Professional Documents
Culture Documents
Brugha, T. S., Spiers, N., Bankart, J., Cooper, S.-A., McManus, S., Scott, F. J., Tyrer, F. (2016) - Epidemiology of Autism in Adults Across
Brugha, T. S., Spiers, N., Bankart, J., Cooper, S.-A., McManus, S., Scott, F. J., Tyrer, F. (2016) - Epidemiology of Autism in Adults Across
Background
The epidemiology of autism in adults has relied on untested those with moderate to profound intellectual disability (odds
projections using childhood research. ratio (OR) = 63.5, 95% CI 27.4–147.2). Male gender was a
strong predictor of autism only in those with no or mild
Aims intellectual disability (adjusted OR = 8.5, 95% CI 2.0–34.9;
To derive representative estimates of the prevalence of interaction with gender, P = 0.03).
autism and key associations in adults of all ages and ability
levels. Conclusions
Method Few adults with autism have intellectual disability; however,
Comparable clinical diagnostic assessments of 7274 Adult autism is more prevalent in this population. Autism measures
Psychiatric Morbidity Survey participants combined with a may miss more women with autism.
population case-register survey of 290 adults with intellectual
disability. Declaration of interest
None.
Results
The combined prevalence of autism in adults of all ages in Copyright and usage
England was 11/1000 (95% CI 3–19/1000). It was higher in B The Royal College of Psychiatrists 2016.
Globally, in 2010, there were an estimated 52 million people with capacity to consent or to take part in a standard survey interview,
autism that accounted for more than 111 disability adjusted or who were living in care settings such as institutions or care
life years (DALYs) per 100 000 population.1 Until recently, homes for people with intellectual disabilities, adults with
information on the epidemiology of autism was based on moderate to profound intellectual disability were unrepresented.
childhood studies.2,3 A complete understanding of the nature, Given the strong association between intellectual disability and a
causes and public health impact of autism should consider the childhood13 and adulthood9 diagnosis of autism, knowledge of
interplay of genetic, epigenetic and environmental associations the overall prevalence and age11 and gender2,3 profile of autism
throughout the life course.4 There is a widespread but largely in adults requires adults of all ability levels to be examined using
uninformed assumption5,6 that as many as a half of all adults with comparable methods. This paper reports on the epidemiology of
autism have intellectual disability, which, if untrue, could distort autism drawing on samples combined to reflect the full range of
planning a balanced range of services for the whole population ability levels in the adult general population. The sample from
with autism. As childhood diagnoses of autism (or of Asperger the first general population study11 was extended with the
syndrome) have increased,7,8 parental fears remain undiminished inclusion of representative samples of adults with intellectual
concerning the future care of their offspring with little prospect of disability omitted from the earlier survey. The aims were to
funded services when they can no longer provide support. provide an estimate of the overall prevalence of autism and
Two UK studies limited to adults with intellectual disability9,10 to examine key associations in adults at all intellectual ability
suggest autism rates between 70/1000 and 210/1000 but lacked levels.
a validated systematic diagnostic assessment. Recently the
prevalence of autism was reported as 9.8/1000 in the Adult
Psychiatric Morbidity Survey (APMS), a nationally representative Method
sample of adults capable of giving informed consent and of taking
part in a survey interview, living in private households.11 That Data from a multiphase survey of adults in private households
study11 found autism was associated with reduced verbal IQ, throughout England (APMS;11 fieldwork 2007) and a single-phase
low educational achievement, male gender and epilepsy.12 survey of a representative group of adults with intellectual disability
However, by excluding people without the decision-making drawn from intellectual disability case registers in three areas of
England (the Intellectual Disability Case Register study (IDCR);14
fieldwork 2010) were combined. The APMS employed a stratified
*Findings from this project were presented to: the Annual Meeting of the two-phase design based on a random probability sample of one
Royal College of Psychiatrists, Birmingham, July 2015; IMFAR Conference,
adult per private household,15 throughout England (as already
at Donostia, San Sebastian, Spain, 2013. Summary findings and methods of
described11) followed by diagnostic assessments of respondents
the IDCR study have appeared in the HSCIC Government Report: Brugha T,
Cooper SA, McManus S, Purdon S, Scott FJ, Spiers NA, et al. Estimating the
at an increased risk of autism.16
Prevalence of Autism Spectrum Conditions in Adults: Extending the 2007 For the IDCR, adults not considered in the APMS, by design,
Adult Psychiatric Morbidity Survey. Leeds: The NHS Information Centre, 2012, and living in communal care establishments and private house-
which is cited in this article. Tables in the present article do not duplicate holds, were randomly sampled from three adult intellectual
tables in that report. disability registers in England, in Leicestershire, Lambeth and
498
Epidemiology of autism in adults across ability levels
Sheffield, stratified by age, gender and type of residence (detailed (T.S.B.) and a qualified ADOS trainer (F.J.S.). Training experience
in online supplement DS1). For the adults living in private was gained through assessing adults living in settings in which
households, those judged sufficiently able to have taken part in fieldwork subsequently took place. Field interviews did not
the APMS were then excluded. All adults living in communal care commence until the interviewers achieved at least 90% agreement
establishments were included, as these establishments were on ratings of jointly observed ADOS examinations. During
excluded from the APMS, yet a lot of people with intellectual fieldwork, interviewers received supervision sessions and prepared
disabilities live in such establishments. The sample size in APMS case vignette reports. They took part in post fieldwork debriefing
phase two was chosen to reflect the sample sizes and precision to add further contextual information.
