Establishment Of A New Prediction Model For Short-term Prognosis Of Acute-on-chronic Liver Failure

Objective:To retrospectively analyze the influencing factors of short-term poor prognosis in patients with Acute-on-chronic liver failure（ACLF）,and to construct a new and efficient short-term prognosis prediction model for ACLF according to the influencing factors of short-term prognosis of ACLF.Methods:The clinical data of hospitalized patients with acute-on-chronic liver failure diagnosed in the First Hospital of Lanzhou University from July 2017 to July 2022were collected.A total of 238 patients were finally included after inclusion and exclusion criteria screening.Among them,125 ACLF patients hospitalized from July2017 to December 2020 were selected as the modeling group,and 113 ACLF patients hospitalized from January 2021 to July 2022 were selected as the validation group.The survival of patients 90 days after discharge was determined according to the results of telephone and outpatient follow-up.The general clinical data,laboratory indicators and complications of ACLF patients were collected from the hospital case system.The Child-Turcotte Pugh（CTP）score,model for end-stage liver disease（MELD）and its derived MELD-Na score,and integrated MELD（i MELD）score were calculated according to the corresponding indicators.Univariate and multivariate Logistic regression methods were used to analyze the prognostic factors of ACLF patients and establish a prediction model.Hosmer-Lemeshow test combined with calibration curve was used to evaluate the model calibration.By drawing the receiver operating characteristic（ROC）curve of the new prediction model,the area under the curve（AUC）was calculated to evaluate the discrimination of the model.The area under the ROC curve was compared using the De Long test,net reclassification improvement（NRI）and integrated discrimination improvement（IDI）were calculated to evaluate the prediction performance of the model.Decision-curve analysis（DCA）was used to evaluate the clinical application prospect of the new model.Results:1.The etiology of liver underlying diseases in 238 patients with ACLF were analyzed.The main etiology of ACLF patients in both the modeling group and the validation group was viral hepatitis B,and the proportion of etiology between the two groups was not statistically significant（X²=19.666,P=0.058）.2.The comparison of clinical data in general between the modeling group and the validation group was performed,and there were no statistical significant differences in age,gender share,BMI,base of cirrhosis,and incidence of related complications between the two groups（P>0.05）.Compared with the modeling group,the laboratory indexes of TBIL,PT,INR,APTT,TT,CTP score,MELD score,and MELD-Na score in the validation group were lower（P<0.05）,and MPV and CL were higher（P<0.05）.There was no significant difference in the remaining indicators between the two groups（P>0.05）.3.In the modeling group,univariate and multifactorial Logistic regression analysis showed that MLR（OR=5.746,95%CI:1.279 to 25.808）and BUN（O R=1.779,95%CI:1.302 to 2.432）were independent risk factors for 90-day m ortality in ACLF patients;MPV（OR=0.431,95%CI:0.264 to 0.703）,ALB（O R=0.804,95%CI:0.696 to 0.928）,and TG（OR=0.239,95%CI:0.119 to 0.480）were independent protective factors for 90-day mortality in patients with AC LF.A new prognostic model combining various indicators:P_pre=13.915-0.842×MP V（f L）+1.749×MLR-0.219×ALB（g/L）+0.576×BUN（mmol/L）-1.430×TG（mmol/L）.4.In the modeling group,the Hosmer-Lemeshow test X²=6.247,P=0.620,and the calibration curve showed that the actual value and the predicted value were in good agreement;The ROC curve showed that the AUC of the new prognostic model was 0.928,and the 95%CI was 0.868 to 0.967.In the validation group,Hosmer-Lemeshow test X²=7.849,P=0.448;ROC curve shows that the AUC of the new model is 0.801,95%CI was 0.715 to 0.870.5.Comparing the difference of AUC,NRI and IDI in the modeling group and validation group,it was suggested that the new model P_prehad better predictive efficacy than CTP,MELD,MELD-Na,and i MELD scores.6.The DCA curves were drawn in the modeling group and verification group,which showed that the new model had higher clinical practical value than CTP,MELD,Meld-NA and i MELD scores,and had better clinical application prospects.Conclusions:1.MLR and BUN are independent risk factors for 90-day death of patients with ACLF;MPV,ALB and TG are independent protective factors for the 90-day death of patients with ACLF.2.The new model built by combining independent influencing factors has good predictive performance and better clinical prospects...