Addressing bias in gold-standard autism assessments to improve the early identification of young Black and Latinx children

摘要 对于自闭症谱系障碍（ASD）儿童，接受早期干预（EI）需要准确的早期诊断， 通过提供高质量的ASD评估工具进行识别，这些工具告知临床医生能够提供准确的 1--11及时接受ASD的EI对于确保最佳的长期发育至关重要 12 -21黑人和西班牙裔儿童的严重、漏诊或晚诊通常报告为 那些被确定为智力残疾率较高和发育迟缓更严重的人。 一小部分文献表明ASD评估可能会将黑人或西班牙裔儿童错误分类; 错误的分类可能会导致ASD儿童的家庭承担不必要的评估费用 或错过重要的EI服务。20，30 -34关键的是，很少有研究表明，这两个最广泛的 推荐和研究ASD评估工具，自闭症诊断计划（ADOS）和自闭症 诊断访谈-修订版（ADI-R），对于黑人或西班牙裔人群同样有效和可靠 35 -122大多数ADOS和ADI-R研究甚至没有提供描述性资料， 种族或族裔数据42 - 67，110 -121;即使有报告，通常也不加以分析。 是一个相当强大的因素分析文献，这些工具，大多数研究包括广泛的年龄组， 并且没有人试图确定这些工具是否在心理测量学上等同于用于白色，黑色， 100 -121迄今为止，一些较小的临床研究已经调查了ADOS和ADI-R评分 在西班牙裔和黑人儿童中91 -98;两项单项把握度良好的研究确定了ADOS项目水平差异 在白色、黑人和西班牙裔儿童之间，有些项目显示出很大的效果。 心理测量学研究确定ADOS和ADI-R的因素结构，诊断准确性，可靠性， 聚合效度和发散效度对于识别年轻黑人和西班牙裔儿童的ASD是公平的。 文献在很大程度上是沉默的心理测量等效的ADOS和ADI-R的因素方面 结构，收敛和发散的有效性，可靠性和诊断的准确性，以确定年轻的黑人和 西班牙裔儿童为了解决这些问题，本研究的目的是：1）确定是否因素结构， ADOS和ADI-R对于3至5岁的白色、黑人和西班牙裔儿童是相等的; 2）创建 适应量表显示等因素结构，收敛效度，发散效度和内部信度 3）比较已建立和适应的ADOS和ADI-R量表的诊断准确性， 白色、黑人和西班牙裔儿童;以及4）进行敏感性分析，以确定其他相关变量 (e.g.,儿童性别、母亲教育）对已确定成果的影响。所有分析将根据来自 探索早期发育研究（SEED）数据集，一个大型的、人口统计学上多样化的、多州病例对照 3 - 5岁ASD和DD儿童的研究，具有良好的表型特征。

Abstract For children with autism spectrum disorder (ASD), receipt of early intervention (EI) requires accurate early identification via provision of high quality ASD assessment tools that inform clinicians ability to provide accurate and timely diagnoses.1--11 Timely receipt of EI for ASD is critical for ensuring optimal long term developmental outcomes.12-21 Critically, missed or later identification of Black and Hispanic children are commonly reported with those identified often having higher rates of intellectual disability and more severe developmental delays.22-29 A small body of literature suggests that ASD assessments may misclassify Black or Hispanic children; misclassification could result in families of children with ASD taking on the costs of unnecessary assessments or missing vital EI services.20,30-34 Critically, there are few research studies showing that the two most widely recommended and researched ASD assessment tools, the Autism Diagnostic Schedule (ADOS) and the Autism Diagnostic Interview-Revised (ADI-R), are equally valid and reliable for use with Black or Hispanic populations as they are for White populations.35-122 The majority of ADOS and ADI-R studies do not even provide descriptive race or ethnicity data42-67,110-121; even when reported these are not usually analyzed.68-90,100-109 Further, while there is a reasonably robust factor analytic literature on these tools, most studies include a wide range of age groups, and none have sought to establish whether these tools are psychometrically equivalent for use with White, Black, and Hispanic children.100-121 To date, a few smaller clinical studies have investigated ADOS and ADI-R scoring in Hispanic and Black children91-98; two single well powered studies identified ADOS item level differences between White, Black and Hispanic children, with some items displaying large effects.122,123 There is a clear need for psychometric studies establishing that the ADOS and ADI-R’s factor structure, diagnostic accuracy, reliability, and convergent and divergent validity are equitable for identifying ASD in young Black and Hispanic children. The literature is largely silent on the psychometric equivalence of the ADOS and ADI-R in terms of the factor structure, convergent and divergent validity, reliability, and diagnostic accuracy for identifying young Black and Hispanic children. To address these limitations, the current study aims to: 1) determine if the factor structures of the ADOS and ADI-R are equal for White, Black, and Hispanic children between the age of 3 and 5; 2) create adapted scales displaying equal factor structures, convergent validity, divergent validity, and internal reliability across groups; 3) compare the diagnostic accuracy, of established and adapted ADOS and ADI-R scales across White, Black and Hispanic children; and 4) conduct sensitivity analyses to determine if other relevant variables (e.g., child sex, maternal education) impact identified outcomes. All analyses will be conducted on data from the Study to Explore Early Development (SEED) dataset, a large, demographically diverse, multi-state case control study of children with ASD and DD between 3 and 5 years old with excellent phenotypic characterization.