Sleep Quality in Autism from Adolescence to Old Age

Abstract

Background:

Sleep problems are common in autism from early childhood. Although research suggests that poor sleep continues at least into early middle age, the pattern of sleep problems has rarely been characterized beyond childhood. The aim of this study was to examine sleep quality from adolescence to old age in autistic individuals as compared with age-matched non-autistic comparison groups from the general population.

Methods:

Participants ranged from 15 to 80 years; there were 297 participants in the autistic group (mean [M]_age = 34.36 years, standard deviation [SD] = 15.24), and the comparison group had 233 participants (M_age = 33.01 years, SD = 15.53). Sleep quality, sleep onset latency (SoL), total night sleep, and sleep efficiency as measured by Pittsburgh Sleep Quality Index were compared between groups and across age groups (15–19, 20–39, 40–59, 60+ years). Five predictors of sleep quality (autistic traits, mental health condition, medication, employment, and sex) were also examined.

Results:

Overall, problematic sleep was more common for the autistic participants (63.7%) than the comparison group (46.4%), and autistic participants had poorer sleep quality and longer SoL (all p < 0.001). In early adulthood and middle age, autistic adults had significantly poorer sleep quality and longer SoL than similar age comparison group adults; autistic and comparison group adolescents and the elderly did not differ. In the autistic group, predictors accounted for 21% of sleep quality variance. Sex (p < 0.001) was the strongest predictor, with all predictors except employment contributing unique variance. In the comparison group, predictors accounted for 25% of the variance in sleep quality. The strongest predictor was mental health condition (p < 0.001), with all predictors except sex contributing unique variance.

Conclusions:

Autistic adolescents and adults, particularly females, remain vulnerable to sleep problems, with early and middle adulthood being at times of particular risk. Targeted sleep interventions are required.

Lay summary

Why was this study done?

Difficulty sleeping is a common occurrence among autistic individuals, but we know very little about sleep in autistic adults.

What was the purpose of the study?

To compare self-reported sleep quality in autistic and non-autistic people aged 15 to 80 years.

What did the researchers do?

Online surveys were completed by 297 autistic individuals (average age 34.36 years) and 233 non-autistic individuals (average age 33.01 years). Participants were asked questions about their sleep quality, the time it takes them to fall asleep (sleep latency), and the number of hours of sleep they usually get each night (total sleep). Using information about how long they slept and their responses to questions about their bedtime and wake time we calculated the percentage of time they spent in bed asleep (sleep efficiency [SE]). We compared these sleep measures between the autistic and non-autistic participants. We also split the participants into four age groups (15–19, 20–39, 40–59, and 60+ years) to look at any differences at specific age points. Finally, we looked to see whether autistic symptoms, having a mental health problem, being on medication, being unemployed, and/or sex (male/female) predicted sleep quality.

What were the results of the study?

Poor sleep quality was more common for the autistic participants (63.7%) than non-autistic participants (46.4%). On average, autistic participants also had poorer sleep quality scores and it took them longer to fall asleep than non-autistic participants. Autistic participants in early adulthood (20–39) and middle age (40–59) had poorer sleep quality and took longer to fall asleep than non-autistic adults of the same age. There were no differences between autistic and non-autistic adolescents (15–19) or older adults (60+). For autistic participants, the best predictor of poor sleep quality was being female; other predictors of poor sleep quality were having a mental health problem, more autistic symptoms, and being on medication. Among non-autistic participants, the best predictor of poor sleep quality was having a mental health problem; other predictors were more autistic symptoms, being on medication, and being unemployed.

What do these findings add to what is already known?

Similar to the findings in autistic children, autistic adults are more likely to have poor sleep quality compared with non-autistic adults. Autistic females are particularly at risk for poor sleep, and autistic adults aged 20 to 59 years are more at risk for poor sleep quality.

What are potential weaknesses in the study?

Sleep was measured by using a self-report questionnaire, which is not as reliable as using a sleep diary or other objective measures of sleep (e.g., actigraphy). In addition, this study only looked at data collected at one point in time, and as such it is not possible to examine changes over time in sleep quality among autistic adults.

How will these findings help autistic adults now or in the future?

The findings in this study identified that sleep difficulties persist across the lifespan for autistic adults. Therefore, there is a critical need for future research to focus on understanding the cause of poor sleep quality in autism and develop sleep interventions for autistic adults.

Introduction

Inadequate sleep can negatively affect neurobiological, physical, and psychological health, and it can impair daytime functioning.^1–3 Sleep difficulties are common in the general population, with 33%–45% of Australians having poor sleep,⁴ whereas 16% to 27% of adults across six countries reported that they rarely or never got a good night's sleep.⁵ Non-autistic adults with poor sleep quality generally report insomnia symptomatology (sleep onset and maintenance difficulties) and daytime fatigue.⁶ Of greater concern is that up to 80% of autistic children up to age 18 years⁷ and 54% of autistic adults aged 18 to 65 years⁸ have sleep difficulties. The primary sleep parameters reported as problematic in autistic adults include reduced total sleep time (TST) and sleep efficiency (SE; the proportion of time in bed spent sleeping), and increased sleep onset latency (SoL) and night waking, which are associated with insomnia, circadian rhythm sleep disorders, and daytime fatigue.^8–17 A considerable amount is now known about sleep difficulties in autistic children, but the body of sleep research for autistic adolescents and adults remains limited.

