Assessing the risk of mortality in paediatric cancer patients admitted to the paediatric intensive care unit : a novel risk score?

Intensive front-line protocols have improved survival in children with malignancies; however, intensive multimodal therapy of paediatric malignancies can be associated with a significant risk of serious adverse events. Common risk scores (PRISM, PRISM III, APACHE-II) fail to predict mortality in these patients. A retrospective chart analysis of 32 paediatric cancer patients admitted to the Paediatric Intensive Care Unit (PICU) at the University Hospital of Saarland between January 2001 and December 2003 for life-threatening complications was performed. The aim of this study was to assess risk factors for short-term outcome (survival vs. non-survival when leaving the PICU) and to develop a risk score to estimate outcome in these patients. Overall survival was good (25 of 32 patients). Mortality rate was significantly related to leukaemia/lymphoma ( P = 0.029), to the number of organ failures ( P < 0.0001), neutropenia ( P = 0.001), septic shock ( P = 0.025), mechanical ventilation ( P = 0.01) and inotropic support ( P = 0.01). Employing multiple logistic regression, the strongest predictor for poor outcome was the number of organ failures ( P < 0.05). A risk score (cut-off value: >3 points for non-survival) which included the following risk factors (non-solid tumour, number of organ failures ( n > 2), neutropenia, septic shock, mechanical ventilation, and inotropic medication) yielded a sensitivity of 7/7 (95% CI: 4.56-7.00), a specificity of 23/25 (95% CI: 18.49-24.75), a positive predictive value of 23/23 (95% CI: 19.80-23.00), and a negative predictive value of 7/9 (95% CI: 3.60-8.74) for the time of admission to the PICU. Although our risk of mortality score is of prognostic value in assessing short-term outcome in these patients, prospective validation in a larger study cohort is mandatory. Furthermore, it must be emphasised that this risk score must not be used for decision-making in an individual patient.