Mortality prediction models for pediatric intensive care: comparison of overall and subgroup specific performance.

AIM: To validate paediatric index of mortality (PIM) and pediatric risk mortality (PRISM) models within the overall population as well as in subgroups in pediatric intensive care units (PICUs). METHODS: Variants PRISM prediction models were compared with respect to calibration between predicted risks and observed mortality) and discrimination (area the receiver operating characteristic curve, AUC). We considered the overall study population and in subgroups, defined by diagnoses, age urgency at admission, and length of stay (LoS) at the PICU. We analyzed consecutive patients younger than 16 years admitted to the eight PICUs Netherlands between February 2006 and October 2009. Patients referred to ICU or deceased within 2 h after admission were excluded. RESULTS: A 12,040 admissions were included, with 412 deaths. Variants of PIM2 were calibrated. All models discriminated well, also in patients <28 days of (neonates), with overall higher AUC for PRISM variants (PIM = 0.83, PIM2 PIM2-ANZ06 = 0.86, PIM2-ANZ08 = 0.85, PRISM = 0.88, PRISM3-24 = 0.90). discrimination for PRISM3-24 was confirmed in 13 out of 14 subgroup After recalibration PRISM3-24 predicted accurately in most (12 out of categories. Discrimination was poorer for all models (AUC < 0.73) after days at the PICU. CONCLUSION: All models discriminated well, also in subgroups including neonates, but had difficulties predicting mortality patients >6 days at the PICU. In a western European setting both the or a recalibrated version of PRISM3-24 are suited for overall prediction.