Sleep and Challenging Behaviors in the Context of Intensive Behavioral Intervention for Children with Autism

Abstract

This study examined the associations between sleep and challenging behaviors with average and day-to-day fluctuations in sleep, for 42 children with autism spectrum disorder receiving intensive behavioral intervention. Child sleep was recorded (via actigraphy) for five nights in conjunction with clinician-reported observations of challenging behaviors. Results indicated that on average, poor sleep was associated with higher rates of repetitive behavior, negative affect and overall challenging behaviors. However, only one significant association emerged for day-to-day fluctuations—children who woke more at night engaged in higher rates of self-injurious behaviors the following day. These findings suggest that average sleep patterns are more influential for challenging behaviors (when compared to daily fluctuations). Interventions aimed at improving sleep may have important cascading effects on challenging behaviors for children with ASD.

Sleep problems are highly comorbid in autism spectrum disorder (ASD), and up to 80% of individuals with ASD will experience at least one indicator of poor sleep by adulthood (Richdale and Schreck 2009). Most commonly, these problems include increased sleep onset latency, short sleep duration, frequent and extended night waking, and early morning rise times (Allik, Larsson, and Smedje 2006; Glickman 2010; Goodlin-Jones et al. 2008; Humphreys et al. 2013; Krakowiak et al. 2008; Moore, Evan, Hanvey, and Johnson 2017; Richdale and Schreck 2009; Souders et al. 2017)—among other parent-reported problems (e.g., bedtime resistance, daytime sleepiness; Aathira et al. 2017). The behavioral cascade associated with these sleep problems may have life-long developmental consequences, especially within the context of Intensive Behavioral Interventions (IBI). Within IBI, sleep disturbance and associated challenging behaviors may hinder therapeutic progress. Building on this premise, the present study addressed two research questions. First, are sleep behaviors (on average) associated with challenging behaviors within IBI? Second, are night-to-night fluctuations in sleep associated with day-to-day fluctuations in challenging behaviors (within IBI)? These questions build on three areas of research, including: sleep and challenging behaviors for children with ASD, field standards for sleep assessment, and the importance of the IBI context.

Intensive Behavioral Intervention

Behaviorally based strategies are often the first line of treatment for children with ASD, and IBI is the most common behavioral model used today (Bertoglio and Hendren 2009; National Autism Center 2009). IBI refers to a variety of high-intensity behaviorally-based interventions for children with ASD. IBI in the current study used an Applied Behavior Analysis (ABA) framework. ABA is an evidence-based approach and the most common approach used in Indiana. Within this study, all included IBI programs were center-based and each child had customized one-on-one programing (generated by a board-certified behavioral analyst).

For many young children with ASD, center-based IBI programs constitute a proximal context for their development—offering systematic, one-on-one treatment to facilitate gains in social interactions, communication, and adaptive behavior (Doehring, Reichow, Palka, Phillips, and Hagopian 2014; Peters–Scheffer, Didden, Korzilius, and Sturmey 2011; Lovaas 1987; Smith, Eikeseth, Klevstrand, and Lovaas 1997). Within this context, child challenging behaviors are especially salient as they can hinder developmental and educational progress. Additionally, center-based IBI programs provide a unique opportunity to capture observer-rated day-long observations of challenging behaviors.

Within IBI programs, several factors may influence child challenging behaviors (e.g., intellectual abilities or adaptive functioning) but child sleep is rarely considered (in research). In practice, sleep is often a behavior tracked and communicated across home and IBI contexts but research directly documenting its impacts in IBI is lacking. Building on a solid research base linking sleep and challenging behaviors (detailed below), this study will add the novel and developmentally consequential context of IBI center-based programs.

Sleep and Behavior in Children with ASD

Numerous studies have established correlational associations between sleep problems and daytime behavior in children with ASD (e.g., Schreck, Mulick, and Smith 2004). For children and young adults with ASD, sleep problems are commonly associated with core ASD features (Goldman et al. 2011; Hoffman et al. 2005; Hundley, Shui, and Malow 2016; Phung et al. 2017; Schreck, Mulick, and Smith 2004; Tudor, Hoffman, and Sweeney 2012; Veatch et al. 2017) and comorbid challenging behaviors, such as irritability (or negative affect expressions), aggression, and self-injury (Adams, et al. 2014; Cohen et al. 2017; Goldman et al. 2009; Hirata et al. 2016; Mazurek et al. 2016; Sikora et al. 2012; Sannar et al. 2017).

