* Correspondence to last author at Centre for Child and Adolescent Health, Hampton House, Bristol BS6 6JS, UK. E-mail: email@example.com
The aim of this study was to determine the prevalence of autistic spectrum disorder (ASD) within a large representative population sample: the Avon Longitudinal Study of Parents and Children (ALSPAC). Cases of ASD were identified from the clinical notes of children in the ALSPAC with a suspected developmental disorder and from the Pupil Level Annual Schools Census (PLASC) for England in 2003. Seventy-one cases of ASD diagnosed after a multidisciplinary assessment were identified from health records. There were an additional 15 cases from PLASC data in which ASD was mentioned as a principal difficulty, thus giving a total of 86 children diagnosed by the age of 11 years. Prevalence of ASD per 10 000 population at 11 years was 51.1 for those with a multi-professional diagnosis, and 61.9 if cases from education were included, made up of 21.6 for childhood autism, 10.8 for atypical autism, 16.6 for Asperger syndrome, and 13.0 for unspecified ASD. The male:female ratio was 6.8:1. Median age at diagnosis ranged from 45 months in childhood autism to 116 months in Asperger syndrome. A comorbid developmental disorder was recorded in 33.8% of cases, including learning disability* in 14.7%, epilepsy in 10.3%, and mixed developmental disorder in 4.4%. We conclude that the prevalence of ASD diagnosed at 11 years in a UK representative population-based sample is at least 51.1/10 000.
Autistic spectrum disorder (ASD) is a pervasive developmental disorder characterized by abnormal development of social interaction, communication, and behaviour. In recent years, estimates of prevalence have ranged from 12.2 to 67.4 cases per 10 000 population.1 A large population-based study from South Thames, UK,2 has recently reported a prevalence of ASD of 116 per 10 000, and another study3 has estimated the prevalence of the wider autistic spectrum to be up to 2.7% of children. Other developmental disorders and medical conditions are recognized to co-occur in individuals with a diagnosis of ASD, although precise comorbidities and prevalence rates vary widely in the literature. A review of studies investigating disorders co-existing with ASD concluded that studies of large general population samples where comorbidity patterns can be analyzed without bias should receive high priority because this information can be used to increase understanding of the possible subtypes of childhood autism and the genetic links.4
The Avon Longitudinal Study of Parents and Children (ALSPAC),5 a large population-based birth cohort study, specifically designed to investigate how an individual’s genotype interacts with environmental pressures to influence health and development, is an ideal sample to use for this research. The aim of this study was to report on the prevalence of ASD in the ALSPAC cohort, and to describe comorbid developmental impairments. We also report on the median age at diagnosis for each of the ASDs found in the ALSPAC cohort, and the social and demographic characteristics of children diagnosed with ASD. This information is vital not only so that we can better understand and provide services for children with ASD, but also as a basis for elucidating the contribution of environmental and genetic influences to autistic disorders.
ALSPAC is a longitudinal cohort study following the health and development of children who had an expected date of delivery between April 1991 and December 1992, and who were resident in the Avon area of southwest England at the time of their birth. The initial ALSPAC sample consisted of 14 541 pregnancies, which resulted in 14 062 live births. This study population had social and demographic characteristics in common with the 1991 UK national census.6
Ethical permission was granted for this study by the local research ethics committees, and Caldicott Guardian approval was gained for linkage of UK National Health Service (NHS) data for this investigation. All research was within the guidelines set out by these bodies, and the study was monitored by the ALSPAC Law and Ethics Committee. The ALSPAC ethical framework does not allow researchers to use information from ALSPAC to make contact with individual study members or their families. It was, therefore, not possible to use ALSPAC questionnaires to screen for autism, and cases had to be identified through health and education records. Permission to search their child’s medical records for research purposes was granted by the children’s mothers at the time of enrolment to the ALSPAC study.
