Skip Navigation
  Print Page


Skip sharing on social media links
Share this:

E.J. Costello

Epidemiology is the branch of medical science that deals with the incidence, distribution, and control of disease in a population (Webster’s). Prevalence refers to the rate of the disorder present in the population at a given point in time. Incidence refers to the number of new cases occurring in the population during a given period.

Response to NIH Questions

Question 1: What is the best, empirically substantiated estimate of the prevalence of autism spectrum disorders in children and adults in the United States, and in other countries?

There are no prevalence estimates specifically for the United States, but recent studies from Canada and from Japan indicate that autism is not a rare disorder. Both studies found prevalence rates of autism greater than 10 per 10,000. However, these studies used fairly small (<100,000) samples, and the confidence limits are wide (±5 per 10,000). A rate of at least 22 per 10,000 has been estimated for the broader autism spectrum disorders. Because of similarities between the United States and Canada, the Canadian data are likely to be reasonable estimates of prevalence in the United States for most purposes. Given the available data, there is little justification for the potential costs of a national prevalence study of autism in the United States merely to estimate the prevalence of autism spectrum disorders. Money would be better spent on developmental risk-assessment or cost-benefit studies.

Question 2: What efforts are currently underway to improve estimates of prevalence of autism in children? How can these studies as planned, or with modification, answer the Congressional question regarding prevalence in the United States?

There are a few studies of the prevalence of childhood disorders underway or in the planning stage. It is possible that autism could be included in one or more of these studies. The difficulties of doing so are formidable.

There are two main approaches to obtain prevalence estimates of autism: population-based studies (finding persons with autism in the whole number of people in a country or region), or studies of treated populations (finding the number of persons with autism in the populations of treatment settings such as hospitals, clinics, special education settings). Population-based studies provide unbiased samples since everyone is potentially included, but require very large samples to identify reasonable estimates of disorders like autism (e.g., 500-1,000 households would have to be surveyed to identify 1 child with autism). It is estimated that a sample of over 100,000 children would be needed to produce reliable estimates of the prevalence of autism with greater precision than is currently available in the international epidemiologic literature. Children with autism could be counted in smaller treatment-based studies since they are usually referred for services. However, problems include the following: (a) it is difficult to be certain you have found all treatment settings; (b) not all settings will be willing to participate; and (c) well-functioning children will be underestimated since they may not be served in the special treatment settings. The NIMH has recently funded a network of university sites to conduct epidemiologic research on childhood disorders. This network of centers, called UNOC-CAP will study the Use, Need, Outcomes, and Cost of Child and Adolescent Populations. Both population-based and treated-sample studies will be carried out in UNOC-CAP and specific diagnostic measures will be included. These sites are already funded and procedures specific to the screening and diagnosis of complex disorders in children have been developed.

The National Health Interview Survey is too small to yield valid national estimates of autism. It also uses lay interviewers who may have difficulty in administering a respondent-based interview to identify a disorder like autism. The only U.S. study other than UNOC-CAP known to use diagnostic measures, the Project on Human Development in Chicago neighborhoods, is also too small and geographically too localized. Other ongoing national surveys are too small, do not include the relevant information, or could not be adapted easily to identify autism reliably.

Question 3: What contributions can epidemiologic research make to understanding the etiology, and/or treatment of autism spectrum disorders?

Knowing how many and where cases of a disorder occur in the population ( descriptive epidemiology) is useful for assessing: (a) the number of individuals and families affected by a problem; (b) the size of the financial costs to be expected; (c) the relative cost burden to families, states, and the federal government, or to different service agencies (e.g., education, health, child welfare, juvenile justice); (d) the distribution of the cost and need for services in various geographic, ethnic, or socioeconomic groups; (e) the rise and fall in rates of the disorder over time, and, potentially, the impact of new social policies and treatments on prevalence and outcomes.

