Cirrhotic patients who need critical care support show high morbidity and mortality rates compared with other critically ill patients. Their prognosis is, in fact, influenced by both the severity of the underlying hepatic disease and the worsening of extrahepatic organ function. Clinicians and investigators have been persistently looking for objective scoring systems capable of providing accurate information on disease severity and short-term prognosis. Risk stratification helps differentiate patients who would not benefit from admission to the intensive care unit (ICU) from those who could achieve better outcomes once aggressively treated. The most common scores, ie, multiple organ dysfunction score, sequential organ failure assessment, and acute physiology and chronic health evaluation, developed in general ICUs to evaluate illness severity, have also been validated to predict the prognosis of cirrhotic patients admitted to the ICU. However, their absolute predictive value has been questioned. A weakness of common prediction models consists in not recognizing the continuum of physiological changes in critically ill decompensated cirrhotic patients. In addition, the predictive power to stratify individual risk is relatively low due to the great variability of liver dysfunction stages, the severity of related manifestations, and the number of nonfunctioning organs on admission. Probability models are not capable of predicting whether a patient will live or die with 100% accuracy, nor can they deny or confirm the indications for mechanical ventilation, vasopressor support or renal replacement therapy, or help to decide when to withhold or withdraw support. Because there are no absolute criteria to predict which cirrhotic decompensated patients will improve with normalization of organ function or deteriorate progressively, a scoring system should be regarded as an adjunct rather than a substitute for clinical judgment in the decision process concerning whether a patient should be admitted to the ICU.
Copyright © 2011 Elsevier Inc. All rights reserved.