Predictive accuracy of Sepsis-3 definitions for mortality among adult critically ill patients with suspected infection

Background: Sepsis-3 definitions have been published recently;however, their diagnostic value remains controversial. This study was to assess the accuracy of Sepsis-3 definitions compared to Sepsis-1 definitions by stratifying mortality among adult critically ill patients with suspected infection. Methods: A multicenter, prospective cohort study was conducted from November 10, 2017 to October 10, 2018, in five Intensive Care Units (ICUs) at four teaching hospitals. Thirty-day mortality was compared across categories for both Sepsis-3 definitions and Sepsis-1 definitions, which were evaluated by logistic regression analysis followed by measurement of the area under the receiver operating characteristic curve (AUROC) for predicting 30-day mortality rates. Results: Of the 749 enrolled patients, 644 (85.9%) were diagnosed with sepsis according to the Sepsis-1 definitions. Among those patients, 362 were diagnosed with septic shock (362/749, 48.3%). However, according to the Sepsis-3 definitions, there were 483 patients with a diagnosis of sepsis (483/749, 64.5%), among whom 299 patients were diagnosed with septic shock (299/749, 39.9%). According to the Sepsis-3 definitions, sepsis (sepsis and septic shock) patients had higher 30-day mortality (41.8%) than sepsis patients according to the Sepsis-1 definitions (31.8%,χ^2 = 5.552, P = 0.020). The AUROC of systemic inflammatory response syndrome (SIRS) and quick sequential organ failure assessment (qSOFA) scores with regard to 30-day mortality rates were 0.609 (0.566-0.652) and 0.694 (0.654-0.733), respectively. However, the AUROC of SOFA scores (0.828 [0.795-0.862]) were significantly higher than that of SIRS or qSOFA scores (P < 0.001). Conclusion: In adult critically ill patients with suspected infection, the Sepsis-3 definitions were relatively accurate in stratifying mortality and were superior to the Sepsis-1 definitions.