Quick chronic liver failure-sequential organ failure assessment: an easy-to-use scoring model for predicting mortality risk in critically ill cirrhosis patients.

Abstract

Critically ill cirrhosis patients have an increased risk of morbidity and mortality, even after admission to the ICU. Our objectives were to compare the predictive accuracy of model for end-stage liver disease (MELD), MELD-Na, UK model for end-stage liver disease, and chronic liver failure-sequential organ failure assessment (CLIF-SOFA) by the development and validation of an easy-to-use prognostic model [named quick CLIF-SOFA (qCLIF-SOFA)] for early risk prediction in critically ill patients with cirrhosis. Overall, 1460 patients were extracted from the MIMIC-III database and enrolled in this study at 30-day and 90-day follow-up. qCLIF-SOFA was developed in the established cohort (n=730) and a performance analysis was completed in the validation cohort (n=730) using area under the receiver operating characteristic curve. Results were compared with CLIF-SOFA. The performance of CLIF-SOFA was significantly better than that of MELD, MELD-Na, and UK model for end-stage liver disease for predicting both 30-day and 90-day mortality (all P<0.05). qCLIF-SOFA consisted of five independent factors (bilirubin, creatinine, international normalized ratio, mean arterial pressure, and vasopressin) associated with mortality. In the established cohort, CLIF-SOFA and qCLIF-SOFA predicted mortality with area under the receiver operating characteristic curve values of 0.768 versus 0.743 at 30-day, 0.747 versus 0.744 at 90-day, and 0.699 versus 0.706 at 1 year, respectively (all P>0.05). A similar result was observed in the validation cohort (0.735 vs. 0.734 at 30 days, 0.723 vs. 0.737 at 90 days, and 0.682 vs. 0.700 at 1 year, respectively, all P>0.05). The utility of CLIF-SOFA was further shown to predict mortality for critically ill cirrhosis patients. The novel and simpler qCLIF-SOFA model showed comparable accuracy compared with existing CLIF-SOFA for prognostic prediction.