Using Big Data to Understand Sepsis in an Immunocompromised Population

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT Through this proposal, I will develop and evaluate a sepsis prediction tool targeted to allogeneic hematopoietic cell transplant (HCT) recipients. Allogeneic HCT recipients are an immunocompromised population that is disproportionately affected by sepsis, a life-threatening dysregulated immunologic response to an infection. While it is well established that early detection and treatment of sepsis with fluids and broad-spectrum antibiotics reduce the risk of mortality, recent data suggests early broad-spectrum antibiotic use in allogeneic HCT recipients may have microbiota-mediated detrimental effects on morbidity and mortality. Because of the risks associated with both missed and falsely identified sepsis events among allogeneic HCT recipients, early and accurate sepsis diagnosis is crucial. However, sepsis is generally challenging to diagnose and is made more complicated in allogeneic HCT recipients by the fact that sepsis presents differently following transplantation and common complications of the transplant procedure present like sepsis. In previous work, we demonstrated that current sepsis clinical criteria have low predictive value among allogeneic HCT recipients and concluded that population specific prediction tools are needed. Recently developed single algorithm, machine learning sepsis prediction tools have shown promising results in general, immuno-competent populations. However, few studies have tested the ability of machine learning workflows to predict sepsis in high-risk, immunocompromised patients, such as allogeneic HCT recipients. Additionally, current sepsis prediction tools rely on the assumption that the true relationship between the predictors and the outcome is contained within a single algorithm. This proposed work has two main objectives. The first is to develop an automated sepsis prediction tool for allogeneic HCT recipients using a state-of-the-art ensemble-based machine learning workflow (the super learner) that relaxes the single algorithm assumption of current sepsis prediction tools. The second is to estimate the utility of this tool in comparison to currently available tools in both traditional (accuracy methods) and novel ways (mathematical modeling of health outcomes). Both aims will be completed with the ultimate goal of improving sepsis prediction among allogeneic HCT recipients and in such, reducing sepsis related mortality and inappropriate antibiotic use among this hard to diagnose population. Further, this research will advance the methodological discussion around the usefulness of machine learning prediction tools in clinical practice and the use of ensemble modeling for prediction of rare, high-case fatality diseases. Such advances have the potential to improve the prediction of health outcomes beyond sepsis and reduce the burden of treatable diseases among immunocompromised populations.