Predictors of Mortality in the Critically Ill Cirrhotic Patient: Is the Model for End-Stage Liver Disease Enough?

Alagappan Annamalai, Megan Y. Harada, Melissa Chen, Tram Tran, Ara Ko, Eric J. Ley, Miriam A Nuno, Andrew Klein, Nicholas Nissen, Mazen Noureddin

Research output: Contribution to journalArticle

9 Scopus citations

Abstract

Background Critically ill cirrhotics require liver transplantation urgently, but are at high risk for perioperative mortality. The Model for End-stage Liver Disease (MELD) score, recently updated to incorporate serum sodium, estimates survival probability in patients with cirrhosis, but needs additional evaluation in the critically ill. The purpose of this study was to evaluate the predictive power of ICU admission MELD scores and identify clinical risk factors associated with increased mortality. Study Design This was a retrospective review of cirrhotic patients admitted to the ICU between January 2011 and December 2014. Patients who were discharged or underwent transplantation (survivors) were compared with those who died (nonsurvivors). Demographic characteristics, admission MELD scores, and clinical risk factors were recorded. Multivariate regression was used to identify independent predictors of mortality, and measures of model performance were assessed to determine predictive accuracy. Results Of 276 patients who met inclusion criteria, 153 were considered survivors and 123 were nonsurvivors. Survivor and nonsurvivor cohorts had similar demographic characteristics. Nonsurvivors had increased MELD, gastrointestinal bleeding, infection, mechanical ventilation, encephalopathy, vasopressors, dialysis, renal replacement therapy, requirement of blood products, and ICU length of stay. The MELD demonstrated low predictive power (c-statistic 0.73). Multivariate analysis identified MELD score (adjusted odds ratio [AOR] = 1.05), mechanical ventilation (AOR = 4.55), vasopressors (AOR = 3.87), and continuous renal replacement therapy (AOR = 2.43) as independent predictors of mortality, with stronger predictive accuracy (c-statistic 0.87). Conclusions The MELD demonstrated relatively poor predictive accuracy in critically ill patients with cirrhosis and might not be the best indicator for prognosis in the ICU population. Prognostic accuracy is significantly improved when variables indicating organ support (mechanical ventilation, vasopressors, and continuous renal replacement therapy) are included in the model.

Original languageEnglish (US)
Pages (from-to)276-282
Number of pages7
JournalJournal of the American College of Surgeons
Volume224
Issue number3
DOIs
StatePublished - Mar 1 2017
Externally publishedYes

    Fingerprint

ASJC Scopus subject areas

  • Surgery

Cite this