| Objectives The aim of this article is to establish a logistic regression model forpredicting the prognosis of patients with liver failure and in order to provide a guidancein options of treatment and improve the prognosis.Methods198patients with hepatic failure admitted to the department of infectiousdiseases in our hospital between January2006to April2013were enrolled in this study.Univariate and multivariate logistic regression analysis were used to explore theassociations between mortality and prognostic factors by establishing a regressionmodels. Linear regression was adopted for diagnosis of multicollinearity betweenprognostic factors. ROC curve was drawn to determine the optimal threshold of theprognostic model.Results Univariate logistic regression analysis showed that increased levels of serumalpha fetoprotein and total cholesterol (TC) were the protective factors for the patientswith liver failure; alderly patients, elevated levels of serum total bilirubin (TBIL) andurea, increased international normalized ratio (INR), occurrence of hepaticencephalopathy (HE), upper gastrointestinal bleeding and ascites were the risk factorsfor the prognosis. Prognostic factors included in the study had no multicollinearity. Inmultivariate logistic regression analysis, the independent factors predicting prognosiswere age (OR1＝7.207, OR2＝21.251, P＜0.001), serum TBIL (OR＝0.347, P＝ 0.002), serum TC (OR＝3.769, P＜0.001) and ascites (OR＝0.142, P＝0.002). TheLogistic regression model was P＝1／1＋e-y，Y＝－2.726＋1.975X21＋3.056X22－1.059X3＋1.327X7－1.952X10.The area under the ROC curve was0.893and thepredicted accuracy rate of the model was86.96％.Conclusions The main clinical types of hepatic failure in our hospital wereacute-on-chronic and chronic liver failure，mainly caused by hepatitis B virus infection.The regression model in our study suggest that the patients’ age, serum TBIL, TC andascites are useful for the judgment of prognosis in patients with liver failure. |
PDF Full Text Request