Development and Validation of a Prognostic Nomogram for Predicting 3-Month Mortality After Liver Transplantation

Abstract

Background and aims

Identifying patients at high risk of mortality following liver transplantation and implementing timely interventions are essential to reducing fatality rates. This study aimed to develop a prognostic model to predict mortality after liver transplantation, using a nomogram based on specific risk factors.

Methods

The study included 1951 liver transplant recipients, divided into a training group (n = 1366) and a validation group (n=585) through stratified cluster sampling. Nine key factors were considered for the prognostic model: Child-Pugh score, MELD score, extended Intensive Care Unit stays over seven days, catheter-related bloodstream infections, delayed graft function, intraperitoneal hemorrhages, pulmonary infections, renal failure, and vascular complications. Lasso regression and Boruta feature selection were applied to identify the most important predictors. The model's effectiveness was evaluated using ROC curves, calibration plots, decision curve analysis, and internal validation.

Results

The model achieved an area under the curve of 0.824 in the training group and 0.774 in the validation group. Calibration plots showed good alignment with the ideal diagonal line, indicating robust predictive accuracy. Decision curve analysis demonstrated various levels of clinical benefit across a range of risk thresholds (3%-81%).

Conclusions

This nomogram, despite its limitations from the retrospective cohort, proves to be a valuable tool for predicting mortality outcomes in liver transplant recipients. It helps healthcare professionals identify high-risk patients who require urgent interventions following transplantation.