=772
Level-2
(individual) N=158 N=156 N=159

BS WS BS WS BS WS

OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI
Duration
1.47 0.88-2.44 1.44* 1.02-2.02 0.96 0.72-1.30 0.88 0.75-1.03 1.09 0.92-1.31 1.15 0.99-1.33

Bedtime
0.70 0.44-1.11 0.90 0.64-1.27 0.83 0.62-1.10 1.06 0.88-1.28 0.99 0.84-1.18 0.94 0.80-1.10
Note: Each combination of sleep health predictor and dietary outcome tested as a separate model
All models control for L1 covariates: burst #, and day of week (weekend vs. week day), and L2 covariates child sex and child age at baseline
Models also control for any of the following baseline covariates that were significantly associated with the outcome of interest at p <0.10: BMI
category (mother or child), maternal education (college graduate vs. not college graduate), maternal marital status (married vs. not married), maternal
work status (full time vs. not full time), household income quartile, household size, child race (white vs. non-white), child ethnicity (Hispanic vs. non-
Hispanic), and Pubertal Development Category
BS= Between-subjects (centered on the group mean); WS= Within-subjects (centered on the person mean)
The sample size for each model differs due to missing covariates
a
Breakfast consumption =1 on days when child reported eating breakfast
b
>4 eating events =1 on days when child reported greater than 4 eating occasions (excluding non-caloric intake)
c
Evening overconsumption=1 on days when child consumed highest proportion of kilocalories in the evening (17:00-24:00) as compared to mid-day
or morning.
Bold font indicates that the finding remained significant after adjusting for multiple comparisons using Benjamini-Hochberg FDR
*p<0.05; **p<0.01, ***p<0.001

109
Table 10. Results of multilevel regression models predicting children's dietary quality and eating behavior as a combined function of
their previous night's EMA-reported sleep duration and bedtime aggregated across measurement bursts 2-6

HEI-2015 Total
Score
a

HEI-2015
Moderation Sub-
Score
b

HEI-2015
Adequacy Sub-
Score
c

Breakfast
Consumption
d

> 4 Eating Events
e
Evening
Overconsumption
f

! (SE) ! (SE) ! (SE) OR (95% CI) OR (95% CI) OR (95% CI)
BS duration 0.22 (0.82) 0.47 (0.44) -0.32 (0.66) 1.30 (0.70-2.42) 0.78 (0.54-1.14) 1.15 (0.92-1.44)
WS duration -0.52 (0.57) -0.33 (0.29) -0.20 (0.37) 1.62 (1.04-2.53)* 0.86 (0.70-1.05) 1.18 (0.99-1.41)
BS bedtime -3.25 (0.95)*** -0.95 (0.43)* -2.43 (0.65)*** 0.82 (0.46-1.46) 0.71 (0.49-1.02) 1.08 (0.87-1.34)
WS bedtime -0.88 (0.65) -0.59 (0.33) -0.32 (0.43) 1.23 (0.79-1.92) 0.95 (0.76-1.20) 1.05 (0.86-1.29)
Note: All four sleep predictors simultaneously entered into the model with each dietary outcome
All models control for L1 covariates: burst #, and day of week (weekend vs. week day), and L2 covariates child sex and child age at baseline
Models also control for any of the following baseline covariates that were significantly associated with the outcome of interest at p <0.10: BMI
category (mother or child), maternal education (college graduate vs. not college graduate), maternal marital status (married vs. not married),
maternal work status (full time vs. not full time), household income quartile, household size, child race (white vs. non-white), child ethnicity
(Hispanic vs. non-Hispanic), and Pubertal Development Category
BS= Between-subjects (centered on the group mean); WS= Within-subjects (centered on the person mean)
a
HEI-2015 Total Score ranges from 0 - 100 (higher=better)
b
Moderation Sub-Score is the sum of all moderation components; scores range from 0 - 40 (higher=better)
c
Adequacy Sub-Score is the sum of all adequacy components; scores range from 0 - 60 (higher=better)
d
Breakfast consumption =1 on days when child reported eating breakfast
e
>4 eating events =1 on days when child reported greater than 4 eating occasions (excluding non-caloric intake)
f
Evening overconsumption=1 on days when child consumed highest proportion of kilocalories in the evening (17:00-24:00) as compared to mid-
day or morning.
Bold font indicates that the finding remained significant after adjusting for multiple comparisons using Benjamini-Hochberg FDR
*p<0.05; **p<0.01, ***p<0.001

110
Table 11. Results of multilevel regression models predicting children's dietary quality and eating behavior as a combined function of
their previous night's EMA-reported sleep duration, bedtime, and interaction of duration*bedtime, aggregated across measurement
bursts 2-6
HEI-2015 Total
Score
a

HEI-2015
Moderation
Sub-Score
b

HEI-2015
Adequacy Sub-
Score
c

Breakfast
Consumption
d

> 4 Eating
Events
e

Evening
Overconsumption
f

! (SE) ! (SE) ! (SE) OR (95% CI) OR (95% CI) OR (95% CI)
WS duration*
WS bedtime
0.37 (0.40) 0.23 (0.20) 0.11 (0.27) NS NS NS
BS duration 0.25 (0.98) 0.49 (0.44) -0.32 (0.66) 1.31 (0.70-2.45) 0.78 (0.54-1.14) 1.15 (0.92-1.44)
WS duration -0.56 (0.57) -0.35 (0.29) -0.21 (0.37) 1.61 (1.03-2.51)* 0.85 (0.70-1.04) 1.18 (0.99-1.42)
BS bedtime -3.17 (0.95)** -0.90 (0.43)* -2.40 (0.65)*** 0.83 (0.46-1.49) 0.72 (0.49-1.02) 1.09 (0.87-1.36)
WS bedtime -0.78 (0.65) -0.52 (0.33) -0.32 (0.43) 1.28 (0.80-2.06) 0.96 (0.76-1.21) 1.06 (0.86-1.30)
Note: All five sleep predictors (including interaction term) simultaneously entered into model with each dietary outcome
All models control for L1 covariates: burst #, and day of week (weekend vs. week day), and L2 covariates child sex and child age at baseline
Models also control for any of the following baseline covariates that were significantly associated with the outcome of interest at p <0.10:
BMI category (mother or child), maternal education (college graduate vs. not college graduate), maternal marital status (married vs. not
married), maternal work status (full time vs. not full time), household income quartile, household size, child race (white vs. non-white), child
ethnicity (Hispanic vs. non-Hispanic), and Pubertal Development Category
BS= Between-subjects (centered on the group mean); WS= Within-subjects (centered on the person mean)
a
HEI-2015 Total Score ranges from 0 - 100 (higher=better)
b
Moderation Sub-Score is the sum of all moderation components; scores range from 0 - 40 (higher=better)
c
Adequacy Sub-Score is the sum of all adequacy components; scores range from 0 - 60 (higher=better)
d
Breakfast consumption =1 on days when child reported eating breakfast
e
>4 eating events =1 on days when child reported greater than 4 eating occasions (excluding non-caloric intake)
f
Evening overconsumption=1 on days when child consumed highest proportion of kilocalories in the evening (17:00-24:00) as compared to
mid-day or morning.
Bold font indicates that the finding remained significant after adjusting for multiple comparisons using Benjamini-Hochberg FDR
*p<0.05; **p<0.01, ***p<0.001

111
CHAPTER 4: INDEPENDENT AND COMBINED EFFECTS OF WITHIN-
SUBJECT MEAN AND VARIABILITY IN SLEEP DURATION ON
LONGITUDINAL WEIGHT TRAJECTORIES AND ATTAINED BMI IN YOUTH

Abstract

Introduction: Short sleep duration has been linked to elevated obesity risk among youth.
Beyond mean sleep duration, recent studies suggest a unique role of within-subject variability, or
the extent to which sleep varies across days, as a risk factor for obesity. However, only a handful
of studies have studied the independent role of within-subject variability in sleep duration on
children’s weight, and few studies have considered the combined effect of short mean sleep
duration and high sleep duration variability on weight outcomes. This study sought to determine
the independent and combined effects of within-subject mean and variability in sleep duration on
children’s BMI trajectory and attained BMI across 1.5 years.
Methods: Children (N=153; mean age: 10.5 years; 51% female) completed 8-day measurement
bursts every six months (4 bursts and 1.5 years total). At each burst, children used a smartphone
application to report on daily bed and wake times, which were used to calculate each child’s
baseline within-subject mean and variability (SD) in sleep duration, as well as a sleep risk factor
score representing the combined effects of within-subject mean and variability in duration. At
each burst, children’s height and weight were measured to determine BMI. Multilevel growth
curve models tested the independent and combined associations of baseline within-subject mean
sleep duration, variability in sleep duration, and sleep risk factor score on rate of BMI change
(slope) across 1.5 years and attained BMI (intercept) at 1.5 years, controlling for age, sex,
pubertal development, and other relevant covariates.
Results: At baseline, 66.4% of children obtained the minimum recommended ≥ 9 hours of sleep
on average, while 30.9% obtained 8-9 hours, and 2.7% obtained < 8 hours. Average day-to-day

112
sleep duration variability was 50 minutes, and 36% were classified as either overweight or obese.
At baseline, children with one or two sleep risk factors trended toward significantly higher BMI
than children with zero risk factors (F-value= 3.01, p= 0.053). Regardless of sleep duration
characteristics, children’s BMI increased by approximately 0.5 units at each measurement burst,
for a total average increase of approximately 1.5 kg/m
2
across the study period. There was no
significant association of baseline within-subject mean or within-subject variability in sleep
duration, or sleep risk factor score on rate of BMI change across 1.5 years or attained BMI at 1.5
years (p’s all >0.05).
Conclusions: Among a sample of healthy youth, baseline sleep duration and sleep duration
variability were not significantly associated with BMI trajectory or attained BMI at the end of
the study. A potential reason for the lack of association is the relatively short follow-up period
compared to other studies, and the high rate of meeting sleep recommendations within this
sample.

113
Introduction

Over the past several decades, the US has seen staggering increases in the proportion of
individuals who are overweight or obese. Current estimates suggest that approximately one in
every three US children ages 2-19 is classified as overweight or obese, (Ogden, Carroll, Kit, &
Flegal, 2014) defined as body mass index (BMI) z-score ≥ 85
th
percentile and ≥ 95
th
percentile
for sex and age, respectively. Obesity is a complex, multifaceted disease associated with a
multitude of adverse health outcomes, including as cardiovascular disease, diabetes, and
decreased overall life expectancy. The development of overweight and obesity in youth has the
potential to negatively impact lifelong health and contribute to morbidity and mortality. (Dietz,
1998; A. S. Singh et al., 2008) While two energy balance behaviors - physical activity and eating
– are the primary behavioral contributors to childhood obesity, other behavioral factors,
including sleep health, have also emerged as potential obesity risk factors. (M. Miller,
Kruisbrink, Wallace, Ji, & Cappuccio, 2018)
Along with steady increases in obesity rates, there has been a marked decline in sleep
health over the past decade, including among children and adolescents. (Iglowstein, Jenni,
Molinari, & Largo, 2003; Matricciani, Olds, & Petkov, 2012) Individuals today are sleeping less
than previous generations, a trend that cuts across ages, genders, and geographical locations,
which amounts to an overall reduction in children’s nightly sleep duration by approximately of
one hour over the past century. (Matricciani, Olds, & Petkov, 2012) It is speculated that these
global shifts in sleep health are driven in part by technological advances and modernization,
including increased exposure to artificial light which can disrupt circadian regulation of sleep,
and availability of television and smartphones, which contribute to a 24hr society. (Matricciani,
Olds, & Petkov, 2012; Matricciani, Olds, Blunden, Rigney, & Williams, 2012) The decrease in
sleep duration among youth is especially troubling, given that adequate sleep is essential for

114
numerous cognitive and developmental processes, (Buysse, 2013; Wolfson & Carskadon, 2003)
in addition to its importance for overall health, well-being and longevity. (Colten, Altevogt, &
Colten., 2006) According to the National Sleep Foundation, school-aged children (i.e., ages 6-
13) require 9-11 hours of sleep per night, while adolescents (i.e., ages 14-17) require 8-10 hours
per night. (Hirshkowitz et al., 2015) Despite this, it is estimated that fewer than half of children
and adolescents meet guidelines for minimum nightly sleep duration. (Basch et al., 2014)
The Role of Mean Sleep Duration in Obesity

The high prevalence of both short sleep and obesity may not be independent phenomena;
indeed, evidence from observational, experimental, and epidemiological studies supports an
inverse association between sleep duration and obesity risk in children. (Börnhorst et al., 2012;
Gonnissen et al., 2013; Nielsen et al., 2011) A meta-analysis of longitudinal studies found that
children with short usual sleep duration have twice the odds of overweight or obesity (OR: 2.15;
95% CI: 1.64–2.81), (Fatima et al., 2015) and other meta-analyses have echoed the protective
effect of adequate sleep duration on weight status. (M. Miller et al., 2018)
Potential mechanisms linking mean sleep duration to obesity. The strong association
of sleep duration and obesity underscores the need to better understand how various aspects of
sleep health affect weight gain in children. Several biological and behavioral mechanisms have
been posited to explain the association between short sleep and elevated obesity risk. Insufficient
sleep may increase activation of the hypothalamic pituitary adrenal (HPA) axis, (Kumari et al.,
2009; Ly et al., 2015) producing stress hormones known to increase appetite, and suppressing the
expression of leptin, a hormone known to produce feelings of satiety. (Pervanidou & Chrousos,
2012) Short sleep may promote increased brain reactivity to highly palatable (e.g., high-fat, high-
sugar) food cues, (St-Onge et al., 2014) leading to increased reward response to weight-
promoting foods. In addition to biological mechanisms, many of the hypothesized pathways

115
linking poor usual sleep with childhood obesity involve behavioral alterations, such as decreased
physical activity, (Stone et al., 2013) increased sedentary behavior, (Hale & Guan, 2015) poorer
dietary quality, (Westerlund et al., 2009) and altered eating behavior. (Spaeth et al., 2019)
The Role of Within-Subject Variability in Sleep Duration on Obesity

Beyond short within-subject mean (i.e., average) sleep duration, several recent studies
have highlighted the role of high within-subject variability in duration (i.e., the extent to which
an individual’s sleep duration varies across repeatedly assessed days) as a novel risk factor for
obesity. Some degree of variability in sleep across repeatedly assessed days is common. (Dillon
et al., 2015) However, a high degree of within-subject variability in sleep is associated with less
healthy lifestyle choices, such as poorer diet and elevated sedentary behavior. (Duncan et al.,
2016) Growing evidence suggests that, even after adjusting for within-subject mean sleep
duration, greater within-subject variability in sleep duration may be independently associated
with children’s adiposity outcomes, including both BMI (Kjeldsen et al., 2014; Moore et al.,
2012) and visceral fat. (He, Bixler, Liao, et al., 2015)
However, while there is a broad body of evidence supporting the relationship of within-
subject mean sleep duration on obesity outcomes, (Fatima et al., 2015; M. Miller et al., 2018)
very few studies to date have assessed the role of within-subject variability in sleep duration as it
relates to children’s obesity. Among the handful of existing studies, findings have been mixed,
which highlights the need for further investigation. (Bei et al., 2016; D. Jarrin et al., 2013) With
one known exception, all of the studies that repeatedly assessed children’s and adolescents’ sleep
across days to examine the association of within-subject variability in sleep duration (controlling
for within-subject mean duration) on with obesity risk have been cross-sectional. This is an
important limitation, because measuring both sleep and adiposity at the same point in time limits
the ability to determine the effect of sleep on subsequent changes in weight.

116
To date, four known studies (two in the same cohort) have found a positive association
between within-subject sleep duration variability and adiposity in youth. (He, Bixler, Berg, et al.,
2015; He, Bixler, Liao, et al., 2015; Moore et al., 2011; Spruyt et al., 2011) For example, one
study found a significant association of within-subject variability (i.e., SD) in sleep duration with
abdominal adiposity in adolescents, (He, Bixler, Liao, et al., 2015) and an effect such that a one-
hour increase in within-subject sleep duration variability was associated with higher
android/gynoid fat ratio and android/whole body fat ratio. (He, Bixler, Berg, et al., 2015)
Another study in adolescents found a significant positive correlation of actigraphy-assessed
within-subject variability in sleep duration and BMI. (Moore et al., 2011) In younger children
(ages 4-10), obese children had higher within-subject variability in sleep duration compared to
non-obese children. (Spruyt et al., 2011)
Five studies (two in the same parent cohort) have found a null association between
within-subject variability in repeatedly assessed sleep duration and obesity outcomes in youth.
(Cespedes Feliciano, Oken, Taveras, & Redline, 2018; Jansen et al., 2018; Kuo et al., 2014;
McHale, Kim, Kan, & Updegraff, 2011; Park et al., 2016) Of these null reports, two were
conducted within a population of Mexican-American adolescents who completed 7-day sleep
reports with daily telephone interviews at two timepoints separated by 2 years. Within this
cohort, the cross-sectional analysis (McHale et al., 2011) found a non-significant linear
association between sleep duration variability and BMI (b= 0.77 [SE: 1.41]), while the
longitudinal analysis (Kuo et al., 2014) also determined a non-significant prospective association
of sleep duration variability and BMI two years later (b= - 0.22 [SE: 0.31]). Other studies found
non-significant correlation of within-subject sleep variability and BMI (Park et al., 2016) or with
several adiposity indicators (e.g., BMIz, waist circumference, DXA fat mass) in multivariate
analyses. (Cespedes Feliciano et al., 2018) A recent study in n=528 youth ages 9-17 in Mexico

117
City, participants were classified into four groups based on mean sleep duration (meeting
recommendation vs. not) and sleep duration variability (above or below the group median for
standard deviation of sleep duration). This study found that youth with insufficient/stable sleep
duration had significantly higher levels of adiposity (i.e., BMIz, % body fat, triceps skinfolds,
waist circumference, p’s all <0.001) than youth with sufficient/stable sleep duration. However,
there was no additional detrimental effect of insufficient/unstable (i.e., highly variable) sleep on
adiposity outcomes. (Jansen et al., 2018)
Potential mechanisms linking within-subject variability in sleep duration to obesity.
A number of biological and physiological mechanisms may underlie the unique effect of greater
within-subject variability in sleep duration on obesity risk. Controlling for mean levels, greater
day-to-day variability in adolescent sleep duration was prospectively associated with decreased
white matter integrity, (Telzer et al., 2015) which can lead to deficits in information processing
and cognitive control, and difficulty in planning and executing healthy behavior across the
lifespan. Strikingly, the detrimental impact on white matter was not observed for short within-
subject mean sleep duration, indicating a unique role of within-subject variability in sleep health
for brain development. (Telzer et al., 2015) Greater within-subject variability in sleep may also
lead to circadian misalignment, as previous research has demonstrated that variability in sleep
and wake schedules (and subsequent changes to overall duration) may lead to deregulation of
hormones and temperatures that can take several days to regulate. (Sadeh et al., 2009) Circadian
misalignment promotes caloric intake and disrupts glucose metabolism, which both may
independently lead to weight gain. (McHill & Wright, 2017) Furthermore, elevated within-
subject sleep duration is linked to psychosocial stress (Mezick et al., 2009) and negative
affectivity, (Fuligni & Hardway, 2006) both of which have been independently associated with
weight-related behaviors and obesity risk. (M. Singh, 2014; Torres & Nowson, 2007)

118
Limitations of previous studies examining within-subject variability in sleep
duration and obesity. Although several studies have provided evidence for an association
between within-subject variability in sleep duration and obesity, these studies are limited by their
cross-sectional nature. Cross-sectional designs cannot determine causality, nor can they rule out
the possibility of reverse causation, or a bi-directional association, such that elevated BMI
increases an individual’s risk for sleep disordered breathing, shorter sleep duration, and poorer
overall sleep health. (Vgontzas et al., 2010) Cross-sectional design can also be particularly
problematic for use in studying multiple health behaviors, as the underlying relationship between
the two processes may be related through a specific developmental trajectory that is not captured
through a static, single time-point assessment (Kraemer et al., 2000) An understanding of the
longitudinal association of within-subject sleep duration variability is especially important due to
evidence from studies finding that mean sleep duration in childhood predicts adult obesity risk,
suggesting that the effect of poor sleep may accumulate and compound over time to impact
health outcomes. (Landhuis, Poulton, Welch, & Hancox, 2008)
Potential Combined Effect of Within-Subject Mean and Variability in Sleep Duration on
Obesity

Given that both within-subject mean and within-subject variability in sleep duration have
both been independently implicated in elevated obesity risk, it is possible that these two sleep
health characteristics may exhibit a combined effect, whereby individuals with the shortest
within-subject mean duration and the greatest within-subject variability in duration may exhibit
the poorest overall weight outcomes. To date, there is some limited evidence for a combined
effect. In the previously described study by Spruyt et al. (2011), children (ages 4-10) with the
shortest mean sleep duration and most irregular sleep schedules exhibited the poorest metabolic
health (elevated BMI z-score, altered insulin, LDL, and CRP levels), while children with the
longest mean sleep and most regular sleep schedules had the healthiest metabolic profile. (Spruyt

119
et al., 2011) In another study, parents completed a questionnaire, reporting on their child’s usual
sleep on weekends and weekdays (i.e., not repeated assessment across days, as in the other
reviewed studies above). This study found that longer mean sleep duration was associated with
smaller increases in BMI z-score over time, but only in children with the most consistent
bedtimes (b= -0.12, 95% CI: -0.23, -0.01). (Miller et al., 2014) However, other studies have
found no combined effect of short mean sleep duration and elevated variability in sleep duration
on weight outcomes. For example, the study of adolescents in Mexico City found that youth with
insufficient/stable sleep duration had higher levels of adiposity than youth with sufficient/stable
sleep duration, but there was no additional detrimental effect of insufficient/unstable (i.e., highly
variable) sleep on adiposity outcomes. (Jansen et al., 2018)
The hypothesized effect in which a child’s overall sleep health is a function of both the
within-subject mean duration level and the degree of within-subject variability in duration, is
illustrated in Figure 5. In this conceptual diagram, a given individual is a classified according to
(a) within-subject mean (i.e., average) sleep duration, which is either adequate (greater than or
equal to the recommended 9 hours for youth up to age 14) or short (less than the recommended 9
hours), and (b) the degree of within-subject variability in sleep duration present across repeatedly
assessed days, classified as low variability (less than the group mean), or high variability (greater
than or equal to the group mean). According to this conceptual diagram, children with adequate
within-subject mean and low within-subject variability in sleep duration have the ‘ideal’ sleep
health, while children with both short mean and high variability have the ‘poorest’ sleep health,
and children with one sleep health risk factor have moderate relative sleep health.
In summary, cross sectional and longitudinal literature supports a strong association
between short mean sleep duration and elevated risk of obesity in youth. (Cappuccio et al., 2008;
Fatima et al., 2015) However, only a handful of studies have studied the unique role of within-

