Share this article on:

Estimating the Incidence of Autism

Hertz-Picciotto, Irva; Delwiche, Lora

doi: 10.1097/EDE.0b013e3181a81ef9
Letters to the Editor

University of California, Davis; Davis, CA;

Back to Top | Article Outline

The authors respond:

e appreciate the concerns raised by Grinker and Leventhal1 regarding limitations of the California Department of Developmental Services (DDS) databases. Indeed, nearly all these issues are discussed in our article2—some in considerable detail (underestimation of total autism or autism spectrum disorders incidence, inclusion of milder cases over time).

However, several of their statements are incorrect. They allege that we “conclude there is a true increase that cannot be accounted for...” and they call for “temper[ing] overly ambitious conclusions.” We stand by our original conclusion: “the possibility of a true increase in incidence deserves serious consideration.”

Grinker and Leventhal incorrectly claim that we used quarterly DDS caseload data. We pointed out “caseload statistics do not provide valid information about individuals newly diagnosed.” To calculate measures of incidence, DDS provided us with files of individual cases and we obtained state birth records and US census figures for denominators.

They criticize our use of DDS Early Start Reports on children younger than 36 months of age, correctly noting that, in general, diagnoses at younger ages can be unstable. Because of our concern over the same issue, we checked and determined that 87% of these early designations were later confirmed when the child was 36 months or older,2 possibly reflecting cautious use of this diagnosis on Early Start Reports by Regional Center staff. Exclusion of the 1.5% diagnoses found only on an Early Start Reports, among all 30,832 autism cases, did not alter the study's results.

Another argument is that we placed “undue confidence on age of first appearance in the DDS dataset as a proxy for date of onset.” Recognizing this problem, we had initially used the term “pseudo-incidence,” but the reviewers pointed out that use of date of first contact with a provider as a proxy is standard practice and asked us to drop such language. Our concern that age at first appearance is not necessarily related to age of onset (which by DSM IV criteria occurs before 3 years of age) was in fact the basis for our analyzing “age at diagnosis” as an artifact.

The use of administrative data for epidemiologic purposes is certainly debatable. Hospital discharge data, vital records, information collected by health maintenance organizations, etc. are routinely used by epidemiologists in health services research. Responsible scientists consider weaknesses in their data sources, particularly how those might have affected the results, which we believe we have done. Grinker and Leventhal raise excellent questions about the California DDS system, and we invite them and others to carry out investigations that could provide answers.

Irva Hertz-Picciotto

Lora Delwiche

University of California

Davis Davis, CA

Back to Top | Article Outline


1. Grinker RR, Leventhal BL. Estimating the incidence of autism. [Letter] Epidemiology. 2009; 20:620–621.
2. Hertz-Picciotto I, Delwiche LD. The rise of autism and the role of age at diagnosis. Epidemiology. 2009;20:84–90.
© 2009 Lippincott Williams & Wilkins, Inc.