Autism spectrum disorders (ASD) prevalence estimates derived from a single data source under-identify children and provide a biased profile of case characteristics. We analyzed characteristics of 1,919 children with ASD identified by the Autism and Developmental Disabilities Monitoring Network. Cases ascertained only at education sources were compared to those identified at health sources. 38% were education-only. These were older at their earliest evaluation (54.5 vs. 42.0 months, p < 0.001) and earliest ASD diagnosis (62 vs. 53 months, p < 0.001). More lived in census blocks with lower adult education (p < 0.001). Lower educational attainment of adults in census blocks of residence of education-only cases suggests disparities in access to clinical services with the schools providing crucial services to many families.