Although both insomnia and poor sleep quality are reported in autism and are often used interchangeably, they are not exactly equivalent. Insomnia is a sleep disorder defined by the presence of sleep onset and maintenance difficulties, with impact on daytime functioning over at least 3 months.⁹ However, sleep quality is complex and more difficult to define¹⁸ and poor sleep quality can be reported despite adequate sleep being shown on polysomnography (PSG).¹⁹ Good sleep quality includes (1) that at least 85% of time in bed is spent sleeping (SE)¹⁸; (2) taking less than 30 minutes to fall asleep (SoL)¹⁸; (3) waking for 20 minutes or less after falling asleep¹⁸; and (4) getting sufficient total sleep for age (TST).^18,20 Environmental factors such as noise and light can also affect sleep quality.¹⁸ In addition to these features of sleep quality, sleep also has an internal structure, which is measured by using PSG and defined by sleep stages (i.e., rapid eye movement sleep [REM] and light sleep and deep sleep [Non-REM]). Sleep structure also shows age-related changes and can be used as an indicator of sleep quality.¹⁸ For example, PSG measurement of SE and TST, sleep stage transitions, and usual sleep quality reported on the Pittsburgh Sleep Quality Index (PSQI)²¹ contributed to self-reported overnight sleep depth and restfulness in older adults, with SE being the best indicator of sleep quality.¹⁹ Thus, a range of sleep variables can be indicators of poor sleep quality.

Age can also differentially affect sleep quality, but to date studies have not examined age as a factor in sleep quality for autistic adults through to old age. The limited number of studies in autistic adolescents and adults currently report rates of poor sleep quality or insomnia ranging from 18% to 90%^8,9,12,16,22 using varying age ranges, sample size and composition, definitions of poor sleep, and sleep measurement. The first meta-analysis of sleep in autistic adults identified only eight data sets (sample size range 15–41 autistic adults; mean age range 21–47 years) for inclusion, concluding that on all subjective and objective measures autistic adults' sleep was impaired compared with non-autistic individuals.²³ In Australian general population samples, sub-optimal sleep (<8 hours) on school nights is reported by 38% of adolescents, with 32% reporting sleep onset difficulties and 48% reporting unrefreshing sleep²⁴; and adults aged 18–24 years were less likely to report adequate sleep than older adults.⁴ A population study of adults age 18–65 years found that the lowest odds of reporting insufficient sleep was being older than 50 years, whereas overall 24% of adults reported insufficient sleep that was related to chronic disease, lifestyle factors, and being female.²⁵ In old age, medical conditions and pain, mental health conditions, and circadian changes also can contribute to poor sleep.^26,27 A review of studies from across the world comprising adults aged 60 years or older reported insomnia symptoms in up to 62% of individuals, with a few reported rates less than 25%.²⁸

Sex differences in sleep in autism are largely unexplored, and nothing is known about autistic adolescent girls or autistic women. One child study suggested that autistic girls may be less susceptible to sleep problems than autistic boys, but there were only 6 girls and 22 boys.²⁹ Conversely, a study of autistic pre-schoolers showed that girls aged 1 to 4 years exhibited more sleep problems than boys.³⁰ Potential sex differences in sleep quality are reported in the general population, with women reporting poorer sleep quality than men.^6,25,29–34 However, Kaplan and colleagues¹⁹ found that although older women tended to report sleeping more deeply and being more rested than older men, conversely, they also reported poorer sleep quality on the PSQI. Thus, additional factors may contribute to women's perception of sleep quality.¹⁹ In women, increased risk for poor sleep also can be associated with menopause symptoms, particularly hot flushes.^35,36 Hot flushes are associated with neurovascular control by the autonomic nervous system and vary across females in both age of onset and severity. They are indicative of “autonomic neurovascular dysregulation” and may be due to individual differences in genes associated with relevant hormone and neurotransmitter pathways.³⁵

Poor sleep in autistic adolescents or adults has been related to anxious and depressive symptomatology,^16,37 pre-sleep arousal,^37,38 reduced quality of life,¹³ unemployment,³⁹ and core traits of autism,^8,32 which is concerning. It is well established that similar relationships exist in the general population where individuals with poor sleep are at risk of significant negative health consequences,^19,20 including poor mental health,^26,40,41 increased autistic traits,²⁷ and unemployment.^31,42 Further, medications associated with mental health conditions also can influence sleep quality,⁴³ and for non-autistic individuals with comorbid insomnia and depression, treating both conditions results in better outcomes, including susceptibility to depression relapse.²⁶

Thus, poor sleep is an emerging lifespan issue for autistic adults, who, like autistic children, appear to have poorer sleep than non-autistic individuals, which is potentially related to a range of negative health consequences. However, we know of no research that examines a range of sleep quality variables across the lifespan, or that specifically examines sleep in older autistic adults or women. Consequently, a better understanding of the extent and presentation of sleep problems in individuals diagnosed with autism spectrum disorder across the lifespan is needed. Using two large Australian cohorts, our aims were to examine: (1) sleep quality in a cross-sectional sample of people on the autism spectrum with no intellectual disability from mid-adolescence to old age, compared with a similar age, non-autistic comparison group; and (2) the influence of autistic traits, mental health, medication, employment, and sex (male, female) on sleep across this age range compared with a comparison group.

