At T1, the study set out to explore (i) the extent of autistic traits and co-occurring physical and mental health difficulties in autistic adults; (ii) age-related differences in these areas through comparisons between younger and older autistic adults; and (iii) how these factors are associated with daily living and QoL.

At T2, we followed-up participants from the T1 study. In addition to replicating the aims of Study 1, the principal aim at T2 was (i) to evaluate the profile of potential age-related changes over time in younger and older autistic adults, related to autistic traits, co-occurring mental health conditions, and QoL. Regression analyses explored (ii) how these factors were associated with QoL over time.

The study was advertised via the United Kingdom National Autistic Society website and online via the Twitter network of the first author. We aimed to recruit older adults and autistic women as these groups are greatly underrepresented in most autism research. Participants from the research databases at the Autism Research Group and online research recruitment portal at City, University of London were also invited to take part in the study. Informed consent was obtained from all participants prior to enrolment in the study and they were informed of their right to withdraw from the study at any time, without being disadvantaged. Participants were offered £25 as a gratuity for taking part in the research at each time point and received full reimbursement of travel expenses.

The relative lack of studies on ageing in autism means there is no consensus on the definition of ‘older’ age in this group. For the purpose of the present study, we included autistic adults across a wide age range, with ‘older’ adults being defined as those aged ≥50 years and ‘younger’ adults as those aged 18–49 years. Participants were assigned to ‘younger’ and ‘older’ age groups at T1 and remained assigned to those same groups at T2 follow-up.

To ensure that participants were able fully to understand all the task requirements English language proficiency was assessed using the Comprehensive Receptive and Expressive Vocabulary Test–Third Edition (CREVT-3; Wallace and Hammill, 2013). Two participants were excluded at this stage as they did not meet the standardised assessment criteria for English fluency (CREVT-3 overall language score > 70; population mean 100, SD 15; see Appendix 1 in Supplementary material).

Participants at T1 comprised 68 autistic adults aged 19–80 years (mean 44.1 years, SD 15.5 years), including 37 younger (mean 31.9 years, 10 female) and 31 older adults (mean 58.6 years; seven female). All participants had a formal diagnosis of autism, confirmed by a copy of clinical diagnostic reports obtained at enrolment. Age groups were matched on gender ratio (reported as male and female in this study, no participants identified as transgender or non-binary), years of formal education and general intellectual ability (IQ; measured by Wechsler Adult Intelligence Scales–Fourth Edition; WAIS-IV; Wechsler, 2008; see Table 1 for data, and see Appendix 1 in Supplementary material). Following the T1 study, participants were asked if they would be willing to take part in a subsequent follow-up study (T2).

A total of 49 individuals (72.1% of the T1 sample) agreed to take part at T2. Their ages ranged from 24 to 74 years (Mean 48.4 years), including 25 younger (mean 36.2 years; five female) and 24 older adults (mean 60.9 years; five female; for details see Table 2).

Reasons for non-participation at T2 (n = 19, 28% of the T1 sample) were: chronic or terminal illness (n = 7), or death (n = 2); lost to follow-up, moved to different town, city or country (n = 2); work commitments, personal difficulties, or family commitments (n = 5); administrative reasons, or self-exclusion or withdrawal from the project for other reasons (n = 3).

Participant well-being was central to all stages of the research. During the study scoping and design, advice was sought from autistic peer researchers in co-creation discussion forums. Participants were consulted throughout the data collection about any adaptations they might require to the study procedures and materials, and every effort was made to incorporate adaptations to meet their individual needs. These adaptations included easy read formats for information about study aims and task instructions, changes to sensory stimuli in the research laboratory, e.g., noise reduction and soft lighting, and frequent breaks between tasks as needed. Every effort was made to ensure that these adaptations did not compromise the methodology or quality of the data.

The Wechsler Adult Intelligence Scales–Fourth Edition (WAIS-IV) was administered at T1 and T2 follow-up. The WAIS-IV is a widely used standardised measure to assess intellectual ability profiles in adults aged 16–90 years (Wechsler, 2008). It comprises 10 core and five supplemental subtests, providing Index scores for Verbal Comprehension (VCI), Perceptual Reasoning (PRI), Working Memory (WMI) and Processing Speed (PSI), and a composite Full-Scale IQ (FSIQ) score.