120
subject variability in sleep duration on children’s obesity risk, nearly all of which have been
cross-sectional. Furthermore, few studies have considered the combined effect of short sleep
duration and high sleep variability as it relates to obesity in youth. The present study builds upon
previous research as the first known study to examine the associations of both within-subject
mean and variability in children’s sleep duration on the longitudinal change in BMI and attained
BMI levels across a nearly two-year period.
Specific Aims and Hypotheses

7. To test the association of baseline within-subject mean sleep duration on the rate of
change of BMI across 1.5 years and intercept of BMI at 1.5 years.
Hypothesis: It was hypothesized that children with shorter mean sleep duration at
baseline would have steeper rates of increase in BMI from baseline to 1.5 years, and
higher BMI intercept at the end of the study.
8. To test the association of baseline within-subject variability in sleep duration on the
rate of change of BMI across 1.5 years and BMI intercept at 1.5 years.
Hypothesis: It was hypothesized that, controlling for one’s mean sleep duration,
greater 7-night variability in sleep duration at baseline would be associated with
steeper rates of increase in BMI across 1.5 years and higher BMI intercept at the end
of the study.
9. To test the combined effect of baseline shorter within-subject mean sleep duration
and higher sleep duration variability (i.e., using a sleep risk factor score, with possible
scores of 0 [adequate duration/low variability], 2 [either adequate duration/high
variability; or short duration/low variability], and 4 [short duration/high variability])
on the rate of change of BMI across 1.5 years and attained BMI at 1.5 years.
Hypothesis: It was hypothesized that, compared to children with adequate duration

121
and low variability (i.e., sleep risk factor score of 0), children with either short
duration or high variability (i.e., score of 2) would have steeper rates of increase in
BMI across 1.5 years and higher attained BMI at 1.5 years, and that children with
both short duration and high variability (i.e., score of 4) would have the steepest
increase in BMI across 1.5 years and highest attained BMI at the end of the study.
Methods

This study draws upon data from the Mothers’ and their Children’s Health (MATCH)
Study, a longitudinal study with semi-annual assessments (i.e., measurement bursts). Children
used an ecological momentary assessment (EMA) smartphone application to report on their daily
sleep for one week and underwent in-person anthropometric assessments at each burst. Analyses
explored the longitudinal effect of children’s baseline within-subject mean and variability in
sleep duration on rate of BMI change across 1.5 years and attained BMI at the end of the study.
Participants

Mother-child dyads in the MATCH Study were recruited from schools and community
centers in Los Angeles County area of southern California. A total of 202 dyads enrolled in the
study. Children were in 3
rd
– 6
th
grade, spoke and wrote in English or Spanish, and had no heath
conditions limiting physical activity or requiring inhaler. Full inclusion and exclusion criteria
and detailed information on the MATCH study design and sample is published elsewhere.
(Dunton et al., 2015)
Procedures

Children provided assent, and mothers provided consent, for self and child to participate
in the MATCH Study. MATCH dyads completed measurement bursts approximately every six
months for three years, for a total of six semi-annual bursts. At each burst, mothers and children
attended an in-person visit, during which they completed paper and pencil surveys, were

122
assessed for anthropometric measurements, and received written and verbal instructions on the
proper use of all study devices. Children were trained on use of the study application (‘app’),
each on their own device. Participants who had their own Android compatible phone are asked to
download the study app on their own phone, and participants without a phone or with a non-
Android phone borrowed a MotoG study phone (Motorola Mobility, USA) for use during the
study period. Upon leaving the visit, children completed eight consecutive days of ecological
momentary assessment (EMA) surveys. Beginning with the third measurement burst (i.e.,
‘baseline’ for the purposes of the current study), a set of sleep questions were included in the
daily EMA surveys. Because sleep data is available beginning from the third burst, the present
analysis draws upon data from measurement bursts 3-6 (4 assessment total), and for the purposes
of this analysis the third burst is referred to as ‘baseline’.
Measures

EMA sleep. At baseline, children completed up to 7 days of EMA sleep reports using a
smartphone application. Sleep questions were included in the first answered prompt of each day;
on weekend days, the first survey occurred between 7:30-8:00am, and on weekdays the first
survey of the day occurred between 3:30-4:00pm. If the sleep items were skipped or incomplete,
the sleep questions were repeated at subsequent windows, allowing participants multiple
opportunities to respond. Children were asked to report on their previous night’s sleep each
morning via EMA survey, including “What time did you fall asleep last night?” and “What time
did you wake up this morning?” Day-level sleep duration was calculated as the total number of
minutes between self-reported bedtime and wake-time. All available baseline sleep reports were
aggregated to calculate each participant’s within-subject mean sleep duration (i.e., average across
days) and within-subject variability in sleep duration (i.e., standard deviation across days).
Standard deviation of the mean sleep duration as an indicator of within-subject sleep variability

123
has been used in previous studies. (Ogilvie et al., 2013)
Sleep risk factor score. An overall baseline sleep risk factor score was created to
categorize each child based on their usual level and the degree of variability in sleep duration.
The score was calculated by taking the product of the values for short sleep duration and high
sleep variability, such that higher scores indicated poorer sleep health. Short sleep duration at
baseline was defined as a mean sleep duration of < 540 minutes (9 hours). This was selected in
accordance with current National Sleep Foundation recommendations for sleep duration.
(Hirshkowitz et al., 2015) Participants with sleep duration ³540 min were assigned a score of 1
(adequate duration), and participants with sleep duration <540 min were assigned a score of 2
(short duration), such that a higher number signified shorter 7-night mean sleep duration. High
sleep variability at baseline was defined using the group level average standard deviation of
sleep duration (i.e., 50 minutes). Participants with sleep variability less than 50 min were
assigned a score of 1 (low variability), and participants with sleep variability greater than or
equal to 50 min were assigned a score of 2 (high variability), so that a higher number signified
more 7-night variability in sleep duration.
To calculate each child’s sleep risk factor score, the values for short sleep duration and
high sleep variability were multiplied. Children with both adequate duration and low variability
(i.e., those with the ‘healthiest’ sleep) received a sleep risk factor score of 1 (i.e., 1*1=1);
participants with either short duration or high variability received a sleep risk factor score of 2
(i.e., 1*2=2), and participants with both short duration and high variability (i.e., those with the
‘unhealthiest’ sleep) received a sleep risk factor score of 4 (i.e., 2*2=4). To illustrate, a child
with adequate sleep duration (short sleep duration = 1), and high sleep variability (high sleep
variability = 2) would have a sleep risk factor score of 2 (1*2 = 2). This coding was selected in
order to approximate an interaction effect, in which an increasing number of sleep risk factors is

124
reflected by increasingly higher sleep risk score value.
Anthropometric measures. At each measurement burst, children’s weight and height
were assessed using a digital scale (Tanita WB-110A) and stadiometer. Anthropometric
measurements were taken in duplicate and averaged when discrepant. Body mass index (BMI;
kg/m
2
) and weight classification were calculated using EpiInfo 2005, Version 3.2 (CDC, Atlanta,
GA). In the present study, continuous BMI was used in the main models predicting the change in
children’s weight status over time. BMI categories (normal weight, overweight, obese) were
used for the purpose of descriptive statistics in preliminary analyses.
Demographic measures. Children completed surveys reporting on their sex, age and
pubertal development. Pubertal development was measured using the Pubertal Development
Scale (PDS), a 5-item, sex-specific self-report scale, which assessed the presence of physical
changes associated with puberty in various domains. (Petersen et al., 1988) Scores across items
were summed to assign the PDS category, which ranges from 1 (pre-pubertal) – 5 (post-
pubertal). At each measurement burst, participants’ mothers completed surveys reporting on their
marital status (married vs. non-married), number of individuals living in the household
(continuous), annual household income (<$35,000, $35,001-$74,999, $75,000-$104,999, and
$105,000+), maternal education level (college vs. no college) as well as their child’s race (white
vs. non-white) and ethnicity (Hispanic vs. non-Hispanic).
Statistical Analysis

The current analysis includes children who 1) completed the baseline assessment and
provided both sleep and BMI data, and 2) completed at least one additional BMI measurement
period after baseline (i.e., had at least two available BMI measurements). Children who were
missing sleep data at baseline were not included in any analytic models, resulting in a L2 sample
size of N=145 participants and a total L1 sample size of n=580 measurement bursts.

125
To describe the study population and assess potential model covariates, descriptive
summary statistics were calculated using mean (SD) for continuous and N (%) for categorical
variables. Participant demographic characteristics were calculated at both baseline and at the end
of the study 1.5 years later, and sleep descriptive statistics were calculated at baseline to examine
group-level distribution of sleep health. T-tests and correlation was used to compare sleep
indices by demographic characteristics.
Mixed effects linear growth models were used to examine the age- and sex-adjusted
associations between baseline sleep health characteristics and the rate of change in BMI across
the study period and the attained value of BMI at the end of the study, 1.5 years after baseline.
The effect of time was represented by measurement burst, which was coded as -3 (baseline), -2
(6 months), -1 (1 year), and 0 (end of study, 1.5 years after baseline); thus every one-unit
increase in time reflects the change in BMI across an approximately 6-month period, and the
estimate for intercept reflects the effect of predictors on the BMI value at the end of study. All
analyses were conducted using SAS v. 9.4 (SAS Institute, Cary, NC). Statistical significance was
determined at p <0.05.
Research Question 7. To test the association of baseline within-subject mean sleep
duration on the rate of change of BMI across 1.5 years and intercept of BMI at 1.5 years.
Mixed effect linear growth models were conducted using SAS PROC MIXED. Multilevel
models account for the clustering of observations (i.e., measurement bursts) within participants
(children). Intercept and slope were treated as random variables in models to allow for subject-
level differences in BMI (intercept) at the end of study and in BMI trajectory (slope) from
baseline to end of study. The two parameters of interest were estimated in a single combined
model: specifically, the effect of baseline within-subject mean sleep duration on (1) the change
in BMI (slope) from baseline to end of study (a period of approximately 1.5 years) and (2) the

126
attained BMI (intercept) at the end of the study. A stepwise approach was used to add predictors
into the model, as follows: (a) time only; (b) time and baseline mean sleep duration only; (c)
time, baseline mean sleep duration and an interaction term for baseline mean sleep
duration*time; (d) time, baseline mean sleep duration and an interaction term for baseline mean
sleep duration*time, baseline age and sex; and (e) the final model including other covariates that
significantly altered the effect of sleep on BMI.
A priori covariates included child sex and baseline age (Level 2). Additional baseline
covariates associated with children’s sleep and BMI were screened for inclusion into the final
model, and for the sake of parsimony, only a priori covariates (i.e., age and sex) and those that
altered the effects of interest by 10% or more were retained in the final model. Due to a high ICC
(0.95) and evidence that time-invariant characteristics were the most relevant predictor of BMI in
children, only person-level (L2) covariates were considered for inclusion as covariates. Screened
baseline covariates included: maternal marital status (married vs. not married), maternal
employment status (fulltime vs. not fulltime) maternal education (college graduate vs. not
college graduate), annual household income (quartiles: ≤ $35,000; $35,001 - $75,000; $75,001 -
$105,000l; > $105,000); child race, ethnicity, pubertal development category, and number of
days of sleep reports provided at baseline.
Prior to analyses, baseline within-subject mean sleep duration was grand-mean centered,
so that each participant’s baseline mean duration reflected the deviation of their mean sleep
duration from the group mean. All other continuous covariates (e.g., child age) were also grand-
mean centered, so that their interpretation reflects the effect of the variable on BMI outcomes for
a child at the average level of that variable (e.g., the effect of child age on BMI for a child of the
average age). In order to estimate the effect of baseline within-subject mean sleep duration on the
rate of change (slope) of BMI across the study period, models include an interaction term for

127
time*baseline mean sleep duration. The interaction term estimates the effect of baseline within-
subject mean sleep duration on the change in BMI over the four measurement bursts. The effect
of time was represented by measurement burst, which was coded as -3 (baseline), -2 (6 months),
-1 (1 year), and 0 (end of study, 1.5 years after baseline). Every one-unit increase in time reflects
the change in BMI across an approximately 6-month period, and the estimate for time intercept
reflects not the mean starting BMI at baseline, but instead the mean attained BMI at the end of
study (i.e., when the re-coded measurement burst is equal to 0). (Dunton, Intille, Wolch, &
Pentz, 2012; Mcconnell et al., 2015) Unstandardized ß coefficients are reported, representing the
effect of predictors on actual BMI units.
Level 1: !
"#
=%
&#
+ (
)#
+
"#
+ ,
"#

Level 2: %
-"
=.
&&
+ .
&)
%/0
#
+ 1
&#

%
"#
=.
)&
+ .
))
%/0
#
+ 1
)#

Basic Combined Model: !
"#
=.
&&
+ .
&)
%/0
#
+ .
)&
+ .
))
%/0
#
+
"#
+ 1
&#
+ 1
)#
T + ,
"#

The above equations illustrate the basic Level 1, Level 2, and combined equations for the overall
statistical approach. Here, in the Level 1 equation, !
"#
denotes the dependent variable - BMI -
for individual j at time i; %
&#
represents the intercept for individual j; (
)#
denotes the slope for
individual j; +
"#
denotes the time of measurement i for individual j; ,
"#
denotes the within-subject
residual term, or the difference between the predicted and actual BMI value for individual j at
time i.
In the Level 2 equation, the fixed effects are as follows: .
&&
denotes the BMI for
individuals at the mean level of baseline sleep duration at the end of the study; .
&)
denotes how
much higher or lower the BMI at the end of the study is for individuals with higher than average
levels of baseline sleep duration; .
)&
represents the change in BMI for each measurement burst

128
for individuals with the mean sleep duration; and .
))
denotes how much higher or lower the rate
of change for BMI is for individuals with higher than average baseline sleep duration. The
random effects refer to the distribution of subjects’ intercepts and slopes above or below the
group average levels. Here, 1
&#
is the subject-specific intercept deviation, and 1
)#
is the subject-
specific slope deviation.
Research Question 8. To test the association of baseline within-subject variability in
sleep duration on the rate of change of BMI across 1.5 years and BMI intercept at 1.5 years.
To test RQ8, an identical MLM approach to RQ7 was taken, with within-subject
variability in sleep duration as the main predictor to estimate the effect of sleep duration
variability on BMI slope and intercept, controlling for within-subject mean sleep duration. Mixed
effect linear growth models were conducted using SAS PROC MIXED. Intercept and slope were
treated as random variables in models to allow for subject-level differences in attained BMI
(intercept) and in BMI trajectory (slope). Two outcomes were estimated in a single combined
model: specifically, the effect of baseline within-subject sleep duration variability on (1) the
change in BMI (slope) from measurement burst 0 – 3 (re-coded as -3 to 0) and (2) the attained
BMI (intercept) at measurement burst 3 (1.5 years post-baseline). A stepwise approach was used
to add predictors into the model, as follows: (a) time only; (b) time and baseline sleep duration
variability only; (c) time, baseline sleep duration variability, and baseline mean sleep duration;
(d) time, baseline sleep duration variability, sleep duration, and an interaction term for baseline
sleep duration variability*time, baseline age and sex (e) the final model including other
covariates that significantly altered the effect of sleep on BMI. For parsimony across models, the
same covariates that were identified in RQ7 were included in models for RQ8 and RQ9.
Research Question 9. To test the combined effect of baseline shorter within-subject
mean sleep duration and higher within-subject sleep duration variability (i.e., using a sleep risk

129
factor score, with possible scores of 0 [adequate duration/low variability], 2 [either adequate
duration/high variability; or short duration/low variability], and 4 [short duration/high
variability]) on the rate of change of BMI across 1.5 years and attained BMI at 1.5 years.
To test RQ9, an identical approach to RQ7 and RQ8 was employed. However, the terms
for sleep duration and sleep duration variability were removed, and instead the indicator for sleep
risk factor score was included in the model (i.e., the product of short duration and high
variability, with possible values of 1, 2, and 4), in order to estimate the combined effect of low
duration and high variability on BMI slope and intercept. To estimate the effect of baseline sleep
risk factor score on (1) the change in BMI (slope) from baseline to end of study (a period of
approximately 1.5 years) and (2) the attained BMI (intercept) at the end of study (1.5 years post-
baseline), the following models were tested: (a) time only; (b) time and baseline sleep risk factor
score only; (c) time, baseline sleep risk factor score, and an interaction term for baseline sleep
risk factor score *time, baseline age and sex (e) the final model including other covariates that
significantly altered the effect of sleep on BMI. All other modeling approaches described for
RQ7 and RQ8 were also applied in RQ9.
Results

Data Availability

The initial analytical sample included 153 (participants) x 4 (measurement bursts), or 612
observations. Of those, four participants did not have any available sleep data, and four
participants had only one sleep report, and were excluded from the analytical sample. There were
no significant differences in baseline BMI, mean sleep duration, sex, pubertal status, race, family
income, maternal education between the included sample as compared to the excluded sample.
However, excluded children were more likely to be of Hispanic ethnicity (100% of excluded vs.
57% of included, p< 0.05). The final L2 sample consisted of N=145 participants, who provided a

130
total L1 sample size of n=580 measurement bursts. Among the included sample, children
provided an average of 5.48 days (SD: 1.51, range: 2-7) of sleep reports.
Descriptive Results

Demographics. Participant characteristics at baseline and at the end of the study are
reported in Table 12. At baseline, the mean age was 10.5 years (SD: 0.91) and 43.8% of the
sample was male. The mean baseline BMI was 19.67 (SD: 4.20); 64% of youth were of normal
weight, 19% were classified as overweight, and 17% were obese. Between baseline and the end
of the study 1.5 years later, mean BMI increased from 19.67 kg/m
2
to 21.25 kg/m
2
, for an
average increase of approximately 1.58 kg/m
2
over the 1.5-year study period. Although older
children had marginally higher BMI at the end of study compared to younger children (b=
0.8277 [SE: 0.4244], p= 0.053), there was no significant effect of baseline age on the rate of
change of BMI across 1.5 years (p> 0.05).
Baseline sleep descriptive characteristics are presented in Table 13. The majority (66.4%)
of children met minimum recommended sleep duration (9 hours) at baseline, and 95% of
children had an average sleep duration between 480 – 660 minutes (8 – 11 hours). Across days,
the average sleep duration variability was approximately 49.84 minutes. A total of 56 (38.62%)
children had neither sleep risk factor, 68 (46.90%) participants had one sleep risk factor, and 21
(14.48%) had both sleep risk factors.
Preliminary Analyses

Preliminary analyses revealed no significant difference in baseline within-subject mean
or variability in sleep duration between boys vs. girls (p’s >0.05), or between children classified
as overweight/obese vs. normal weight at baseline. There was a non-significant correlation
between baseline BMI and sleep duration (r = -0.110, p= 0.188), and children with short baseline
duration (<9 hrs) had non-significantly higher BMI (short duration: 20.5 kg/m
2
vs. adequate

131
duration: 19.2 kg/m
2
; p= 0.12). Similarly, there was a non-significant positive correlation
between baseline BMI and sleep variability (r = 0.039, p= 0.646), and children with high
baseline within-subject variability had non-significantly higher BMI (high variability: 20.15
kg/m
2
vs. low variability: 19.38 kg/m
2
; p= 0.28). Additionally, at baseline, sleep risk factor score
and BMI were non-significantly correlated (r= 0.136, p= 0.107), but children with one or two
sleep risk factors had trending toward significantly higher BMI at baseline than children with no
risk factors (F-statistic: 3.01, p= 0.053).
Growth Curve Models

Preliminary visualization was conducted to inspect the fixed and random effects in initial
BMI and change in BMI across the study period as a function of baseline mean sleep duration
(Figure 6) and sleep duration variability (Figure 7). In these plots, the bold red line represents the
average effect for each group, while the black thin lines represent the individual trajectories of
BMI for individuals in each group. In preliminary visualization of the relationship between
adequate vs. short sleep duration (Figure 6), participants with adequate (≥9 hrs) baseline sleep
duration had lower initial BMI value and attained value (i.e., intercept), but similar rate of
change (i.e., slope) of BMI across the study period. Taken altogether, the initial and attained
values for individual BMI trajectories among children with low sleep duration demonstrated
greater variance and higher upper values overall. Preliminary visualization of BMI trajectory by
the degree of within-subject variability (Figure 7) in sleep duration suggested that participants
with high baseline variability (>50 min) had both higher initial and attained values, as well as
steeper slope of BMI increase across the study compared to those with low baseline sleep
variability; however, these effects were visually modest. Next, formal linear growth curve
models predicting BMI change and attained values as a function of sleep were tested in order to
determine the significance of these visual trends.

