Transitions Among Discrete Clinical States During ICU Stays in Patients with SARS-CoV-2 Pneumonia

A. Project Summary and Abstract The goals of this NRSA postdoctoral fellowship proposal are: 1) to facilitate Dr. Catherine Gao's development as an independent physician-scientist and an expert in the handling, integration, and computational analyses of complex datasets, and 2) to model transitions between discrete clinical states during the ICU stays of patients with SARS-CoV-2 pneumonia. This proposal takes advantage of a unique dataset generated as part of the Successful Clinical Response in Pneumonia Therapy (SCRIPT) Systems Biology Center led by Dr. Wunderink, the candidate's primary sponsor. SCRIPT contains the electronic health record data, as well as rich expert clinician adjudication of outcomes. Leveraging those data, in Aim 1, the applicant will use machine learning approaches to cluster and model distinct clinical states over the course of ICU stays. In Aim 2, the candidate will identify features associated with transitions towards favorable or unfavorable clinical states, looking specifically at the administration of specific pharmaceuticals and the development of ventilator associated pneumonia. These data will further inform other cores within SCRIPT to optimize the high resolution but sparsely available multiomic data. The candidate and her mentors have used the unique research environment provided by SCRIPT to design a detailed training plan tailored to the candidate's specific needs and goals. The plan includes a rigorous research component that lays the foundation for a successful career: 1) formalized coursework (including a Master's in Health and Biomedical Informatics) to learn computational skills to manage large electronic medical record datasets and analyze multiomic data, 2) hands-on training through research plan and feedback from a multidisciplinary team of mentors to become an independent physician-scientist.