of psychosis prevalence estimates required to monitor trends in Intellectual disability was defined as a significant intellectual
each APMS survey since the first APMS in 1993.17 The IDCR impairment with onset before adulthood and deficits in skills
sample was designed to achieve similar precision. needed for daily functioning23–25 assessed in the IDCR by the
APMS participants gave informed consent directly to APMS carer-report version of the Vineland II Adaptive Behaviour
phase-one interviewers. In the IDCR, following the English Mental Scales.26 In the APMS, predicted Verbal IQ (V-IQ; range estimate
Capacity Act 2005, consent was taken wherever possible with 70–130) was derived using the National Adult Reading Test
input from consultees as appropriate. In keeping with the require- (NART).27 The NART requires a high reading age, leaving gaps
ments of the ethics committees, participants in Leicestershire were in its completion for adults with literacy problems of a wide range
telephoned by the research team (‘opt-out consent procedure’); of causes, including mild intellectual disability, dementia, dyslexia,
those in Lambeth and Sheffield contacted the research team only lack of education. Given this limitation (online supplement
if they wished to take part in the study (‘opt-in consent DS2), we were unable to identify those in the APMS with mild
procedure’). Ethical approval for the APMS was obtained from intellectual disability, so they all were included in a category of
the Royal Free Medical School Research Ethics Committee, none to mild intellectual disability. This assumption is reasonable,
London, UK. Ethical approval for the IDCR was obtained in as ability to participate in the APMS would be extremely unlikely
Leicestershire from the Derbyshire Ethics Committee and for at an ability level of moderate intellectual disabilities or lower. In
Sheffield and Lambeth from the ‘Essex 2’ Research Ethics both surveys questionnaires were completed covering participant’s
Committee, UK. physical and mental health, socioeconomic factors and use of
The 20-item self-completion Autism-spectrum Quotient services, using comparable measures.15
(AQ)15,18 was used in phase one of the APMS to select
participants for a second-phase evaluation using detailed clinical Statistical analysis
assessments based on Module 4 of the Autism Diagnostic
Observation Schedule (ADOS-Mod4).16 In the IDCR study, Online supplement DS1 describes how the APMS and IDCR
most participants were assessed at first interview with Module 1 samples were combined for analysis, which is illustrated in online
of ADOS (ADOS-Mod1),19 which is designed for individuals Fig. DS1. The svytabulate procedure (Stata 12.0 for Windows) was
who do not consistently use phrase speech. The ADOS-Mod4 used to estimate prevalence of autism by intellectual disability, age
was used for verbally fluent adults living in communal care and gender; svylogistic was used to fit logistic regression models
establishments. for autism by age and gender, taking the complex survey design
Threshold scores of 12+ on the ADOS-Mod1 and 10+ on into account and adjusting for the presence of epilepsy; confidence
the ADOS-Mod4 were used to define an individual as having intervals were calculated using Taylor linearisation.28 To examine
autism. Both the ADOS-Mod1 and ADOS-Mod4 were subject whether predictors of autism are the same in those with and
to validation and calibration work (online supplement DS2) without moderate intellectual disability, models were fitted for
within the study general population samples11,14 based on univariable predictors with an interaction term, allowing odds
developmental assessments using the Autism Diagnostic Interview ratios to vary by disability level. The significance of interaction terms
– Revised (ADI-R)20 and the Diagnostic Interview for Social and was tested using an adjusted Wald test29 and where significant
Communication Disorders (DISCO)21 and, in the APMS, a (P50.05) was included in the final multivariable model.
consensus clinical diagnosis evaluation (n = 200).22 In the IDCR
a random sample of 30 carers of individuals who scored high Results
on the ADOS-Mod1 (57) and a random sample of 30 carers
of individuals who scored low on the ADOS-Mod1 (57) were Achieved sample and response rate
invited to take part in an interview by a senior research Of 13 171 households identified as potentially eligible in the
psychologist (J.S.) using the DISCO and ADI-R, to test the APMS, 7461 (57%) provided a complete phase-one interview of
accuracy of the ADOS-Mod1 in identifying individuals with whom 849 were selected for phase-two interviews. Of these 630
autism. Both studies confirmed the diagnostic thresholds for (74%) completed phase-two assessments: 618 full ADOS-Mod4
autism originally recommended by the developers of the ADOS.19 { assessments were carried out in the APMS. Analyses reported
Diagnostic interviewers were experienced in psychological previously11 found no evidence of non-response bias.
research, and received an induction and training programme, In the IDCR study, response rates were much higher in
run by a senior research psychologist (J.S.), a psychiatrist Leicestershire under the opt-out ethical approval procedure than
{
Researchers and clinicians using the ADOS-Mod4 in clinical populations should note certain differences in the method we have been using to calculate
ADOS-Mod4 total score (all scores available for secondary analysis in the data archive; see Acknowledgements). The published ADOS-Mod4 algorithm protocol that
was developed in clinical populations19 includes subdomain rules (e.g. a threshold of 10 or greater (10+) on the ADOS-Mod4 total score for identifying cases
of autism,19only if there are scores of 3+ on the Communication subdomain and 6+ on the Reciprocal Social Interaction subdomain). However, calibration by us of
the ADOS-Mod4 score22 for general population research was performed across ADOS-Mod4 score thresholds ranging from 5+ to 13+; for all but two of these
thresholds (7+ and 10+) no subdomain rules exist.10 Therefore, no subdomain rules were used in calibrating the ADOS-Mod4 total score, which we based on the
sum total of the subdomain ratings of Communication and Reciprocal Social Interaction. However, the subdomain rule for identifying cases of ’autism spectrum’
(7+) (scores of 2+ on the Communication subdomain and 4+ on the Reciprocal Social Interaction subdomain) does apply to all cases of ’autism’ (10+ threshold) in
our general population studies using the ADOS.
