The overall aims of this thesis were to improve prediction, diagnostics and knowledge on epidemiology of sepsis.
In paper I, we developed and evaluated an integrated platform for rapid analysis of sepsis-causing organisms directly from blood samples. Testing with blood samples spiked with bacteria and samples from septic patients indicate that the detection limit of the system is in the upper part of clinically relevant bacteria concentration range. The paper describes proof-of-principle for the integrated system for faster sepsis diagnostics.
In paper II, we assessed the incidence of hospital-treated sepsis in an entire population based on clinical findings. The annual incidence for severe sepsis (sepsis-2) was 687/100 000 person years (95% CI 549-824) and the annual incidence for sepsis-3 was 780/100 000 person years (95% CI 633-926). These estimates are closer to the true incidence of sepsis compared to estimates based on ICD-codes.
In paper III & IV, we evaluated different early warning scores for sepsis prediction and detection. We also developed and evaluated a candidate warning score for sepsis based on vital signs and heparin-binding protein. NEWS2 was superior to qSOFA and RETTS for screening for sepsis. Even with a statistical approach, we could not construct better warning scores for sepsis than NEWS2.
In paper V, patients with sepsis admitted to an ICU were retrospectively studied in a clinical chart review. We found a high proportion of bacteremic patients, probably due to that clinical chart review minimizes the misdiagnosis of other conditions. We also demonstrated higher mortality among bacteremic patients, than in non-bacteremic patients.