Objective: The Pediatric Index of Mortality 2 (PIM2), one of the key mortality prediction models for children in intensive care units, has not been validated in Japan. The purpose of this study was to validate the performance of PIM2 in a population of patients admitted to one pediatric intensive care unit (PICU) in Japan.
Methods: This was a prospective cohort study involving consecutive patients admitted to the largest multidisciplinary PICU in Japan between 1 January 2008 and 31 December 2010. There were no interventions.
Results: A total of 2,536 patients were included in this study of whom 67 (2.6%) died. Discrimination between survival and death assessed by the area under the receiver operating characteristic curve was 0.92 [95% confidence interval (CI) 0.89-0.96]. Calibration across the five risk intervals according to the Hosmer-Lemeshow goodness-of-fit test showed a chi-square value of 4.8 (df = 5, p = 0.44). The standardized mortality ratio for the whole population was 0.77 (95% CI 0.59-0.96).
Conclusions: At the largest PICU center in Japan, the PIM2 was found to have excellent discriminatory power and good calibration, although it over-predicted deaths. Based on these results, PIM2 can be used as a good prediction model for pediatric mortality, which is a tool used to assess the overall quality of care in a PICU.