499
Brugha et al
Prevalence (%)
few family carers of adults living in private households in Lambeth
or Sheffield responded to the written invitation so, under ‘opt-in’
40 –
procedures, almost all could not be contacted further. Never-
theless, the achieved communal care establishments sample in
Leicestershire compared well with the case-register population
(see online supplement DS3), although the participants in the 20 –
private household sample were more likely to be male and have
more severe intellectual disability.
Of 290 individuals interviewed, 276 were assessed for autism. 0–
Assessments with the remaining 14 were attempted but could not Profound intellectual Moderate/severe None to mild
be completed because participants had profound and multiple disability disability disability
disabilities and assessors were unable to give a confident
assessment. Missing values in the APMS were minimal (51% Fig. 1 Gradient of autism prevalence by intellectual ability;
on all variables): there were 12 (4.3%) individuals in the IDCR combined sample.
study who had no Vineland assessment but were assessed for Intellectual ability is classified using the Vineland II caregiver-rating form for the
Intellectual Disability Case Register (IDCR) sample; those in the Adult Psychiatric
autism. Sensitivity analyses with these sequentially counted as Mobility Survey (APMS) sample are assumed to have no or mild intellectual disability.
having and not having intellectual disability had no effect on
the findings. Other missing values in the IDCR study were
infrequent and are shown in the tables where they amount to
more than 5% of n.
online Table DS4 for a version of Table 1 covering a larger number
of characteristics). The increased prevalence of South Asian
Participant characteristics by intellectual ability ethnicity reflects the location of most of the IDCR sample in
Participants with moderate to profound intellectual disability were Leicestershire (online Table DS4). Those in the sample with
more likely to be male, younger, and were more ethnically diverse moderate to profound intellectual disability were more likely to
than those with no or mild intellectual disability (see Table 1 and be disabled and less likely to have ever worked.
Gender, n (%)
Male 121 (55.8) 19 (40.4) 3130 (43.0)
Female 96 (44.2) 28 (59.6) 4144 (57.0)
Age group, n (%)
18–29 38 (17.5) 13 (27.7) 921 (12.7)
30–44 62 (28.6) 18 (38.3) 1966 (27.0)
45–64 97 (44.7) 10 (21.3) 2409 (33.1)
65+ 20 (9.2) 6 (12.8) 1978 (27.2)
Intellectual ability, n (%)b
Profound intellectual disability 125 (57.6) – –
Severe intellectual disability 58 (26.7) – –
Moderate intellectual disability 34 (15.7) – –
Mild/borderline intellectual disability – 47 (100) –
IQ 70–85 – – 1006 (13.8)
IQ 86–100 – – 1829 (25.1)
IQ 101+ – – 3916 (53.8)
IQ not assessed – – 523 (7.2)
Activities of daily living (ADL)
ADL difficulties, median (IQR)c 7 (7, 7) 5 (4, 7) 0 (0, 1)
ADL with a lot of difficulty, median (IQR) 6 (4, 7) 2 (0, 3) 0 (0, 0)
Participants with missing data on ADL, n (%) 13 (6.0) 8 (17.0) 18 (0.2)
Work, n (%)
Never in paid work 185 (85.3) 30 (63.8) 230 (3.2)
Ever in paid work 10 (4.6) 11 (23.4) 6975 (95.9)
Missing 22 (10.1) 6 (12.8) 69 (0.9)
IDCR, Intellectual Disability Case Register; APMS, Adult Psychiatric Mobility Survey; IQR, interquartile range.
a. See online Table DS4 for a more detailed version of this table covering a larger number of variables.
b. Classified using the Vineland II caregiver-rating form26 for the Intellectual Disability Case Register (IDCR) sample and the National Adult Reading
Test (NART) for Adult Psychiatric Morbidity Survey (APMS) sample. Twelve adults from the IDCR study were excluded because they could not be classified.
c. Difficulty with seven ADL including personal care, getting out and about and using transport, medical care, household activities, practical activities, paperwork and managing money.
500
Epidemiology of autism in adults across ability levels
Autism prevalence by age, gender and intellectual significant gender6intellectual disability interaction on autism
ability prevalence, with men and women with at least moderate
There were 14 men and 4 women with autism in the APMS intellectual disability having similar prevalence. Previous studies
subsample, and 49 men and 40 women with autism in the IDCR of adults with intellectual disability have found a higher rate of
subsample. The prevalence of autism in England, estimated from autism in men than women9,30,31 although not as high as for
the combined reweighted sample, was 1.1% (95% CI 0.3–1.9%). the rest of the population.32 Childhood population estimates33
Because people with moderate to profound intellectual disability have reported a male:female ratio of 2.1 : 1 for children with
make up just 0.3% of the total population, overall associations IQ570 and 3.7 : 1 for those without; an administrative study34
of autism with age and gender for the population as a whole are also found that the gender ratio diminished with increasing
unchanged by the inclusion of rates for people with intellectual disability level in children; the Global Burden of Disease (GBD)
disability. There was a gradient of autism prevalence by project estimate,1 based on childhood, incidence and mortality
intellectual ability (Fig. 1), with prevalence considerably higher data, was three times commoner in males than females with
in those with moderate to profound intellectual disability autistic disorders (autism with delay in language or cognitive
(39.3%, 95% CI 31.0–48.4, compared with 1.0%, 95% CI 0.4– development) and over four times commoner for other forms of
2.2 in those with no or mild intellectual disability (OR = 63.5, autism.
