Sepsis is a complex, heterogeneous spectrum disorder with significant morbidity and mortality. Like many inflammatory processes, predictive models for sepsis are derived primarily from phenotypic snapshots of physiologic or laboratory data at one or two isolated timepoints using conventional modeling techniques such as logistic regression or survival analyses. Changes in physiologic and laboratory data over time can be used to predict progression to single and multiple organ failure such as the multiple organ dysfunction syndrome (MODS). Physiological biomarker data collected at the highest available frequencies per unit time could both identify and forecast clinically subtle increased risk for disease progression. Models that use this data can aid in understanding the evolution of sepsis, as well as prognostication and timely treatment.
Dr. Andre Holder’s research focuses on whether temporal trends in physiologic and laboratory biomarkers, and complex measures of physiologic variability, can improve prediction of clinical progression in patients with sepsis. He is interested in employing advanced data-driven techniques such as nonlinear prediction modeling and machine learning approaches to select variables with the highest yield to predict organ failure in sepsis. The ultimate goal of his research is to develop prediction models that forecast outcome trajectories to optimally guide clinical management as well as to inform discussions between providers and patients or their families.