Improving Universal Screening and Modeling the Effects on Referral and Diagnosis for Autism Spectrum Disorder

PROJECT SUMMARY/ABSTRACT Universal screening for autism spectrum disorder (ASD) has been recommended by the American Academy of Pediatrics in order to improve early diagnosis and facilitate access to early intervention for children with ASD. Despite the widespread support for this policy, the optimal approach for universal screening remains unknown, and the current approach may be limited in several ways. First, emerging evidence has highlighted the low sensitivity and positive predictive value of autism-specific screening (i.e., M-CHAT/F) alone when applied in real- world screening practice. The fixed schedule of screenings at 18 and 24 months, along with the exclusion of important risk indicators (e.g., sex, prematurity, family history, developmental delays, medical concerns) may also contribute to the overall low detection rate. Second, the current approach has overlooked the important downstream effects of universal screening on the diagnostic process, where real-world resource constraints of limited diagnostic services and the prolonged waiting time for diagnosis also critically affect the age at diagnosis. To bridge these gaps, an innovative analytic framework will be developed to integrate a large real-world health record dataset, data analytics and simulation modeling, with the overarching goal of identifying more effective universal screening policies that could further lower the age at diagnosis under practical resource constraints. In particular, this project will first incorporate an existing autism-specific screening tool with clinical variables related to known ASD risk factors to develop a comprehensive risk model for improving the screening accuracy (Aim 1). Then a discrete-event simulation model will be built to simulate the chain process from screening to diagnosis for any given screening policy, which is specified by risk threshold for referral, age range for screening, and interval for repeated screening. The simulation will also explicitly model the waiting process for the diagnostic evaluation under a limited-service capacity (Aim 2). Parameterized and calibrated based on the real-world clinical data, the simulation model will then be used to systematically evaluate and compare a rich set of alternative screening policies, which will allow policy makers to identify the optimal universal screening policy that maximizes the detection of ASD while lowering the age at diagnosis given the limited diagnostic service capacity (Aim 3). This proposed study will present a novel systemic framework for evaluating the effects of autism screening policies, which directly responds to the United States Preventive Services Task Force’s recent review calling for “a broader analytic framework that considers the process chain in its entirety.” The findings anticipated from this study will provide first-of-its-kind evidence in evaluating alternative universal screening policy designs to inform more effective policies to further facilitate early diagnosis.