[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Views 455
Citations 0
Comment & Response
July 1, 2019

Caution About Displaying State-Level Differences in the Prevalence of Autism Spectrum Disorder

Author Affiliations
  • 1Health Resources and Services Administration, Maternal and Child Health Bureau, Rockville, Maryland
  • 2Harvard Medical School, Harvard University, Boston, Massachusetts
JAMA Pediatr. 2019;173(9):887-888. doi:10.1001/jamapediatrics.2019.1692

To the Editor The article by Xu et al1 on the prevalence and treatment patterns of autism spectrum disorder (ASD) in the United States displayed a Figure on the state-level weighted prevalence of ever-diagnosed ASD using data from the 2016 National Survey of Children’s Health (NSCH). We have serious concerns about this Figure. The information is either incomplete or misleading for a number of reasons. First, the Figure (and attendant analyses) is not described in the Methods section and provides no rationale for the selection of cut points to create the category boundaries. Second, the state-level map covers parent report of ever diagnosed rather than the more conservative parent report of “currently have the condition”; the reasons 7% of the children’s parents reported ever diagnosed but did not report current condition are unknown and could include a preliminary diagnosis of ASD that was then ruled out.2 Third, the Results section provides prevalence estimates and 95% confidence intervals for only the 2 states having the highest and lowest prevalence of ASD. This obscures the fact that while the overall sample for the NSCH is large (N = 50 212), the numbers within each state for a low prevalence condition, such as ASD, can be small, often resulting in wide confidence intervals and few statistically significant differences between other states. In fact, publicly available data on this topic indicate that nearly half of the ASD state-level estimates using only 2016 data are statistically unreliable3,4 because the confidence interval width exceeded 120% of the estimate. Fourth, the Discussion section mentioned that ASD prevalence varied appreciably by state yet did not discuss limitations given the unreliability of state-level estimates using small numbers nor was there an acknowledgment that the observed state-level differences mainly reflect artifacts of the small numbers.