CTSA Administrative Supplement for Informatics Core: A novel AI/ML system to predict respiratory failure and ARDS in Covid-19 patients

PROJECT Summary: The Einstein-Montefiore Institute for Clinical and Translational Research (ICTR) proposes an AdministrativeSupplement pursuant to NOT-TR-20-011, CTSA Program Applications to Address 2019 Novel Coronavirus(Covid-19). Specifically, this application addresses the urgent need for research on the coronavirus pandemicwith a project focusing on informatics and data science to preemptively identify patients with the life-threatening complications of SARS-CoV-2, using CTSA-supported core resources. Characterized by severehypoxemia, tachypnea, and decreased lung compliance, the diagnosis of acute respiratory failure (ARF) is abad prognostic sign, and in a subset, leads to development of acute respiratory distress syndrome (ARDS).The rates of Covid-19 infection and death in the Bronx have been higher than any other borough of NYC. Asthe major regional health system, our experience with Covid-19 provides guideposts that may prevent futurevictims of this pandemic. The bleak picture for ARDS in the 4,452 patients admitted showed that 78% of ourintubated Covid-19 patients developed ARDS, with 42% mortality. The overall goal of this proposal is toleverage our novel informatics and analytics platforms enabled by the Einstein-Montefiore CTSA (NIH/NCATS1ULTR002556), and extensive Artificial Intelligence and Deep Learning resources to implement a novel,situational awareness and clinical decision support system for ARF and ARDS (SA-ARDS). We will re-train ourexisting deep learning models with data collected from Covid-19 patients and contextualize its implementationwith data from the Covid-19 response during the pandemic in NYC. The SA-ARDS data platform will providelongitudinally integrated clinical data for research and multi-institutional and national collaborations, with thefollowing specific aims: Aim 1: To integrate, re-train, and validate our novel, near real-time, Electronic RiskAssessment System (ERAS 1.0) optimized for early recognition of ARF, ARDS, and inpatient mortality; Aim 2:To develop an evidence based, real-time, and context appropriate Situational Awareness clinical decisionsupport system targeting ARF and ARDS response (SA-ARDS); and Aim 3: Through our partner CTSAorganizations, to standardize and disseminate ERAS 1.0 and the SA-ARDS to other health systems, includingthe NYC consortium of CTSA hubs and the PCORI INSIGHT network. We will use the clinical data underlyingthe SA-ARDS to support research in local, regional, and national collaborations. All the methods and toolsdeveloped will be shared with the CTSA community via NCATS' National Center for Data to Health (CD2H).