Statistical Methods in COVID-19/PASC Clinical Research

点击翻译按钮获取中文摘要

Project Summary This proposal aims to develop, evaluate and disseminate novel statistical tools for rigorous investigation of the clinical spectrum, biological underpinnings and social determinants of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection, COronoVIrus Disease (COVID-19), and Post-Acute Sequelae of SARS- CoV-2 infection (PASC). Given the enormous scale of the COVID-19 pandemic, the potential severity of PASC, the complexity of available and anticipated data streams, and the paucity of biological and clinical knowledge, there is an urgent need for novel and robust statistical methods to address the most pressing PASC related research questions. Advanced statistical methods for observational data can be leveraged to address many of these questions; however, the identification, rigorous application and advancement of apropos methods requires sophisticated understanding of both the clinical context and the nuanced capabilities of available methods. To this end, we propose statistical innovations and novel translation of existing methods to significantly advance PASC clinical research, bringing together a team of physician-scientists and biostatisticians who are deeply embedded in COVID-19 clinical research to achieve the following specific aims: Aim 1: Develop and evaluate a causal mediation analysis framework for investigating the mechanistic pathways from SARS-CoV-2 infection to PASC and PASC recovery, including methods to accommodate time-varying and unevenly spaced mediators and time-to-event outcomes; Aim 2: Apply, evaluate, and extend marginal structural models as a framework to assess the impact of interventions on likelihood of PASC associated severe outcomes, in the context of time- varying confounding, competing and semi-competing risks, interval censoring and unobserved disease subtypes; Aim 3: Develop and apply methods for positive unlabeled data, using an expectation-maximization approach that leverages measured covariates and information on patient-level outcomes. The proposed methods will be applied using local and national emerging observational and EHR data resource. These statistical innovations will transform our understanding of the clinical course of PASC as we lead rigorous application to several leading-edge data resources.