The Spatial Structure of Autism Spectrum Disorders In Utah

Background:   The existence of environmental risk factors for Autism Spectrum Disorders (ASDs) has been a controversial topic.  One step toward possible discovery is a spatial analysis.  A non-random distribution of Autism Spectrum Disorders (ASDs) in Utah could reflect an underlying spatial structure in either the risk factors associated with ASD or in the services providing ASD education, awareness, and treatment. Spatial analysis of ASD patterns may lead to improvements in the development of hypotheses concerning ASD etiology or in the planning of ASD services.

Objectives:   The objective of this study was to determine the presence, location and size of ASD clusters in a population of 8 year old children residing in a three county surveillance region in Utah.  In addition, we investigated if the presence, location and size of ASD clusters change after controlling for confounding sociodemographic risk factors associated with ASD.

Methods:   ASD  (n = 132) was identified in children born in 1994 and residing in a three county surveillance region at age 8 by the Utah Registry of Autism and Developmental Disabilities using the methods developed by the Metropolitan Atlanta Developmental Disabilities Surveillance program.  Control population (n = 21,935) included children from the 1994 birth cohort born in the three county surveillance region lacking a developmental disability. Maternal residential birth addresses of cases and controls were geocoded as latitude and longitude point locations.  Clusters were detected using the scan statistic assuming a Bernoulli model as implemented by SaTScan software (Kulldorf 1997).  Following the initial analysis, clusters were adjusted for sociodemographic factors including mother’s age at birth and mother’s race.

Results:   One primary cluster was identified in the three county surveillance region.  The primary cluster included 87 ASD cases, and the relative risk of ASD inside the primary cluster was 2.17. The secondary cluster contained 3 ASD cases and its relative risk was 15.07.  The size of the primary cluster was reduced and the relative risk (RR = 35) increased considerably after adjusting for confounding sociodemographic factors.  Controlling for confounding factors produced multiple secondary clusters that were not present before adjusting for covariates.

Conclusions:   ASD clusters exist in the 1994 Utah dataset and controlling for sociodemographic risk factors altered the size and shape of ASD clusters. The persistence of these clusters across multiple surveillance years will test for further support that these clusters are real.  Further investigation with statistical models may elucidate potential risk or sampling factors responsible for ASD clusters.