Study design: a descriptive, prospective cohort study. Objectives: identify the risk factors associated with SAPS 3 and APACHE II score to improve mortality prediction in elderly admitted in the intensive care unit (ICU). Methodology: a cohort study with 174 elderly patients admitted in ICU from July 2016 to June 2017. The primary independent variables analyzed were: presence of stress hyperglycemia, the need for invasive mechanical ventilation (IMV), acute renal failure, level of consciousness and use of vasopressors. The outcomes studied were death and hospital discharge. Results: ICU and hospital mortality were 17.8% and 29.8%, respectively. The multivariate analysis of factors associated with hospital death not contemplated in the obtaining of SAPS 3 demonstrated that IMV, Chronic Obstructive Pulmonary Disease (COPD) and Mean Arterial Pressure (MAP) <70 mmHg improved the prediction of mortality. The Hosmer-Lemeshow test (p=0.472) and area under ROC curve=0.80 (CI95%:0.731-0.872) confirmed, respectively, an appropriate calibration and acceptable-to-excellent discriminatory power for this model. IMV and COPD were also factors associated with the APACHE II score in prediction of mortality with appropriated calibration (p=0.471) and acceptable-to-excellent discriminatory power (area under ROC curve=0.80 [CI95%:0.730-0.868]). Conclusion: IMV, COPD, and MAP <70 mmHg were associated with SAPS3 and APACHE II to predict mortality in elderly admitted to ICU.