95% CI 27.4–147.2)).
In the population with moderate to profound intellectual dis- Strengths and limitations
ability, prevalence of autism was not specifically associated with The strength of this study lies in the comprehensive epidemiological
gender, being 42.3% (95% CI 31.1–54.3) in men and 35.2% sampling of adults of all ability levels in defined geographic areas
(95% CI 23.5–49.0 ) in women, P = 0.43 (Table 2). However, in and the use of direct diagnostic assessments of autism carefully
the population with no or mild intellectual disability, prevalence validated in the study samples with the aim of achieving comparable
was considerably higher in men at 1.9% (95% CI 0.8–4.2) than measurement across intellectual ability levels. However, there is
in women 0.2% (95% CI 0.0–0.7). The interaction between potential for selection bias on the estimate of autism prevalence
intellectual disability and gender on the prevalence of autism in the IDCR study as a result of the low response in the IDCR
was statistically significant (P = 0.02; Wald test) and remained private household sample. Detailed investigation of the pattern
statistically significant when adjusted for age and presence of of non-response by age, gender, residence and presence of autistic
epilepsy (Table 3). There was evidence of a small decline in the traits in Leicestershire (online supplement DS3), makes type II
prevalence of autism with age, statistically significant only in those error unlikely (i.e. failure to find a relationship between gender
with moderate to profound intellectual disability (Table 2). and prevalence of intellectual disability where it really exists).
We used moderate intellectual disability assessed by the
Discussion Vineland II caregiver-rating form in the IDCR as a threshold for
intellectual disability in the logistic regression, with none or mild
This standardised whole population sample case-finding study has intellectual disability imputed for the APMS sample. This measure
yielded new understanding of the prevalence of autism and its is consistent with other recent prevalence studies of adults,35,36
associations in adults with intellectual disability, gender and age. giving a standardised but more exclusive measure of intellectual
The usual male gender excess for autism in childhood2,3 was not disability. Our results were substantially unchanged when we
evident among adults with intellectual disability, showing a reanalysed with intellectual disability defined pragmatically as lack
Gender
Women 1.00 1.00
Men 1.35 (0.64–2.83) 0.43 8.97 (2.20–36.52) 0.002 0.02
Age (year) 0.96 (0.93–1.00) 0.008 0.98 (0.92–1.04) 0.51 0.61
a. Weighted to represent the English population by age, gender, intellectual disability and type of residence.
b. Classified using the Vineland II caregiver-rating form in the Intellectual Disability Case Register (IDCR); those in the Adult Psychiatric Mobility Survey (APMS) sample are assumed to
have no or mild intellectual disability.
Table 3 Multivariate predictors of autism by intellectual disability, with gender6intellectual disability interaction a
OR (95% CI)
Moderate to profound No or mild/borderline P-value for variable6intellectual
Characteristic All intellectual disabilityb intellectual disability disability interaction
Gender
Women – 1.00 1.00 –
Men – 1.31 (0.58–2.99) 8.46 (2.05–34.80)c 0.03
Age (year) 0.98 (0.92–1.05) – – –
a. Reweighted to represent the English population by age, gender, intellectual disability and type of residence and adjusted for carer or self-reported epilepsy or fits since age 16.
b. Classified using the Vineland II caregiver-rating form in the Intellectual Disability Case Register (IDCR); those in the Adult Psychiatric Mobility Survey (APMS) sample are assumed
to have no or mild intellectual disability.
c. P50.01.
501
Brugha et al
of decision-making capacity to consent and to participate in a independent work on this association using case-finding population
household survey. This is closer to a threshold of mild intellectual research methods is needed.
disability, but with unavoidable undercounting of those with mild It was noted that research-identified individuals with autism
disability in the APMS. reported previously in the able household population11 had not
Analysis was limited by the small number of individuals with been recognised or diagnosed by health services. New findings
autism, particularly in the APMS sample. The presented analyses reported here suggest that the research case-finding measures used
are weighted to represent the national population by age, gender, may also fail to identify women with autism who do not have
intellectual disability and type of residence. Calculation of the intellectual disabilities, possibly adding further to the invisibility
IDCR weights was subject to error as it relied upon incomplete of autism in society. The picture that emerges is of a large
official statistics, and on the assumption that the three case-register population of adults who are significantly disabled whose needs
areas represent the English population as a whole. Detailed remain unmet because they are not recognised, particularly when
sensitivity analyses found that the effects of estimating unknown they do not have intellectual disabilities. The clinical, health and
population quantities on the overall prevalence estimates was economic implications are potentially enormous and urgently
minimal, giving prevalence of between 1.1 and 1.2%, regardless merit the attention of further research.
