External validation and comparison of two versions of simplified sequential organ failure assessment scores to predict prognosis of septic patients

Abstract

Background Evidence shows that simplified SOFA scoring system has better clinical practice. Objective This study aimed to validate and compare the scores acquired with simplified organ dysfunction criteria optimized for electronic health records (eSOFA), and simplified and accurate sequential organ failure assessment (sa‐SOFA) for their accuracies in predicting the prognosis of septic patients. Methods This retrospective observational study was conducted at three major academic hospitals. Clinical data from 574 patients diagnosed with sepsis following the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis‐3)were retrospectively retrieved and analysed. Scores from the quick sequential organ failure assessment (qSOFA) and sequential organ failure assessment (SOFA) were used as reference scores. The area under the receiver operating characteristic curve (AUROC) was used to assess the performance of eSOFA and sa‐SOFA scores in predicting in‐hospital mortality. Results AUROC analysis demonstrated the predictability of the four scoring systems for sepsis surveillance, listed in descending order as: sa‐SOFA, 0.790 (95% confidence interval [CI]: 0.754‐0.822); SOFA, 0.774 (95% CI: 0.738‐0.808); eSOFA, 0.729 (95% CI: 0.691‐0.765); and qSOFA, 0.618 (95% CI: 0.577‐0.658). Moreover, sa‐SOFA and SOFA scores (Z = 1.950, P = .051) did not significantly differ from each other in discriminatory power, but the sa‐SOFA score had a higher power than eSOFA score (P values < .001). Conclusion sa‐SOFA appeared to have performed better than eSOFA score for predicting in‐hospital mortality in patients’ sepsis. Further large prospective studies are needed to externally validate.

A comparison of qSOFA, SIRS and NEWS in predicting the accuracy of mortality in patients with suspected sepsis: A meta-analysis

“…Eight studies [8,12,15,17,20,24,30,35] were prospective studies and the others [13, 14, 16, 18, 19, 21-23, 25-29, 31-34, 36] were retrospective studies. Five studies [12,15,23,24,33] were multi-center studies, and the others [8, 13, 14, 16-22, 25-32, 34-36] were described as single-center. The average age varied between 57 and 81, and the proportion of men varied between 46% and 65% in these studies.…”

“…Except for the study of Brink, et al [14], all other studies included consecutive or random cases, avoiding inappropriate exclusion. Because the cut-off values of the 11 studies [8,16,17,25,26,28,[32][33][34][35][36] were not determined in advance, there were high risks in the index test. Dick's funnel plot suggested that there was no publication bias (Fig 2 ).…”