We predicted that the proportion of autistic adults with poor sleep quality on the PSQI would exceed a comparison group, with poorer sleep quality, increased SoL, and reduced TST and SE for the autistic group. Across the lifespan (15–19, 20–39, 40–59, 60+ years), we predicted that autistic age groups would have poorer sleep than comparison age groups on all sleep variables. We also predicted that for both groups, sleep would be poorer in adolescence and old age compared with middle age, and that autistic traits, mental health, medication, employment, and sex would significantly predict sleep quality, with mental health being the strongest predictor of sleep quality for both groups.

Methods

Participants

Participants aged from 15 to 80 years were from the baseline data sets of two longitudinal studies, the Study of Australian School Leavers with Autism (SASLA), and the Australian Longitudinal Study of Adults with Autism (ALSAA). The autistic group included 150 males and 163 females; 11 individuals selected an LGBTI (lesbian, gay, bisexual, trans or intersex) category. The comparison group had 59 males and 196 females. No participants reported an intellectual disability. Majority of participants were not compensated at baseline. Those recruited into SASLA via an online recruitment site (primarily comparison group participants) received a $10 gift voucher.

Materials

Demographics

This included current autism diagnosis, age, gender, highest level of education, employment, studying, current mental health condition, and current medication. In addition, ALSAA participants also reported whether they were married or were retired.

The autism quotient-short.⁴⁴

The AQ-Short is a self-report measure consisting of 28 items, which assess autistic traits in individuals with normal intelligence. Participants respond to each item on a four-point Likert Scale; a score of >65 indicates a high likelihood of an autism spectrum disorder diagnosis (sensitivity = 0.97; specificity = 0.82).

Pittsburgh Sleep Quality Index.²¹

This self-report questionnaire assesses sleep habits over the past month and a global sleep quality score is calculated from 7 subscale scores, including SE; SoL, wake time, and sleep duration (TST) are reported in questions 2–4. A global score >5 indicates poor sleep quality (specificity = 89.6%, sensitivity = 86.5%). High internal consistency is reported (α = 0.83). Recently, the single-factor PSQI global score for adolescents (13–18 years) was confirmed, with good internal consistency (α = 0.73).⁴⁵ The PSQI also has shown good internal consistency (α = 0.68) for autistic adults (21–41 years).⁹

Procedure

Both SASLA and ALSAA have ethics approval from the relevant university ethics committees. SASLA was designed to follow autistic individuals aged 15.0 to 25.0 years at entry (baseline), and 1 and 2 years after baseline, and an age-matched group of non-autistic individuals across a range of demographic, mental, and physical health, social, social support, and adaptive behavior variables. ALSAA was similarly designed to follow autistic and non-autistic individuals aged from 25 years at baseline, and 2 years after baseline.⁴⁶ To allow the lifespan to be examined, questions and questionnaires across the two studies were designed, where possible, to be compatible. Both studies are conducted online by using Qualtrics ©. Longitudinal data collection is ongoing for both studies, and ALSAA has a rolling baseline. Recruitment occurred Australia-wide through a variety of sources, including state autism associations, autism support groups (autistic participants only), online recruitment (including a recruitment site, Rulo), advertisements, and word of mouth. Participants expressed their interest in participating by contacting either the SASLA or the ALSAA research team, usually via email. They then received relevant information about the study and informed consent. All ALSAA participants provided consent online; SASLA participants provided either written or online consent depending on recruitment route and age. Parents also provided consent for participants younger than 18 years. After consent, each participant received a unique link to the relevant online survey.

Data screening, cleaning, and analysis

Demographic variables across the datasets were compared, matched, and recoded and the data files were merged. Where relevant, demographic variables were dichotomized, and new variables were created from existing variables. Mental health conditions (e.g., anxiety, mood disorders, psychosis, post-traumatic stress disorder) are frequently associated with poor sleep^26,40,41,47; however, participants reported one or more of a range of mental health diagnoses. Thus, it was not feasible to examine any effect of individual conditions and mental health was dichotomized as presence/absence of a current mental health condition, “yes” (0) or “no” (1). Medication and employment were dichotomized “yes” (0) or “no” (1) based on participant responses, and sex was entered as male (0) or female (1). The PSQI global score, SoL, TST, and SE categories (≥85%, <85%) are the sleep variables analyzed here. When participants provided a time range for SoL (e.g., 30–60 minutes) or TST (5–7 hours), the mid-point was used as per scoring instructions. To establish a developmental picture, participants were grouped into four developmental age bands: 15–19 (adolescence), 20–39 (early adulthood), 40–59 (middle adulthood), and 60+ (older adulthood).

