ABSTRACT/PROJECT SUMMARY SARS-CoV-2, the novel coronavirus resulting in COVID19 disease, has caused a global pandemic of unprecedented impact. In just over three months, SARS-CoV-2 spread has infected more than 2 million individuals and resulted in at least 150,000 deaths. Non-pharmacologic interventions (NPIs), like social distancing and shelter-in-place measures, have proven to be the only effective strategy available today to mitigate rapidly growing outbreaks. However, the effectiveness of social distancing depends on early identification of viral spread, since even short delays in NPIs can result in overwhelming surges in acute illness and healthcare demand. Unfortunately, current prediction models of SARS-CoV-2 viral spread are based on lagging or incomplete indicators of infections like COVID19 case positivity, hospitalization, or death rates. As a result, these prediction models may have limited efficacy during the earliest stages of viral spread, when NPIs can have the greatest impact. This project will use methods my laboratory has developed to predict sepsis – a life-threatening infectious disease marked by a dysregulated host response – that incorporate novel real-time data to identify and compare the value of early indicators of SARS-CoV-2 viral spread. We will compare the predictive utility of these data in SARS-CoV-2 with influenza, a seasonal viral disease that can cause sepsis while also resulting in surges in healthcare demand. We will use a unique source of highly-detailed electronic health record data arising from an integrated health system with more than 200 medical offices and 21 hospitals caring for 4.4 million patients. Our findings will have broad and immediate impact for predicting SARS-CoV-2 viral spread that can inform effective strategies for COVID19 mitigation by patients, clinicians, public health agencies, researchers, and health systems.