776. Effect of the SEP-1 Sepsis Bundle on Mortality in Hospital-Onset v. Community-Onset Sepsis

Abstract

BackgroundThe SEP-1 sepsis bundle is a performance measure from the Centers for Medicare and Medicaid Services that requires blood cultures, serum lactate, broad-spectrum antibiotics, and IV fluids (in some cases) within 3 hours of onset of sepsis. Published evidence regarding an effect of SEP-1 on mortality is mixed and largely excludes cases of hospital-onset sepsis.MethodsRetrospective cohort study using clinical data from 4 University of California hospitals. Sepsis-related admissions from 2014–2017 were identified by diagnosis codes. We compared the effect of the SEP-1 sepsis bundle on in-hospital mortality in cohorts with community-onset and hospital-onset sepsis. To control for selection bias, patients who did and did not receive the SEP-1 bundle from each cohort were balanced on key variables related to likelihood of treatment using Mahalanobis distance matching.Results5,034 out of 6,005 sepsis-related patient encounters were matched, including 1,770 (35%) patients who received the SEP-1 bundle and 3,264 (65%) who did not. The SEP-1 bundle was not associated with an effect on mortality in the unmatched (Table 2) or matched analyses (Table 3). Point estimates from the matched analysis suggested a greater potential benefit associated with SEP-1 and its components in community-onset sepsis, but differences in effect size between community-onset and hospital-onset were nonsignificant. Among bundle components, timely blood cultures, lactate, and antibiotics were not associated with an effect on mortality, while IV fluids were associated with a 3-percentage-point increase in mortality risk. In subgroup analysis, IV fluids were associated with increased mortality in patients with pneumonia and immunosuppression.ConclusionThe use of the SEP-1 sepsis bundle was not associated with an effect on mortality. Subgroup analyses suggested situations in which IV fluids may be harmful, though prospective data are needed. Disclosures All authors: No reported disclosures.