A nomogram model based on albumin-bilirubin score for predicting 90-day prognosis in patients with acute-on-chronic liver failure

Objective: To develop a nomogram model based on the albumin-bilirubin (ALBI) score for predicting the 90-day prognosis of patients with acute-on-chronic liver failure (ACLF) and to evaluate its predictive efficacy.

Methods: Clinical data of 290 ACLF patients at the Third People's Hospital of Nantong City, collected from December 2020 to December 2023, were analyzed. The data were divided into a training set (n = 200) and a validation set (n = 90), with August 2022 as the cut-off date. Patients in the training set were categorized into an improvement group (n = 133) and a mortality group (n = 67) based on their 90-day outcomes. The predictive power of baseline parameters was assessed using univariate and multivariate logistic regression to construct model. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA) and the Hosmer-Lemeshow test.

Results: Creatinine (CR) [odds ratio (OR) = 1.013, 95% confidence interval (CI): 1.004-1.022], ALBI (OR = 10.831, 95% CI: 4.009-33.247), Gender (OR = 1.931, 95% CI: 0.973-3.870) and ascites (OR = 3.032, 95% CI: 1.249-8.178) were identified as independent prognostic factors. The prognostic model formula was derived as prognostic index (PI) = -0.591 + 0.658 × Gender + 1.109 × ascites + 0.012 × CR + 2.382 × ALBI. The area under the curve (AUC) was 0.804 (95% CI: 0.741-0.866), with a specificity of 85.0% and a sensitivity of 65.7% at a cut-off of 0.425. The AUC of the validation set was 0.811 (95% CI: 0.697-0.926). The Hosmer-Lemeshow test indicated a good model fit with a p-value of 0.287 for the training set and 0.423 for the validation set. Calibration curves demonstrated the accuracy of the model, and DCA results suggested that the model was clinically useful when the threshold was below 0.6.

Conclusion: The nomogram model incorporating ALBI with CR, Gender and ascites can predict the 90-day prognosis of ACLF patients, potentially helping to optimize treatment strategies and improve patient outcomes.