Improved diagnosis and management of sepsis and bloodstream infection

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.

List of scientific papers

I. John Karlsson Valik, Logan Ward, Hideyuki Tanushi, Kajsa Müllersdorf, Anders Ternhag, Ewa Aufwerber, Anna Färnert, Anders F Johansson, Mads Lause Mogensen, Brian Pickering, Hercules Dalianis, Aron Henriksson, Vitaly Herasevich, Pontus Nauclér. Validation of automated sepsis surveillance based on the Sepsis-3 clinical criteria against physician record review in a general hospital population: observational study using electronic health records data. BMJ Quality & Safety. 2020, vol 29, 735-745.
https://doi.org/10.1136/bmjqs-2019-010123

II. John Karlsson Valik, Lisa Mellhammar, Jonas Sundén-Cullberg, Logan Ward, Christian Unge, Hercules Dalianis, Aron Henriksson, Kristoffer Strålin, Adam Linder, Pontus Nauclér. Peripheral Oxygen Saturation Facilitates Assessment of Respiratory Dysfunction in the Sequential Organ Failure Assessment Score With Implications for the Sepsis-3 Criteria. Critical Care Medicine. 2022, vol 50, e272-e283.
https://doi.org/10.1097/CCM.0000000000005318

III. John Karlsson Valik*, Logan Ward*, Hideyuki Tanushi, Anders F Johansson, Anna Färnert, Mads Lause Mogensen, Brian W Pickering, Vitaly Herasevich, Hercules Dalianis, Aron Henriksson, Pontus Nauclér. Predicting onset of sepsis in the emergency department and non-intensive care unit wards using a machine learned Bayesian network model in electronic health records: a cohort study. *Shared first authorship. [Manuscript]

IV. Jasper Van Heuverswyn*, John Karlsson Valik*, Pontus Hedberg, Suzanne Desirée van der Werff, Christian Giske, Pontus Nauclér. Association between time to appropriate antimicrobial treatment and 30-day mortality in patients with bloodstream infections: a retrospective cohort study. Clinical Infectious Diseases. 2022, Sep, Epub ahead of print. *Shared first authorship.
https://doi.org/10.1093/cid/ciac727

V. John Karlsson Valik, Christian Giske, Badrul Hasan, Mónica Gozalo Margüello, Luis Martínez Martínez, Manica Mueller Premru, Žiga Martinčič, Bojana Beović, Sofia Maraki, Maria Zacharioudaki, Diamantis Kofteridis, Kate McCarthy, David Paterson, Maria de Cueto, Isabel Morales, Leonard Leibovici, Tanya Babich, Fredrik Granath, Jesús Rodríguez-Baño, Antonio Oliver, Dafna Yahav, Pontus Nauclér. Association of Pseudomonas aeruginosa virulence genotype with patient characteristics, septic shock and mortality in bloodstream infection. [Manuscript]