The SASLA had 357 participants (141 autistic, 216 comparison groups), and there were 399 participants in the ALSAA (253 autistic, 146 comparison group). Parents reported on sleep for 21 participants with ID, and there is also some evidence that sleep in those with autism and ID differs from autistic individuals with no ID^11,48; thus, the 33 autistic individuals with ID were removed. Participants with no reported formal autism spectrum disorder diagnosis (n = 31), or who did not complete the AQ-short (a school cohort, n = 99) also were removed. Screening for missing values and outliers revealed 10 participants with 80% or more missing data who were then removed.

Missing data were not replaced for the AQ-Short. PSQI scoring instructions preclude replacing missing data; however, participants without a global score may have a valid response for individual items or have valid subscales scores. Thus, participants who had a score for at least one sleep outcome variables of interest, sleep quality (PSQI global score), SoL (single item), TST (single item), or SE (subscale) were retained for analyses. Age was missing for one autistic participant who was retained for total group analyses where age was not a variable of interest.

Results

The comparison group had 233 participants (mean [M]_age = 33.01 years, standard deviation [SD] = 15.53), and 297 participants self-reported an autism spectrum disorder diagnosis (M_age = 34.36 years, SD = 15.24). There were more females in the comparison group, χ² (1) = 30.30, p < 0.001, ϕ = 0.246. As expected, the autistic group had a higher AQ-short score (N = 286, M = 83.77, SD = 12.53) than the comparison group (N = 224, M = 58.31, SD = 12.55), t(508) = 22.75, p < 0.001. The autistic group was more likely to be unemployed, have no partner have a mental health condition and be on medication, and generally had less formal education than the comparison group (Table 1).

Table 1.

Demographic Variables Compared Across Diagnostic Groups

Characteristic	Comparison group		Autistic group		X (df)	p	Φ
	N	% (n)	N	% (n)
English spoken at homea	232	97 (225)	294	99.7 (293)	—	0.024	—
Currently employed	221	76.5 (169)	259	52.5 (136)	28.53 (1)	<0.001	0.248
Currently studying	179	59.8 (107)	226	54.4 (123)	0.96 (1)	0.328	0.054
Highest level of education	233		294		30.69 (4)	<0.001	0.235
High school not completed		3.9 (9)		13.6 (40)
Completed high school		28.3 (66)		23.1 (68)
TAFE/diploma/trade		14.6 (34)		25.2 (74)
Undergraduate studies		34.8 (81)		24.8 (67)
Postgraduate studies		18.5 (43)		15.3 (45)
Retiredb	69	18.8 (13)	111	16.2 (18)	0.06 (1)	0.802	0.034
Current partnerb	125	75.2 (94)	180	46.1 (83)	24.45 (1)	<0.001	0.290
Current mental health condition	224	27.7 (62)	291	64.3 (187)	66.37 (1)	<0.001	0.363
Current medication	231	39.4 (91)	295	64.4 (190)	31.58 (1)	<0.001	0.249

Note that some participants may be both working and studying.

^a Fisher's exact.

^b Only answered by ALSAA participant.

ALSAA, Australian Longitudinal Study of Adults with Autism.

Sleep quality

In the autistic group, 63.7% (n = 278) had poor sleep quality compared with 46.4% (n = 220) of the comparison group χ² (1, N = 498) = 14.23, p < 0.001, ϕ = −0.173. The two diagnostic groups also differed significantly on their global sleep quality score (Table 2). The prevalence of poor sleep quality did not differ across age groups within the comparison group, but it did vary significantly with age within the autistic group. Autistic individuals aged 40–59 years had the highest frequency of sleep problems, though no standardized residual was ≥2. Comparing the autistic and comparison groups across age groups, significantly more autistic participants in the 20–39-year-old [χ² (1, N = 238) = 12.13, p < 0.001, ϕ = −0.234] and 40–59-year-old [χ² (1, N = 121) = 9.58, p = 0.002, ϕ = −0.299] age groups reported poor sleep quality (Table 3).

Table 2.

Pittsburgh Sleep Quality Index, Sleep Onset Latency (Minutes), and Total Night Sleep (Hour) for Autistic and Comparison Groups

PSQI	Group		t (df)	p	r
	Comparison group	Autistic group
Global score	5.94 (SD = 3.41)	7.45 (SD = 3.95)	4.58 (492.19)	<0.001	0.20
	n = 220	n = 278
SoL (minutes)	29.03 (SD = 29.42)	41.11 (SD = 42.71)	3.83 (513.24)	<0.001	0.17
	n = 231	n = 293
TST (hour)	7.08 (SD = 1.34)	7.16 (SD = 1.78)	0.55 (520.02)	0.586	0.02
	n = 231	n = 292

PSQI, Pittsburgh Sleep Quality Index; SD, standard deviation; SoL, sleep onset latency; TST, total sleep time.

Table 3.

Prevalence (%) of Poor Sleep Quality and Poor Sleep Efficiency Across Diagnostic Groups Within Each Age Group

Sleep variable	Group	Age group (years)
		15–19, % (n)	20–39, % (n)	40–59, % (n)	60+, % (n)
Sleep problem	Comparisona	61.5 (39)	41.4 (116)	44.7 (47)	47.1 (17)
	Autisticb	53.3 (60)	64.8 (122)	74.3 (74)	50.0 (22)
SE	Comparisona	44.2 (43)	36.7 (120)	37.5 (48)	50 (18)
	Autistica	43.8 (64)	46.4 (125)	51.3 (76)	41.7 (24)

PSQI >5.

