Comparison on performance of the pediatric index of mortality (Pim)-2 and pim-3 scores in the pediatric intensive care unit of a high complexity institution

Abstract

Objective: To determine the performance of each of the available pediatric index of mortality (PIM) scores, by assessing the capability for discrimination and calibration in patients admitted to a pediatric intensive care unit in Bogotá. Design and setting: We designed a retrospective, observational cohort study, which included all patients aged between a month and 17 years and 364 days, admitted to the pediatric intensive care unit of a high complexity university hospital between April 1, 2016 and December 31, 2018. We analyzed the standardized mortality ratio, discrimination, calibration, and net reclassification index (NRI) for each model. Results: A total of 722 patients were included, the mortality rate was 3.74%, and for PIM-3, the ratio between expected and observed mortality was 0.66 [confidence interval (CI) 0.40–1.05] for PIM-2 and 1.00 (CI 0.59–1.68) for PIM-3. The Hosmer–Lemeshow (HL) test suggests inadequate calibration for PIM-2 (HL = 13.18, p = 0.11) and adequate calibration for PIM-3 (HL = 28.08, p < 0.01). The area under the diagnostic performance curves for PIM-2 and PIM-3 were 0.87 (95% CI 0.80–0.94) and 0.89 (95% CI 0.82–0.95), respectively. The NRI was −27.1%. PIM-3 classified survivors better than PIM-2, but inadequately classified nonsurvivors. Conclusion: Although both models show adequate discrimination ability, PIM-3 shows a better correlation between predicted risk score and observed mortality. Thus, it may be a useful tool for measuring the internal processes of intensive care units in Colombia and for making comparisons between groups of similar characteristics.