Acquired bloodstream infection in the intensive care unit: incidence and attributable mortality

INTRODUCTION To estimate the incidence of intensive care unit (ICU)-acquired bloodstream infection (BSI) and its independent effect on hospital mortality. METHODS We retrospectively studied acquisition of BSI during admissions of >72 hours to adult ICUs from two university-affiliated hospitals. We obtained demographics, illness severity and co-morbidity data from ICU databases and microbiological diagnoses from departmental electronic records. We assessed survival at hospital discharge or at 90 days if still hospitalized. RESULTS We identified 6339 ICU admissions, 330 of which were complicated by BSI (5.2%). Median time to first positive culture was 7 days (IQR 5-12). Overall mortality was 23.5%, 41.2% in patients with BSI and 22.5% in those without. Patients who developed BSI had higher illness severity at ICU admission (median APACHE III score: 79 vs. 68, P < 0.001). After controlling for illness severity and baseline demographics by Cox proportional-hazard model, BSI remained independently associated with risk of death (hazard ratio from diagnosis 2.89; 95% confidence interval 2.41-3.46; P < 0.001). However, only 5% of the deaths in this model could be attributed to acquired-BSI, equivalent to an absolute decrease in survival of 1% of the total population. When analyzed by microbiological classification, Candida, Staphylococcus aureus and gram-negative bacilli infections were independently associated with increased risk of death. In a sub-group analysis intravascular catheter associated BSI remained associated with significant risk of death (hazard ratio 2.64; 95% confidence interval 1.44-4.83; P = 0.002). CONCLUSIONS ICU-acquired BSI is associated with greater in-hospital mortality, but complicates only 5% of ICU admissions and its absolute effect on population mortality is limited. These findings have implications for the design and interpretation of clinical trials.