Objective: To assess the validity of the Pediatric Risk of Mortality score (PRISM), the Pediatric Index of Mortality (PIM) and the PIM 2 in two Spanish pediatric intensive care units.
Patients and methods: We prospectively studied 241 critically ill children consecutively admitted over a 6-month period. The overall performance of the scoring systems was assessed by the Standardized Mortality Ratio (SMR), comparing observed deaths with expected deaths by each index. Discrimination (the ability of the model to distinguish between patients who live and those who die) was quantified by calculating the area under the receiver operating characteristic (ROC) curve. Calibration (the accuracy of mortality risk predictions) was calculated with the Hosmer-Lemeshow goodness-of-fit test, in which statistical calibration is evidenced by p > 0.05.
Results: The mortality rate was 4.1 %. PRISM overestimated mortality (SMR = 0.44). Discrimination was better for PRISM and PIM 2 than for PIM (areas under ROC curves: 0.883, 0.871, and 0.800 respectively), with no significant differences. Finally, calibration was acceptable for PIM 2 (x2 (8) = 4.8730, p 0.8461) and for PIM (x2 (8) = 8.0876, p 0.5174), but no statistical calibration was found for PRISM (x2 (8) = 15.0281, p 0.0133).
Conclusions: PIM and PIM 2 showed better discrimination and calibration than PRISM in a heterogeneous group of children in Spanish critical care units. However, these results should be confirmed in a larger study.