Higher Mortality in Bacteremic Sepsis - A Propensity Score Matched Study
BackgroundSepsis is a highly heterogenous disease which needs to be thoroughly mapped. The aim of this study was to describe characteristics and outcome for critically ill patients with sepsis-3 with either culture-positive or -negative sepsis.Methods Patients with severe sepsis or septic shock were retrospectively identified in the local quality registry from a general mixed Intensive Care Unit (ICU) at a University Hospital in 2007-2014. Data were collected through manual review of medical
... rts. Patients were included if they fulfilled sepsis-3 criteria and at least one blood culture was sampled ±48h from ICU admission. In a propensity score analysis bacteremic and non-bacteremic patients were matched 1:1 with regard to age, comorbidities, site of infection and antimicrobial therapy prior to blood cultures. A Latent Class Analysis (LCA) was performed to identify unmeasured class membership.Results784 patients were identified as treated in the ICU with a sepsis diagnosis. Blood cultures were missing in 140 excluded patients and additionally 95 patients did not fulfill a sepsis diagnosis and were also excluded. In total 549 patients were included, 295 (54%) with bacteremia, 90 (16%) were non-bacteremic but had relevant pathogens detected from another body location and in 164 (30%) no relevant pathogen was detected in microbial samples. After the propensity score analysis (n=172 in each group) 90-day mortality was higher in bacteremic patients, 47%, than in non-bacteremic patients, 36%, p =0.04. Patients without antibiotic treatment before sample collection (n=352) were more often bacteremic, 63%. Non-bacteremic patients without prior antibiotic treatment had lower mortality, 31% (n=129), compared to non-bacteremic patients with prior antibiotic treatment, 51% (n=124), p <0.01.The LCA identified 8 classes, with different mortality rates, where pathogen detection in microbial sampleswere important factors for class distinction andoutcome.ConclusionsBacteremic patients had higher mortality than their non-bacteremic counter-parts. Bacteremia is more common in sepsis than previously reported, when studied in a clinical review. Clinical chart review should be considered gold standard since a significant proportion of the patients in the proposed sepsis cohort, did not have sepsis, but would have been included in ICD- or electronic health record (EHR) algorithm approaches.