132
In empty model predicting BMI, the fixed effects estimate for group mean BMI at
baseline was 20.58 kg/m
2
. The intra-class correlation (ICC) for BMI was 0.95, indicating that
95% of the variability in children’s BMI across measurements was attributable to individual-
level (L2) differences. The remaining 5% of variance was attributable to time-varying (L1)
factors. This provided rationale for including only L2 covariates in subsequent models.
Next, time was added as a predictor to the model. The indicator for time was the
measurement burst number, coded from -3 to 0 so that each subsequent measurement burst was
coded with a successively higher number, and the final measurement at the end of the study
occurred when time was equal to zero. This model revealed a significant fixed effect estimate of
time (b= 0.493 (SE: 0.044), p< 0.0001), indicating a positive linear effect of time on the slope of
change in BMI. For every one-unit increase in time (one measurement burst), the mean increase
in BMI was nearly half of a unit (0.5 kg/m
2
).
The final model testing the effect of baseline within-subject mean sleep duration on the
rate of change of BMI (slope) across 1.5 years and attained BMI (intercept) at the end of the
study is presented in Table 14. Results revealed that the attained BMI for a child for whom all
covariates were equal to zero or the group mean was 22.96 kg/m
2
. Baseline within-subject mean
sleep duration was not significantly associated with the rate of change in BMI across 1.5 years
(b= -0.001 [SE: 0.001], p= 0.753) or with level of attained BMI at 1.5 years (b= -0.006 [SE:
0.007], p= 0.69). Controlling for all predictors in the model, at each successive measurement
point children’s BMI significantly increased (b= 0.505 [SE: 0.047], p< 0.0001). Higher age was
associated with significantly higher attained BMI (b= 0.789 [SE: 0.373], p= 0.036), and maternal
college education was associated with significantly lower attained BMI (b= - 2.671 [SE: 0.68],
p< 0.0001) at 1.5 years. There were no differences in attained BMI by sex or by pubertal status.
Table 15 shows the results of the final model testing the effect of baseline within-subject

133
variability in sleep duration on the rate of change of BMI (slope) across 1.5 years and attained
BMI (intercept) at the end of the study, controlling for baseline within-subject mean sleep
duration. There was no significant effect of baseline sleep duration variability on the rate of
change in BMI across 1.5 years (b= - 0.002 [SE: 0.002], p= 0.296) or on the level of attained
BMI at 1.5 years (b= 0.005 [SE: 0.013], p= 0.688). Lower baseline child age and maternal
college education were associated with significantly lower attained BMI, as in the previous
model.
The final model testing the effect of baseline sleep risk factor score on the rate of change
of BMI (slope) across 1.5 years and attained BMI (intercept) at 1.5 years is presented in Table
16. There was a positive but non-statistically significant effect of baseline sleep risk factor score
on attained BMI level at the end of the study (b= 0.569 [SE: 0.396], p= 0.153), indicating a trend
toward a higher sleep risk factor score contributing to elevated attained BMI. However, there
was not a significant effect of baseline sleep risk factor score on the rate of change of BMI across
the study (b= 0.028 [SE: 0.049], p= 0.565). As in the previous models, child age and maternal
education were significantly associated with attained BMI.
Discussion

Increasing rates of childhood obesity represent a major public health issue, especially
because obesity in youth tends to persist into adulthood, producing negative effects throughout
the lifespan. Currently, one in every three youth are classified as overweight or obese, and these
youth are at risk for developing high blood pressure, high cholesterol, and impaired glucose
tolerance- health issues which are implicated in cardiovascular disease and type 2 diabetes.
While energy balance behaviors (i.e., activity and eating) are primary contributors to weight
status, obesity is a complex, multifactorial disease with numerous contributors at various levels,
including the environment, family, and individual. Recent evidence has suggested that sleep,

134
which is essential for cognitive development, learning and memory, and mood and emotion
regulation, (Baum et al., 2014) may also act as a risk or protective factor for obesity in youth.
In light of the importance of sleep health for obesity outcomes in youth, the present study
tested the association of 7-day within-subject mean and variability in baseline sleep duration on
the rate of BMI change across the 1.5-year study period, and the level of attained BMI at follow
up. This was the first known study to examine the longitudinal effects of within-subject
variability in sleep duration on the rate of BMI change across time. Results revealed that,
regardless of sleep duration characteristics, children’s BMI increased by approximately 0.5 units
each 6 months, for a total average increase of approximately 1.5 kg/m
2
across the study period.
Contrary to hypotheses, there was no effect of baseline within-subject mean sleep duration on
either the rate of BMI change or the level of attained BMI at study close. Nor was there a
significant effect of baseline within-subject variability in sleep duration on rate of BMI change or
attained BMI. Though a higher sleep risk factor score was marginally associated with higher
BMI at baseline, there was also no significant effect of baseline sleep risk factor score on BMI
growth or attained value at 1.5 years.
The result revealing no significant effect of baseline within-subject mean sleep duration
on weight is surprising, given the strong effects observed for short mean sleep duration and
elevated risk of obesity across the literature. (Cappuccio et al., 2008; Fatima et al., 2015; M.
Miller et al., 2018) One potential explanation for the null findings for within-subject mean sleep
duration on BMI is the overall adequate sleep duration reported by the youth in this sample. As
compared to national averages, in which a sizable subgroup of children obtains less than
recommended sleep, the majority (66%) of children in the sample met the minimum
recommended guideline of 9 hours of sleep on average at baseline, and only 4 (3%) children
obtained less than 8 hours of sleep at baseline. This is in stark contrast to a recent study in

135
Mexican youth, in which 66% of children ages 9-12 and 43% of children ages 12-14 did not
meet sleep duration guidelines. (Jansen et al., 2018) Given the overall adequate sleep health
among this population, it is possible that children’s relatively good sleep health and lack of
extreme between-subject sleep values contributed to the lack of significant findings.
Although the finding for a lack of effect of baseline within-subject variability on weight
outcomes was inconsistent with study hypotheses, it is not inconsistent with previous literature,
in which effects have been mixed. Of the nine identified studies examining the effect of within-
subject variability in sleep duration on children’s weight and adiposity, only four found a
significant positive association, while the remaining half of studies found a null association.
Through nearly all of these studies have been limited by cross-sectional design, the one
longitudinal study found a non-significant prospective relationship between within-subject
variability in sleep duration and BMI two years later (b= - 0.22 [SE: 0.31]). (Kuo et al., 2014)
One potential reason for the null result is the overall level of sleep duration variability observed
in this sample. Although some children had shorter or more variable sleep duration than the
group average, these levels may still be within a normative level, and may not be expected to
have a great impact on weight status. Currently, there is a lack of consensus in regards to how
much variability in sleep is ‘normal’ or ‘healthy’. (Gooley, 2016) In the current sample, the
average within-subject variability in sleep duration was 50±32 minutes, and only one in three
children had sleep duration variability of 60 minutes or more. In contrast, the study in Mexican
youth, the mean within-subject variability in sleep duration was 78±37 minutes in youth ages 9-
12, and 84±39 min in youth ages 12-14, for an overall higher variability of about 30 minutes.
(Jansen et al., 2018) Yet even within this cohort, elevated sleep duration variability did not have
a unique effect above and beyond that of mean sleep duration to influence weight outcomes.
(Jansen et al., 2018) It is possible that the relatively lower degree of variability present is

136
‘healthy’ and thus might not have detrimental effects on weight.
Alternatively, it is possible that this study simply did not follow BMI trajectories of
children in this sample for long enough to capture the accumulated effects of poor sleep on
weight gain, which may operate over a longer period of time. In the current study, children’s
sleep health was measured at baseline and was examined in relation to the change in BMI across
1.5 years, which may be insufficient for detecting the cumulative effects that high within-subject
mean and variability in sleep may produce. In contrast, a systematic literature review of n= 7
longitudinal studies examining children’s mean sleep duration and subsequent weight gain, the
follow-up period ranged from a minimum of 3 years to a maximum of 27 years, with a median of
5.5 years follow-up. (Magee & Hale, 2012) All of the 7 studies found a significant inverse
association between sleep duration and weight gain. In fact, one study assessed sleep in 5-year-
old children and found a significant inverse association between childhood sleep duration and
adult BMI at age 32, which remained significant even after controlling for adult sleep duration as
well as other sociodemographic characteristics. (Landhuis et al., 2008) Findings from this study
suggest that childhood sleep habits can have a long-term effect on life-long health (however, in
the Magee & Hale literature review, it is important to note that sleep was reported by parents,
and not assessed across multiple days, which is a methodological limitation). Evidence of the
negative effects of within-subject sleep variability on other outcomes, such as white matter
integrity (Telzer et al., 2015) also suggests that there may be a lagged, cumulative impact of
sleep variability on subsequent health impacts. It is possible that a longer follow-period within
this sample might have revealed stronger effects.
Contrary to hypotheses, this study did not reveal a combined effect of short mean
duration and high variability in duration on the rate of growth or attained level of BMI. Children
with a sleep risk factor score of 4 are those with short within-subject mean duration (<9 hours),

137
and high within-subject variability in duration (>50 min). These children are likely those who
obtain substantially less than the recommended 9 hours of sleep during the week on school
nights, but who deviate greatly from this weekday trend by sleeping in on the weekends to
‘catch-up’ on lost sleep, contributing to high variability across weeks. This pattern of chronic
sleep deprivation during work or school days, and sleep extension on free days, is characteristic
of ‘social jetlag’. which refers to the misalignment between internal body clock and external time
experienced by travelers to a different time zone. Adolescents also commonly extend their sleep
(i.e., ‘sleeping in’) on weekends, when schedules are less constrained by academic demands, in
order to compensate for the sleep debt that accumulated across the week, resulting in social
jetlag. In this way, elevated sleep duration variability in this sample may actually reflect
weekend catch-up sleep, (Carskadon et al., 1993; Malone et al., 2016) which has recently been
suggested as a protective against obesity among youth. For example, in one sample of Chinese
youth, those did not compensate for short weekday sleep by extending sleep on weekends and
holidays had greater odds of overweight or obesity. (Kim et al., 2012) One potential reason for a
protective effect is that this compensatory weekend catch-up sleep contributes to higher overall
time spent in Stage 3, during which growth hormones are released, a process thought to be
protective against weight gain. (Lewitt, 2017)
The present analysis used repeatedly assessed sleep duration across one week to
determine each child’s overall degree of sleep duration variability. However, findings from some
studies in children suggest that it may be important to consider not just the overall level of
variability present across the week, but when and to what degree the variability is present. For
example, in a sample of n=308 children ages 4-10, researchers found a complex pattern of
within-subject sleep duration variability by weight status: children with the shortest mean and
highest variability in sleep duration had the poorest obesity and metabolic outcomes. Whereas

138
normal weight children were characterized by stable weeknight sleep duration and longer sleep
duration on weekend nights, obese children were characterized by within-weeknight variability
in sleep duration that increased throughout the week, with greatest variability on weekends.
Overweight children were characterized by yet another variability pattern, with increasing sleep
duration throughout the week and longest duration on weekend nights. (Spruyt et al., 2011)
Although limited by cross-sectional design, this pattern of findings suggests a dose-response like
relationship between sleep variability and adiposity. Others have also singled out within-week
(i.e., Sunday – Thursday) variability in sleep as more harmful for overall health and well-being
than the commonly observed weekend-to-weekday variability that is suggested to be normative
and compensatory for social and academic demands. (Fuligni & Hardway, 2006) These findings
point to a future direction of classifying not just the overall degree of variability present, but also
to classifying children by different patterns of variability across the week. For example, children
can be classified based on the degree of weekday-to-weekend day difference in duration, or the
degree of variability present during the school week only, to determine whether these different
variability subtypes have stronger ties to weight gain and obesity risk than others.
Limitations

This study used a repeated self-report EMA sleep measure to determine children’s
baseline within-subject mean and variability in sleep duration. As with all self-report measures,
the EMA sleep assessment may be subject to self-report biases, such as retrospective bias, in
comparison to more objective approaches for assessing sleep health, including accelerometers.
(Shiffman et al., 2008) However, preliminary analyses of the EMA sleep assessment compared
to wrist-worn actigraphy within a subset of the present study population demonstrated moderate
to high correlation for EMA and Actigraphy-assessed sleep duration (r= 0.71, p< 0.001) and
sleep duration variability (r= 0.44, p< 0.01). There were also no significant differences between

139
EMA and Actigraphy for determining within-subject mean sleep duration (mean difference: -
13.8 min, p= 0.090) or with within-subject variability in sleep duration (mean difference: 8.54
min, p= 0.190). (O’Connor, unpublished) These findings suggest that EMA is an acceptable
alternative for measuring day-to-day sleep health among youth, with comparable estimates of
sleep characteristics as compared to actigraphy. However, while the EMA approach may garner
more ecologically valid results than the traditional retrospective, survey-based approach, it is
unknown how children’s reporting of sleep duration in the study week may compare to their
typical levels of sleep duration and its variability. If participation occurs in a non-typical week,
estimates could be non-representative of children’s usual sleep health at baseline.
A limitation of the current study is reliance on BMI as the primary measure of children’s
adiposity. Although the repeated assessment of children’s BMI by trained research assistants
across 1.5 years is a major strength, the use of BMI as an indicator of adiposity is not without
limitation. The accuracy of BMI in capturing underlying adiposity (i.e., obesity risk) differs by
children’s body fat percentage, and may be less accurate for children with lower body fat.
(Freedman, 2008) Nevertheless, estimates from a global analysis of over 7,000 children ages 9-
11 find that BMI is highly correlated (i.e., r > 0.90) with total body fat, and thus this limitation is
minimal. (Katzmarzyk et al., 2015) Future studies might consider the use of alternative body
composition measures, such as waist circumference or waist-to-height ratio, dual x-ray
absorptiometry (DXA), fat mass (FM) or percent body fat (% BF) to determine children’s
adiposity levels in relation to sleep duration and variability.
Among the current sample, BMI status and its change over time was highly influenced by
higher-level, or L2 (person-level) factors, as opposed to lower-level, or L1 (measurement burst-
level) factors. In the current study, the intra-class correlation (ICC) for children’s BMI across the
study period was .95, indicating that 95% of the variance in BMI attributable to the participant

140
level and less than 5% attributable to time-varying, or measurement burst-level factors. Given
the numerous child-, family-, community-level factors that may impact both sleep health and
obesity risk, (Brown et al., 2016; Moore et al., 2011) it is possible that not all relevant L2
covariates were accounted for, which is a limitation of the current study.
In contrast to previous cross-sectional studies, the current study used longitudinal design
to estimate the effect of sleep on the change in BMI and attained BMI across the study period.
However, despite the strength of this approach, there are also some limitations. By using baseline
sleep to determine each child’s overall sleep health, it is possible that this study did not capture
the full effect of sleep duration and variability on the change. For example, this study did not
examine how children with greater rates of change (slope) in within-subject mean or variability
in sleep duration across the study period may have also experienced greater rates of BMI change.
It may be that children whose mean duration or variability in duration continues to increase
across study period are at greatest risk for excessive weight gain. Future studies may consider
using a path modeling approach to estimate both the intercept and the slope for each variable of
interest (within-subject mean duration, within-subject variability in duration, and BMI) at each
measurement burst, and using the slopes and intercepts themselves to predict other outcomes.
This would allow one to answer questions such as: “Are the rates of change of within-subject
mean and variability in sleep duration correlated?” or, “Does the rate of change of within-subject
variability in duration predict the rate of change of BMI across the study period?”.
Another potential limitation of the present study is missing sleep and BMI data. Previous
EMA literature has revealed that non-compliance and missingness is often attributable to other
factors, either measured or unmeasured, and this systematic noncompliance is a threat to validity.
For example, children who are more stressed than other children may be less likely to be
compliant with EMA surveys; while stress is linked to compliance, it may also be linked to

141
poorer weight-related behaviors and elevated obesity risk. In the present analysis, imputation was
not employed, which is a limitation. However, as the amount of missing data was low (95% of
children were included in the analytical sample, and the average EMA compliance rate 79%) it is
unlikely that missing data significantly altered the effect estimates.
Finally, although this analysis is strengthened by its longitudinal design, there is still a
possibility of reverse causation or a bi-directional association which was not explored. Youth
with overweight and obesity may have poorer sleep health than youth with normal weight, due to
chronic sleep restriction stemming from obesity-related sleep disordered breathing, such as
obstructive sleep apnea. Though substantial longitudinal evidence suggests that insufficient sleep
confers increased risk of obesity, there is also evidence that excess weight and adiposity may
also lead to sleep disruptions. (Vgontzas et al., 2010)
Implications

The present study is one of the first to use a longitudinal design to examine the role of
sleep duration variability on children’s rate of BMI change and attained level of BMI across a
period of approximately one and a half years. Although there is strong evidence in the literature
for an association between short sleep duration and elevated risk of obesity, the present study
found a non-significant association between both within-subject mean and within-subject
variability in sleep duration as well as a sleep risk factor score on children’s weight status.
Though there was no evidence for an effect of baseline sleep health on rate of BMI change or
attained BMI, there was some evidence for a trend toward higher initial BMI among children
with shorter sleep duration and elevated variability at baseline. Children with short baseline
duration (<9 hrs) had non-significantly higher BMI (short duration: 20.5 kg/m
2
vs. adequate
duration: 19.2 kg/m
2
; p= 0.12), that children with high baseline within-subject variability had
non-significantly higher BMI (high variability: 20.15 kg/m
2
vs. low variability: 19.38 kg/m
2
),

142
and that children with one or two sleep risk factors had trending toward significantly higher BMI
at baseline than children with no risk factors (F-statistic: 3.01, p= 0.053). Taken together, these
findings suggest that higher BMI and shorter/more variable sleep duration may be related,
though this is not statistically supported within in the current study.
Future Directions

One potential direction for this field is the systematic review and interpretation of results
for the association of within-subject variability in sleep duration and adiposity outcomes in
children and adolescents to date. Understanding the state of the literature for the relationship
between variability in sleep duration and risk of obesity among youth can help to (1) illuminate
the most commonly used methodologies, statistical approaches, definitions; and (2) determine
overall trends in findings across studies; and (3) uncover any person-level or study-level
moderators of the association (e.g., different effects for boys vs. girls, or in cross-sectionally vs.
prospective analyses).
As mentioned in the previous section, future studies should also consider children by
different patterns of variability across the week; aside from the degree of weekday-to-weekend
day difference in duration, other aspects such as the variability present during the school week
only, or the mean difference between weekday and weekend sleep health, may be more strongly
linked to weight gain and obesity risk. Additionally, given the dearth of longitudinal studies of
the unique effect of within-subject variability in sleep duration on youth weight trajectories,
future studies should continue to examine this effect, ideally with longer (i.e., >3 years) follow-
up period.
Conclusions

In a sample of healthy youth with overall adequate sleep duration, the present study
found no significant longitudinal association of baseline within-subject mean sleep duration,

143
sleep duration variability, or combined effect of mean and variability in duration on the rate of
change of BMI or attained BMI 1.5 years later. Future studies may consider examining the role
of within-subject variability in youth sleep on more proximal indicators of obesity (e.g., insulin
sensitivity, high cholesterol), or on alternative indicators of adiposity (e.g., % body fat, waist-to-
height ratio).

144

Figure 5. Hypothesized effects of within-subject mean and variability in sleep duration on overall
sleep health

Within-Subject
Mean Sleep Duration
Adequate Mean Duration
(≥9 hrs)
Short Mean Duration
(<9 hrs)
Degree of Within-
Subject Variability in
Sleep Duration
Low Variability
(< 50 min)
Ideal Sleep
(Adequate Mean,
Low Variability)
Mixed Sleep
(Unhealthy Mean,
Low Variability)
High Variability
(≥ 50 min)
Mixed Sleep
(Healthy Mean,
High Variability)
Poorest Sleep
(Unhealthy Mean,
High Variability)

145
Table 12. Baseline and follow-up characteristics for 153 children enrolled in the MATCH Study

Baseline Follow-up

M/N SD/% M/N SD/%

Age 10.54 0.91 11.97 1.03
Male 67 43.79 68 44.74
Hispanic
a
90 59.21 88 58.28
White 67 44.37 67 44.67
Pubertal Score
a

Pre-Pubertal
34 24.46 21 14.69
Early Pubertal
30 21.58 25 17.48
Mid Pubertal
50 35.97 48 33.57
Late Pubertal
12 8.63 13 9.09
Post Pubertal
13 9.35 36 25.17
BMI
a,b
19.67 4.20 21.25 4.88
BMI Category
a,c

Normal weight 95 64.00 98 64.00
Overweight 29 19.00 24 16.00
Obese 25 17.00 30 20.00
Household Income
a

≤ $35,000
40 26.67 32 21.62
35,001 - 75,000
40 26.67 43 29.05
75,001 - 105,000
29 19.33 29 19.59
> $105,000
41 27.33 44 29.73
Mother Married
a

102 68.00 98 65.33
Mother College
a

85 58.22 86 59.31
Note: Baseline refers to the 3rd measurement burst, and follow-up refers to the 6th burst
a
data missing on variable
b
BMI=body mass index (kg/m2)
c
BMI Categories calculated according to CDC cutoffs for children

146
Table 13. Baseline descriptive statistics for 7-day EMA sleep among N=153 MATCH Children

M/N SD/%

Duration (min)
a
558.41 44.67
<480 4 2.68
480 - 539 46 30.87
540 - 599 71 47.65
600 - 559 26 17.45
≥660 2 1.34
Variability (min)
b
49.84 31.82
< 15 11 7.59
15 - 29 24 16.55
30 - 44 44 30.34
45 - 59 26 17.93
> 60 40 27.59
Low Duration
c
50 33.56
High Variability
d
62 42.76
Sleep Risk Factor Score
e

No risk factors (1) 56 38.62
One risk factor (2) 68 46.90
Two risk factors (4) 21 14.48
Note: Baseline refers to the 3rd measurement burst

a
sleep duration (min) is the mean daily elapsed duration between EMA-reported sleep and wake
time
b
sleep variability (min) is the standard deviation of sleep duration across daily EMA-reported sleep
c
low duration (0= no; 1=yes) defined as mean sleep less than 550 minutes (9 hours)
d
high variability (0=no; 1=yes) defined as sleep variability above the group mean (50 min)
e
sleep risk factor is the product of 'low duration' and 'high variability' variables; sleep score of 1
indicates adequate duration and low variability; sleep score of 2 indicates either short duration or
high variability, and 4 indicates both short duration and high variability.

147

Figure 6. Spaghetti plot of average (red) and child-specific (black) trajectories of BMI across
measurement bursts 0 - 3, for children with high (left) and low (right) baseline within-subject
mean sleep duration

148

Figure 7. Spaghetti plot of average (red) and child-specific (black) trajectories of BMI across
measurement bursts 0 - 3, for children with high (left) and low (right) within-subject variability
in baseline sleep duration.

