Development and Validation of an Equitable Computable Phenotype for Classifying Pediatric Sleep Deficiency in Electronic Health Records

项目总结

PROJECT SUMMARY Sleep deficiency remains one of the most prominent and unaddressed public health concerns in pediatric healthcare settings. Pediatric sleep disparities are prominent across minoritized racial subpopulations in the dimensions of sleep duration, timing, alertness, behaviors, and quality/disorders. Despite the evidence of sleep deficiency burdening minoritized youth, these susceptible subpopulations are underrecognized in the clinical workflow leading to sleep medicine specialty services. Ignoring this underlying bias has yielded poorly defined pediatric sleep cohorts in clinical contexts (e.g., historical overrepresentation of White patients). A computable phenotype offers an efficient way to examine a large amount of data from many health systems, specifically electronic health record (EHR) data. Developing a computable phenotype for pediatric sleep deficiency will help us to target sleep screening and care where it is needed the most. However, to do this we will have to ensure the computable phenotype is designed to capture traditionally missed groups and is not biased in a way which harms historically marginalized subpopulations. This K01 will address these equity gaps by identifying potential biases inherent in EHR datasets, understanding their causes, and mitigating them using rigorous methods. The proposed K01 award will allow me to conduct the following aims: 1) the development and validation of a computable phenotype algorithm for classifying pediatric sleep deficiency; and 2) application of postprocessing bias mitigation methods to build and test an equitable computable phenotype model. My primary goal is to become an independent investigator focused on detecting pediatric sleep deficiency and translating that knowledge into effective strategies to improve sleep health in underserved communities. Achieving this goal requires training and research mentorship in specific content areas to (1) learn advanced biomedical informatics approaches for leveraging EHR (e.g., computable phenotyping) and develop an automated screening tool for use by pediatric health systems, (2) develop expertise in population-level sleep disparities research and SDH measurement, and (3) employ responsible conduct of research skills in developing unbiased artificial intelligence (AI) and applying machine learning. My proposed research and training plan will equip me with the skills necessary to become an independent investigator in pediatric sleep research and population health science, prepared to work in interdisciplinary clinical and technical teams. An exceptional interdisciplinary team has been assembled to complete the aims of this K01 research, as well as to mentor me in the training areas critical to my long-term career development. My K01 mentorship team includes both mid-career (Drs. Azizi Seixas, Jennifer Cooper, Christopher Bartlett) and senior mentors/collaborators (Drs. Deena Chisolm, Hongfang Lui, Kelly Kelleher, Lauren Hale), ensuring that I have access to researchers utilizing the latest cutting-edge methods, as well as mentors with large collaborative networks and resources to help launch my career.