Abstract

Background: Sepsis, a life-threatening syndrome with escalating mortality per treatment delay, requires prognostic tools beyond current Sepsis-3 criteria. This study investigated the dual predictive capacity of Nutrition Risk Screening 2002 (NRS-2002) and Human Leukocyte Antigen-DMB (HLA-DMB), proposing a novel early-warning framework for sepsis risk stratification.
Methods: This case-control study enrolled 90 patients with acute infections from the Department of Pulmonary and Critical Care Medicine at Hefei Second People's Hospital. Participants were stratified into sepsis (n = 45) and non-sepsis (n = 45) groups according to Sepsis-3 diagnostic criteria. Clinical baseline characteristics, laboratory parameters, nursing-assessed nutritional risk scores, and HLA-DMB gene expression levels were systematically collected through standardized case report forms. Binary logistic regression identified independent predictors, while ROC curve analysis was employed to construct a combinatorial prediction model.
Results: In patients with sepsis, HLA-DMB gene expression levels were significantly lower, while NRS-2002 scores were higher; both were independent predictors of sepsis (p < 0.001). Restricted cubic spline analysis indicated that higher HLA-DMB levels might reduce the risk of sepsis, whereas lower NRS-2002 scores were associated with an increased risk. Furthermore, receiver operating characteristic curve analysis demonstrated that the combined predictive efficacy of HLA-DMB expression and NRS-2002 scores surpassed that of either variable alone (AUC = 0.8430).
Conclusions: HLA-DMB gene expression levels and NRS-2002 scores have been utilized to assess the risk of developing sepsis. Their combined evaluation has enhanced predictive accuracy, facilitating the rational allocation of medical resources in the early stages.

DOI: 10.7754/Clin.Lab.2025.250557