Abstract
Sepsis is a common condition with significant morbidity, mortality, and annual costs-of-care in the billions of
dollars. Despite innumerable studies on the causes of, and therapies for, sepsis, the mortality rate has not
changed substantially in the last 20 years. Treatments remain generic, with current guidelines recommending
the same approach for all patients, regardless of the obvious litany of differences that exist at baseline.
Moreover, the blanket administration of 30cc/kg of intravenous fluid (IVF) to all patients is recognized as being
directly harmful to some. Patient-level heterogeneity in prior sepsis trials is recognized as a substantial
contributor to all of these problems, yet no prior investigation has attempted to identify septic phenotypes, a
necessary first step towards precision care. This project will address these critical knowledge gaps by
systematically characterizing profiles of patients with sepsis using a combination of sonographic and biomarker
variables. This will begin at the time of presentation to the Emergency Department and continue at specified
time points during hospitalization, thereby creating a rich longitudinal database from which identification of
unique phenotypes will be possible. Our central hypothesis is that central venous hypertension: (i) is
deleterious to the function of the lungs, liver kidneys, and vascular endothelium; (ii) is worsened by cardiac
dysfunction and IVF administration; and (iii), contributes to adverse organ-specific and overall outcomes.
Cardiac function will be assessed with echocardiography while congestion in the lungs, and kidneys will be
assessed using previously validated sonographic markers of congestion. Biomarkers for each organ will be
collected concurrently thereby increasing the fidelity of our phenotypic profiles by pairing indicators of macro-
and microscopic stress and dysfunction. Classical and machine learning approaches will be used to analyze
our large data-stream and develop a rule-based system to identify distinct sub-populations of sepsis patients
who have greater risk/likelihood of both organ-specific and overall adverse outcomes. Completion of the
project will result in generation of the first comprehensive description of septic phenotypes at the level of
individual organs and the organismal level, including the effects of resuscitative interventions. The combination
of longitudinal imaging and laboratory data is novel and offers potential short-term (outcome-based guidance of
IVF administration and prognosis via sonographic data) and long-term (development of biomarker panels and
targeted therapeutics) patient-oriented benefits.