Established Clinical Information Techonlogy System For Liver Failure Patients And A Prognosis Model For Acute-on-chronic Liver Failure (ACLF)

Background:With the coming of the information age, better use of information technology to optimize clinical processes and to mine clinical data has become an important direction of today’s medicine. Acute-on-chronic liver failure (ACLF) is one of the most deadly, prevalent, and costly clinical syndromes in Asia. However, no prognostic model has been developed that is based specifically on non-liver-transplant data gathered from Asian patients with ACLF. The aim of the present study was to make full use of information technology to build a consistent clinical research database for more efficient data collation and classified analysis. And developing a prognostic model to estimate the probability of death related to ACLF and pre-ACLF patients not undergoing liver transplant, which is based on the above database.Methods:After carding and planning the clinical diagnosis and treatment process on liver failure, we established a clinical research system on liver failure, which included clinical pathway and HER, multi-center clinical trial electronic data capture systems (electronic data capture, EDC), the follow-up system of liver failure, the artificial liver treatment and follow-up system. Using the system, we conducted a retrospective observational cohort study to analyze clinical data from857patients with ACLF/pre-ACLF, who were hospitalized in the First Affiliated Hospital of Zhejiang University from1December2008to1February2012and did not undergo liver transplantation. The enrolled patients were divided into subgroups according to the extent of the disease and cirrhosis. The actual survival of each group was analyzed by Kaplan-Meier curves. And Kaplan-Meier and Cox proportional hazards regression model were used to estimate survival rates and survival affected factors to build a prognostic model. The area under the receiver operating characteristic curve (auROC) was used to evaluate the performance of the model for predicting early mortality.Results:The mortality rates among patients with pre-ACLF at12weeks and24weeks after diagnosis were30.5%and33.2%, respectively. The mortality rates among patients with early-stage ACLF at12weeks and24weeks after diagnosis were33.9%and37.1%, respectively. The mortality rates among patients with intermediate-stage ACLF at12weeks and24weeks after diagnosis were49.5%and53.8%, respectively. The mortality rates among patients with late-stage ACLF at12weeks and24weeks after diagnosis were77.2%and78.5%, respectively. The difference in survival between the pre-ACLF patients and the early-stage ACLF patients was not statistically significant (P>0.05). The mortality rates of the pre-ACLF patients within12weeks and24weeks were lower than the mortality rates of the intermediate-stage ACLF patients. The difference in survival between the pre-ACLF patients and the intermediate-stage ACLF patients was statistically significant (P<0.0001). The mortality rates of the intermediate-stage ACLF patients within12weeks and24weeks were lower than the mortality rates of the late-stage ACLF patients. The difference in survival between the intermediate-and late-stage ACLF patients was statistically significant (P<0.0001).The actual survival of all patients with cirrhosis (n=455) was compared to that of patients without cirrhosis (n=402). The mortality of patients with cirrhosis was63.1%within12weeks and65.5%within24weeks. The mortality of patients without cirrhosis was45.5%within12weeks and46.5%within24weeks. The difference in survival of12weeks and24weeks between the above two groups were statistically significant (P <0.0001).Bivariate analysis and COX analysis showed five independent factors to be associated with survival among patients with ACLF and pre-ACLF; namely, MELD score, age, hepatic encephalopathy, triglyceride level and platelet count. The mortality increases as the exit of hepatic encephalopathy, the increase of MELD score/age/triglyceride levels, and the decrease of platelet count levels. We established the Li-ACLF model in which the risk scores (R) for individual patients can be calculated by combining their5prognostic values with the regression coefficients. That is, R=0.021×(age in years)+0.279x(hepatic encephalopathy score)+0.513×(MELD score)-0.210xloge (platelet count109/L)-0.176×loge (triglyceride levels mg/dL)Conclusion:We established a consistent clinical research database and solved the problem of data consistency, which optimized the clinical workflow and made more efficient use of clinical analytical data. On this basis, we concluded that pre-ACLF patients had a poor prognosis as early-stage ACLF patients did. So it is recommended that the original2006Chinese diagnostic criteria of ACLF should be broadened and the pre-ACLF patients should receive system diagnosis and treatment as soon as possible. The Li-ACLF model can be used to predict ACLF-related death of patients in non-liver transplant period to determine whether an emergency liver transplant therapy was needed, which may be helpful for improving the prognosis of patients.