PTU-039 Mortality and utility of prognostic scoring models in cirrhotic patients admitted to a tertiary non-transplant intensive care unit (ICU) in the UK

Introduction Patients with liver cirrhosis admitted to ICU have a poor prognosis with reported ICU, in-hospital and 1-year mortality rates of 37%>65%, 49–74% and 69–81% respectively. Prognostic models may enable accurate discrimination at admission of those who will benefit from ICU admission. Organ failure scoring models have been shown to be good predictors of ICU mortality, but there are few reports of their utility in non-transplant centres in Europe. We aimed to describe mortality and the prognostic value of commonly used predictors in patients with cirrhosis admitted to the ICU of a tertiary non-transplant London teaching hospital ICU. Methods Retrospective analysis of the Intensive Care National Audit and Research Centre database of all patients with cirrhosis admitted to ICU from January 2006 to November 2010. Statistical analysis was performed with SPSS V. 19. The area under receiver operating curve (AUROC) was used to assess the prognostic accuracy of the scoring models. Results 135 out of 4890 (2.8%) admissions had diagnosed cirrhosis. Mean age was 53. 77% were white, 16% were Asian, and 7% were black. 74% were male. Cause of cirrhosis was alcohol in 70.4%, HCV in 9.6%, HBV in 4.4% and NASH in 5.2%. Admission was for sepsis in 34% of patients, GI bleeding in 24%, and encephalopathy in 12%. Mortality was 39% in ICU, 55% at 30 days and 59% at 1 year. 30-day mortality in those who required renal replacement therapy (RRT), advanced cardiovascular support, mechanical ventilation or CPR was 75%, 65%, 58% and 70% respectively. Overall all-cause ICU mortality was 19.4% for this period. Conclusion Mortality in this series compares favourably with published rates, particularly at 1 year. Mortality in patients who require CPR or RRT is high at 70% and 75%, but is appreciably better than the 100% and 89% mortality previously reported. SOFA score has the best predictive value for mortality, but is not strongly predictive. All predictive scores have a worse AUROC for mortality than previously seen suggesting a reduced prognostic validity in the non-transplant UK setting. Competing interests None declared.