At T1, clinical reports confirmed autism diagnosis (see the section Sample Characteristics). Since those reports incorporated a variety of measures, the Autism Diagnostic Observation Schedule–2nd Edition (ADOS-2; Lord et al., 2012) and the Social Responsiveness Scale–2nd Edition (SRS-2; Constantino and Gruber, 2012) were administered to confirm diagnostic reports. The Module 4 (adult) ADOS assessment is reported to have sensitivity of 0.61 (Bastiaansen et al., 2011; de Bildt et al., 2016); specificity is between 0.50 and 0.84 (Maddox et al., 2017). The first author was trained in ADOS-2 administration to 0.89 reliability and overseen by a certified ADOS trainer. The ADOS-2 was administered to 50 participants (74%) who consented to complete this assessment. Over a third (37.2%) of assessments in this study were double-coded for inter-rater reliability which was maintained at 0.84 or above. We note that subsequent development of calibrated severity scores (CSS; Hus and Lord, 2014) are now more commonly used as more sensitive measures of autism symptom severity in adults. However, because the CSS algorithm was not available when the present study data were collected, the ADOS data reported here are according to the algorithm in Lord et al. (2012).

The SRS-2 is a self-rated measure of autism-related traits and difficulties based on the DSM-5 (American Psychiatric Association, 2013) criteria for autism. It provides a Total score and separate domain index scores (T-scores; range 30 to >90, mean 50, SD 10) for Social Communication, Social Motivation, Restricted Interests and Repetitive Behaviours (RRBs), Social Awareness, and Social Cognitive functioning. Studies with autistic adults have reported sensitivity and specificity of 0.85 and 0.83, respectively (Bölte, 2012). The SRS-2 was administered at T1 and T2 explore possible age-related changes over time.

At the time this study began, very few investigations had evaluated QoL in autism and even fewer had explored QoL in older autistic adults (e.g., Geurts and Vissers, 2012). Accordingly, two QoL measures were used (see below); these are designed to capture similar domains but use different methods of calculating outcome scores. Thus, a conversion formula (see International Well-being Group, 2013) was applied to the World Health Organisation Quality of Life–Short Form (WHOQOL-BREF) scores for Overall-QoL, Health-QoL, and degree of Support received, for comparable reporting in relation to the PWI-A scores. Both the WHOQOL-BREF and the PWI-A were administered at T1 and T2 to evaluate any (positive or negative) change in quality of life and well-being.

Table 1 describes the sample characteristics at T1. Given the study design, younger (n = 37) and older (n = 31) groups differed significantly on chronological age. Groups did not differ on years of formal education, despite five older adults holding fewer. There were also no group differences in gender [χ²(1) = 0.005, p = 0.94], or any IQ scale scores (see Table 1).

Table 4 summarises autistic profile scores for younger and older adults. There were no age group differences in ADOS-2 or SRS-2 scores (Table 4). As previously mentioned, all participants had existing clinical diagnoses of an autism spectrum condition. Although 12 participants (younger, n = 7; older, n = 5) did not meet the cut-off for ADOS-2 Total scores for ‘Autism Spectrum’ (≥7), they did meet or exceeded the cut-off for both index scores (Communication ≥2; Social Interaction ≥4). These findings are consistent with variable sensitivity and specificity of the ADOS for adults who also have co-occurring mental health conditions (Maddox et al., 2017).

Figure 2 summarises the PIAS self-reported data related to everyday difficulties with social skills, mental and physical health indicated high rate of co-occurring conditions in younger and older adults. Overall, participants reported high rates of symptoms related to anxiety (n = 50 [74%]) and depression (n = 47 [69%]), and related difficulties in identifying and describing emotions (alexithymia, n = 28 [41%]; for further information see Appendix 3 in Supplementary material; Figure 1). Sleep disturbances (e.g., difficulty falling asleep; frequent waking) were common in more than half to two-thirds of adults (n = 48 [71%]) as were sensory hypersensitivities (n = 58 [85%]). Conditions related to sensory sensitivities (e.g., visual, auditory, touch, taste/texture, and olfactory, n = 27 [40%]) were reported more by younger than older adults, whilst only five adults (younger, n = 1; older, n = 4) reported hyposensitivity or sensory-seeking. Social difficulties and stresses were related to social conversation (n = 26 [39%]), social anxiety (n = 28 [42%]), and loneliness or social isolation (n = 26 [39%]). Additionally, somatic conditions that presented greatest difficulties were related to neurophysiological symptoms (heart racing, n = 10 [14%]; shortness of breath, n = 14 [21%]); digestive problems (bowel, n = 14 [21%]; indigestion, n = 28 [41%]); and pain (stomach, n = 19 [28%]; back, n = 10 [14%]; and joints or limbs, n = 28 [41%]).

