Development and validation of a nomogram for early prediction of sepsis-associated liver injury: a multicenter study.

Background: Sepsis-associated liver injury (SALI) is an important manifestation of sepsis-associated organ dysfunction and is associated with adverse clinical outcomes. However, practical tools for estimating the risk of subsequent SALI in critically ill patients with sepsis remain limited. This study aimed to develop and geographically validate a nomogram based on routinely available clinical variables using multicenter retrospective data. Methods: Sepsis patients admitted to the ICUs of Hebei General Hospital and Handan Central Hospital, East District, between March 1, 2021, and June 1, 2024, were retrospectively enrolled. A total of 847 eligible patients from these two centers constituted the model development and internal validation cohort and were randomly divided into a training cohort and a testing cohort at a 6:4 ratio. An independent cohort of 424 sepsis patients admitted to the ICU of Chongqing General Hospital between March 1, 2021, and January 1, 2026, was included for external geographic validation. Demographic characteristics, comorbidities, laboratory parameters, and clinical interventions were collected. Candidate predictors were initially screened using univariable analysis and least absolute shrinkage and selection operator (LASSO) regression, followed by multivariable logistic regression to derive the final prediction model for subsequent in-hospital SALI. A nomogram was constructed using the rms package. Model discrimination was assessed using receiver operating characteristic (ROC) curves and the area under the curve (AUC), calibration was evaluated using calibration plots and quantitative calibration metrics, and potential decision-analytic value was assessed using decision curve analysis (DCA). Results: A total of 1,271 patients with sepsis from three centers were included in the final analysis. The development and internal validation cohort comprised 847 patients, of whom 352 developed SALI (41.6%), and was randomly divided into a training cohort of 508 patients and a testing cohort of 339 patients. The independent external geographic validation cohort comprised 424 patients, of whom 183 patients developed SALI (43.2%). After feature selection and multivariable analysis, age, direct bilirubin, aspartate aminotransferase, prothrombin time-international normalized ratio, and vasoactive-drug use were retained as predictors associated with subsequent in-hospital SALI. The nomogram demonstrated favorable discrimination, with AUCs of 0.925 (95% CI, 0.901-0.948), 0.918 (95% CI, 0.888-0.948), and 0.884 (95% CI, 0.849-0.918) in the training, testing, and external geographic validation cohorts, respectively. Calibration was close to the ideal values in the training cohort but deteriorated in the testing and external geographic validation cohorts, particularly in the external cohort, in which the calibration-in-the-large was 0.945 and the calibration slope was 0.600. Decision curve analysis suggested potential net benefit within an illustrative threshold-probability range of 0.10-0.40. Conclusion: The nomogram demonstrated favorable discrimination across the training, testing, and external geographic validation cohorts, although calibration deteriorated outside the training cohort. The model may provide a practical approach for estimating subsequent in-hospital SALI risk and identifying patients who may warrant closer clinical assessment. However, local recalibration, prospective validation, and clinical-impact evaluation are required before application in new clinical settings. This retrospective prediction study does not establish that implementation of the nomogram guides effective interventions, improves patient prognosis, or optimizes ICU resource allocation.