of assumptions made.37 We acknowledge that for some there may be a difficulty in
placing the present prevalence work in a familiar autism diagnostic
context. The instruments used in our survey, including the ADOS,
Interpretation of our findings in many ways are reflective of, but are not directly derived from
There are various hypotheses that could account for the gender either of the classification systems current when we developed the
pattern we found: women with autism could be more severely study. These were the ICD-1049 and the DSM-IV-TR50 diagnostic
impaired38 or there could be more ‘missed’ cases of autism in criteria for pervasive developmental disorder (including childhood
women without intellectual disability. Missed cases could result from autism and Asperger disorder) both primarily developed for use
male bias in autism diagnostic markers;39 female presentation of in childhood. We believe that the strength of our approach to
autism may differ from male presentation and measures may be assessing autism in adulthood is that it is achieved in such a
less able to detect the female presentation.39,40 Autism in women way as to make future replication and reliable comparison by
of average or above average intelligence may be ‘masked’ by other independent researchers possible in other countries and
conditions, such as eating disorders,41 anxiety disorders42 and populations. All such research requires a trade-off between
borderline personality disorders.43 Such women may be better reproducibility and clinical familiarity and usage. We believe that
than women with intellectual disability at hiding their difficulties for a first-ever epidemiological study in adults we have taken the
by imitating social interactions,44 having better language skills, most appropriate line using the most relevant instruments, with
different special interests, less hyperactivity and aggression.45 If clearly established reliability, supported by validation studies.51
more able women with autism are not diagnosed or are incorrectly In future work additional clinical information will be collected
diagnosed, then the prevalence of autism could be underestimated with the intention of making it possible to describe findings in
and their needs unmet. Biological theories for the male excess relation to more recent developments in the classification of
of autism46 may also benefit from reconsideration. Gender autism such as the DSM-552 and ICD-11 autism spectrum
determination and autism could have a common cause, i.e. disorder criteria currently being finalised and possibly also to
intrauterine levels of testosterone. Differences in gender ratio future such developments in classification.
could reflect different causes for autism. One small study (n = 94
Traolach S. Brugha, MD(NUI) FRCPsych, Nicola Spiers, MSc PhD, Department
individuals with autism)47 reported a male:female gender ratio that of Health Sciences, University of Leicester, Leicester; John Bankart, MSc PhD,
was closer to 2 : 1 in individuals with genetic, magnetic resonance Department of Primary Care and Health Sciences, University of Keele, Stoke-on-Trent;
Sally-Ann Cooper, MD FRCPsych, Institute of Health and Wellbeing, University of
imaging or clinical morphological abnormalities and found gender Glasgow, Gartnavel Royal Hospital, Glasgow; Sally McManus, MSc, NatCen Social
ratios closer to 8 : 1 in individuals without these abnormalities. Research, London; Fiona J. Scott, PhD Cpsychol, Jane Smith, MSc, Freya Tyrer,
MSc, Department of Health Sciences, University of Leciester, Leciester, UK
Although almost two in five adults with moderate to profound
intellectual disability had autism, higher than expected based Correspondence: Traolach S. Brugha, Department of Health Sciences,
on previous research,9,32,48 only 1% of adults with no or mild University of Leicester, Leicester General Hospital, Gwendolen Road, Leicester
LE5 4PW, UK. Email: tsb@le.ac.uk
intellectual disability had autism. But because moderate to
profound intellectual disability affects only 0.3% of all adults9 First received 31 Aug 2015, final revision 13 Dec 2015, accepted 21 Jan 2016
502
Epidemiology of autism in adults across ability levels
16 Lord C, Rutter M, Le Couteur A. The autism diagnostic observation schedule- 43 Rydén G, Rydén E, Hetta J. Borderline personality disorder and autism
generic: a standard measure of social and communication deficits associated spectrum disorder in females – a cross-sectional study. Clin Neuropsychiatry
with the spectrum of autism. J Autism Dev Disord 2002; 30: 205–23. 2008; 5: 22–30.
17 Jenkins R, Lewis G, Bebbington P, Brugha T, Farrell M, Gill B, et al. The 44 Holliday-Willey L. Pretending to be Normal: Living with Asperger’s Syndrome.
national psychiatric morbidity surveys of great Britain– initial findings from Jessica Kingsley Publishers, 1999.
the household survey. Psychol Med 1997; 27: 775–89. 45 Gillberg C. The Biology of the Autistic Syndromes (3rd edn). Cambridge
18 Baron-Cohen S, Wheelwright S, Skinner R, Martin J, Clubley E. The autism- University Press, 2000.
spectrum quotient (AQ): evidence from Asperger syndrome/high-functioning 46 Baron-Cohen S, Lombardo MV, Auyeung B, Ashwin E, Chakrabarti B,
autism, males and females, scientists and mathematicians. J Autism Dev Knickmeyer R. Why are autism spectrum conditions more prevalent in
Disord 2001; 31: 5–17. males? PLoS Biol 2011; 9: e1001081.
19 Lord C, Rutter M, DiLavore PC, Risi S. Autism Diagnostic Observation 47 Miles JH, Hillman RE. Value of a clinical morphology examination in autism.
Schedule ADOS Manual. Western Psychological Services, 2002. Am J Med Genet 2000; 91: 245–53.
20 Lord C, Rutter M, Le Couteur A. Autism Diagnostic Interview – Revised: a 48 Saemundsen E, Juliusson H, Hjaltested S, Gunnarsdottir T, Halldorsdottir T,
revised version of a diagnostic interview for caregivers of individuals with Hreidarsson S, et al. Prevalence of autism in an urban population of adults
possible pervasive developmental disorders. J Autism Dev Disord 1994; 24: with severe intellectual disabilities – a preliminary study. J Intellect Disabil
659–85. Res 2010; 54: 727–35.
21 Wing L, Leekam SR, Libby SJ, Gould J, Larcombe M. The Diagnostic Interview 49 World Health Organization. The ICD-10 Classification of Mental and
for Social and Communication Disorders: background, inter-rater reliability Behavioural Disorders: Diagnostic Criteria for Research. WHO, 1993.
and clinical use. J Child Psychol Psychiatry 2002; 43: 307–25. 50 American Psychiatric Association. Diagnostic and Statistical Manual of
22 Brugha TS, McManus S, Smith J, Scott FJ, Meltzer H, Purdon S, et al. Mental Disorders (4th edn, text revision) (DSM-IV-TR). American Psychiatric
Validating two survey methods for identifying cases of autism spectrum Association, 2000.
disorder among adults in the community. Psychol Med 2012; 42: 647–56. 51 Brugha TS, McManus S, Smith J, Scott FJ, Meltzer H, Purdon S, et al.