SE <85%.

^a Proportions NS within group, across age groups.

^b χ² (3, N = 278) = 8.29, p = 0.040, ϕ = 0.172.

SE, sleep efficiency.

There was a significant age group by diagnostic group interaction for the global sleep quality score, F(3, 489) = 3.29, p = 0.020, partial η² = 0.020. The main effect for the diagnostic group was also significant, F(1, 483) = 7.85, p = 0.005, partial η² = 0.016, but the main effect for the age group was not, F(3, 483) = 1.62, p = 0.183, partial η² = 0.010. Analysis of simple effects indicated a significant difference in mean global score for the 20–39 years age groups, F(1, 236) = 15.09, p < 0.001, partial η² = 0.060 and the 40–59 years age groups, F(1, 119) = 14.36, p < 0.001, partial η² = 0.108. That is, these two autistic age groups had significantly poorer sleep quality scores than similar age comparison groups. There was also a significant age group effect within the autistic group, F(3, 274) = 3.05, p = 0.029, partial η² = 0.032. However, post hoc Tukey tests were not significant though the oldest and youngest age groups tended to have better sleep quality than middle-aged participants (both p < 0.10) (Table 4).

Table 4.

Sleep Variables by Diagnostic Group and Age Group (Mean and Standard Deviation)

Age group (years)	Comparison group			Autistic group
	PSQIa	SoL (minutes)	TST (hours)	PSQIa	SoL (minutes)	TST (hours)
15–19	7.15 ± 3.67	45.55 ± 46.58	7.24 ± 1.51	6.77 ± 3.98	45.05 ± 42.56	7.86 ± 1.48
	(n = 39)	(n = 44)	(n = 43)	(n = 60)	(n = 64)	(n = 65)
20–39	5.58 ± 3.37	25.76 ± 22.41	7.15 ± 1.36	7.42 ± 3.91	41.56 ± 42.62	7.30 ± 1.93
	(n = 116)	(n = 120)	(n = 121)	(n = 122)	(n = 126)	(n = 126)
40–59	5.85 ± 3.28	23.78 ± 22.64	6.74 ± 1.14	8.46 ± 3.90	42.58 ± 46.90	6.39 ± 1.56
	(n = 47)	(n = 48)	(n = 48)	(n = 74)	(n = 78)	(n = 77)
60+	5.59 ± 3.41	24.44 ± 20.23	7.09 ± 1.35	5.92 ± 3.41	24.22 ± 24.05	7.16 ± 1.60
	(n = 17)	(n = 18)	(n = 18)	(n = 22)	(n = 25)	(n = 24)

^a PSQI global score >5 indicates poor sleep quality.

Sleep efficiency

SE ≥85% was found for 60.9% of comparison adults (n = 230) and for 53.3% of autistic individuals (n = 289), χ² (1, N = 519) = 3.00, p = 0.083, p = 0.08. There was no significant difference between the two groups within each age group for SE, all p > 0.10, or any difference across age groups within the comparison or autistic groups (Table 3).

Sleep latency

Autistic individuals had significantly longer SoL than the comparison group (Table 2). As SoL was positively skewed, when examining SoL across group and age group, a two-way analysis of variance (ANOVA) with 1000 bootstrapped samples was conducted. There was no significant interaction, F(3, 515) = 1.91, p = 0.126, partial η² = 0.011. The main effects for diagnostic group, F(1, 515) = 4.54, p = 0.034, partial η² = 0.009, and age group, F(3, 515) = 4.034, p = 0.007, partial η² = 0.023, were significant. The Levene's test was significant (p < 0.001), indicating that the autistic group had more variable SoL than the comparison group.

There was a significant difference across age groups in the comparison group, F(3, 226) = 6.14, p < 0.001, partial η² = 0.075, but not in the autistic group, F(3, 289) = 1.53, p = 0.21, partial η² = 0.016. Levene's test was significant (p < 0.001) for the comparison group, but not the autism group (p = 0.128), indicating that there was more variance across age groups in the comparison group than in the autism group. Post Hoc comparisons using the Tukey HSD test indicated that the 15–19 years age group had longer sleep onset latencies (M = 45.54, SD = 46.58) compared with the 20–39 (M = 25.74, SD = 22.41; 95% BCa: 6.55–34.77), 40–59 (M = 23.78, SD = 22.64; 95% BCa: 7.44–38.65), and 60+ (M = 24.44, SD = 20.23; 95% BCa: 4.34–38.17) age groups. That is, adolescents in the comparison group had longer SoL than adults (Table 4).

Sleep duration

Both groups averaged about 7 hours of sleep (Table 2). Examining sleep duration, there was no significant interaction F(3, 514) = 1.97, p = 0.117, partial η² = 0.011 or main effect for group, F(1, 514) = 0.07, p = 0.792, partial η² = 0. There was a significant main effect for age group, F(3, 514) = 7.98, p < 0.001, partial η² = 0.045 with small effect size. Post hoc comparisons using the Tukey HSD test indicated that the mean score for the 40–59 years age group (M = 7.12, SD = 1.59) was significantly different from the 15–19 years age group (M = 7.61, SD = 1.67), and the 20–39 years age group, both p < 0.001. Individuals aged 40–59 slept less than individuals younger than 40 years (Table 4).