Previous ASD studies report that children who sleep less (or have more sleep problems) exhibit higher levels of social impairment (Phung and Goldberg 2017; Schreck, Mulick, and Smith 2004; Tudor, Hoffman, and Sweeney 2012; Veatch et al. 2017) and increased restricted and repetitive behaviors (Goldman et al. 2011; Hundley, Shui, and Malow 2016; Schreck, Mulick and Smith 2004; Tudor, Hoffman, and Sweeney 2012). When considering challenging behaviors in children with ASD, however, study findings are more mixed. For example, Goldman et al. (2011) and Didden et al. (2002) report sleep problems were not associated with parent–reported self–injury. Conversely, Goldman et al. (2009) noted moderate associations between poor sleep and self–injurious behaviors. These findings area part of a growing line of studies that explore the links between sleep problems and a variety of challenging behaviors like aggression (e.g., Cohen et al. 2017; Mazurek and Sohl 2016) and irritability (e.g., Mazurek et al. 2016; Sannar et al. 2017).

Apart from a few recent studies (e.g., Cohen et al. 2017; Goldman et al. 2009), associations between sleep and challenging behaviors are most commonly assessed using parent-report indices of both sleep and daytime behaviors. Although this practice is understandable from a practical standpoint, shared-reporter bias may inflate the presented associations. To avoid this bias, the present study uses direct observations of challenging behaviors (i.e., negative affect expressions, aggression, self-injury, and repetitive behaviors) and more objective indices of sleep.

Sleep Assessment

Assessments of sleep in children with ASD may include parent-reports, actigraphy, videosomnography (VSG), or polysomnography (PSG; Aathira et al. 2017; Schwichtenberg, Hensle, Honaker, Miller, Ozonoff, and Anders 2016; Schwichtenberg, Young, Hutman, Iosif, Sigman, Rogers, and Ozonoff 2013). Parent report measures most commonly index perceptions of sleep problems (Owens, Spirito, and McGuinn 2000), while actigraphy or VSG directly assess sleep behaviors (e.g., sleep onset latency sleep duration, and night waking; Ipsiroglu et al. 2015; Meltzer, Montgomery-Downs, Insana, and Walsh 2012). PSG is the field gold standard and includes the categorization of sleep states based on EEG activity. However, for most studies, PSG is not a feasible option because of rigorous training requirements, equipment expense, and time. Additionally, because of the invasive nature of PSG, it is difficult to capture typical sleep patterns for children with ASD using this method. For these reasons, actigraphy is regularly used to index sleep. Actigraphy uses an accelerometer to index movement and established algorithms are used to estimate several elements of sleep—most commonly, sleep duration and time awake after sleep onset (WASO; Sadeh, Alster, Urbach, and Lavie 1989; Sadeh, Sharkey, and Carskadon 1994).

Actigraphy guidelines recommend at least three, 24-hour periods of recording when generating average-level estimates of sleep duration and WASO (Ancoli–Israel et al. 2015). Although average-level estimates can inform general sleep patterns over time, they do not capture night-to-night sleep variability. Several studies have documented that children with ASD (or other developmental disabilities) have more night-to-night variability in their sleep (e.g., Anders, Iosif, Schwichtenberg, Tang, and Goodlin-Jones 2011). Additionally, night-to-night fluctuations in sleep have been linked with several health and daytime behavior concerns (Bei et al. 2016; Fuligni, Arruda, Krull, and Gonzales 2017). However, assessing both average-level and night-by-night sleep for children with ASD is not common. Using both indices may directly inform what elements of sleep are consequential for daytime challenging behaviors.

Present Study

The aims of this study are two-fold. First, we aim to replicate previous sleep and ASD studies within a new context – IBI programs. This will include looking at the relationships between average sleep patterns with ASD symptom severity and challenging behaviors. Second, we will explore if night-to-night sleep variations are influential in day-to-day fluctuations in challenging behaviors. Based on previous research, we hypothesize less optimal sleep (e.g., short sleep duration, frequent night waking) both as an average pattern and on a given night will be associated with greater ASD symptom severity and more challenging behaviors.

Methods

Participants

Participants included 42 children with ASD, recruited from five IBI centers within a two-hour radius of Purdue Univeristy. Children were eligible to participate if they had a documented ASD diagnosis and attended the IBI center five days per week. Children ranged from 2 to 10 years of age (M = 5.52, SD = 2.26), and the majority of enrolled children were male (81%), Caucasian (88%), and not Hispanic or Latino (97%). See Table 1 for additional child and family characteristics. Medications and comorbid diagnoses are also described in Table 2. Caregiver respondents included biological mothers (81%), biological fathers (9.5%), and biological grandmothers (9.5%).

Table 1.

