INTRODUCTION:: Few data can be found about cardiac arrest in the intensive care unit outside reference centers in third world countries. OBJECTIVE:: To study epidemiology and prognostic factors associated with cardiac arrest in the intensive care unit (ICU) in an average Brazilian center. METHODS:: Between June 2011 and July 2014, 302 cases of cardiac arrest in the intensive care unit were prospectively evaluated in 273 patients (age: 68.9 ± 15 years) admitted in three mixed units. Data regarding cardiac arrest and cardiopulmonary resuscitation were collected in an "Utstein style" form and epidemiologic data was prospectively obtained. Factors associated with do not resuscitate orders, return of spontaneous circulation and survival were studied using binary logistic regression. Statistical package software used was SPSS 19.0 (IBM Inc., USA). RESULTS:: Among 302 cardiac arrests, 230 (76.3%) had their initial rhythm recorded and 141 (61.3%) was in asystole, 62 (27%) in pulseless electric activity (PEA) and 27 had a shockable rhythm (11.7%). In 109 (36.1%) cases, cardiac arrest had a suspected reversible cause. Most frequent suspected cardiac arrest causes were hypotension (n=98; 32.5%), multiple (19.2%) and hypoxemia (17.5%). Sixty (19.9%) cardiac arrests had do not resuscitate orders. Prior left ventricle dysfunction was the only predictor of do not resuscitate order (OR: 3.1 [CI=1.03-9.4]; P=0.04). Among patients that received cardiopulmonary resuscitation, 59 (24.4%) achieved return of spontaneous circulation and 12 survived to discharge (5.6%). Initial shockable rhythm was the only return of spontaneous circulation predictor (OR: 24.9 (2.4-257); P=0.007) and survival (OR: 4.6 (1.4-15); P=0.01). CONCLUSION:: Cardiopulmonary resuscitation rate was high considering ICU patients, so was mortality. Prior left ventricular dysfunction was a predictor of do not resuscitate order. Initial shockable rhythm was a predictor of return of spontaneous circulation and survival.