149
Table 14. Results of multilevel growth curve model showing the effects of baseline sleep
duration on BMI growth (slope) and attained BMI (intercept)
a
at the end of the study

95% CI
Fixed effects Estimate SE Lower Upper p-value
Intercept (attained BMI)
a
22.960 0.690 21.595 24.324 <0.0001
Time
b
0.505 0.047 0.413 0.597 <0.0001
Sleep Duration
c
-0.006 0.007 -0.022 0.007 0.690
Duration*Time
b,c,d
-0.001 0.001 -0.002 0.002 0.753
Sex
e
-0.562 0.68 -1.906 0.783 0.410
Age
c
0.789 0.373 0.052 1.527 0.036
Maternal Education
f
-2.671 0.659 -3.974 -1.367 <0.0001
Pubertal Status
g
0.397 0.538 -0.668 1.462 0.462

Note: Level 2 sample size is N=153 participants; BMI=body mass index (kg/m2)
a
Intercept represents effect on attained BMI at the final study assessment (1.5 years after
baseline)
b
Time is coded as baseline=-3; 6 months=-2; 1 year=-1; 1.5 years=0.
c
Sleep duration (min) and age (years) are grand-mean centered
d
Sleep Duration*Time is an interaction term, representing the effect of baseline sleep on rate
of BMI change across time
e
Child sex is coded as Male=1, Female=0
f
Maternal education coded as college graduate=1 vs. non-college graduate=0
g
Pubertal Status is coded as pre or early puberty=0; mid, late, or completed puberty=1
*p<0.05; **p<0.01, ***p<0.001

150
Table 15. Results of multilevel growth curve model showing the effects of baseline sleep
duration variability on BMI growth (slope) and attained BMI (intercept)
a
at the end of the study

95% CI
Fixed effects Estimate SE Lower Upper p-value
Intercept (attained BMI)
a
23.026 0.712 21.617 24.435 <0.0001
Time
b
0.503 0.048 0.409 0.598 <0.0001
Sleep Duration Variability
c
0.005 0.013 -0.022 0.033 0.688
Variability*Time
b,c,d
-0.002 0.002 -0.005 0.002 0.296
Sleep Duration
c
-0.008 0.007 -0.023 0.007 0.290
Sex
e
-0.531 0.711 -1.937 0.875 0.456
Age
c
0.786 0.379 0.035 1.536 0.04
Maternal Education
f
-2.809 0.672 -4.139 -1.479 <0.0001
Pubertal Status
g
0.459 0.554 -0.636 1.554 0.409

Note: Level 2 sample size is N=153 participants; BMI=body mass index (kg/m2)
a
Intercept represents effect on attained BMI at the final study assessment (1.5 years after
baseline)
b
Time is coded as baseline=-3; 6 months=-2; 1 year=-1; 1.5 years=0.
c
Sleep duration variability (min), sleep duration (min), and age (years) are grand-mean
centered
d
Variability*Time is an interaction term, representing the effect of baseline sleep variability
on rate of BMI change across time
e
Child sex is coded as Male=1, Female=0
f
Maternal education coded as college graduate=1 vs. non-college graduate=0
g
Pubertal Status is coded as pre or early puberty=0; mid, late, or completed puberty=1
*p<0.05; **p<0.01, ***p<0.001

151
Table 16. Results of multilevel growth curve model showing the effects of baseline sleep risk
factor score on BMI growth (slope) and attained BMI (intercept)
a
at the end of the study

95% CI
Fixed effects Estimate SE Lower Upper p-value
Intercept (attained BMI)
a
22.031 1.050 19.927 24.079 <0.0001
Time
b
0.451 0.105 0.245 0.657 <0.0001
Sleep Risk Factor Score
c
0.569 0.396 -0.214 1.352 0.153
Sleep Score*Time
b,c,d
0.028 0.049 -0.068 0.125 0.565
Sex
e
-0.695 0.695 -2.070 0.681 0.319
Age
f
0.786 0.378 0.037 1.535 0.038
Maternal Education
g
-2.748 0.673 -4.078 -1.417 <0.0001
Pubertal Status
h
0.400 0.544 -0.675 1.476 0.463

Note: Level 2 sample size is N=153 participants; BMI=body mass index (kg/m2)
a
Intercept represents effect on attained BMI at the final study assessment (1.5 years after
baseline)
b
Time is coded as baseline=-3; 6 months=-2; 1 year=-1; 1.5 years=0.
c
Sleep risk factor score is the product of high variability (1=no, 2=yes) and low duration
(1=no, 2=yes), resulting in the following possible scores: no risk factors=1; one risk
factor=2; two risk factors=4.
d
Sleep Score*Time is an interaction term, representing the effect of baseline sleep score
on rate of BMI change across time
e
Child sex is coded as Male=1, Female=0
f
Age (years) is grand-mean centered
g
Maternal education coded as college graduate=1 vs. non-college graduate=0
h
Pubertal Status is coded as pre or early puberty=0; mid, late, or completed puberty=1
*p<0.05; **p<0.01, ***p<0.001

152
CHAPTER 5: DISCUSSION AND CONCLUSIONS

Summary of Findings

This dissertation includes three interrelated studies that leveraged real-time data capture
methodologies to explore the relationships among sleep health and its variability across days,
patterns of dietary quality and eating behavior, and weight trajectories among youth. In brief,
these studies sought to understand: (1) the agreement of EMA-assessed within-subject mean,
variability, and day-to-day sleep health as compared to actigraphy, as well as moderators of the
day-level agreement; (2) the daily effects of sleep duration and timing on next-day dietary
quality and eating behavior; and (3) the longitudinal relationship of baseline within-subject mean
and variability in sleep duration on rate of change in body mass index (BMI) across 1.5 years and
attained BMI at follow-up.
Study 1 was the first known study to examine the agreement of self-report EMA sleep
health assessment with objective actigraphy sleep. It was also the first study to examine different
moderators (e.g., sex, pubertal development status, body mass index [BMI] category, and day of
the week) of agreement at the day-level, in comparison to previous studies which examined
agreement of mean levels across groups. This study provides preliminary evidence that EMA
sleep assessment is an acceptable alternative to actigraphy for measuring day-to-day sleep health
among free-living youth within their natural environments. This study found that the majority of
EMA sleep variables (i.e., bedtime, wake time, duration) had moderate-to-high agreement with
actigraphy for estimates of within-subject mean (i.e., the average values across days) and
variability (i.e., the standard deviation across days) of duration and sleep timing that were not
significantly different between the two methods. However, EMA estimates of wake-after-sleep
onset (WASO) were not in agreement with actigraphy. The differences in estimation of WASO
were likely due to a combination of misclassification of children’s limb movement as of ‘awake’

153
time in actigraphy and difficulty in ascertaining aspects of within-night sleep quality through
EMA. The innovative day-level analysis revealed good daily agreement between the two
methods: on days when actigraphy detected higher levels of each of the four sleep variables (i.e.,
bedtime, wake time, duration, WASO) than a child’s own mean level, the child’s EMA self-
report of that same sleep variable was also significantly higher. Further, with minor exceptions
(e.g., higher WASO agreement for males and earlier pubertal development) this day-to-day
agreement did not differ by the participant’s sex, pubertal development, BMI category, or by the
day of the week (weekend vs. week day). Overall, this study provides strong evidence that EMA
sleep assessment can garner estimates of children’s sleep that are in agreement with actigraphy,
indicating that EMA may be used as a cost- and resource-effective alternative to actigraphy in
studies measuring day-to-day sleep.
In Study 2, the EMA sleep assessments were used to disentangle the daily vs. usual
effects of sleep duration and bedtime on next-day dietary quality and eating behaviors. This was
the first known study to examine the relationship between children’s sleep health and the Healthy
Eating Index (HEI-2015), an overall dietary quality score that was calculated from repeated 24hr
dietary assessments. It was also the first known study to take a within-subjects approach to assess
the daily effects of sleep on children’s next-day dietary quality and eating behaviors. This study
revealed that on nights when children slept longer than their own average duration, they were
nearly one and a half times more likely to consume breakfast the following day. This finding
suggests that daily deviations from one’s usual sleep duration may play a role in eating behaviors
the next day. Additionally, Study 2 revealed consistent between-subject effects, such that
children with longer mean sleep duration and earlier mean bedtimes than their peers had higher
overall HEI-2015 scores, including lower intake of ‘moderation’ items, such as foods and
beverages with added sugars, refined grains, and higher intake of ‘adequacy’ items, including

154
fruits and vegetables. The finding that overall sleep health impacts overall dietary quality
provides additional rationale for targeting sleep within obesity prevention programs.
In Study 3, the relationship of sleep and obesity was examined by assessing the effect of
baseline within-subject mean and variability in sleep duration on the rate of body mass index
(BMI) change across 1.5 years, and level of attained BMI at the end of the study. This was the
first known study to examine the longitudinal effects of within-subject sleep duration variability
on the rate of BMI change across time. It was also one of the first studies to assess how the
interplay of short mean sleep duration (< 9 hrs.) and high mean variability in sleep duration
(above the group mean) may affect BMI over time. Although children’s BMI increased by
approximately half a unit at each measurement burst, this study determined that there was no
significant effect of baseline within-subject mean or within-subject variability in sleep duration
on BMI, nor was there a combined effect of mean and variability in duration on BMI outcomes.
The lack of association of within-subject mean sleep duration on weight outcomes, which is
strongly established in the scientific literature, may be due to the relatively short follow-up time
in this study, whereas other studies have followed youth for several years. (Magee & Hale, 2012)
Another potential explanation is the healthy levels of overall sleep in this sample, in which the
majority of youth met sleep duration recommendations. The lack of findings for a role of within-
subject variability in sleep duration on change in BMI, though contrary to study hypotheses, is
consistent with results from five previous studies which also found null associations with youth
adiposity. (Cespedes Feliciano et al., 2018; Jansen et al., 2018; Kuo et al., 2014; McHale et al.,
2011; Park et al., 2016) For example, a recently published cross-sectional study of youth ages 9-
17 in Mexico City found that youth with insufficient/stable sleep duration had significantly
higher levels of adiposity (i.e., BMIz) than youth with sufficient/stable sleep duration, but that
there was no additional detrimental effect of insufficient/unstable (i.e., highly variable) sleep on

155
adiposity outcomes. (Jansen et al., 2018) Compared to this and other cross-sectional and
longitudinal studies of within-subject sleep duration variability, the current project was
strengthened by repeated BMI measurements, which allowed for the examination of children’s
trajectory of weight change.
Overall, this dissertation found that, while children with shorter within-subject mean
sleep duration and later within-subject mean bedtimes have poorer overall dietary quality, the
effects of sleep on diet do not necessarily occur at the day-level, and the negative impacts of poor
sleep health on diet may not translate to changes in weight gain or obesity status. Reconciling the
findings from Study 2 and 3, it appears that while sleep may be an important correlate of weight-
related behaviors (e.g., dietary quality), poor sleep may not directly impact children’s weight
status. One potential reason for the lack of finding for sleep on obesity is that obesity is a
complex, multifaceted disease that is affected by several spheres of influence. (Ebbeling et al.,
2002) Beyond a child’s individual behavior, such as sleep and diet, there are a multitude of other
factors that interact to influence weight gain and propensity for overweight and obesity. Within
the current project, these other factors may have been more salient than sleep in terms of their
overall contribution to children’s weight trajectories.
According to social ecological models, obesity risk is influenced by individual
characteristics, interpersonal/familial characteristics, household factors, neighborhood, and
societal factors; these varying levels of influence can also interact in numerous ways to influence
weight status. (Franzini et al., 2009; Glanz, Rimer, & Viswanath, 2008) Thus, although previous
childhood obesity prevention efforts have largely focused on the role of modifiable behaviors
(e.g., diet and activity), there are a number of other layers of influence that also impact children’s
weight status. For example, in the present dissertation, maternal college education, a marker of
socioeconomic status which is highly correlated with family income and two-parent household

156
status within this sample, was more highly associated with child BMI than was sleep duration:
Study 3 revealed that children with college-educated mothers had lower BMI by nearly 3.0
kg/m
2
on average, which translates to a substantial reduction in obesity risk. Beyond family
characteristics, neighborhood-level factors, such as access to green outdoor space, availability of
grocery stores or fast food outlets, crime levels, and perceived safety are also associated with
obesity risk in youth. (Brown et al., 2015; Ebbeling et al., 2002) Because the current dissertation
focused primarily on the role of sleep behavior on weight and did not incorporate the various
other levels of influence known to impact obesity, the lack of a significant effect is unsurprising.
Results from this project suggest that, despite the association of late bedtimes and short
sleep duration with children’s overall dietary patterns, improvements in youth sleep health may
not necessarily produce positive impacts on dietary quality. Similarly, improvements to dietary
quality may not translate to favorable impacts on weight status. This is because previous obesity
prevention and interventions programs with a narrow focus on behavioral targets that are only
modestly associated with weight status, such decreasing intake of dietary fat (vs. decreasing
overall caloric intake), or increasing planned physical activity (vs. increasing overall lifestyle
activity) have had limited success to date. (Ebbeling et al., 2002) For example, although many
interventions have attempted to improve diet in order to improve obesity outcomes in children,
the majority have been ineffective in producing sustained, long-term dietary improvements and
subsequent improvements in weight outcomes. (Gibson et al., 2006) Beyond the single-behavior
approach, there is some evidence for a positive impact of multi-behavior obesity prevention
programs, which incorporate sleep, diet, and activity into a single intervention. This is a
worthwhile area of investigation, because beyond diet quality and eating behaviors, previous
studies have suggested a link between insufficient and decreased physical activity, (Stone et al.,
2013) increased sedentary behavior, (Hale & Guan, 2015) both of which are implicated in

157
obesity risk. Thus, targeting multiple related health behaviors associated with obesity may be
more effective in preventing weight gain. Children with chronic insufficient sleep may exhibit
altered behavioral patterns consisting of low physical activity, high sedentary behavior, and poor
diet. (Felső et al., 2017) These behavioral alterations may become ingrained and habitual, and the
cumulative effects of these poor health behaviors may negatively impact obesity risk over time.
(Ievers-Landis et al., 2016) Previous laboratory studies have found that chronic, sub-sufficient
sleep duration has a cumulative, dose-response effect on overall functioning similar to the effect
of complete sleep restriction next-day functioning. (Dongen et al., 2003)
This dissertation project did not reveal a unique role of within-subject variability in sleep,
or of day-to-day deviations in sleep on weight-related behaviors and outcomes. In Study 2, day-
level deviations from one’s own usual sleep did not negatively impact next-day diet and eating
behavior. This finding is surprising, given the evidence for the effect of a single night of
curtailed sleep on biological processes linked to dietary alterations, such as increased ghrelin
production. (Schmid, Hallschmid, & Jauch, 2008) One potential reason for the lack of within-
subject finding in Study 2 is the use of a dietary quality indicator, as opposed to total caloric
intake, as the outcome measure. It is possible that on days following nights with poorer than
usual sleep, children might have consumed more calories without altering their pattern of eating.
Because the HEI-2015 is a reflection of dietary quality (and not of total caloric intake), an effect
of WS deviations in sleep on total caloric intake would not have been detected in the current
project. Similarly, the non-significant effect of within-subject variability in sleep duration
observed in Study 3 is surprising, given the evidence that sleep variability is linked to disrupted
glucose metabolism, (McHill & Wright, 2017) elevated psychosocial stress, (Mezick et al., 2009)
and greater negative affectivity, (Fuligni & Hardway, 2006) all of which are risk factors for
obesity. (M. Singh, 2014; Torres & Nowson, 2007) However, due to the lack of significant

158
finding for both within-subject mean (which is strongly linked in the literature) and within-
subject variability in sleep duration, it is possible that other factors, such as insufficient follow-
up time, or adequate overall sleep health (i.e., no participants with poor sleep health) contributed
to this null association.
Across the three studies, an innovative, real-time data capture approach was used to
repeatedly assess children’s sleep. While previous studies comparing self-report sleep logs to
actigraphy in youth have found a wide range of agreement, (Short et al., 2012; Tremaine et al.,
2010; Werner et al., 2008) Study 1 revealed an overall high agreement, in which children’s EMA
report of sleep was not significantly different from actigraphy for the majority of sleep variables
(such as bedtime) and conceptualizations (such as within-subject mean) tested. This EMA sleep
assessment tool was practical to deploy within the larger MATCH Study, in that it could be
added into the existing EMA protocol with minimal additional financial investment (e.g.,
purchasing equipment), or additional practical burden to participants (e.g., wearing an additional
device, remembering to complete paper logs each day). The EMA sleep measures allowed for the
determination of several sleep health conceptualizations, including within-subject mean and
variability, as well as day-to-day sleep health. Although traditional sleep logs can provide similar
information at a low cost, EMA provides additional benefits, including minimal disruption of
daily routines, customized prompting schemes, and limited sampling windows with time stamps
for completed sleep reports, which allow researchers to verify that sleep information is reported
with specified windows, potentially decreasing recall bias. (Shiffman et al., 2008) Although not
leveraged in the current study, EMA sleep assessment may also be uniquely suited for studies
that seek to use day-to-day sleep health information to tailor behavioral sleep interventions,
which is not feasible with traditional paper sleep logs.
Figure 8, below, presents the full conceptual model for this dissertation project. This

159
model builds upon the simplified conceptual model presented in Figure 2 of the Introduction. In
the full conceptual model, the role of sleep, diet, and weight gain are placed into a larger context,
with hypothesized links among the various levels of influence thought to influence children’s
long-term obesity risk. In this model, distinct aspects of children’s sleep health may have
downstream effects on biological, cognitive, psychosocial, and behavioral processes underlying
weight gain. These effects may unfold at the day-level (i.e., sleep on a given night affects activity
levels the next day), in a cumulative manner (i.e., a week of insufficient sleep can produce a
short-term [e.g., one week] elevation in ghrelin production), or at the usual level (i.e., habitually
late bedtimes are associated with elevated overall snack food intake). Over time, these varying
levels of sleep-related alterations may lead to a positive energy balance and weight gain that
leads to overweight or obesity and ultimately increases the risk for chronic diseases, including
type 2 diabetes, cardiovascular disease, and certain cancers. This model also includes various
child, family, and environmental characteristics known to affect sleep, behavior, and obesity risk,
which are important to consider as moderators or independent risk factors in future studies. This
model posits that over time, poor sleep health may directly or indirectly lead to elevated weight
gain and risk of obesity and other chronic diseases; however, this model also provides insight
into potential intervention targets, to improve sleep or buffer the effects of poor sleep on
downstream processes.

160

Figure 8. Overall conceptual model of the relationships among sleep health and variability, alterations to biological, cognitive,
psychosocial, and behavioral processes, increased adiposity, obesity and other health consequences among youth.

DISCUSSION & CONCLUSIONS

161
161
Overall Limitations

In addition to the study-specific limitations that were described within each of the
individual study chapters, there are also overall limitations for this dissertation. One important
finding that cuts across this body of work is the overall adequate sleep health present among this
sample. In Study 3, 66% of youth obtained the recommended 9 hours of sleep, and 95% of youth
within the sample obtained at least 8 hours of sleep. Recommendations from the various sleep
foundations and pediatric medicine agencies differ slightly. For example, the American
Academy of Sleep Medicine recommends 9-12 hours for children aged 6-12 years, and 8-10
hours for adolescents aged 13-18 years, (Paruthi et al., 2016) while the National Sleep
Foundation recommends 9-11 hours for youth 6-13, and 8-10 hours for adolescents 14-17.
(Hirshkowitz et al., 2015) The finding of high prevalence of adequate sleep is in contrast with
other samples such as a sample of Mexican youth, in which 66% of 9-12 year-old and 43% of
12-14 year-old participants did not meet sleep guidelines for sleep duration. (Jansen et al., 2018)
Similarly, the overall degree of within-subject variability of 50 minutes was fairly low in
comparison to this other cohort, in which within-subject variability was 78±37 minutes in youth
ages 9-12, and 84±39 min in youth ages 12-14. (Jansen et al., 2018). Due to the overall healthy
sleep profile among this population, it is possible that children classified as the ‘poorest’ sleepers
may still have been within a generally healthy range, contributing to the null finding in Paper 3.
Another potential limitation for this dissertation project is that children were not screened
for the presence of for sleep disorders. It is estimated that approximately 30% of US children
suffer from sleep disturbances that the child or child’s parent considers to be significant. (Stores,
2009) Additionally, clinical sleep disorders, such as obstructive sleep apnea, restless leg
syndrome, or insomnia may be undiagnosed or undetected among this population. Failure to
screen and control for sleep disturbances would result in an uncontrolled confounder in the

162
relationship between sleep and obesity in Study 3, as sleep disorders may both negatively impact
sleep health and contribute to weight-gain over time. However, if sleep disorders were present
among this population, they would be expected to bias results toward the alternative hypothesis;
as Study 3 found null results, this limitation is unlikely to have played a major role.

Implications

Methodological Implications

Many of the previous studies examining sleep, diet, and weight among youth have been
subject to various methodological limitations. The present project, which attempted to address
many of these limitations, suggests several methodological implications.
Study 1 revealed good agreement of EMA with actigraphy for assessing children’s
within-subject mean, variability, and day-to-day sleep, suggesting that EMA is an acceptable
alternative for measuring day-to-day sleep health among youth. EMA methods provide an
alternative, low-cost, and low burden methodology for free-living sleep assessment among
youth. Further, the high overall compliance rate of EMA sleep was 85%, identical to the
compliance rate for actigraphy-assessed sleep, suggesting that EMA was not overly burdensome
to participants. The finding that EMA can be used to estimate sleep health levels that are in
agreement with actigraphy is an important methodological implication for future studies seeking
to utilize this approach. Smartphone ownership among children and adolescents is at an all-time
peak, and 95% of adolescents (ages 13-17) either own or have access to a smartphone, (Pew,
2018) while children ages 8-12 increasingly have access to tablets and other devices. The
accessibility of the mobile platform for youth and the ease of reaching research participants
(especially those growing up during the digital age) thorough EMA highlights the importance of
this methodology for future sleep health research efforts.

