Although there is increasing recognition that viruses are an important cause of sepsis, most clinicians and policymakers tend to equate sepsis with bacterial infection. As a result, current guidelines emphasize immediate administration of antibiotics for all patients with suspected sepsis, even though antibiotics will not benefit patients with pure viral sepsis and may cause harm at both the individual and population levels. However, the contribution of viruses to the overall burden of sepsis is poorly characterized, and important questions remain about antibiotic utilization patterns and potential antibiotics-associated harms in this population. The proposed project seeks to provide rigorous estimates of the proportion of community-onset sepsis caused by respiratory viruses, describe current antibiotic utilization in viral sepsis, quantify potential antibiotic-associated harms, and identify patient subgroups in whom antibiotics can safely be limited. The ultimate goal of the proposal is to promote safer, more nuanced treatment for patients with sepsis, in line with AHRQ’s stated priorities of improving patient safety and avoiding harm. The proposed research will accomplish this goal by using large datasets with rich clinical data from electronic health records (EHR) at over 200 diverse hospitals, a validated EHR-based approach to identifying sepsis, and sophisticated causal inference statistical techniques to address confounding in order to pursue three aims: (1) Determine the fraction of community- onset sepsis attributable to respiratory viruses and characterize antibiotic use and predictors of prolonged antibiotics in this population; (2) Assess hospital-level variation in antibiotic prescribing for patients with viral sepsis and evaluate risk-adjusted outcomes for patients at high- versus low- utilization hospitals; and (3) Assess the potential utility of the combination of positive viral assay and low procalcitonin level as an indicator that antibiotics can safely be withheld or rapidly discontinued. The candidate, Dr. Claire Shappell, is an advanced research fellow and Pulmonary and Critical Care physician at Brigham and Women’s Hospital (BWH) with experience using EHR data to gain actionable insights into problems in critical care. Dr. Shappell’s goals during the K08 period are to develop advanced proficiency with the acquisition, preparation, and handling of large healthcare datasets; learn statistical methods for causal inference analysis including propensity score methods and approaches to multilevel data; acquire expertise in the use of R statistical computing software; and strengthen abilities in scientific communication. To achieve these goals, she has assembled a multidisciplinary mentorship team led by Drs. Michael Klompas and Chanu Rhee, experts in the use of EHR data for sepsis surveillance and outcomes research. By the completion of the K08 award period, Dr. Shappell will be well-positioned to achieve her long-term goal of beco...