UCSF

Background

Sepsis progresses rapidly and is associated with considerable morbidity and mortality. Bedside risk stratification scores can quickly identify patients at greatest risk of poor outcomes; however, there is lack of consensus on the best scale to use.

Objective

To compare the ability of quick Sequential Organ Failure Assessment (qSOFA), the National Early Warning System (NEWS2), and the Shock Index-which does not require mental status assessment-to predict poor outcomes among patients with suspected sepsis during triage.

Design, setting, and participants

Retrospective cohort study of adults presenting to an academic emergency department (ED) from June 2012 to December 2018 who had blood cultures and intravenous antibiotics within 24 hours.

Main outcomes and measures

Clinical data were collected from the electronic health record. Patients were considered positive at qSOFA ≥2, Shock Index >0.7, or NEWS2 ≥5 scores. We calculated test characteristics and area under the receiver operating characteristics curves (AUROCs) to predict in-hospital mortality and ED-to-intensive care unit (ICU) admission.

Results

We included 23,837 ED patients; 1,921(8.1%) were qSOFA-positive, 4,273 (17.9%) Shock Index-positive, and 11,832 (49.6%) NEWS2-positive. There were 1,427 (6.0%) deaths and 3,149 (13.2%) ED-to-ICU admissions in the sample. NEWS2 had the highest sensitivity for in-hospital mortality (76.0%) and ED-to-ICU admission (78.9%). qSOFA had the highest specificity for in-hospital mortality (93.4%) and ED-to-ICU admission (95.2%). Shock Index exhibited the highest AUROC for in-hospital mortality (0.648; 95 CI, 0.635-0.662) and ED-to-ICU admission (0.680; 95% CI, 0.617-0.689). Test characteristics were similar among those with sepsis.

Conclusions

Institution priorities should drive score selection, balancing sensitivity and specificity. In our study, qSOFA was highly specific and NEWS2 was the most sensitive for ruling out patients at high risk. Performance of the Shock Index fell between qSOFA and NEWS2 and could be considered because it is easy to implement.