Improved diagnosis and management of sepsis and bloodstream infection

Abstract

Sepsis is a severe organ dysfunction triggered by infections, and a leading cause of hospitalization and death. Concurrent bloodstream infection (BSI) is common and around one third of sepsis patients have positive blood cultures. Prompt diagnosis and treatment is crucial, but there is a trade-off between the negative effects of over diagnosis and failure to recognize sepsis in time. The emerging crisis of antimicrobial resistance has made bacterial infections more difficult to treat, especially gram-negative pathogens such as Pseudomonas aeruginosa.

The overall aim with this thesis was to improve diagnosis, assess the influence of time to antimicrobial treatment and explore prognostic bacterial virulence markers in sepsis and BSI. The papers are based on observational data from 7 cohorts of more than 100 000 hospital episodes. In addition, whole genome sequencing has been performed on approximately 800 invasive P. aeruginosa isolates collected from centers in Europe and Australia.

Paper I showed that automated surveillance of sepsis incidence using the Sepsis-3 criteria is feasible in the non-ICU setting, with examples of how implementing this model generates continuous epidemiological data down to the ward level. This information can be used for directing resources and evaluating quality-of-care interventions. In Paper II, evidence is provided for using peripheral oxygen saturation (SpO2) to diagnose respiratory dysfunction in sepsis, proposing the novel thresholds 94% and 90% to get 1 and 2 SOFA points, respectively. This has important implications for improving sepsis diagnosis, especially when conventional arterial blood gas measurements are unavailable. Paper III verified that sepsis surveillance data can be utilized to develop machine learning screening tools to improve early identification of sepsis. A Bayesian network algorithm trained on routine electronic health record data predicted sepsis onset within 48 hours with better discrimination and earlier than conventional NEWS2 outside the ICU. The results suggested that screening may primarily be suited for the early admission period, which have broader implications also for other sepsis screening tools. Paper IV demonstrated that delays in antimicrobial treatment with in vitro pathogen coverage in BSI were associated with increased mortality after 12 hours from blood culture collection, but not at 1, 3, and 6 hours. This indicates a time window where clinicians should focus on the diagnostic workup, and proposes a target for rapid diagnostics of blood cultures. Finally, Paper V showed that the virulence genotype had some influence on mortality and septic shock in P. aeruginosa BSI, however, it was not a major prognostic determinant.

Together these studies contribute to better understanding of the sepsis and BSI populations, and provide several suggestions to improve diagnosis and timing of treatment, with implications for clinical practice. Future works should focus on the implementation of sepsis surveillance, clinical trials of time to antimicrobial treatment and evaluating the prognostic importance of bacterial genotype data in larger populations from diverse infection sources and pathogens.