Correlations

Relationships between sleep variables (sleep quality, SoL, TST) and sex, AQ-short, mental health condition, medication, and employment were examined. Significant weak-to-moderate correlations were found between one or more sleep variables and AQ-short, mental health, medication, and employment for both groups. In the autistic group, poorer sleep quality also was associated with sex (Table 5). In the autistic group, sex was also associated with AQ-short, r (n = 294) = 0.179, p = 0.002; mental health condition, χ² (1) = 4.973, p = 0.026, ϕ = −0.146 and medication, χ² (1) = 5.062, p = 0.024, ϕ = −0.146. Autistic females were more likely to have poorer sleep quality, higher AQ-short scores, have a mental health condition, and be on medication.

Table 5.

Correlations Between Sleep Variables and AQ-Short, Sex, Demographic, Mental Health, and Medication

Variable	Sleep quality		SoL		Total night sleep
	Comparison group	Autistic group	Comparison group	Autistic group	Comparison group	Autistic group
AQ-short	0.260a	0.245a	—	—	−0.152b	−0.154c
	(n = 211)	(n = 269)			(n = 222)	(n = 281)
Sex	—	0.297a	—	0.156c(n = 282)	—	—
		(n = 268)
Employed	0.189c	—	0.147b	—	—	0.164c
	(n = 209)		(n = 219)			(n = 254)
Mental health condition	−0.408a	−0.310a	−0.209c	−0.195a	0.136b	—
	(n = 212)	(n = 272)	(n = 223)	(n = 287)	(n = 222)
Medication	−0.229c	−0.292a	—	−0.132b	—	—
	(n = 218)	(n = 276)		(n = 291)

Only significant correlations are reported.

^a p ≤ 0.001.

^b p < 0.05.

^c p ≤ 0.01.

Regression

The predictors of sleep quality (PSQI global score) for each group were examined by using hierarchical regression. AQ-short score was entered at Step 1; at Step 2, current mental health condition (yes, no); at Step 3, current medication (yes, no); at Step 4, employment (yes, no); and at Step 5, sex (male, female). Based on P-P plots and scatterplots, the distribution was approximately normal; multicollinearity was not a problem, and Cooks distance and standardized residuals showed no outliers; Mahalanobis distance identified one outlier in the autistic group; and omitting this participant did not alter the total amount of variance explained, and they were retained. Three people (one comparison, two autistic) had global sleep quality scores just exceeding three SDs from the mean; they were retained, as sleep quality was expected to range from excellent to very poor.

Comparison group

The final model was significant, R² = 0.254, F(5, 203) = 13.85, p < 0.001, accounting for 25% of variance in sleep quality. At Step 1, autistic traits accounted for 6.8% of variance, F(1, 207) = 14.99, p < 0.001. The addition of current mental health condition added 13.3% of variance, F(1, 206) = 34.23, p < 0.001. Current medication added a further 2.8% of variance, F(1, 205) = 7.33, p = 0.007, and employment at Step 4 contributed a further 2.1% of variance, F(1, 204) = 5.67, p = 0.018. Finally, the addition of sex accounted for 0.5% of variance, F(1, 203) = 1.50, p = 0.223. After the final step, the strongest, unique predictor of sleep quality was having a mental health condition (β = −0.331, p < 0.001). Increased autistic traits (β = 0.191, p = 0.003), medication (β = −0.159, p = 0.012), and being unemployed (β = 0.153, p = 0.014) also contributed significantly to the final model.

Autistic group

The final model was significant, R² = 0.205, F(5, 238) = 12.31, p < 0.001, accounting for 21% of variance in sleep quality. At step 1, autistic traits accounted for 6.0% of variance, F(1, 242) = 15.43, p < 0.001. The addition of current mental health problem added 7.7% of variance, F(1, 241) = 21.56, p < 0.001, and at step 3 current medication added a further 1.9% of variance, F(1, 240) = 5.51, p = 0.020. The addition of employment contributed 0.4% of variance, F(1, 239) = 1.18, p = 0.279, and at step 5 sex accounted for 4.5% of variance, F(1, 238) = 13.42, p < 0.001. After the final step, the strongest, unique predictor of sleep quality was being female (β = 0.218, p < 0.001). Having a mental health condition (β = −0.191, p = 0.003), increased autistic traits (β = 0.145, p = 0.018), and medication (β = −0.146, p = 0.027) also contributed significantly to the final model.

Discussion

We compared self-reported sleep quality, SoL, TST, and SE from mid-adolescence to old age in autistic people and a similar age comparison group, and also examined the influence of sex. As hypothesized, and consistent with the poor sleep quality or insomnia previously reported for autistic adolescent and adult samples,^8,9,14,16 the autistic group were more likely to have poor sleep quality than non-autistic individuals in this investigation.

