Accounting for selection and information bias in studies of Autism Spectrum Disorder - PROJECT ABSTRACT In the United States, approximately 1 in 44 children are diagnosed with Autism Spectrum Disorder (ASD). Many children with ASD experience a reduced quality of life due to adverse health outcomes across the life course. Early diagnosis and intervention services are critical for improving long-term outcomes for individuals with ASD. However, because of historical practices of racial segregation and discrimination in the US, access to an early diagnosis and intervention services are not equitable. Identifying and diagnosing children with ASD facilitates access to appropriate services, but racial and ethnic disparities in healthcare services cause many Non-Hispanic Black and Hispanic children with ASD symptomology to be undiagnosed and under-treated in the healthcare system and under-represented in research studies generated from those clinical databases. Even though these racial and ethnic disparities in ASD diagnosis are well documented, many epidemiologic studies use billing code algorithms and healthcare databases to examine etiologic questions about ASD, often without necessary consideration of potential structural biases. Thus, this proposed dissertation research will explore the impact of information and selection bias on ASD research, with an application to the association between epidural analgesia use and ASD. The dissertation research will address the following specific aims: (1) Assess the validity (i.e., sensitivity, predictive values) of an ASD diagnosis documented in the medical record compared to a diagnosis derived from gold standard clinical assessments (stratified by child's race and ethnicity); (2) Use internal validation data to conduct a bias analysis to examine the impact of outcome misclassification on the association between epidural use during childbirth and ASD; and (3) Conduct Monte Carlo simulations to assess the impact of selection bias on the association between epidural use during childbirth and ASD. Data from the Study to Explore Early Development (SEED), a large US multi-site case control study, will be used to conduct the proposed research. During the grant period, the fellowship applicant will achieve the following goals: (1) gain more in-depth experience in scientific writing and results dissemination activities; (2) obtain in-depth training in responsible conduct of research; (3) develop analytical skills in methods for quantitative bias analysis and simulation studies; (4) enhance professional and personal development; and (5) develop a strong understanding of child development and the etiology of Autism Spectrum Disorders. These goals would allow the applicant to establish a solid foundation in the etiology of ASD, health inequities, and quantitative bias analysis approaches through additional coursework, seminars, guided readings, and consultation with the mentorship team. Furthermore, they will give the applicant the skillset needed to achieve their goal of becoming a tenure-track professor focused on bridging the gap between epidemiologic research and public health practice to improve maternal and child health outcomes among racial and ethnic minorities.
