A comprehensive, predictive mortality score for patients with bloodstream infections (PROBAC): a prospective, multicentre cohort study

Abstract Objectives Bloodstream infections (BSI) are an important cause of mortality, although they show heterogeneity depending on patients and aetiological factors. Comprehensive and specific mortality scores for BSI are scarce. The objective of this study was to develop a mortality predictive score in BSI based on a multicentre prospective cohort. Methods A prospective cohort including consecutive adults with bacteraemia recruited between October 2016 and March 2017 in 26 Spanish hospitals was randomly divided into a derivation cohort (DC) and a validation cohort (VC). The outcome was all-cause 30-day mortality. Predictors were assessed the day of blood culture growth. A logistic regression model and score were developed in the DC for mortality predictors; the model was applied to the VC. Results Overall, 4102 patients formed the DC and 2009 the VC. Mortality was 11.8% in the DC and 12.34% in the CV; the patients and aetiological features were similar for both cohorts. The mortality predictors selected in the final multivariate model in the DC were age, cancer, liver cirrhosis, fatal McCabe underlying condition, polymicrobial bacteraemia, high-risk aetiologies, high-risk source of infection, recent use of broad-spectrum antibiotics, stupor or coma, mean blood pressure <70 mmHg and PaO2/FiO2 ≤ 300 or equivalent. Mortality in the DC was <2% for ≤2 points, 6%–14% for 3–7 points, 26%–45% for 8–12 points and ≥60% for ≥13 points. The predictive score had areas under the receiving operating curves of 0.81 (95% CI 0.79–0.83) in the DC and 0.80 (0.78–0.83) in the VC. Conclusions A 30 day mortality predictive score in BSI with good discrimination ability was developed and internally validated.