Child and Family Characteristics

N = 42		N (%)
Sex	Male	34 (81)
Age (Years)		2–10 (M = 5.52)
Race	Asian	2(4.8)
	African-American	1(2.4)
	Caucasian	37(88.1)
	Multiracial	2(4.8)
Ethnicity	Hispanic	1(2.4)
	Non-Hispanic	39(92.9)
	Unknown/Unreported	2(4.8)
Caregiver Education	High School/GED	9(21.4)
	Trade or Vocational	3(7.1)
	Associates or 2 Year Degree	5(11.9)
	Some College	8(19.0)
	College Degree	9(21.4)
	Master’s Degree	4(9.5)
	Professional Degree	3(7.1)
	Other/Unreported	1(2.4)
Family Income	Below $20,000	10(23.8)
	$20,001–$40,000	8(19.0)
	$40,001–60,000	6(14.3)
	$60,001–$80,000	5(11.9)
	$80,001–$100,000	2(4.8)
	100,001–$125,000	2(4.8)
	$125,001 and above	4(9.6)
	Unreported	5(11.9)
Marital Status	Married	24(57.1)
	Single (Never Married)	6(14.3)
	Divorced/Separated	10(23.8)
	Living with Partner (not married)	1(2.1)
	Unknown/Unreported	1(2.4)

Table 2.

Comorbid Diagnoses and Sleep Medications

N = 42		N (%)
Comorbidities	None	30(71.4)
	Genetic Disorder/Condition	4(9.5)
	ADHD	3(7.1)
	Developmental Delay	3(7.1)
	Seizure Disorder/Epilepsy	2(4.8)
	Celiac Disease	2(4.8)
	Speech Delay	2(4.8)
	Cerebral Palsy	1(2.4)
	Anxiety	1(2.4)
	Multiple Comorbidities	4(9.5)
Sleep Medications	None	29(69.0)
	Melatonin	9(21.4)
	Clonidine	2(4.8)
	Other Sleep Medication	2(4.8)
	Total Children Taking Sleep Medications	13(31.0)

Note. For children with multiple comorbidities, each diagnosis is also individually accounted for in the categories listed above.

Procedure

Following an initial screening email to determine eligibility, families completed a one-hour enrollment visit in their home (n =35) or the child’s IBI center (n = 7) prior to the start of data collection. Institutional Review Board approval was granted from Purdue University. At the enrollment visit, parents or legal guardians provided written informed consent and completed a series of questionnaires on their child’s ASD features, daytime behaviors, and sleep problems. Caregivers also received an actigraph (described below). Caregivers were instructed to keep the actigraph on their child at all times unless the child was bathing or playing in water. Participating families were compensated with $50 (U.S. dollars) for enrolling their child and completing all questionnaires. Participating families also received a detailed clinical report of their child’s sleep patterns after the completion of the study. Daytime challenging behaviors were observed at the child’s IBI center for the five consecutive days that he/she wore the actigraph.

Measures

ASD Features

The Social Communication Questionnaire (SCQ; Berument et al. 1999) was administered to confirm the presence of core ASD features. Item scores were summed to yield a total SCQ score, with higher scores indicating more ASD symptoms. Scores 11 or above are indicative of possible ASD (Wiggins, Bakeman, Adamson, and Robins 2007). Caregivers endorsed a variety of ASD symptoms in their children, with SCQ scores ranging from 11 to 37 (M = 21.67, SD = 7.79). All scores were above the suggested cutoff and reflect a diverse range of ASD symptoms (highest possible SCQ score = 40).

Sleep

Sleep was measured using the micromini–motionlogger® actigraph (Ambulatory Monitoring, Incorporated), which provides a minute–by–minute recording of motion. Recent guidelines state that a minimum of 72 hours of recording are needed to obtain valid sleep estimates (Ancoli–Israel et al. 2015). Thus, actigraphs were worn for five consecutive 24–hour periods (Sunday through Friday) and data were excluded if a child had fewer than three full nights of valid recording. For the present study, sleep parameters included total sleep time (TST) and wake after sleep onset (WASO).

All children with severe sensory sensitivities (based-on parent report) completed an actigraph band desensitization protocol, which included wearing a practice actigraph band during the week prior to data collection (33%, n = 14). Of the children who completed the desensitization protocol, 93% (13 children) went on to wear the device for five full nights of recording. Due to prolonged sensory sensitivities, one child wore the sensor at night only. To confirm undetected sensory sensitivities were not systematically biasing our sleep estimates, a series of ANOVAs were completed comparing night one to nights two through five, these revealed no group differences (p > .05). Additionally, comparisons across children who completed the desensitization protocol and those who did not were also explored. No significant group differences emerged (p > .05).

