Abstract
INTRODUCTION: The prognostic value of D-dimer (DD) in sepsis remains controversial. This study aimed to investigate the performance of DD for predicting sepsis mortality in the hospital and for identifying its potential correlates. MATERIALS AND METHODS: The clinical and laboratory data of adult sepsis patients were extracted from the Medical Information Mart for Intensive Care III (MIMIC III, v1.4) database using the structured query language (SQL). The database contains critical illness admitted to the intensive care unit at Beth Israel Deaconess Medical Center between June 2001 and October 2012. The association between DD and mortality was investigated with receiver operating characteristic (ROC) curve, restricted cubic spline and logistic regression analysis. Subgroup analysis was also used for identifying DD correlates. RESULTS: The study population consisted of 358 sepsis patients. Those who died during hospital stay (N = 160) had significantly higher DD values than those who survived (N = 198). The area under the ROC curve (AUC) of DD was 0.59 (P < 0.010). In subgroup analysis, white blood cell (WBC) count > 18 x10(9)/L and vasopressor therapy significantly decreased DD diagnostic performance. Categorical DD value was independently associated with hospital mortality after sequential organ failure score (SOFA) and blood lactate adjustment. Restricted cubic spline analysis revealed a U-shape relationship between DD and in-hospital mortality. DISCUSSION: We conclude that the accuracy of DD for predicting in-hospital sepsis mortality depends on WBC count and vasopressor therapy. Both low and extremely elevated DD values are associated with higher risk of death.