ABSTRACT INTRODUCTION Early assessment of prognosis following major abdominal surgery is associated with decreased risk of complications and death. While scoring systems are useful in this regard, there is no index that enables comprehensive individual patient assessment and is also applicable in ICUs with limited resources. OBJECTIVES Demonstrate that a model based on intra-abdominal pressure is effective in predicting death after major abdominal surgery. METHODS A prospective observational study was done of 300 post–abdominal-surgery patients admitted to the ICU of a university hospital affiliated with the General Calixto Garcia Medical Faculty, in January 2008 through January 2010. Patients were randomly assigned (2:1) to two groups: test and validation. The independent variable was vital status at discharge (alive or deceased); independent variables were age, sex, malignancy, APACHE II score and intra-abdominal pressure. In the test group, three mathematical models were fit to predict death (APACHE II, intra-abdominal pressure, and APACHE II plus intraabdominal pressure), which were later validated in the second group. Each model’s capacity to discriminate between living and deceased was evaluated according to sensitivity and specificity of receiver operating characteristic curves. Calibration was assessed with the Hosmer-Lemeshow goodness-of-fit test and comparison of receiver operating characteristic curves by chi-square test of homogeneity. Each patient was followed until hospital discharge or death. RESULTS The three mortality prediction models displayed excellent calibration and discrimination, very similar predictive power, and no differences among their respective areas under the curve (chi square 2.802, p = 0.094). Variables with the most influence on probability of death were age, APACHE II score and intra-abdominal pressure. CONCLUSIONS The three models show good capacity and similar effectiveness to predict death after major abdominal surgery. The model based on intra-abdominal pressure is the most feasible in limited-resource settings.