A Machine Learning-Based Model for Cirrhosis Risk Stratification Incorporating Noninvasive Markers and Clinical Variables.

BACKGROUND

This study aimed to develop and validate a nomogram model integrating routine laboratory parameters, non-invasive fibrosis markers, and liver elastography parameters for cirrhosis risk stratification in patients with chronic liver disease.

METHODS

A total of 344 patients with chronic liver disease were retrospectively enrolled and randomly divided into a training set (n=241) and a validation set (n=103) in a 7:3 ratio. Independent predictors were identified using univariate analysis, LASSO regression, and multivariate logistic regression. Machine learning algorithms, including Random Forest, Support Vector Machine, and Logistic Regression were constructed using these predictors. Internal validation was performed using the Bootstrap method. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA).

RESULTS

Multivariable logistic regression identified age, platelet count, aspartate aminotransferase/alanine aminotransferase (AST/ALT) ratio, total bilirubin, abnormal international normalized ratio (>1.2), and liver stiffness measurement as independent predictors of cirrhosis. The Random Forest model demonstrated slightly superior performance, with AUCs of 0.835 (95% CI: 0.767-0.904) and 0.745 (95% CI: 0.597-0.893) in the training and validation sets, respectively. The calibration curves demonstrated good consistency between the predicted probabilities and the actual risks (Hosmer-Lemeshow test, P > 0.05). Decision curve analysis indicated that the Random Forest model provided a superior net clinical benefit across a threshold probability range of 0.1-0.3.

CONCLUSION

This study developed and validated a cirrhosis risk prediction model. The Random Forest model offers marginally better accuracy, while the nomogram provides a simple, interpretable tool for bedside clinical use. With further external validation, this model could potentially assist clinicians in stratifying cirrhosis risk among patients with chronic liver disease.