LiFe: a liver injury score to predict outcome in critically ill patients

Christin Edmark, Mark J. W. McPhail, Max Bell, Tony Whitehouse, Julia Wendon, Kenneth B. Christopher

To develop a liver function-related risk prediction tool to identify acute-on-chronic liver failure patients at greatest risk of in-hospital mortality. The LiFe (liver, injury, failure, evaluation) score, was constructed based on the opinions of 157 intensivists within the European Society for Intensive Care Medicine. Experts were surveyed and instructed to weigh the diagnostic importance of each feature of a proposed prediction model. We performed a retrospective cohort study of 1916 patients with chronic liver disease admitted to a medical or surgical ICU between 1997, and 2011 in three large hospitals in Boston, USA, and London, UK, with arterial lactate, total bilirubin and INR drawn at ICU admission. The derivation cohort consisted of ICU patients from Brigham and Women's Hospital and Massachusetts General Hospital in Boston (n = 945), and the validation cohort comprised patients from Kings College Hospital, London, admitted to the Liver Intensive Therapy Unit (n = 971). A clinical prediction model was derived and validated based on a logistic regression model describing the risk of in-hospital mortality as a function of the predictors (arterial lactate 0-1.9, ≥2.0-3.9, ≥4.0-5.9, ≥6.0 mg/dL; total bilirubin 0-1.9, ≥2.0-3.9, ≥4.0-5.9, ≥6.0 mg/dL; INR 0-1.9, ≥2.0-3.9, ≥4.0-5.9, ≥6.0) at ICU admission. Performance analysis of the LiFe score against SOFA, CLIF-SOFA, APACHE II and SAPS II was completed in the validation cohort of critically ill cirrhotic patients. The derivation cohort (n = 941) was 53 % male with a mean age of 65 years and an in-hospital mortality rate of 30 %. The validation cohort (n = 971) was 63 % male with mean age of 51 years and an in-hospital mortality rate of 52 %. The C statistic for the prediction model was 0.74 (95 % CI 0.70-0.77) in the derivation cohort and 0.77 (95 % CI 0.74-0.80) in the validation cohort. In the validation cohort, in-hospital mortality was 17 % in the low-risk group (0 risk score points), 28 % in the intermediate-risk group (1-3 points), 47 % in the high-risk group (4-8 points), and 77 % in the very high-risk group (>8 points). In the validation cohort, the C statistics for SOFA, CLIF-SOFA, APACHE II, and SAPS II were 0.80, 0.81, 0.77, and 0.78, respectively. Further, a significant positive correlation exists between LiFe score and acute-on-chronic liver failure grade, (r = 0.478, P < 0.001). Our LiFe score calculated from arterial lactate, total bilirubin and INR at ICU admission is a simple, quick and easily understandable score that may increase clinical utility for risk prediction in ICU patients with acute-on-chronic liver failure. The LiFe score can be used in place of physiological based scores for early risk prediction in patients with chronic liver disease but is not intended to replace CLIF-SOFA as a benchmark for prognostication.