Figure 3 summarises the percentage of adults who met the threshold for at least one co-occurring condition (measured by PHQ) and experienced everyday difficulties related to those conditions. Overall, at T1 59% of the younger (n = 22) and 30% of older adults (n = 9) met the criteria for at least one other condition. The number of co-occurring mental health conditions ranged from 0 to 4, with almost half (46%) of all autistic adults having multiple co-occurring conditions. Although there were no significant differences between younger and older adults on any mental health measures (Table 5), 37% of younger (n = 14) and 22% of older adults (n = 7) met the criteria for three or more co-occurring mental health conditions. Both groups reported being on multiple pharmacological treatments for those conditions, which aligns with the self-report background data collected (using the PIAS; see Appendix 3 in Supplementary material).

The most common conditions reported by younger adults were Anxiety (27%; of which 16.2% other Anxiety syndromes; 10.8% Panic syndrome), Major Depressive syndrome (21.6%), Eating disorders (21.6%; of which 18.9% Binge Eating; 2.7% Bulimia Nervosa), Somatic disorders (16.2%, e.g., bodily pain), and Alcohol abuse (8.1%). For older adults, the most common conditions were Depression (15.6%; of which 12.5% Major Depressive syndrome; 3.1% other Depression syndrome), Binge Eating disorder (12.5%), Anxiety (9.4%; of which 6.3% other Anxiety syndromes; 3.1% Panic syndrome); Alcohol abuse (6.3%), and Somatic disorders (3.1%). Moreover, both groups reported difficulties in everyday functioning (e.g., doing housework, employment, and social relationships), as “somewhat” to “very difficult,” related to these conditions.

Table 5 summarises the statistical analyses for T1 comparisons between younger and older adults, on the standardised assessments of mental health and quality of life. Although the standardised measures captured a lower rate of mental health concerns than those self-reported in background descriptive information (using the PIAS), these were nevertheless still predominant for the majority of adults, indicating at least “moderate anxiety” symptoms (as measured by BAI-II), and “mild mood disorder” to “borderline clinical depression” symptoms (as measured by BDI-II), which were also observed on the PHQ (reported in Table 5).

As set out in Table 5, scores across SWB and QoL domains were, overall, poor for both younger and older adults, indicated by below-average scores (<50; scale 0–100) on the PWI and WHOQOL-BREF scales, respectively (see Roestorf and Bowler, 2016; Roestorf, 2018 for reports of significant lower QoL for autistic adults compared to non-autistic groups). Moreover, both groups reported low degree of Support received for their everyday needs. The most common indicators of low SWB were related to lack of Personal Relationships and feeling isolated from the Community; lack of Achievement; and concerns about Health and Future. Scores for these factors were also below normative population mean scores of 70–80 points (see Cummins et al., 2003; International Well-being Group, 2013). Standard of Living and feeling safe (Safety) were amongst the highest SWB indicators.

Table 6 sets out the main predictors of QoL domains across all participants. Age was not consistently related to QoL [all r(44) < 0.17, all p > 0.25] as low QoL scores, across domains, were observed across the lifespan. The only exceptions to this were Overall QoL and Social QoL domains, for which older adults reported greater satisfaction. Overall, depression and anxiety symptoms (as measured by BDI-II and PHQ, and the BAI-II, respectively) were the strongest consistent predictors of Global Life Satisfaction and Subjective Well-being, and most QoL domains including Overall-QoL, Health-QoL, Physical-QoL, Psychological-QoL, and Environmental-QoL. However, these symptoms did not predict Social-QoL scores [R² = 0.07; F(2,46) = 1.83, p > 0.05].

Difficulties related to autistic traits (as measured by SRS-2 Total scores), predicted Subjective Well-Being, and to a much lesser extent Overall QoL, Global Life Satisfaction, Psychological-QoL, and Environmental-QoL.

By contrast, factors related to age, gender, and autistic traits (as measured by ADOS-2 and SRS-2 Communication and RRB scores) did not predict any QoL or SWB outcomes [all F(8,30) < 1.60, all p > 0.05].

Table 2 summarises the characteristics T2 participants (25 younger and 24 older adults). The mean interval between T1 and T2 assessments was 2.4 years (range 1.2–3.8 years) and was not significantly different between age groups [t(25) < 1.00, p > 0.05]. Comparisons of the profiles of participants who continued to the T2 follow-up with those who did not, revealed significantly greater T1 depression symptoms in the non-continuing group [t(36) = 2.33, p = 0.03, d = 0.78], but no other differences in T1-derived cognitive or health profiles.

Autistic traits profiles (as measured by the SRS-2; Constantino and Gruber, 2012) were re-assessed in younger and older participants. As at T1, there were no age-related group differences at T2.

As indicated in Table 2, there were no significant Age Group differences in anxiety or depression symptoms (measured by the BAI-II and BDI-II, respectively). Nor were there any differences in PHQ-measured symptoms of somatic complaints (e.g., bodily pain), or eating disorders, or alcohol abuse. Once again, daily difficulties related to co-occurring conditions were rated by both groups as “somewhat” to “very difficult.”

Table 2 summarises the T2 QoL scores. As at T1, QoL at T2 was low across domains. The T2 data followed a similar pattern to T1 observations, with older adults once again reporting greater satisfaction in Subjective Well-being, Overall-QoL, and Social-QoL. However, in contrast to T1 data, older adults also reported greater Psychological-QoL than younger adults.

Table 3 indicates the change in T1-T2 scores that were assessed by comparing individual scores for autistic traits, mental health (anxiety, depression), and QoL, following the procedure set out earlier (see section “Statistical analysis”). There were no significant differences between T1 and T2 scores on any mental health or QoL domain scores [all t(26) < 1.96, all p > 0.05]. Regarding autistic traits, only RRBs showed significant change presented as increased at T2 [t(26) = 2.22, p < 0.04]. However, the analysis was not significant after applying Bonferroni corrections (p < 0.001) for multiple analysis (p = 0.16, η_p² = 0.13).

Table 7 summarises the main predictors of T2 QoL for all autistic adults. Once again, Age did not predict any QoL outcomes, across domains [all r(29) ≤ 0.17, all p ≥ 0.18], which followed the pattern observed at T1. However, Age at T2 was significantly correlated with Subjective well-being [r(29) = 0.40, p = 0.015] and Global Life Satisfaction [all r(29) = 0.34, p = 0.036], which was explained by higher satisfaction in these domains reported by older autistic adults (see Table 2).

Autistic traits were significantly negatively correlated with all QoL domains, except for Health-QoL, (all r(29) ≥ 0.32, all p < 0.05). Given the marginal increase in autistic traits observed at T2, these were also explored in relation to QoL outcomes. RRBs, but not social communication or total scores (measured by SRS-2), were a significant predictor of Subjective Well-being (R² = 0.14, F(1,27) = 4.43, p < 0.05), Physical-QoL (R² = 0.2, F(1,27) = 4.35, p = 0.05), and Environmental-QoL (R² = 0.14, F(1,28) = 4.62, p < 0.05).

Difficulties related to anxiety, depression, and eating disorders were consistent predictors of Health-QoL [R² = 0.60, F(3,22) = 11.16, p < 0.001], Physical-QoL [R² = 0.65, F(3,22) = 13.34, p < 0.001; Table 7]. Whilst depression alone predicted Overall-QoL [R² = 0.51, F(2,23) = 12.16, p < 0.001], and anxiety and difficulty in everyday living predicted Psychological-QoL [R² = 0.50, F(2,23) = 11.42, p < 0.001]. Social-QoL, however was predicted by multiple factors including anxiety, depression, and autistic traits difficulties related to social communication, and difficulty with everyday living [R² = 0.71, F(6,19) = 7.80, p = 0.005]. The significant predictors of respective QoL outcomes are presented in Table 7.