163
Despite the utility of repeated assessment for understanding the day-to-day impacts of
sleep on other behavioral consequences (i.e., diet), the vast majority of studies to date have
aggregated repeated sleep assessments to calculate the within-subject mean (i.e., average across
all days) value of sleep variables across observations, (Weiss, Johnson, et al., 2010; Werner et
al., 2008) and have not leveraged the rich within-subject data to understand the potentially
differing within- and between- subject effects of sleep on outcomes. (Hart et al., 2013; Kjeldsen
et al., 2014) As the first known study to disentangle the within- and between-subject effects,
Study 2 yielded additional methodological insight relating to the use of multilevel models to
statistically disentangle the effects of a child’s usual sleep health from their sleep on a given
night.
The current study suggested that there was no unique effect of within-subject sleep
variability on children’s weight gain or attained BMI. In the current study, sleep variability was
conceptualized as the standard deviation of mean duration across 7 days. However, findings from
other studies in children suggest that it may be important to consider not just the overall level of
variability present in children’s sleep across a given week, but also when and to what degree the
variability is present. For example, in a sample of n=308 children ages 4-10, researchers found a
complex pattern of within-subject sleep duration variability by weight status: children with the
shortest mean and highest variability in sleep duration had the poorest obesity and metabolic
outcomes. While normal weight children were characterized by stable weeknight sleep duration
and longer sleep duration on weekend nights, obese children were characterized by elevated
weeknight variability in sleep duration that increased throughout the week and the greatest
degree of variability on weekends. Overweight children were characterized by yet another
variability pattern, with increasing sleep duration throughout the week and longest duration on
weekend nights. (Spruyt et al., 2011) Although limited by cross-sectional design, this pattern of

164
findings suggests a dose-response like relationship between sleep variability and adiposity.
Others have singled out within-week (i.e., Sunday – Thursday) variability in sleep as more
harmful for overall health and well-being than the commonly observed weekend-to-weekday
variability that is suggested to be normative and compensatory for social and academic demands.
(Fuligni & Hardway, 2006) These findings suggest the importance of carefully considering the
operationalizing of sleep health variability in future studies. For example, children might be
classified based on the degree of weekday-to-weekend day difference in duration, or the degree
of variability present during the school week only, to determine whether these different
variability subtypes have stronger ties to weight gain and obesity risk than others.
Theoretical Implications

This dissertation has several theoretical implications. Most of the existing health behavior
theories were developed using data gathered from cross-sectional or infrequent longitudinal
designs, (Hekler et al., 2016) outside of the complex real-world setting in which behavior is
enacted. (Hekler et al., 2016; Spruijt-metz et al., 2015) While these theories may be useful for
explaining how traits and behavior are related at the subject-level, they fall short when applied to
more micro-level timescales and phenomena. (Dunton, 2018) Because sleep is a high-frequency
behavior (i.e., occurs daily), static measurements are unable to fully capture its complexity.
(Dunton, 2018) However, real-time data capture techniques can allow for naturalistic assessment
of sleep and related behavior.
As the first known study to disaggregate the within- and between-subject effects of sleep
duration and bedtime on next-day dietary quality and eating, Study 2 yielded important insight
into the time scale in which sleep may impact diet. This approach allowed us to discover that,
although usual sleep duration was not related to overall breakfast intake, breakfast intake was
more likely on days following nights with longer than one’s own usual sleep duration. This

165
suggests a unique window of opportunity for intervening upon children’s breakfast consumption,
which may in turn have positive effects on overall dietary quality. (Medin et al., 2019) In
contrast, Study 2 found only a between-subjects effect for the role of sleep duration and timing
on overall dietary quality, suggesting a lack of effects of a given night’s sleep on the quality of
diet consumed the next day. This suggests that the effects of poor sleep may pool or accumulate
over time to negatively impact child behavioral outcomes. The finding that the effects of poor
sleep on diet may not operate at the daily level is important for theory development, because it
suggests that the pathways leading from sleep to diet may play out at day-level. Instead,
cumulative alterations to biological, cognitive, or psychological processes may in time lead to
change in diet. Future studies can continue to use repeated sleep assessments and within-subject
analytical approaches to determine on what timescale sleep may impact diet, as well as the
mechanisms underlying this association. For example, future studies should seek to determine
different time scales for the effect of sleep on children’s diet and eating, including cumulative
effects, where poor sleep on a given night may trigger short-term (e.g., a few days) alterations in
biological or behavioral trajectories; or lagged effects, where sleep on a given day is related to
diet consumed a few days later.
When sleep duration and bedtime were included in a single model in Study 2, bedtime
(not duration) was inversely associated with overall dietary quality. This suggests that bedtime
might play an even more important role on weight-related behaviors that total sleep duration.
This is important for theory development, as it suggests a role of not just overall time spent
sleeping but the time of day in which one spends sleeping, for obesity-related outcomes. It may
be that children with habitually late bedtimes, either due to late chronotype or later stage of
pubertal development (in which sleep phase becomes delayed), may have unique dietary
alterations that accumulate over time. For example, youth with late usual bedtimes may both

166
consume more food in the evenings (due to longer time spend awake at the end of the day) and
less food in the mornings (potentially due to heightened sleepiness and less time to prepare and
consume breakfast in the morning). This pattern of eating has implications for overall dietary
quality, as breakfast tends to be high in fruits, fiber, and nutrients, while foods consumed in the
evening after dinner are more likely to be high-fat, high-sugar snack foods. Although Study 3
focused on the effects of sleep duration and its variability on weight trajectory (which was
selected due to the stronger state of evidence for duration on obesity risk compared to bedtime),
it is possible that late or more variable bedtimes may play a role in children’s weight trajectories
or attained BMI, which may be examined in future studies.
Additionally, there appears to be differing effects of sleep on intake of healthy and
unhealthy foods; in combined models testing the effect of bedtime and duration on diet, late
usual bedtimes had the strongest effect on HEI-2015 Total score, and specifically on the
Adequacy sub-score. While there were initial effects on the Moderation sub-score, this was no
longer significant when adjusting for multiple comparisons. This result suggests that children
with habitually late bedtimes tend to have diet characterized by insufficient intake of healthy
foods, as opposed to excessive intake of unhealthy foods. This finding can inform theories by
emphasizing adequate intake of nutrient-dense foods, as opposed to focusing on limiting intake
of high-fat, high-sugar foods and beverages.
Intervention and Policy Implications

Several studies to date have attempted to improve child and adolescent sleep and related
behavior in the context of obesity interventions. (Yoong et al., 2016) While these studies have
demonstrated mixed success for actually enforcing a sleep improvement protocol, studies that
successfully produce a change children’s sleep health have noted improvements in biomarkers,
(Hart et al., 2013) weight-related behaviors, (Asarnow et al., 2017; Hart et al., 2013) and weight

167
status. (Hart et al., 2013) A recent literature review and meta-analysis focused on the
effectiveness of randomized controlled behavioral sleep intervention trials (RCT) for improving
children’s diet, physical activity, and weight outcomes. (Yoong et al., 2016) Of the 8 studies
identified, five targeted sleep as part of a multi-dimensional behavioral intervention, while three
were specific sleep improvement protocols. Of the three trials included in the meta-analysis
examining the effect of sleep interventions on BMI, two failed to actually improve sleep.
However, in the single RCT that was effective at modifying sleep duration (by increasing
duration by 0.75 hrs), there were positive impacts on children’s BMI between baseline and
follow-up (0.40kg/m2, p=0.05). (Yoong et al., 2016) Of note, this intervention used motivational
coaching with parents of youth ages 2-5 to improve household routines. (Haines et al., 2019)
This review highlights several key points: (1) sleep interventions for improving weight and
weight-related behaviors are still in their infancy; (2) there is some evidence for an effect of
sleep improvements on diet; and (3) fidelity and adherence to sleep improvement protocols is
challenging, yet essential for producing an effect on weight status.
Intervention settings. There are a number of settings amenable to interventions targeting
sleep health and obesity in youth. The home is an ideal setting for behavioral sleep interventions,
given the critical role of parents and other family members in influencing children’s health
behaviors, including sleep. (Agaronov, Ash, Sepulveda, Taveras, & Davison, 2018; Tikotzky,
2017) Parents have a strong influence on child sleep health, and factors such as parental
monitoring, (Gunn et al., 2018) as well as parental stress, mood, inter-parent relationship, as well
as parents’ own bedtime and sleep behaviors have been found to affect the development of sleep
health habits in childhood. (Tikotzky, 2017) Additionally, in light of the high concordance in
sleep health between parents and their adolescent children (Fuligni et al., 2015) addressing sleep
health within the context of the whole family will likely prove to be more effective.

168
Alternatively, school-based sleep interventions have been somewhat effective in improving child
sleep health. (Gruber, Somerville, Bergmame, Fontil, & Paquin, 2016; Rigney et al., 2015)
Integrating sleep promotion education into school curriculum may be more feasible than other
levels of intervention, such as policy level change of school start times, which are often met with
pushback from stakeholders. (Cassoff, Knäuper, Michaelsen, & Gruber, 2013)
Targets of intervention. In general, sleep health interventions may attempt to increase
sleep health knowledge, or improve sleep behavior. (Arora & Taheri, 2017; Cassoff et al., 2013)
Thus, programs and interventions to improve children’s sleep should focus on improving
children’s knowledge and promoting the practice of sleep hygiene. Sleep hygiene behavioral
targets include obtaining the recommended age-appropriate nightly sleep duration, maintaining
consistent sleep and wake times, limiting caffeine intake, limiting screen time before bed, and
maintaining a dark and quiet sleep environment. (Hirshkowitz et al., 2015) As this dissertation
found an effect of late bedtimes and short duration on dietary quality, the current project
provides empirical support for the importance meeting sleep duration recommendations.
However, beyond improving sleep health knowledge, it is important to consider the
multiple interacting spheres of influence on sleep behavior. (Glanz et al., 2008) For example,
interpersonal family factors (e.g., older siblings with later bedtimes, lack of household rules
regarding sleep, household chaos), social factors (e.g., perceptions of friends’ sleep habits),
community factors (e.g., neighborhood crime and safety) environmental factors (e.g., artificial
light at night [ALAN] exposure from street lights), and public policy factors (e.g., school start
times) may all interact to influence a child’s sleep behavior. This is important because early
childhood exposures influence sleep health habits, which may then persist into adulthood.
Relatedly, behavioral interventions often observe an initial effect, followed by a relapse to
ingrained behavioral patterns (e.g., staying up past bedtime to watch TV in bed). (Kwasnicka,

169
Dombrowski, & White, 2016) Thus, interventions to improve sleep must consider the important
role long-term behavior maintenance and habit formation. Factors such as individual motivations
for maintenance, enjoyment of the new behavior, availability of resources needed to maintain the
behavior, automaticity or habitualness of the new behavior, and social/environmental support of
the new behavior are key for long-term maintenance. (Kwasnicka et al., 2016)
Types of interventions. One specific type of intervention that may be an important area
of future research is the use of just-in-time adaptive interventions (i.e., JITAI). JITAIs constitute
a suite of interventions delivered through mobile technologies, that adapt to an individual’s
specific needs in real-time. (Nahum-Shani, Hekler, & Spruijt-Metz, 2015; Nahum-shani et al.,
2014) JITAIs might attempt to improve sleep health or buffer the effects of poor sleep health on
downstream behaviors (e.g., diet and eating behavior) by using information on sleep and diet to
inform the timing, content, and other aspects of an intervention. In the context of the current
dissertation findings, JITAI might be used to buffer the effects of shorter than one’s own usual
sleep on the odds of breakfast consumption the following day. To do so, JITAIs could monitor
fluctuations in day-to-day sleep health, providing real-time, tailored messaging via EMA app at
key times throughout the day based on establish sleep behavior change approaches. However, as
theory and empirical data regarding the social and cognitive influences on daily sleep health
motivations and behaviors are lacking, (Tagler, Stanko, & Forbey, 2017) considerable
preliminary work is still needed before effective JITAIs can be developed, although personalized
adaptive approaches may be important future areas of intervention. (Nahum-Shani et al., 2016)
Policy implications. Although challenging to enact, policy changes based on the growing
body of evidence for the negative impacts of insufficient sleep on children and adolescents’
overall health have the potential to positively impact a large number of individuals. (Buscemi et
al., 2017) Policy changes at the district, state, or national level have the potential to improve

170
sleep and potentially diet and weight outcomes. (Arora & Taheri, 2017) The most salient policy-
level recommendation for improving sleep and down-stream weight-related behaviors in children
and adolescents is delaying school start times to 8:30am or later, which has been advocated by
numerous national groups. (American Academy of Pediatrics, 2014; Watson et al., 2017) Aside
from changing school start times, other educational policy changes may include limiting
homework so that youth can adhere to earlier bedtimes (however, this may not be as effective for
older adolescents who exhibit delayed sleep preference and have difficulty adhering to earlier
bedtimes), (Jenni et al., 2005) or allowing youth to take short (i.e., <20 min) naps during the
school day as a way to temporarily relieve the biological drive to sleep and improve alertness in
the classroom (though, the effect of naps on nocturnal sleep health is still under investigation).
(Brooks & Lack, 2006; Jakubowski, Hall, Lee, & Matthews, 2017)
Additionally, given the mounting evidence for the importance of sleep for human health
and particularly diet and obesity risk, governmental agencies can affect large-scale change by
incorporating sleep health recommendations into existing public health guidance, thereby
increasing public awareness. This would not be unprecedented, as the US Dietary Guidelines for
Americans includes physical activity guidelines along with dietary guidelines, and emphasizes
the importance of regular physical activity in addition to a healthy diet for optimal health. (U.S.
Department of Health and Human Services and US Department of Agriculture, 2015) Raising
awareness of the importance of sleep health for overall energy balance can improve knowledge
and potentially impact sleep health attitudes and behaviors.
Future Research Directions

Future studies should consider mechanisms underlying the relationships between sleep
with diet and eating, as well as between sleep and obesity risk. Potential mechanisms include the
biological, cognitive, and psychosocial pathways outlined in Figure 8. Generally, the

171
mechanisms linking sleep with diet and eating can be categorized into homeostatic and non-
homeostatic drives to eat. (Chaput, 2014) Short sleep duration is thought to lead to increased
caloric intake through non-homeostatic drive to eat, such as through greater potential time during
the day for eating opportunities, and greater sensitivity to food rewards, which have been well-
supported in the literature. (Chaput, 2014) Additionally, homeostatic mechanisms, which include
to hunger hormones (i.e. leptin, ghrelin, cortisol) may also play an important role in the
relationship between sleep, diet/eating, and obesity. (Broussard et al., 2016; Hart et al., 2013)
Future studies should also consider the potential time scales for the WS effect of sleep on
diet. The present analysis was limited to exploring the effects with an 24hr period (i.e.,
association of a given night’s sleep on the immediately following day of diet), based on evidence
that a single night of sleep restriction alters biological processes the following day. (Donga et al.,
2010; Schmid, Hallschmid, & Jauch, 2008) However, it is possible that the effects of short
duration or later bedtime may affect dietary and eating patterns at a longer lag, such as two or
even several days later. Future studies could test varying time scales beyond one day, such as the
cumulative effects of a given night of poorer than usual sleep on diet and eating behaviors over
the span of several days following poor sleep, to determine whether there is a differential effect.
Additionally, while diet seems to be strongly linked to sleep health, previous studies have
also highlighted the relationship of children’s sleep health with physical activity (Krietsch et al.,
2016; Stone et al., 2013) and sedentary behavior. (Hale & Guan, 2016) To build upon the
findings from Study 2, future studies should consider the broader interrelationships and patterns
of children’s energy balance behaviors. These studies can be used to inform multicomponent
behavioral obesity prevention and intervention programs. (Yoong et al., 2016)
Findings from the current project suggest that future studies should explore different
developmental windows during which sleep may have the strongest effects on weight-related

172
behaviors and weight gain. Existing evidence for the relationship of short sleep duration with
obesity suggests stronger effects for children as compared to adults. (Fatima et al., 2015)
However, it is unclear whether the strength of effect differs within childhood; although
systematic reviews across age ranges appear to suggest a trend for stronger prospective
longitudinal effects of sleep on weight status among younger children (Magee & Hale, 2012) as
compared to adolescents, (Guidolin & Gradisar, 2012a) a recent systematic review by Miller et
al. (2018) found a similar effect of short sleep on elevated risk of overweight and obesity across
all stages of childhood, from infancy to adolescence. (M. Miller et al., 2018) Due to the rapidly
changing developmental stages and varying sleep needs across childhood and adolescence, there
may be specific periods of youth in which sleep health may play a stronger risk or protective
role. In the future, studies should attempt to pinpoint the timescale through which sleep may
produce effects on BMI. For example, longitudinal cohort studies with multiple intensive
measurement timepoints may be ideal settings in which to examine the different strength of
association of sleep with weight. A study of this type may reveal that sleep health at a specific
age (e.g., 5-6 years old) or developmental period (e.g., early-pubertal) may be the strongest
predictor of weight trajectory and attained BMI in adulthood; this would provide strong rationale
for designing interventions to improve sleep health during this critical period.
Another intriguing avenue of future investigation concerns children’s sleep health, energy
balance behaviors, and weight gain during school vacation. For many families, summer vacation
is akin to an extended weekend, during which time is less structured and behavioral patterns are
shifted. According to the Structured Days Hypothesis, (Brazendale et al., 2017) a lack of a
structured environment during summer vacation may have a detrimental effect on children’s
obesogenic behaviors (e.g., physical activity, sedentary screen time, diet, sleep), one potential
reason why children’s weight gain tends to accelerate over the summer months. (Brazendale et

173
al., 2017) This hypothesis is supported by a recent study of a nationally representative sample of
18,170 US children in kindergarten through second grade, which found that increases in the
prevalence of overweight and obesity occurred essentially exclusively during the summer
months, with no increases observed during the school year. (Hippel & Workman, 2016) The
MATCH Study did not assess participants over summer vacation; however, as vacation
comprises a large proportion of the year (e.g., 3 months, or ¼ of the year), it is essential to
understand the influence of sleep on weight-related behaviors and obesity risk during this time.
Understanding the role of day-to-day sleep health and variability during these less restricted (i.e.,
no early school times) months may yield new insights into the role of sleep and obesity
development or support development of behavioral interventions to be delivered during the
summer months, a potentially critical window of obesity prevention.
Conclusions

To summarize, this dissertation used real-time data capture techniques including EMA to
repeatedly assess sleep among youth and examine its relationship with dietary intake and weight
outcomes. Findings suggest that EMA is an acceptable alternative for capturing daily, mean, and
variability in sleep across nights for several key sleep health variables that have been implicated
in overall health and wellbeing. Leveraging the EMA sleep assessments, this dissertation
examined the daily and usual effects of sleep duration and bedtime on dietary quality, intake of
moderation and adequacy foods, and several aspects of eating behavior. Results revealed lower
overall dietary quality among children with short usual duration and late bedtime, and evidence
for a day-level effect of longer than usual sleep duration on next day breakfast consumption.
Finally, in a longitudinal analysis, there was no significant effect of within-subject mean or
variability in sleep duration on rate of BMI change or attained BMI at the study end; this null
finding may be attributed to a relatively short follow-up duration or to relatively healthy sleep

174
habits in the sample overall. This project adds to the current literature for the role of sleep on
youth obesity in several ways, including the use of within-subject approaches to examine the
day-level effects of sleep on diet, and the examination of children’s weight trajectories in relation
to the degree of variability in their sleep duration across nights. Results from this dissertation
supports the use of EMA to measure free-living sleep among youth, and incorporation of
behavioral sleep interventions into obesity prevention and intervention programs.

175
REFERENCES

Acebo, C., Sadeh, A., Seifer, R., Tzischinsky, O., Wolfson, A. R., Hafer, A., & Carskadon, M.
A. (1999). Estimating sleep patterns with activity monitoring in children and adolescents:
How many nights are necessary for reliable measures? Sleep, 22(1), 95–103.
https://doi.org/10.1016/0029-7844(94)00457-O
ActiLife. (2012). ActiLife 6 User’s Manual.
Adam, E. K., & Kumari, M. (2009). Assessing salivary cortisol in large-scale , epidemiological
research. https://doi.org/10.1016/j.psyneuen.2009.06.011
Agaronov, A., Ash, T., Sepulveda, M., Taveras, E. M., & Davison, K. K. (2018). Inclusion of
Sleep Promotion in Family-Based Interventions To Prevent Childhood Obesity, 14(8).
https://doi.org/10.1089/chi.2017.0235
American Academy of Pediatrics. (2014). Policy Statement: School Start Times for Adolescents.
Pediatrics (Vol. 134). https://doi.org/10.1542/peds.2014-1697
Ancoli-Israel, S., Cole, R., Alessi, C., Chambers, M., Moorcroft, W., & Pollak, C. P. (2003). The
Role of Actigraphy in the Study of Sleep and Circadian Rhythms. Sleep, 26(3), 342–392.
Anderson, B., Storfer-Isser, A., Taylor, H. G., Rosen, C. L., & Redline, S. (2009). Associations
of Executive Function With Sleepiness and Sleep Duration in Adolescents. Pediatrics,
123(4), e701–e707. https://doi.org/10.1542/peds.2008-1182
Archbold, K. H., Pituch, K. J., Panahi, P., & Chervin, R. D. (2002). Symptoms of sleep
disturbances among children at two general pediatric clinics. Journal of Pediatrics, 140(1),
97–102. https://doi.org/10.1067/mpd.2002.119990
Arora, T., Broglia, E., Pushpakumar, D., Lodhi, T., & Taheri, S. (2013). An Investigation into
the Strength of the Association and Agreement Levels between Subjective and Objective
Sleep Duration in Adolescents, 8(8), 1–6. https://doi.org/10.1371/journal.pone.0072406
Arora, T., Choudhury, S., & Taheri, S. (2015). The Relationships Among Sleep, Nutrition, and
Obesity. Current Sleep Medicine Reports. https://doi.org/10.1007/s40675-015-0030-z
Arora, T., & Taheri, S. (2015). Associations among late chronotype, body mass index and
dietary behaviors in young adolescents. International Journal of Obesity (2005), 39(1), 39–
44. https://doi.org/10.1038/ijo.2014.157
Arora, T., & Taheri, S. (2017). Is sleep education an effective tool for sleep improvement and
minimizing metabolic disturbance and obesity in adolescents? Sleep Medicine Reviews, 36,
3–12. https://doi.org/10.1016/j.smrv.2016.08.004
Asarnow, L. D., Greer, S. M., Walker, M. P., & Harvey, A. G. (2017). The Impact of Sleep
Improvement on Food Choices in Adolescents With Late Bedtimes. Journal of Adolescent
Health, 60(5), 570–576. https://doi.org/10.1016/j.jadohealth.2016.11.018
Asarnow, L. D., McGlinchey, E., & Harvey, A. G. (2015). Evidence for a Possible Link between
Bedtime and Change in Body Mass Index. Sleep, 1523–1527.
Banfield, E. C., Liu, Y., Davis, J. S., Chang, S., & Frazier-Wood, A. (2016). Poor adherence to
U.S. dietary guidelines for children and adolescents in the NHANES population. J Acad
Nutr Diet., 428(4), 709–719. https://doi.org/10.1016/j.jmb.2016.01.029
Baranowski, T., Cullen, K. W., Nicklas, T., Thompson, D., Baranowski, J., Cullen, K. W., &
Thompson, D. (2003). Are Current Health Behavioral Change Models Helpful in Guiding
Prevention of Weight Gain Efforts ?
Baron, K. G., Reid, K. J., Kern, A. S., & Zee, P. C. (2009). Role of Sleep Timing in Caloric
Intake and BMI. Obesity, 19(7), 1374–1381. https://doi.org/10.1038/oby.2011.100
Basch, C. E., Basch, C. H., Ruggles, K. V, & Rajan, S. (2014). Prevalence of sleep duration on
an average school night among 4 nationally representative successive samples of American

