# Accuracy of the Identification and Prognosis Prediction of SOFA-Based Sepsis-3 for Septic Patients in the Emergency Department Compared With Sepsis-2

## TL;DR

This study compares the Sepsis-3 criteria with Sepsis-2 for identifying sepsis and predicting mortality in emergency department patients, finding that Sepsis-3 identifies more cases but has lower mortality prediction accuracy than qSOFA and NEWS.

## Contribution

The study evaluates the performance of Sepsis-3 in the ED and introduces variable combinations to improve mortality prediction accuracy.

## Key findings

- Sepsis-3 identified more sepsis cases than Sepsis-2 with moderate consistency.
- qSOFA and NEWS scores outperformed SOFA in predicting mortality.
- Combining T, RR, and SBP improved mortality prediction accuracy.

## Abstract

Aim: To evaluate the value of the Sequential Organ Failure Assessment (SOFA) score, a Sepsis-3 criterion, for identification and prognosis prediction among adult patients with sepsis in the emergency department (ED) compared with the Sepsis-2.

Methods: Adult patients with suspected sepsis presenting to the ED were retrospectively identified via Sepsis-2/Sepsis-3 criteria. The vital signs, laboratory test results, etc., were collected, and the SOFA/quick SOFA (qSOFA) scores and National Early Warning Score (NEWS) were calculated accordingly. ROC curves were generated to evaluate mortality prediction accuracy.

Results: Among the 481 patients included, 288/339 met the Sepsis-2/Sepsis-3 criteria, respectively, with moderate between-protocol consistency (Kappa = 0.507, p < 0.001; concordance = 77.3%); 115 patients (23.9%) died in hospital or within 28 days. SOFA/qSOFA scores and NEWS were significantly greater in the sepsis and death groups (p < 0.001), but there was no between-group difference for Sepsis-2/Sepsis-3. The temperature (T) and respiratory rate (RR) increased in the death group, whereas the systolic blood pressure (SBP) decreased. The usefulness of the SOFA score (AUC = 0.644) for predicting mortality was lower than that of qSOFA score (AUC = 0.716) and NEWS (AUC = 0.718), which could be improved (AUC = 0.701–0.721) by combining with two/three of variables (T, RR, and SBP).

Conclusion: Compared with Sepsis-2, Sepsis-3 identified more patients with sepsis and was suitable for ED use. The SOFA score had lower mortality prediction accuracy than the qSOFA score and NEWS, which could be significantly improved by combining with two/three variables (T, RR, and SBP).