At least five nights were recorded and scored for 90% of total participants (n = 38). Of 42 total participants, 41 children completed the actigraphy phase of the study. Full actigraphy data were missing for one child due to device malfunction. Partial data were missing for three additional participants due to prolonged actigraph removals on some nights of recording.

Following actigraphy guidelines, sleep diaries were also used to aid in interpreting actigraphy data (Ancoli–Israel et al. 2015). Caregivers were instructed to indicate when their child was awake and asleep during five consecutive 24–hour periods. Caregivers were also asked to note when the device was removed and replaced, if the child was sick or used medications, and if his/her sleep was atypical. If the child was sick during the recording week (n = 2), caregivers were asked to record on the subsequent week (Schwichtenberg, Shah, and Poehlmann 2013). Actigraphy data were scored using the Sadeh actigraph algorithm (Sadeh, Alster, Urbach, and Lavie 1989; Sadeh, Sharkey, and Carskadon 1994) in the Action W-2 version 2.4.20 software (Ambulatory Monitoring, Inc.). Down intervals, sensor removals, and sleep in a moving object were manually scored using the sleep diary. At least two raters with extensive actigraphy experience addressed validity concerns (via group consensus) prior to data analysis. Thirty-nine caregivers completed the accompanying sleep diary.

Daytime Challenging Behavior

The child’s behavioral clinician¹ recorded daytime behaviors while he/she received center-based IBI. Observations occurred for the duration of one standard treatment week (Monday–Friday). Child therapy hours ranged from 4 to 8 hours per day. If the child was absent for more than one day (or if the center was closed due to inclement weather), caregivers were given the option to record again during a subsequent week of their choosing (n = 1). Accuracy of data collection was ensured using a systematic and tiered-system of training and ongoing supervision. First, all clinicians were trained to record behavior using a standardized data collection protocol. This one hour training included a brief introduction to the data collection procedures, use of study materials, and behavioral observations. Clinicians then completed a reliability set of 15 videos (~ 1 minute in length) with children portraying challenging behaviors. Master codes were confirmed by a PhD level investigator, and clinician reliability ranged from 80–100%. Finally, all clinicians received continuous supervision from their supervisor, who were proficient in data collection, at their respective places of employment.

Partial-interval recording (Cooper, Heron, and Heward 2007) was implemented at five–minute intervals for the entirety of the treatment day to index repetitive behavior, negative affect, aggression, and self–injury. With a partial interval recording system, a behavior is scored as an occurrence if it is observed at least once within a given interval (in this case, a 5-min interval). Similar methods of data collection have been used to evaluate sleep and behavior in children with disabilities (Scheithauer and Zarcone 2015). Clinicians utilized a small vibrating timer (MotivAider®) to signal the end of each five–minute interval, which auto-reset throughout the day. See Table 4 for operational definitions of each challenging behavior.

Table 4.

Models of Total Sleep Time (TST) Predicting Behavioral Outcomes (n = 41)

	Repetitive Behavior		Negative Affect		Aggression		Self-Injury		General Behavior
	Effect	s.e.	Effect	s.e.	Effect	s.e.	Effect	s.e.	Effect	s.e.
Fixed Effects
child level characteristics
Age	.04*	.02	−.02*	.01	.00	.00	.00	.00	.02	.02
Level of functioning=1	−.42***	.08	−.05	.03	−.01	.01	−.02	.01	−.41***	.06
Level = 2	−.38***	.07	−.10**	.03	−.01	.01	−.01	.01	−.39***	.06
Level = 3	0		0		0		0		0
TST within	.05	.05	.00	.05	−.01	.01	−.01	.04	.02	.06
TST between	−1.00***	.29	−.17	.13	.08*	.03	−.01	.01	−.79***	.24
Intercept	1.14***	.28	.36**	0.14	−.06	0.03	.01	.03	1.20***	.24
AIC	−137.63		−299.40		−599.07		−798.16		−113.53
BIC	−122.07		−277.78		−583.51		−782.60		−97.97
−2LL	−147.98		−314.06		−609.42		−808.51		−123.87
ICC within	23%		53.33%		50%		50%		35%
ICC between	77%		46.66%		50%		50%		65%

Note. Level 3 is set to zero because it serves as a reference group; CBC = challenging behavior composite

^*p < .05

^**p < .01

^***p < .001

We computed a challenging behavior composite (CBC) from the four challenging behaviors listed above to indicate whether any behaviors occurred within each interval. These scores were dichotomous and were used to index an occurrence of any challenging behavior (in general), rather than the occurrence of specific behaviors. Further, we calculated a ratio (or percentage) for each challenging behavior (repetitive behavior, negative affect, aggression, self-injury, and the CBC) by dividing the number of five minute intervals in which the behavior occurred by the total number of five minute intervals for each child. These ratios provided an index of each child’s target behaviors per day, and are consistent with previous summative methods used for partial interval recording (specifically in studies evaluating behavior in children with ASD; e.g., Crutchfield et al. 2015).

