Journal of Environmental and Public Health
Volume 2011 (2011), Article ID 202783, 5 pages
http://dx.doi.org/10.1155/2011/202783
Administrative Censoring in Ecological Analyses of Autism and a Bayesian Solution
1Program in Public Health and Department of Statistics, University of California, Irvine, 2241 Bren Hall, Irvine, CA 92697-1250, USA
2Gradient, Seattle, WA 98101-1248, USA
3Department of Health and Nutrition Sciences, Brooklyn College, Brooklyn, NY 11210, USA
Received 22 September 2010; Accepted 2 March 2011
Academic Editor: Pam R. Factor-Litvak
Copyright © 2011 Scott M. Bartell and Thomas A. Lewandowski. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
Widely cited ecological analyses of autism have reported associations with mercury emissions, with precipitation, and race at the level of counties or school districts. However, state educational agencies often suppress any low numerical autism counts before releasing data—a phenomenon known as “administrative censoring.” Previous analyses did not describe appropriate methods for censored data analysis; common substitution or exclusion methods are known to introduce bias and produce artificially narrow confidence intervals. We apply a Bayesian censored random effects Poisson model to reanalyze associations between 2001 Toxic Release Inventory reported mercury emissions and 2000-2001 autism counts in Texas. Relative risk estimates for autism decreased from 4.44 (95% CI: 4.16, 4.74) per thousand lbs. of air mercury emissions using a naive zero-substitution approach to 1.42 (95% CI: 1.09, 1.78) using the Bayesian approach. Inadequate attention to censoring poses a serious threat to the validity of ecological analyses of autism and other health outcomes.