Sepsis survivors face myriad challenges in the post-sepsis recovery period. Both sepsis survivors and healthcare systems are motivated to reduce the strikingly high rate of hospital readmission following an initial hospitalization for sepsis. Unfortunately, current one-size-fits-all approaches to risk prediction and treatment strategies are inadequate for optimizing care for this heterogenous population. The overarching goal of our work is to improve health outcomes and reduce healthcare utilization for sepsis survivors. The objectives of this study are to apply latent class analysis (LCA) to identify important new phenotypes of sepsis survivors with distinct characteristics and risk profiles, and to demonstrate the application of LCA in determining whether phenotype membership moderates the effectiveness of interventions designed to reduce readmission. To achieve this objective we will 1) use our team's large, feature-rich sepsis datamart to identify distinct phenotypes of sepsis survivors based on patients' predisposing characteristics, illness factors, and treatments, 2) validate the predictive utility of the developed method in a separate sample of 1200 sepsis survivors, and 3) determine whether phenotype is a moderator of the effectiveness of a sepsis recovery program currently being tested by our team in a clinical trial. Sepsis survivors represent a vulnerable population with high morbidity, mortality, and hospital readmissions, and strategies to improve transition and recovery are urgently needed. By transitioning away from whether post-sepsis treatment strategies work toward which strategies work best for whom, our project will be the first rigorous examination of phenotypes of sepsis survivors and their association with readmission risk and differential treatment effects. Ultimately, these results will provide clinicians, researchers, and policy makers with immediate, actionable data about how to target interventions to specific group vulnerabilities, leading to effective and efficient reduction in post-sepsis morbidity.
This project addresses a fundamental knowledge-practice gap in the management of 14 million annual sepsis survivors that can result in population-level reductions in post-sepsis morbidity and healthcare utilization. Our study will apply latent class analysis to identify distinct phenotypes of sepsis survivors and use phenotype membership to predict risk of hospital readmission as well as responsiveness to a targeted sepsis recovery intervention. Results will advance our understanding and approach to post-sepsis recovery to improve care and healthcare utilization for millions of sepsis survivors across the United States.