Identifying cases of ASD
Information on which children in the ALSPAC may have been diagnosed with an ASD was gathered both from NHS and education sources. First, we requested data on all individuals with a date of birth in the ALSPAC cohort who had a diagnosis relating to any form of developmental delay for the period 1991 to 2003 inclusive (the diagnosis being based on the World Health Organization’s International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10]) from the computer systems of local NHS trusts (North Bristol Trust, United Bristol Healthcare Trust, Weston Area Health Trust and Royal United Hospital, Bath). Second, we requested information from the Child Health computer system in Bristol (shared across all the NHS trusts in the area) about all children who were identified as having special educational needs between 1993 and 2003 inclusive. These two lists were then matched against the ALSPAC cohort to confirm that the child was a member of the study and that permission had been given to search their health records. A team of three experienced researchers then searched hospital medical records (in-patient and outpatient) and community child-health records (including child development team records) to identify children who had a diagnosis of ASD made after a multidisciplinary assessment. No direct contact was made with any of the children, parents, or clinicians. The researchers searched for ICD-10 diagnoses according to a structured proforma, based on information that was available in the records written by the multiprofessional team involved in the care of the child. The date a diagnosis was confirmed by a multidisciplinary assessment was also recorded from the notes. The project’s principal investigator (AE), a consultant paediatrician with 30 years of clinical experience, reviewed all information collected from the notes and confirmed that the diagnostic information was consistent with ICD-10 criteria.
The data collected from NHS sources (both hospital and community notes) formed the core NHS dataset for this study. All the NHS Trusts in the ALSPAC area have specialist autism teams in children’s services, trained to use standardized assessment tools such as the Diagnostic Interview for Social and Communication Disorders,7 the Autism Diagnostic Observation Schedule,8 and the Asperger Syndrome Diagnostic Interview.9 Only cases diagnosed by multidisciplinary teams were included. Cases where a lone consultant or speech therapist applied a label of ASD were excluded. All personal identifiers were stripped from the NHS dataset and further work was performed by an ALSPAC statistician (KT), not by the researchers who had collected the data from the notes.
The second source of data on children with developmental disabilities came from national education records. ALSPAC supplied information on all study children from the Pupil Level Annual Schools Census (PLASC) dataset for 2003/4, supplied by the Department of Education for England. These data included children attending state schools (both mainstream and special schools) in England who were recorded as having some form of special educational needs (SEN) within that academic year. For each child, the PLASC data provided information on the level of extra help being provided at school (School Action, School Action Plus, or Statement of Special Educational Needs), and one of two principal difficulties for children at School Action Plus and Statement of Special Needs level. We searched the complete PLASC dataset for children with ASD listed as either a primary or a secondary concern. However, it was not possible to link this dataset with the children’s health records (hospital or community) because the ethical framework of ALSPAC does not allow data to be given to a researcher that identify individual children as having specific problems.6 Therefore, health records for those children whose ASD label was obtained from the PLASC database could not be searched to confirm these diagnoses, or to gain information on comorbidities.
All data manipulation and statistical analyses were performed using SPSS (version 12.0.1). Approximation to the normal distribution was used to calculate 95% confidence intervals (CI) for the prevalence rates. Between-group comparisons of ages and dates used the non-parametric Mann–Whitney U and Kruskal–Wallis tests, because the data were not normally distributed. Fisher’s exact test was used to test for an association in a two-by-two contingency table when expected counts were small.
Cases of ASD
A total of 86 children in the ALSPAC were diagnosed with ASD by 11 years of age, from an original cohort of 14 062 births. Searches of NHS notes produced 71 cases of a confirmed ASD diagnosis, and there were 51 children on the PLASC database who had a diagnosis of ASD listed as one of the two principal difficulties. There were 31 children that were only identified from the NHS data, 15 that were identified only from the PLASC data, and 40 that were common to both sources (Fig. 1).
The estimated prevalence of ASD at age 11 years in the ALSPAC cohort was 51.1 per 10 000 (95% confidence interval [CI] CI 39.2–62.9) for those diagnosed after multidisciplinary assessment, and 61.9 per 10 000 (CI 48.8–74.9) if the extra cases identified by education were included.
We also calculated an estimated prevalence for different autism diagnoses. The 15 cases identified only by the PLASC data were assigned the ‘unspecified ASD’ code, as we were not able to check their health records. The prevalence of childhood autism was estimated as 21.6 per 10 000 (CI 13.9–29.3; n=30), atypical autism as 10.8 per 10 000 (CI 5.3–16.3; n=15), Asperger syndrome as 16.6 per 10 000 (CI 9.8–23.3; n=23), and other or unspecified ASD as 13.0 per 10 000 (CI 4.8–15.3; n=18). There was no case of Rett syndrome or of ‘other childhood disintegrative disorder’.
