823 Epigenetic Regulation of Gene Expression in NAFLD

Abstract

Background: Efforts to reduce healthcare costs have brought attention to high readmission rates following hospitalization for chronic diseases. Although short term readmission rate in cirrhosis patients were recently reported, information on readmission rates beyond 30 days is lacking. Our aim was to determine the risk and factors for shortand intermediateterm readmissions at a large tertiary care center. Methods: We retrospectively reviewed medical records of all patients with an established diagnosis of cirrhosis admitted to the primary University of Pittsburgh Medical Center Hospital between May 2008 and May 2009. Information on demographics, etiology, reason for admission and readmission (cirrhosisrelated vs. other causes), MELD score, length of stay, and outcome (death, liver transplant) were noted. We calculated time to first readmission and readmission rates at 30, 60, 90, and 120 days. Multivariate Cox proportional hazard modeling was performed to determine predictors of time to first hospital readmission. Results: A total of 798 cirrhotic patients (mean age 57, 56% male, 86% white) had 1923 total admissions during the study period. During amedian follow up of 6months (interquartile range= 76, 282), at least one readmission was noted in 426 (53.4%). The rates of readmissions at 30, 60, 90, and 120 days were 29%, 37%, 42%, and 46% respectively. Among those with cirrhosis-related index admissions, the rates were significantly higher at 41%, 52%, 58%, and 64%, with 74% being readmitted over the study period. In univariate analyses, significant predictors (p<0.05) of readmission included a higher MELD score (19 vs. 12), cirrhosis-related reason for index admission (73% vs. 30%) and longer initial length of stay (6 vs. 5 days). On multivariate Cox regression analysis, MELD score (HR for each point increase =1.1, p<0.0001) and cirrhosis-related index admission (HR=2.09, p<0.0001) (Figure 1) were the significant predictors of readmission. A relatively small proportion of the study population (90 patients; 11%) accounted for a surprisingly large percentage of total hospital readmissions (685; 35%). Significant predictors for 5 or more readmissions included non-white race, higher MELD score, and cirrhosisrelated index admission. Conclusions: In this cohort, a relatively small percentage of patients accounted for a disproportionately large number of readmissions. MELD score and cirrhosisrelated complications at the index admission are strong predictors of subsequent hospital readmission. Studies directed at understanding the specific causes of readmissions in these patients will help design interventions to reduce readmissions rates in cirrhosis patients.