Child Level of Functioning

Level of functioning was assessed using the severity levels for ASD in the DSM-5: level 1: requiring support (n = 13), level 2: requiring substantial support (n = 14), and level 3 requiring very substantial support (n = 15). The first author assessed level of functioning based on caregiver-report (via an informal interview) and informal observations in the home and clinic settings. Child level of functioning was normally distributed and was used as a covariate in all analyses.

Data Analyses

In the current study, the repeated measures of behavioral and sleep data were modeled using linear mixed effects models implemented with the SPSS GENLINMIXED command. This analytic framework allowed us to analyze how well sleep predicted behavior both within and between children (levels 1 and 2, respectively). That is, we assessed how much an individual’s day-to-day fluctuations in sleep impacted day-to-day variations in behavior (modeled at level 1) in addition to how much a child’s average sleep impacted their overall, average levels of behavior (modeled at level 2). In order to accomplish this, we created two sleep variables from the actigraph data. The sleep variable included at level 1, was the child’s day-to-day fluctuations, centered (Hox & Roberts 2011). Centering at the person-level allowed all level 1 sleep predictors to express differences from each child’s mean sleep estimate. The sleep variable included at level 2, was each child’s average sleep estimate (e.g., average TST or average WASO based on their five nights of recording).

The Restricted Maximum Likelihood Method (REML) was used when estimating all models in order to minimize bias in variance and covariance estimates. Goodness of Fit indices (e.g., AIC, BIC) were used to assess model fit, or the ability of a model to predict a given behavior better than a null model. Proportional reduction in error (PRE), was calculated to provide an estimate of variance explained when predictors were added to the model. The PRE indicates the proportional reduction in residual variance between models (e.g., null model, model with controls, model with controls and predictors). Larger PRE-values are indicative of a larger reduction in error (better model fit).

Results

Descriptive Statistics

Children slept for an average of 493 minutes (or approximately 8 ¼ hours) across all nights of their recording. On average, children woke for 70 minutes (or approximately 1 hour) after sleep onset. Child TST ranged from 278 to 670 minutes (roughly over 4 ½ hours to 11 ¼ hours) and WASO ranged from zero to 171 minutes (almost 3 hours).

In the current sample, children engaged in challenging behavior (regardless of type) at an average rate of 41% (SD = .29) of their total treatment sessions. These behaviors included repetitive behaviors 27% (SD = .33), negative affect 12% (SD = .11), aggressive behaviors 3% (SD = .03), and self-injurious behaviors 2% (SD = .03), all reported as a ratio of behavior/treatment length. Behavior ratios varied by the child’s level of functioning (F (2, 40) = 3.12, p < .05), with the highest overall occurrence (ratio) of challenging behaviors in children requiring very substantial support (M = .62, SD = .31) vs. requiring support (M = .24, SD = .15). Similarly, repetitive behaviors were most common in children requiring very substantial support (F (2, 40) = 3.60, p < .05; M = .49, SD = .40) in comparison to children requiring support (M = .07, SD = .11). Sleep and challenging behavior did not significantly differ by use of sleep medications or other comorbid diagnoses (Table 2), and these variables were therefore not included as covariates in subsequent statistical models.

Multilevel Models

Linear mixed effects models were estimated in a series of steps for each behavioral outcome (starting with the unconditional model). For all behavioral outcomes, every step in the model resulted in a reduction in −2LL, indicating better overall model fit with the addition of new predictors at each step. Results for all models, including fit indices, are provided in tables 4 (TST models) and 5 (WASO models).

Table 5.

Models of Wake After Sleep Onset (WASO) Predicting Behavioral Outcomes (n = 41)

	Repetitive Behavior		Negative Affect		Aggression		Self-Injury		General Behavior
	Effect	s.e.	Effect	s.e.	Effect	s.e.	Effect	s.e.	Effect	s.e.
Fixed Effects
child level characteristics
Age	.06*	.02	−.01	.01	−.00	.00	.00	.00	.03*	.02
Level of functioning=1	−.42***	.09	−.06	.04	−.01	.01	−.02	.01	−.43***	.06
Level = 2	−.30**	.10	−.10**	.03	−.02	.01	−.03*	.01	−.36***	.07
Level = 3	0		0		0		0		0
WASO within	.04	.20	−.02	.07	.02	.03	.40*	.02	−.02	.10
WASO between	.04	.08	.31*	.02	−.02	.07	−.06	.04	.42	.24
Intercept	.28	.18	.16*	.07	.04	.02	.06	.04	.55***	.15
AIC	−115.81		−248.61		−571.08		−591.19		−99.44
BIC	−100.43		−233.23		−555.70		−581.89		−84.06
−2LL	−126.17		−258.97		−581.44		−597.33		−109.80
ICC within	21.33%		56.25%		50%		100%		32.1%
ICC between	78.66%		43.75%		50%		0%		67.9%

Note. Level 3 is set to zero because it serves as a reference group; CBC = challenging behavior composite

^*p < .05

^**p < .01

^***p < .001

Challenging behavior composite.

For TST, children who slept less (on average) engaged in more challenging behaviors (Table 4). For each hour of missed sleep, children engaged in .79 more intervals of challenging behavior (regardless of type per hour; p < .001). For child level of functioning, children identified as requiring support (level 1) engaged in .41 fewer intervals of challenging behaviors per hour on average than those identified as requiring very substantial support (level 3), and children requiring substantial support (level 2) engaged in .39 fewer intervals of challenging behaviors per hour on average than those requiring very substantial support (all p < .01). Additionally, for each year increase in child age, children engaged in .03 more intervals of challenging behaviors per hour (p < .05). Child night-by-night TST (within-child effects) was not associated with CBC scores (Table 4). Child sleep fragmentation (WASO) was not associated with CBC scores on average or when considering night-by-night WASO (Table 5).

Repetitive behaviors.

Average TST was associated with child repetitive behaviors during treatment (Table 4). On average, for each hour of missed sleep (1 SD unit increase), children engaged in repetitive behavior during approximately one additional interval per hour (p < .01). Thus, the less a child slept on average, the more they tended to exhibit repetitive behaviors. Children requiring support (level 1) had .42 fewer intervals of repetitive behaviors per hour on average than children requiring very substantial support (level 3; p < .01). Similarly, children requiring substantial support (level 2) engaged in .38 fewer intervals of repetitive behaviors on average than children in requiring very substantial support level 3 (p < .01). Additionally, for each year increase in age, children engaged in .04 more repetitive behaviors per hour (p < .05). Contrary to hypothesis 2, there were no within-person differences for TST and repetitive behavior.

We also examined relations between WASO and repetitive behavior; however, contrary to hypotheses, there were no significant between- or within-child effects (Tables 4 and 5). Consistent with previous models, child level of functioning and age were positively associated with repetitive behaviors.

Negative affect.

The TST models yielded no significant between-or within-child effects in TST and behavior (Table 4). Children requiring substantial support engaged in .10 fewer intervals of negative affect expressions per hour on average than children requiring very substantial support (p < .05). Additionally, for each year increase in age, children engaged in .02 fewer intervals of negative affect expressions per hour (p < .05). On average during the week, children who woke more at night (higher WASO) engaged in .31 more intervals of negative affect per hour (p < .05). There were no daily (within-child) associations between sleep and behavior. Children requiring substantial support engaged in negative affect during .10 fewer intervals per hour than children requiring very substantial support (p < .05).

Aggression.

There were no significant associations (between- or within-child) between TST or WASO and aggression (Tables 4 and 5).

Self-injurious behavior.

There were no between- or within-child effects for TST and self-injurious behavior. There were also no between-child effects for WASO and self-injurious behavior. However, in line with hypothesis 2, children who spent more of their night awake were more likely to engage in self-injurious behavior the following day. Specifically, for each hour of night waking, children engaged in self-injurious behavior during .40 more intervals the following day (p < .05). Children requiring substantial support engaged in .03 fewer intervals of self-injurious behavior per hour on average than children requiring very substantial support (p < .05).

Discussion

Sleep is a critical component of healthy development in children—promoting cognitive, physical, and emotional growth (El-Sheikh 2011; Mindell and Owens 2015). As noted in our findings and a growing body of literature, sleep difficulties contribute to ASD symptoms and challenging behaviors (Cohen et al. 2017; Hirata et al. 2016; Hundley, Shui, and Malow 2016; Mazurek and Sohl 2016; Scheithauer and Zarcone, 2015; Sannar et al. 2017; Veatch et al. 2017). Within the present study, IBI programs are assessed as a proximal context for development; wherein, sleep concerns could inhibit treatment progress (e.g., Smith, Carr, and Moskowitz 2016), and potentially, a child’s long-term developmental progress.

