Use of the N3C enclave and machine learning to create generalizable algorithms that predict patient outcomes at (a) diagnosis, and (b) time of hospitalization
Over the last three years, many studies have sought to identify risk factors for more severe SARS-CoV-2 infection and adverse outcomes of COVID-19. Since the onset of the pandemic, progressive insights have been gained into demographic and clinical features, co-morbidities, various biomarkers, therapeutic strategies, and social factors that affect acute disease outcomes among outpatients and inpatients. The fundamental question being asked is: Based on knowledge concerning prognostic factors that has accrued since 2019, can the N3C enclave be leveraged to identify generalizable algorithms that predict hospitalization, disease severity, and mortality associated with SARS-CoV-2 infection?
Aim 1: Leveraging data in the N3C enclave and supervised machine learning, create generalizable algorithms that predict outcomes at (a) diagnosis, and (b) time of hospitalization. Ideally these algorithms should be interpretable, leveraging selected features that can be identified clinically or through diagnostic testing with relative ease.
Aim 2: Assess the temporal and geographic generalizability of the algorithms optimized in Aim 1. An underlying goal is to identify algorithms that may be clinically helpful in a broad context, taking key predictive features into account
Analysis Plan / Research Method
Be explicit about how your study design, research methodology, and analytical plan will address the overall research question and aims. See the detailed guidance on expectations at the submission portal when completing the relevant fields in your proposal.
Be explicit about how your inclusion strategy will rigorously address the overall question and aims.
Be explicit about how your exclusion strategy will rigorously address the overall question and aims.
See inclusion and exclusion sections.
Written report within the N3C Enclave detailing the findings from analyses that support your conclusions and that address the overall research questions and aims. Provide sufficient methodologic detail for independent replication of the work, and an interpretation of the findings and study limitations. All code generated to complete the work will be made available in a deployable format with sufficient documentation/annotation for independent validation. Figures and/or tables in a format specified by NCATS suitable for uploading to the N3C Public Health web site.