Sepsis has high mortality, early affects the liver, specialists in Intensive Care and Emergency tools used to predict patient outcomes. Objective: To design a scale to unify criteria for the diagnosis of the severity of liver damage from sepsis. Methods: Retrospective, observational, descriptive 200 deaths from sepsis study; liver histological patterns and epidemiological and clinical variables were determined and a scale was designed to assess the severity of liver injury in sepsis, which was validated internally. Results: liver damage was classified as severe and not severe, severe was three times higher in patients with non-white skin color and 17 times higher in patients with APACHE II (mortality risk) greater than 76% and abdominal infection; the best model identified severe liver damage with accuracy of 89.5%, sensitivity of 97.6% and specificity of 52.8% and the area under the receiver operating characteristic curve was 0.78, with a statistic significance associated with less than 0.01 calculated and predicted better that chance, with confidence interval from 0.71 to 0.85. Conclusions: The model had high sensitivity, better identified severe liver damage and predicted better than chance.