Epidemiology of autism


The epidemiology of autism is the study of the incidence and distribution of autism spectrum disorders. A 2022 systematic review of global prevalence of autism spectrum disorders found a median prevalence of 1% in children in studies published from 2012 to 2021, with a trend of increasing prevalence over time. However, the study's 1% figure may reflect an underestimate of prevalence in low- and middle-income countries.
Socioeconomic barriers also affect access to treatment. Due to the high cost of individualized therapies such as applied behavior analysis, speech therapy, and occupational therapy; approximately 36% of children with ASD face difficulty affording care or remain untreated.
ASD averages a 4.3:1 male-to-female ratio in diagnosis, not accounting for ASD in gender diverse populations, which overlap disproportionately with ASD populations. The number of children known to have autism has increased dramatically since the 1980s, at least partly due to changes in diagnostic practice; it is unclear whether prevalence has actually increased; and as-yet-unidentified environmental risk factors cannot be ruled out. In 2020, the Centers for Disease Control and Prevention's Autism and Developmental Disabilities Monitoring Network reported that approximately 1 in 54 children in the United States are diagnosed with an autism spectrum disorder, based on data collected in 2016. This estimate is a 10% increase from the 1 in 59 rate in 2014, 105% increase from the 1 in 110 rate in 2006 and 176% increase from the 1 in 150 rate in 2000. Diagnostic criteria of ASD has changed significantly since the 1980s; for example, U.S. special-education autism classification was introduced in 1994.
ASD is a complex neurodevelopmental disorder, and although what causes it is still not entirely known, efforts have been made to outline causative mechanisms and how they give rise to the disorder. The risk of developing autism is increased in the presence of various prenatal factors, including advanced paternal age and diabetes in the mother during pregnancy. In rare cases, autism is strongly associated with agents that cause birth defects. It has been shown to be related to genetic disorders and with epilepsy. ASD is believed to be largely inherited, although the genetics of ASD are complex and it is unclear which genes are responsible. ASD is also associated with several intellectual or emotional gifts, which has led to a variety of hypotheses from within evolutionary psychiatry that autistic traits have played a beneficial role over human evolutionary history.

Frequency

Different measures can be used to examine the frequency of autism, including incidence rates, point or period prevalence, or cumulative incidence.

Incidence and prevalence

defines several measures of the frequency of occurrence of a disease or condition:
  • The incidence rate of a condition is the rate at which new cases occurred per person-year, for example, "2 new cases per 1,000 person-years".
  • The cumulative incidence is the proportion of a population that became new cases within a specified time period, for example, "1.5 per 1,000 people became new cases during 2006".
  • The point prevalence of a condition is the proportion of a population that had the condition at a single point in time, for example, "10 cases per 1,000 people at the start of 2006".
  • The period prevalence is the proportion that had the condition at any time within a stated period, for example, "15 per 1,000 people had cases during 2006".
When studying how conditions are caused, incidence rates are the most relevant measure of condition frequency as they quantify new cases. However, incidence can be difficult to measure with rarer conditions such as autism. In autism epidemiology, incidence can be difficult to quantify because the condition starts long before it is diagnosed, possibly as early as conception for strong genetic causes. The gap between initiation and diagnosis is influenced by many factors unrelated to chance. Research focuses mostly on whether point or period prevalence is increasing with time; cumulative incidence is sometimes used in studies of birth cohorts.

Estimation methods

The three basic approaches used to estimate prevalence differ in cost and in quality of results. The simplest and cheapest method is to count known autism cases from sources such as schools and clinics, and divide by the population. This approach is likely to underestimate prevalence because it does not count children who have not been diagnosed yet, and it is likely to generate skewed statistics because some children have better access to treatment.
The second method improves on the first by having investigators examine student or patient records looking for probable cases, to catch cases that have not been identified yet. The third method, which is arguably the best, screens a large sample of an entire community to identify possible cases, and then evaluates each possible case in more detail with standard diagnostic procedures. This last method typically produces the most reliable, and the highest, prevalence estimates.

