The ADDM Network was composed of 11 sites for surveillance year 2016 (Arizona, Arkansas, Colorado, Georgia, Maryland, Minnesota, Missouri, New Jersey, North Carolina, Tennessee, and Wisconsin). Children included in ADDM surveillance year 2016 were born in 2008 and had a parent or guardian who lived in one of 11 surveillance sites in 2016. Each site selected a portion of its state (except Arkansas, which included the entire state) to monitor ASD among children aged 8 years in 2016. All sites functioned as public health authorities under the Health Insurance Portability and Accountability Act of 1996 Privacy Rule and met applicable local institutional review board, privacy, and confidentiality requirements under 45 CFR 46 (24). The racial and ethnic composition of populations in ADDM Network sites is provided (Supplementary Table 1, https://stacks.cdc.gov/view/cdc/85386).
The ADDM Network uses a multiple-source, records-based surveillance methodology developed by CDC's Metropolitan Atlanta Developmental Disabilities Surveillance Program (2,24). The ADDM Network ASD surveillance methodology is a two-phase process that has been described previously (3). In brief, in the first phase, ADDM Network staff review records from medical, education, and service providers (e.g., autism specialty clinics or intervention providers) in the community after requesting records that include various billing codes from the International Classification of Disease, Ninth Revision (ICD-9) or International Classification of Diseases, Tenth Revision (ICD-10) or special education exceptionalities (Supplementary Table 2, https://stacks.cdc.gov/view/cdc/85386). If any record contains an indication of ASD, the child's evaluations and other information (e.g., intelligence quotient [IQ] tests) are abstracted and compiled from all available sources in the community. Although all ADDM Network sites use records from medical and service providers, not all sites have complete access to education records.
In the second phase, an ADDM Network clinician reviews the deidentified, compiled record for each child to determine ASD case status. The ADDM Network ASD case definition is based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), and the process for scoring the features of the surveillance case definition have been described previously (3,25,26). ADDM Network clinicians might assign ASD case status if documented evidence satisfies the behavioral criteria for the ASD case definition, or if the child has an established ASD diagnosis. ADDM Network clinicians might decide a child who otherwise meets ASD surveillance criteria should not be included as a case because of insufficient or conflicting information or if other conditions better account for the child's symptoms. Another clinician performs a secondary review if the first reviewer indicates uncertainty. To monitor interrater reliability, 10% of records were randomly selected for an independent review (ASD case status kappa = 0.89) (Supplementary Table 3, https://stacks.cdc.gov/view/cdc/85386). At most ADDM Network sites, clinicians also applied the previous ASD case definition based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) for at least a portion of the children with abstracted information.
Population denominators were obtained from the National Center for Health Statistics vintage 2018 bridged-race postcensal population estimates for 2016 (27). For study areas comprising subcounty school districts, a standardization process using public school enrollment counts was used to adjust the population estimates (Supplementary Methods, https://stacks.cdc.gov/view/cdc/85386). Each site linked each child to birth certificate information from their state. When successful, this linkage indicates which children were born in the state that they lived in at age 8 years and provides additional demographic information. Information about race and ethnicity came from information abstracted from the medical or education records, which was augmented by data from birth certificates and data from administrative or billing information. Children with race coded as "other" or "multiracial" were excluded from race-specific estimates, as were American Indian/Alaskan Native children because of small numbers.
Age at first developmental evaluation on record was based on each child's abstracted evaluation information and restricted to children born in the state (or ADDM Network surveillance area in Minnesota) where the ADDM Network site is located. Age at first ASD diagnosis was based on the age of a child when an examiner recorded an ASD diagnostic statement or noted the child's age when another provider previously diagnosed ASD. Intellectual disability status was based on IQ scores ≤70 on a child's most recent test available through 2016. A child without an IQ score also could be classified as having intellectual disability on the basis of an examiner's statement of intellectual disability in a developmental evaluation. Children were considered to have community-identified ASD if their records contained any of the following: 1) a diagnostic statement from a qualified professional of autistic disorder, pervasive developmental disorder not otherwise specified (PDD-NOS), Asperger disorder, or ASD; 2) any ASD ICD billing code at any time from birth through 2016; or 3) receipt of (or met eligibility for) special education services under the autism classification in public school.
Prevalence was calculated as the number of children with ASD per 1,000 children aged 8 years in the defined population or subgroup. Overall prevalence estimates included all children identified with ASD. Results for the combined (overall) total include data from all sites unless otherwise noted. Ninety-five percent confidence intervals (CIs) for prevalence, proportions, and prevalence ratios were calculated using the Wilson score method. Pearson chi-square tests were performed for comparison of proportions, and the Mantel-Haenszel (Woolf) test of homogeneity was used to compare prevalence ratios across sites. Permutation tests were conducted to test differences in medians. Statistical tests with p values <0.05 were considered statistically significant, as were 95% CIs that excluded 1.0 for prevalence ratios. Cumulative incidence of ASD diagnoses was calculated as the total children with ASD diagnosed during or before a given month of age, divided by the total population of children aged 8 years in the surveillance area. R software (version 3.5.3; R Foundation) and additional packages were used to conduct analyses. Additional information about the statistical software is available (Supplementary Table 4, https://stacks.cdc.gov/view/cdc/85386).