Some of the most powerful uses of epidemiology in medicine are as an analytic methodology, that is, as a way of testing hypotheses about causes of disease and the consequences of prevention or treatment strategies. The working group believes that analytic epidemiology can make important contributions to improving understanding of etiology, diagnosis, and treatment.


Genetic epidemiology has already shown its importance for understanding etiology, and this importance will grow, not only as more genes are identified but as the functional roles of those genes are understood developmentally. Preventive interventions have to be based on etiologic theory; thus every intervention study is an implicit test of theory. Descriptive epidemiology can provide the basic hypotheses for interventions, as well as the methodology for testing the theory.


The process of turning a taxonomy such as DSM-IV into instruments for epidemiologic studies helps to tighten the diagnostic criteria, making them more reliable for both epidemiologic and clinical purposes. The process of developing screening instruments helps to refine diagnosis by identifying the core symptoms and the range of variability of the diagnosis. Developmental epidemiology helps to track the developmental sequencing of patterns of symptoms and the impact of symptoms at one time on functioning at a later stage. Longitudinal studies also help to identify the development of compensatory strategies to cope with earlier symptoms.


In an epidemiologic context, every treatment is an experiment, testing the validity of a causal theory. Clinical epidemiology, a highly developed aspect of research in many areas of medicine, has hardly gained a foothold yet in psychiatry or child development, but can provide a framework for comparing the cost-effectiveness of various treatment approaches, and examining the outcome of treatment trials for what they say about the causes of disorder and functional impairment.

Recommendations of the Working Group on Epidemiology

  1. The Canadian data on prevalence are adequate for most U.S. uses. Rather than funding a national prevalence study of autism, autism should be included as one of the childhood disorders in the screening stage of the NIMH UNOC-CAP studies. A follow-up of all potential cases could then be done through UNOC-CAP with a more intensive evaluation, perhaps using experienced clinicians and one of the standard assessment packages currently in use in autism research. The UNOC-CAP population sample will not exceed 20,000 in all, but can establish rates of autism spectrum disorders for the general population. It will be too small to produce reliable rates for minority populations or to allow comparisons of stable rates for different age groups. The treated samples will not exceed 4,000, and are likely to yield localized rather than national estimates. However, the data from UNOC-CAP would provide significant, cost-effective additions to current U.S. information on autism, particularly since prevalence data will be collected in the context of service use and need, cost, and treatment outcome.
  2. Research should be implemented to address the following issues: (a) Variations in the longitudinal course of autism, from early childhood into adulthood: Why do some children do well and some poorly? (b) "Boundary conditions" around autism: What is the rate of strictly defined autism relative to the rates of other types of pervasive developmental disorders, and learning disabilities? (c) Patterns of autism-like deficits in families of children or adults with autism: What is the prevalence of these problems? (d) At what age do children who will develop autism become identifiable? (e) Sociodemographic correlates of autism: What are they? (f) What other disorders (e.g., seizures, depression) may occur during different developmental stages and in which subgroups in autism? (g) Costs associated with the appropriate treatment of children with autism: What are the costs associated with different types of lifetime support of persons with autism and their families? The working group believes that these issues can best be addressed in developmentally focused, longitudinal epidemiologic studies that follow families over time. Such studies need to include not only the child but also the family.
  3. There was support in the working group for a national autism registry. Registries have proved to be invaluable tools in clinical epidemiology. Such registries have been very useful in some other branches of medicine (e.g., tumor registries, birth defect registries).

first | previous | next | last

Last Updated Date: 08/15/2006
Last Reviewed Date: 08/15/2006
Vision National Institutes of Health Home BOND National Institues of Health Home Home Storz Lab: Section on Environmental Gene Regulation Home Machner Lab: Unit on Microbial Pathogenesis Home Division of Intramural Population Health Research Home Bonifacino Lab: Section on Intracellular Protein Trafficking Home Lilly Lab: Section on Gamete Development Home Lippincott-Schwartz Lab: Section on Organelle Biology