Promoting Universal Screening and Early Identification of Child ADHD via Integrated Automatic EHR Supports in Primary Care

摘要 ADHD是儿科最常见的行为健康状况之一， 初级保健.如果不治疗，ADHD会产生负面后果，包括 自杀、犯罪行为和严重的物质使用。美国儿科学会 建议在初级保健中对4-18岁的儿童进行ADHD筛查。不幸的是， 遵守实践指南和行为健康的实际实施 筛选是高度可变的。即使有普遍的行为健康筛查基础设施， 地方，筛查率可以保持在50%以下。电子健康记录（EHR） 识别ADHD风险儿童的算法有可能实现普遍筛查 并促进早期识别和与护理的联系。 拟议的项目将：1）描述ADHD筛查频率的差异， 诊断和保健利用ADHD儿童，2）开发一种算法来预测 ADHD表型早于使用EHR结构化和文本数据诊断的典型年龄， 和3）与利益相关者合作，制定实施路线图， 表型分型算法在儿科初级保健。研究人员成功地将自然 语言处理（NLP）技术对EHR数据进行识别，以确定患者的行为健康 条件，包括自杀行为，自闭症和双相情感障碍，但NLP还没有 用于ADHD的诊断。由此产生的表型分析算法有可能成为 集成到儿科初级保健的EHR中，自动标记有ADHD风险的儿童， 实时触发更密切的监测，减少筛查和诊断的差异，并启动 早期治疗。由此产生的表型分型算法和实现路线图将设置 R 01试验的阶段，以评估自动EHR表型分型算法在 儿科初级保健。

Abstract ADHD is among the most common behavioral health conditions presented in pediatric primary care. When left untreated, ADHD is associated with negative consequences including suicide, criminal behavior, and serious substance use. The American Academy of Pediatrics recommends screening for ADHD in primary care for children ages 4-18. Unfortunately, compliance with practice guidelines and real-world implementation of behavioral health screening is highly variable. Even with universal behavioral health screening infrastructure in place, screening rates can remain below 50%. Developing an electronic health record (EHR) algorithm to identify children at risk for ADHD has the potential to realize universal screening and facilitate early identification and linkage to care. The proposed project will: 1) Describe disparities in the frequency of ADHD screening, diagnosis, and healthcare utilization for children with ADHD, 2) Develop an algorithm to predict ADHD phenotypes earlier than the typical age of diagnosis using EHR structured and text data, and 3) Collaborate with stakeholders to develop an implementation roadmap for the phenotyping algorithm in pediatric primary care. Researchers have successfully applied Natural Language Processing (NLP) techniques to EHR data to identify patients with behavioral health conditions, including suicidal behaviors, autism, and bipolar disorder, but NLP has not been applied to the identification of ADHD. The resulting phenotyping algorithm holds potential to be integrated into EHR in pediatric primary care to automatically flag children at risk for ADHD in real-time to trigger closer monitoring, reduce disparities in screening and diagnosis, and initiate earlier treatment. The resulting phenotyping algorithm and implementation roadmap will set the stage for a R01 trial to evaluate the clinical utility of an automated EHR phenotyping algorithm in pediatric primary care.