OBJECTIVE To develop a nomogram model for prediction of the risk of sepsis in the patients with acute liver failure (ALF) and validate its clinical value.

METHODS Totally 228 patients with ALF who were treated in the Fifth Medical Center of Chinese PLA General Hospital from Jan. 2009 to Mar. 2023 were recruited as the research subjects and were grouped according to the occurrence of sepsis, the clinical characteristics and results of laboratory tests were collected from the patients. The subjects were brought into multivariate logistic regression analysis after the primary screening with univariate regression analysis, the independent predictive factors were screened out, and the nomogram model was established. The accuracy, calibration accuracy and clinical practicability of the model were assessed by means of receiver operating characteristic (ROC) curves, calibration curves and decision curve analysis (DCA).

RESULTS Among the 228 patients with ALF, 159 (69.74%) were diagnosed with sepsis. Totally 6 independent predictive factors were screened out by the multivariate analysis, including age odd ratio(OR)=1.098, 95%CI: 1.030 to 1.220, aspartate transaminase(OR=0.998, 95%CI: 0.996 to 0.999), white blood cells counts(OR=1.037, 95%CI: 1.020 to 1.064), hemoglobulin(OR=0.981, 95%CI: 0.962 to 0.998), lactic acid(OR=1.187, 95%CI: 1.022 to 1.426) and mechanical ventilation(OR=3.463, 95%CI: 2.340 to 5.125). The area under the curve (AUC) of the nomogram model was 0.864(95%CI: 0.807 to 0.921) in the training set, 0.817(95%CI: 0.717 to 0.918) in the validation set, remarkably better than that of sequential organ failure assessment (SOFA) scores 0.710(95%CI: 0.625 to 0.795) and 0.647(95%CI: 0.515 to 0.779). Hosmer-Lemeshow test(P=0.512) and the calibrated curves showed that the predication probability of the model was highly consistent with the actual risk, and DCA indicated that the net clinical benefit was brought more from the model than from SOFA score.

CONCLUSION The age, hemoglobulin, aspartate transaminase, white blood cells counts, lactic acid and mechanical ventilation are the independent predictive factors for the sepsis in the ALF patients. The nomogram model established based on the factors has high predictive efficiency and clinical application value.