Analysis Of Influencing Factors And Establishment Of Estimation Model Of Minimal Hepatic Encephalopathy In Patients With Cirrhosis

Objective: To investigate the current situation of minimal hepatic encephalopathy(MHE)in patients with liver cirrhosis and to explore the influencing factors.Aim to apply data mining methods to analysis the factors associated with MHE and to establish estimation model for early discovery of patients with high risks to develop HE,and to lay the foundation for the implementation of targeted nursing intervention.Methods: Determine the observational indicators of this study by means of literature review and expert consultation,totally 300 patients with liver cirrhosis were selected from a tertiary hospital in Shanghai from June2017 to December 2018,and screened MHE by psychometric hepatic encephalopathy score,clinical data were analyzed by the logistic regression model to determine the risk factors of MHE.Establish the logistic regression model and artificial neural network model,validate the models in the validation group(104 cirrhosis patients included from January to November 2019).Results:1.The incidence of MHE was 48.67% in liver cirrhosis inpatients,14factors(age,etiology,diabetes,infection,renal insufficiency,nutritional risk,abdominal volume,Child-pugh,MELD score,hemoglobin,total serum bilirubin,serum albumin,INR,blood ammonia)are related to the occurrence of MHE.2.The results of logistic regression analysis: the probability of occurrence of MHE(Logit P)=-3.152+0.039×age+0.549×history of diabetes+0.683 concurrent infection+0.682×nutritional risk+ 0.848INR(≥1.5)+0.507 blood ammonia(>47μmol/L),the model was statistically significant(P<0.01),the Hosmer and Lemeshow test(P >0.05)indicated that the model fitted well.3.Established 2 kind of artificial neural network model,ANN17 with17 input layer variables,and ANN6 with 6 variables.ANN17 includes 2hidden layers,the number of nodes in hidden layer1 is 6,and 5 in hidden layer2,under the curve area AUC=0.808,sensitivity=72.5%,specificity=79.0%.ANN17 includes 2 hidden layers,the number of nodes in hidden layer 1 is 6,and 5 in hidden layer 2,under the curve area AUC=0.808,sensitivity=72.5%,specificity=79.0%.ANN6 includes 2hidden layers,the number of nodes in hidden layer 1 is 7,and 5 in hidden layer 2,under the curve area AUC=0.783,sensitivity=67.6%,specificity=73.6%.4.The validation group data were substituted into the three models,the AUC of logistic regression model was 0.636,and the AUC of ANN17 was 0.768,the difference was statistically significant(P<0.05).Conclusions: The incidence of minimal hepatic encephalopathy is high in patients with cirrhosis,the logistic regression analysis explained the risk factors clearly,and the artificial neural network model is better at disease screening,the advantages of different models should be synthetically evaluated based on the practical application needs.