Leveraging routinely collected health data to improve early identification of autism and co-occurring conditions

点击翻译按钮获取中文摘要

ABSTRACT – Project 2 The overall goal of the Duke Autism Center of Excellence (ACE) is to use a translational digital health and computational approach to address the critical need for more effective autism screening tools, objective outcome measures, and brain-based biomarkers that can be used in clinical trials with young autistic children. This Project will develop and evaluate a novel digital health approach to autism screening. Universal autism screening is recommended for children at 18 months. This is typically achieved via a caregiver questionnaire. However, research has shown that a commonly used autism screening questionnaire has reduced accuracy when used in real-world settings, such as primary care. By leveraging health data related to early medical conditions collected as part of clinical care, Project 2 aims to develop an automatic, objective tool for autism prediction at 18 months that can be implemented in primary care settings. We will use routinely collected health data to develop a prediction model for autism and use the model to design a clinical decision support tool for providers that can be integrated into pediatric primary care and includes actionable guidance regarding referrals and linkage to services. We will first develop and validate a generalizable, off-the-shelf model to predict autism for use at 18 months of age using longitudinal claims data (Medicaid and Blue Cross Blue Shield) from a diverse sample of children across North Carolina with continuous coverage from birth to age 6 years (N ~ 230,000) to predict likelihood of an autism diagnosis (N ~ 6,000). We will then adapt the autism prediction model to the Duke University Health System (DUHS) clinical environment and augment it with granular electronic health record (EHR) data by using machine learning-based natural language processing to embed provider notes. Through engagement with stakeholders both within and outside of DUHS and in collaboration with Project 1, we will use the prediction model to design a clinical decision support prototype that could assist providers in making appropriate and timely referrals. Through the design process, we will identify a set of key priority factors to consider when choosing a clinical decision support for autism screening that are applicable across a broad range of stakeholders in different health care settings. Finally, leveraging our robust data on early health encounters, we will describe the nature and prevalence of patterns of medical conditions during early life. We will test the specific hypothesis that gastrointestinal problems during early life are associated with higher rates of psychiatric conditions by age 6.