Is the prevalence of autism increasing in the United States?

BMJ 2014; 348 doi: http://dx.doi.org/10.1136/bmj.g3088 (Published 09 May 2014) Cite this as: BMJ 2014;348:g3088
  1. Stephanie Blenner, assistant professor of pediatrics,
  2. Marilyn Augustyn, professor of pediatrics
  1. 1Division of Developmental Behavioral Pediatrics, Boston University School of Medicine, Boston, MA 02118, USA
  1. Stephanie.Blenner{at}bmc.org

A consistent approach to identification and studies in a stable representative population are needed to decide

Few reports from the Centers for Disease Control and Prevention (CDC) garner as much media attention as the periodic updates from the United States Autism and Developmental Disabilities Monitoring (ADDM) network.1 Perhaps this reflects the reported increases in the prevalence of autism, which are worrying to all those who care for affected children and their families.

Considering the ADDM’s findings, an understanding of its methodology is important. This multistate autism surveillance network was established in 2000 by congressional mandate and the CDC. The study sample comes from states selected through a competitive grant, so participating states vary over time, and it is not representative of the US population. The most recent report from March 2014 is based on data from 11 sites in the 2010 surveillance year.2 Data derive from a two phase approach to identifying 8 year olds with autism spectrum disorder. The first phase involves abstraction of medical and educational (if available) evaluations. The second phase is identification of children who on review of the abstracted records by trained clinicians meet criteria for any of the following DSM-IV-TR (Diagnostic and Statistical Manual of Mental Disorders, fourth edition, revised) diagnoses: autistic disorder, pervasive developmental disorder not otherwise specified, or Asperger’s disorder. The recent ADDM report does not evaluate prevalence using DSM-5 criteria,3 although this is discussed for future reports.

The study documents several noteworthy findings, most importantly a prevalence of 14.7 per 1000 or one in 68 children in the network identified as having symptoms consistent with an autism spectrum disorder. This continues a trend of increasing prevalence from the initial 2000 data (6.7/1000), through 2006 (9.0/1000), and 2008 (11.3/1000). In keeping with previous studies of autism spectrum disorder, boys were more likely to be identified, with a total male to female prevalence ratio of 4.5. Importantly, the median age of diagnosis is unchanged from previous years, at 4.5 years. The proportion of children with the disorder and average to above average intelligence increased to 46%; seven of the 11 sites had intelligence data on at least 70% of their cohort, which qualified them for inclusion in this analysis. Combining data from these seven sites, the prevalence of co-occurring intellectual disability autism spectrum disorder was 4.7 per 1000 children, whereas prevalence without intellectual disability was 10.2 per 1000. Racial differences in this co-occurrence were striking, with 48% of black children with the disorder classified in the range of intellectual disability compared with 38% of Hispanic and 25% of white children.

These data have several important limitations. Although often quoted as such, the measured prevalence is not intended for use as actual prevalence of autism spectrum disorder in the US. Prevalence varies greatly by state, ranging from one in 175 in Alabama to one in 45 in New Jersey. This relates to whether educational records were available (5/11 states did not have full access to records), as well as differences between medical records and special education eligibility reports, and differences in availability of specialty medical evaluation. Special education reports are not designed for medical diagnosis or intended as public health measures, and both special education and specialty medical care vary by geographic region regarding children assessed, assessors’ backgrounds, and what constitutes assessment. This also may contribute to differential ascertainment of children who have not been formally evaluated.

So what might this latest report mean for those who diagnose and care for children with autism spectrum disorder?

The main message is the continued late age of evaluation and diagnosis. Only 44% of these children had a comprehensive evaluation by age 3 years and 36% did not have one until after 4 years. This is despite documentation of developmental concerns in abstracted records by 36 months of age for most of them. Furthermore, late diagnosis continues, with the median age of first known diagnosis of autism spectrum disorder at 52 months for white and black children and 56 months for Hispanic children. This is well beyond the age that many children can be identified and would benefit from intervention. With the marked attention in the medical and popular media, why has so little progress been made? Despite substantial public focus and research, these children still wait far too long to receive a diagnosis.

A second important point to glean from the report is the increasing number of children identified with average to above average cognitive abilities. Current intervention and social service programs often focus on those with intellectual disability. Our current systems of care are challenged to provide support to those with social and adaptive needs but higher cognitive functioning, particularly as they move into adolescence and adulthood.

The final point lies with the overall prevalence data over time. This report highlights that the prevalence of this disorder continues to rise but change is perhaps the most solid conclusion to be drawn. While each report identifies increasing numbers of children, the network design and study methods do not answer the question of by how much autism spectrum disorder is actually increasing in the population. The term “epidemic” continues to resurface with each of these reports. But is autism spectrum disorder truly crossing the “epidemic” threshold or are we just counting children differently? This report is not designed to answer that question and trying to force it to do so does not recognize the inherent limitations of the dataset.

However, these data do support the need to identify children early and to devise ways to target underserved populations and children both with and without concomitant intellectual disability. A better understanding of the underpinnings of this disorder, which still needs a stable definition, and ongoing prevalence studies that use a consistent representative population are also required.


Cite this as: BMJ 2014;348:g3088


  • Competing interests: We have read and understood BMJ policy on declaration of interests and declare the following interests: None.

  • Provenance and peer review: Commissioned; not externally peer reviewed.