Machine-Assisted Interdisciplinary Approach For Early Clinical Evaluation of Neurodevelopmental Disorders

摘要 神经发育迟缓是大多数罕见疾病的特征，通常是第一个表现。 罕见疾病的非特异性早期表现对患者和照顾者都提出了挑战， 多年没有诊断，医生必须区分常见疾病和罕见疾病。 早期评估可以简化诊断过程，并导致快速实施有针对性的 治疗在这项建议中，我们的主要目标是缩短全面遗传学的途径， 通过初级保健电子医疗评估疑似神经发育障碍（NDD） 基于电子病历（EMR）的机器学习算法识别临床上适合遗传学治疗的患者 评价我们讨论了我们的计划，为一体化的预测试遗传咨询在初级保健设置，通过 视频和远程医疗，并将开发一个范例，可以适应儿科初级保健 工作流我们将在UG 3阶段通过以下方式实现并迭代改进我们的算法： 学术遗传学家，神经发育儿科医生和初级保健之间的密切合作关系 儿科医生的儿童健康中心（CHC）在华盛顿DC，并过渡到成熟的计划， CNH Goldberg中心的所有业务。因此，我们将把早期的基因评估带到最大的 通过利用我们的多学科团队，在哥伦比亚特区大都会地区建立初级儿科实践网络 致力于NDD的早期识别和表征。我们将致力于实现以下目标： 目标1（UG 3）：评估可扩展的机器辅助管道在早期识别患者方面的效用 基于EMR的自动特征提取的NDD。我们将训练并反复改进一台机器- 学习算法，根据儿童的电子病历识别遗传性NDD高风险儿童。 目标2（UG 3）：评估初级保健医生发起的多学科评价的效用，以加快 遗传评估和神经发育表型。我们的工作流程从自动化图表开始 识别将允许初级保健提供者访问我们的多学科神经发育遗传学 团队包括远程医疗、基于应用的视频和电子摄入在内的技术创新将 促进这一进程。 目标3（UH 3）：通过以下方式评价EMR中机器辅助识别NDD的通用性： 扩大对整个戈德堡中心儿科实践网络的访问。我们将扩展到所有CNH 初级保健诊所服务于高度多样化的华盛顿大都市区，并确保方法是 对不同地点的特定人口统计学和流行病学因素具有鲁棒性。 我们的方法将在初级保健环境中识别发育迟缓的患者， 诊断奥德赛和加快深入的表型和遗传调查，以及重新评估 测序结果用于不同DC大都市人群的早期诊断。

ABSTRACT Neurodevelopmental delay is a feature of a majority of rare diseases and is often the first presenting sign. Nonspecific early presentations of rare disorders challenge both patients and caregivers who often struggle for years without diagnoses, and physicians who must distinguish between common concerns and rare disease. Early evaluations can streamline the diagnostic process and lead to rapid implementation of targeted therapies. In this proposal, our primary objective is to shorten the pathway to comprehensive genetic evaluations for suspected neurodevelopmental disorders (NDDs) through primary care electronic medical record (EMR) based machine-learning algorithmic identification of patients clinically eligible for genetic evaluation. We discuss our plan for integration of pretest genetic counseling in the primary care setting through video and telemedicine, and will develop a paradigm that can be adapted to the pediatric primary care workflow. We will implement and iteratively improve upon our algorithms during the UG3 Phase through a close partnership between academic geneticists, neurodevelopmental pediatricians, and the primary care pediatricians of Children’s Health Center (CHC) in Washington DC, and transition the mature program during the UH3 to all CNH Goldberg Center practices. We will thus bring early genetic evaluations to the largest network of primary pediatric practices in the D.C. Metropolitan area by leveraging our multidisciplinary team dedicated to early identification and characterization of NDDs. We will address the following aims: Aim 1 (UG3): Assess utility of a scalable machine-assisted pipeline for early identification of patients with NDDs based on automated feature extraction from EMR. We will train and iteratively refine a machine- learning algorithm to identify children at high risk of genetic NDDs based on their EMR. Aim 2 (UG3): Assess utility of a primary care clinician-initiated multidisciplinary evaluation to expedite genetic evaluation and neurodevelopmental phenotyping. Our workflow starting with automated chart identification will permit primary care providers access to our multidisciplinary neuro-developmental-genetics team. Technical innovations including telemedicine, application based videos, and electronic intakes will facilitate this process. Aim 3 (UH3): Evaluate generalizability of machine-assisted identification of NDDs from EMR by expanding access to entire network of Goldberg Center Pediatric practices. We will expand to all CNH primary care clinics serving the highly diverse Washington DC metropolitan area and ensure approach is robust to the specific demographic and epidemiologic factors of different sites. Our approach will identify patients with developmental delay in the primary care setting at the beginning of a diagnostic odyssey and expedite deep phenotyping and genetic investigations, as well as reevaluate sequencing results for early diagnosis in the diverse DC metropolitan population.