Establishment and evaluation of a model for predicting 3-month mortality in Chinese patients with hepatic encephalopathy.

ABSTRACT

Hepatic encephalopathy (HE) is a serious complication of liver disease. To establish a model for predicting 3-month mortality in patients with HE in China. This retrospective study included 609 patients with HE admitted to the Peoples' Hospital, Liaocheng City, China (August 2006 to January 2016). Patients were allocated to a modeling (n = 409) or validation (n = 200) group. Demographic/clinical characteristics, laboratory test results, Model for End Stage Liver Disease (MELD) score and Child-Turcotte-Pugh (CTP) score were extracted from medical records. A model for predicting death within 3 months after admission was established using logistic regression analysis (modeling group). Model validity (validation group) was assessed using receiver operating characteristic (ROC) curve analysis. 270/409(66.0%) patients died in the modeling group and 142/203(70.0%) died in the validation group. Compared with survivors, patients who died had more severe HE, and higher MELD score, CTP score, incidence of complications including hepatorenal syndrome (HRS) and upper gastrointestinal bleeding, and values for laboratory parameters including red blood cell count(RBC) and total bilirubin(TBIL)(P < 0.05). Regression analysis revealed RBC, TBIL, HE stage, HRS and upper gastrointestinal bleeding as independent factors associated with death (P < 0.05). The area under the ROC curve (AUC) for the model was 0.931.The model had a higher Youden index than MELD or CTP scores and predicted death in the validation group with a sensitivity of 83.1% and specificity of 93.4%. The established model has superior performance to MELD and CTP scores for predicting mortality in patients with HE.