基于血清HBP、HE4构建的XGBoost模型对脓毒症患者发生急性肾损伤的预测价值OA

Objective To investigate the predictive value of an extreme gradient boosting(XGBoost)model based on serum heparin-binding protein(HBP)and human epididymis protein 4(HE4)for the acute kidney injury(AKI)occurrence in the patients with sepsis.Methods A total of 120 sepsis patients treated in Taiyuan Municipal Central Hospital from September 2023 to March 2025 were enrolled as the study subjects.The pa-tients were divided into the AKI group and non-AKI group according to whether AKI occurred within 7 d after admission.The data of the patients were collected,and serum HBP and HE4 levels at admission were meas-ured and compared between the two groups.The Lasso regression and multivariate Logistic regression were used to analyze the risk factors for the AKI occurrence in sepsis patients.The XGBoost algorithm was applied to rank the importance of the factors related to AKI,and the XGBoost model was constructed based on the se-lected important feature subset.The receiver operating characteristic(ROC)curve was drawn to analyze the efficiency of serum HBP,HE4 and the XGBoost model for predicting the AKI occurrence in sepsis patients.Results There were 46 cases in the AKI group and 74 cases in the non-AKI group.The proportion of the pa-tients complicating diabetes at admission,Acute Physiology and Chronic Health Evaluation Ⅱ(APACHE Ⅱ)score,Sequential Organ Failure Assessment(SOFA)score,proportion of septic shock occurrence,and serum C-reactive protein(CRP),interleukin-6(IL-6),procalcitonin(PCT),HBP and HE4 levels in the AKI group were all higher than those in the non-AKI group(P<0.05).The Lasso regression analysis screened the six influencing factors:complicating diabetes,septic shock,CRP,IL-6,HBP and HE4.The multivariate Logistic regression analysis results showed that complicating diabetes,septic shock occurrence and increased serum CRP,IL-6,HBP and HE4 levels were the independent risk factors for the AKI occurrence in sepsis patients(P<0.05).The restricted cubic spline(RCS)analysis revealed a positive nonlinear relationship between ser-um HBP,HE4 and the risk of AKI occurrence in sepsis patients(P<0.05).The XGBoost algorithm indicated that septic shock,complicating diabetes,HBP,IL-6,HE4 and CRP as the predictive indicators could maximize the model's predictive efficiency.The ROC curve analysis results demonstrated that the areas under the curves(AUCs)of serum HBP,HE4 and XGBoost model for predicting the AKI occurrence in sepsis patients were 0.782,0.789 and 0.943,respectively.The AUC of the XGBoost model was significantly higher than that of HBP(Z=3.470,P=0.001)or HE4(Z=3.295,P=0.003)alone.Conclusion Serum HBP and HE4 lev-els are significantly elevated in sepsis patients with AKI occurrence,meanwhile which shows a positive nonlin-ear relationship with the AKI occurrence.The XGBoost model constructed by septic shock,complicating dia-betes,HBP,IL-6,HE4 and CRP as predictive indicators demonstrates the high predictive efficiency for the AKI occurrence in sepsis.