Age at diagnosis
Age at diagnosis was only known for the 71 cases of ASD found by researchers through examination of the health records. Median age at diagnosis of ASD was 81.9 months (interquartile range [IQR] 42.7–116.6), or 6 years 7 months. Age at diagnosis was lowest for children with childhood autism, at a median age of 44.9 months (IQR 36.5–85.6), followed by atypical autism at 75.5 months (IQR 42.7–117.9), then Asperger syndrome at 115.9 months (IQR 93.4 to 130.6). Age at diagnosis for children with Asperger syndrome was significantly later than for children diagnosed with other forms of ASD (Kruskal–Wallis test, p<0.001).
Of the 86 cases of ASD in the ALSPAC cohort, 75 (87.2%) were males, and 11 (12.8%) females, a male:female ratio of 6.8:1 (Table I). Of the cases where the specific ASD diagnosis was known (childhood autism, atypical autism, or Asperger syndrome), there appeared to be a trend towards females being over-represented at the more severe end of the spectrum (childhood autism) compared with males (Table II). However, the total numbers of females with each ASD diagnosis were too low for statistical comparisons.
Table I. Comparison of demographic and other factors between children with ASD and the rest of the ALSPAC cohort
ASD present (%)
ASD absent (%)
Sample sizes vary because of missing data. Variable with the most missing data is maternal occupation (n=10 088). aOrdinary level pass at national assessments at 16 years. ALSPAC, Avon Longitudinal Study of Parents and Children, ASD, autism spectrum disorder.
Maternal age, y
Paternal age, y
Characteristics at birth
Sex of child
Highest educational attainment
Table II. Frequency of females and males having each subtype of ASD
Male, n (%)
Female, n (%)
ASD, autism spectrum disorder.
Other and unspecified ASD
Across all ASD diagnoses derived from NHS data, the median age at diagnosis for females was 81.9 months (IQR 40.4–119.5), and for males 83.8 months (IQR 43.0–116.5). This difference between the sexes for age at diagnosis was not statistically significant (Mann–Whitney U test, p=0.923).
There was no difference in the proportion of children with ASD who came from a non-white background compared with the rest of the cohort. Mean age of the mothers of children with ASD was slightly higher than the rest of the cohort, with more mothers of children with ASD aged 30 to 34 years; there was no association with paternal age. No difference was apparent in the educational background of the parents of children with ASD compared with the rest of the parents in the cohort. No consistent trend was found with maternal occupational social class, but fathers of children with ASD were more likely to have a non-manual occupation (Table I).
Median age at diagnosis for the children identified from NHS data was 89.2 months for children of mothers educated to less than Ordinary Level (O-level; high school) standard, 68.9 months for children with mothers educated to O-level, and 92.9 months for children with mother educated to greater than O-level. There was no statistical evidence of a difference (Kruskal–Wallis test, p=0.866).
The children with ASD did not differ from the rest of the cohort at birth for birthweight, gestation, or whether they had been in a multiple pregnancy.
Special educational needs status
Of all the ASD cases, 94.4% had a statement of SEN, 4.2% were on School Action Plus, and 1.4% (one child with Asperger syndrome) had no special provision. Data on special needs support were missing for 10 children with ASD: these children attended a private school, were home educated, or attended a school outside England.
Comorbid developmental impairments
Some children had other developmental impairments, as well as having a diagnosis of ASD. Data were available on specific conditions for the children who had a diagnosis of ASD confirmed by a multidisciplinary assessment (n=71): 33.8% had at least one associated developmental disorder. The figure was 36.7% for cases of childhood autism, 26.7% for atypical autism, and 34.8% for Asperger syndrome.
Table III summarizes the frequency and percentage of cases of comorbid developmental impairments that were found in association with diagnoses of ASD. Because children had between one and four developmental impairments each, the total number of developmental impairments does not relate to the number of children.