176
high school students, 2007-2013. Preventing Chronic Disease, 11(7), E216.
https://doi.org/10.5888/pcd11.140383
Baum, K. T., Desai, A., Field, J., Miller, L. E., Rausch, J., & Beebe, D. W. (2014). Sleep
restriction worsens mood and emotion regulation in adolescents. Journal of Child
Psychology and Psychiatry and Allied Disciplines, 55(2), 180–190.
https://doi.org/10.1111/jcpp.12125
Becker, S. P., Sidol, C. A., Van Dyk, T. R., Epstein, J. N., & Beebe, D. W. (2017).
Intraindividual variability of sleep/wake patterns in relation to child and adolescent
functioning: A systematic review. Sleep Medicine Reviews, 34, 94–121.
https://doi.org/10.1016/j.smrv.2016.07.004
Beebe, D. W., Simon, S., Summer, S., Hemmer, S., Strotman, D., & Dolan, L. M. (2013).
Dietary intake following experimentally restricted sleep in adolescents. Sleep, 36(6), 827–
834. https://doi.org/10.5665/sleep.2704
Bei, B., Manber, R., Allen, N. B., Trinder, J., & Wiley, J. F. (2017). Too long, too short, or too
variable? sleep intraindividual variability and its associations with perceived sleep quality
and mood in adolescents during naturalistically unconstrained sleep. Sleep, 40(2).
https://doi.org/10.1093/sleep/zsw067
Bei, B., Wiley, J. F., Trinder, J., & Manber, R. (2016). Beyond the mean: A systematic review on
the correlates of daily intraindividual variability of sleep/wake patterns. Sleep Medicine
Reviews, 28, 104–120. https://doi.org/10.1016/j.smrv.2015.06.003
Bel, S., Michels, N., De Vriendt, T., Patterson, E., Cuenca-García, M., Diethelm, K., …
Huybrechts, I. (2013). Association between self-reported sleep duration and dietary quality
in European adolescents. British Journal of Nutrition, 110(05), 949–959.
https://doi.org/10.1017/S0007114512006046
Belenky, G., Wesensten, N. J., Thorne, D. R., Thomas, M. L., Sing, H. C., & Redmond, D. P.
(2003). Patterns of performance degradation and restoration during sleep restriction and
subsequent recovery : a sleep dose-response study. J Sleep Res, 12, 1–12.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and
Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B
(Methodological).
Biggs, S. N., Lushington, K., van den Heuvel, C. J., Martin, a. J., & Kennedy, J. D. (2011).
Inconsistent sleep schedules and daytime behavioral difficulties in school-aged children.
Sleep Medicine, 12(8), 780–786. https://doi.org/10.1016/j.sleep.2011.03.017
Bland, J. M., & Altman, D. G. (1999). Measuring agreement in method comparison studies.
Statistical Methods in Medical Research, 8(2), 135–160.
https://doi.org/10.1191/096228099673819272
Borbély, A. A., Daan, S., Wirz-Justice, A., & Deboer, T. (2016). The two-process model of sleep
regulation: A reappraisal. Journal of Sleep Research, 25(2), 131–143.
https://doi.org/10.1111/jsr.12371
Börnhorst, C., Hense, S., Ahrens, W., Hebestreit, A., Reisch, L., Barba, G., … Bayer, O. (2012).
From sleep duration to childhood obesity—what are the pathways? European Journal of
Pediatrics, 171(7), 1029–1038. https://doi.org/10.1007/s00431-011-1670-8
Brazendale, K., Beets, M. W., Weaver, R. G., Pate, R. R., Turner-McGrievy, G. M., Kaczynski,
A. T., … von Hippel, P. T. (2017). Understanding differences between summer vs. school
obesogenic behaviors of children: The structured days hypothesis. International Journal of
Behavioral Nutrition and Physical Activity, 14(1), 1–14. https://doi.org/10.1186/s12966-
017-0555-2
Brooks, A., & Lack, L. (2006). A brief afternoon nap following nocturnal sleep restriction:

177
which nap duration is most recuperative? Sleep, 29(6), 831–840.
https://doi.org/10.1093/sleep/29.6.831
Broussard, J. L., Kilkus, J. M., Delebecque, F., Abraham, V., Day, A., Whitmore, H. R., &
Tasali, E. (2016). Elevated ghrelin predicts food intake during experimental sleep
restriction. Obesity, 24(1), 132–138. https://doi.org/10.1002/oby.21321.Elevated
Brown, C. L., Halvorson, E. E., Cohen, G. M., Lazorick, S., & Skelton, J. a. (2015). Addressing
Childhood Obesity. Pediatric Clinics of North America, 62, 1241–1261.
https://doi.org/10.1016/j.pcl.2015.05.013
Brown, C. L., Halvorson, E. E., Cohen, G. M., Skelton, J. A., Hill, C., Children, B., & Forest, W.
(2016). Addressing Childhood Obesity : Opportunities for Prevention, 62(5), 1241–1261.
https://doi.org/10.1016/j.pcl.2015.05.013.Addressing
Burrows, T. L., Martin, R. J., & Collins, C. E. (2010). A Systematic Review of the Validity of
Dietary Assessment Methods in Children when Compared with the Method of Doubly
Labeled Water. Journal of the American Dietetic Association, 110(10), 1501–1510.
https://doi.org/10.1016/j.jada.2010.07.008
Buscemi, J., Janke, E. A., Kugler, K. C., Duffecy, J., Mielenz, T. J., St. George, S. M., &
Sheinfeld Gorin, S. N. (2017). Increasing the public health impact of evidence-based
interventions in behavioral medicine: new approaches and future directions. Journal of
Behavioral Medicine, 40(1), 203–213. https://doi.org/10.1007/s10865-016-9773-3
Buysse, D. J. (2013). Sleep health: can we define it? Does it matter? Sleep, 37(1), 9–17.
https://doi.org/10.5665/sleep.3298
Calamaro, C. J., Yang, K., Ratcliffe, S., & Chasens, E. R. (2012). Wired at a Young Age: The
Effect of Caffeine and Technology on Sleep Duration and Body Mass Index in School-
Aged Children. Journal of Pediatric Health Care, 26(4), 276–282.
https://doi.org/10.1016/j.pedhc.2010.12.002
Capaldi, V. F., Handwerger, K., Richardson, E., Laura, R., & Capaldi, V. F. (2010). Associations
Between Sleep and Cortisol Responses to Stress in Children and Adolescents : A Pilot
Study Associations Between Sleep and Cortisol Responses to Stress in Children and
Adolescents : A Pilot Study, 2002. https://doi.org/10.1207/s15402010bsm0304
Cappuccio, F. P., Cooper, D., Delia, L., Strazzullo, P., & Miller, M. A. (2011). Sleep duration
predicts cardiovascular outcomes: A systematic review and meta-analysis of prospective
studies. European Heart Journal, 32(12), 1484–1492.
https://doi.org/10.1093/eurheartj/ehr007
Cappuccio, F. P., D’Elia, L., Strazzullo, P., & Miller, M. A. (2010a). Quantity and Quality of
Sleep and Incidence of Type 2 Diabetes A systematic review and meta-analysis. Diabetes
Care, 33(2), 414–420. https://doi.org/10.2337/dc09-1124.
Cappuccio, F. P., D’Elia, L., Strazzullo, P., & Miller, M. A. (2010b). Sleep duration and all-
cause mortality: a systematic review and meta-analysis of prospective studies. Sleep, 33(5),
585–592. https://doi.org/10.1111/j.1365-2869.2008.00732.x
Cappuccio, F. P., Taggart, F. M., Kandala, N.-B., Currie, A., Peile, E., Stranges, S., & Miller, M.
A. (2008). Meta-analysis of short sleep duration and obesity in children and adults. Sleep,
31(5), 619–626.
Carskadon, M. A. (1999). When worlds collide: adolescent need for sleep versus societal
demands. Phi Delta Kappan, 80(Jan), 348–353.
Carskadon, M. A., & Dement, W. C. (2011). Normal Human Sleep: An Overview. In Principles
and practice of sleep medicine (5th ed., pp. 16–26). Elsevier. https://doi.org/10.1016/B978-
1-4160-6645-3.00141-9
Carskadon, M. A., Harvey, K., Duke, P., Anders, T. F., Litt, I. F., & Dement, W. C. (1980).

178
Pubertal changes in daytime sleepiness. Sleep, 2(4), 453–460.
Carskadon, M. A., Vieira, C., & Acebo, C. (1993). Association between puberty and delayed
phase preference. Sleep.
Cassoff, J., Knäuper, B., Michaelsen, S., & Gruber, R. (2013). School-based sleep promotion
programs: Effectiveness, feasibility and insights for future research. Sleep Medicine
Reviews, 17(3), 207–214. https://doi.org/10.1016/j.smrv.2012.07.001
Cespedes, E. M., Hu, F. B., Redline, S., Rosner, B., Gillman, M. W., Rifas-Shiman, S. L., &
Taveras, E. M. (2016). Chronic insufficient sleep and diet quality: Contributors to childhood
obesity. Obesity, 24(1), 184–190. https://doi.org/10.1002/oby.21196
Cespedes Feliciano, E. M., Oken, E., Taveras, E. M., & Redline, S. (2018). Objective Sleep
Characteristics and Cardiometabolic Health in Young Adolescents, 142(1).
https://doi.org/10.1542/peds.2017-4085
Chaput, J.-P. (2014). Sleep patterns, diet quality and energy balance. Physiology & Behavior,
134, 86–91. https://doi.org/10.1016/j.physbeh.2013.09.006
Cole, R. J., Kripke, D. F., Mullaney, D. J., & Gillin, J. C. (1992). Technical Note Automatic
Sleep / Wake Identification From Wrist Activity, 15(5), 461–469.
Colrain, I. M., & Baker, F. C. (2011). Changes in sleep as a function of adolescent development.
Neuropsychology Review, 21(1), 5–21. https://doi.org/10.1007/s11065-010-9155-5
Colten, H., Altevogt, B. B. M., & Colten., H. (2006). Sleep Disorders and Sleep Deprivation: An
Unmet Public Health Problem. Committee on Sleep Medicine and Research.
https://doi.org/10.1097/01.CHI.0000270812.55636.3b
Cox, R. C., Sterba, S. K., Cole, D. A., Upender, R. P., & Olatunji, B. O. (2018). Time of day
effects on the relationship between daily sleep and anxiety: An ecological momentary
assessment approach. Behaviour Research and Therapy, 111(September), 44–51.
https://doi.org/10.1016/J.BRAT.2018.09.008
Curran, P. J., & Bauer, D. J. (2011). The Disaggregation of Within-Person and Between-Person
Effects in Longitudinal Models of Change. Annual Review of Psychology, 62(1), 583–619.
https://doi.org/10.1146/annurev.psych.093008.100356
Dahl, R. E., & Lewin, D. S. (2002). Pathways to adolescent health: Sleep regulation and
behavior. Journal of Adolescent Health, 31(6 SUPPL.), 175–184.
https://doi.org/10.1016/S1054-139X(02)00506-2
Demos, K. E., Sweet, L. H., Hart, C. N., McCaffery, J. M., Williams, S. E., Mailloux, K. A., …
Wing, R. R. (2017). The effects of experimental manipulation of sleep duration on neural
response to food cues,. Sleep.
Dietz, W. H. (1998). Health consequences of obesity in youth: childhood predictors of adult
disease. Pediatrics, 101(3 Pt 2), 518–525. https://doi.org/10.1007/s12098-011-0489-7
Dillon, H. R., Lichstein, K. L., Dautovich, N. D., Taylor, D. J., Riedel, B. W., & Bush, A. J.
(2015). Variability in self-reported normal sleep across the adult age span. Journals of
Gerontology - Series B Psychological Sciences and Social Sciences, 70(1), 46–56.
https://doi.org/10.1093/geronb/gbu035
Donga, E., Van Dijk, M., Van Dijk, J. G., Biermasz, N. R., Lammers, G. J., Van Kralingen, K.
W., … Romijn, J. A. (2010). A single night of partial sleep deprivation induces insulin
resistance in multiple metabolic pathways in healthy subjects. Journal of Clinical
Endocrinology and Metabolism, 95(6), 2963–2968. https://doi.org/10.1210/jc.2009-2430
Dongen, H. P. A. Van, Maislin, G., Mullington, J. M., & Dinges, D. F. (2003). The Cumulative
Cost of Additional Wakefulness: Dose-Response Effects on Neurobehavioral Functions and
Sleep Physiology From Chronic Sleep Restriction and Total Sleep Deprivation. SLEEP,
26(2), 117–126.

179
Douglas, N., Thomas, S., & Jan, M. (1992). Clinical value of polysomnography. The Lancet,
339, 347–350.
Driscoll, D. L., Salib, P., & Rupert, D. J. (2007). Merging Qualitative and Quantitative Data in
Mixed Methods Research : How To and Why Not. Ecological and Environmental
Anthropology, 3(1), 18–28.
Duncan, M. J., Kline, C. E., Rebar, A. L., Vandelanotte, C., & Short, C. E. (2016). Greater bed-
and wake-time variability is associated with less healthy lifestyle behaviors: a cross-
sectional study. Journal of Public Health (Germany), 24(1), 31–40.
https://doi.org/10.1007/s10389-015-0693-4
Dunton, G. F. (2018). Sustaining Health-Protective Behaviors Such as Physical Activity and
Healthy Eating. JAMA - Journal of the American Medical Association, 1–2.
https://doi.org/10.1249/mss
Dunton, G. F., Intille, S. S., Wolch, J., & Pentz, M. A. (2012). Children’s perceptions of physical
activity environments captured through ecological momentary assessment: a validation
study. Preventive Medicine, 55(2), 119–121. https://doi.org/10.1016/j.ypmed.2012.05.015
Dunton, G. F., Liao, Y., Dzubur, E., Leventhal, A. M., Huh, J., Gruenewald, T., … Intille, S.
(2015). Investigating Within-day and Longitudinal Effects of Maternal Stress on Children’s
Physical Activity, Dietary Intake, and Body Composition: Protocol for the MATCH Study.
Contemporary Clinical Trials, 43, 142–154. https://doi.org/10.1016/j.cct.2015.05.007
Duraccio, K. M., Zaugg, K., & Jensen, C. D. (2019). Effects of Sleep Restriction on Food-
Related Inhibitory Control and Reward in Adolescents, 1–11.
https://doi.org/10.1093/jpepsy/jsz008
Ebbeling, C. B., Pawlak, D. B., & Ludwig, D. S. (2002). Childhood obesity: Public-health crisis,
common sense cure. Lancet, 360(9331), 473–482. https://doi.org/10.1016/S0140-
6736(02)09678-2
El-Sheikh, M., Buckhalt, J. A., Keller, P. S., & Granger, D. A. (2008). Children’s objective and
subjective sleep disruptions: links with afternoon cortisol levels. Health Psychology :
Official Journal of the Division of Health Psychology, American Psychological Association,
27(1), 26–33. https://doi.org/10.1037/0278-6133.27.1.26
Fatima, Y., Doi, S. A. R., & Mamun, A. A. (2015). Longitudinal impact of sleep on overweight
and obesity in children and adolescents: A systematic review and bias-adjusted meta-
analysis. Obesity Reviews, 16(2), 137–149. https://doi.org/10.1111/obr.12245
Felden, É. P. G., Leite, C. R., Rebelatto, C. F., Andrade, R. D., & Beltrame, T. S. (2015). Sleep
in adolescents of different socioeconomic status: A systematic review. Revista Paulista de
Pediatria, 33(4), 467–473. https://doi.org/10.1016/j.rppede.2015.08.009
Felső, R., Lohner, S., Hollódy, K., Erhardt, É., & Molnár, D. (2017). Relationship between sleep
duration and childhood obesity: Systematic review including the potential underlying
mechanisms. Nutrition, Metabolism and Cardiovascular Diseases, 27(9), 751–761.
https://doi.org/10.1016/j.numecd.2017.07.008
Foster, R. G., & Kreitzman, L. (2017). Circadian Rhythms: A Very Short Introduction. Oxford,
England.
Franzini, L., Elliott, M. N., Cuccaro, P., Schuster, M., Gilliland, M. J., & Grunbaum, J. A.
(2009). Influences of Physical and Social Neighborhood Environments on Children’s
Physical Activity and Obesity. Am J Public Health, 99(2), 271–278.
https://doi.org/10.2105/AJPH.2007.128702
Freedman, C. D. S. (2008). The Validity of BMI as an Indicator of Body Fatness and Risk
Among Children, 124(September 2009). https://doi.org/10.1542/peds.2008-3586E
Friedlander, S. L., Larkin, E. K., Rosen, C. L., Palermo, T. M., & Redline, S. (2003). Decreased

180
Quality of Life Associated With Obesity in School-aged Children. Archives of Pediatrics &
Adolescent Medicine, 157(12), 1206. https://doi.org/10.1001/archpedi.157.12.1206
Fuligni, A. J., & Hardway, C. (2006). Daily variation in adolescents’ sleep, activities, and
psychological well-being. Journal of Research on Adolescence, 16(3), 353–378.
https://doi.org/10.1111/j.1532-7795.2006.00498.x
Fuligni, A. J., Tsai, K. M., Krull, J. L., & Gonzales, N. A. (2015). Daily concordance between
parent and adolescent sleep habits. Journal of Adolescent Health, 56(2), 244–250.
https://doi.org/10.1016/j.jadohealth.2014.09.013
Gibson, L. J., Peto, J., Warren, J. M., & dos Santos Silva, I. (2006). Lack of evidence on diets for
obesity for children: A systematic review. International Journal of Epidemiology, 35(6),
1544–1552. https://doi.org/10.1093/ije/dyl208
Giles, C. M., Cradock, A. L., Sc, D., Gortmaker, S. L., & Ph, D. (2017). Simulation of Growth
Trajectories of Childhood Obesity into Adulthood, 2145–2153.
https://doi.org/10.1056/NEJMoa1703860
Girschik, J., Fritschi, L., Heyworth, J., & Waters, F. (2013). Validation of self-reported sleep
against actigraphy. Journal of Zanjan University of Medical Sciences and Health Services,
21(88), 462–468. https://doi.org/10.2188/jea.JE20120012
Glanz, K., Rimer, B. K., & Viswanath, K. (2008). Health Behavior and Health Education:
Theory Research & Practice. https://doi.org/10.1016/S0033-3506(49)81524-1
Golley, R. K., Maher, C. a, Matricciani, L., & Olds, T. S. (2013). Sleep duration or bedtime?
Exploring the association between sleep timing behaviour, diet and BMI in children and
adolescents. International Journal of Obesity, 37(4), 546–551.
https://doi.org/10.1038/ijo.2012.212
Gong, Q. H., Li, H., Zhang, X. H., Zhang, T., Cui, J., & Xu, G. Z. (2017). Associations between
sleep duration and physical activity and dietary behaviors in Chinese adolescents: results
from the Youth Behavioral Risk Factor Surveys of 2015. Sleep Medicine, 37, 168–173.
https://doi.org/10.1016/j.sleep.2017.06.024
Gonnissen, H. K. J., Adam, T. C., Hursel, R., Rutters, F., Verhoef, S. P. M., & Westerterp-
Plantenga, M. S. (2013). Sleep duration, sleep quality and body weight: Parallel
developments. Physiology and Behavior, 1211. Gonn, 112–116.
https://doi.org/10.1016/j.physbeh.2013.04.007
Gooley, J. J. (2016). How Much Day-To-Day Variability in Sleep Timing Is Unhealthy ? Sleep,
39(2), 457–465. https://doi.org/10.5665/sleep.5424
Gradisar, M., Gardner, G., & Dohnt, H. (2011). Recent worldwide sleep patterns and problems
during adolescence: A review and meta-analysis of age, region, and sleep. Sleep Medicine,
12(2), 110–118. https://doi.org/10.1016/j.sleep.2010.11.008
Gruber, R., Somerville, G., Bergmame, L., Fontil, L., & Paquin, S. (2016). School-based sleep
education program improves sleep and academic performance of school-age children. Sleep
Medicine, 21, 93–100. https://doi.org/10.1016/j.sleep.2016.01.012
Guenther, P. M., Casavale, K. O., Reedy, J., Kirkpatrick, S. I., Hiza, H. A. B., Kuczynski, K. J.,
… Krebs-Smith, S. M. (2013). Update of the Healthy Eating Index: HEI-2010. Journal of
the Academy of Nutrition and Dietetics, 113(4), 569–580.
https://doi.org/10.1016/j.jand.2012.12.016
Guenther, P. M., Kirkpatrick, S. I., Reedy, J., Krebs-Smith, S. M., Buckman, D. W., Dodd, K.
W., … Carroll, R. J. (2014). The Healthy Eating Index-2010 is a valid and reliable measure
of diet quality according to the 2010 Dietary Guidelines for Americans. The Journal of
Nutrition, 144(3), 399–407. https://doi.org/10.3945/jn.113.183079
Guidolin, M., & Gradisar, M. (2012a). Is shortened sleep duration a risk factor for overweight

