Construction And Evaluation Of 28-Day Survival Prediction Model Of Liver Cirrhosis Patients In Intensive Care Unit

Objective: Liver cirrhosis refers to various chronic liver diseases that lead to hepatocyte necrosis,diffuse proliferation of liver fibrous tissue,and then form regenerative nodules and pseudolobules,resulting in the destruction of normal hepatic lobule structure and blood supply.With the progress of the disease,many complications can occur in the late stage,including ascites,rupture and bleeding of esophageal and gastric varices,hepatic encephalopathy,hepatorenal syndrome,hepatopulmonary syndrome,hyponatremia,infection,cirrhotic cardiomyopathy and so on.Patients with liver cirrhosis are admitted to ICU due to various complications in the later stage,and the mortality is high.Therefore,early evaluation of the severity and mortality of patients is of great significance to the treatment and prognosis of patients.The purpose of this study was to establish a nomogram to predict the 28 day survival rate of cirrhotic patients in intensive care unit.Methods: Based on the American medical information mart for intensive care(mimic),the mimic-III database collected the demographic information,general vital signs,drug use,laboratory indicators,nursing staff’s medical order records,fluid balance,and other information of more than 40000 patients admitted to the intensive care unit from 2001 to 2012 Diagnosis code,imaging report,length of stay in hospital,survival data,etc.1265 patients diagnosed with liver cirrhosis were selected from the database,and their gender,age,marital status,weight,social security type,complications,electrolyte,hemoglobin,leukocyte count,platelet count,albumin,blood creatinine,blood urea nitrogen,lactic acid Anion gap,alkali residue,systolic blood pressure,respiratory rate,heart rate,urinary specific gravity,survival outcome and other indicators.Through single factor regression and multi factor Cox regression,the indexes with better prediction efficiency are selected.The prediction model was constructed by Cox regression.The 28 day survival rate of patients with liver cirrhosis was predicted by the model,and the differentiation,calibration and clinical effectiveness of the model were tested respectively.SPSS and R software were used for statistical analysis.Results:1.Through univariate analysis and multivariate Cox regression analysis,10 statistically significant clinical indexes were selected: organ dysfunction(HR:1.4536,95%CI: 1.1340～1.8633,P<0.01)、SASP II(HR:1.0209,95%CI: 1.0124～1.0294,P<0.001)、Lac(HR:1.0576,95%CI:1.0142 ～1.1028,P<0.01)、AST(HR:1.0001,95%CI:1.0000～1.0002,P=0.0225)、BUN(HR:1.0047,95%CI:1.0001～1.0094,P=0.045)、PT(HR:1.0138,95%CI:1.0030～1.0247,P=0.0123)、PTT(HR:1.0050,95%CI:1.0005～1.0095,P=0.028)、Age(HR:1.0038,95%CI:1.0003～1.0073,P=0.0345)、Mean BP(HR:0.9832,95%CI:0.9743～0.9922,P<0.001)、Resp Rate Mean(HR:1.0254,95%CI:1.0050～1.0461,P=0.0145)、Alb(HR:0.7260,95%CI:0.6235～0.8452,P<0.001).Using these indexes to construct a 28-day survival prediction model.2.Model evaluation results:(1)discrimination: training set(AUC = 0.673)and test set(AUC = 0.71),indicating that the model has good discrimination efficiency.(2)calibration: the calibration curve shows that the predicted value is in good agreement with the observed value.(3)Clinical effectiveness: both training set and test set decision curves have net benefits.Conclusion:In this study,Cox regression method was used to establish and verify a valuable prediction model of 28 days survival of liver cirrhosis.The results have certain guiding significance for clinical practice and can evaluate the prognosis of cirrhotic patients in ICU.