[Development and evaluation of a risk prediction model for sepsis-associated liver injury in the intensive care unit].

OBJECTIVE: To investigate independent risk factors for sepsis-associated liver injury (SALI) in intensive care unit (ICU) patients with sepsis, establish an effective risk prediction model, and validate it.

METHODS: A retrospective cohort study was conducted using patients diagnosed with sepsis from the Medical Information Mart for Intensive Care IV (MIMIC- IV) v2.2 database. Patients were divided into SALI and non-SALI groups based on the occurrence of SALI. Baseline characteristics, comorbidities, infection sites, vital signs upon ICU admission, initial laboratory parameters within 24 hours of ICU admission, treatment interventions, 24-hour urine output, disease severity scores, and prognostic indicators collected and compared between the two groups. The entire cohort was divided into training and validation sets using a 7 : 3 random sampling ratio. Independent risk factors for SALI occurrence were identified in the training set using least absolute shrinkage and selection operator (LASSO) regression analysis and multivariate Logistic regression analysis. Based on this, a risk prediction model was constructed and presented as a nomogram. Model performance was evaluated in both training and validation sets. Predictive capability was assessed using receiver operator characteristic curve (ROC curve), model calibration was determined via calibration curves, and net benefit was evaluated through decision curve analysis (DCA).

RESULTS: The final cohort included 8 549 sepsis patients, comprising 1 067 in the SALI group and 7 482 in the non-SALI group. Significant differences were observed between the two groups regarding gender, type of ICU admission, hypertension, organ failure status, site of infection (bloodstream infection, pulmonary infection, abdominal infection), vital signs, and laboratory tests [white blood cell count (WBC), hemoglobin (Hb), red cell distribution width (RDW), platelet count (PLT), albumin (Alb), total bilirubin (Tbil), aspartate aminotransferase (AST), alanine aminotransferase (ALT), creatinine, blood urea nitrogen, serum calcium, serum potassium, serum sodium, bicarbonate, anion gap (AG), prothrombin time (PT), activated partial thromboplastin time (APTT), international normalized ratio (INR), pH, arterial partial pressure of carbon dioxide (PaCO2), lactate], therapeutic interventions [use of albumin, norepinephrine, vasopressin, continuous renal replacement therapy (CRRT), mechanical ventilation, central venous catheterization], 24-hour urine output after ICU admission, disease severity scores, and prognostic indicators were statistically significant (all P < 0.05). After regression analysis screening of these factors in the training set, eight indicators were ultimately included in the predictive model: intra-abdominal infection [odds ratio (OR) = 4.046, 95% confidence interval (95%CI) was 3.236-5.059, P < 0.001], vasopressin (OR = 2.188, 95%CI was 1.778-2.694, P < 0.001), CRRT (OR = 1.928, 95%CI was 1.506-2.469, P < 0.001), RDW (OR = 1.109, 95%CI was 1.069-1.151, P < 0.001), PLT (OR = 0.996, 95%CI was 0.995-0.997, P < 0.001), PT (OR = 1.042, 95%CI was 1.034-1.051, P < 0.001), lactate (OR = 1.154, 95%CI was 1.107-1.202, P < 0.001), and sequential organ failure assessment (SOFA; OR = 1.119, 95%CI was 1.089-1.150, P < 0.001). ROC curves demonstrated an area under the curve (AUC) of 0.830 (95%CI was 0.814-0.847) for the training set and 0.847 (95%CI was 0.825-0.869) for the validation set. Calibration curves demonstrated good agreement with the ideal curve in both the training and validation sets. DCA curve indicated that the model could effectively distinguish between patients with and without disease progression.

CONCLUSIONS: Intra-abdominal infection, vasopressin, CRRT, RDW, PLT, PT, lactate, and SOFA score are independent risk factors for SALI in patients with sepsis. The SALI prediction model constructed based on these factors demonstrates certain clinical predictive value.