Frequency estimates

Estimates of the prevalence of autism vary widely depending on diagnostic criteria, age of children screened, and geographical location. Most recent reviews tend to estimate a prevalence of 1–2 per 1,000 for classic autism and close to 27.6 per 1,000 for ASD;
PDD-NOS is the vast majority of ASD, Asperger syndrome is about 0.3 per 1,000 and the atypical forms childhood disintegrative disorder and Rett syndrome are much rarer.
A 2006 study of nearly 57,000 British nine- and ten-year-olds reported a prevalence of 3.89 per 1,000 for autism and 11.61 per 1,000 for ASD; these higher figures could be associated with broadening diagnostic criteria. Studies based on more detailed information, such as direct observation rather than examination of medical records, identify higher prevalence; this suggests that published figures may underestimate ASD's true prevalence. A 2009 study of the children in Cambridgeshire, England used different methods to measure prevalence, and estimated that 40% of ASD cases go undiagnosed, with the two least-biased estimates of true prevalence being 11.3 and 15.7 per 1,000.
A 2009 U.S. study based on 2006 data estimated the prevalence of ASD in eight-year-old children to be 9.0 per 1,000. A 2009 report based on the 2007 Adult Psychiatric Morbidity Survey by the National Health Service determined that the prevalence of ASD in adults was approximately 1% of the population, with a higher prevalence in males and no significant variation between age groups; these results suggest that prevalence of ASD among adults is similar to that in children and rates of autism are not increasing.

Increased diagnoses over time

Attention has been focused on whether the prevalence of autism is increasing with time. Earlier prevalence estimates were lower, centering at about 0.5 per 1,000 for autism during the 1960s and 1970s and about 1 per 1,000 in the 1980s, as opposed to today's 23 per 1000.
The number of reported cases of autism increased dramatically in the 1990s and 2000s, prompting ongoing investigations into two main potential reasons:
  • More children may have autism; that is, the true frequency of autism may have increased.
  • The apparent increase may be illusory, caused by some form of measurement bias such as improved diagnosis or widened diagnostic criteria.

    Increased diagnoses as measurement bias

Many explanations suggest that the increased diagnosis of autism might be caused by changes in the definition of the condition, improved methods of diagnosis, decreased stigma, and widespread mainstream public awareness of the condition.
  • There may be more complete pickup of autism, as a result of increased awareness and funding. For example, attempts to sue vaccine companies may have increased case-reporting.
  • The diagnosis may be applied more broadly than before, as a result of the changing definition of the disorder, particularly changes in DSM-III-R and DSM-IV.
  • An editorial error in the description of the PDD-NOS category of Autism Spectrum Disorders in the DSM-IV, in 1994, inappropriately broadened the PDD-NOS construct. The error was corrected in the DSM-IV-TR, in 2000, reversing the PDD-NOS construct back to the more restrictive diagnostic criteria requirements from the DSM-III-R.
  • Successively earlier diagnosis in each succeeding cohort of children, including recognition in nursery, may have affected apparent prevalence but not incidence.
  • A review of the "rising autism" figures compared to other disabilities in schools shows a corresponding drop in findings of intellectual disability.
The reported increase is largely attributable to changes in diagnostic practices, referral patterns, availability of services, age at diagnosis, and public awareness. A widely cited 2002 pilot study concluded that the observed increase in autism in California cannot be explained by changes in diagnostic criteria, but a 2006 analysis found that special education data poorly measured prevalence because so many cases were undiagnosed, and that the 1994–2003 U.S. increase was associated with declines in other diagnostic categories, indicating that diagnostic substitution had occurred.
A small 2008 study found that a significant number of people diagnosed with pragmatic language impairment as children in previous decades would now be given a diagnosis as autism. A study of all Danish children born in 1994–99 found that children born later were more likely to be diagnosed at a younger age, supporting the argument that apparent increases in autism prevalence were at least partly due to decreases in the age of diagnosis.
A 2007 study that modeled autism incidence found that broadened diagnostic criteria, diagnosis at a younger age, and improved efficiency of case ascertainment, can produce an increase in autism diagnoses ranging up to 29-fold depending on the frequency measure, suggesting that methodological factors may explain what appears to be an increase in autism over time. This observation is often mistakenly described as an "autism epidemic".