23 American Psychiatric Association. Diagnostic and Statistical Manual of Validating two survey methods for identifying cases of autism spectrum
Mental Disorders (4th edn). American Psychiatric Association, 1994. disorder among adults in the community. Psychol Med 2012; 42: 647–56.
24 Department of Health. Valuing People. A new strategy for Learning Disability 52 American Psychiatric Association. Diagnostic and Statistical Manual of
for the 21st century (White Paper). TSO (The Stationery Office), 2001. Mental Disorders (5th edn) (DSM-5). APA, 2013.
503
Data supplement to Brugha et al. Epidemiology of autism in adults across age
groups and ability levels. Br J Psychiatry doi: 10.1192/bjp.bp.115.174649
Online supplement DS1: Sampling details of the Intellectual Disability Case Register
study (IDCR)
In order to achieve a complete sample of adults in England two groups of adults not
identified: those living in a private household but incapable of taking part in a survey
interview because of intellectual disability, and those not included in the APMS sample due
to being resident in a communal care establishment (or ‘institution’). The IDCR was based
on samples drawn from three registers of adults with intellectual disability in England,
located in Leicestershire, Lambeth and Sheffield, although the achieved sample in Sheffield
was very small. The adult prevalence of intellectual disability, defined by significant
intellectual impairment (WHO definition) with adaptive skill deficits, using the registers at the
three sites was within the expected range (4.9, 4.3 and 5.4/ 1000 of the population of
Leicestershire, Lambeth and Sheffield respectively (National Statistics (1)). These registers
A random sample of individuals on the registers from private households with addresses
eligible for the APMS, was selected, stratified by age, sex and type of residence. Participants
were then excluded if they were judged to have decision-making capacity to consent and
sufficient ability to participate in the APMS. Interviewers initially made this judgement over
the telephone and again when visiting the potential participant in their own home (if they had
not already been excluded following the initial telephone conversation). The study also
included residents of communal care establishments, comprising any domestic group setting
with an intrinsic care function, such as a residential home or nursing home. For
Leicestershire and Sheffield, the IDCR study adopted a two stage sampling design for
communal care establishments. The first stage involved randomly selecting establishments
1
with four or more residents, with selection probabilities proportional to the number of eligible
residents. The second stage involved randomly selecting four participants from each of the
chosen communal care establishments stratified by sex and age. For Lambeth, which had
fewer communal care establishments, all establishments with 3 or more residents were
Consent was obtained following the English mental capacity act, 2005, for adults who did not
participants who could not speak English and had non-English speaking carers. Sampled
participants, carers or managers were sent a letter of invitation, with information about the
study. Participants in Leicestershire were telephoned by the research team ('opt-out consent
procedure'); carers and participants in Lambeth contacted the research team only if they
In the IDCR study the primary sampling unit was the communal care establishment or, in the
case of those in private households, the individual. In the APMS the primary sampling units
were individual or small groups of postcode sectors. The datasets were appended, giving a
total of 268 strata, 260 by socioeconomic status within region from the APMS and 8 by age,
sex and residence (communal establishment or private household) from the IDCR study.
APMS data were weighted to account for the complex sampling design, including two-stage
sampling for autism as described elsewhere (5). The combined data (figure) were weighted
to represent the English population by age, sex, intellectual disability and type of residence.
Post-stratification weights were calculated using data from official statistics, including the
mid-year population estimates derived from the 2001 census and data on intellectual
disability from the Social Services Activity Statistics and the case registers. Full details are
2
Figure: Illustrating the calculation of the combined autism prevalence estimate
intellectual disability across the APMS and IDCR samples and in the assessment of
Both ADOS-Mod1 and ADOS-Mod4 have been carefully validated in the study samples with
the aim of achieving comparable measurement across intellectual ability levels. There is
some concern that autism assessment in the intellectual disability study with the ADOS-
Mod1 achieved only moderate sensitivity on validation. Assessors suggested that this may
be due to assessment tasks that are age-inappropriate for the adult population, for example
using bubbles, dolls and toy cars. It is possible that missed cases at older ages may
3
account for the decline in prevalence of autism with age found in the intellectual disability
study; alternatively, there may be an earlier age of death in the group with both autism and
intellectual disability (14;15). It is also possible that older people may do better in
Online Supplement DS3: Examination of Potential for Selection Bias in the IDCR
The 276 adults who were assessed for autism had a similar age structure to the non-
participants from the case registers, but with some oversampling of women (Table DS1).
Further information was recorded for 97% (n=3080) of those on the Leicestershire case
register only. This included a measure of autistic traits previously validated against clinical
diagnoses (16).
The traits were poor quality of social interaction, lack of empathy, simple stereotypies and
elaborate routine, usually carer-reported, but self-reported for a small minority who were
establishments had similar intellectual disability severity, ethnicity, epilepsy and autistic trait
profiles (Table DS1). Participants in private households were younger, had more severe
intellectual disability and greater autism, prevalence than non-participants. On restricting the
sample to adults with moderate to profound ID, the prevalence of epilepsy was marginally
higher among participants (30.6% and 24.3% for participants and non-participants
respectively) and the mean number of autistic traits was similar (0.72 and 0.74 for
Prevalence of two or more autistic traits in communal care establishments was 22.6% in
men and 21.4% in women, compared to 14.4% and 9.9% respectively in the private
with two or more autistic traits were somewhat oversampled compared to those with no or
one trait, while the proportion of women with two or more traits was identical in participants
4
and non-participants (Table DS2). In the private household sample numbers were too small
to assess possible non-response bias. For the unexpected number of female autism cases
sampling of autistic men and oversampling of autistic women in the IDCR achieved sample.
