Comparison of accuracy of pediatric index of mortality-3 (PIM-3) and pediatric index of mortality-2 (PIM-2) scores in predicting outcome of children admitted to pediatric ICU of a tertiary care hospital

TL;DR

This study compares two scoring systems, PIM-3 and PIM-2, for predicting mortality in children admitted to a pediatric ICU in Pakistan.

Contribution

The study evaluates and compares the predictive accuracy of PIM-3 and PIM-2 in a specific regional ICU setting.

Findings

PIM-3 showed better discrimination with an AU-ROC of 0.858 compared to PIM-2's 0.842.

PIM-3 had higher specificity (82.57%) than PIM-2 (73.48%).

Both scores demonstrated good calibration according to the Hosmer-Lemeshow test.

Abstract

To compare the accuracy of pediatric index of mortality-3 and pediatric index of mortality-2 scores in predicting the mortality of children admitted to the pediatric intensive care unit of a tertiary care hospital in Pakistan.

This observational study conducted in the pediatric intensive care unit of Mayo Hospital, Lahore, Pakistan. In this study variables of pediatric index of Mortality-2 and pediatric index of Mortality-3 scores, along with baseline characteristics of children aged one month to 12 years, were recorded prospectively from 1st May, 2024 to 31st October, 2024. Sensitivity, specificity and positive and negative predictive values were determined for the two scores. The standardized mortality rate was calculated to compare the observed and predicted mortality of two scores. Area under the Receiver Operating Characteristic (AU-ROC) curves and Hosmer-Lemeshow Goodness-of-fit tests were used to determine which model had better discrimination and calibration.

A total of 188 patients (57.5% males, 42.5% females) with a median age of 10 months (Interquartile range: 4-36 months) were included in the study. The sensitivity of pediatric index of Mortality-2 and pediatric index of Mortality-3 was 85.71 % and 80.35%, respectively, while the specificity was 73.48% and 82.57%, respectively. The standardized mortality rate of pediatric index of Mortality-2 and pediatric index of Mortality-3 scores were 0.67 and 0.82, respectively. Both scores had good discrimination, while pediatric index of Mortality-3 performed better with area under the Receiver Operating Characteristic curve of 0.858 compared to pediatric index of Mortality-2, which was 0.842. For both tests, the Hosmer-Lemeshow goodness-of-fit tests showed good calibration (p > 0.05).

The predictive accuracy of the pediatric index of Mortality-3 score was better than that of the pediatric index of Mortality-2 score, although both scores performed well and can be used for mortality prediction.

Figures