Advances in acute care have decreased short-term mortality of sepsis, resulting in an increasing number of survivors who experience significant morbidity and mortality. Recurrent infections account for 60% of hospital readmissions after sepsis and are an important cause of long-term mortality. Identifying sepsis survivors at risk for infections and understanding predisposing factors are therefore important first steps to develop personalized interventions to improve long-term outcomes. The overall research goal is to develop innovative targeted interventions to improve long-term outcomes after sepsis. This proposal focuses on identifying subgroups of sepsis survivors (phenotypes) at high risk for infection-related hospital readmissions and deaths. It is hypothesized that the combination of host factors, disease characteristics, and interventions prior to and during sepsis will identify distinct phenotypes that are captured by clinical data. These clinical phenotypes likely have distinct pathophysiologic mechanisms (e.g., immunosuppression), predispose to different outcomes (e.g., recurrent infections), and patients with the same phenotype may respond similarly to targeted interventions such as immunomodulation.
In Aim 1, three nationally representative datasets of sepsis survivors will be assembled and analyzed by machine learning techniques to identify valid clinical sepsis phenotypes.
In Aim 2, phenotypes at high-risk for infection-related readmissions and deaths at 6 months will be identified, and underlying biomarker profiles for these phenotypes analyzed.
In Aim 3, these high-risk phenotypes will be used to conceptualize and simulate an adaptive platform trial designed to test the efficacy of three different immunomodulatory drugs. An adaptive trial design was chosen, because it can test multiple interventions simultaneously and reduces the chances of exposing patients to ineffective or harmful interventions. The research plan is augmented by expert mentoring and rigorous didactic training. Together, this will provide the candidate with essential career development skills in the science of precision medicine, including machine learning methodologies, Bayesian statistics, and innovate clinical trial design. This proposal has potentially groundbreaking implications for the future treatment of sepsis survivors, because it deviates from the current ?one-size fits all? approach and attempts to create the framework for personalized interventions. This framework sets the stage for future independent investigator (RO1) applications evaluating personalized treatments in adaptive clinical trials, and will uniquely position the candidate as a future leader in the field of long-term outcomes after critical illness.
This project is important for public health because sepsis is a life-threatening infection that affects 30 million people worldwide each year. Long-term outcomes for patients who survive a hospital admission for sepsis are poor, and many patients develop new disabilities or worsening of preexisting health conditions. This proposal supports the NIH?s mission to enhance health and reduce illness and disability by trying to identify interventions for sepsis survivors at highest risk for poor outcomes.
