Screening for autism spectrum disorder (ASD) during well-child pediatric check-ups reduces the age of diagnosis, allowing more time for critical ASD-specific early intervention (EI), which greatly improves outcomes. Furthermore, universal screening mitigates disparities in ASD detection for children who are underrepresented minorities and from economically disadvantaged backgrounds. The American Academy of Pediatrics (AAP) recommends autism-specific screening at 18 and 24 months for all children. However, current literature suggests low fidelity of ASD screening, including (1) Inconsistent use of universal screening as opposed to selecting which children to screen, (2) Screening children at only one instead of both 18 and 24 month well- child visits, and (3) Not referring children for an evaluation and EI when positive screens indicate risk for ASD. There is currently lack of knowledge regarding factors that predict screening fidelity and age of diagnosis, critical in order to address the public health challenge of delayed detection of ASD. The current project uses a pseudo-trial design to assess factors that relate to high-fidelity screening and to age of ASD diagnosis in a diverse sample of at least 250 pediatric providers and electronic health records from a minimum of 27,000 toddlers seen for 18 month visits. Factors include child sociodemographics (sex, race, ethnicity, insurance status); pediatric providers’ beliefs and attitudes, sex, and years in practice; and practice factors, including size, resources, location (urban, suburban, and rural), and whether or not they are affiliated with an academic medical center. The first specific aim investigates predictors of high-fidelity screening, with the primary outcomes of referrals for evaluation and EI for children who screen at risk for ASD, and secondary fidelity outcomes of universal (vs. selective) screening and repeat screening at both 18- and 24-month visits (vs. single timepoint). The second aim investigates predictors of age of diagnosis in children with autism including child, provider, and practice factors as well as fidelity of screening. The first two aims will be addressed using random forests, a supervised machine learning algorithm that assesses not only which factors predict outcomes, but their relative importance. The third aim uses mixed methods to identify potentially modifiable healthcare provider and practice factors that relate to early detection of autism in primary care, through quantitative surveys and semi-structured interviews with providers. This Project contributes to the Public Health and Autism Science advancing Equitable Strategies across the life course (PHASES) Center’s aims of investigating modifiable health determinants, inequities in health services and opportunities for mitigation, especially in racially/ethnically diverse and economically disadvantaged children, and impact of health services delivery on subsequent health outcomes, through exploration of fa...