There is no evidence of this in the comparison with the Leicestershire populations (Table
DS3), although assessment was limited by small numbers in the private household sample.
While there is some potential for selection bias on the estimate of autism prevalence in the
IDCR study private household sample, bias sufficient to undermine our finding of an
interaction between sex and intellectual disability on the prevalence of autism is implausible.
The IDCR sample is drawn largely from Leicestershire, a population that compares closely to
the English population on socio-demographic characteristics (Table DS2), with the exception
of greater numbers of South Asian ethnicity, mainly resident in the City of Leicester. The
proportion of school pupils with an official statement of special education needs, and the
suggesting that a honeypot effect is unlikely. There are approximately 0.5m people living in
the City of Leicester and surrounding built up area, 0.3m living in other urban areas (market
towns or other conurbations with population>10,000), and 0.2m living in rural areas.
Leicestershire lacks a major conurbation, which may be relevant given evidence from the
Leicestershire case register that autism is more prevalent in rural areas (7), but our sample
In conclusion, the ‘opt-in’ consent procedure in Lambeth and Sheffield, proved an almost
insuperable barrier to recruitment, meaning that the IDCR sample is largely based upon the
Leicestershire was successfully surveyed, with 64% of establishments and 69% of eligible
individuals approached taking part. However, the response rate in private households in
Leicestershire was low at 35%. Difficulties in recruiting adults with intellectual disability into
research studies have been well documented (8-10). Many refusals were because carers
5
were worn out with a lifetime of caring. However, studies of autism in adults are very rare,
and the response rate is consistent with others (11;12). Only, the Glasgow cohort (13)
achieved a response rate of 70.6%, based on exceptional quality of routine data coding, and
collaborative working between the clinical team and research team in providing information
and consenting, but it did not have a validated systematic diagnostic assessment.
Additional References
(2) Roeleveld N, Zielhuis GA, Gabreels F. The prevalence of mental retardation: a critical review
of recent literature. Dev Med Child Neurol 1997 Feb;39(2):125-32.
(3) Institute of Public Care. Estimating the prevalence of severe learning disability in adults
(working paper). Oxford: Institute of Public Care via Projecting Adult Needs and Service
Information (PANSI); 2009.
(4) Parrott R, Emerson E, Hatton C, Wolstenholme J. Future demand for residential provision for
people with learning disabilities. Manchester: Hester Adrian Research Centre, University of
Manchester; 1997.
(5) Brugha TS, McManus S, Bankart J, Scott F, Purdon S, Smith J, et al. Epidemiology of autism
spectrum disorders in adults in the community in England. Arch Gen Psychiatry 2011
May;68(5):459-65.
(6) Brugha T, Cooper SA, McManus S, Purdon S, Scott FJ, Spiers NA, et al. Estimating the
Prevalence of Autism Spectrum Conditions in Adults: Data quality and methodology
document. Leeds: The NHS Information Centre; 2012 Jan 31.
(7) Kiani R, Tyrer F, Hodgson A, Berkin N, Bhaumik S. Urban-rural differences in the nature and
prevalence of mental ill-health in adults with intellectual disabilities. J Intellect Disabil Res
2013 Feb;57(2):119-27.
(8) Evenhuis H, van SJ, Vink M, Weerdenburg C, van ZB, Stilma J. Obstacles in large-scale
epidemiological assessment of sensory impairments in a Dutch population with intellectual
disabilities. J Intellect Disabil Res 2004 Nov;48(Pt 8):708-18.
(9) Lacono T. Issues of informed consent in conducting medical research involving people with
intellectual disabilities. Journal of Applied Research in Intellectual Disabilities 2003;16:41-51.
(10) Oliver PC, Piachaud J, Done J, Regan A, Cooray S, Tyrer P. Difficulties in conducting a
randomized controlled trial of health service interventions in intellectual disability:
implications for evidence-based practice. J Intellect Disabil Res 2002 May;46(Pt 4):340-5.
6
(11) Bhaumik S, Tyrer FC, McGrother C, Ganghadaran SK. Psychiatric service use and psychiatric
disorders in adults with intellectual disability. J Intellect Disabil Res 2008 Nov;52(11):986-95.
(13) Cooper SA, Smiley E, Morrison J, Williamson A, Allan L. Mental ill-health in adults with
intellectual disabilities: prevalence and associated factors. Br J Psychiatry 2007 Jan;190:27-
35.
(14) Gillberg C, Billstedt E, Sundh V, Gillberg IC. Mortality in autism: a prospective longitudinal
community-based study. J Autism Dev Disord 2010;40(3):352-7.
(15) Bilder D, Botts EL, Smith KR, Pimentel R, Farley M, Viskochil J, et al. Excess mortality and
causes of death in autism spectrum disorders: a follow up of the 1980s Utah/UCLA autism
epidemiologic study. J Autism Dev Disord 2013 May;43(5):1196-204.
