Objective: To compare patient outcomes using the Pediatric Index of Mortality-3 (PIM-3) model with PIM-2 model for children admitted to the intensive care unit.
Methods: We prospectively recorded the baseline characteristics, variables of PIM-3 and PIM-2 at admission, and outcomes of children ≤17 years over a period of 11 months. We used Area Under Receiver Operating Characteristics (AU-ROC) curves and Goodness-of-fit (GOF) tests to determine which of the two models had better discrimination and calibration.
Results: Out of 202 children enrolled, 69 (34%) died. Sepsis and pneumonia were the common admitting diagnoses. The AU-ROC was better for PIM-3 (0.75) as compared to PIM-2 (0.69; P=0.001). The GOF-P value was 0.001 for both models, that indicated poor calibration of both (P<0.001). The AU-ROC curves were acceptable across different age and diagnostic sub-groups.
Conclusion: PIM-3 had better discrimination when compared to PIM-2 in our unit. Both models had poor calibration across deciles of risk.