Quantitative methods for improving medical decision-making

ABSTRACT

Innovation in causal inference and implementation of electronic health record systems are rapidly transforming medical care. In this dissertation, we present three examples in which use of methods in causal inference and large electronic health record data address existing challenges in medical decision-making. First, we use principles of causal inference to examine the structure of randomized trials of biomarker targets, which have produced divergent results and controversial clinical guidelines for management of hypertension and other chronic diseases. We discuss four key threats to the validity of trials of this design. Second, we use methods in causal inference for adjustment of time-varying confounding to estimate the effect of time-varying treatment strategies for hypertension. We report the results of a study which used longitudinal electronic health record data from a prospective virtual cohort of veterans. Third, we use individual-level electronic health record data to predict the need for critical care resources during surges in COVID-19 cases, to aid hospital administrators with resource allocation in periods of crisis. (PsycInfo Database Record (c) 2022 APA, all rights reserved)