Assessment of Racial and Ethnic Bias in Autism Spectrum Disorder Prevalence Estimates from a U.S. Surveillance System

Poster Presentation
Saturday, May 12, 2018: 11:30 AM-1:30 PM
Hall Grote Zaal (de Doelen ICC Rotterdam)
P. Imm1, T. C. White2 and M. S. Durkin3, (1)University of Wisconsin-Madison, Madison, WI, (2)Center for Health and Environmental Data, Colorado Department of Public Health and Environment, Denver, CO, (3)Population Health Sciences, University of Wisconsin School of Medicine and Public Health, Madison, WI
Background: The Autism and other Developmental Disabilities Monitoring (ADDM) Network is a multiple-source, population-based, active surveillance system for monitoring Autism Spectrum Disorder (ASD) in the U.S. Children with ASD are included in the surveillance system only if their residence in specific geographic areas can be confirmed in a given surveillance year. Since 2000, the Network has consistently reported disparities in ASD prevalence by race and ethnicity, with the ASD prevalence higher among non-Hispanic white relative to both non-Hispanic black and Hispanic children.

Objectives: The purpose of this study was to assess potential under-ascertainment of ASD in Hispanic and black children due to differential missing information in the surveillance system on residency and race/ethnicity. Two hypotheses tested were: 1) relative to children included in prevalence estimates, those excluded based on inability to confirm residence within the surveillance area were more likely to be black or Hispanic; 2) imputation of missing information on residency and race/ethnicity will result in less racial and ethnic disparity in ASD prevalence than when prevalence estimation is restricted to cases with complete information.

Methods: The Colorado and Wisconsin ADDM Network sites reviewed combined records from Surveillance Years 2012 and 2014. For hypothesis 1, we conducted a case-control analysis to determine whether race/ethnicity of children excluded due to missing residency information (and determined likely to be classified as ASD cases based on a confirmed diagnosis or evaluation/treatment at an Autism Clinic) differed from ASD cases included in the surveillance system. For hypothesis 2, we evaluated the impact on racial and ethnic disparities in ASD prevalence of various approaches to imputation of missing information, such as race/ethnicity imputation based on notes in records, surname data, and/or census block demographics.

Results: Compared to randomly selected ASD cases included in the surveillance system (N=81), those excluded due to missing residency (N=27) were significantly less likely to be white non-Hispanic (48% vs 69%; p < 0.05) and more likely to be Hispanic (44% vs 15%; p < 0.01). Additionally, inclusion of children with unconfirmed residency and imputation of race/ethnicity information for confirmed cases resulted in slight increases in ASD prevalence overall (from 12.4 to 12.6), but did not affect the ratio of ASD prevalence in white non-Hispanic versus other groups. For example, the ratio of prevalence in white compared to black children was 1.5 (95% CI: 1.25-1.73) in the complete case-only analysis and was unchanged after inclusion of cases with missing residency confirmation and imputed race/ethnicity.

Conclusions: Although potential ASD cases excluded from the surveillance system due to missing residency information were significantly more likely to be from under-represented racial/ethnic groups than the included cases, the strength of this bias and the number of excluded cases were insufficient to account for the observed racial and ethnic disparities in ASD prevalence. This evaluation provides evidence of the robustness of the surveillance system while suggesting the need for continued research into the disparities in ASD prevalence.

See more of: Epidemiology
See more of: Epidemiology