Table III. Frequency of other developmental conditions in children with a diagnosis of ASD
Childhood autism (n=30) n (%)
Atypical autism (n=15) n (%)
Asperger syndrome (n=23) n (%)
aMental retardation on ICD-10. ASD, autism spectrum disorder.
Specific motor function disorder and mixed developmental disorder
As there were few comorbid developmental impairment conditions associated with specific ASD diagnoses, no firm statistical conclusions can be drawn from these data. However, learning disability did appear to be more strongly associated with atypical autism (26.7% of cases) compared with Asperger syndrome and childhood autism (8.7% and 13.3% of cases respectively). Epilepsy was more strongly associated with childhood autism (16.7% of cases) than either atypical autism (no cases) or Asperger syndrome (8.7% of cases). Although classifications according to ICD-1010 and the DSM-IV11 do not allow the simultaneous labelling of autism and attention-deficit–hyperactivity disorder, it is well recognized in clinical practice that there is an overlap. In the ALSPAC sample, hyperkinetic disorders appeared to be more associated with Asperger syndrome (13% of cases), than either childhood autism or atypical autism (3.3% and 6.7% of cases respectively).
Although we have used a large population-based sample, our case-finding methods were biased towards children with a statement of SEN who were attending English state schools in 2003. We are likely to have under-reported children at the higher-functioning end of the autistic spectrum (Asperger syndrome), who, typically, may not have a statement of SEN, and who may be diagnosed after 11 years of age. We also potentially missed children with diagnoses of ASD who attended private or independent schools, schools outside England, or children who were educated at home.
Although most cases had a good diagnostic assessment by a specialist multidisciplinary team, different tools were used by different teams, and the ASD label was applied by clinical consensus and not after a research evaluation. This may have biased our results in either direction: the lack of a rigorously applied research diagnostic scale may have resulted in an overestimate of the true prevalence by including children with autistic traits who do not fulfil strict diagnostic criteria. Alternatively, the reliance on a clinically referred population may have underestimated the true prevalence by not including milder cases on the spectrum that were never referred. For this reason, we included the education data, but these presented their own difficulties such as, most importantly, the lack of detailed diagnostic information. The PLASC database only records a primary and a secondary educational difficulty, so unless an ASD has been identified as one of the two major concerns in a child’s ability to access the national curriculum, it will not be listed. Interestingly, the 17% of cases identified only by educational sources in the present study is small compared with the 40% of ASD cases identified only by educational and not medical sources in a US-based study.12 These results highlight the importance of searching for ASD cases using multiple methods,13 to ensure as complete an ascertainment as possible.
The most secure estimate of ASD prevalence in the ALSPAC cohort of 51 per 10 000 is within the range of estimates from other population-based studies. The wide range of estimates in the literature is likely to be a result of different methods used to find cases, rather than genuine differences in prevalence in different populations.1,14 Chakrabarti and Fombonne15 highlight that four UK surveys on similar paediatric populations reported a fourfold difference in prevalence, which suggests that the rigorousness of the methods used to find cases is likely to have the strongest influence on the result, rather than the characteristics of the study population.
The prevalence estimate for childhood autism reported in the present study (21.6 per 10 000) is high compared with figures quoted in 23 studies before 1999,14 and figures quoted in more recent studies15,16 which ranged from 0.7 to 21.1 cases per 10 000. However, it is similar to the Special Needs and Autism Project,2 which calculated a prevalence of 24.8 per 10 000 for a narrow definition of childhood autism, and 38.9 per 10 000 for a prevalence of childhood autism estimated using a sample weighting procedure.
For atypical autism, the prevalence of 10.8 cases per 10 000 in this study is comparable to that published by Lingam et al.16 of 10.5 per 10 000 cases. The prevalence of Asperger syndrome in the present study of 16.6 cases per 10 000 was two to three times that reported previously.15,17 Identification of Asperger syndrome is often made later in childhood or adolescence, so the point prevalence would be expected to increase with age after 11 years.14,17 In addition, the prevalence rate will depend on which diagnostic classification is used. A recent study of 8-year-olds from Finland reported prevalence rates of Asperger syndrome to range from 16 to 29 per 10 000 with different diagnostic criteria used on the same sample of children.18
In addition, all 15 cases identified only by the PLASC data were labelled as ‘unspecified ASD’ by this study. It is likely that some of these cases belong to a more specific ASD diagnosis. Therefore, although the prevalence of individual ASD diagnoses in the present study appears high, in reality these are likely to be underestimates.
