Lasso-LR Model Identified as Best Predictor for AKI Mortality in Alcoholic Cirrhosis Patients

A recent study published in Frontiers in Medicine highlights the effectiveness of the Lasso-LR (least absolute shrinkage and selection operator-logistic regression) model in predicting in-hospital mortality among adult patients suffering from alcoholic cirrhosis complicated by severe acute kidney injury (AKI). Conducted by researchers led by Meina Sun from the First Affiliated Hospital of Jinzhou Medical University in China, the study used a retrospective analysis of 856 patients, with training and validation cohorts to ensure robustness.

The researchers developed and compared eight different machine learning models, utilizing feature selection through LASSO regression, to identify the most significant variables influencing mortality risk. The nine key features incorporated into the optimal model included total bilirubin, presence of acute respiratory failure, vasopressin use, septic shock, oliguria, AKI stage, lactate levels, transfusion of fresh frozen plasma, and norepinephrine administration.

The Lasso-LR model demonstrated superior predictive accuracy with an area under the receiver operating characteristic curve (AUROC) of 0.809, outperforming other models. Its effectiveness was further validated through external cohorts, with AUROCs of 0.836 and 0.809. To assist clinical decision-making, the team also developed static and web-based dynamic nomograms visualizing the model's predictions.

According to the study, "The Lasso-LR model exhibits robust predictive capability for in-hospital mortality among patients with alcoholic cirrhosis complicated by AKI, providing valuable insights and supporting personalized patient management." These findings suggest that integrating this model into clinical practice could enhance prognosis assessments and treatment strategies.

For more details, see the full study here: Frontiers in Medicine, 2025.