181
and obesity during adolescence ? A review of the empirical literature. Sleep Medicine,
13(7), 779–786. https://doi.org/10.1016/j.sleep.2012.03.016
Guidolin, M., & Gradisar, M. (2012b). Is shortened sleep duration a risk factor for overweight
and obesity during adolescence? A review of the empirical literature. Sleep Medicine, 13(7),
779–786. https://doi.org/10.1016/j.sleep.2012.03.016
Gunn, H. E., O’Rourke, F., Dahl, R. E., Goldstein, T. R., Rofey, D. L., Forbes, E. E., & Shaw, D.
S. (2018). Young adolescent sleep is associated with parental monitoring. Sleep Health,
(xxxx), 6–11. https://doi.org/10.1016/j.sleh.2018.09.001
Guo, X., Warden, B. A., Paeratakul, S., & Bray, G. A. (2004). Healthy eating index and obesity.
European Journal of Clinical Nutrition, 58(12), 1580–1586.
https://doi.org/10.1038/sj.ejcn.1601989
Hagenauer, M. H., Perryman, J. I., Lee, T. M., & Carskadon, M. A. (2009). Adolescent changes
in the homeostatic and circadian regulation of sleep. Developmental Neuroscience, 31(4),
276–284. https://doi.org/10.1159/000216538
Haines, J., Mcdonald, J., Brien, A. O., Sherry, B., Bottino, C. J., Schmidt, M. E., & Taveras, E.
M. (2019). Healthy Habits, Happy Homes Randomized Trial to Improve Household
Routines for Obesity Prevention Among Preschool-Aged Children, 02114(11), 1072–1079.
https://doi.org/10.1001/jamapediatrics.2013.2356
Hale, L., & Guan, S. (2015). Screen time and sleep among school-aged children and adolescents:
A systematic literature review. Sleep Medicine Reviews, 21, 50–58.
https://doi.org/10.1016/j.smrv.2014.07.007
Hale, L., & Guan, S. (2016). Screen Time and Sleep among School-Aged Children and
Adolescents: A Systematic Literature Review. Sleep Medicine Reviews, 50–58.
https://doi.org/10.1016/j.smrv.2014.07.007.Screen
Hales, C. M., Carroll, M. D., Fryar, C. D., & Ogden, C. L. (2017). Prevalence of Obesity Among
Adults and Youth: United States, 2015–2016. NCHS Data Brief, (288), 1–8.
https://doi.org/10.1017/S1368980017000088
Han, J. C., Lawlor, D. A., & Kimm, S. Y. (2010). Childhood obesity. The Lancet, 375(9727),
1737–1748. https://doi.org/10.1016/S0140-6736(10)60171-7
Hanlon, E. C., Tasali, E., Leproult, R., Stuhr, K. L., Doncheck, E., Wit, H. De, … Cauter, E.
Van. (2016). Sleep Restriction Enhances the Daily Rhythm of Circulating Levels of
Endocannabinoid 2-Arachidonoylglycerol. Sleep, 39(3), 653–664.
Harnack, L. (2013). Nutrition Data System for Research (NDSR). In M. D. Gellman & J. R.
Turner (Eds.), Encyclopedia of Behavioral Medicine (pp. 1348–1350). New York, NY:
Springer New York. https://doi.org/10.1007/978-1-4419-1005-9_1683
Hart, C. N., Carskadon, M. a, Considine, R. V, Fava, J. L., Lawton, J., Raynor, H. a, … Wing, R.
(2013). Changes in children’s sleep duration on food intake, weight, and leptin. Pediatrics,
132(6), e1473-80. https://doi.org/10.1542/peds.2013-1274
He, F., Bixler, E. O., Berg, A., Kawasawa, Y. I., Vgontzas, A. N., Fernandez-Mendoza, J., …
Liao, D. (2015). Habitual sleep variability, not sleep duration, is associated with caloric
intake in adolescents. Sleep Medicine, 16(7), 856–861.
https://doi.org/10.1016/j.sleep.2015.03.004
He, F., Bixler, E. O., Liao, J., Berg, A., Imamura Kawasawa, Y., Fernandez-Mendoza, J., …
Liao, D. (2015). Habitual sleep variability, mediated by nutrition intake, is associated with
abdominal obesity in adolescents. Sleep Medicine, 16(12), 1489–1494.
https://doi.org/10.1016/j.sleep.2015.07.028
Hekler, E. B., Michie, S., Pavel, M., Rivera, D. E., Collins, L. M., Jimison, H. B., … Spruijt-
metz, D. (2016). Advancing Models and Theories for Digital Behavior Change

182
Interventions. American Journal of Preventive Medicine, 51(5), 825–832.
https://doi.org/10.1016/j.amepre.2016.06.013
Heron, K. E., Everhart, R. S., McHale, S. M., & Smyth, J. M. (2017). Systematic Review: Using
mobile-technology-based Ecological Momentary Assessment (EMA) methods with youth: a
systematic review and recommendations. Journal of Pediatric Psychology, (September), 1–
21. https://doi.org/10.1093/jpepsy/jsx078
Hippel, P. T. Von, & Workman, J. (2016). From Kindergarten Through Second Grade , U . S .
Children ’ s Obesity Prevalence Grows Only During Summer Vacations, 24(11), 2296–
2300. https://doi.org/10.1002/oby.21613
Hirshkowitz, M., Whiton, K., Albert, S. M., Alessi, C., Bruni, O., DonCarlos, L., … Adams
Hillard, P. J. (2015). National sleep foundation’s sleep time duration recommendations:
Methodology and results summary. Sleep Health, 1(1), 40–43.
https://doi.org/10.1016/j.sleh.2014.12.010
Ievers-Landis, C. E., Kneifel, A., Giesel, J., Rahman, F., Narasimhan, S., Uli, N., … Riordan, M.
O. (2016). Dietary Intake and Eating-Related Cognitions Related to Sleep Among
Adolescents Who Are Overweight or Obese. Journal of Pediatric Psychology, 41(March),
670–679. https://doi.org/10.1093/jpepsy/jsw017
Iglowstein, I., Jenni, O. G., Molinari, L., & Largo, R. H. (2003). Sleep Duration From Infancy to
Adolescence: Reference Values and Generational Trends. Pediatrics, 111(2), 302–307.
https://doi.org/10.1542/peds.111.2.302
Jakubowicz, D., Barnea, M., Wainstein, J., & Froy, O. (2013). High Caloric Intake at Breakfast
vs . Dinner Differentially Influences Weight Loss of Overweight and Obese Women,
21(12), 2504–2512. https://doi.org/10.1002/oby.20460
Jakubowski, K. P., Hall, M. H., Lee, L., & Matthews, K. A. (2017). Temporal Relationships
Between Napping and Nocturnal Sleep in Healthy Adolescents. Behavioral Sleep Medicine,
15(4), 257–269. https://doi.org/10.1080/15402002.2015.1126595
Jan, J. E., Reiter, R. J., Bax, M. C. O., Ribary, U., Freeman, R. D., & Wasdell, M. B. (2010).
Long-term sleep disturbances in children: A cause of neuronal loss. European Journal of
Paediatric Neurology, 14(5), 380–390. https://doi.org/10.1016/j.ejpn.2010.05.001
Jansen, E. C., Dunietz, G. L., Chervin, R. D., Baylin, A., Baek, J., Banker, M., … Peterson, K. E.
(2018). Adiposity in Adolescents: The Interplay of Sleep Duration and Sleep Variability.
The Journal of Pediatrics, 1–8. https://doi.org/10.1016/j.jpeds.2018.07.087
Jarrin, D. C., McGrath, J. J., & Drake, C. L. (2013). Beyond sleep duration: distinct sleep
dimensions are associated with obesity in children and adolescents. International Journal of
Obesity, 37(4), 552–558. https://doi.org/10.1038/ijo.2013.4
Jarrin, D., Mcgrath, J., & Drake, C. (2013). Beyond sleep duration: distinct sleep dimensions are
associated with obesity in children and adolescents. Sleep Medicine, 14(4), e159.
https://doi.org/10.1016/j.sleep.2013.11.367
Javaheri, S., Storfer-Isser, A., Rosen, C. L., & Redline, S. (2008). Sleep Quality and Elevated
Blood Pressure in Adolescents. Circulation, 118(10), 1034–1040.
https://doi.org/10.1161/CIRCULATIONAHA.108.766410
Jenni, O. G., Achermann, P., & Carskadon, M. A. (2005). Homeostatic sleep regulation in
adolescents. Sleep, 28(11), 1446–1454. https://doi.org/10.1093/sleep/28.11.1446
Jr, A. L. C., Kapur, V., Maganti, R., Owens, J., Pancer, J., Swick, T. J., & Zak, R. (2007).
Practice Parameters for the Clinical Evaluation and Treatment of Circadian Rhythm Sleep
Disorders An American Academy of Sleep Medicine Report Circadian Rhythm Sleep
Disorders : Part I , Basic Principles , Shift Work and Jet Lag Disorders An American Acad,
3(11), 757–760.

183
Kahleova, H., Lloren, J. I., Mashchak, A., Hill, M., & Fraser, G. E. (2017). Meal Frequency and
Timing Are Associated with Changes in Body Mass Index in Adventist Health Study 2. The
Journal of Nutrition, (11), jn244749. https://doi.org/10.3945/jn.116.244749
Kamath, C. C., Vickers, K. S., Ehrlich, A., Mcgovern, L., Johnson, J., Singhal, V., … Montori,
V. M. (2008). Behavioral Interventions to Prevent Childhood Obesity : A Systematic
Review and Metaanalyses of Randomized Trials, 93(December), 4606–4615.
https://doi.org/10.1210/jc.2006-2411
Katzmarzyk, P. T., Barreira, T. V, Broyles, S. T., Chaput, J., Fogelholm, M., Hu, G., … Maia, J.
(2015). Association between body mass index and body fat in 9 – 11-year-old children from
countries spanning a range of human development, 43–46.
https://doi.org/10.1038/ijosup.2015.18
Kim, C. W., Choi, M. K., Im, H. J., Kim, O. H., Lee, H. J., Song, J., … Park, K. H. (2012).
Weekend catch-up sleep is associated with decreased risk of being overweight among fifth-
grade students with short sleep duration. Journal of Sleep Research, 21(5), 546–551.
https://doi.org/10.1111/j.1365-2869.2012.01013.x
Kjeldsen, J. S., Hjorth, M. F., Andersen, R., Michaelsen, K. F., Tetens, I., Astrup, a, … Sjödin,
a. (2014). Short sleep duration and large variability in sleep duration are independently
associated with dietary risk factors for obesity in Danish school children. International
Journal of Obesity, 38(1), 32–39. https://doi.org/10.1038/ijo.2013.147
Klingenberg, L., Chaput, J.-P., Holmbäck, U., Visby, T., Jennum, P., Nikolic, M., … Sjödin, A.
(2013). Acute Sleep Restriction Reduces Insulin Sensitivity in Adolescent Boys. Sleep,
36(7), 1085–1090. https://doi.org/10.5665/sleep.2816
Kopasz, M., Loessl, B., Hornyak, M., Riemann, D., Nissen, C., Piosczyk, H., & Voderholzer, U.
(2010). Sleep and memory in healthy children and adolescents - A critical review. Sleep
Medicine Reviews, 14(3), 167–177. https://doi.org/10.1016/j.smrv.2009.10.006
Kraemer, H. C., Yesavage, J. A., Taylor, J. L., & Kupfer, D. (2000). How can we learn about
developmental processes from cross-sectional studies, or can we? American Journal of
Psychiatry, 157(2), 163–171. https://doi.org/10.1176/appi.ajp.157.2.163
Krietsch, K. N., Armstrong, B., McCrae, C. S., & Janicke, D. M. (2016). Temporal Associations
Between Sleep and Physical Activity Among Overweight/Obese Youth. Journal of
Pediatric Psychology, jsv167. https://doi.org/10.1093/jpepsy/jsv167
Kruger, A. K., Reither, E. N., Peppard, P. E., Krueger, P. M., & Hale, L. (2014). Do sleep-
deprived adolescents make less-healthy food choices? The British Journal of Nutrition,
111(10), 1898–1904. https://doi.org/10.1017/S0007114514000130
Kuczmarski, R., Ogden, C., & Guo, S. (2002). 2000 CDC Growth Charts for the United States:
Methods and Development. Vital Health Stat, 11(246), 1–190.
Kumari, M., Badrick, E., Ferrie, J., Perski, A., Marmot, M., & Chandola, T. (2009). Self-
Reported Sleep Duration and Sleep Disturbance are Independently Associated with Cortisol
Secretion in the Whitehall II study. The Journal of Clinical Endocrinology and Metabolism,
94(12), 4801–4809. https://doi.org/10.1210/jc.2009-0555
Kuo, S. I. C., Updegraff, K. A., Zeiders, K. H., McHale, S. M., Umaña-Taylor, A. J., & De Jesús,
S. A. R. (2014). Mexican American Adolescents’ Sleep Patterns: Contextual Correlates and
Implications for Health and Adjustment in Young Adulthood. Journal of Youth and
Adolescence, 44(2), 346–361. https://doi.org/10.1007/s10964-014-0156-1
Kwasnicka, D., Dombrowski, S. U., & White, M. (2016). Theoretical explanations for
maintenance of behaviour change : a systematic review of behaviour theories, 7199.
https://doi.org/10.1080/17437199.2016.1151372
Landhuis, C. E., Poulton, R., Welch, D., & Hancox, R. J. (2008). Childhood Sleep Time and

184
Long-Term Risk for Obesity: A 32-Year Prospective Birth Cohort Study. Pediatrics,
122(5), 955–960. https://doi.org/10.1542/peds.2007-3521
Lewitt, M. S. (2017). The Role of the Growth Hormone / Insulin-Like Growth Factor System in
Visceral Adiposity. https://doi.org/10.1177/1178626417703995
Longo, V. D., & Panda, S. (2016). Perspective Fasting , Circadian Rhythms , and Time-
Restricted Feeding in Healthy Lifespan. Cell Metabolism, 23(6), 1048–1059.
https://doi.org/10.1016/j.cmet.2016.06.001
Luyster, F. S., Strollo, P. J., Zee, P. C., & Walsh, J. K. (2012). Sleep: A Health Imperative.
Sleep, 35(6), 727–734. https://doi.org/10.5665/sleep.1846
Ly, J., McGrath, J. J., & Gouin, J.-P. (2015). Poor sleep as a pathophysiological pathway
underlying the association between stressful experiences and the diurnal cortisol profile
among children and adolescents. Psychoneuroendocrinology, 57, 51–60.
https://doi.org/10.1016/j.psyneuen.2015.03.006
Magee, L., & Hale, L. (2012). Longitudinal associations between sleep duration and subsequent
weight gain: A systematic review. Sleep Medicine Reviews, 16(3), 231–241.
https://doi.org/10.1016/j.smrv.2011.05.005
Malone, S. K., Zemel, B., Compher, C., Souders, M., Chittams, J., Thompson, A. L., … Lipman,
T. H. (2016). Social jet lag, chronotype and body mass index in 14-17-year-old adolescents.
Chronobiology International, 33(9), 1–12. https://doi.org/10.1080/07420528.2016.1196697
Markwald, R. R., Melanson, E. L., Smith, M. R., Higgins, J., Perreault, L., Eckel, R. H., &
Wright, K. P. (2013). Impact of insufficient sleep on total daily energy expenditure, food
intake, and weight gain. Proceedings of the National Academy of Sciences, 110(14), 5695–
5700. https://doi.org/10.1073/pnas.1216951110
Matricciani, L., Olds, T., & Petkov, J. (2012). In search of lost sleep: Secular trends in the sleep
time of school-aged children and adolescents. Sleep Medicine Reviews, 16(3), 203–211.
https://doi.org/10.1016/j.smrv.2011.03.005
Matricciani, L., Olds, T. S., Blunden, S., Rigney, G., & Williams, M. T. (2012). Never Enough
Sleep: A Brief History of Sleep Recommendations for Children. Pediatrics, 129(3), 548–
556. https://doi.org/10.1542/peds.2011-2039
Mcconnell, R., Shen, E., Gilliland, F. D., Jerrett, M., Wolch, J., & Chang, C. (2015). Research |
Children ’ s Health A Longitudinal Cohort Study of Body Mass Index and Childhood
Exposure to Secondhand Tobacco Smoke and Air Pollution : The Southern California
Children ’ s Health Study, 360(4), 360–366.
McDonald, L., Wardle, J., Llewellyn, C. H., & Fisher, A. (2015). Nighttime sleep duration and
hedonic eating in childhood. International Journal of Obesity (2005), 39(10), 1463–1466.
https://doi.org/10.1038/ijo.2015.132
McHale, S. M., Kim, J. Y., Kan, M., & Updegraff, K. A. (2011). Sleep in Mexican-American
Adolescents: Social Ecological and Well-Being Correlates. Journal of Youth and
Adolescence, 40(6), 666–679. https://doi.org/10.1007/s10964-010-9574-x
McHill, A. W., & Wright, K. P. (2017). Role of sleep and circadian disruption on energy
expenditure and in metabolic predisposition to human obesity and metabolic disease.
Obesity Reviews, 18(February), 15–24. https://doi.org/10.1111/obr.12503
Mcreynolds, A., Trivedi, Z. B., Roberts, A. L., Sy, M., & Hirsch, J. (2012). Sleep restriction
leads to increased activation of brain regions sensitive, (7), 818–824.
https://doi.org/10.3945/ajcn.111.027383.818
Medin, A. C., Myhre, J. B., Diep, L. M., & Andersen, L. F. (2019). Diet quality on days without
breakfast or lunch – Identifying targets to improve adolescents ’ diet. Appetite,
135(January), 123–130. https://doi.org/10.1016/j.appet.2019.01.001

185
Meerlo, P., Sgoifo, A., & Suchecki, D. (2008). Restricted and disrupted sleep: Effects on
autonomic function, neuroendocrine stress systems and stress responsivity. Sleep Medicine
Reviews, 12(3), 197–210. https://doi.org/10.1016/j.smrv.2007.07.007
Meltzer, L. J., Montgomery-Downs, H. E., Insana, S. P., & Walsh, C. M. (2012). Use of
actigraphy for assessment in pediatric sleep research. Sleep Medicine Reviews, 16(5), 463–
475. https://doi.org/10.1016/j.smrv.2011.10.002
Meltzer, L. J., Walsh, C. M., Traylor, J., & Westin, A. M. L. (2012). Direct Comparison of Two
New Actigraphs and Polysomnography in Children and Adolescents. Sleep, 35(1), 159–166.
Mezick, E. J., Matthews, K. A., Hall, M., Kamarck, T. W., Buysse, D. J., Owens, J. F., & Reis,
S. E. (2009). Intra-individual variability in sleep duration and fragmentation: Associations
with stress. Psychoneuroendocrinology, 34(9), 1346–1354.
https://doi.org/10.1016/j.psyneuen.2009.04.005
Miller, A. L., Kaciroti, N., LeBourgeois, M. K., Chen, Y. P., Sturza, J., & Lumeng, J. C. (2014).
Sleep Timing Moderates the Concurrent Sleep Duration–Body Mass Index Association in
Low-Income Preschool-Age Children. Academic Pediatrics, 14(2), 207–213.
https://doi.org/10.1016/j.acap.2013.12.003
Miller, M., Kruisbrink, M., Wallace, J., Ji, C., & Cappuccio, F. P. (2018). Sleep duration and
incidence of obesity in infants, children, and adolescents: a systematic review and meta-
analysis of prospective studies. Sleep, (May), 1–19. https://doi.org/10.1093/sleep/zsy018
Moore, M., Kirchner, H. L., Drotar, D., Johnson, N., Rosen, C., Ancoli-Israel, S., & Redline, S.
(2009). Relationships among sleepiness, sleep time, and psychological functioning in
adolescents. Journal of Pediatric Psychology, 34(10), 1175–1183.
https://doi.org/10.1093/jpepsy/jsp039
Moore, M., Kirchner, H. L., Drotar, D., Johnson, N., Rosen, C., & Redline, S. (2011). Correlates
of adolescent sleep time and variability in sleep time: The role of individual and health
related characteristics☆. Sleep Medicine, 12(3), 239–245.
https://doi.org/10.1016/j.sleep.2010.07.020
Moore, M., & Meltzer, L. J. (2008). The sleepy adolescent: causes and consequences of
sleepiness in teens. Paediatric Respiratory Reviews, 9(2), 114–121.
https://doi.org/10.1016/j.prrv.2008.01.001
Moore, M., Ph, D., Kirchner, H. L., Drotar, D., Johnson, N., Rosen, C., & Redline, S. (2012).
Correlates of Adolescent Sleep Time and Variability in Sleep Time : The Role of Individual
and Health Related Characteristics, 12(3), 239–245.
https://doi.org/10.1016/j.sleep.2010.07.020.CORRELATES
Morselli, L., Leproult, R., Balbo, M., & Spiegel, K. (2010). Role of sleep duration in the
regulation of glucose metabolism and appetite. Best Practice & Research. Clinical
Endocrinology & Metabolism, 24(5), 687–702. https://doi.org/10.1016/j.beem.2010.07.005
Mukaka, M. M. (2012). Statistics Corner : A guide to appropriate use of Correlation coefficient
in medical research, 24(September), 69–71.
Nahum-Shani, I., Hekler, E. B., & Spruijt-Metz, D. (2015). Building health behavior models to
guide the development of just-in-time adaptive interventions: A pragmatic framework.
Health Psychology, 34(0), 1209–1219. https://doi.org/10.1037/hea0000306.Building
Nahum-Shani, I., Smith, S. N., Spring, B. J., Collins, L. M., Witkiewitz, K., Tewari, A., &
Murphy, S. A. (2016). Just-in-Time Adaptive Interventions (JITAIs) in Mobile Health: Key
Components and Design Principles for Ongoing Health Behavior Support. Annals of
Behavioral Medicine, 1–17. https://doi.org/10.1007/s12160-016-9830-8
Nahum-shani, I., Smith, S. N., Witkiewitz, K., Collins, L. M., Spring, B., & Murphy, S. A.
(2014). Just-in-time adaptive interventions (JITAIs): An organizing framework for ongoing

