Accuracy of the Paediatric Index of Mortality 3 Score in Predicting Mortality in the Paediatric Intensive Care Unit (PICU).

Accurate prediction of mortality in the Pediatric Intensive Care Unit (PICU) is essential for risk stratification, clinical decision-making, benchmarking, and quality improvement. Several severity-of-illness scoring systems have been developed for this purpose, among which the Pediatric Index of Mortality (PIM) and Pediatric Risk of Mortality (PRISM) scores are the most widely used. The original PIM score, developed by Shann et al., was designed to estimate mortality risk based on physiological and diagnostic variables collected at the time of PICU admission. Subsequent revisions led to PIM2 and the currently used PIM3, which incorporates updated diagnostic categories and coefficients to improve predictive accuracy and calibration.

PIM3 has been validated across multiple international PICU settings and has generally demonstrated good discriminatory ability, with reported area under the receiver operating characteristic curve (AUROC) values ranging from approximately 0.75 to 0.90. Its main advantage lies in its simplicity and early applicability, as it relies on data obtained within the first hour of PICU admission. This makes PIM3 particularly useful for early outcome prediction, audit, and inter-unit comparisons. However, studies have shown variability in calibration and predictive performance across different regions, especially in low- and middle-income countries, likely due to differences in case mix, disease severity, healthcare resources, and admission practices.

In contrast, the PRISM III score, developed by Pollack et al., uses a more comprehensive set of physiological variables collected over the first 12–24 hours of PICU stay. While PRISM III is often considered highly accurate, its complexity and reliance on later data limit its utility for immediate risk assessment. Comparative studies between PIM3 and PRISM III have shown that both scores have comparable discrimination, though PRISM III may offer slightly better calibration in some settings, whereas PIM3 remains more practical for routine use.

Given these variations, several authors have emphasized the importance of local and center-specific validation of mortality prediction models before their routine application. Such validation ensures that the scoring system accurately reflects local patient populations and care practices. In this context, evaluating the performance of PIM3 in a tertiary care PICU setting is justified to determine its accuracy, calibration, and applicability for mortality prediction, quality assurance, and benchmarking against established tools such as PRISM III.