Lasso-LR Model Best for Predicting AKI Mortality in Alcoholic Cirrhosis

TUESDAY, May 27, 2025 (HealthDay News) — The least absolute shrinkage and selection operator-logistic regression (Lasso-LR) model is optimal for predicting in-hospital mortality for adult patients with alcoholic cirrhosis complicated by severe acute kidney injury (AKI), according to a study published online May 7 in Frontiers in Medicine.

Meina Sun, from the First Affiliated Hospital of Jinzhou Medical University in China, and colleagues developed and validated machine learning models to predict mortality risk in a retrospective analysis conducted on 856 adult patients with alcoholic cirrhosis complicated by severe AKI. A total of 627 patients from 2008 to 2016 were designated as the training cohort, while the temporal external validation cohort included 229 patients from 2017 to 2019. Eight distinct machine learning models were developed following feature selection conducted using LASSO regression. To determine the optimal model, performance was assessed using the area under the receiver operating characteristic curve (AUROC).

Nine key features were identified in LASSO regression: total bilirubin, acute respiratory failure, vasopressin, septic shock, oliguria, AKI stage, lactate, fresh frozen plasma transfusion, and norepinephrine. The Lasso-LR model achieved the highest AUROC value (0.809), establishing it as the optimal model. A static nomogram and web-based dynamic nomogram were developed for visualization purposes. For the training cohort and temporal external validation cohort, the AUROCs were 0.836 and 0.809, respectively, in the nomogram model. Decision curve analysis curves indicated certain clinical application value for the model.

“The Lasso-LR model exhibits robust predictive capability for in-hospital mortality among patients with alcoholic cirrhosis complicated by AKI, offering valuable prognostic insights and individualized treatment decision support for health care professionals,” the authors write.

Abstract/Full Text