Poor sleep quality occurred for 50% or more of autistic participants across age groups, with autistic adults aged 20 to 59 years being most vulnerable to poor sleep quality. Poor sleep quality tended to be lower in the comparison group, exceeding 40% of participants across age groups. The prevalence of poor sleep quality in autistic participants aged 20–59 years (i.e., early and middle adulthood) exceeded similar age comparison participants. Similarly, PSQI global scores were higher in the autistic group than the comparison group in early and middle adulthood, but within the autistic group, adolescents and older adults tended to have better sleep quality than individuals in early and middle adulthood. Our findings are largely consistent with Hohn and colleagues,⁸ who reported that 54% of autistic adults aged 18–65 years had at least sub-threshold insomnia, but that age did not predict insomnia. However, insomnia and poor sleep quality are not strictly equivalent.¹⁸ The percentage of autistic adolescents with poor sleep quality was within the range previously reported for autistic children and adolescents,^7,15,22 but poor sleep quality did not differ between our two adolescent groups. This is inconsistent with previous reports showing that autistic adolescents (13–17 years) were three times more likely to report having a sleep problem than non-autistic adolescents²² and that autistic adolescents and young adults (11–26 years) had poorer sleep on a sleep questionnaire.¹⁴ Differences may be due to age and sample size; our adolescents were aged 15 to 19 years, our sample size was approximately double those in earlier studies, and we used a global sleep quality score (PSQI) rather than a single question as in one study.²²

SoL averaged a significant 12 minutes longer in the autistic group and exceeded the recommended 30 minutes indicative of good sleep quality.¹⁸ SoL was also more variable in autistic individuals than in comparison individuals, which has been reported previously in adults.⁹ Within the autistic group, average SoL exceeded 40 minutes from adolescence to late middle age (i.e., 15–59 years), which is consistent with previous reports of increased SoL in autistic individuals across the lifespan.^7,9,15,22 Although adolescents in both groups had a similarly long average SoL, through adulthood the average SoL for comparison age groups was <30 minutes. Increased SoL from childhood⁷ through early adulthood^14–16,22 has been reported previously for autistic individuals. Using actigraphy, two studies also found that SoL in younger autistic adults was longer than a matched comparison group, but average SoL in both groups was <30 minutes.^9,17 However, increased SoL through to late middle adulthood in autistic individuals has not been reported previously. It appears that difficulties with SoL may be a lifetime issue for those on the autism spectrum.

Individuals in both groups averaged about 7 hours sleep per night, except those aged 40–59 years, who slept less. Methodology (self-report questionnaires or sleep diaries, or objective measurement using actigraphy or PSG) can result in differences in reported TST. Co-occurring psychopathology may also be important; compared with a matched comparison group, autistic adults with no co-occurring psychopathology had reduced TST, but those autistic adults medicated for a psychiatric condition did not.⁹ Regardless, both groups were at or below the lower end of recommended sleep requirements for age.²⁰

Most participants in both groups had good SE (SE ≥85%) This contrasts with previous results that identified poorer SE among autistic adolescents or adults^9,11,14,22 compared with controls. Here, we used self-report with a dichotomous classification; whereas in three of the former studies, SE was examined by using actigraphy and both autistic and non-autistic groups had mean SE ≤85%.^9,11,14 Thus, it is likely that poor SE is common in both autistic and non-autistic populations, but those on the spectrum may have poorer measured SE, on average, than other population groups.

Autistic females were more likely to have poor sleep quality, increased autistic traits, or a mental health condition and be medicated than were autistic males. No similar sex relationships occurred in the comparison group, though the low number of males may have influenced the results. There are no studies examining sex differences in autistic adults, and reports of sex differences in sleep in two child studies are contradictory.^29,30 The current results highlight that autistic females are more vulnerable to developing sleep problems than autistic males, possibly because they are more vulnerable to anxiety and depression as also has been reported recently.⁴⁹ Poor sleep and mental health problems are related in the general population^40,47 and in autistic samples.^7,16,37

Predictors of sleep quality

Autism severity, mental health condition, medication, employment, and sex together accounted for more than 20% of variance in sleep quality in both groups. Autism severity, mental health, and medication contributed unique variance to sleep quality for both autistic and comparison groups. This is consistent with research showing associations between poor sleep and autism severity,²⁷ relationships between psychopathology and autistic traits,⁵⁰ and that poor sleep quality and poor mental health are bi-directionally related.^40,51

Poor mental health and/or pre-sleep arousal may cause sleep onset difficulties and poor sleep maintenance. Two-thirds of autistic participants reported a mental health condition, which is a little below the reported 79% lifetime prevalence of any mental health disorder in autism.⁵² Further, pre-sleep arousal has been related to poor sleep in autistic adolescents³⁷ and adults.³⁸ Rumination associated with poor mental health increases arousal and may contribute to SoL difficulties.¹⁶ Supporting this, autistic adolescents have reported that they “have difficulty switching thoughts off,” affecting sleep onset.⁵³ Disruption to bedtime rituals influences autistic children's sleep⁷ and sensory hyper-reactivity predicted insomnia in autistic adults,⁸ suggesting that the sleep environment also is important.

Employment status only contributed unique variance to the comparison group's sleep quality. This contrasts with a report that autistic adults with a sleep disorder were more likely to be unemployed.³⁹ However, among autistic individuals in this study, unemployment was related to shorter TST rather than sleep quality. Therefore, sleep difficulties may still influence employment among autistic adults.

