Validation of the Utah Registry of Autism and Developmental Disabilities: A State-Wide Registry for Autism Spectrum Disorder

Background: State-based autism spectrum disorder (ASD) registries are becoming more common in the U.S. and provide an important public health surveillance service. Eight state-level ASD registries currently exist including the Utah Registry of Autism and Developmental Disabilities (URADD). URADD, established in 2002, tracks changes in ASD prevalence in Utah, investigates ASD epidemiology, informs public policy, and supports ASD-related service planning. The multiple uses of ASD registry data necessitate validation of information acquired by these registries, especially ASD case status.

Objectives: 1) Validate URADD’s ascertainment of children with ASD against the Utah Autism and Developmental Disabilities Monitoring (UT-ADDM) project’s ascertainment of children with ASD. 2) Determine the impact of imposing a more stringent definition of ASD on URADD prevalence estimates, sensitivity, and specificity.

Methods: URADD uses two ascertainment methods. The first is a passive surveillance system in which individuals with ASD are identified by birth cohort based on a community medical ASD diagnosis and/or autism special education eligibility (URADD method). Second, URADD previously participated in the CDC’s ADDM Network in surveillance years (SY) 2002, 2008, 2010, and 2012. ADDM uses a retrospective record review approach to identify children who meet ASD case definition based on DSM-IV-TR criteria (UT-ADDM method). For this study, eight-year-old children identified with ASD, residing in the three county UT-ADDM surveillance area in SY2010 or SY2012 were categorized by ascertainment method: URADD, UT-ADDM, or both. ASD prevalence was estimated by ascertainment method and compared using a chi-square test. The sensitivity and specificity of URADD ASD case status was estimated using UT-ADDM as the gold standard. A more stringent URADD definition was subsequently imposed by limiting ascertainment to children with ≥ 2 medical ASD diagnoses and/or an autism special education eligibility. URADD-based ASD prevalence, sensitivity, and specificity analyses were subsequently repeated.

Results:  Collectively, 872 and 1141 eight-year-old children were identified with ASD by UT-ADDM and URADD, respectively for SYs 2010/2012. 74% of eight-year-old children identified with ASD in SYs 2010/2010 were ascertained by both UT-ADDM and URADD (N=647). URADD’s ASD prevalence estimate (23/1,000 eight-year-old children (95% Confidence Interval (CI)): 22.1-24.8) was statistically significantly higher than UT-ADDM’s (18/1,000 eight-year-old children (95% CI: 16.8-19.1); p-value < 0.0001). Sensitivity and specificity of URADD case ascertainment was 81.7% and 96.4%, respectively. After removing the 263 cases ascertained by URADD based on a single medical diagnosis, URADD’s prevalence was reduced to 14.0/1,000 (95% CI: 12.9-14.9) eight-year-old children, and sensitivity and specificity of URADD versus UT-ADDM changed to 58.5% and 98.6%, respectively.

Conclusions:  URADD’s high sensitivity and specificity compared with UT-ADDM validates URADD’s community-based approach to ASD case ascertainment. URADD’s approach results in higher ASD prevalence estimates than the UT-ADDM method suggesting that the URADD approach may be more susceptible to identifying persons who are false-positive for ASD. However, imposing a stricter definition of ASD on URADD resulted in a marked decrease in sensitivity and only a minimal increase in specificity. From a public health perspective, URADD’s capacity to conduct complete population-wide ASD ascertainment outweighs the risk of mis-classifying a small proportion of unaffected persons.