Construction and validation of liver injury risk prediction model in patients with sepsis
Objective:To investigate the influencing factors of the risk of sepsis-associated liver injury(SALI)in patients with sepsis,and establish and verify a model for predicting the risk of SALI.Methods:A total of 415 patients with sepsis admitted to the Intensive Care Unit(ICU),the Affiliated Huai'an No.1 People's Hospital of Nanjing Medical University,from January 2019 to January 2022 were enrolled and divided into SALI group(n=97)and non-SALI group(n=318),according to clinical diagnosis.Basic information and clinical data of patients were collected.Least absolute shrinkage and selection operator(LASSO)regression was used for univariate screening.Then,multivariate Logistic regression was used to further select the risk factors and construct the model based on the nomogram.The performance of the nomogram,including differentiation,accuracy and clinical utility,was evaluated through internal validation by using bootstrap method.Results:Nine variables were selected by LASSO regression for multivariate analysis.Multivariate Logistic regression further showed that total bilirubin,alanine aminotransferase,γ-glutamyl transpeptidase,mechanical ventilation,renal failure,international normalized ratio,and acute respiratory failure were independent risk factors for SALI.The internal validation of the constructed model showed that the area under the ROC curve was 0.823(95%CI:0.773-0.873),and the model demonstrated satisfactory accuracy(P>0.05).Decision curve analysis indicated that the prediction model could generate a net benefit within the threshold range of 5%to 100%for predicting the risk of developing SALI in sepsis patients.Conclusion:Total bilirubin,alanine aminotransferase,γ-glutamyl transpeptidase,mechanical ventilation,renal failure,international normalized ratio,and acute respiratory failure were independent risk factors for SALI in patients with sepsis,and the nomogram model established based on above factors had good predictive value.
万方数据
维普
