Construction of a predictive model for the risk of sepsis-related liver injury in the intensive care unit: a retrospective cohort study over 10 years

Abstract OBJECTIVE: Sepsis-associated liver injury (SALI) seriously affects the prognosis of patients in the intensive care unit, and there is still a lack of effective predictive tools. The aim of this study was to investigate the independent risk factors of SALI and to establish and validate an effective risk prediction model. METHODS: Patients diagnosed with sepsis by the Medical Information Marketplace for Intensive Care (MIMIC-IV v2.2) were included. The least absolute shrinkage and selection operator (LASSO) logistic regression was used to identify independent risk factors for developing SALI. Risk prediction models were constructed and column-line plots were drawn. The predictive model was validated using receiver operating characteristic (ROC) curves, Hosmer-Lemeshow (H-L) test and decision curve analysis (DCA). RESULTS: A total of 8549 sepsis patients were included in the final cohort. There were 4834 males and 3715 females; the mean age of the patients was 66.72 ± 16.18. A total of 1067 (12.48%) patients developed SALI. All patients were randomized into a training cohort and a validation cohort in a 7:3 ratio. The training cohort was analyzed using LASSO logistic regression with the final inclusion of eight indicators used to construct the predictive model: abdominal infection (OR =4.046,95% CI:3.236-5.059, P<0.001), vasopressin (OR=2.188, 95% CI:1.778-2.694 , P<0.001), continuous renal replacement therapy (CRRT) (OR=1.928, 95%CI:1.506-2.469, P<0.001), red blood cell distribution width (RDW) (OR=1.109, 95%CI:1.069-1.151, P<0.001), platelet ( OR=0.996, 95%CI:0.996-0.996, P< 0.001), prothrombin time (PT) (OR=1.042, 95%CI:1.034-1.051, P<0.001), lactate (OR=1.154, 95%CI:1.107-1.202 ,P<0.001) and SOFA score (OR=1.119, 95%CI:1.089-1.150 ,P <0.001). The area under the ROC curve (AUC) of the predictive model was 0.830 (95% CI:0.814-0.847) in the training cohort and 0.847 (95% CI:0.825-0.869) in the validation cohort. The calibration curves were in good agreement with the ideal curves in both the training cohort and the validation cohort. Decision curve analysis (DCA) showed the ability to obtain a net benefit from this model with good clinical applicability. CONCLUSION: Abdominal infection, vasopressin, CRRT, RDW, platelet, PT, lactate, and SOFA score are independent risk factors for patients with SALI. The SALI prediction model constructed based on the above factors has certain clinical predictive value.