186
health behavior support. The Methodology Center Technical Report, 073975(14), 1–37.
National Sleep Foundation. (2006). 2006 Sleep in America: Teens and Sleep. National Sleep
Foundation, (202), 0–76. https://doi.org/10.1017/CBO9781107415324.004
Nielsen, L. S., Danielsen, K. V., & Sørensen, T. I. a. (2011). Short sleep duration as a possible
cause of obesity: critical analysis of the epidemiological evidence. Obesity Reviews, 12(2),
78–92. https://doi.org/10.1111/j.1467-789X.2010.00724.x
O’Connor, S. G., Huh, J., Schembre, S. M., Lopez, R. D. N. V, & Dunton, G. F. (2019). The
Association of Maternal Perceived Stress With Changes in Their Children’s Healthy Eating
Index ( HEI-2010 ) Scores Over Time. Annals of Behavioral Medicine, 1–9.
https://doi.org/10.1093/abm/kay095
Ogden, C. L., Carroll, M. D., Kit, B. K., & Flegal, K. M. (2012). Prevalence of Obesity and
Trends in Body Mass Index Among US Children and Adolescents, 1999-2010. Jama,
307(5), 483. https://doi.org/10.1001/jama.2012.40
Ogden, C. L., Carroll, M. D., Kit, B. K., & Flegal, K. M. (2014). Prevalence of childhood and
adult obesity in the United States, 2011-2012. Jama, 311(8), 806–814.
https://doi.org/10.1001/jama.2014.732
Ogden, C. L., & Flegal, K. M. (2010). Changes in terminology for childhood overweight and
obesity. National Health Statistics Reports, (25), 1–5.
Ogilvie, R. P., Redline, S., Bertoni, A. G., Chen, X., Ouyang, P., Szklo, M., & Lutsey, P. L.
(2013). Actigraphy Measured Sleep Indices and Adiposity : The Multi-Ethnic Study of
Atherosclerosis ( MESA ), 1701–1708.
Paavonen, E. J., Fjällberg, M., Steenari, M.-R., & Aronen, E. T. (2002). Actigraph placement
and sleep estimation in children. Sleep, 25(2), 235–237.
Parent, A. S., Teilmann, G., Juul, A., Skakkebaek, N. E., Toppari, J., & Bourguignon, J. P.
(2003). The Timing of Normal Puberty and the Age Limits of Sexual Precocity: Variations
around the World, Secular Trends, and Changes after Migration. Endocrine Reviews, 24(5),
668–693. https://doi.org/10.1210/er.2002-0019
Park, H., Tsai, K. M., Dahl, R. E., Irwin, M. R., McCreath, H., Seeman, T. E., & Fuligni, A. J.
(2016). Sleep and inflammation during adolescence. Psychosomatic Medicine, 78(6), 677–
685. https://doi.org/10.1097/PSY.0000000000000340
Paruthi, S., Brooks, L. J., Ambrosio, C. D., Hall, W. A., Kotagal, S., Lloyd, R. M., … Troester,
M. M. (2016). Recommended Amount of Sleep for Pediatric Populations: A Consensus
Statement of the American Academy of Sleep Medicine, 12(6), 785–786.
Patel, S. R., Zhu, X., Storfer-Isser, A., Mehra, R., Jenny, N. S., Tracy, R., & Redline, S. (2009).
Sleep duration and biomarkers of inflammation. Sleep, 32(2), 200–204.
https://doi.org/10.1093/sleep/32.2.200
Pervanidou, P., & Chrousos, G. P. (2012). Metabolic consequences of stress during childhood
and adolescence. Metabolism: Clinical and Experimental, 61(5), 611–619.
https://doi.org/10.1016/j.metabol.2011.10.005
Petersen, A. C., Crockett, L., Richards, M., & Boxer, A. (1988). A self-report measure of
pubertal status: Reliability, validity, and initial norms. Journal of Youth and Adolescence,
17(2), 117–133. https://doi.org/10.1007/BF01537962
Peuhkuri, K., Sihvola, N., & Korpela, R. (2012). Diet promotes sleep duration and quality.
Nutrition Research, 32(5), 309–319. https://doi.org/10.1016/j.nutres.2012.03.009
Pew. (2018). Teens, Social Media & Technology 2018.
Piantadosi, S., Byar, D., & Green, S. (1988). The ecological fallacy. American Journal of
Epidemiology, 127(5), 893–904. https://doi.org/10.1086/222356
Piernas, B. C., & Popkin, B. M. (2010). Trends In Snacking Among U.S. Children. Health

187
Affairs, 3(3), 398–404. https://doi.org/10.1377/hlthaff.2009.0666
Quante, M., Kaplan, E. R., Cailler, M., Rueschman, M., Wang, R., & Redline, S. (2018).
Actigraphy-based sleep estimation in adolescents and adults : a comparison with
polysomnography using two scoring algorithms, 13–20.
Quante, M., Kaplan, E. R., Cailler, M., Rueschman, M., Wang, R., Weng, J., … Redline, S.
(2018). Actigraphy-based sleep estimation in adolescents and adults: a comparison with
polysomnography using two scoring algorithms. Nature and Science of Sleep, 10, 13–20.
https://doi.org/10.2147/NSS.S151085
Reedy, J., & Krebs-Smith, S. M. (2010). Dietary Sources of Energy, Solid Fats, and Added
Sugars among Children and Adolescents in the United States. Journal of the American
Dietetic Association, 110(10), 1477–1484. https://doi.org/10.1016/j.jada.2010.07.010
Rigney, G., Blunden, S., Maher, C., Dollman, J., Parvazian, S., Matricciani, L., & Olds, T.
(2015). Can a school-based sleep education programme improve sleep knowledge , hygiene
and behaviours using a randomised controlled. Sleep Medicine, 16(6), 736–745.
https://doi.org/10.1016/j.sleep.2015.02.534
Sadeh, A. (2011). The role and validity of actigraphy in sleep medicine: An update. Sleep
Medicine Reviews, 15(4), 259–267. https://doi.org/10.1016/j.smrv.2010.10.001
Sadeh, A., & Acebo, C. (2002). The role of actigraphy in sleep medicine. Sleep Medicine
Reviews, 6(2), 113–124. https://doi.org/10.1053/smrv.2001.0182
Sadeh, A., Dahl, R. E., Shahar, G., & Rosenblat-Stein, S. (2009). Sleep and the transition to
adolescence: a longitudinal study. Sleep, 32(12), 1602–1609.
https://doi.org/10.1016/j.ypsy.2010.09.033
Schmid, S. M., Hallschmid, M., & Jauch-, K. (2008). Sleep loss A single night of sleep
deprivation increases ghrelin levels and feelings of hunger in normal-weight healthy men,
331–334. https://doi.org/10.1111/j.1365-2869.2008.00662.x
Schmid, S. M., Hallschmid, M., & Jauch, K. (2008). Sleep loss A single night of sleep
deprivation increases ghrelin levels and feelings of hunger in normal-weight healthy men,
331–334. https://doi.org/10.1111/j.1365-2869.2008.00662.x
Schwartz, J. E., & Stone, A. A. (1998). Strategies for analyzing ecological momentary
assessment data. Health Psychology, 17(1), 6–16. https://doi.org/10.1037/0278-6133.17.1.6
Schwimmer, J. B., Burwinkle, T. M., & Varni, J. W. (2003). Health-Related Quality of Life of
Severely Obese Children and Adolescents. Journal of the American Medical Association,
289(14), 1813–1819. https://doi.org/10.1001/jama.289.14.1813
Shiffman, S. (2007). Designing Protocols for Ecological Momentary Assessment. The Science of
Real-Time Data Capture, 27–53.
Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological Momentary Assessment.
Annual Review of Clinical Psychology, 4, 1–32.
https://doi.org/10.1146/annurev.clinpsy.3.022806.091415
Short, M. A., Gradisar, M., Lack, L. C., Wright, H., & Carskadon, M. A. (2012). The
discrepancy between actigraphic and sleep diary measures of sleep in adolescents. Sleep
Medicine, 13(4), 378–384. https://doi.org/10.1016/j.sleep.2011.11.005
Sims, J., Scarborough, P., & Foster, C. (2015). The Effectiveness of Interventions on Sustained
Childhood Physical Activity : A Systematic Review and Meta-Analysis of Controlled
Studies, 1–21. https://doi.org/10.1371/journal.pone.0132935
Singh, A. S., Mulder, C., Twisk, J. W. R., Van Mechelen, W., & Chinapaw, M. J. M. (2008).
Tracking of childhood overweight into adulthood: A systematic review of the literature.
Obesity Reviews, 9(5), 474–488. https://doi.org/10.1111/j.1467-789X.2008.00475.x
Singh, M. (2014). Mood, food, and obesity. Frontiers in Psychology, 5(September), 1–20.

188
https://doi.org/10.3389/fpsyg.2014.00925
Sisk, C. L., & Foster, D. L. (2004). The neural basis of puberty and adolescence. Nature
Neuroscience, 7(10), 1040–1047. https://doi.org/10.1038/nn1326
Slater, J. A., Botsis, T., Walsh, J., King, S., Straker, L. M., & Eastwood, P. R. (2015). Assessing
sleep using hip and wrist actigraphy. Sleep and Biological Rhythms, 13(2), 172–180.
https://doi.org/10.1111/sbr.12103
Spaeth, A. M., Hawley, N. L., Raynor, H. A., Jelalian, E., Greer, A., Crouter, S. E., … Hart, C.
N. (2019). Sleep, energy balance, and meal timing in school-aged children. Sleep Medicine.
https://doi.org/10.1016/j.sleep.2019.02.003
Spiegel, K., Tasali, E., Leproult, R., & Van Cauter, E. (2009). Effects of poor and short sleep on
glucose metabolism and obesity risk. Nature Reviews. Endocrinology, 5(5), 253–261.
https://doi.org/10.1038/nrendo.2009.23
Spring, B., King, A., Pagoto, S., Horn, L. Van, & Fisher, J. (2015). Fostering Multiple Healthy
Lifestyle Behaviors for Primary Prevention of Cancer. American Psychological Association,
70(2), 75–90. https://doi.org/10.1037/a0038806.Fostering
Spruijt-metz, D., Hekler, E., Saranummi, N., Intille, S., Korhonen, I., Nilsen, W., … Pavel, M.
(2015). Building new computational models to support health behavior change and
maintenance: new opportunities in behavioral research. Translational Behavioral Medicine,
5(3), 335–346. https://doi.org/10.1007/s13142-015-0324-1
Spruyt, K., Molfese, D. L., & Gozal, D. (2011). Sleep duration, sleep regularity, body weight,
and metabolic homeostasis in school-aged children. Pediatrics, 127(2), e345–e352.
https://doi.org/10.1542/peds.2010-0497
St-Onge, M., Ard, J., Baskin, M. L., Chiuve, S. E., Johnson, H. M., Kris-Etherton, P., & Varady,
K. (2017). Meal Timing and Frequency: Implications for Cardiovascular Disease
Prevention: A Scientific Statement From the American Heart Association. Circulation,
135(9), e96–e121. https://doi.org/10.1161/CIR.0000000000000476
St-Onge, M., Roberts, A. L., Chen, J., Kelleman, M., O’Keeffe, M., RoyChoudhury, A., & Jones,
P. J. H. (2011). Short sleep duration increases energy intakes but does not change energy
expenditure in normal-weight individuals. American Journal of Clinical Nutrition, 94(2),
410–416. https://doi.org/10.3945/ajcn.111.013904
St-Onge, M., Wolfe, S., Sy, M., Shechter, A., & Hirsch, J. (2014). Sleep restriction increases the
neuronal response to unhealthy food in normal-weight individuals. International Journal of
Obesity (2005), 38(3), 411–416. https://doi.org/10.1038/ijo.2013.114
Stone, M. R., Stevens, D., & Faulkner, G. E. J. (2013). Maintaining recommended sleep
throughout the week is associated with increased physical activity in children. Preventive
Medicine, 56(2), 112–117. https://doi.org/10.1016/j.ypmed.2012.11.015
Stoner, L., Castro, N., Signal, L., Skidmore, P., Faulkner, J., Lark, S., … Harrex, H. (2018).
Sleep and Adiposity in Preadolescent Children: The Importance of Social Jetlag. Childhood
Obesity, X(X), chi.2017.0272. https://doi.org/10.1089/chi.2017.0272
Stores, G. (2009). Aspects of sleep disorders in children and adolescents. Dialogues in Clinical
Neurosciences, 11(1), 81–90.
Story, M., Neumark-Sztainer, D., & French, S. (2002). Individual and environmental influences
on adolescent eating behaviors. Journal of the American Dietetic Association, 102(3), S40–
S51.
Szajewska, H., & Ruszczyński, M. (2010). Systematic Review Demonstrating that Breakfast
Consumption Influences Body Weight Outcomes in Children and Adolescents in Europe
that Breakfast Consumption Influences Body Weight Outcomes in Children and
Adolescents in Europe. Critical Reviews in Food Science and Nutrition, 8398.

189
https://doi.org/10.1080/10408390903467514
Tagler, M. J., Stanko, K. A., & Forbey, J. D. (2017). Predicting sleep hygiene: a reasoned action
approach. Journal of Applied Social Psychology, 47(1), 3–12.
https://doi.org/10.1111/jasp.12411
Tande, D. L., Magel, R., & Strand, B. N. (2010). Healthy Eating Index and abdominal obesity.
Public Health Nutrition, 13(02), 208. https://doi.org/10.1017/S1368980009990723
Tasali, E., Chapotot, F., Wroblewski, K., & Schoeller, D. (2014). The effects of extended
bedtimes on sleep duration and food desire in overweight young adults: A home-based
intervention. Appetite, 80, 220–224. https://doi.org/10.1016/j.appet.2014.05.021
Tate, E. B., Spruijt-Metz, D., O’Reilly, G., Jordan-Marsh, M., Gotsis, M., Pentz, M. A., &
Dunton, G. F. (2013). mHealth approaches to child obesity prevention: successes, unique
challenges, and next directions. Translational Behavioral Medicine, 3(4), 406–415.
https://doi.org/10.1007/s13142-013-0222-3
Tek, N. A., Yildiran, H., Akbulut, G., Bilici, S., Koksal, E., Karadag, M. G., & Sanlier, N.
(2011). Evaluation of dietary quality of adolescents using Healthy Eating Index. Nutrition
Research and Practice, 5(4), 322–328. https://doi.org/10.4162/nrp.2011.5.4.322
Telzer, E. H., Goldenberg, D., Fuligni, A. J., Lieberman, M. D., & Gálvan, A. (2015). Sleep
variability in adolescence is associated with altered brain development. Developmental
Cognitive Neuroscience, 14, 16–22. https://doi.org/10.1016/j.dcn.2015.05.007
Tikotzky, L. (2017). Parenting and sleep in early childhood. Current Opinion in Psychology, 15,
118–124. https://doi.org/10.1016/j.copsyc.2017.02.016
Torres, S. J., & Nowson, C. a. (2007). Relationship between stress, eating behavior, and obesity.
Nutrition, 23(11–12), 887–894. https://doi.org/10.1016/j.nut.2007.08.008
Tremaine, R. B., Dorrian, J., & Blunden, S. (2010). Subjective and objective sleep in children
and adolescents: Measurement, age, and gender differences. Sleep and Biological Rhythms,
8(4), 229–238. https://doi.org/10.1111/j.1479-8425.2010.00452.x
Troxel, W. M., & Wolfson, A. (2017). The intersection between sleep science and policy:
Introduction to the special issue on school start times. Sleep Health, 3, 419–422.
https://doi.org/10.1016/j.sleh.2017.10.001
U.S. Department of Health and Human Services and US Department of Agriculture. (2015).
2015 – 2020 Dietary Guidelines for Americans. 2015 – 2020 Dietary Guidelines for
Americans (8th Edition), 18. https://doi.org/10.1097/NT.0b013e31826c50af
United States Department of Agriculture. (2000). The Healthy Eating Index: 1999-2000.
US Department of Health and Human Services. (2010). Healthy People 2020.
Vgontzas, A., Bixler, E. O., & Basta, M. (2010). Obesity and Sleep: A Bidirectional
Association? SLEEP, 33, 593–598.
Vgontzas, A., Lin, H.-M., Papaliaga, M., Calhoun, S., Vela-Bueno, A., Chrousos, G. P., &
Bixler, E. O. (2008). Short sleep duration and obesity: the role of emotional stress and sleep
disturbances. International Journal of Obesity (2005), 32(5), 801–809.
https://doi.org/10.1038/ijo.2008.4
Waters, E., de Silva-Sanigorski, A., Burford, B. J., Brown, T., Campbell, K. J., Gao, Y., …
Summerbell, C. D. (2011). Interventions for preventing obesity in children. The Cochrane
Database of Systematic Reviews, 7(12), CD001871.
https://doi.org/10.1002/14651858.CD001871.pub3.www.cochranelibrary.com
Watson, N. F., Martin, J. L., Wise, M. S., Carden, K. A., Kirsch, D. B., Kristo, D. A., … Weaver,
T. E. (2017). Delaying Middle School and High School Start Times Promotes Student
Health and Performance : An American Academy of Sleep Medicine Position Statement,
13(4), 11–13. https://doi.org/10.5664/jcsm.6558

190
Weiss, A., Johnson, N. L., Berger, N. A., & Redline, S. (2010). Validity of activity-based
devices to estimate sleep. Journal of Clinical Sleep Medicine, 6(4), 336–342.
Weiss, A., Xu, F., Storfer-Isser, A., Thomas, A., Ievers-Landis, C. E., & Redline, S. (2010). The
association of sleep duration with adolescents’ fat and carbohydrate consumption. Sleep,
33(9), 1201–1209.
Wells, N. M., Evans, G. W., Beavis, A., & Ong, A. D. (2010). Early childhood poverty,
cumulative risk exposure, and body mass index trajectories through young adulthood.
American Journal of Public Health, 100(12), 2507–2512.
https://doi.org/10.2105/AJPH.2009.184291
Wen, C. K. F., Schneider, S., Stone, A. A., & Spruijt-Metz, D. (2017). Compliance With Mobile
Ecological Momentary Assessment Protocols in Children and Adolescents: A Systematic
Review and Meta-Analysis. Journal of Medical Internet Research, 19(4), e132.
https://doi.org/10.2196/jmir.6641
Werner, H., Molinari, L., Guyer, C., & Jenni, O. G. (2008). Agreement rates between actigraphy,
diary, and questionnaire for children’s sleep patterns. Archives of Pediatrics & Adolescent
Medicine, 162(4), 350–358. https://doi.org/10.1001/archpedi.162.4.350
Westerlund, L., Ray, C., & Roos, E. (2009). Associations between sleeping habits and food
consumption patterns among 10–11-year-old children in Finland. British Journal of
Nutrition, 102(10), 1531. https://doi.org/10.1017/S0007114509990730
Willett, W. (1998). Nutritional Epidemiology. New York, NY: Oxford University Press.
Wilson, K. G., Watson, S. T., & Currie, S. R. (1998). Daily diary and ambulatory activity
monitoring of sleep in patients with insomnia associated with chronic musculoskeletal pain.
Pain, 75(1), 75–84. https://doi.org/10.1016/S0304-3959(97)00207-8
Wittmann, M., Dinich, J., Merrow, M., Roenneberg, T., Diego, S., Centrum, B., & Germany, M.
(2006). Social jetlag: misalignment of biological and social time, 23, 497–509.
https://doi.org/10.1080/07420520500545979
Wolfson, A. R., & Carskadon, M. a. (2003). Understanding adolescents’ sleep patterns and
school performance: a critical appraisal. Sleep Med Rev, 7(6), 491–506.
https://doi.org/10.1016/S1087-0792(03)90003-7
Wong, W. W., Ortiz, C. L., Lathan, D., Moore, L. a, Konzelmann, K. L., Adolph, A. L., …
Butte, N. F. (2013). Sleep duration of underserved minority children in a cross-sectional
study. BMC Public Health, 13(1), 648. https://doi.org/10.1186/1471-2458-13-648
Wright, K. P., Lowry, C. A., & LeBourgeois, M. K. (2012). Circadian and wakefulness-sleep
modulation of cognition in humans. Frontiers in Molecular Neuroscience, 5(April), 1–12.
https://doi.org/10.3389/fnmol.2012.00050
Yoong, S. L., Chai, L. K., Williams, C. M., Wiggers, J., Finch, M., & Wolfenden, L. (2016).
Systematic review and meta-analysis of interventions targeting sleep and their impact on
child body mass index, diet, and physical activity. Obesity, 24(5), 1140–1147.
https://doi.org/10.1002/oby.21459
Zarrinpar, A., Chaix, A., & Panda, S. (2016). Daily Eating Patterns and Their Impact on Health
and Disease. Trends in Endocrinology and Metabolism, 27(2), 69–83.
https://doi.org/10.1016/j.tem.2015.11.007

Abstract (if available)

Abstract This dissertation consists of three unique, yet interrelated, studies that leveraged real-time data capture methodologies to examine the influence of sleep health and variability on patterns of dietary quality and eating behavior, as well as weight trajectories over time, in a sample of youth. The aims of this project were to: (1) determine the agreement of ecological momentary assessment (EMA) with actigraphy for assessment of within-subject mean, variability, and day-to-day sleep health, as well as potential moderators of day-level agreement

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The acute and longitudinal associations between sedentary behaviors, affective states, and emotional disorder symptoms among youth

PDF

Understanding the methodological limitations In the ecological momentary assessment of physical activity

PDF

System dynamics of in-home family eating behavior: insights from intensive longitudinal data using Ecological Momentary Assessment and wearable sensors

PDF

Developing and testing novel strategies to detect inattentive responding in ecological momentary assessment studies

PDF

Understanding the dynamic relationships between physical activity and affective states using real-time data capture techniques

PDF

Free-living and in-lab effects of sedentary time on cardiac autonomic nervous system function in youth with overweight and obesity

PDF

Prenatal sleep health, cortisol, and gestational weight gain

PDF

Effects of near-roadway air pollution exposure on obesity, obesity-related behavior, and neurobehavioral deficits during peripuberty

PDF

Influences of specific environmental domains on childhood obesity and related behaviors

PDF

Effects of the perceived and objectively assessed environment on physical activity in adults and children

PDF

The Internet activities, gratifications, and health consequences related to compulsive Internet use

PDF

Psychosocial and cultural factors in the primary prevention of melanoma targeted to multiethnic children

PDF

A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use

Asset Metadata

Creator O'Connor, Sydney Grace (author)

Core Title Sleep health and variability in youth: a real-time data capture study to examine influences on daily dietary intake patterns and longitudinal weight trajectories

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Preventive Medicine (Health Behavior Research)

Publication Date 07/15/2019

Defense Date 04/23/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag BMI,body mass index,Children,dietary quality,eating behavior,ecological momentary assessment,EMA,OAI-PMH Harvest,obesity,Sleep

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Dunton, Genevieve (committee chair), Huh, Jimi (committee chair), Belcher, Britni (committee member), Page, Kathleen (committee member)

Creator Email oconnor.sy@gmail.com,sgoconno@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-183412

Unique identifier UC11660650

Identifier etd-OConnorSyd-7549.pdf (filename),usctheses-c89-183412 (legacy record id)

Legacy Identifier etd-OConnorSyd-7549.pdf

Dmrecord 183412

Document Type Dissertation

Format application/pdf (imt)

Rights O'Connor, Sydney Grace

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA