Introduction
Good quality sleep is vital for cognition, affect, behaviour and life satisfaction (National Sleep Foundation, NSF
2004). A survey conducted in America by the NSF (
2004) found that 40 million people suffer some form of sleep disorder. Additionally 60% of adults experience sleep disturbance three or more nights per week. Sleep disorders affect 25–40% of children (Owens
2007), 84% of whom will continue to exhibit them at a 3 year follow-up (Kataria et al.
1987). A longitudinal study conducted by Zuckerman et al. (
1987) noted sleep disorders persist throughout lifetime if untreated. Though prevalent, sleep disorder rates increase for those with developmental disorders such as autism spectrum disorder (ASD). Increased sleep onset latency and night wakings, decreased sleep efficiency and variable sleep patterns have been found in comparison to typically developing peers (Vriend et al.
2011). While 20–40% of typically developing peers exhibit sleep problems, this increases to 40–80% for children with ASD (Souders et al.
2009). The National Sleep Foundation identified children with ASD as a high priority population for sleep research (Mindell et al.
2006).
Sleep disturbances negatively impact a variety of daytime behaviours with a reduction of only 30 min hampering functioning (Sadeh et al.
2003). Poor sleep affects cognitive functions such as memory, learning, attention and mood regulation (Gozal
1998), as well as health, behaviour and overall wellbeing (Mindell et al.
2006). Children with ASD appear to suffer additional side-effects such as increased stereotypy, decreased social skills and increased aggressive behaviours (Schreck et al.
2004). Increased stereotypy has been found to be related to anxiety for individuals with ASD (Schlaggar and Mink
2003) suggesting that anxiety may be connected with sleep disorders.
In order to understand why sleep difficulties occur, one must consider biology. All individuals adhere to the circadian cycle of sleep and wake (Finger
1982). Infant sleep differs greatly from adult sleep in that an infant experiences quicker transitions between rapid eye movement (REM) and non-REM phases than an adult (Jin et al.
2013). Electroencephalogram studies show REM sleep precedes arousal and cause wakings (Anders
1979). As a result of increased REM frequency infants are more likely to wake. These cyclic-REM arousals represent a universal vulnerability in all infants which if combined with a maladaptive parenting style (i.e., a style that aids the child in accessing contingencies of reinforcement for sleep incompatible behaviours) make infants prone to sleep difficulties (Anders et al.
1992).
When seeking help for sleep disturbances, often, the first port of call is the paediatrician (Jin et al.
2013). Of paediatricians, 96% feel sleep management falls within their role (Faruqui et al.
2011) while only 18% have received formal training. Medication results from 80% of doctor visits for sleep (Stojanovski et al.
2007). Medication presents a number of limitations, such as poor parental acceptance, long term inefficacy (France et al.
1991), day time hangover effects (Wiggs and France
2000) and poor empirical support (Rosen et al.
2002). Additionally, medication per se does not facilitate access to the environmental contingencies surrounding sleep and as such does not lead to learning of appropriate sleep onset behaviours. One form of medication which has demonstrated some success for children with ASD is melatonin (Cortesi et al.
2010). Melatonin has been found effective at increasing sleep duration and reducing sleep onset latency (Kuhn and Weidinger
2000). However melatonin lacks controlled research findings (Rossignol and Frye
2011). Benefits of medication include its ability to establish sleep as a potent reinforcer (Mindell
1999) and to reduce crying and conditioned vomiting (France et al.
2003). Overall, although there seems to be some evidence to support drug use, this is not sufficient and makes pharmaceuticals unnecessary to obtain desired results.
An alternative to medications are behavioural interventions. A systematic review of behavioural sleep interventions found that 94% of all studies resulted in clinically significant improvements (Mindell et al.
2006). One of the most common interventions is stimulus control, which posits that falling asleep is an operant behaviour reinforced by sleep itself (Bootzin
1977). Falling asleep can be thought of as the final step in an operant behaviour chain (Ferster et al.
1975). This chain commences with pre-bed behaviours and finishes with behavioural quietude prior to sleep onset (Blampied and France
1993). Behavioural quietude allows for internal cues such as sleepiness or tiredness to be discriminated (Blampied and France
1993). Activity acts as a competing response and may interfere with the discrimination of natural cues for sleep onset thus delaying sleep. Halting activities allows exposure to sleep-preparatory discriminative stimuli (SDs) increasing sleep onset likelihood. This perspective is supported by sleep latency data (Carskadon and Dement
1987).
A central assumption of this theory is that if consistent sleep is to occur, steps in the behavioural chain must come under stimulus control of suitable SDs (Bootzin
1977). As is the case with any behaviour, discriminative properties are obtained by stimuli in the environment which reliably signal the availability of reinforcement, in this case sleep. Typical SDs for behavioural quietude are dark lighting, cool temperatures and bedding, the absence of which can lead to individuals experiencing difficulties with sleep onset.
Motivating operations (MOs) are also implicated in sleep onset. MOs such as quality of previous sleep, previous sleep duration and time since previous sleep all alter the reinforcing value of sleep and the likelihood of behaviours which previously gained access to sleep (Michael
1982). As such the value altering and behaviour altering aspects of MOs are important to sleep onset. A number of interventions have been established from this theoretical perspective, two of which hold numerous empirical analyses, positive routines and bedtime fading.
Positive routines consist of a series of pleasurable calming activities undertaken during wakefulness to facilitate sleep onset (Gruber et al.
2011). A routine is defined as a repetitive and observable behaviour pattern (Koulougliti et al.
2008), which contains five to seven activities and takes between 30 and 40 min to complete (Durand
1998). Completion of each step of a routine is praised informing the child of the transition to the next step (Morgenthaler et al.
2006). Activities move from rich to lean reinforcement and from active (e.g. running) to passive (e.g., reading,). Positive routines attempt to establish appropriate sleep onset SDs by establishing a behavioural chain terminating in behavioural quietude. The terminal reinforcement for completing this chain is sleep onset. Studies have posited routine occurrence rather than content is important (Kodak and Piazza
2008). Sleep resumption requires the presence of these sleep associated SDs (Ferber
1990). Thus if these SDs are encountered during wakings, independent sleep resumption is likely. Parents often implement positive routine interventions (Owens
2006). While positive routines are often combined with other intervention components, minimal research to date has evaluated its individual efficacy (Mindell et al.
2006).
Positive routines were first utilised by Milan et al. (
1981) as a means to combat negative side effects experienced with extinction. This was found to be effective in reducing night-time tantrums for three children with severe intellectual disabilities within the home setting. However the inclusion of a fading aspect prevents conclusions to be drawn on the efficacy of the positive routines in isolation. A number of follow up studies utilised positive routines within multicomponent interventions (e.g., Adams and Rickert
1989; Knight and Johnson
2014; Christodulu and Durand
2004). Only two studies have evaluated positive routines in isolation. Mindell et al. (
2009) evaluated the efficacy of positive routines for 405 typically developing toddlers aged between 7 and 36 months and found them to be effective at reducing sleep onset latency and night wakings. A more recent evaluation by Mindell et al. (
2015) examined the impact of routine ‘dose’ on efficacy. This study included over 10,000 typically developing participants and found positive routines to be effective with routine exposure negatively correlating with bedtime difficulties. However both of the above studies involved typically developing participants therefore conclusions cannot be drawn for individuals with ASD.
A second stimulus control intervention is bedtime fading, the central aim of which is to manipulate the sleep wake cycle to increase sleep likelihood. Bedtime fading involves temporarily moving bedtime to more closely coincide with the child’s natural sleep onset, so as to ensure rapid sleep initiation, and then fading this earlier if sleep onset latency remains short according to developmental norms and parental wishes (Morgenthaler et al.
2006). A scheduled wake time is also established and sleep is not permitted outside of these times. This acts to establish sleep as a potent reinforcer. Bedtime fading relies on internal cues of sleepiness caused by sleep deprivation. These serve as establishing operations which increase the reinforcing value of sleep for the child. This facilitates a smooth transition and quick sleep onset, preventing inappropriate behaviours. A central assumption is that sleep resumption following arousal will be facilitated as the SDs associated with sleep onset are readily available (Ferber
1990).
The use of bedtime fading was first shown effective with a 6 year old girl with attention deficit hyperactivity disorder experiencing multiple sleep difficulties (Piazza and Fisher
1989). Piazza and Fisher (
1991) also used bedtime fading to increase total sleep duration for two typically developing children within an inpatient facility. The present study was based on Adams and Rickert (
1989) which combined positive routines with a simplistic version of bedtime fading. Substantial research into bedtime fading within multi-componential packages has been conducted (e.g., Piazza and Fisher
1991; Ashbaugh and Peck
1998; Christodulu and Durand
2004), while one study has evaluated bedtime fading in isolation (DeLeon et al.
2004) noting an 81% reduction in night wakings and an 82% reduction in self-injurious behaviour for a 4 year old boy with ASD.
While positive routines and bedtime fading have demonstrated efficacy with typically developing children, individual efficacy and efficacy with ASD cannot be established (Mindell et al.
2006). The present study aims to answer three questions. First, it seeks to evaluate if bedtime fading and positive routines are effective in decreasing sleep onset latency and night wakings in young children with ASD. Second, it aims to respond to the question of whether bedtime fading and positive routines are effective in increasing the total sleep duration for young children with ASD. The third question is whether primary caregivers can successfully implement positive routine and bedtime fading interventions. In the same line, the study aims to examine whether these interventions are acceptable to parents.
Discussion
This study sought to evaluate the impact of two stimulus control sleep interventions on total sleep duration, night wakings and sleep onset latency for young children with ASD. The interventions investigated were positive routines and bedtime fading.
Results obtained suggest the efficacy of parent implemented bedtime fading for young children with autism. Bedtime fading was found to increase total sleep duration and decrease sleep onset latency for all participants. As minimal night wakings were exhibited no conclusions can be made on intervention effects. Secondary dependent variable measures also demonstrated increased learning opportunities and decreased challenging behaviour following intervention. Results obtained for positive routines note decreased sleep onset latency for all participants following intervention implementation. Total sleep duration was found to increase for two of the three participants. However these effects were minimal. Both interventions were also found to have high social validity, treatment fidelity and acceptability by parent participants.
Results are consistent with those of Piazza and Fisher (
1989,
1991). This study expanded on these findings by demonstrating efficacy within the home environment using parents as active change agents. Further, Piazza and Fisher (
1989,
1991) examined bedtime fading within a multicomponent intervention whereas this study demonstrated efficacy as individual interventions. It should be noted that though participants demonstrated increased total sleep durations, none met target sleep durations within the study. However as Sadeh et al. (
2003) noted an increase of only 30 min can significantly impact a child’s daily behaviours. As such, the gains made are socially significant for participants and were further confirmed by results obtained in the TEI-SF and TAI. Total sleep duration may have been hampered by the strict sleep/wake times which form the intervention. Decreased frequency of challenging behaviour is also consistent with previous research (Piazza et al.
1991).
Though the results obtained support the efficacy of bedtime fading as an independent intervention for young children with ASD, positive routines results were not as robust. Introduction of the positive routines intervention did result in decreased sleep onset latencies for all participants. However, target sleep onset latencies were not consistently met and mean latencies remained high. It should be noted that baseline sleep onset latency was higher within positive routines than bedtime fading. This differs somewhat from Mindell et al. (
2009) who noted decreased sleep onset latency (to target levels) following positive routines implementation for typically developing young children. The lack of an ASD diagnosis for participants within Mindell et al. (
2009) study however may account for the discrepancies in results obtained. Total sleep duration data also present variability. Though duration increased for both Martin and Alan, John noted a small decrease. However John’s parents reported that he was ill throughout the intervention and that this may have impacted sleep. These results somewhat support the findings of Milan et al. (
1981) and Adams and Rickert (
1989) who found positive routines effective in increasing total sleep duration within the home environment. These studies however also included a bedtime fading element. This may account for the slight difference in results obtained and prevents direct comparison of effects. Night wakings data were only available for Alan with neither Martin nor John exhibiting any night wakings during the implementation of the study. Decreased frequency and duration of night wakings were noted with the decrease in duration occurring quickly. As results are not conclusive and no previous studies have evaluated the impact of positive routines on night wakings this area is particularly in need of further empirical analysis.
Secondary dependent variables also present mixed results for positive routines. No notable change to challenging behaviour was noted. With regards to educational units both Martin and Alan reported minimal increases with no change noted for John. However this is to be expected considering the lack of effects on primary dependent variables. While positive routines appear somewhat effective it is clear that further empirical analysis is warranted to fully understand its effects as an individual intervention rather than as a one of several components within an intervention package (Mindell et al.
2006).
The behavioural principles underlying both interventions may shed light on these differing results. The key assumption of a stimulus control approach to sleep is that for consistent sleep to occur, steps in the behavioural chain must come under stimulus control of appropriate discriminative stimuli (Bootzin
1977). While positive routines rely on stimulus control as the central mechanism of behaviour change, bedtime fading also utilises motivating operations. Research has noted the role of establishing operations (EOs) in sleep onset (Michael
1982). EOs such as quality, duration and time since previous sleep all alter the potency of sleep as a reinforcer. Bedtime fading functions by placing the child in bed at a time at which sleep onset is hypothesised to be quick. Pushing bedtime later may have had both value altering (sleep becomes a more potent reinforcer) and behaviour altering effects (increased likelihood of engaging in sleep appropriate behaviours). As such, the use of EOs within bedtime fading may have positively impacted intervention effects. Future research should further examine the role of EOs within sleep interventions.
This study further adds to existing research by demonstrating that slight intervention alterations can be made without impacting efficacy. For example, within the bedtime fading intervention the response cost element was removed. In previous studies had the child remained awake in bed for 15 min following the parent bidding them goodnight and leaving the room a response cost procedure was implemented (Piazza et al.
1997). This study differed in that response cost was omitted entirely. This was consistent with DeLeon et al. (
2004) who noted bedtime fading as effective in the absence of a response cost element. A second slight alteration within this study was that bedtime was faded by only 15 min per night if the child demonstrated short sleep onset latency. This differed to Piazza and Fisher (
1991) and Ashbaugh and Peck (
1998) who faded bedtime by 30 min. This alteration may have increased the nights required to meet target sleep/wake times. This may in turn have prevented participants reaching target sleep durations within the timeframe of the research. Positive routines also permitted alterations in that routine content and duration was individualised to match child preferences and family wishes. Visual schedules were used to facilitate child participants transitioning through stages on the schedule. This differed from Mindell et al. (
2009) and Christodulu (
2000) where no visual supports were used.
Though the results of this study support positive routines and bedtime fading for young children with autism, limitations are apparent. First, the absence of objective measures of sleep present. The primary method of data collection was parent completed sleep diary. Research notes diaries have high validity, internal consistency, and agreement with objective measures (Mindell et al.
2006). Their self-report nature however presents caveats to results obtained. Participant bias can occur with self-report (Wade et al.
2007). Sleep diaries only contain data of which parents are aware (Sadeh
1994). As such total sleep duration may have been overestimated as parents would have been asleep, thus unable to record data. Night wakings may have been underestimated for similar reasons (Ashbaugh and Peck
1998; Sadeh
2011). Schreck and Mulick (
2000) however noted that although parents may underestimate wakenings those missed cannot be considered problematic as they omit overt signals of wake (e.g. crying, calling out, etc.).
Self-report may have also impacted IOA data reliability. While parents may have influenced each other during data collection, there are no data indicating this. Future studies could use videotaping or digital devices registering physiological activity as an alternative method of collecting IOA data. Sleep diaries do present a number of benefits. The primary benefit was inclusion of data on events outside the bedroom (e.g. illness, staying in grandparents, etc.). This was valuable in terms of identifying extraneous variables which may have impacted adherence and efficacy. Future research should investigate the use of more objective measures of sleep and the reliability, accuracy and validity of parent-completed sleep diaries.
A further limitation in this study is the use of the CSHQ (Owens et al.
2000). Recent research completed by Johnson et al. (
2016) noted inconsistent internal consistency ratings across subsets of the CSHQ. Additionally they raised queries on the applicability of the subsets to individuals with ASD. However as the CSHQ was used to establish the degree of severity of the sleep difficulties rather than as a pre/post measure its use within this study appeared warranted. It is clear that further research into the CSHQ is required to fully establish its effectiveness.
Inclusion criteria present a further limitation. To enrol, participants were required to be aged 2–7, sleep less than 7 h per night, present with a sleep difficulty on the SATT and have ASD. Inclusion criteria may not have been stringent enough and may have impacted generalisability. Firstly no reference to severity of ASD diagnosis or IQ was made. Though sleep disturbances are common across the spectrum, functioning level is pivotal when analysing behavioural interventions (Williams et al.
2004). Children with ASD who are non-verbal may require additional supports or prompting to adhere to interventions while higher functioning children may not. Though level of functioning was referenced by including developmental age (derived from VB-MAPP) this may not have been sufficiently precise.
A final limitation was the absence of function based interventions. Participants were randomly allocated to an intervention regardless of SATT or CSHQ scores. Further functional relations surrounding sleep were not investigated. Functional assessments are the hallmark of behavioural intervention (Hanley et al.
2003). Sleep problems have been exempt from such analyses though they represent a severe challenge for child and family. Though effective interventions have been documented, conditions under which each should be applied have not been established. Jin et al. (
2013) reported the efficacy of individualised function based sleep interventions for young children. The SATT was utilised to provide the researchers with a hypothesised function of the sleep competing behaviours. The lack of environmental contingency evaluation of participants sleep difficulties may have adversely impacted results. Future research should examine the functional relations and use these to guide intervention.
These results impact current clinical approaches in a number of ways. The apparent efficacy of stimulus control techniques is beneficial as it minimises the need to utilise extinction based protocols which often have multiple side-effects (Jin et al.
2013). Reid et al. (
1999) noted a drop-out rate of 20% of parent participants from extinction based interventions due to a reluctance to ignoring crying. Stimulus control interventions in comparison do not result in prolonged periods of crying or increased rates of challenging behaviours (Burke et al.
2004). No parent participants within this study noted increased challenging behaviour associated with sleep related stimuli. Further challenging behaviour decreased for participants within the bedtime fading intervention. A key advantage of stimulus control interventions is the facilitation of skill acquisition where extinction does not permit this (Cortesi et al.
2010). Finally, intervention implementation at the child’s bedtime is advantageous. At this time parental fatigue is minimised increasing implementation fidelity (Kodak and Piazza
2008). The high social validity, treatment fidelity and acceptability scores obtained within this study also support the positive effects of implementing sleep interventions at this time. This suggests that future research on stimulus control interventions rather than extinction based approaches is warranted.