The current study used real-time objective approaches to measure both sleep (actigraphy) and relations with daytime challenging behavior in children with ASD. TST and WASO were associated with some challenging behaviors, on average—providing partial support for our first hypothesis. However, somewhat unexpectedly, children’s sleep and challenging behaviors did not exhibit significant daily associations. Indeed, sleep is influential in children’s challenging behaviors; however, average patterns over time appear more influential than daily fluctuations.

As hypothesized, children who slept less (per 24-hour period on average) exhibited significantly higher rates of repetitive behavior while receiving intensive behavioral intervention. This finding is consistent with previous research using both parent-report sleep estimates (Gabriels et al. 2005; Hundley, Shui, and Malow 2016; Hoffman et al. 2005; Mayes and Calhoun 2009; Schreck, Mulik, and Smith 2004; Tudor, Hoffman, and Sweeney 2012) and objective sleep estimates (Goldman et al. 2009). However, it should be noted that Goldman and colleagues used the repetitive behavior scale (RBS-R), which includes several subscales of parent-reported repetitive behaviors (e.g., stereotyped, ritualistic and compulsive). In Goldman et al. (2009) children with more parent-reported sleep problems, longer sleep onset latency, and longer WASO also had significantly higher scores on the compulsive and ritualistic behavior scale when compared to children with better sleep, although we did not directly observe these subtypes of repetitive behavior in our study. In contrast to our findings, Goldman and colleagues also noted relations between WASO and overall repetitive behaviors in their sample. We ultimately found no significant associations between WASO and repetitive behaviors, although this may be due, at least in part, to differences in our operationalization and measurement of repetitive behavior.

Consistent with the repetitive sensory motor (RSM) subtype used by Hundley et al. (2016), our operational definition of repetitive behavior included repetitive use of language, movements, and objects. However, we did not measure adherence to functional routines and rituals, sensory-seeking behaviors, or sensory aversions (APA 2013). Although other studies have examined specific subscales of the RBS-R, Hundley and colleagues (2016) were the first to stress the importance of teasing apart this construct into its diagnostic subtypes when studying sleep. They found that, in a large sample of children with ASD, RSM was associated with total sleep problems, while insistence on sameness was not. Hundley et al (2016) also sheds light on the importance of objective, standardized measures of behavior for consistent reporting in the literature.

As expected, on average during the week, children who woke more at night engaged in higher rates of negative affect (Didden et al. 2002; Lukowski and Milojevich 2017). This finding is consistent with Didden and colleagues (2002). However, this association has not been established in studies using either objective measures of sleep or direct assessments of behavior. This lack of research may reflect that negative affect is an extensive and somewhat subjective construct, which is often measured using self/parent report methods.

In addition to between-person effects, we aimed to assess intraindividual changes in sleep and behavior. In the current study, daily associations only emerged for WASO and self-injurious behavior. This broad finding suggests that between-person differences in sleep and challenging behavior are more robust than daily fluctuations. While contrary to our second hypothesis, this finding is consistent with some previous research in children with severe disabilities and low functioning autism (Cohen et al. 2017; Scheithauer and Zarcone 2015). For example, in Scheithauer and Zarcone (2015) there was no significant difference in challenging behavior occurring after a night characterized by low sleep (defined as two or more hours of sleep fewer than recommended by the National Sleep Foundation) compared to nights characterized by average sleep. Similarly, Cohen et al. (2017) found that one night of sleep (i.e., the previous night) did not significantly predict challenging behaviors occurring the following day.

Although there are few published ASD studies assessing day-to-day fluctuations in sleep, similar methods have been used in other populations and suggest that, perhaps, one night of poor sleep (in isolation) may not be particularly concerning (e.g., Wild-Hartman et al. 2013). Rather, overall patterns of poor sleep (or chronic sleep loss) are likely more predictive of negative daytime behaviors (Becker et al. 2016; Van Dyk et al. 2016).

Particularly when examining day-level effects, it is important to acknowledge individual vulnerability and variation in sleep need, circadian rhythm, and response to sleep debt (which may have a genetic component; Kamphuis et al. 2012; Sletton et al. 2015). For example, naturally short sleepers (a sleep characteristic for many children with ASD) have a higher threshold for homeostatic pressure than do individuals who require longer sleep episodes (Aeschbach et al. 2001). Relatedly, individuals who have similar circadian patterns and who sleep for similar amounts each night are differentially affected (in their daytime functioning) when experiencing sleep loss (Doran, Van Dongen and Dinges 2001). Children with ASD may exhibit higher levels of individual variation in response to sleep loss given the heterogeneity of ASD (particularly in our study given the range of ASD severity; Kamphuis et al. 2012).

Consistent with current literature, cumulative sleep debt (or consistently short sleep durations) seems to be driving our between-person effects (and lack of daily associations between sleep and behavior; Goel et al. 2009). Children who receive one night of poor sleep but are able to ‘recover’ or regulate back to a typical sleep schedule (consistent with their individual needs) may not exhibit observable daytime difficulties. It is currently unclear whether such a child would instead experience delayed challenging behaviors after a night of poor sleep (indicative of a lag effect). The possibility of a lag effect depends on individual factors including the magnitude of the child’s sleep disturbance and their ability to recover from initial sleep debt (Kerkhof and Van Dongen 2010).

Strengths and Limitations

To our knowledge, this is the first study to explore observational relations between sleep and behavior in the context of behavioral treatment. Children’s sleep was measured using actigraphy, which allowed us to record data continuously over five treatment days and ultimately glean objective estimates of sleep duration and night waking. Training behavioral clinicians to record data provided a unique and important opportunity to measure key behavioral outcomes, in real time, across full treatment sessions. Although the context and observational approach of this study provide crucial strengths, several limitations of our research also warrant further discussion.

The first limitation in our study is the use of the Social Communication Questionnaire as the sole ASD screening measure for study inclusion. The validity of children’s diagnoses should be supported through standardized diagnostic assessments, such as the Autism Diagnostic Observation Schedule (Lord et al. 2000) and the Autism Diagnostic Interview-Revised (Lord et al. 1994). However, in all of the participating IBI centers in this study, all children were required to supply documentation of a medical diagnosis of ASD prior to the start of IBI. While we asked parents to report all comorbid diagnoses (as part of a brief demographic questionnaire), we did not request further medical or genetic records. As highlighted in Hundley et al. (2016), the Standford Binet Intelligence Scale, the Weschler Intelligence Scale for Children, and the Mullen Scales of Early Learning are among the gold standards for cognitive assessment in children. Additionally, due to the duration and intensity of our data collection, direct interobserver agreement was not feasible. We instead opted to use a standardized training and video-based reliability method. While not frequently used in the context of treatment, this method is common practice in social science research (Chorney, McMurtry, Chambers, and Bakeman 2014; Haidet et al. 2009). The target behaviors observed in this study are routinely monitored in behavioral interventions, and data collectors were subject to ongoing supervision and received continuous feedback at their respective places of employment. Accurate data collection is part of the standard set of skills in people who implement behavioral therapy. Altogether, we are confident that the data they collected accurately reflects participant behavior. Finally, it is possible that compensation for participating families may have biased our results because this package may have appealed to a group of, and not the whole, population of families at the IBI centers. Given that caregivers had a critical role in our study (e.g., ensuring their child wore the actigraph), compensation was likely unavoidable. However, future research should work to mitigate any biases that may occur during the participant recruitment process.

Implications

This study offers important implications for future research, practice, and policy. First, given the high prevalence of sleep problems in individuals with ASD (50–80%) and the critical role of IBI in facilitating optimal development, this research breaks new ground in understanding how improved sleep may benefit children and their families in both the short-term and long-term.

Within the treatment context, challenging behaviors can influence child treatment progress and ultimately their developmental trajectory. Findings from the present study therefore highlight sleep as a potential mechanism to reduce these behaviors (particularly repetitive behavior). While preliminary, this research serves as a crucial starting point for conversations about promoting healthy sleep in children with ASD receiving IBI. Our overall findings are consistent with some previous research (e.g., Cohen et al. 2017; Goldman et al. 2009) and support the use of sleep training or sleep education programs as a part of a child’s treatment plan. Some sleep treatment studies in children with ASD already suggest that improving child sleep will result in improved daytime outcomes (Hodge et al. 2014; Malow et al. 2014). Ultimately, in line with the recent call for action regarding translational research in pediatric sleep (Gruber et al. 2016), our findings will help drive future work aimed to improve sleep and related daytime outcomes in children with ASD.

Table 3.

Operational Definitions of Challenging Behaviors

Variable	Operational Definition
Repetitive Behavior	Any movements without purpose that are repeated in a similar manner and in close succession (e.g., body rocking, swaying, hand flapping, jumping, bouncing, head nods, and spinning/twirling objects) OR any vocal repetition of words or phrases that occur in close succession.
Negative Affect	Any expression of negative emotions (e.g., frowning, crying, whining, or screaming).
Aggression	Any inappropriate physical contact (aimed at a person or object) with force that may leave a mark or result in physical or tangible harm/injury (e.g., hitting, kicking, grabbing, pinching, head butting, biting, or throwing).
Self-injurious Behavior	Any actions that cause or have the potential to cause physical harm or injury to self (e.g., self-hitting, head banging, self-pinching, hair pulling, or biting).