Higher mortality and morbidity are well established in children with malignancies in whom intensive care admissions are required. A retrospective cohort study was conducted to assess the risk factors for children with cancer in the pediatric intensive care unit (PICU) for short-term outcome (survival vs. nonsurvival when leaving the PICU). The records of 36 children with a median age of 5 years (range: 0.5 to 21) between August 2004 and August 2007 were reviewed. Mortality rate was 55%, higher than the yearly overall PICU mortality rate of 12% (P<0.0001). The mean Pediatric Risk of Mortality Score (PRISM) III score among survivors was lower than that among nonsurvivors (9.4+/-5.7 vs. 16.4+/-5.3, P=0.001). Comparison of observed and predicted mortality derived from the PRISM III score showed that distribution of outcome was not different and the prediction model performed well (goodness of fit test: chi=3.64, df=6, P=0.725). The mortality rates were 66.6% and 33.3% in patients with high (>10 points) and low (< or =10 points) PRISM III score, respectively (P=0.05). Mortality rate was significantly related to presence and number of organ system dysfunction (P=0.031 and P=0.013, respectively), sepsis (P=0.05), mechanical ventilation (P=0.005), and positive inotropic support (P=0.003). By using multiple logistic regressions, the independent risk factor was PRISM III score at the time of admission to PICU (P=0.05). The PRISM III score performed well as a predictor of outcome. For decision to admit such patients to the PICU or to forgo life-sustaining therapies, other factors such as need for mechanical ventilation and positive inotropic support, presence and numbers of organ system dysfunction should be taken into consideration as well.