ABSTRACT
STUDY OBJECTIVE: To predict severe sepsis/septic shock in ED patients. METHODS: We conducted a retrospective case-control study of patients ≥18 admitted to two urban hospitals with a combined ED census of 162,000. Study cases included patients with severe sepsis/septic shock admitted via the ED. Controls comprised admissions without severe sepsis/septic shock. Using multivariate logistic regression, a prediction rule was constructed. The model's AUROC was internally validated using 1000 bootstrap samples. RESULTS: 143 study and 286 control patients were evaluated. Features predictive of severe sepsis/septic shock included: SBPâ¯≤â¯110â¯mmâ¯Hg, shock index/SIâ¯≥â¯0.86, abnormal mental status or GCSâ¯<â¯15, respirationsâ¯≥â¯22, temperatureâ¯≥â¯38C, assisted living facility residency, disabled immunity. Two points were assigned to SI and temperature with other features assigned one point (mnemonic: BOMBARD). BOMBARD was superior to SIRS criteria (AUROC 0.860 vs. 0.798, 0.062 difference, 95% CI 0.022-0.102) and qSOFA scores (0.860 vs. 0.742, 0.118 difference, 95% CI 0.081-0.155) at predicting severe sepsis/septic shock. A BOMBARD scoreâ¯≥â¯3 was more sensitive than SIRSâ¯≥â¯2 (74.8% vs. 49%, 25.9% difference, 95% CI 18.7-33.1) and qSOFAâ¯≥â¯2 (74.8% vs. 33.6%, 41.2% difference, 95% CI 33.2-49.3) at predicting severe sepsis/septic shock. A BOMBARD scoreâ¯≥â¯3 was superior to SIRSâ¯≥â¯2 (76% vs. 45%, 32% difference, 95% CI 10-50) and qSOFAâ¯≥â¯2 (76% vs. 29%, 47% difference, 95% CI 25-63) at predicting sepsis mortality. CONCLUSION: BOMBARD was more accurate than SIRS and qSOFA at predicting severe sepsis/septic shock and sepsis mortality.