Importantly, the current results highlight that autistic females may be more vulnerable to developing sleep and mental health problems than autistic males. Sex was the strongest predictor of sleep quality for the autistic group but contributed no unique variance for the comparison group. As autistic females were more likely to have a mental health condition and higher autism severity than autistic males, one might have expected any sex influence to be accounted for by these variables. However, we did not measure the severity of mental health symptoms, and autistic women may have more severe symptomatology. Autistic women have previously been reported to have more severe symptoms of anxiety and depression across the lifespan.⁴⁹ As there were more autistic females than males 25 years and older, menopause also may be a factor as menopausal women are at higher risk for disturbed sleep, and middle age women in general are also at higher risk of poor sleep quality.³⁶ Nothing is known about the experience of menopause in autistic women. However, majority of the comparison group were women, which does not support a menopause explanation. Further exploration of male/female differences in autism is imperative, as our results suggest that what is known about autistic males cannot be assumed for autistic females.

Clinical implications

Given the high lifetime prevalence of anxiety and depression associated with an autism diagnosis,⁵⁴ treating insomnia should be a priority, yet treatment research is scant. Education within the autism community about poor sleep and its potential impacts on health and well-being should be a primary focus. Second, awareness is needed about increased risk for poor sleep quality in autistic women. Third, targeted, effective remedies for sleep problems for autistic adolescents and adults are needed. To achieve this, greater attention must be given to understanding the causes of poor sleep in autism spectrum disorder, sleep's transdiagnostic nature,⁵⁵ and the usefulness of transdiagnostic interventions.^56,57 For example, in non-autistic populations, transdiagnostic interventions can be as clinically effective as diagnosis-specific treatments in treating anxiety disorders,⁵⁸ whereas insomnia treatment results in improvement in mental health conditions,⁵⁵ including depression.⁴¹ Further, insomnia may precede the onset of poor mental health.⁴⁷ At least two studies have reported that treatment of mental health conditions in autistic individuals also improved sleep.^59,60

Strengths and limitations

This study had several strengths. First, it addressed a significant gap in the autism sleep literature by examining sleep quality from late adolescence to old age (15–80 years). No previous study has examined the range of sleep variables reported in this study from adolescence to old age. Second, our study had a large sample size, and it included an age-matched comparison group. Third, this is the first study to examine male–female differences in sleep in autistic adolescents and adults. Finally, as SASLA and ALSAA are not designed as sleep studies, results are unlikely to be biased by individuals with poor sleep preferentially choosing to participate.

Nevertheless, limitations are present. First, sleep was measured by using a self-report, retrospective questionnaire, which is not as reliable as measuring sleep prospectively, using sleep diaries or actigraphy. Actigraphy also is objective, reducing the chance of bias, whereas recollection may be susceptible to under- or over-estimation.⁶¹ It is also possible that autistic adults and non-autistic adults differ in how they self-report their sleep. However, studies that have used both self-report and actigraphy have found group differences in sleep on both measures.^9,14 Second, the study was cross-sectional; thus, long-term trends for sleep quality in the context of chronicity or aging are not captured. Third, autistic adults with intellectual disability were not included. Fourth, about half the autistic sample was female, but the generally accepted male:female ratio in autism spectrum disorder is 4:1. The comparison group was also biased toward females. Finally, the sample consisted of volunteers who responded to an advertisement, and not all participants had responses for all variables. Thus, we may not have accurately captured sleep in men and our results may not be generalized to the autistic or general communities overall.

Conclusions

There are several key findings. First, poor sleep quality is highly prevalent from late adolescence to old age in autistic individuals, with those in early and middle adulthood being most vulnerable. Second, poor global sleep quality and increased SoL (>30 minutes) are the most salient sleep issues. Third, autistic women are at higher risk for poor sleep quality than autistic men. Finally, autism trait severity and mental health conditions are significant contributors to poor sleep quality. Future research should focus on understanding the cause of poor sleep quality in those with a diagnosis of autism spectrum disorder to prevent the persistence of poor sleep across the lifespan; understanding why autistic women are more at risk for poor sleep quality is a high priority; and intervention research is sorely needed.

Authors' Contributions

S.J. contributed to recruitment, produced the clean data file from the original SASLA and ALSAA data sets, produced a draft of the article in the form of an Honors thesis, and provided feedback on later drafts of the article. A.L.R. conducted the analyses presented in the article, redrafted the article, and edited in response to feedback. L.P.L. assisted with the original draft and provided feedback on later drafts. M.U., S.R.C.A., and J.N.T. provided feedback on later drafts of the article. A.L.R., J.N.T., M.U., S.R.C.A., and L.P.L. were members of the team designing and/or recruiting for the SASLA and ALSAA projects from which the data were drawn. All team members read and approved the final submission.

Disclaimer

This paper is submitted solely to Autism in Adulthood; it has not been submitted for publication elsewhere.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

The authors acknowledge the financial support of the Cooperative Research Centre for Living with Autism (Autism CRC), established and supported under the Australian Government's Cooperative Research Centres Program.