To describe patients admitted to intensive care unit (ICU) for hypothermia, evaluate prognostic factors, and test the hypothesis that patients found indoors have a worse outcome. Retrospective clinical investigation in a medical ICU. Eighty-one consecutive patients admitted to ICU, with a body temperature of 35 degrees C or lower and rewarmed passively or with minimally invasive techniques, over a 17-year period. Patients were analyzed by age, gender, and causes of hypothermia and split into two groups (indoors and outdoors), according to the location where hypothermia occurred. Prognostic factors were determined by univariate method and stepwise logistic regression. The major complications were acute renal failure (43 %), aspiration pneumonia (22 %), rhabdomyolysis (22 %), and acute respiratory distress syndrome (12%). Principal comorbidities in the outdoor patients (21%) were alcohol and drug intoxication, and those in the indoor patients (79 %) were sepsis and neuropsychiatric disorders. Stepwise logistic regression identified two variables predictive of death: illness severity at admission (SAPS II > or = 40) and the location where hypothermia occurred (indoors versus outdoors). With equivalent body temperature, patients found indoors were more severely affected and died more frequently than those found outdoors.