Comparing scoring systems for prediction of mortality in patients with bloodstream infection

Abstract

Summary Background Blood stream infections (BSIs) are associated with significant short-term mortality. There are many different scoring systems for assessing the severity of BSI. Aim We studied confusion, urea, respiratory rate, blood pressure, age 65(CURB65), Confusion Respiratory Rate, Blood pressure, age 65(CRB65), quick sequential organ failure assessment (qSOFA), systemic inflammatory response syndrome (SIRS) and National Early Warning Score (NEWS) and assessed how effective they were at predicting 30-day mortality across three separate BSI cohorts. Design A retrospective analysis was performed on three established BSI cohorts: (i) All cause BSI, (ii) Escherichia coli and (iii) Streptococcus pneumoniae. Methods The performance characteristics (sensitivity, specificity, positive predictive value, negative predictive value and area under receiver operating curve [AUROC]) for the prediction of 30-day mortality were calculated for the 5 scores using clinically relevant cut-offs. Results 528 patients were included: All cause BSI—148, E. coli—191 and S. pneumoniae—189. Overall, 30-day mortality was 22%. In predicting mortality, the AUROC for CURB65 and CRB65 were superior compared with qSOFA, SIRS and NEWS in the all cause BSI (0.72, 0.70, 0.66, 0.51 and 0.53) and E. coli cohorts (0.81, 0.76, 0.73, 0.55 and 0.71). In the pneumococcal cohort, CURB65, CRB65, qSOFA and NEWS were broadly equal (0.63, 0.65, 0.66 and 0.62), but all were superior to SIRS (0.57). CURB65, CRB65 and qSOFA had considerably higher accuracy than SIRS or NEWS across all cohorts. Conclusion CURB65 was superior to other scores in predicting 30-day mortality in the E. coli and all cause BSI cohorts. Further research is required to assess the potential of broadening the application of CURB65 beyond pneumonia.