The predominance (87.2%) of male children having a diagnosis of ASD was similar to figures for five recently published UK studies,2,13,15,16,19 which calculated the percentages of males with ASD to be between 79.2 and 87.4%. Although there was no evidence of a difference in age at diagnosis between males and females in ALSPAC, Yeargin-Allsop et al. reported that males were being diagnosed significantly earlier than females (3.6 years and 4.1 years respectively).12
The lack of association of ASD with maternal education is consistent with previous studies,14,19 although Baird et al. found that the rate of identification of ASD by local services was lowest for children of less educated parents.2
In the present study, 98.6% of children with ASD were known to have special educational needs, with 94.4% of children having a full statement of SEN. These figures are in line with those quoted in the Mental Health of Children and Young People in Great Britain report,21 which stated that 97% of children with a diagnosis of ASD have special educational needs.
Comorbid developmental impairments
In this study, 33.8% of ASD cases had an identified comorbid developmental disorder: previous studies14,21,22 have quoted figures of between 9% and 37% for the co-existence of developmental and/or neurological disorders, although Yeargin-Allsop et al.12 reported a much higher figure of 62%. This suggests we achieved relatively good rates of ascertainment of comorbid developmental conditions, despite this research being based on health records rather than systematic assessment of children with a diagnosis of ASD to find possible comorbidities. However, the percentage of children (14.7%) labelled as having learning disability in this study may be an underestimate. An early study23 quoted a prevalence of learning disability in childhood autism to be as high as 80%, although more recently, better diagnostic services have meant that more higher-functioning cases of ASD are now being diagnosed, and the estimates of ASD cases with learning disability now range from 25.8 to 30%.12,16,21,22 One reason why the figure for a diagnosis of learning disability may be low in our study is that researchers only coded this diagnosis if a child had undergone IQ testing and the results were filed in the clinical records. When a child has severe autism and is diagnosed very young, they may go straight into a special school, and their IQ may never be measured. This may explain why our results show a higher proportion of learning disability in atypical autism, rather than classical childhood autism as would be expected from the literature.
Epilepsy is a commonly reported comorbidity in children with ASD. It was found in 16.7% of cases of childhood autism and 8.7% of children with Asperger syndrome, similar to previous estimates of epilepsy in ASD in the literature ranging from 5.6% to 29% of cases.12,17,21,22,24 Although Gillberg and Billstedt4 report that Tourette syndrome and other tic disorders, blindness, deafness, and chromosomal abnormalities also show high rates of comorbidity with ASD, there were too few cases of these disorders found in ALSPAC. We did find hyperactivity to be a relatively common comorbidity, especially in children with a diagnosis of Asperger syndrome (13% of cases). Many authors4,17,21,22,24 have commented that symptoms of attention-deficit–hyperactivity and conduct disorders are over-represented in children with ASD, despite the diagnostic limitations of the classification systems of ICD-10 and DSM-IV. It is also important to recognize that there is significant overlap between different subtypes of ASD and that the developmental difficulties associated with autism may change as a child gets older.
The prevalence of ASD in this representative sample is consistent with recent reports and probably reflects improved ascertainment by multidisciplinary teams in health and education. The identification of cases of ASD in the ALSPAC study will facilitate further studies using this cohort to elucidate the genetic and environmental causes of ASD, and these children will be followed up into adolescence and adult life.
We are grateful to all the families who took part in this study, the midwives for help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. The UK Medical Research Council, the Wellcome Trust, and the University of Bristol provide core support for ALSPAC. This study was supported by a steering group consisting of the authors and Dr Jon Pollock, Ms Sue Bonnell, Ms Karen Birmingham, and Professor Jean Golding. We thank Andy Boyd and Mike Crawford for their help with data management and linkage, and Debbie Johnson for her help with data extraction from notes. This study was funded by the Wellcome Trust, (grant 59579).