Sepsis is a leading cause of in-hospital death in high-income countries, and it likewise causes a formidable burden of disease in low-income countries, where in-hospital mortality for severe sepsis can exceed 60%. Building upon Duke University’s strong collaborative clinical research platform in Kilimanjaro, Tanzania, these studies will use data-driven clustering methods and Bayesian latent class models to define clinically meaningful subtypes of sepsis that are specific to the infectious disease epidemiology and population sub-structures of sub- Saharan Africa (sSA). In doing so, we seek to advance the long-term goal of improving detection, risk stratification and, eventually, tailored interventions for sepsis among adults in resource-limited settings. The rationale driving this project is that sepsis subtype characterization holds great promise for improving the evaluation, management and clinical investigation of sepsis in sSA. To perform our characterizations of adult sepsis subtypes, we will leverage existing samples and data from our research platform’s 2016-2019 severe febrile illness cohort to inform a two-year prospective observational study of sepsis admissions at district hospitals in Kilimanjaro. By developing a precision medicine-based approach to classify the key pathophysiologic subtypes of sepsis in sSA, this project promotes the US National Institutes of Health’s mission to uncover new knowledge that will lead to better health for everyone—in this case, better health for the most severely ill in the region with the highest burden of sepsis in the world.

To achieve this, the project has set out SPECIFIC AIMS that will develop clinical phenotype clusters of adult sepsis derived from clinical bioinformatics using Bayesian statistics (Aim 1) as well as immunologic sepsis clusters based upon the molecular characterization of the host immune response to infection (Aim 2). We will integrate the approaches in Aim 1 and Aim 2 in order to identify robust and clinically meaningful subtypes of sepsis in Kilimanjaro (Aim 3). In Year 1, we will use the existing samples and data collected 2016-2019 to develop and refine the statistical and analytical models for our Aims. This will inform the analytical framework for the prospective sepsis patient cohort in Years 2-3, which will be the basis for both derivation and validation of the clinical and molecular sepsis subtype classifications. The clinical clusters and molecular characterizations discovered in Aim 1 and Aim 2 will also be compared to findings from clinical bioinformatic and gene expression signature analyses that have described sepsis subtypes in Europe and North America. The disease epidemiology of sepsis in sSA—high prevalence of advanced HIV infection and more diverse sepsis etiologies—as well as potential host genetic differences compared to European and North American sepsis patients necessitate that subtype identification be specifically derived and validated for application in sSA. Not only will this project identify subtypes that improve triage and tailored intervention design for sepsis in sSA—it will also establish a framework for sepsis research in a setting where the greatest gains are needed and where the greatest improvements in sepsis outcomes can indeed be made.