(16) Bhaumik S, Tyrer F, Barrett M, Tin N, McGrother CW, Kiani R. The relationship between
carers' report of autistic traits and clinical diagnoses of autism spectrum disorders in adults
with intellectual disability. Res Dev Disabil 2010 May-Jun;31(3):705-12
7
Table DS1. Comparison of Case Register Characteristics between Participants
and Non-Participants, by residence and setting
Participants Non-participants
with an autism enrolled on case
assessment registers when
sampled for IDCR
Private households
All (N=79) (N=4444)
Age (when sampled) 36.0±14.1 38.0±14.5
Sex
Male 41 (51.9) 2551 (57.4)
8
Plus-minus values are means±SD
a. Poor quality of social interaction, limited empathy, presence of elaborate routines
and presence of stereotypies (Bhaumik 2010; Holmes et al. 1982)
b. As measured using the Leicestershire ID Scale (Tyrer et al. 2008)
c. Self- or carer-reported epilepsy or fits since age 16.
9
Table DS2. Comparison of Leicestershire Case Register Characteristics between
Participants and Non-Participants by Sex.
Participants Non-participants
with an autism enrolled on case
assessment registers when
sampled for IDCR
Private Households
Males (N=38) (N=1122)
Autistic traits†
≥2 traits 3 7.9 160 14.3
≤1 trait 33 86.8 936 83.4
Missing 2 5.3 26 2.3
Females (N=36) (N=898)
Autistic traits†
≥2 traits 5 13.9 86 9.6
≤1 trait 30 83.3 795 88.5
Missing 1 2.8 17 1.2
10
Table DS3. Leicestershire and England: Sociodemographic Characteristics and
Special Educational Needs (2011 Census).
Characteristic Leicestershire England
Usual resident population (%,N=1017697) (%,N=53,012,456)
Age – yr
0 to 15 19.0 18.9
16 to 44 39.7 39.4
45 to 64 25.5 25.4
65 and over 15.8 16.3
Sex
Male 49.4 49.2
Ethnic group
White 78.4 85.4
South Asian 16.1 7.8
Black/African/Caribbean/Black 2.4 3.5
British
Other/mixed 3.1 3.3
Geography
Urban (built-up area >10,000 77.9 82.4
inhabitants)
Usual resident population aged 16 (%,N=824,351) (%,N=42,989,620)
years and over
Highest Educational Qualification
None 24 22.5
GCSE or equivalent *, 27.9 28.5
Post-school† 42.2 43.4
Other ‡ 6 5.7
11
Table DS4: Sample characteristics.a
Characteristic Moderate to profound intellectual No or mild/borderline intellectual disability
disability
IDCR (n=217) IDCR (n=47) APMS (n=7274)
Gender, n (%)
Male 121 (55.8) 19 (40.4) 3130 (43.0)
Female 96 (44.2) 28 (59.6) 4144 (57.0)
Age group
18–29 38 (17.5) 13 (27.7) 921 (12.7)
30–44 62 (28.6) 18 (38.3) 1966 (27.0)
45–64 97 (44.7) 10 (21.3) 2409 (33.1)
65+ 20 (9.2) 6 (12.8) 1978 (27.2)
Ethnic group, n (%)
White 176 (81.1) 42 (89.4) 6700 (92.1)
South Asian 29 (13.4) 2 (4.3) 185 (2.5)
Black 8 (3.7) 0 191 (2.6)
Other/missing 4 (1.8) 3 (6.4) 198 (2.7)
Residence
Private household 68 (31.3) 9 (19.2) -
Communal establishment 149 (68.7) 38 (80.9) -
Intellectual ability, n (%)b
Profound intellectual disability 125 (57.6) - -
Severe intellectual disability 58 (26.7) - -
Moderate intellectual disability 34 (15.7) - -
Mild/borderline intellectual disability - 47 (100) -
IQ 70–85 - - 1006 (13.8)
IQ 86–100 - - 1829 (25.1)
IQ 101+ - - 3916 (53.8)
IQ not assessed - - 523 (7.2)
Activities of daily living (ADL)
ADL difficulties, median (IQR)c 7 (7, 7) 5 (4, 7) 0 (0, 1)
ADL with a lot of difficulty, median (IQR) 6 (4, 7) 2 (0, 3) 0 (0, 0)
12
Participants with missing data on ADLs, n (%) 13 (6.0) 8 (17.0) 18 (0.2)
Mobility, n (%)
No difficulty 19 (8.8) 20 (42.6) 6253 (86.0)
Some difficulty 66 (30.4) 18 (38.3) 657 (9.0)
A lot of difficulty 132 (60.8) 9 (19.1) 364 (5.0)
Work, n (%)
Never in paid work 185 (85.3) 30 (63.8) 230 (3.2)
Ever in paid work 10 (4.6) 11 (23.4) 6975 (95.9)
Missing 22 (10.1) 6 (12.8) 69 (0.9)
a. This is a more detailed version of Table 1 in the main text.
b. Classified using the Vineland II caregiver rating form26 for the Intellectual Disability Case Register (IDCR) sample, and the National Adult
Reading Test for Adult Psychiatric Morbidity Survey (APMS) sample. 12 adults from the IDCR study are excluded because they could not be
classified.
c. Difficulty with seven ADL including personal care, getting out and about and using transport, medical care, household activities, practical
activities, paperwork and managing money.
13
Epidemiology of autism in adults across age groups and ability
levels
Traolach S. Brugha, Nicola Spiers, John Bankart, Sally-Ann Cooper, Sally McManus, Fiona J. Scott, Jane
Smith and Freya Tyrer
BJP 2016, 209:498-503.
Access the most recent version at DOI: 10.1192/bjp.bp.115.174649
References This article cites 34 articles, 2 of which you can access for free at:
http://bjp.rcpsych.org/content/209/6/498#BIBL
Reprints/ To obtain reprints or permission to reproduce material from this paper, please
permissions write to permissions@rcpsych.ac.uk