Informing the Emergency Care of Septic Shock Patients: A Novel Application of Data-Driven Analytics

PROJECT SUMMARY
Background: Septic shock is a commonly, costly, and deadly condition. There is increasing recognition that
septic shock patients vary significantly in terms of (1) clinical presentation, (2) response to treatments, and (3)
clinical outcomes. This patient-level heterogeneity may explain why optimal early septic shock management
remains poorly understood. Patients with septic shock are four times more likely to die than septic patients
without shock. Our preliminary data shows that Black patients have higher odds of mortality from septic shock
compared to White patients. Current studies do not characterize patient heterogeneity among septic shock
patients and do not explicate pharmacogenetic factors that may influence disparities in outcomes.
Objective: Insights from synergistic data types are necessary to provide a more complete understanding of
septic shock heterogeneity and hereditable factors that may influence vasopressor response and disparities in
outcomes. The overall objective of the proposed research is to characterize both phenotypic and genetic
aspects of heterogeneity in septic shock. This work is organized into two aims: (1) Identify Septic Shock
Phenotypes Using Advanced Analytic Methods and (2) Quantify Vasopressor Pharmacogenetic
Polymorphisms by Race and Vasopressor Response. Our overall hypothesis is that advanced analytic
methods applied to clinical and genetic data can identify defining features of septic shock heterogeneity that
are relevant to early septic shock management in the Emergency Department and disparities in outcomes.
Methods: (Aim 1) We will use a national dataset of septic patients with hypotension refractory to initial
Emergency Department fluid resuscitation and apply unsupervised machine learning clustering methods to
define clinically relevant phenotypes of early septic shock patients. We will analyze phenotypic variation in
clinical characteristics and outcomes. Then, we will develop a supervised model for phenotype classification.
(Aim 2) We will perform targeted pharmacogenomics of 100 samples balanced for race, 73 of which are part of
an existing research biobank of septic shock patients from our urban, safety-net hospital. We will enroll an
additional 27 patients to complete the sample. We will examine the presence of risk alleles of single nucleotide
polymorphisms for vasopressor-relevant genes by race. We will also examine the association between the
targeted genetic polymorphisms and shock reversal.
Career Development: During the proposed Career Development Award, I will work with my mentorship team
to build the skills necessary to achieve independence as a clinical researcher. Specifically, I will 1) receive
hands-on experience in the design and conduct of translational clinical research studies, 2) take didactic
coursework in data science, biomedical informatics, and implementation science, 3) receive training and
education in translational data science, clinical decision support, pharmacogenomics, and precision public
health, and 4) become a leader and an effective mentor in academic emergency medicine.

Grant Number: 5K23GM144802-04
NIH Institute/Center: NIH
Principal Investigator: Lauren Black